1. No category

 How to Write a Scientific Article:
From First Draft to Publication
Su-­‐Ling Yeh and Art Woodward Department of Psychology National Taiwan University Preface
The main purpose of this manual is to provide writing instruction for researcher
assistants and graduate students who spent much time conducting research, but have
little time or training to write a manuscript in order to submit it for publication – even
though it is well recognized that publication is the key for success in academia. A
second purpose is to help busy advisors by increasing the efficiency of their advising
activities, not only during manuscript writing, but also during revision after receiving
reviews from a journal.
By filling out the checklists as you follow the guidelines and procedures, you
will be able to write your own ideas in a logical step-by-step way, using the numerous
research articles that are included as “living examples” to imitate. Although the
example papers in this manual are mainly perception research, we hope they will be
useful in other related fields as well.
In the fall semester of 2009, we taught a course named “Scientific Writing in
Perception Research” in the Psychology Department at the National Taiwan
University. It was meant to be a PI-centered lab writing course, after the students have
taken two previous courses that emphasize more general-level basic English writing
skills. In this course, we developed the basic structure of this manual. The course
continued for three semesters under the direction of Professor Yeh, with the second
semester emphasizing writing a new article and the third semester, revision. Thus we
developed two separate checklists and instructions for these two important
components of getting an article accepted in a peer-review journal. We used both
checklists for instructions in the fourth semester before finalizing this version of the
manual.
We express our gratitude to Kuan-Ming Chen and San-Yuan Lin who were the
teaching assistants for these courses. Without their technical support, this manual
would not be possible. We also thank the professors and students in the writing
courses who provided feedback useful for improvement of the manual.
Enjoy, and we hope you have a fruitful writing experience.
Su-Ling Yeh and Art Woodward
2011, 7, 27
Please use ‘Bookmarks’ and red underlined links to navigate. Part I: Checklists Checklist for New Article
Please print out this checklist and check each box when you have begun each step
Title of Article ________________________________________________________
Authors _____________________________________________________________
Date begun ___________________________________________________________
☐ Target journal 1 ______________________________________________
☐ Review instructions to authors (insert the link)
___________________________________________________________
☐ Target journal 2 ______________________________________________
☐ Review instructions to authors (insert the link)
___________________________________________________________
☐ Select the category of importance of the paper.
It belongs to the category of __________
0. Overall structure
☐ Prepare figures (procedure and data) and tables for writing and communication with peers
(be precise on correctness, ignore artistic aspects).
☐ Write the 4 Ws for the paper and the title
(W1 = Why was the research done; W2 = What was done; W3 = What was discovered; W4 = Why is the discovery
important?)
Discuss with advisor. Date: ___________________
A. First draft (follow the first draft writing tips)
☐ For each experiment separately and repeat for multiple experiments:
☐ Write the W1 for the introduction of experiment.
☐ Write the method section. Do not edit except to possibly insert some details
not available during writing.
☐ Write the results section.
☐ Write the discussion section.
Date completed________________
☐ Write the General Discussion.
Suggestion: write only the topic sentences and check with advisor before writing the
paragraphs. Add headings if the GD is long.
☐ Write the Introduction – make the introduction mainly introduce what you wrote in the
discussion section.
Suggestion: write only the topic sentences and check with advisor before writing the
paragraphs.
☐ Write the general conclusion section and cut, paste, and revise to make the abstract.
☐ Highlight the 4 Ws and give the first draft to one peer.
(Peer name____________________)
Date completed ____________________
B. Second draft
☐ Follow the second draft tips and revise the entire paper.
Date completed ____________________
C. Third draft
☐ Do (imagined) audience assessment (highlight all terminologies and ask yourself if the
audience would understand).
☐ Check completeness and decoding time for all tables and figures.
☐ Revise and complete the title.
☐ Check accuracy of references and cross check with citation in the text.
☐ Verify APA style is correct.
☐ Verify the consistency of the terms used (name of effect, factor or the proposed
hypothesis)
Note: These details, though not directly related to the research, they will create an impression of completeness of this
article for the editor or the reviewers.
Date completed _____________________
D. Give the third draft to your advisor
☐ Highlight the 4 Ws and with your advisor compare (explicitly) to the four examples.
(W1 = Why was the research done; W2 = What was done; W3 = What was discovered; W4 = Why is the discovery
important?)
☐ Submit to Papercheck.com (including statement requesting excellent editor)
☐ Check the files with the advisors that all of them are ready to submit.
☐ Submit to target journal __________________________________
Action editor: _________________
ID No.: _________________
Date of submission: _______________
Checklist for Revision
The date for receiving the decision letter: ______________________________
Title: ___________________________________________________________
Journal: _________________________________________________________
Editor: __________________________________________________________
Reviewer: _______________________________________________________
1. Read Editor’s message:
□ Accept
Congratulations! Go celebrate!
□ Minor revision
Follow the suggestion without spending too much time on considering
what to do
□ Major revision
□ Reject but can be resubmitted
□ Reject
2. Read the opinions of reviewers:
□ Highlight the key points in the decision letter. Save it to
‘JournalName_decision_marked.doc’
□ Discuss with other authors about the next step (link to manuscript triage)
□ Do the “divide and conquer” procedure
3. Divide and Conquer:
□ Save the results of ‘Divide and Conquer’ to another file, named as
‘JournalName_ Divide and Conquer.doc’
□ Using review mode to edit the text
□ Delete sentences that are comments or descriptions
□ Keep sentences with question marks, and is an obvious demand
□ Deal with major comments
□ Deal with minor comments (statistics, method descriptions…)
□ Do as reviewer suggested
□ Appropriately use footnote
□ Not do as reviewer suggested
□ Provide counter evidence
4. Write the cover letter and responses to reviewers,
□ Edit the ‘JournalName_decision_marked.doc’ to become cover letter and
responses to reviewers
□ Mark the text you changed in the revision (mark the changes in red as is
shown here)
□ Check the consistency between the letter/responses and the revision
5. Before submission, check How to respond appropriately):
□ Responded completely (back and forth between text marked in red and the
responses to reviewers)
□ Responded politely
□ Responded with evidence
□ Send the changed text for papercheck.com (only the changed part): specify to the
editor that this is a re-submission (e.g., “This is a major revision of a previous
manuscript that was submitted to a top-tier journal: XXX. I would like to have
the same editor with a Doc ID XXX to review this paper.”)
□ Write the cover letter to specify what was answered, and what was not (rebuttal),
according to the divide and conquer procedure.
Submitted to __________________________________
Action editor: _________________
ID No.: _________________
Date of submission: _________________
Part II: Guidelines for Writing Inside-out Procedure for Writing an Article
Su-Ling Yeh and J. Arthur Woodward
National Taiwan University
Department of Psychology
November 2009
Revised: June 2010
Note 1: Consult with your advisor at any time the need arises.
Note 2: Cross-referenced articles are attached as examples of each section of an article,
along with papers illustrating where and how the 4 Ws are written into an article.



Select target journal 1 and target journal 2.
Download the journal policy and instructions to authors. Read them.
Check the example of APA style APA style example
Follow Inside out procedure (do not edit until the entire first draft is completed)
In writing the first draft of each section, open one of the articles provided in this
document (e.g., the most similar to your paper). If necessary, find another article
you are familiar with that is suitably similar to your paper. Then use it as a
model for what to say first, second, etc, at what level, and in what amount. To
the extent it is useful, transfer the schema to your paper.
A. First draft (Follow the first draft writing tips.) Steps First draft writing tips
1) For each experiment separately, write (a) through (c):
a) Write a first draft of the method section. Do not edit except to possibly
insert some details not available during writing.
b) Write a first draft of the results section (including tables and figures).
c) Write a first draft of the discussion section.
2) Now stop and write the 4 Ws for the paper and the title. 4W1 4W2 W4 Titles
3) Write general discussion.
4) Review with at least one peer.
5) Write the general introduction – make the introduction mainly introduce what you
wrote in the discussion section.
6) Write the general conclusion section and (mainly) cut and paste the abstract.
B. Second draft Second draft writing tips
Follow the second draft tips and revise the entire paper.
C. Third draft
1) Do the (imagined) audience assessment (For example, highlight all technical
terms and decide if your audience would understand them as written.)
2) Do audience assessment to check completeness and decoding time for all tables
and figures.
7
3) Finish the title. Titles
D. Give the third draft to your advisor (the English should be sloppy)
1) Highlight the 4 Ws; author and advisor compare (explicitly) with the four
examples. 4Ws1 4Ws2
2) Submit to Papercheck.com.
(Be sure to include the statement to the editors of paper check.)
For example, modify the following example:
(This paper will be submitted to a top-tier journal: “Attention Perception &
Psychophysics” for brief report. In a previous review by a journal, the editor
requested more proof reading to get the English as c orrect as possible. He wrote “If
possible, the authors should seek the editing help of a professional English language
editor to correct and smooth usage. Please assign a senior editor familiar with
editing for top tier journals.)
3) Submit to journal.
~ End of Guidelines ~
8
Appendices
Steps for Writing the First Draft
Note 1: Use these guidelines and cross references to the attached articles:
(1) (C-Y): Chen & Yeh;
(2) (Th): Theeuwes, et al;
(3) (Ca): Cavanagh, et al;
(4) (M-E): Moore and Egeth; and
(5) (R):Rowe
to follow the steps from Part A above.
Note 2: Use the highlighted sections from attached articles to see examples of
how to write the 4 Ws into your article.
4W1 4W2
Experiment 1
Introduction to Experiment 1
1) Write the main purpose of this particular experiment (w1).
Th1 Th2 M-E
2) Critical manipulation needed to answer w1 (w2).
M-E
3) Do not list details of method or procedure unless it is necessary.
Ca
Method for Experiment 1
1) Write the details of experiment 1 so that others could replicate the
experiment (participants, stimuli, apparatus, design, and experimental
procedure including practice trials).
See for examples: Th M-E C-Y
2) Complete the figure to show the procedure from the point of view of the
participant.
M-E C-Y POV
Results for Experiment 1
1) Statistics (e.g., ANOVA) should be stated as supplements to main findings.
Th
2) Do not list unnecessary statistics (e.g., irrelevant high-level interactions); if
possible, condense them into one sentence, uses Fs > xx, ps < xx for
summary.
Th
3) Do not list the non-significant results unless it is critical.
M-E
4) Complete the tables for Experiment 1.
Discussion for Experiment 1
1) State the main finding first (w3).
Th1 Th2 M-E
2) Talk about possible confounding before concluding this experiment, and
provide a transition to the next experiment (not necessarily placed here; it
could be in the introduction of the next experiment too).
M-E Th
9
Return to Inside-out Procedure
Return to Checklist for New Article
3) Mention the other findings but do not stress too much if irrelevant to W1.
4) Note the difference between small w and big W. Ideally each experiment
should have their own 4 small ws.
Repeat the above for each experiment
General Discussion
Opening paragraph of the general discussion
1) State the research purpose and summarize the main findings (short paper
may go directly to the main findings).
C-Y Th
2) Combine the purpose and finding of each experiment in short sentences.
Th
3) This is the 3rd Ws, but not the 4th one (which should appear later).
4)
5)
6)
7)
Middle paragraph of the general discussion
Optional discussion of limitations and scope.
R
Relate to literature.
R
Final paragraph of general discussion
Write mainly the 4th W.
M-E
If possible, end with a conclusion that fits the topic sentence of the first
paragraph in the introduction.
R C-Y
General Introduction
Opening two paragraphs of the Introduction
1) Write the first paragraph in a general tone, better come with a daily example.
(general and familiar).
C-Y Ca R
2) If possible, write the research purpose in the first paragraph, and then
elaborate it in the following paragraphs.
C-Y
Additional topics to be covered in the introduction
3) Cover those that mentioned in the general discussion.
10
Return to Inside-out Procedure
Return to Checklist for New Article
4) The scope should be from general to specific, in a reverse order as in the
general discussion.
5) Write 1st, 2nd Ws about here.
Final paragraph of the introduction
8) Write an oververview of the experiments.
Ca
9) Write predictions, if applicable.
Ca
General Conclusion (conclusion section is optional)
The general conclusion section should include
1) Cover the 1st, 3rd, and 4th Ws here in a short paragraph.
Ca
2) If possible, end with a sentence that is related to the first paragraph and/or
the title.
Ca
Return to Inside-out Procedure
Return to Checklist for New Article
11
Describing the experimental procedure and creating figures
The most important principle in describing the experimental procedure, including tables and figures is this:
Describe the procedure from the point of view of the participant
Participant POV
1) Describe in words what the participants actually saw, in the order they saw it. Different
experimental conditions should be included so the critical differences in what the participants
experienced in the different conditions will be clear.
2) Describe in an accompanying figure a representation of what the participant actually saw and
when they saw it.
a) The figure should show the actual stimuli if possible, but usually they will be schematic
representations of what they saw. For example, when the stimuli are white on a black
background this consumes too much journal ink, so the background can be made white
and the figure black, with an accompanying explanation;
b) The figure should show the sequence and time line;
c) All symbols in the table should be defined (there should be no undefined symbols)
d) The labels and figure title should contain the units (inches, seconds, average milliseconds etc);
e) Different experiment conditions should be represented in the figure using the same names
of the independent variables as used in the text;
f) The dependent variables should be referred to by the same names as in the text;
g) The figure caption should closely match the words in the text section that refers to the
figure;
h) In figures and tables with several independent variables see the following examples form
published article (adopted from Fortin et al., 2010);
Figure 7. Experiment 4. (a) Mean reaction times to the
second stimulus as a function of memory set size in
switch and no-switch trials. Error bars represent the
SEM computed with the MSE.
Return to Inside-out Procedure
12
The 4 Ws
Art Woodward
Fall 2009
The four Ws are:
W1: Why did you do the research?
W2: What did you do? (Note this is a verbal explanation of what was done that
appears in the introduction)
W3: What did you discover?
W4: Why is the discovery important? (Note in answering this question, do not
simply restate the discovery. Instead,
explain why the discovery is important.)
Five articles are used to illustrate the 4 Ws:
(1) (S): Shams, L., Kamitani, Y., & Shimojo, S. (2000). What you see is what you
hear. Nature, Vol. 408, pp.788.
W1 W2 W3 W4(1) W4(2)
(2) (T-E): Tong, F. H., & Engel, S. A. (2001). Interocular rivalry revealed in the
human cortical blind-spot representation. Nature, 411, 195-199.
W1 W2 W3(1) W3(2) W4(1) W4(2) W4(3) W4(4)
(3) (S-K): Saxe, R. & Kanwisher, N. (2003). People thinking about thinking people:
The role of the temporo-parietal junction in theory of mind. NeuroImage. 19
1835-1842
W1 W2(1) W2(2) W2(3) W2(4) W3(1) W3(2) W4(1) W4(2)
(4) (W-K): Wojciulik, E., & Kanwisher, N. (1999). The Generality of Parietal
Involvement in Visual Attention. Neuron. 23 747-764
W1(1) W1(2) W2 W3 W4(1) W4(2) W4(3)
(5) (R): Rowe, G., Hirsh, J.B., & Anderson, A.K., (2006). Positive affect increases
the breadth of attentional selection. Proceedings of the National Academy of
Science, 104(1): 383-8.
W1 W2(1) W2(2) W2(3) W3(1) W3(2) W3(3) W4
Return to Inside-out Procedure
Return to Steps for Writing 1st Draft
Return to Checklist for New Article
1
13
First draft Writing Tips
Please do NOT do these while writing the first draft:
Do not insert any un-needed details like names, dates, etc, unless automatic.
1) Do not try to improve the English as you write – write in Chinese-English.
2) If you don’t know the English word, insert Chinese characters and move on.
3) Do not edit your first draft as you write. Complete the entire first draft before editing.
Instead, please DO these things as much as you can (even if it slows you down):
Before you write, spend 5 minutes planning what you will write.
Think about:
(1) An attractive title;
G
(2) Intro: Write General
Specific
and Familiar Not Fam.
Intro
Discussion: Do the reverse and
Step 1
Cross reference
Step 2
(Do not forget the 4 Ws)
F
Not F
Step 3
Attempt to take two steps
Gen.
Discussion
S
(3) Pause and think hard about the first sentence in each paragraph (called the topic
sentence). The topic sentence should name the topic of the paragraph and state the
controlling idea (see example on the next page);
(4) Concentrate on conciseness (use no unnecessary words and do not repeat);
(5) Concentrate on cohesion (connect the sentences logically);
(6) Concentrate on coherence (make all sentences in a paragraph be about the same topic
and obey the controlling idea of the topic sentence);
(7) Use headings to help the reader identify the different main topics in your document;
(8) Take the perspective of the reader (do not use words the reader may not understand; do
not expect readers to remember what they read on previous pages.)
14
Return to Inside-out Procedure
Practice for topic sentences and controlling idea
Use bubbl.us (http://bubbl.us/)
Write clear and well defined topic sentences. See the following examples.
1. There are three reason NTU is the best university in Taiwan.
Topic: _NTU is the best university in Taiwan_;
Controlling idea: _three reasons it’s the best__;
Give some specific examples of the controlling idea:
2. The food in Taiwan is unique in East Asia because of the special spices used.
Topic: __________________________________;
Controlling idea: __________________________;
Give some specific examples of the controlling idea:
3. Taiwanese food is unique in East Asia because of the many fresh ingredients.
Topic: __________________________________;
Controlling idea: __________________________;
Give some specific examples of the controlling idea:
Return to Inside-out Procedure
15
Get off to a good start with W1
Make sure sentences similar to these examples appear in the first several paragraphs of
your paper. Before submitting to a journal, review explicitly your 4 Ws in the paper by
comparing them to the examples below.
Example 1: Shams, Nature (ISI; 12/3/09;144)
“Vision is believed to dominate our multi-sensory perception of the world. Here we overturn this
established view by showing that auditory information can qualitatively alter the perception of an
unambiguous visual stimulus to create a striking visual illusion.”
Example 2: Tong & Engel. Nature (ISI; 12/3/09;155)
“Despite extensive research, the neural basis of binocular rivalry has remained highly
controversial. Specifically, it is debated whether discrepant monocular patterns rival because of
interocular competition or pattern competition. To resolve this issue, we used functional magnetic
resonance imaging (fMRI) to monitor rivalry-related activity in a monocular region of human V1
corresponding to the blind spot.”
Example 3: Saxes and Kanwisher, NeuroImage (ISI; 12/3/09; 213)
“The remarkable human facility with social cognition depends on a fundamental ability to reason
about other people. Specifically, we predict and interpret the behavior of people based on an
understanding of their minds: that is, we use a “theory of mind.” In this study we show that a
region of human temporo-parietal junction is selectively involved in reasoning about the contents
of other people’s minds.”
Example 4: Wojciulik and Kanwisher, Neuron (ISI; 12/3/09; 273)
“However, very little work has been directed to the crucial question of whether visual attention in
fact consists of a single general-purpose mechanism or whether it instead consists of a
heterogeneous set of different mechanisms, each involved in a different kind of selection. In the
present study, we used functional magnetic resonance imaging (fMRI) to address this question,
asking whether there is any region of the human brain that is activated by each of three very
different attention-requiring tasks yet not activated by a language task that is difficult but does not
place heavy demands on visual attention.”
Return to Inside-out Procedure
Return to Steps for Writing the 1st Draft
16
Categories of importance for documenting W4
Category of Importance
Develop a new theory or model
Example
Modify a famous theory or model


Rummelhart D.E. and McClelland J. L.
(Distributed Processing Models)
David Marr. (Level of analysis; computational
model of visual processing)
Young, T., and Helmholtz, H. (Trichromatic
Theory of Color Vision)
Gibson, J. J. (Direct perception)
Wolfe J.M. (Guided Search)
Overturn a widely held belief

Shams, L (Double Click Illusion)
Describe or explain a puzzling mystery

Develop a new method or paradigm



Ramachandran V.S. (e.g., explains phantom
pain)
“In this paper we present a new model for…”



“Here we overturn the widely held belief that…”
Discover a new phenomenon and rule out
alternative explanations

Pearl J. (Bayes Net Models)
Lauterbur, P., and Mansfield, P. (Magnetic
Resonance Imaging): they applied MR physics to
develop MRI scanner for human body; they won
a Nobel prize in Physiology/ Medicine, 2003.
Egly et al., 1994 (The double rectangle cueing
paradigm that can probe location- and objectbased attention in a single experimental
framework)
Tsuchiya and Koch (Continuous Flash
Suppression)
Kanwisher, N. (Repetition Blindness)
Resolve an existing debate or controversy

Tong, F.H. and Engle S.A. (Binocular Rivalry)
Discover the underlying connection
between two existing phenomena

Theory that influences other disciplines

Maxwell’s equations in Electronic and Magnetic.
[There may be no integral theories like this in
psychology]
Kahnmann & Tversky (Prospect Theory): Two
psychologists won the Nobel prize in economics
Useful for practical applications

“This study presents a new paradigm for..”


“This is the first study to …”
“Thus our study contributes to resolving the
controversy about…”
Karremans, Stroebe, & Claus, 2006 (Subliminal
priming that can be used in movies or commercial
advertisements)
Return to Inside-out Procedure
Return to Checklist for New Article
17
Some Ideas for Good Titles
A good title will attract the attention of potential readers, arouse their curiosity, and
correctly prime them for the content of the article.
1) Some guidelines and suggested characteristics are:
a) should contain familiar (older) information;
b) may contain a clear statement of the main finding “Conditions where implicit and
explicit cross-modal information are processed independently”
c) may contain a question that will be answered by the research, so curiosity is
aroused “Can implicit and explicit cross-modal sensory information be processed
independently”
d) may be a “play on words” or a metaphor, “What you hear is what you see”
e) The title could be some reflection of the first W, third W, or fourth W
2) Examples: Which of the above characteristics apply to the following real article titles?
“What you see is what you hear”
“Interocular rivalry revealed in the human cortical blind-spot representation”
“People thinking about thinking people: The role of the temporo-parietal junction in
theory of mind.”
“The Generality of Parietal Involvement in Visual Attention”
Return to Inside-out Procedure
18
Top-Down Second Draft Writing Tips
Creating your second draft
1) Revise your title if you can make it more attractive and interesting to your
audience.
2) Look at all your heading; add and remove headings as necessary;
3) Paragraph coherence:
a) look at each topic sentence and decide if it needs to be made more clear
for the reader;
b) Make sure all the sentences in the paragraph match the topic sentence. If
they don’t remove them, or create another paragraph.
4) Sentences
a) Re write any run on sentences;
b) Combine any sets of too short sentences;
c) Insert any missing details, such as dates, names, or English words you did
not know.
Note: You do not need to improve the English grammar very much. Let the editor show
you how to do that so you can practice revising according to professional edits. However
you do need to be sure the writing is concise, cohesive, and coherent. The editors at
Papercheck.com cannot improve the quality your thinking.
Here are some useful sites for practice concise, cohesive, and coherent:
Concise
http://grammar.ccc.commnet.edu/grammar/quizzes/wordy_quiz.htm
http://grammar.ccc.commnet.edu/grammar/quizzes/nova/nova8.htm
http://grammar.ccc.commnet.edu/grammar/quizzes/nova/nova11.htm
Cohesion
http://www.uefap.com/writing/exercise/parag/refer.htm#top
Coherence
http://lrs.ed.uiuc.edu/students/fwalters/cohex3d.html
Return to Inside-out Procedure
Return to Checklist for New Article
Some Useful Information When Having the Paper Edited
1.
Have the paper edited (sent to papercheck.com) after everything is ready (references,
figures, tables included)
2.
If you have a preferable editor, add the following: “Please have the editor who
edited document XXXXXX (ID) work on this document".
3.
Examples: “This paper will be submitted to a top-tier journal: xxx.” If applicable,
add: “In a previous review by a journal, it was suggested that we seek the editing
help of a native English speaker to correct and smooth usage.” Or “This is the final
version and we will submit the manuscript after revising according to your English
edits.”
Return to Inside-out Procedure
20
Figure 2.1. Sample One-Experiment Paper (The numbers refer to numbered
sections in the Publication Manual.)
Running head: EFFECTS OF AGE ON DETECTION OF EMOTION
1
Establishing a title, 2.01; Preparing the
manuscript for submission, 8.03
Effects of Age on Detection of Emotional Information
Christina M. Leclerc and Elizabeth A. Kensinger
Boston College
Formatting the author name (byline) and
institutional affiliation, 2.02, Table 2.1
Elements of an author note, 2.03 Author Note
Christina M. Leclerc and Elizabeth A. Kensinger, Department of Psychology,
2
EFFECTS OF AGE ON DETECTION OF EMOTION
Boston College.
Abstract
Writing the abstract, 2.04
This research
arch was supported by National Science Foundation Grant BCS 0542694
Age differences were examined in affective processing, in the context of a visual search task.
awarded to Elizabeth
beth A. Kensinger.
Young and older adults were faster to detect high arousal images compared with low arousal and
Correspondence
ndence concerning this article should be addressed to Christina M. Leclerc,
neutral items. Younger adults were faster to detect positive high arousal targets compared with
Department of Psychology,
sychology, Boston College, McGuinn Hall, Room 512, 140 Commonwealth
other categories. In contrast, older adults exhibited an overall detection advantage for emotional
Avenue, Chestnut
ut Hill, MA 02467. Email: [email protected]
images compared with neutral images. Together, these findings suggest that older adults do not
display valence-based effects on affective processing at relatively automatic stages.
Keywords: aging, attention, information processing, emotion, visual search
Double-spaced manuscript,
Times Roman typeface,
1-inch margins, 8.03
Paper adapted from “Effects of Age on Detection of Emotional Information,” by C. M. Leclerc and E. A. Kensinger,
2008, Psychology and Aging, 23, pp. 209–215. Copyright 2008 by the American Psychological Association.
21
Return to Inside-out procedure
Figure 2.1. Sample One-Experiment Paper (continued)
EFFECTS OF AGE ON DETECTION OF EMOTION
3
Writing the introduction, 2.05
Effects of Age on Detection of Emotional Information
Frequently, people encounter situations in their environment in which it is impossible to
attend to all available stimuli. It is therefore of great importance for one’s attentional processes to
select only the most salient information in the environment to which one should attend. Previous
research has suggested that emotional information is privy to attentional selection in young
adults (e.g., Anderson, 2005; Calvo & Lang, 2004; Carretie, Hinojosa, Marin-Loeches, Mecado,
& Tapia, 2004; Nummenmaa, Hyona, & Calvo, 2006), an obvious service to evolutionary drives
Selecting to approach rewarding situations and to avoid threat and danger (Davis & Whalen, 2001; Dolan
the correct
tense, 3.18 & Vuilleumier, 2003; Lang, Bradley, & Cuthbert, 1997; LeDoux, 1995).
For example, Ohman, Flykt, and Esteves (2001) presented participants with 3 × 3 visual
Numbers
arrays with images representing four categories (snakes, spiders, flowers, mushrooms). In half
expressed
in words, the arrays, all nine images were from the same category, whereas in the remaining half of the
4.32
Ordering citations within
the same parentheses, 6.16
Numbers that represent
statistical or mathematical
functions, 4.31
arrays, eight images were from one category and one image was from a different category (e.g.,
Use of hyphenation for
compound words, 4.13,
discrepant
ant stimulus. Results indicated that fear
fear-relevant
r- relevant images were more quickly detected than Table 4.1
eight flowers and one snake). Participants were asked to indicate whether the matrix included a
elevant items, aand larger search facilitation effects were observed for participants who
fear-irrelevant
arful of the stimuli. A similar pattern of results has been observed when examining the
were fearful
EFFECTS OF AGE ON DETECTION OF EMOTION
4
n-grabbing
(includ ing those
attention-grabbing
nature of negative facial expressions, with threatening faces (including
Calvo & Lang, 2004; Carretie et al., 2004; Juth, Lundqvist, Karlsson, & Ohman, 2005;
nded to) identified more quickly than positive or neutral faces (Eastwood, Smilek, &
not attended
Nummenmaa
et 1988).
al., 2006).
e, 2001; Hansen
Merikle,
& Hansen,
The enhanced detection of emotional information is
From this
research,
it seems
that younger adults
show
benefits for
ited to threatening stimuli;
there
is evidence
thatclear
any high-arousing
stimulus
candetection
be
not limited
arousing of
information
environment.
It is lessvalenced
clear whether
these 2005;
effects
d rapidly, regardless
whether itinis the
positively
or negatively
((Anderson,
5 are preserved
detected
across the adult life span. The focus of the current research is on determining the extent to which
Continuity in presentation aging influences the early, relatively automatic detection of emotional information.
of ideas, 3.05
Regions of the brain thought to be important for emotional detection remain relatively
intact with aging (reviewed by Chow & Cummings, 2000). Thus, it is plausible that the detection
of emotional information remains relatively stable as adults age. However, despite the
preservation of emotion-processing regions with age (or perhaps because of the contrast between
the preservation of these regions and age-related declines in cognitive-processing regions; Good
et al., 2001; Hedden & Gabrieli, 2004; Ohnishi, Matsuda, Tabira, Asada, & Uno, 2001; Raz,
No capitalization in
naming theories, 4.16
2000; West, 1996), recent behavioral research has revealed changes that occur with aging in the
regulation and processing of emotion. According to the socioemotional selectivity theory
Citing one
work by six
or more
authors, 6.12
(Carstensen, 1992), with aging, time is perceived as increasingly limited, and as a result, emotion
regulation becomes a primary goal (Carstensen, Isaacowitz, & Charles, 1999). According to
socioemotional selectivity theory, age is associated with an increased motivation to derive
emotional meaning from life and a simultaneous decreasing motivation to expand one’s
knowledge base. As a consequence of these motivational shifts, emotional aspects of the
22
Return to Inside-out procedure
Figure 2.1. Sample One-Experiment Paper (continued)
EFFECTS OF AGE ON DETECTION OF EMOTION
5
Using the colon between
two grammatically
complete clauses, 4.05
To maintain positive affect in the face of negative age-related change (e.g., limited time
remaining, physical and cognitive decline), older adults may adopt new cognitive strategies. One
such strategy, discussed recently, is the positivity effect (Carstensen & Mikels, 2005), in which
older adults spend proportionately more time processing positive emotional material and less
time processing negative emotional material. Studies examining the influence of emotion on
memory (Charles, Mather, & Carstensen, 2003; Kennedy, Mather, & Carstensen, 2004) have
found that compared with younger adults, older adults recall proportionally more positive
information and proportionally less negative information. Similar results have been found when
examining eye-tracking patterns: Older adults looked at positive images longer than younger
adults did, even when no age differences were observed in looking time for negative stimuli
Capitalization of words
beginning a sentence after
a colon, 4.14
(Isaacowitz, Wadlinger, Goren, & Wilson, 2006). However, this positivity effect has not gone
uncontested; some researchers have found evidence inconsistent with the positivity effect (e.g.,
Grühn, Smith, & Baltes, 2005; Kensinger, Brierley, Medford, Growdon, & Corkin, 2002).
Hypotheses and their
correspondence to research
design, Introduction, 2.05
Based on this previously discussed research, three competing hypotheses exist to explain
motional processing associated with the normal aging process. First,
age differences in emotional
emotional informationn may remain important throughout the life span, leading to similarly
OFin
AGE
ON DETECTION
OF Second,
EMOTION
younger
and older adults.
with aging,
facilitated detection of emotionalEFFECTS
information
Using the semicolon to
separate two independent
6
clauses not joined
by
a conjunction, 4.04
emotional informationn may take on additional importance, resulting in older adults’ enhanced
rapidly detect emotional information. We hypothesized that on the whole, older adults would be
al information in their environment. Third, older adults may focus
detection of emotional
slower to detect information than young adults would be (consistent with Hahn, Carlson, Singer,
principally on positivee emotional information and may show facilitated detection of positive, but
& Gronlund, 2006; Mather & Knight, 2006); the critical question was whether the two age
nal information.
not negative, emotional
groups would show similar or divergent facilitation effects with regard to the effects of emotion
The primary goal in the present experiment was to adjudicate among these alternatives.
on item detection. On the basis of the existing literature, the first two previously discussed
ed a visual search paradigm to assess young and older adults’ abilities to
To do so, we employed
hypotheses seemed to be more plausible than the third alternative. This is because there is reason
Using the comma between
elements in a series, 4.03
Punctuation with citations
in parenthetical material,
6.21
to think that the positivity effect may be operating only at later stages of processing (e.g.,
strategic, elaborative, and emotion regulation processes) rather than at the earlier stages of
processing involved in the rapid detection of information (see Mather & Knight, 2005, for
discussion). Thus, the first two hypotheses, that emotional information maintains its importance
across the life span or that emotional information in general takes on greater importance with
age, seemed particularly applicable to early stages of emotional processing.
Indeed, a couple of prior studies have provided evidence for intact early processing of
emotional facial expressions with aging. Mather and Knight (2006) examined young and older
Citing references in text,
inclusion of year within
paragraph, 6.11, 6.12
adults’ abilities to detect happy, sad, angry, or neutral faces presented in a complex visual array.
Mather and Knight found that like younger adults, older adults detected threatening faces more
quickly than they detected other types of emotional stimuli. Similarly, Hahn et al. (2006) also
found no age differences in efficiency of search time when angry faces were presented in an
array of neutral faces, compared with happy faces in neutral face displays. When angry faces,
compared with positive and neutral faces, served as nontarget distractors in the visual search
Prefixes and
suffixes that
do not require
hyphens,
Table 4.2
arrays, however, older adults were more efficient in searching, compared with younger adults,
23
Return to Inside-out procedure
Figure 2.1. Sample One-Experiment Paper (continued)
EFFECTS OF AGE ON DETECTION OF EMOTION
7
negative stimuli were not of equivalent arousal levels (fearful faces typically are more arousing
than happy faces; Hansen & Hansen, 1988). Given that arousal is thought to be a key factor in
modulating the attentional focus effect (Hansen & Hansen, 1988; Pratto & John, 1991; Reimann
& McNally, 1995), to more clearly understand emotional processing in the context of aging, it is
necessary to include both positive and negative emotional items with equal levels of arousal.
In the current research, therefore, we compared young and older adults’ detection of four
categories of emotional information (positive high arousal, positive low arousal, negative high
arousal, and negative low arousal) with their detection of neutral information. The positive and
Prefixed words that
require hyphens,
Table 4.3
negative stimuli were carefully matched on arousal level, and the categories of high and low
arousal were closely matched on valence to assure that the factors of valence (positive, negative)
and arousal (high, low) could be investigated independently of one another. Participants were
presented with a visual search task including images from these different categories (e.g., snakes,
cars, teapots). For half of the multi-image
g arrays,
y , all of the images
g were of the same item,, and for
the remaining half of the arrays, a single target image of a different type from the remaining
Using abbreviations, 4.22; Explanation
of abbreviations, 4.23; Abbreviations
used often in APA journals, 4.25;
Plurals of abbreviations, 4.29
items was included. Participants were asked to decide whether a different item was included in
EFFECTS OF AGE ON DETECTION OF EMOTION
8
the array, and their reaction times were recorded for each decision. Of primary interest were
for the arousing items than shown by the young adults (resulting in an interaction between age
differences in response times (RTs)) based on the valence and arousal levels of the target
and arousal).
ung and older adults were equally focused on emotional
categories. We reasoned that if young
Method
information, then we would expectt similar degrees of facilitation in the detection of emotional
older adults were more affectively focused than
stimuli for the two age groups. By contrast, ifParticipants
Elements of the Method
section, 2.06; Organizing
a manuscript with levels
of heading, 3.03
adults (14 women, 10 men, Mage = 19.5 years, age range: 18–22 years) were
were younger adults, older adults should show eitherYounger
faster detection
speeds for all of the
recruited with flyers posted on the Boston College campus. Older adults (15 women, nine men,
utral items) than shown by young adults or greater facilitation
emotional items (relative to the neutral
Mage = 76.1 years, age range: 68–84 years) were recruited through the Harvard Cooperative on
Identifying
subsections
within the
Method
section, 2.06
Aging (see Table 1, for demographics and test scores).1 Participants were compensated $10 per
hour for their participation. There were 30 additional participants, recruited in the same way as
described above, who provided pilot rating values: five young and five old participants for the
assignment of items within individual categories (i.e., images depicting cats), and 10 young and
10 old participants for the assignment of images within valence and arousal categories. All
Using numerals to express
numbers representing age, 4.31
participants were asked to bring corrective eyewear if needed, resulting in normal or corrected
to normal vision for all participants.
Materials and Procedure
Participant (subject)
characteristics,
Method, 2.06
The visual search task was adapted from Ohman et al. (2001). There were 10 different
types of items (two each of five Valence × Arousal categories: positive high arousal, positive low
arousal, neutral, negative low arousal, negative high arousal), each containing nine individual
exemplars that were used to construct 3 × 3 stimulus matrices. A total of 90 images were used,
each appearing as a target and as a member of a distracting array. A total of 360 matrices were
presented to each participant; half contained a target item (i.e., eight items of one type and one
target item of another type) and half did not (i.e., all nine images of the same type). Within the
24
Return to Inside-out procedure
Figure 2.1. Sample One-Experiment Paper (continued)
EFFECTS OF AGE ON DETECTION OF EMOTION
9
matrix. Within the 180 target trials, each of the five emotion categories (e.g., positive high
arousal, neutral, etc.) was represented in 36 trials. Further, within each of the 36 trials for each
emotion category, nine trials were created for each of the combinations with the remaining four
other emotion categories (e.g., nine trials with eight positive high arousal items and one neutral
item). Location of the target was randomly varied such that no target within an emotion category
was presented in the same location in arrays of more than one other emotion category (i.e., a
negative high arousal target appeared in a different location when presented with positive high
arousal array images than when presented with neutral array images).
The items within each category of grayscale images shared the same verbal label (e.g.,
mushroom, snake), and the items were selected from online databases and photo clipart
Latin abbreviations, 4.26
Numbers expressed in words
at beginning of sentence, 4.32
packages. Each image depicted a photo of the actual object. Ten pilot participants were asked to
write down the name corresponding
to eachhead:
object;
any object
didON
notDETECTION
consistently generate
Running
R
EFFECTS
OFthat
AGE
OF EMOTION
10
the intended response was eliminated from the set. For the remaining images, an additional 20
selected such that the arousal difference between positive low arousal and positi
positive high arousal
pilot participants rated the emotional valence and arousal of the objects and assessed the degree
was equal to the difference between negative low arousal and negative high arou
arousal.
of visual similarity among objects within a set (i.e., how similar the mushrooms were to one
between
Similarity ratings. Each item was rated for within-category and between-categories
another) and between objects across sets (i.e., how similar the mushrooms were to the snakes).
similarity. For within-category similarity, participants were shown a set of exemplars
exem
(e.g., a set
Valence and arousal ratings. Valence and arousal were judged on 7-point scales (1 =
of mushrooms) and were asked to rate how similar each mushroom was to the re
rest of the
negative valence or low arousal and 7 = positive valence or high arousal). Negative objects
mushrooms, on a 1 (entirely dissimilar) to 7 ((nearly identical
identical)) scale. Participants made these
received mean valence ratings of 2.5 or lower, neutral objects received mean valence ratings of
ratings on the basis of overall similarity and on the basis of the specific visual di
dimensions in
3.5 to 4.5, and positive objects received mean valence ratings of 5.5 or higher. High arousal
which the objects could differ (size, shape, orientation). Participants also rated hhow similar
objects received mean arousal ratings greater than 5, and low arousal objects (including all
objects of one category were to objects of another category (e.g., how similar the mushrooms
neutral stimuli) received mean arousal ratings of less than 4. We selected categories for which
were to the snakes). Items were selected to assure that the categories were equate
equated on withinboth young and older adults agreed on the valence and arousal classifications, and stimuli were
category and between-categories similarity of specific visual dimensions as well as for the
Italicization of anchors
of a scale, 4.21
overall similarity of the object categories ((p
(pss > .20). For example, we selected pa
particular
h
ti l cats
t so that
th t the
th mushrooms
h
i il to
t one another as were the
mushrooms
andd particular
were as similar
cats (i.e., within-group similarity was held constant across the categories). Our object selection
also assured that the categories differed from one another to a similar degree (e.g., that the
mushrooms were as similar to the snakes as the cats were similar to the snakes).
Procedure
Each trial began with a white fixation cross presented on a black screen for 1,000 ms; the
matrix was then presented, and it remained on the screen until a participant response was
recorded. Participants were instructed to respond as quickly as possible with a button marked yes
if there was a target present, or a button marked no if no target was present. Response latencies
and accuracy for each trial were automatically recorded with E-Prime (Version 1.2) experimental
25
Return to Inside-out procedure
Figure 2.1. Sample One-Experiment Paper (continued)
EFFECTS OF AGE ON DETECTION OF EMOTION
11
software. Before beginning the actual task, participants performed 20 practice trials to assure
compliance with the task instructions.
Results
Elements of the
Results section, 2.07
Analyses focus on participants’ RTs to the 120 trials in which a target was present and
was from a different emotional category from the distractor (e.g., RTs were not included for
Abbreviations
accepted as
words, 4.24
arrays containing eight images of a cat and one image of a butterfly because cats and butterflies
are both positive low arousal items). RTs were analyzed for 24 trials of each target emotion
category. RTs for error trials were excluded (less than 5% of all responses) as were RTs that
Symbols, 4.45;
Numbers, 4.31
were ±3 SD from each participant ’s mean (approximately 1.5% of responses). Median RTs were
then calculated for each of the five emotional target categories, collapsing across array type (see
Table 2 for raw RT values for each of the two age groups). This allowed us to examine, for
example, whether participants were faster to detect images of snakes than images of mushrooms,
Nouns followed
by numerals or
letters, 4.17
regardless of the type of array in which they were presented. Because our main interest was in
examining the effects of valence and arousal on participants’ target detection times, we created
scores for each emotional target category that controlled for the participant’s RTs to detect
neutral targets (e.g., subtracting the RT to detect neutral targets from the RT to detect positive
high arousal targets). These difference scores were then examined with a 2 × 2 × 2 (Age [young,
older] × Valence [positive, negative] × Arousal [high, low]) analysis of variance (ANOVA). This
ANOVA revealed only a significant main effect of arousal, F(1, 46) = 8.41, p = .006, ηp2 = .16,
with larger differences between neutral and high arousal images (M = 137) than between neutral
and low arousal images (M = 93; i.e., high arousal items processed more quickly across both age
Reporting
p values,
decimal
fractions,
4.35
Statistical symbols,
4.46, Table 4.5
groups compared with low arousal items; see Figure 1). There was no significant main effect for
valence, nor was there an interaction between valence and arousal. It is critical that the analysis
Numbering and discussing
figures in text, 5.05
26
Return to Inside-out procedure
Figure 2.1. Sample One-Experiment Paper (continued)
EFFECTS OF AGE ON DETECTION OF EMOTION
12
revealed only a main effect of age but no interactions with age. Thus, the arousal-mediated
effects on detection time appeared stable in young and older adults.
The results described above suggested that there was no influence of age on the
influences of emotion. To further test the validity of this hypothesis, we submitted the RTs to the
five categories of targets to a 2 × 5 (Age [young, old] × Target Category [positive high arousal,
Statistics
in text, 4.44
positive low arousal, neutral, negative low arousal, negative high arousal]) repeated measures
2
ANOVA. Both the age group, F(1, 46) = 540.32, p < .001,
ηp2
= .92, and the ta rget category,
Spacing, alignment,
and punctuation of
mathematical copy, 4.46
F(4, 184) = 8.98, p < .001, η p2 = .16, main effects were significant, as well as the Age Group ×
Target Category interaction, F(4, 184) = 3.59, p = .008, η p2 = .07. This interaction appeared to
reflect the fact that for the younger adults, positive high arousal targets were detected faster than
targets from all other categories, ts(23) < –1.90,p < .001, with no other target categories
differing significantly from one another (although there were trends for negative high arousal
Capitalize effects
or variables when
they appear with
multiplication
signs, 4.20
and negative low arousal targets to be detected more rapidly than neutral targets (p < .12). For
older adults, all emotional categories of targets were detected more rapidly than were neutral
targets, ts(23) > 2.56, p < .017, and RTs to the different emotion categories of targets did not
differ significantly from one another. Thus, these results provided some evidence that older
adults may show a broader advantage for detection of any type of emotional information,
whereas young adults’ benefit may be more narrowly restricted to only certain categories of
Elements of the
Discussion section, 2.08
emotional information.
Discussion
As outlined previously, there were three plausible alternatives for young and older adults’
performance on the visual search task: The two age groups could show a similar pattern of
enhanced detection of emotional information, older adults could show a greater advantage for
27
Return to Inside-out procedure
Figure 2.1. Sample One-Experiment Paper (continued)
EFFECTS OF AGE ON DETECTION OF EMOTION
13
emotional detection than young adults, or older adults could show a greater facilitation than
young adults only for the detection of positive information. The results lent some support to the
first two alternatives, but no evidence was found to support the third alternative.
Clear statement of support or
nonsupport of hypotheses,
Discussion, 2.08
In line with the first alternative, no effects of age were found when the influence of
valence and arousal on target detection times was examined; both age groups showed only an
arousal effect. This result is consistent with prior studies that indicated that arousing information
can be detected rapidly and automatically by young adults (Anderson, Christoff, Panitz, De
Rosa, & Gabrieli, 2003; Ohman & Mineka, 2001) and that older adults, like younger adults,
continue to display a threat detection advantage when searching for negative facial targets in
nd neutral distractors (Hahn et al., 2006; Mather & Knight, 2006).
6 Given the
arrays of positive and
relative preservationn of automatic processing with aging (Fleischman, Wilson, Gabrieli, Bienias,
EFFECTS OF AGE ON DETECTION OF EMOTION
nnings & Jacoby, 1993),
3 it makes sense that older adults would remain able
& Bennett, 2004; Jennings
14
to take advantage off these automatic
alerting
systems
foreffects
detecting
high arousal
information.
processing,
given
that no
of valence
were observed
in older adults’ detection speed. In the
espite the similarity
in arousal-mediated
on detection
between
the two
However, despite
effects
present study,
older adults were
equally
fast to detect
positive
and negative information,
ent study did provide some evidence for age-related change (specifically,
age groups, the present
consistent with prior research that indicated that older adults often attend equally to positive and
ment) in the detection of emotional information. When examining RTs for
age-related enhancement)
negative stimuli (Rosler et al., 2005). Although the pattern of results for the young adults has
in detecting
positive
the five categories of emotionaldiffered
targets,across
younger
adults were
efficient
studies—in
the more
present
study and
in some past
research, young adults have shown
2 whereas
high arousal images (as presented
in Table
2),
older adults
displayed
an Anderson,
overall 2005; Calvo & Lang, 2004; Carretie
facilitated
detection
of positive
information
(e.g.,
ting all emotional
This2006),
pattern
advantage for detecting
images
compared
withNummenmaa
neutral images.
et al., 2004;
Juth
et al., 2005;
et al.,
whereas in other studies, young adults
nfluence ofhave
suggests a broader influence
emotion
onan
older
adults’ for
detection
of information
stimuli, providing
support & Dolan, 2002; Hansen &
shown
advantage
negative
(e.g., Armony
hat as individuals
age,
emotional
becomes
more salient.
for the hypothesis that
Hansen,
1988;
Mogg, information
Bradley, de Bono,
& Painter,
1997; Pratto & John, 1991; Reimann &
ng that thisMcNally,
second set
of findings
It is interesting
is clearly
inconsistent
with 1996)—what
the hypothesis
1995;
Williams,
Mathews,
& MacLeod,
is important to note is that the
ffect in older
information
that the positivity effect
adults
operates
at relatively
automatic
stages of
older
adults
detected
both positive
and negative
stimuli
at equal rates. This equivalent detection
of positive and negative information provides evidence that older adults display an advantage for
Use of an em dash to
indicate an interruption the detection of emotional information that is not valence-specific.
in the continuity of a
Thus, although younger and older adults exhibited somewhat divergent patterns of
sentence, 4.06;
emotional detection on a task reliant on early, relatively automatic stages of processing, we
Description of an
em dash, 4.13
found no evidence of an age-related positivity effect. The lack of a positivity focus in the older
adults is in keeping with the proposal (e.g., Mather & Knight, 2006) that the positivity effect
does not arise through automatic attentional influences. Rather, when this effect is observed in
older adults, it is likely due to age-related changes in emotion regulation goals that operate at
later stages of processing (i.e., during consciously controlled processing), once information has
been attended to and once the emotional nature of the stimulus has been discerned.
Although we cannot conclusively say that the current task relies strictly on automatic
processes, there are two lines of evidence suggesting that the construct examined in the current
28
Return to Inside-out procedure
Figure 2.1. Sample One-Experiment Paper (continued)
EFFECTS OF AGE ON DETECTION OF EMOTION
15
Use of parallel construction
with coordinating conjunctions
used in pairs, 3.23
research examines relatively automatic processing. First, in their previous work, Ohman et al.
(2001) compared RTs with both 2 × 2 and 3 × 3 arrays. No significant RT differences based on
the number of images presented in the arrays were found. Second, in both Ohman et al.’s (2001)
study and the present study, analyses were performed to examine the influence of target location
on RT. Across both studies, and across both age groups in the current work, emotional targets
were detected more quickly than were neutral targets, regardless of their location. Together,
these findings suggest that task performance is dependent on relatively automatic detection
Discussion section ending
with comments on
importance of findings, 2.08
processes rather than on controlled search processes.
Although further work is required to gain a more complete understanding of the agehe early processing of emotional information, our findings indicate that
related changes in the
ults are similar in their early detection of emotional images. The current
young and older adults
EFFECTS OF AGE ON DETECTION OF EMOTION
her evidence that mechanisms associated with relatively automatic processing
study provides further
the life span
of emotional imagess are well maintained throughout the latter portion of References
16
Construction of an accurate and
complete reference list, 6.22;
General desciption of references, 2.11
3 Leclerc
(Fleischman et al., 2004; Jennings
& Jacoby,
& Hess,
2005). It
critical
that, dynamics supporting awareness.
Anderson,
A. K.1993;
(2005).
Affective
influences
onisthe
attentional
idence for a positive
focusofinExperimental
older adults’ controlled
processing
emotional
although there is evidence
Journal
Psychology:
General,of
154,
258– 281. doi:10.1037/00962005; Charles et al., 2003; Mather & Knight, 2005), the
information (e.g., Carstensen & Mikels,
3445.134.2.258
est that theAnderson,
tendency to
on the positive
doesD.,
notDe
always
present results suggest
A. focus
K., Christoff,
K., Panitz,
Rosa ,arise
E., &when
Gabrieli, J. D. E. (2003). Neural
ely automatic and correlates
rapid detection
information
in the environment.
tasks require relatively
of theof
automatic
processing
of threat facial signals. Journal of Neuroscience,
23, 5627– 5633.
Armony, J. L., & Dolan, R. J. (2002). Modulation of spatial attention by fear-conditioned
stimuli: An event-related fMRI study. Neuropsychologia, 40, 817–826.
doi:10.1016/S0028-3932%2801%2900178-6
Beck, A. T., Epstein, N., Brown, G., & Steer, R. A. (1988). An inventory for measuring clinical
anxiety: Psychometric properties. Journal of Consulting and Clinical Psychology, 56,
893– 897. doi:10.1037/0022-006X.56.6.893
Calvo, M. G., & Lang, P. J. (2004). Gaze patterns when looking at emotional pictures:
Motivationally biased attention. Motivation and Emotion, 28, 221–243. doi:
10.1023/B%3AMOEM.0000040153.26156.ed
Carretie, L., Hinojosa, J. A., Martin-Loeches, M., Mecado, F., & Tapia, M. (2004). Automatic
attention to emotional stimuli: Neural correlates. Human Brain Mapping , 22, 290–299.
doi:10.1002/hbm.20037
Carstensen, L. L. (1992). Social and emotional patterns in adulthood: Support for socioemotional
selectivity theory. Psychology and Aging, 7, 331–338. doi:10.1037/0882-7974.7.3.331
Carstensen, L. L., Fung, H., & Charles, S. (2003). Socioemotional selectivity theory and the
regulation of emotion in the second half of life. Motivation and Emotion , 27, 103– 123.
29
Return to Inside-out procedure
Figure 2.1. Sample One-Experiment Paper (continued)
EFFECTS OF AGE ON DETECTION OF EMOTION
17
Carstensen, L. L., & Mikels, J. A. (2005). At the intersection of emotion and cognition: Aging
and the positivity effect. Current Directions in Psychological Science, 14, 117–121. doi:
10.1111/j.0963-7214.2005.00348.x
ather, M., & Carstensen, L. L.. (2003). Aging and emotional memory: The
Charles, S. T., Mather,
forgettablee nat
nature of negative images for older adults. Journal of Experimental
DETECTION OF EMOTION
y: G
Psychology:
General,EFFECTS
132, 310–OF
324.AGE
doi: ON
10.1037/0096-3445.132.2.310
18
Chow, T. W., & Cum
Cummings, J. L. (2000). The amygdala and Alzheimer’s disease. In J. P.
Grühn, D., Smith, J., & Baltes, P. B. (2005). No aging bias favoring memory for positive
Aggleton (Ed.
(Ed.), The amygdala: A functional analysis (pp. 656– 680). Oxford, England:
material: Evidence from a heterogeneity-homogeneity list paradigm using emotionally
niver
Oxford University
Press.
toned words. Psychology and Aging, 20, 579–588. doi:10.1037/0882-7974.20.4.579
Davis, M., & Wha
alen P. J. (2001). The amygdala: Vigilance and emotion. Molecular Psychiatry,
Whalen,
Hahn, S., Carlson, C., Singer, S., & Gronlund, S. D. (2006). Aging and visual search: Automatic
Automa
6, 13–
3 34. doi:
d 10.1038/sj.mp.4000812
13–34.
and controlled attentional bias to threat faces. Acta Psychologica, 123 , 312–
2 336. doi:
312–336.
Dolan, R.
J., & Vu
uille
Vuilleumier,
P. (2003).
Amygdala
EFFECTS
OF AGE ON
DETECTION
OFautomaticity
EMOTION in emotional processing. Annals
10.1016/j.actpsy.2006.01.008
19
of the New
w Yo
York Academy of Sciences, 985, 348–355.
Hansen, C. H., & Hansen , R. D. (1988). Finding the face in the crowd: An anger superiority
Kensinger, E. A., Brierley, B., Medford , N., Growdon, J. H., & Corkin, S. (2002). Effects of
Digital object identifier as
Eastwood, J. D., Smilek,
S
Smil
D., & Merikle, P. M. (2001). Negative facial expression captures
effect. Journal of Personality and Social Psychology , 54, 917–924.
917–
7 924. doi: 10.1037/0022normal aging and Alzheimer’s disease on emotional memory. Emotion, 2, 118– 134. doi: article identifier, 6.31;
attention an
nd ddisrupts performance. Perceptual Physiology, 65, 352–358.
and
Example of reference to a
3514.54.6.917
10.1037/1528-3542.2.2.118
periodical, 7.01
Fleischman, D. A.,
A. , Wilson,
W
R. S., Gabrieli, J. D. E., Bienias, J. L., & Bennett , D. A. (2004). A
Hedden, T., & Gabrieli,i J. D. E. (2004). Insights into the ageing mind: A view from cognitive
Lang, P. J., Bradley, M. M., & Cuthbert, B. N. (1997). Motivated attention: Affect, activation,
longitudina
al st
longitudinal
study of implicit and explicit memory in old persons. Psychology and Aging,
87–
7 96. doi:10.1038/nrn1323
neuroscience. Nature Reviews Neuroscience, 5, 87–96.
and action. In P. J. Lang, R. F. Simons, & M. Balaban (Eds.), Attention and orienting:
19, 617–
7 6225. ddoi: 10.1037/0882-7974.19.4.617
617–625.
Example
of reference
to a
Isaacowitz, D. M., Wadlinger, H. A., Goren, D., & Wilson , H. R.. (2006). Selective
preference
in
Sensory and motivational processes (pp. 97– 135). Mahwah, NJ: Erlbaum.
book chapter, print verison,
Good, C. D., John
srud I. S., Ashburner, J., Henson, R. N. A., Firston, K. J., & Frackowiak, R.
Johnsrude,
eyee trackingno
Psycholo
visual fixation away from negative images in old age? An eye-tracking
study.
DOI,Psychology
7.02, Example 25
Leclerc, C. M., & Hess, T. M. (2005, August). Age differences in processing of affectively
S. J.. (2001)
). A voxel-based morphometric study of ageing in 465 normal adult human
(2001).
0 48. doi:10.1037/08822 7974.21.1.40
of Aging, 21, 40–
doi:10.1037/0882-7974.21.1.40
primed information. Poster session presented at the 113th Annual Convention of the
brains. Neu
uroI
NeuroImage,
14, 21–36. doi:10.1006/nimg.2001.0786
Jennings, J. M., & Jacoby, L. L.. (1993). Automatic versus intentional uses of memory: Aging,
American Psychological Association, Washington, DC.
3 293. doi:10.1037/0882attention, and control. Psychology and Aging, 8, 283–
283–293.
LeDoux, J. E. (1995). Emotion: Clues from the brain. Annual Review of Psychology, 46, 209–
7974.8.2.283
235. doi:10.1146/annurev.ps.46.020195.001233
Juth, P., Lundqvist, D., Karlsson, A., & Ohman, A. (2005). Looking for foes and friends:
Mather, M ., & Knight, M. (2005). Goal-directed memory: The role of cognitive control in older
9 39
Perceptual and emotional factors when finding a face in the crowd. Emotion, 5, 379–
395.
adults’ emotional memory. Psychology and Aging, 20, 554–570. doi:10.1037/0882-
7974.20.4.554
doi:10.1037/1528-3542.5.4.379
Kennedy, Q., Mather, M., & Carstensen, L. L.. (2004). The role of motivation in the age-related
age -relate
Mather, M., & Knight, M. R. (2006). Angry faces get noticed quickly: Threat detection is not
positive bias in autobiographical memory. Psychological Science, 15, 208–
8 214.
208–214.
impaired among older adults. Journals of Gerontology, Series B: Psychological Sciences,
61B, P54–P57.
doi:10.1111/j.09566 7976.2004.01503011.x
doi:10.1111/j.0956-7976.2004.01503011.x
Mogg, K., Bradley, B. P., de Bono , J., & Painter, M. (1997). Time course of attentional bias for
threat information in non-clinical anxiety. Behavioral Research Therapy, 35, 297– 303.
Nelson, H. E. (1976). A modified Wisconsin card sorting test sensitive to frontal lobe defects.
Cortex, 12, 313–324.
30
Return to Inside-out procedure
Figure 2.1. Sample One-Experiment Paper (continued)
EFFECTS OF AGE ON DETECTION OF EMOTION
20
Nummenmaa, L., Hyona , J., & Calvo, M. G. (2006). Eye movement assessment of selective
attentional capture by emotional pictures. Emotion, 6, 257–268. doi:10.1037/1528-
Article with more than
3542.6.2.257
seven authors, 7.01,
Ohman, A., Flykt, A., & Esteves, F. (2001). Emotion drives attention: Detecting the snake in the
Example 2
7/0096
grass. Journal of Experimental Psychology: General, 130, 466–478. doi:10.1037/0096-
EFFECTS OF AGE ON DETECTION OF EMOTION
3445.130.3.466
21
modu
Ohman, A., & Mineka, S. (2001). Fears, phobias, and preparedness: Toward an evolvedd module
Rosler, A., Ulrich, C., Billino, J., Sterzer, P., Weidauer , S., Bernhardt, T., …Kleinschmidt, A.
3of fear and fear learning. Psychological Review, 108, 483–522. doi:10.1037/0033(2005). Effects of arousing emotional scenes on the distribution of visuospatial attention:
295X.108.3.483
Changes with aging and early subcortical vascular dementia. Journal of the Neurological
orphol
Ohnishi, T., Matsuda, H., Tabira, T., Asada, T., & Uno, M. (2001). Changes in brain morphology
Sciences,229, 109–116. doi:10.1016/j.jns.2004.11.007
in Alzheimer’s disease and exaggerated aging process? American Journal of
Shipley, W. C. (1986).Shipley Institute of Living Scale. Los Angeles, CA: Western Psychological
Services.
Neuroradiology, 22, 1680–1685.
emporaalPetrides, M., & Milner, B. (1982). Deficits on subject-ordered tasks after frontal- and temporalSpielberger, C. D., Gorsuch, I., & Lushene, R. E. (1970). Manual for the State–Trait Inventory.
lobe lesions in man. Neuropsychologia, 20, 249–262. doi: 10.1016/0028Palo Alto, CA: Consulting Psychologists Press.
EFFECTS OF AGE ON DETECTION OF EMOTION
3932%2882%2990100-2
). Wechslerr Memory Scale
— Revised. San Antonio, TX: Psychological
Wechsler, D. (1987).
Scale—
Corporation..
22
Placement and format
negati
ve
Pratto, F., & John, O. P. (1991). Automatic vigilance:Footnotes
The attention-grabbing power of of
negative
footnotes,
2.12
1
Analyses Journal
of covariance
were conducted
with
these covariates,
with nodoi:
resulting
social information.
of Personality
and Social
Psychology,
61, 380–391.
). Technical manual for the Wechsler Adult Intelligence and Memory Scale–
Wechsler, D. (1997).
10.1037/0022-3514.61.3.380
influences of these variables on the pattern or magnitude of the results.
rk, NY: The Psychological Corporation.
III. New York,
2 These
of of target
Raz, N. (2000). Aging
ofdata
the brain
and its
impactwith
on cognitive
performance:
Integration
were also
analyzed
a 2 × 5 ANOVA
to examine
the effect
West, R. L. (1996). An application of prefrontal cortex function theory to cognitive aging.
andboo
ok
structural
functional
findings.
F. I. containing
M. Craik &neutral
T. A. Salthouse
(Eds.),
Handbook
categoryand
when
presented
only in In
arrays
images, with
the results
remaining
cal Bulletin, 120, 272–
2 292. doi: 10.1037/0033-2909.120.2.272
Psychological
272–292.
ofqualitatively
aging and cognition
ed.,
pp. 1–90).
Mahwah,
NJ: Erlbaum.
the same.(2nd
More
broadly,
the effects
of emotion
on target detection were not
athews , A., & MacLeod, C., (1996). The emotional Stroop task and
Williams, J. M., Mathews
mation..
Reimann,qualitatively
B., & McNally,
R. (1995).
impacted
by theCognitive
distractorprocessing
category. of personally relevant information.
ology. Psychological Bulletin, 120, 3–
3 24. doi: 10.1037/0033-2909.120.1.3
psychopathology.
Cognition and Emotion, 9, 324–340.
erman, N., Burgess, P. W., Emslie,
e H. C., & Evans, J. J. (1996). The
Wilson, B. A., Alderman,
Behaviourall Assessment of the Dysexecutive Syndrome. Flempton, Bury St. Edmunds,
hames Valley Test Company.
England: Thames
31
Return to Inside-out procedure
Figure 2.1. Sample One-Experiment Paper (continued)
EFFECTS OF AGE ON DETECTION OF EMOTION
23
Table 1
Participant Characteristics
Younger group
Measure
M
SD
Years of education
13.92
1.28
Beck Anxiety Inventory
9.39
5.34
BADS– DEX
20.79
7.58
Selecting effective
STAI–State
45.79
4.44
STAI–Trait
45.64
4.50
presentation, 4.41;
Digit Symbol Substitution
49.62
7.18
Logical and effective
Generative naming
46.95
9.70
table layout, 5.08
Vocabulary
33.00
3.52
Digit Span– Backward
8.81
2.09
Arithmetic
16.14
2.75
Mental Control
32.32
3.82
Self-Ordered Pointing
1.73
2.53
CTS OF AGE
ON
DETECTION
OF
EMOTION
EFFECTS
WCST perseverative errors
0.36
0.66
Table 2
Older group
M
SD
16.33
2.43
6.25
6.06
13.38
8.29
47.08
3.48
45.58
3.15
31.58
6.56
47.17
12.98
35.25
3.70
8.25
2.15
14.96
3.11
23.75
5.13
9.25
9.40
1.83
3.23
F (1, 46)
18.62
3.54
10.46
1.07
0.02
77.52
.004
4.33
0.78
1.84
40.60
13.18
4.39
24
p
<.001
.066
.002
.306
.963
<.001
.951
.043
.383
.182
<.001
.001
.042
Note. The Beck Anxiety Inventory is from Beck et al. (1988); the Behavioral Assessment of the
Dysexecutive
Questionnaire (BADS–DEX) is from Wilson et al.
esponse Time
for Young and Older Adults
Raw Response
(RT) ScoresSyndrome—Dysexecutive
(1996); the State–Trait Anxiety Inventory (STAI) measures are from Spielberger et al. (1970);
ory
y
Young
gg
group
p
Older group
g p
Category
and the Digit Symbol
ve high arousal
825Substitution, Digit Span–Backward,
1,580 and Arithmetic Wechsler Adult
Positive
ve low arousal
899and Wechsler Memory Scale—III
1,636 measures are from Wechsler (1997).
Positive
Intelligence Scale—III
al
912
1,797
Neutral
ive high arousal
1,578
Negative
Generative naming 885
scores represent the total number
of words produced in 60 s each for letter
ive low arousal
896
1,625
Negative
F, A, and S. The Vocabulary measure is from Shipley (1986); the Mental Control measure is
Note. Values represent median response times, collapsing across array type and excluding arrays
from Wechsler (1987); the Self-Ordered Pointing measure was adapted from Petrides and Milner
of the same category as targets (i.e., positive high arousal represents the median RT to respond to
(1982); and the Wisconsin Card Sorting Task (WCST) measure is from Nelson (1976).
positivee high arousal targets, collapsing across positive low arousal, neutral, negative high
All values represent raw, nonstandardized scores.
arousal,, and negative low arousal array categories). The median response time values were
Elements of
table notes, 5.16
ed in milliseconds.
recorded
32
Return to Inside-out procedure
Figure 2.1. Sample One-Experiment Paper (continued)
EFFECTS OF AGE ON DETECTION OF EMOTION
25
Principles of figure use and
construction, types of figures;
standards, planning, and
preparation of figures, 5.20–5.25
.
Figure 1. Mean difference values (ms) representing detection speed for each target category
subtracted from the mean detection speed for neutral targets. No age differences were found in the
arousal-mediated effects on detection speed. Standard errors are represented in the figure by the
Figure legends
and captions, 5.23
error bars attached to each column.
33
Return to Inside-out procedure
Figure 2.2. Sample Two-Experiment Paper (The numbers refer to numbered sections in the Publication Manual. This abridged manuscript illustrates the organizational structure characteristic of
multiple-experiment papers. Of course, a complete multipleexperiment paper would include a title page, an abstract page,
and so forth.)
INHIBITORY INFLUENCES ON ASYCHRONY
1
Inhibitory Influences on Asychrony as a Cue for Auditory Segregation
Auditory grouping involves the formation of auditory objects from the sound mixture
reaching the ears. The cues used to integrate or segregate these sounds and so form auditory
objects have been defined by several authors (e.g., Bregman, 1990; Darwin, 1997; Darwin &
Carlyon, 1995). The key acoustic cues for segregating concurrent acoustic elements are
differences in onset time (e.g., Dannenbring & Bregman, 1978; Rasch, 1978) and harmonic
relations (e.g., Brunstrom & Roberts, 1998; Moore, Glasberg, & Peters, 1986). In an example of
the importance of onset time, Darwin (1984a, 1984b) showed that increasing the level of a
harmonic near the first formant (F1) frequency by adding a synchronous pure tone changes the
phonetic quality of a vowel. However, when the added tone began a few hundred milliseconds
before the vowel, it was essentially removed from the vowel percept.… [section continues].
General Method
Elements of empirical studies, 1.01
Overview
In the experiments reported here, we used a paradigm developed by Darwin to assess the
perceptual integration of additional energy in the F1 region of a vowel through its effect on
phonetic quality (Darwin, 1984a, 1984b; Darwin & Sutherland, 1984).…[section continues].
Stimuli
Amplitude and phase values for the vowel harmonics were obtained from the vocal-tract
transfer function using cascaded formant resonators (Klatt, 1980). F1 values varied in 10-Hz
steps from 360–550 Hz—except in Experiment 3, which used values from 350– 540 Hz—to
produce a continuum of 20 tokens.…[section continues].
Listeners
Paper adapted from “Inhibitory Influences on Asychrony as a Cue for Auditory Segregation,” by S. D.
Holmes and B. Roberts, 2006, Journal of Experimental Psychology: Human Perception and Performance, 32,
pp. 1231–1242. Copyright 2006 by the American Psychological Association.
34
Return to Inside-out procedure
Figure 2.2. Sample Two-Experiment Paper (continued)
INHIBITORY INFLUENCES ON ASYCHRONY
2
Listeners were volunteers recruited from the student population of the University of
Birmingham and were paid for their participation. All listeners were native speakers of British
English who reported normal hearing and had successfully completed a screening procedure
(described below). For each experiment, the data for 12 listeners are presented.…[section
Plural forms of nouns
of foreign origin, 3.19
continues].
Procedure
At the start of each session, listeners took part in a warm-up block. Depending on the
number of conditions in a particular experiment, the warm-up block consisted of one block of all
the experimental stimuli or every second or fourth F1 step in that block. This gave between 85
and 100 randomized trials. … [section continues].
Data Analysis
The data for each listener consisted of the number of /I/ responses out of 10 repetitions
for each nominal F1 value in each condition. An estimate of the F1 frequency at the phoneme
boundary was obtained by fitting a probit function (Finney, 1971) to a listener ’s identification
data for each condition. The phoneme boundary was defined as the mean of the probit function
(the 50% point).…[section continues].
Multiple Experiments, 2.09
Experiment
pe
e t1
iment, we used noise-band captors and compared their efficacy with that of a
In this experiment,
ASYCHRONY
ach noise-band
- INHIBITORY
captor had INFLUENCES
the same energyON
as that
of the corresponding pure
pure-tone captor. Each
pure--
3
nter frequency equal to the frequency of this tonal captor…[section
tone captor and a center
There were nine conditions: the three standard ones (vowel alone, incremented fourth,
continues].
Method
and leading fourth) plus three captor conditions and their controls. A lead time of 240 ms was
Abbreviating units
of measurement,
4.27, Table 4.4
used for the added 500-Hz tone.… [section continues].
Results and Discussion
Figure 4 shows the mean phoneme boundaries for all conditions and the restoration effect
for each captor type. The restoration effects are shown above the histogram bars both as a
boundary shift in hertz and as a percentage of the difference in boundary position between the
incremented-fourth and leading-fourth conditions.… [section continues].
Experiment 2
This experiment considers the case where the added 500-Hz tone begins at the same time
as the vowel but continues after the vowel ends.… [section continues].
Method
There were five conditions: two of the standard ones (vowel alone and incremented
Policy on metrication, 4.39; fourth), a lagging-fourth condition (analogous to the leading-fourth condition used elsewhere),
Style for metric units, 4.40 and a captor condition and its control. A lag time of 240 ms was used for the added 500-Hz
tone.… [section continues]
Results and Discussion
35
Return to Inside-out procedure
Figure 2.2. Sample Two-Experiment Paper (continued)
INHIBITORY INFLUENCES ON ASYCHRONY
4
1984; Roberts & Holmes, 2006). This experiment used a gap between captor offset and vowel
onset to measure the decay time of the captor effect …[section continues].
Method
There were 17 conditions: the three standard ones (vowel alone, incremented fourth, and
leading fourth), five captor conditions and their controls, and four additional conditions
(described separately below). A lead time of 320 ms was used for the added 500-Hz tone. The
captor conditions were created by adding a 1.1-kHz pure-tone captor, of various durations, to
each member of the leading-fourth continuum.…[section continues].
Results
Figure 6 shows the mean phoneme boundaries for all conditions. There was a highly
Use of statistical term rather
than symbol in text, 4.45
significant effect of condition on the phoneme boundary values, F(16, 176) = 39.10, p < .001.
Incrementing the level of the fourth harmonic lowered the phoneme boundary relative to the
vowel-alone condition (by 58 Hz, p < .001), which indicates that the extra energy was integrated
into the vowel percept.…[section continues].
Discussion
The results of this experiment show that the effect of the captor disappears somewhere
between 80 and 160 ms after captor offset. This indicates that the captor effect takes quite a long
time to decay away relative to the time constants typically found for cells in the CN using
physiological measures (e.g., Needham & Paolini, 2003).…[section continues].
Summary and Concluding Discussion
Darwin and Sutherland (1984) first demonstrated that accompanying the leading portion
of additional energy in the F1 region of a vowel with a captor tone partly reversed the effect of
the onset asynchrony on perceived vowel quality. This finding was attributed to the formation of
Running head: INHIBITORY INFLUENCES ON ASYCHRONY
5
a perceptual group between the leading portion and the captor tone, on the basis oof their common
onset time and harmonic relationship, leaving the remainder of the extra energy to integrate into
the vowel percept… .[section continues].
[Follow the form of the one-experiment sample paper to type references, the author note,
footnotes, tables, and figure captions.]
36
Return to Inside-out procedure
Figure 2.3. Sample Meta-Analysis (The numbers refer to numbered sections in the Publication Manual. This abridged manuscript illustrates the organizational structure characteristic of reports of
meta-analyses. Of course, a complete meta-analysis would
include a title page, an abstract page, and so forth.)
THE SLEEPER EFFECT IN PERSUASION
1
The Sleeper Effect in Persuasion:
A Meta-Analytic Review
Persuasive messages are often accompanied by information that induces suspicions of
invalidity. For instance, recipients of communications about a political candidate may discount a
message coming from a representative of the opponent party because they do not perceive the
source of the message as credible (e.g., Lariscy & Tinkham, 1999). Because the source of the
political message serves as a discounting cue and temporarily decreases the impact of the
message, recipients may not be persuaded by the advocacy immediately after they receive the
communication. Over time, however, recipients of an otherwise influential message may recall
the message but not the noncredible source and thus become more persuaded by the message at
Italicize key terms, 4.21
that time than they were immediately following the communication. The term sleeper effect was
used to denote such a delayed increase in persuasion observed when the discounting cue (e.g.,
or “dissociated”
the communication in the
noncredible source) becomes unavailable
THE SLEEPER
EFFECT INfrom
PERSUASION
2
sage recipients (Hovland, Lumsdaine, & Sheffield, 1949).…[section
memory of the message
retention, attitude and decay, and persuasion and decay . Because researchers often use the terms
opinion and belief, instead of attitude, we conducted searches using these substitute terms as
Method
well.
Description of meta-analysis, 1.02;
Sample of Studies
Guidelines for reporting meta-analysis,
Second, … [section continues].
2.10;
d reports related to the sleeper effect that were available by March
2003see
by also Appendix
We retrieved
Selection Criteria
rocedures. First, we searched computerized databases, including PsycINFO
means of multiple procedures.
We used the following criteria to select studies for inclusion in the meta-analysis.
rtation Abstracts International (1861– 2003), ERIC (1967–
7 2003), and the
(1887–2003), Dissertation
(1967–2003),
1. We only included studies that involved the presentation of a communication containing
tion-Index (1956–
6 2003), using the keywords sleeper effect,t delayed-action,
Social-Science-Citation-Index
(1956–2003),
persuasive arguments. Thus, we excluded studies in which the participants played a role or were
redibility,
source
expertise, attitude change, discounting cue, attitude
credibility, source credibility,
asked to make a speech that contradicted their opinions. We also excluded developmental studies
r , attitude and
persistence, attitude maintenance, persuasion, propaganda, attitude and memory
memory,
involving delayed effects of an early event (e.g., child abuse), which sometimes are also referred
to as sleeper effects.…[section continues] .
Identification of elements in a
series within a sentence, 3.04
Moderators
For descriptive purposes, we recorded (a) the year and (b) source (i.e., journal article,
unpublished dissertations and theses, or other unpublished document) of each report as well as
(c) the sample composition (i.e., high-school students, university students, or other) and (d) the
country in which the study was conducted.
We also coded each experiment in terms of .…[section continues].
Studies were coded independently by the first author and another graduate student.
Paper adapted from “The Sleeper Effect in Persuasion: A Meta-Analytic Review,” by G. Kumkale and D.
Albarracin, 2004, Psychological Bulletin, 130, pp. 143–172. Copyright 2004 by the American Psychological
Association.
37
Return to Inside-out procedure
Figure 2.3. Sample Meta-Analysis (continued)
THE SLEEPER EFFECT IN PERSUASION
3
was satisfactory (Orwin, 1994). We resolved disagreements by discussion and consultation with
colleagues. Characteristics of the individual studies included in this review are presented in
Table 1. The studies often contained several independent datasets such as different messages and
different experiments. The characteristics that distinguish different datasets within a report
appear on the second column of the table.
Dependent Measures and Computation of Effect Sizes
We calculated effect sizes for (a) persuasion and (b) recall–recognition of the message
content. Calculations were based on the data described in the primary reports as well as available
responses of the authors to requests of further information.…[section continues].
Analyses of Effect Sizes
wo major models used in meta-analysis:
metaa analysis: fixed-effectss and randomThere are two
ontinues]. THE SLEEPER EFFECT IN PERSUASION
effects.…[section continues].
4
Use at least
two subheadings
in a section, 3.02
To benefit from thee strengths place
of both
models,
we chose to
aggregate the effect sizes and to
over
time.…[section
continues].
ing both approaches.…[section
conduct analyses using
continues].
In light of these requirements,
we first examined whether discounting cues led to a decrease in
agreement withResults
the communication (boomerang effect). Next,.…[section continues].
alysis included a description
The data analysis
experimentsboomerang
we summarized,
an To determine whether or not a delayed
Ruling outofathe
nonpersisting
effect.
estimation of overalll effects, moderator
andrepresents
tests of mediation.
increase inanalyses,
persuasion
an absolute sleeper effect, one needs to rule out a nonpersisting
Sample of Studies and Datasets
boomerang effect, which takes place when a message initially backfires but later loses this
metaa analysis
appear in
Descriptive characteristics
of the
datasets
included
the present
meta-analysis
reverse
effect
(see panel
A ofinFigure
1).…[section
continues].
Table 2.…[section continues].
Average sleeper effect. Relevant statistics corresponding to average changes in
verage Effect
Overview of the Average
Sizes from the immediate to the delayed posttest appear in Table 4, organized by the
persuasion
A thorough understanding
of theconditions
sleeper effect
requires examining
(a) the betweendifferent
we considered
(i.e., acceptance-cue,
discounting-cue, no-message control,
es at each time
condition differences
point as well as
(b) the within-condition
changes
and message-only
control).
In Table 4, positive
effect that
sizestake
indicate increases in persuasion over
time, negative effect sizes indicate decay in persuasion, and zero effects denote stability in
persuasion. Confidence intervals that do not include zero indicate significant changes over time.
The first row of Table 4 shows that recipients of acceptance cues agreed with the message less as
time went by (fixed-effects, d + = –0.21; random-effects, d + = –0.23). In contrast to the decay in
persuasion for recipients of acceptance cues, there was a slight increase in persuasion for
recipients of discounting cues over time (d + = 0.08). It is important to note that change in
discounting-cue conditions significantly differed from change in acceptance-cue conditions,
(fixed-effects; B = –0.29, SE = 0.04), QB (1) = 58.15, p < .0001; QE(123) = 193.82, p <
.0001.…[section continues].
Summary and variability of the overall effect. The overall analyses identified a relative
sleeper effect in persuasion, but no absolute sleeper effect. The latter was not surprising, because
the sleeper effect was expected to emerge under specific conditions.…[section continues].
38
Return to Inside-out procedure
Figure 2.3. Sample Meta-Analysis (continued)
THE SLEEPER EFFECT IN PERSUASION
5
Moderator Analyses
Although overall effects have descriptive value, the variability in the change observed in
discounting-cue conditions makes it unlikely that the same effect was present under all
conditions. Therefore, we tested the hypotheses that the sleeper effect would be more likely (e.g.,
more consistent with the absolute pattern in Panel B1 of Figure 1) when…[section continues].
Format for references included
in a meta-analysis with less
than 50 references, 6.26
THE SLEEPER EFFECT IN PERSUASION
6
References
References marked with an asterisk indicate studies included in the meta-analysis.
Albarracín, D. (2002). Cognition in persuasion: An analysis of information processing in
response to persuasive communications. In M. P. Zanna (Ed.), Advances in experimental
social psychology (Vol. 34, pp. 61–130). doi:10.1016/S0065-2601(02)80004-1
… [references continue]
Johnson, B. T., & Eagly, A. H. (1989). Effects of involvement in persuasion: A meta-analysis.
Psychological Bulletin, 106, 290–314. doi:10.1037/0033-2909.106.2.290
*Johnson, H. H., Torcivia, J. M., & Poprick, M. A. (1968). Effects of source credibility on the
relationship between authoritarianism and attitude change. Journal of Personality and
Social Psychology, 9, 179–183. doi:10.1037/h0021250
*Johnson, H. H., & Watkins, T. A. (1971). The effects of message repetitions on immediate and
delayed attitude change. Psychonomic Science, 22, 101–103.
Jonas, K., Diehl, M., & Bromer, P. (1997). Effects of attitudinal ambivalence on information
processing and attitude-intention consistency. Journal of Experimental Social
Psychology, 33, 190–210. doi:10.1006/jesp.1996.1317
. . . [references continue]
[Follow the form of the one-experiment sample paper to type the author note, footnotes,
tables, and figure captions.]
39
Return to Inside-out Procedure
Part III: Examples Psychonomic Bulletin & Review
2008, 15 (2), 404-408
doi: 10.3758/PBR.15.2.404
Visual events modulated by sound
in repetition blindness
Yi-Chuan Chen and Su-Ling Yeh
National Taiwan University, Taipei, Taiwan
Repetition blindness (RB; Kanwisher, 1987) is the term used to describe people’s failure to detect or report an
item that is repeated in a rapid serial visual presentation (RSVP) stream. Although RB is, by definition, a visual deficit, whether it is affected by an auditory signal remains unknown. In the present study, we added two sounds before,
simultaneous with, or after the onset of the two critical visual items during RSVP to examine the effect of sound on
RB. The results show that the addition of the sounds effectively reduced RB when they appeared at, or around, the
critical items. These results indicate that it is easier to perceive an event containing multisensory information than
unisensory ones. Possible mechanisms of how visual and auditory information interact are discussed.
Open Gen Intro (1) General and Familiar opening Also connect with last paragraph
If a wind shear is encountered by an airplane, the captain will turn on the “fasten seatbelt” sign, the onset of
which is usually accompanied by a sound to avoid this
sign’s being ignored. Is an event containing multisensory
information easier to perceive than a unisensory one? In
this study, we examined whether a sound can ameliorate
Open Gen
the unobserved visual event by adopting the repetition
Intro (2):blindness (RB) paradigm.i.e., W1
Write the RB is the failure to report the second occurrence of a
repeated item in a rapid serial visual presentation (RSVP).
researchFor example, the sentence “I prefer green tea for tea time”
goal in may be misreported as “I prefer green tea for time,” even
though such an error leads to an ungrammatical sentence.
first
RB is considered a general phenomenon in visual perception, since it has been demonstrated with many different
paragraph
visual stimuli, including English words (Harris & Morris,
2001), Chinese characters (Yeh & Li, 2004), colors and
forms (Kanwisher, Driver, & Machado, 1995), and pictures (Bavelier, 1994).
By definition, RB is a visual deficit. Nevertheless, it is
still unknown whether the factors affecting RB are confined to the visual modality or whether, instead, stimuli
from other modalities can also influence RB. To date,
there have been no studies in which this issue has been
directly examined, although some previous studies have
provided a few insights. For example, Kanwisher and
Potter (1989) had their participants either see or hear
sentences and found RB but not repetition deafness.
They concluded that RB is a modality-specific phenomenon that is confined to vision (but see Miller & MacKay, 1994). Soto-Faraco and Spence (2002) found no
evidence for a cross-modal repetition deficit when the
two critical items, C1 and C2, were presented in different
modalities (e.g., visual C1 and auditory C2). This implies that the source of modality could be taken as a part
of episodic representations, so that the factors affecting
RB might be restricted to vision only.
Despite the foregoing, other studies have suggested that
the perception of visual events can be affected by auditory
input. For example, two visual streams crossing each other
can be perceived as streaming or bouncing, depending on
whether a sound is added (Sekuler, Sekuler, & Lau, 1997;
Shimojo, Watanabe, & Scheier, 2001). Another example
is the attentional blink (AB) for the visual target when it
is preceded by an auditory target (Arnell & Larson, 2002).
The proposed mechanism is that the auditory stimulus may
occupy attentional resources or distract attention away
from the visual stimuli (Arnell & Larson, 2002; Shimojo
et al., 2001; but see Duncan, Martens, & Ward, 1997). By
inference from these studies, it seems that adding sounds
to an RSVP stream should suppress the visual process,
due to lack of attention.
Recalling the seatbelt example mentioned earlier, however, makes us wonder whether there is a third possibility:
Sound enhances the visibility of the originally unobserved
visual event (i.e., enhances the visibility of C2 when it suffers from RB). To examine this, we added two different
beep sounds (S1 and S2) to the visual stream.1 If the added
sounds can enhance the visibility of C1 and C2 in RSVP,
they should be perceived more accurately, thus lessening
RB. The onset time of S2 was manipulated in order to probe
the effective time window of the sound facilitatory effect.
Method
Participants Method for Exp (1): Typical method section
Five groups of 20 undergraduates at National Taiwan University
participated in each of the five sound conditions in exchange for
course credit. They were all native speakers of Mandarin and were
naive concerning the purpose of the experiment. All of the participants had normal or corrected-to-normal vision and hearing.
S.-L. Yeh, [email protected]
Copyright 2008 Psychonomic Society, Inc.
404
40
Repetition Blindness Reduced by Sound 405
A
Repeated trial
Unrepeated trial
+
C1
ƌ ᄁ
F
Ҽ $
ჩ Ҽ
Ք
= ᄁ % +
ᄁ
F
Control trial
C2
C2
Ĺ
༮
43 msec/item
B
No sound
–86 msec
Onset
Offset
+86 msec
–86
0
+86
: S1
Method for Exp (2): Complete the figure
msec
: S2
Figure 1. (A) Examples of the rapid serial visual presentation sequence. In
the repeated trial, C1 and C2 were the same character (銘). In the unrepeated
trial, all the displays were the same as those in the repeated trial, except that
another character (輔) was substituted for C1. The control trial contained
only two characters. (B) The five sound conditions. The time scales of panels A
and B are matched. C1, Critical Item 1; C2, Critical Item 2; F, filler character; S1, Sound 1, 500 Hz; S2, Sound 2, 800 Hz.
Stimuli
The stimuli were presented on a 15-in. calibrated EIZO color
monitor and were controlled by a personal computer. The refresh
rate of the monitor was set at 70 Hz. The participants sat at a viewing
distance of 60 cm from the monitor in a dimly lit chamber.
Each trial consisted of seven items in an RSVP sequence, with
three Chinese characters and four symbols all in white and presented
one at a time for 43 msec in the center of a black background. All
the Chinese characters that were selected were ones commonly used.
The three characters included one filler character (F) and two critical
characters (C1 and C2). To avoid any unwanted priming effects, the
characters in each trial were not orthographically similar, except for
repeated items, and did not have any obvious semantic relationship.
Furthermore, the pronunciations did not sound as meaningful twoor three-character words, nor did they rhyme.
The F and C2 were in the Chia font and subtended a visual angle
of 1.15º  1.24º (width  height). C1 was in the Fong font (1.53º 
1.43º). C1 and C2 were chosen to be different in both font and size
in order to avoid the possible confounding of visual masking caused
by physically identical characters appearing at the same position.
A fixation cross (1.15º  1.24º) was shown at the beginning and
41
end of each trial. Symbols (＊) in each trial were randomly selected
from a set of 30 signs, such as &, ▽, ≒, and $, without repeated
symbols occurring in the same trial. The sizes of the 30 symbols
ranged from 1.21º  0.80º to 1.43º  1.24º. In all the trials, the
presented order in RSVP was＊, C1, F,＊,＊, C2,＊(see Figure 1A).
The auditory stimuli (S1 and S2; 500 and 800 Hz, respectively) appeared simultaneously with the onsets of two visual items and lasted
for 10 msec each.
Design
The two factors of sound condition (five conditions of beep
sounds accompanying C1 and C2) and repetition (repeated or unrepeated) were manipulated. The five sound conditions were as
follows. The no-sound condition was solely a visual presentation
and served as the baseline for comparison. In the remaining four
conditions, S1 was always presented at the onset of C1, and S2 was
presented before C2 (86 msec), at the onset of C2, at the offset of
C2, or after C2 (186 msec). Except for the onset time of S2, other
items were the same in the four conditions, except for the no-sound
condition (Figure 1B). Different participants were assigned to these
five conditions.
Repetition Blindness Reduced by Sound 407
Discussion
Open Gen Dis (1): State main findings in short paper
In this study, we found that RB can be reduced when
two critical items are accompanied by different sounds.
We replicated the conventional RB effect in the no-sound
condition, while observing a reduced RB effect in the
86-msec, onset, and offset conditions. The 186-msec
condition revealed a magnitude of RB similar to that in
the no-sound condition.
Regarding the three possibilities mentioned earlier of
how sound would affect RB (independent, suppressed, or
enhanced), our results favor the enhanced account, since
the addition of the sounds reduced the magnitude of the
RB effect by facilitating character identification. Even
though processing the RSVP stream is demanding enough
to prevent influences from peripheral sounds (Santangelo,
Belardinelli, & Spence, 2007), in this study, the task­irrelevant sounds were not filtered out but, rather, still affected the demanding processing of the central stream.
However, unlike the cross-modal AB effect (Arnell &
Larson, 2002), we demonstrated that the presence of the
sound enhanced, rather than suppressed, visual perception. Indeed, there is mounting evidence showing sound
facilitatory effects in visual detection (e.g., Bolognini,
Frassinetti, Serino, & Làdavas, 2005; McDonald, TederSälejärvi, & Hillyard, 2000) and visual filling in (Sheth
& Shimojo, 2004). Nevertheless, these studies adopted
a single LED light or a color patch as the visual target.
Our results thus extend the sound facilitatory effect shown
in the literature in that it can be obtained not only in the
simple display, but also in a rapid and complex visual presentation (i.e., RSVP), which is more similar to our daily
visual world.
Bolognini et al. (2005) reported that sound could enhance the sensitivity of visual perception, whereas Park
and Kanwisher (1994) found that C2 sensitivity was reduced when it suffered from RB. Combining these two
findings, we suggest that the effect of added sounds in the
present study was to boost the lowered sensitivity to the
C2, making it more likely to be immune from RB. These
results imply that multisensory stimuli have a better opportunity to gain visual awareness (Kanwisher, 2001) than
do unisensory ones. Our results may thus have important
implications for the improvement of deficits in the perceiving of sequential events for right-parietal-damaged
patients (Battelli, Pascual-Leone, & Cavanagh, 2007).
The processing of visual items in an RSVP stream can
be understood within the type–token binding framework
(e.g., Bowman & Wyble, 2007; Chun, 1997; Kanwisher,
1987), which suggests that each visual event activates
two kinds of representations: type and token. A type is a
long-term representation used for recognition, whereas a
token is an episodic representation used to specify spatial
and temporal information. To successfully report a visual
event, the activated type needs to link to its token, a process called token individuation. Kanwisher (1987) suggested that RB results from the failure to individuate the
second token of the same activated type, due to the difficulty of binding one type with two tokens within a short
interval. This hypothesis has gained much support from
43
behavioral (e.g., Chun, 1997; Morris & Harris, 1999; Park
& Kanwisher, 1994) and event-related potential (Schendan, Kanwisher, & Kutas, 1997) studies. Attention is required either to facilitate the activation of the type node
(Bowman & Wyble, 2007) or to serve as a supplementary
component for type–token binding (Chun, 1997).
The cross-modal facilitatory effect in RB we found
here can be explained as follows. The presentation of the
sounds may covertly orient participants’ attention to focus
on the upcoming visual items (McDonald et al., 2000),
or it may boost the alertness of participants (Robertson,
Mattingley, Rorden, & Driver, 1998); namely, sounds can
modulate visual attention in the temporal domain. Another
possibility is that sounds help temporal segregation of the
RSVP stream, since they are dominant in the temporal
domain; not only does audition have its advantage in the
representations of temporal structure (Guttman, Gilroy, &
Blake, 2005; Welch, DuttonHurt, & Warren, 1986), but
also the discontinuous nature of sounds makes them more
dominant than visual streams (Shimojo & Shams, 2001).
C1 and C2 can be more distinguishable and, thus, suffer
less RB if they are segregated into two groups, rather than
appearing within the same group in the spatial domain
(Chun & Cavanagh, 1997; Kanwisher & Potter, 1989), so
that they are segregated in the temporal domain (MoreinZamir, Soto-Faraco, & Kingstone, 2003). In this case, the
temporal structure of the visual stream and the sounds
should be coordinated, which may occur after the stimuli
within each modality have been fully processed. A further
possibility is visual–auditory integration’s occurring at an
early perceptual stage. The different sounds adopted in
this study may have integrated with the repeated characters and modified their similarity, which is analogous to
the demonstration of the reduction of RB by making C1
and C2 more visually distinct (e.g., Bavelier, 1994; Chun,
1997). Although the temporal coincident rule with a particular time window of cross-modal integration (Stein &
Meredith, 1993) may account not only for the onset condition, but also for the 86-msec and offset conditions, this
integration account, however, presupposes a close connection between S2 and C2. Yet there is no a priori reason why
this should be the case.
Previous studies have shown that sound can improve
attentional deficits for spatial neglect patients through a
nonspecific alertness effect and a spatially specific crossmodal integration effect (Frassinetti, Pavani, & Làdavas,
2002; Van Vleet & Robertson, 2006). Similarly, all of the
above-mentioned mechanisms may be involved in the
cross-modal facilitatory effect in RB, but each may favor
a certain time window. For example, orienting requires
time to develop (Nakayama & Mackeben, 1989), so the
attentional boost account may best explain the 86-msec
condition. On the other hand, temporal coincidence is critical for the cross-modal integration account (Bolognini
et al., 2005). To tease these possible mechanisms apart,
however, manipulations of factors other than the temporal
relationship between sounds and visual items are required,
especially for the cross-modal integration account. For
example, the effect we observed here is enhancement,
rather than a new representation, as in the McGurk ef-
408 Chen and Yeh
fect in speech perception (McGurk & MacDonald, 1976).
Systematic investigations about whether pitch difference
of sounds is critical may help to clarify these issues.
Nevertheless, regardless of mechanism, we have shown
in this study that RB can be reduced by adding task­irrelevant sounds in RSVP. Multisensory stimuli, as compared with unisensory stimuli, have a better chance of
breaking through visual awareness and making invisible
stimuli visible. Hearing a sound accompanying the “fasten
seatbelt” sign may, indeed, serve as a better warning signal
than does seeing the visual notice alone.
Final Gen Dis (7)
Compare with first paragraph
.
Author Note
This research was supported by Grants NSC93-2413-H-002-017 and
NSC94-2752-H-002-008-PAE from the National Science Council of
Taiwan. We thank Catherine Caldwell-Harris, Anne Hillstrom, Steven
Most, Shinsuke Shimojo, and Charles Spence for their helpful comments on earlier drafts of the manuscript. Correspondence concerning
this article should be addressed to S.-L. Yeh, Department of Psychology,
National Taiwan University, No. 1, Sec. 4, Roosevelt Rd., Taipei, Taiwan,
106 (e-mail: [email protected]).
References
Arnell, K. M., & Larson, J. M. (2002). Cross-modality attentional
blinks without preparatory task-set switching. Psychonomic Bulletin
& Review, 9, 497-506.
Battelli, L., Pascual-Leone, A., & Cavanagh, P. (2007). The “when”
pathway of the right parietal lobe. Trends in Cognitive Sciences, 11,
204-210.
Bavelier, D. (1994). Repetition blindness between visually different
items: The case of pictures and words. Cognition, 51, 199-236.
Bolognini, N., Frassinetti, F., Serino, A., & Làdavas, E.
(2005). “Acoustical vision” of below threshold stimuli: Interaction
among spatially converging audiovisual inputs. Experimental Brain
Research, 160, 273-282.
Bowman, H., & Wyble, B. (2007). The simultaneous type, serial token
model of temporal attention and working memory. Psychological Review, 114, 38-70.
Chun, M. M. (1997). Types and tokens in visual processing: A double
dissociation between the attentional blink and repetition blindness.
Journal of Experimental Psychology: Human Perception & Performance, 23, 738-755.
Chun, M. M., & Cavanagh, P. (1997). Seeing two as one: Linking
apparent motion and repetition blindness. Psychological Science, 8,
74-79.
Duncan, J., Martens, S., & Ward, R. (1997). Restricted attentional
capacity within but not between sensory modalities. Nature, 387,
808-810.
Frassinetti, F., Pavani, F., & Làdavas, E. (2002). Acoustical vision
of neglected stimuli: Interaction among spatially converging audiovisual inputs in neglect patients. Journal of Cognitive Neuroscience,
14, 62-69.
Guttman, S. E., Gilroy, L. A., & Blake, R. (2005). Hearing what the
eyes see. Psychological Science, 16, 229-235.
Harris, C. L., & Morris, A. L. (2001). Identity and similarity in repetition blindness: No cross-over interaction. Cognition, 81, 1-40.
Kanwisher, N. (1987). Repetition blindness: Type recognition without
token individuation. Cognition, 27, 117-143.
Kanwisher, N. (2001). Neural events and perceptual awareness. Cognition, 79, 89-113.
Kanwisher, N., Driver, J., & Machado, L. (1995). Spatial repetition
blindness is modulated by selective attention to color or shape. Cognitive Psychology, 29, 303-337.
Kanwisher, N., & Potter, M. C. (1989). Repetition blindness: The
effects of stimulus modality and spatial displacement. Memory &
Cognition, 17, 117-124.
McDonald, J. J., Teder-Sälejärvi, W. A., & Hillyard, S. A. (2000).
Involuntary orienting to sound improves visual perception. Nature,
407, 906-908.
McGurk, H., & MacDonald, J. (1976). Hearing lips and seeing voices.
Nature, 264, 746-748.
Miller, M. D., & MacKay, D. G. (1994). Repetition deafness: Repeated words in computer-compressed speech are difficult to encode
and recall. Psychological Science, 5, 47-51.
Morein-Zamir, S., Soto-Faraco, S., & Kingstone, A. (2003). Auditory capture of vision: Examining temporal ventriloquism. Cognitive
Brain Research, 17, 154-163.
Morris, A. L., & Harris, C. L. (1999). A sublexical locus for repetition
blindness: Evidence from illusory words. Journal of Experimental
Psychology: Human Perception & Performance, 25, 1060-1075.
Nakayama, K., & Mackeben, M. (1989). Sustained and transient components of focal attention. Vision Research, 29, 1631-1647.
Park, J., & Kanwisher, N. (1994). Determinants of repetition blindness. Journal of Experimental Psychology: Human Perception & Performance, 20, 500-519.
Robertson, I. H., Mattingley, J. B., Rorden, C., & Driver, J. (1998).
Phasic alerting of neglect patients overcomes their spatial deficit in
visual awareness. Nature, 395, 169-172.
Santangelo, V., Belardinelli, M. O., & Spence, C. (2007). The
suppression of reflexive visual and auditory orienting when attention
is otherwise engaged. Journal of Experimental Psychology: Human
Perception & Performance, 33, 137-148.
Schendan, H. E., Kanwisher, N. G., & Kutas, M. (1997). Early brain
potentials link repetition blindness, priming and novelty detection.
NeuroReport, 8, 1943-1948.
Sekuler, R., Sekuler, A. B., & Lau, R. (1997). Sound alters visual
motion perception. Nature, 385, 308.
Sheth, B. R., & Shimojo, S. (2004). Sound-aided recovery from and
persistence against visual filling-in. Vision Research, 44, 1907-1917.
Shimojo, S., & Shams, L. (2001). Sensory modalities are not separate
modalities: Plasticity and interactions. Current Opinion in Neurobiology, 11, 505-509.
Shimojo, S., Watanabe, K., & Scheier, C. (2001). The resolution
of ambiguous motion: Attentional modulation and development. In
J. Braun, C. Koch, & J. L. Davis (Eds.), Visual attention and control
circuits (pp. 243-264). Cambridge, MA: MIT Press.
Soto-Faraco, S. (2000). An auditory repetition deficit under low memory load. Journal of Experimental Psychology: Human Perception &
Performance, 26, 264-278.
Soto-Faraco, S., & Spence, C. (2002). Modality-specific auditory and
visual temporal processing of deficits. Quarterly Journal of Experimental Psychology, 55A, 23-40.
Stein, B. E., & Meredith, M. A. (1993). The merging of the senses.
Cambridge, MA: MIT Press.
Thalheimer, W., & Cook, S. (2002). How to calculate effect sizes from
published research articles: A simplified methodology. Retrieved
May 22, 2006 from work-learning.com/effect_sizes.htm.
Tsai, C. H. (1996). Frequency and stroke counts of Chinese characters.
Retrieved August 31, 1999, from technology.chtsai.org/charfreq/.
Van Vleet, T. M., & Robertson, L. C. (2006). Cross-modal interactions
in time and space: Auditory influence on visual attention in hemispatial neglect. Journal of Cognitive Neuroscience, 18, 1368-1379.
Welch, R. B., DuttonHurt, L. D., & Warren, D. H. (1986). Contributions of audition and vision to temporal rate perception. Perception
& Psychophysics, 39, 294-300.
Yeh, S. L., & Li, J. L. (2004). Sublexical processing in visual recognition of Chinese characters: Evidence from repetition blindness for
subcharacter components. Brain & Language, 88, 47-53.
Notes
1. When two repeated sounds were presented within a short interval,
the performance of the second sound either was facilitated (Soto-Faraco &
Spence, 2002) or deteriorated (Soto-Faraco, 2000), as compared with unrepeated sounds. Hence, we avoided using repeated sounds in our design.
2. The effect size was calculated by using Cohen’s d (Thalheimer &
Cook, 2002).
(Manuscript received June 14, 2006;
revision accepted for publication October 20, 2007.)
44
PSYCHOLOGICAL SCIENCE
Research Article
OUR EYES DO NOT ALWAYS GO WHERE
WE WANT THEM TO GO:
Capture of the Eyes by New Objects
Jan Theeuwes,1 Arthur F. Kramer,2 Sowon Hahn,2 and David E. Irwin2
1
TNO Human Factors Research Institute, Soesterberg, The Netherlands, and 2Beckman Institute and Department of Psychology,
University of Illinois at Urbana-Champaign
Abstract—Observers make rapid eye movements to examine the
world around them. Before an eye movement is made, attention is
covertly shifted to the location of the object of interest. The eyes typically will land at the position at which attention is directed. Here we
report that a goal-directed eye movement toward a uniquely colored
object is disrupted by the appearance of a new but task-irrelevant
object, unless subjects have a sufficient amount of time to focus their
attention on the location of the target prior to the appearance of the
new object. In many instances, the eyes started moving toward the new
object before gaze started to shift to the color-singleton target. The
eyes often landed for a very short period of time (25–150 ms) near the
new object. The results suggest parallel programming of two saccades: one voluntary, goal-directed eye movement toward the colorsingleton target and one stimulus-driven eye movement reflexively
elicited by the appearance of the new object. Neuroanatomical structures responsible for parallel programming of saccades are discussed.
It is well-known that the visual system is sensitive to events that
exhibit sudden change (Breitmeyer & Ganz, 1976). When a person
is engaged in a particular task (e.g., reading a book), something
might happen in the environment (e.g., someone switches on a desk
lamp at a nearby library table) that immediately captures the person’s attention even when he or she had no particular intention to
look for such events. Directing attention to these events may merely seem to distract the person from the intended task. It has been
argued, though, that detecting these sudden changes in the environment is significant for behaving organisms because these events
may require immediate identification and action (Todd & Van
Gelder, 1979; Yantis, in press).
A number of studies have demonstrated that an abrupt-onset object
(an object presented with a transient luminance change) captures
attention automatically (Jonides, 1981; Müller & Rabbitt, 1989;
Theeuwes, 1990; Yantis & Jonides, 1984). These studies show that
even when observers have no intention to look for an onset, an abrupt
onset among other nononset elements is processed first. On the basis
of these findings, it has been argued that sudden luminance changes,
such as the appearance of a new object in a scene, capture attention in
a purely stimulus-driven fashion (note that for attentional capture to
occur, it is not necessary that the new object has a luminance increment; see Yantis & Hillstrom, 1994).
Address correspondence to Jan Theeuwes, TNO Human Factors Research
Institute, P.O. Box 23, 3769 ZG Soesterberg, The Netherlands, e-mail:
[email protected], or Art Kramer, Beckman Institute, University of Illinois,
405 N. Mathews Ave., Urbana, IL 61801, e-mail: [email protected].
Even though there is a large body of evidence which suggests that
the appearance of a new object may capture attention, it is largely
unknown whether such an event also triggers a subsequent eye movement. There is, however, ample evidence to suggest that there is a
close relationship between the oculomotor and attentional systems.
Studies have shown that before a voluntary eye movement is made,
attention is covertly shifted to the location of the object of interest
(e.g., Hoffman & Subramaniam, 1995; Kowler, Anderson, Dosher, &
Blaser, 1995; Shephard, Findlay, & Hockey, 1986). The eye typically
will land at the position at which attention is directed (Deubel &
Schneider, 1996).
EXPERIMENT 1
In the present study, we used a visual search task in which
observers were required to make a voluntary, goal-directed saccade
to a color-singleton target. In half of the trials, simultaneously with
the presentation of the color-singleton target, a new object presented
with an abrupt onset appeared somewhere in the display. In contrast
to previous studies investigating the effect of onsets on attentional
deployment (e.g., Yantis & Jonides, 1984), in the current study, the
onset was never relevant for the task. The question addressed was
whether the appearance of a new yet irrelevant object would disrupt
the planning and execution of the goal-directed saccade toward the
singleton target. Intro Exp (1): Write main purpose
Method
Method for Exp (1): Write the details so others can replicate
Observers viewed displays containing six gray circles (3.7° in
diameter), each containing a small gray figure-eight premask (0.4° ×
0.2°), equally spaced around an imaginary circle with a radius of
12.6°. After 1,000 ms, all circles except one changed to red, and at
the same time, the small premasks inside the circles changed to letters. Observers were instructed that as soon as the colors of the circles changed, they were to quickly and accurately make a saccade
directly to the only gray circle left (a color singleton) and to determine whether the letter inside the gray circle was a c (to which they
responded by pressing a button with their left hand) or a reverse c (to
which they responded by pressing a button with their right hand).
The letters inside the red circles were distractor letters randomly
sampled without replacement from the set S, E, H, P, F, and U. A
pilot study was conducted to ensure that accurate target identification could be achieved only when the letter was fixated. Eye movements were recorded by means of an Eye Link tracker (250-Hz
temporal resolution and 0.2° spatial resolution) from 8 naive
observers each performing 64 practice and 256 experimental trials.
Copyright © 1998 American Psychological Society
VOL. 9, NO. 5, SEPTEMBER 1998
45
379
PSYCHOLOGICAL SCIENCE
Capture of the Eyes by New Objects
Fig. 1. Graphic illustration of the displays and the temporal sequence of an experimental trial (from left to right). Note that the target is defined
simultaneously with the appearance of the onset distractor. The gray circles are indicated by the dashed lines. The red circles are indicated by
the solid lines.
Res Exp (1): Statistics should be stated as supplement
On half of the trials, at one of four possible locations in the visual
field, an additional red circle was added to the display at the same
time that the color singleton was revealed (see Fig. 1 for examples of
the stimulus materials).1
Results and Discussion
Approximately 7.4% of the trials were discarded because
observers were not fixating the fixation dot at the beginning of the
trial (anticipation saccades). Those trials on which observers made
errors (i.e., pressed the wrong response key) were also excluded
(4.4%). Because the error rate was so low, error scores were not analyzed further.
Figure 2 presents the eye movement behavior of one representative
observer. The results displayed are sample points (every 4 ms) of the
path of the first saccade (additional saccades are not shown). When no
new object was added to the display (control condition), saccades generally went directly to the color-singleton target (Fig. 2a). In those trials in which a new object was added to the display, however, the eyes
1. It might be argued that the initial capture of attention by the new object
occurred not because of the sudden appearance of the new object but instead
because the new object appeared at the center of a cluster, or group, of three
closely spaced objects (see, e.g., the location of the new object in Figs. 2b, 2c,
and 2d). We tested this hypothesis in a control study by including an additional circle in the display in the same positions in which the new object appeared
in this study. However, in this control study, the extra circle appeared at the
beginning of each trial along with all of the other circles. In this case, observers
did not start moving their eyes in the direction of this additional circle. Therefore, it was the appearance of a new object and not a cluster of three closely
spaced objects that disrupted saccades to the singleton target.
findings
often went in the direction of the new object, stopped briefly, and then
went on to the target (Figs. 2b, 2c, and 2d).
The appearance of the new object (i.e., the abrupt onset) affected
not only eye movement behavior, but also reaction time (RT) to identify the letter inside the color singleton (RT of 785 ms without the new
object vs. 840 ms with the new object, F[1, 7] = 22.08, p < .01). The
sudden appearance of the new object did not affect the time it took the
eye to start moving from the centrally located fixation point (saccade
latency of 238 ms without the new object vs. 230 ms with the new
object); rather, the increase in manual RT was caused by the need for
an additional saccade to be made from the irrelevant onset stimulus to
the location of the color-singleton target on many trials.
Figure 3 summarizes the effect of the appearance of a new but
task-irrelevant object on the scan path of the eye. The maximum
angular deviation of the eye from a linear path between the fixation
and the target object was calculated over observers. As is clear from
Figure 3a, in the control condition, in which no new object was
added to the display, the eyes tended to go directly to the colorsingleton target. In contrast, Figures 3b, 3c, and 3d show that when
an irrelevant new object was presented, in many instances the eye
went in the direction of the new object. Even when the new object
appeared at a location opposite the singleton target (at 150°), the eye
tended to start moving in the direction of the new object. Overall, the
results indicate that in about half of the trials, the appearance of a
new object disrupted the planning and execution of the goal-directed
saccade toward the singleton target.2 Irrespective of where the new
2. Two different 90° target-distractor separations were used in the study:
The onset distractor could appear 90° away from the target in the same or in the
different hemifield. Given that there was no difference in the eye movement
and performance patterns for these two 90°-separation conditions, they were
collapsed for the purpose of presentation.
380
VOL. 9, NO. 5, SEPTEMBER 1998
46
to main
PSYCHOLOGICAL SCIENCE
J. Theeuwes et al.
A
control (no new object)
B
new object appearing at 30° from the target
C
new object appearing at 90° from the target
object appeared, the eyes tended to move in a direction that corresponded to the location of the new object.
When the eye started to move in the direction of the new object,
in most cases the eye stopped for a brief period before it went on to
the singleton target. Figure 4 shows the fixation durations after the
first saccade for those saccades that went in the direction of the new
object. For this analysis, we used only the 90° and 150° targetdistractor separations because these separations made it easy to discriminate saccades toward the new object from those toward the
target. Note that about 87% of the fixation durations were less than
150 ms, even though a complete change in the direction of the eye
movement was required to redirect the eyes in the direction of the
target.
We also explicitly asked observers after the experiment whether
they were aware that the appearance of the new object affected their
eye movements. Observers indicated that they were sure that their eye
movements were not affected by the appearance of the new object.
On the basis of these results, we conclude that the appearance of a
new object interferes with the planning and execution of a goaldirected eye movement to the color singleton. Earlier research has
demonstrated that the appearance of a new object involuntarily captures attention (Theeuwes, 1994; Yantis & Jonides, 1984). The current
results indicate that a new object not only captures attention, it also
captures the eyes. Dis Exp (1): State main finding
EXPERIMENT 2
Dis Exp (2): Mention posssible confounding in Exp 1...
D
new object appearing at 150° from the target
Fig. 2. Initial tracks that the eyes took as they left the fixation point
until the first fixation near one of the colored circles. Eye position was
digitized at 250 Hz. Thus, the points in the figure represent data points
acquired every 4 ms during the initial eye movement. Eye movement
behavior of an observer is shown for the control condition (a), in
which no new object was presented, and for three onset conditions:
when a new object was presented close to the singleton target (at 30°
of arc, corresponding to a distance of 6.4° of visual angle) (b), when a
new object was presented farther away from the singleton target (at
90° of arc, corresponding to a distance of 19.4° of visual angle) (c),
and when a new object was presented at the opposite side of the visual field (at 150° of arc, corresponding to a distance of 25.4° of visual
angle) (d). The results are collapsed and normalized with respect to the
position of the target singleton (marked here with a double circle) and
the position of the new object.
The results of the first study suggest that both the goal-directed
allocation of attention and the movement of the eyes to a clearly
defined target can be disrupted by the appearance of a new but taskirrelevant object in the visual field, even when this object appears quite
distant from the target. However, the data that we have reported thus
far do not define the boundary conditions for this capture effect on
attention and eye movements. ... to motivate Exp 2
Previous studies of attentional capture have reported that the
appearance of a new object does not influence performance when
observers have a sufficient amount of time to attend to the location of
a target prior to the appearance of the new object. For example, Yantis
and Jonides (1990; see also Theeuwes, 1991) found that as long as a
central arrow cue indicated the position of a target at least 200 ms
before the appearance of an onset elsewhere in the visual field, the
onset had no influence on the time it took subjects to identify the target letter.
In the present study, we examined whether precuing the location of
the color-singleton target would eliminate or reduce the influence of
the task-irrelevant onset on performance and eye movements. Given
the previous literature, which suggested that attention is often shifted
to an area of interest in advance of an eye movement (Deubel &
Schneider, 1996; Hoffman & Subramaniam, 1995), it was our expectation that providing subjects with sufficient time to attend to the location of the subsequent target and to program a goal-directed eye
movement to this location would preclude the misdirection of the eyes
to the task-irrelevant object.
Intro Exp (1): Write main purpose of this Exp &
Method provide prediction
The stimuli and procedures were equivalent to those of Experiment
1 with the following exceptions. A centrally located arrow cue, which
381
VOL. 9, NO. 5, SEPTEMBER 1998
47
PSYCHOLOGICAL SCIENCE
J. Theeuwes et al.
Open Gen Dis (1): State research goal & summarize main
GENERAL DISCUSSION
findings
Fig. 4. Fixation durations after the first saccade for those saccades
that went in the direction of the new object in Experiment 1. The
results displayed here are from the conditions in which the new object
appeared at 90° and 150° from the target and the eye started to move
in the direction of the new object.
Open Gen Dis (2): Combine the
purpose & findings of each Exp
Results and Discussion
The goal of the observers in our task was to make a saccade toward
the color-singleton target. In order to execute such a saccade, attention
first needs to be shifted to the location of the color singleton (Deubel
& Schneider, 1996; Kowler et al., 1995). Furthermore, we assume that
the shift of attention to the location initiated the programming of a saccade to the attended location (e.g., Deubel & Schneider, 1996; Morrison, 1984).3 In Experiment 1, simultaneously with the preparation of
a goal-directed saccade to the color singleton, a new object appeared
in the display, causing attention to be reflexively drawn to the location
of the new object. We assume that the reflexive shift of attention to the
new object also initiated the programming of a saccade, but to the
location of the new object. In fact, the results suggest that there is parallel programming of two saccades, one to the singleton target and one
to the new onset object. Depending on which eye movement program
is ready first, the eyes will start moving in the direction of the onset or
in the direction of the color singleton. If the eyes start moving in the
direction of the onset, as soon as the program for a saccade to the
color-singleton target is ready, inhibition of the reflexive saccade may
occur, causing the eyes to stop somewhere along the trajectory
between the initial center fixation point and the onset. Because the eye
movement program toward the color-singleton target has been prepared, after the eyes stop for a brief period of time, they will move to
the color singleton.4 This interpretation is in line with the very short
fixation durations that we found after the first saccade (see Fig. 4). In
fact, the majority of the fixations were too short to allow the programming of a completely new eye movement, which typically takes 150 to
200 ms (e.g., Becker, 1991; Findlay, 1997; Salthouse & Ellis, 1980).
Overall, these results suggest parallel programming of two saccades (see also McPeek, Skavenski, & Nakayama, 1996): one voluntary, goal-directed eye movement toward the color-singleton target and
one stimulus-driven eye movement reflexively elicited by the appearance of the new object. The parallel programming of two saccades has
been reported before in reading (Henderson & Ferreira, 1990; Morrison, 1984; Reichle, Pollatsek, Fisher, & Rayner, 1998): A reader may
fixate a particular word for a very brief period and immediately move
on to the next word. Parallel programming of two saccades has also
been reported in the double-step paradigm (Becker & Jurgens, 1979),
in which the position of a target changes just prior to a saccade.
Interestingly, shifting attention to the location of the target in
advance of the appearance of the task-irrelevant onset effectively eliminated the misdirection of the eyes to the onset and the performance disruption observed in Experiment 1. Within the theoretical framework
just described, the elimination of the effect of the onset would likely
have resulted from the temporal advantage provided to the programming of the goal-directed saccade by the location precue. However, it
is also conceivable that performance and eye movements (to the target)
The main question addressed in this study was whether cuing the
location of the color-singleton target would eliminate the influence of
the task-irrelevant distractor on target identification and eye movements. Because there were no significant differences between results
for the three SOAs, these results were combined for the analyses
reported here. Figure 5 summarizes the effect of the appearance of the
new but task-irrelevant object on the scan path of the eyes. The maximum angular deviation of the eyes from a linear path between fixation
and the target object was calculated and collapsed across the 7
observers.
As can be seen in the figure, the distribution of trials is quite similar regardless of whether an onset distractor was present (bottom two
panels) or absent (top two panels). These data suggest that onset distractors have little influence on the trajectory of eye movements when
subjects are able to shift their attention to the target location in
advance of the appearance of the onset. This observation was quantified by calculating the number of trials on which the eyes initially
went in the direction of the onset (onset trials) or additional circle
(control trials) and submitting these data to a two-way analysis of variance with target-distractor separation (90° vs. 150°) and condition
(onset vs. control trials) as factors. Neither the main effects nor the
interaction was statistically significant (ps > .40). The average proportion of onset trials on which the eyes initially went toward the onset
circle was 2.3%; in comparison, on 1.7% of the control trials, the eyes
initially went toward the nononset circle that appeared at a comparable spatial position. Res Exp (2): Condense statistics into one sentence
An analysis of the RT data presents a similar picture. The RTs on
the onset and control trials, collapsed across the three SOAs, were 617
3. We are not suggesting that the execution of the saccade necessarily foland 607 ms, respectively. These RTs were not significantly different lowed saccade programming, only that the allocation of attention initiated saccade programming (Hoffman, in press; Klein & Pontefract, 1994). Indeed, on
(p > .30).
In summary, the results of the present study establish an important those trials on which the preparation of the goal-directed saccade was comboundary condition on the attentional and oculomotor capture effects pleted first, the reflexive saccade to the onset was prepared but not executed.
4. On many trials on which the eyes initially moved toward the onset, they
observed in Experiment 1. Capture effects by the appearance of a taskdid not reach its location. Indeed, on average, the saccades toward the onset
irrelevant onset can be overcome when observers have sufficient time traversed only 70% of the distance between fixation and the onset before stopto attend and program an eye movement to the location of a subse- ping for a brief period of time and then moving to the location of the colorquent target stimulus. Dis Exp (1): State main finding
singleton target.
383
VOL. 9, NO. 5, SEPTEMBER 1998
49
P. Cavanagh et al. / Cognition 80 (2001) 47±60
47
COGNITION
Cognition 80 (2001) 47±60
www.elsevier.com/locate/cognit
Attention-based visual routines: sprites
Patrick Cavanagh a,*, Angela T. Labianca a, Ian M. Thornton b,1
a
Vision Sciences Laboratory, Department of Psychology, Harvard University, 33 Kirkland Street,
Cambridge, MA 02138, USA
b
Cambridge Basic Research, Nissan Research and Development Inc., Cambridge, MA 02142, USA
Received 12 October 1999; accepted 17 November 2000
Abstract
A central role of visual attention is to generate object descriptions that are not available
from early vision. Simple examples are counting elements in a display or deciding whether a
dot is inside or outside a closed contour (Ullman, Cognition 18 (1984) 97). We are interested
in the high-level descriptions of dynamic patterns ± the motions that characterize familiar
objects undergoing stereotypical action ± such as a pencil bouncing on a table top, a butter¯y
in ¯ight, or a closing door. We examine whether the perception of these action patterns is
mediated by attention as a high-level animation or `sprite'. We have studied the discrimination of displays made up of simple, rigidly linked sets of points in motion: either pairs of
points in orbiting motion or 11 points in biological motion mimicking human walking. We
®nd that discrimination of even the simplest dynamic patterns demands attention. q 2001
Elsevier Science B.V. All rights reserved.
Keywords: Attention; Tracking; Motion; Visual search
1. Introduction
General & familiar opening and connect with last paragraph
Open Gen Intro (1) Recognizing an item can call on much more than just an analysis of its static form.
Something that moves on a street and makes motor sounds is probably a car or a
truck (or maybe a 3-year-old boy). But in addition to characteristic sounds or properties, many objects have characteristic patterns of movement, revealed only over
* Corresponding author. Fax: 11-617-495-3764.
E-mail address: [email protected] (P. Cavanagh).
1
Present address: Max-Planck-Institut fuer Biologische Kybernetik, Spemannstrasse 38, D-72076
Tuebingen, Germany.
0010-0277/01/$ - see front matter q 2001 Elsevier Science B.V. All rights reserved.
PII: S 0010-027 7(00)00153-0
52
48
P. Cavanagh et al. / Cognition 80 (2001) 47±60
some duration of time. Rubber balls bounce on ¯oors, billiard balls bounce less,
butter¯ies dance up and down, Frisbees ¯y straight, a pencil bouncing off the ¯oor
takes an end over end tumble, and doors swing slowly through an arc when opening
and bang up against their endpoints. Many of these characteristic patterns of motion
are so familiar as to be suf®cient for recognition of the object. The case of biological
motion is perhaps the strongest evidence for this. A human form is easily recognized
from the motions of a set of lights attached to a person ®lmed while walking in the
dark (Johansson, 1973; Neri, Morrone, & Burr, 1998).
How do we accomplish this seemingly effortless recognition of motion patterns?
We are not aware of analyzing components of the motion and coming to intermediate decisions. The human walking just seems to pop out of the display. Johansson
(1973) proposed that the analysis relied on an automatic and spontaneous extraction
of mathematically lawful spatiotemporal relations. But is this act of recognition
really effortless? And what about the continued perception of the motion, an analysis
which continually adapts our impression of the walker's posture and progress to the
moving points in the display. Can that also be as effortless as it seems?
W1(1) We propose that these characteristic motions are analyzed and interpreted by a
special set of operators that we will call `sprites'. In this paper, we will only address
the attentional demands of these operators but we will nevertheless sketch our view
of their properties and the role they play. We consider a sprite to be the set of
routines that is responsible for detecting the presence of a speci®c characteristic
motion in the input array, for modeling or animating the object's changing con®guration as it makes this stereotypical motion, and for ®lling in the predictable details
of the motion over time and in the face of noisy or absent image details. Each
different characteristic motion pattern would have its own `sprite' that would be
built up over many exposures to the pattern. These stored recognition and animation
routines then allow sparse inputs to support rich dynamic percepts. Many others
have stressed that regularities in the world can be captured by ef®cient, higher-order
data structures such as chunks (Miller, 1956), schemata (Bartlett, 1932; Neisser,
1967), frames or scripts (Minsky, 1975; Schank & Abelson, 1977). Characteristic
patterns of motion ought to lend themselves quite well to similarly ef®cient representations.
Separate instances of a characteristic motion are seldom exact repetitions,
however. The path of a bouncing pencil can be quite chaotic, depending strongly
on many factors (starting position, rotation, surface properties, etc.). The regularities
of a bouncing pencil, or a butter¯y's ¯ight, or a walking human, lie at a higher level
of description of the motion. Signi®cant analysis of the motion pattern must precede
any recognition of the regularity and signi®cant computation is then required to use
the knowledge of the regularity to predict or animate subsequent motions.
This procedural aspect of a sprite is closely related to the concept of `visual
routines' addressed by Ullman (1984). These routines act on the representations
emerging from the initial stage of visual analysis to establish properties and relations
that are not explicitly represented in the ®rst stage. Ullman identi®ed elemental
processes such as counting, indexing, tracking, and region-®lling which could be
53
P. Cavanagh et al. / Cognition 80 (2001) 47±60
49
organized (compiled) into visual routines to perform a high-level task such as, for
example, judging whether a point is inside or outside a complex closed curve.
Importantly, in the original work by Ullman (1984), he assigned the critical
operation of shifting the focus of analysis to attentional processes so that his visual
routines were exclusively attention-based. In our case, abstracting the high-level
description of, say, a bouncing pencil or a point-light walker certainly calls on an
analysis of similar or greater complexity than the spatial tasks that Ullman (1984)
W1(2) described. Nevertheless, whether or not these high-level motion descriptors ± sprites
± require attention remains an open question. It is the central question that we
address in this paper.
In the two experiments reported below, we use visual search tasks to examine the
attentional load required to perceive a dynamic motion pattern. The displays present
one to four motion patterns and the subjects report the presence or absence of a
target motion. The relation between reaction time and number of items in the display
allows us to evaluate any increase in attentional load with each additional item. If,
for example, point-light walkers are recognized effortlessly, then there should be no
increase in reaction time as the number of walkers in the display increases from one
to four. Final Gen Intro (8,9): Write overview & prediction of both experiments
Our ®rst experiment examines simple con®gurations of two moving dots.
Although these patterns of motion are relatively simple they are not highly familiar.
If attention is required to discriminate between con®gurations, it may be because we
do not have highly ef®cient routines, sprites, to handle them. Our second experiment
examines highly familiar con®gurations involving human motion. The movements
in these stimuli are more complex but extremely familiar. If any dynamic patterns
can be discriminated without attentional load, we believe it should be these patterns.
2. Experiment 1
This is a special case where the method for an experiment is given
in detail in Intro.
Intro Exp (3) In the ®rst experiment, observers had to discriminate between two different orbital
Give only when
necessary
motions. In each stimulus, two lights rotate around each other while moving around
a central ®xation point. We will describe the two stimuli brie¯y before examining
the similarities between our displays and the classic wheel-generated motions
studied in many previous articles (Duncker, 1937; Johansson, 1973; Prof®tt, Cutting,
& Stier, 1979; Wallach, 1965).
In our ®rst stimulus (Fig. 1a), the motion is like that of a moon orbiting a planet
where the planet itself orbits a central `star' (the ®xation point). The moon traces out
a complex curve (of the cycloid family) around the central point whereas the planet
traces a circle. If the smoothly moving light (the `planet') is turned off, the complex
nature of the moon's motion is immediately evident as a looping path of wildly
varying velocity. With the smoothly moving `planet' turned on, however, the erratic
motion is no longer apparent as the ®rst light is now seen to rotate smoothly at
constant velocity around the `planet'.
In the second stimulus (Fig. 1b), the two lights rotate around a common center and
this central point rotates around the ®xation point. In this case, both lights trace
54
52
End of special case
P. Cavanagh et al. / Cognition 80 (2001) 47±60
one, two, three, or four of the dot pairs moving in the same direction at the same rate
around the ®xation. On a target present trial, one of the pairs is following the target
motion, say, tumbling, whereas the others are following the alternative motion,
orbiting in this case. We recorded the reaction time to respond that the target was
present or absent and analyzed the function relating the reaction time and the
number of dot pairs present. If the reaction time did not increase with the number
of dot pairs present, we would conclude that the extraction of the characteristic
motions calls only minimally on central, attentional resources.
As a control, we also measured the reaction time with only one of the dots present
in each pair. The target would then be the smoothly moving `planet' among singleton tumbling dots or one singleton tumbling dot among smoothly moving planets.
This gives a measure of the distinctiveness of the single dot motions that make up
the characteristic motions of the pairs.
2.1. Methods
2.1.1. Observers
Twelve paid volunteers, with an average age of about 21 years and with normal or
corrected-to-normal vision, participated in this experiment. All participants were
naive to the purpose of the experiment and gave informed written consent before the
experiments, which were approved by the F.A.S. Human Subjects Committee,
Harvard University.
2.1.2. Stimuli
The display was presented on a 14 inch 67 Hz Macintosh display driven by a
Macintosh 7500/100 programmed in Vision Shell. The two motion patterns we used
are shown in Fig. 1. Between one and four of these orbiting or tumbling pairs were
presented rotating around the central ®xation. The distance from the ®xation to the
center of rotation of the pair was 4.38 of visual angle at a ®xed viewing distance of 57
cm. The center±center separation of the two dots was 18 of visual angle. The dots
themselves had a diameter of 0.48 of visual angle. The two dots made one full cycle
around each other every 1.5 s while the pair made a full circuit around ®xation every
7.5 s. The local and global rotations were always in opposite directions and the
direction was set randomly on each trial. The stimuli, when there was more than one,
were spaced evenly around the circular path. The initial position of each local
rotation was set randomly at the beginning of each trial as was the starting location
of the rotation around the circular path. The dots had a luminance of 70 cd/m 2, and
were presented on a 20 cd/m 2 background. The temporal onset and offset of the
motion pairs on each trial was a step function. The ®xation mark was a single static
dot at the center of the display. It was identical in size to the moving dots (0.48 of
visual angle in diameter).
2.1.3. Procedure
In the control conditions, only one dot of each pair was shown. It was always the
smoothly moving dot (the planet) in the orbiting con®guration and either of the dots
57
P. Cavanagh et al. / Cognition 80 (2001) 47±60
Gen Con (1,2): Cover
1st, 3rd, & 4th Ws here
in a short paragraph
59
dif®cult. Motion takes place over time and considerable time appears to be required
to establish the trajectory of each pair. The static versions of the trajectories of our
two motion patterns are much easier to distinguish. The orbiting pattern traces out
something like a dollar sign and the tumbling pair something like a ®gure 8. When
we simply presented these short trajectories as spatial patterns all at once, the
discrimination was much more rapid (slope of less than 100 ms/item, averaged
over four subjects). What makes the real motion case so dif®cult is that the con®guration of dot motions needs to be made explicit, linking each dot to the next and
then to the central dot. This is the only way to discover if one of the dots is actually
maintaining a constant distance from the central dot.
These articulated links of constant length in tumbling and orbiting motion also
form the basis of the structure for the point-light walkers where the light at each joint
is separated from the next by a ®xed distance. However, for the walkers in Experiment 2, the processing time was much faster at about 200 ms per walker (compared
to about 1 s per dot pair). This rate is comparable to some estimates of the dwell time
of attention (Duncan, 1984). This suggests that attention was required to notice the
gait of each walker in turn but that little processing was required once each walker
had been selected. This is evidence that the analysis of a very familiar motion
pattern, despite its complexity, can be very rapid. We suggest that this rapid extraction of the motion pattern is the signature of the `sprite' responsible for recognizing
and animating the percept of a walking human form. Despite this rapid extraction,
our data also show that only one walker at a time can be analyzed. The search rate
was still substantial indicating that the operation of at least the `walker' sprite
requires attention.
The dual task results of Thornton et al. (1999) suggested that, in some cases, the
perception of biological motion can be automatic. However, a search task is perhaps
a more sensitive measure of attentional load because dual task interference only
reveals an interaction between the two tasks if the combined load exceeds the
available capacity. It cannot differentiate between no attentional demands and any
combination that is less than the limit available.
If even familiar patterns of motion require attention to be discriminated, what can
be the advantage of the routines that support the perception of the pattern? Clearly, it
is the same advantage that is offered by any recognition of a familiar pattern. Once
enough of the pattern is acquired to recognize it, the rest can be ®lled in from
memory. Sparse inputs can support rich percepts and in the case of a moving object,
®lling in implies a prediction of likely motions and tracking them with less data than
would be otherwise necessary. These advantages have formed the basis of many
theories of perception from schemata and schema theory (Bartlett, 1932; Neisser,
1967) to frames and scripts (Minsky, 1975; Schank & Abelson, 1977).
To conclude, we suggest that the visual system acquires and uses stored motion
patterns, sprites, which are characteristic of familiar events or objects: the motion of
a wheel, the jump of a ®sh out of water, the way a pencil bounces on the ¯oor when
dropped, and the way a fresh egg does not. We use these stored patterns to recognize
and then animate our perception of familiar events. Our experience of these animation routines might suggest that they are effortless but our study here shows that they
This final paragraph connects with the Introduction
64
60
P. Cavanagh et al. / Cognition 80 (2001) 47±60
are not. We claim that the animations are played out by attentive processes in the
same way that we can animate a mental image. In the visual case, the input image
data act like set points in the progress of the animation but the animation still
requires the support of attentive processes.
Acknowledgements
Thanks to Pawan Sinha for helpful discussions of this work and to Jane Raymond
for sharing her observations on Frisbees and butter¯ies. This research was supported
by grant EY09258 to P.C.
References
Bartlett, S. F. (1932). Remembering: a study in experimental and social psychology. Cambridge:
Cambridge University Press.
Cavanagh, P. (1992). Attention-based motion perception. Science, 257, 1563±1565.
Cutting, J. E. (1978). A program to generate synthetic walkers as dynamic point-light displays. Behavioral
Research Methods and Instrumentation, 10, 91±94.
Duncan, J. (1984). Selective attention and the organization of visual information. Journal of Experimental
Psychology: General, 113, 501±517.
Duncker, K. (1937). Induced motion. In W. D. Ellis (Ed.), A sourcebook of Gestalt psychology. London:
Routledge and Kegan Paul.
Inman, V. T., Ralston, H., & Todd, J. T. (1981). Human walking. Baltimore, MD: Williams & Wilkins.
Johansson, G. (1973). Visual perception of biological motion and a model for its analysis. Perception and
Psychophysics, 14, 201±211.
Miller, G. A. (1956). The magical number seven plus or minus two: some limits on our capacity for
processing information. Psychological Review, 63, 81±97.
Minsky, M. (1975). A framework for representing knowledge. In P. H. Winston (Ed.). The psychology of
computer vision (pp. 211±280). New York: McGraw-Hill.
Neisser, U. (1967). Cognitive psychology. New York: Appleton.
Neri, P., Morrone, M. C., & Burr, D. C. (1998). Seeing biological motion. Nature, 395, 894±896.
Prof®tt, D. R., & Cutting, J. E. (1979). Perceiving the centroid of con®gurations on a rolling wheel.
Perception and Psychophysics, 25, 389±398.
Prof®tt, D. R., Cutting, J. E., & Stier, D. M. (1979). Perception of wheel-generated motions. Journal of
Experimental Psychology: Human Perception and Performance, 5, 289±302.
Schank, R., & Abelson, R. (1977). Scripts, plans, goals and understanding. Hillsdale, NJ: Lawrence
Erlbaum Associates.
Thornton, I. M., Rensink, R. A., & Shiffrar, M. (1999). Biological motion processing without attention.
Perception, 28 (Suppl), 51.
Ullman, S. (1984). Visual routines. Cognition, 18, 97±159.
Wallach, H. (1965). Visual perception of motion. In G. Keyes (Ed.), The nature and the art of motion.
New York: George Braziller.
65
Journal of Experimental Psychology:
Human Perception and Performance
1997, Vol. 23, No. 2, 339-352
Copyright 1997 by the American Psychological Association, Inc.
0096-1523/97/$3.00
Perception Without Attention:
Evidence of Grouping Under Conditions of Inattention
Cathleen M. M o o r e and H o w a r d Egeth
Johns Hopkins University
Many theories of visual perception assume that before attention is allocated within a scene,
visual information is parsed according to the Gestalt principles of organization. This assumption has been challenged by experiments in which participants were unable to identify ~¢laat
Gestalt grouping patterns had occurred in the background of ptimary-t~k displays (A. Mack,
B. Tang, R. Tuma, S. Kahn, & I. Rock, 1992). In the present study, participants reported
which of 2 horizontal lines was longer. Dots in the background, if grouped, formed displays
similar to the Ponzo illusion (Experiments 1 and 2) or the Mtlller-Lyer illusion (Experiment
3). Despite inaccurate reports of what the patterns were, participants' responses on the
line-length discrimination task were clearly affected by the 2 illusions. These results suggest
that Gestalt grouping does occur without attention but that the patterns thus formed may not
be encoded in memory without attention.
An assumption of many theories of visual perception is
that a rich array of perceptual primitives is extracted from
visual scenes without attention, that is, preattentively (e.g.,
Duncan & Humphreys, 1989; Julesz, 1981; Kahneman &
Treisman, 1984; Neisser, 1967; Treisman, 1982, 1988;
Wolfe, 1994). This assumption is tenable on logical
grounds: The visual field must be at least partially organized
at preattentive levels of processing for there to be something
on which to base later--attentive--processing. In addition
to this logical justification, empirical evidence has suggested that perceptual organization occurs preattentively.
Dual-task studies, for example, have demonstrated that
when attention is occupied by a demanding primary task,
substantial perceptual processing can occur within the attentional periphery (e.g., Braun & Sagi, 1990, 1991; but s e e
Ben-Av, Sagi, & Braun, 1992). In addition, within visualsearch experiments, salient items and texture boundaries
seem to "pop out" from visual arrays of surrounding items
with no effort (e.g., Beck, 1982; Treisman, 1985). Results
like these have been taken as evidence that visual scenes
are organized into perceptual primitives before attention is
focused within the scene. Indeed, a substantial amount of
research has been dedicated to identifying what the
preattentive primitives are, based on results from dual-
task and visual-search experiments (e.g., Bravo & Blake,
1990; Brown, Weisstein, & May, 1992; Julesz & Bergen,
1983; Ramachandran, 1988; Treisman & Gormican,
1988).
An important feature of preattentive processing, according to the view just described, is that it organizes the visual
field. Therefore, it is reasonable that preattentive processes
have been assumed to include those mechanisms that give
rise to the Gestalt principles of perceptual organization (see
Pomerantz, 1981, for a review; Figure 1 illustrates two
principles of Gestalt grouping). The Gestalt principle that is
considered most often is grouping by similarity (e.g., Duncan & Humphreys, 1989; Humphreys, Quinlan, & Riddoch,
1989; Neisser, 1967; Treisman, 1982, 1988), but grouping
by proximity, closure, and good continuation have also
been assumed to apply preattentively (Grossberg, Mingolla,
& Ross, 1994; Pomerantz, 1981; Pomerantz & Pristach,
1989).
Object-based theories of perception have been especially
explicit in assuming that the Gestalt principles apply preattentively. These theories maintain that visual scenes are first
parsed, in part according to the Gestalt principles, and that
attention is then directed to the perceptual objects that result
from that parsing process (see Kanwisher & Driver, 1992,
for a review).
There is evidence to support the idea that attention is
allocated to objects. For example, it is difficult to attend to
more than one object simultaneously (e.g., Duncan, 1984;
Kahneman, Treisman, & Burkell, 1983; of. Vecera & Farah,
1994). At the same time, it is diffienlt to ignore distracting
information that is part of an attended object (e.g., Kahneman & Henik, 1981). Similarly, entire objects receive processing benefits when only part of the object is cued or
primed (Egly, Driver, & Rafal, 1994; see also Kahneman,
Treisman, & Gibbs, 1991). Finally, patients who suffer from
the attentional disorder visual neglect sometimes demonstrate object-based frames of reference in neglecting their
Cathleen M. Moore and Howard Egeth, Department of Psychology, Johns Hopkins University.
This research was supported by National Institutes of Health
Grant 5 T32 MH18215, awarded to the Department of Psychology
at Johns Hopkins University, and by National Science Foundation
Grant SBR-9319356. We thank Toby Mordkoff for helpful discussions concerning this work.
Correspondence concerning this article should be addressed to
Cathleen M. Moore, who is now at the Department of Psychology,
Moore Building, Pennsylvania State University, University Park,
Pennsylvania 16802. Electronic mail may be sent via Internet to
cmml 5 @psu.edu.
339
66
PERCEPTION WITHOUT ATI'ENTION
341
cations of how well the patterns could be perceived when
the backgrounds were attended--either partially or fully.
The results from several experiments using the inattention
method suggested that participants failed to group the dots
when they were not attending to the background (Mack et
al., 1992). Report accuracy of the patterns was at chance
following the inattention trials. In fact, participants often
denied that there had been a pattern and professed to have
been guessing about what the pattern might have been. In
contrast, on the divided-attention and full-attention trials,
reports concerning the patterns were quite accurate, which
indicated that the patterns could be seen when they were
attended. This pattern of results was found using several
different types of grouping stimuli in the backgrounds.
In summary, unlike the theories described in the first
section of this article, some theories of attention maintain
that very little perceptual organization occurs preattentively.
In support of this position is evidence that the Gestalt
principles of organization seem not to apply preattentively;
when they were tested under conditions of inattention, participants failed to report what grouping patterns had occurred in the background. This finding conflicts directly
with the assumptions concerning preattentive processing
that are part of many theories of perception, most explicitly
the object-based theories.
Thus, according to this view, assumptions concerning
preattentive perceptual organization must be tested under
conditions of inattention. Toward this end, Mack and Rock
and their colleagues developed an experimental method to
identify what can be perceived under conditions of inattention (Mack et al., 1992; Rock, Linnet, et al., 1992; see
Epstein & Babler, 1989, 1990, for a similar method). In one
study, they applied this method to the specific question of
whether Gestalt grouping can occur under conditions of
inattention (Mack et al., 1992).
The study took the following form. A fairly difficult
perceptual task was superimposed on a background of black
and white dots that were irrelevant to the task. On most
trials, the dots in the background were randomly black and
white. On the critical trial, however--the inattention trial-the dots were colored such that if grouping by similarity
occurred, a salient pattern would appear (e.g., alternating
rows of black and white dots). At the end of this trial,
participants were unexpectedly asked to make a forcedchoice response concerning what pattern had appeared in
the background. The logic was that if grouping occurs
without attention, then despite not having attended to the
background, participants would be able to report what pattern had appeared.
After the inattention trial, participants were no longer in a
state of inattention concerning the background. Therefore
only a single observation under conditions of inattention
could be obtained for each participant. Nonetheless, later in
the experiment a second patterned-background trial was
presented, and again participants were asked to make a
forced-choice response concerning what the pattern had
been. This trial was called the divided-attention trial because participants knew that something might occur in the
background. Finally, before the last trial in the experiment,
participants were asked to ignore the primary task and
simply report what pattern appeared in the background. We
refer to this trial as the full-attention trial. Whereas the
inattention trial provided an indication of how well the
patterns were perceived under conditions of inattention,
the divided-attention and full-attention trials provided indi-
Note this paper has a whole section for W1
Purpose of the Present Research
W1
et al.'s (1992) and Rock, Linnet, et al.'s (1992) definition of
"without attention." In another context, Treisman (1993), for example, did not use the terms preattention and inattention synonymously. She defined preattentive processing as "an inferred stage
of early vision, which [she] attribute[s] to the separate feature
modules. Before any conscious visual experience is possible, some
form of attention is required" (p. 13). By contrast, inattention is
used to refer to a state in which "attention is narrowly focused
elsewhere. In this case, even global feature boundaries should no
longer be available" (19. 15). Treisman's use of preattention and
inattention might correspond to what Mack et al. and Rock, Linnet,
et al. called divided or d/ffuse attention and inattention, respectively. Howevex~Treisman's preattenfion, unlike Mack et al.' s and
Rock, Linnet, et al.'s divided or diffuse attention, refers to processing that occurs without attention. Because this is the primary
concern of this article, we chose to use the terms preattention and
inattention in the manner described in the opening sentences of
this footnote.
68
Although participants were unable to report what grouping patterns had occurred on the inattention trials in Mack et
al.'s (1992) experiments, it does not necessarily follow that
the dots were not grouped. Instead, participants may have
been unable to remember what patterns had appeared. This
might have occurred for two different reasons. First, participants may have forgotten what the patterns were by the
time the question was answered. Alternatively, without having attended to the patterns, participants may never have
successfully encoded them in memory. In either case, participants would have been unable to report what the patterns
were, despite having perceived them at the time. To address
the question of whether Gestalt grouping occurs under conditions of inattention, it would be useful to obtain a measure
of grouping from the time during which the stimuli are
present. This was the purpose of the present research.
We took advantage of two simple perceptual illusionsB
the Ponzo illusion (see Figure 2A) and the MUller-Lyer
illusion (see Figure 2B). As in the experiments reported by
Mack et al. (1992), a difficult line-length discrimination
task was superimposed on a background of black and white
dots, which occasionally formed patterns. The patterns were
designed such that if they were perceived, they could influenceBthrough either the Ponzo illusion or the Mtiller-Lyer
illusion--the perceived lengths of the lines to be discriminated. Using displays like these, we could determine
whether the dots had been grouped on inattention trials
without having to ask the participants directly about what
they had seen. If reports of line length are influenced by the
patterns that are embedded in the dots, then this fmding
PERCEPTION WITHOUT ATTENTION
343
random-matrix trials. Participants simply reported which line segment appeared to be longer. The fourth, seventh, and eighth trials
were pattern-matrix trials; they were the inattention, dividedattention, and full-attention trials, respectively. Following both the
inattention and the divided-attention trials, participants reported
which line segment appeared to be longer. No response was
obtained following the full-attention trial; participants were asked
before the trial to ignore the line-segment task and simply look for
the pattern in the dot-matrix background.
After each of the three pattern-matrix trials, participants were
asked three questions concerning the pattern that had appeared in
the background: the direct query, the forced choice, and the
confidence rating. The direct query was the following: "Did you
notice any pattern in the background of dots on that last trial?" The
specific forced-choice question was different for the two experiments, but in each case it provided a forced-choice discrimination
concerning the pattern that had been presented. Finally, the confidence rating referred to the answer given to the forced-choice
question. The confidence rating was 1, 2, or 3, where 1 meant not
at all sure, 2 meant somewhat sure, and 3 meant very sure. As the
experimenter asked each question, a short form of the question was
presented on the dark background of the monitor (e.g., "pattern?";
"direction [up, down]?"; "confidence rating? [1 = not at all, 2 =
somewhat, 3 = very]"). The participant's responses also appeared
on the monitor as they were being entered by the experimenter.
Finally, at the end of the inattention block, participants were
asked the following question: "Did you notice that some of the
trials in the middle block included patterns like you've seen here?"
The participant's answer was entered into the computer, and that
marked the end of the experiment.
An illustration of the trial events is provided in Figure 5. The
fixation point at the beginning of each trial remained present for
1,000 ms. The trial display was then presented for 200 ms, after
which it was replaced by the mask display. The mask remained
present until a response was entered by the experimenter, after
which the display went blank, with only the gray background
visible. An intertfial interval of approximately 5 s in Experiments
1 and 2 and 8 s in Experiment 3--during which the display was
gray--preceded each trial.
ments 1 and 2, the lengths were chosen randomly from three
short-long pairs: (1.30", 1.50"), (1.50 °, 1.70"), (1.70", 1.90"). In
Experiment 3, the lengths were chosen randomly from three different sbort-long pairs: (5.00", 5.20"), (5.10", 5.30"), (5.20",
5.40*). On pattern-matrix trials, the line segments were the same
length (Experiments 1 and 2: 1.40"; Experiment 3: 5.20*). The
position of the line segments within the matrix was the same on all
trials within a given experiment. In Experiments 1 and 2, one line
segment was presented between the 5th and 6th rows, and the other
was presented between the 8th and 9th rows of the dot matrices. In
Experiment 3, they were presented in the 7th and 16th rows of the
dot matrices, on top of the dots. All line segments were centered
horizontally within the matrix. Finally, for the random-matrix
trials, in which one line segment was longer than the other, the
longer segment was presented equally often in the top and bottom
positions.
Task
The primary task in all three experiments was to report which of
the two black line segments appeared longer by vocally reporting
top or bottom. In addition to this primary task, participants were
sometimes asked questions about the displays. These questions are
described in detail in the Procedure section.
Procedure
Each participant took part in an individual session, which lasted
approximately 15-20 rain. Figure 4 illustrates the structure of the
experiments. Experiments 1 and 3 consisted of one practice block
and two experimental blocks of trials; Experiment 2 consisted of
one practice block and one experimental block of trials.
At the beginning of each session, participants were told that
each trial would begin with a fixation point, followed by a briefly
presented matrix of black and white dots that had two line segments embedded in it. They were asked to report verbally, after
each trial, which of the two line segments appeared longer: the top
or the bottom. If they did not know which was longer, they were
asked to guess.
Following these instructions, a block of 10 practice trials was
provided; these trials were always random-matrix trials. After each
trial, the experimenter entered the participant's response into the
computer and said whether the response was correct or not. Many
of the participants reported that they had been unable to see the
line segments on the first practice trial. All participants, however,
were able to see them and report which was longer by the end of
3 trials.
After the 10 practice trials, participants completed a block of 32
trials, plus 2 warm-up trials at the beginning of the block. We refer
to this block of trials as the illusion block. Participants were told
that the experimenter would no longer provide feedback concerning the participant's responses but would continue to enter the
responses into the computer. A random 16 of the experimental
trials were random-matrix trials, and the remaining 16 were
pattern-matrix trials, on which neither line segment was longer.
(The 2 warm-up trials were always random-matrix trials.) The
details of the patterns are described separately for each experiment.
Experiment 2 was complete at the end of this block.
Participants in Experiments 1 and 3 then completed a third block
of eight trials, which had a design similar to that of the experiments
reported by Mack, Linnet, et al. (1992) and Rock et al. (1992). This
block of trials is referred to as the inattention block. The first three
trials as well as the fifth and sixth trials of this block were
Experiment 1
Intro Exp (1): Write the main purpose
Experiment 1 took advantage of the Ponzo illusion (also
known as the railroad-track illusion): A line segment that is
located near the narrow end of two converging lines appears
longer than an identical line segment that is located nearer
to the diverging end (see Figure 2A). The dots on the
pattern-matrix trims in this experiment were colored such
that if grouped, they would form two converging lines, as in
the Ponzo illusion. Recall that the line segments on the
pattern-matrix trials were actually the same length. If participants are unbiased in their responses, then they should
respond top equally often as bottom. If the dots are grouped,
however, then the pattern of converging lines may cause
participants to perceive the line segment that is near to the
converging end o f the pattern as longer than the other
segment. If this occurs, then it would indicate that grouping
by similarity occurred.
Me~od Method for Exp (1)
Stimuli. Figure 6 provides an illustration of the two types of
pattern-matrix trials that we used in Experiment 1. The black dots
70
344
MOORE AND EGETH
Block 2
Illusion
Block 1
Practice
Block 3
Inattention
/
random-ma~
.~_ s
16p a ~ trials
16
10 random-matrix trials
Trial I
random
matrix
Trial 2
mQ~
random
matrix
IBI
. .[
pattern
matrix
+
Questions
Full
Attention
Trial 8
Trial
• . .
i •..
Inattention
Trial 4
Divided
Attention
7
Trial6
Trial 5
random
matrix
random
matrix
Trial 3
(Experiments '
random
matrix
pattern
matrix
+
Questions
i
•°•
i
ma~Jx
+
Questions
Method for Exp (2): Complete the figure
Figure 4. Structure of the experiments. Each experiment began with a practice block of 10
random-matrix trials. The practice block was followed by the illusion block, in which half of the
trials were random-matrix trials and the other half were pattern-matrix trials. In Experiments 1 and
3, the illusion block was followed by the inattention block, which had a design that was similar to
that of the experiments reported by Mack et al. (1992) and Rock, Linnet, et al. (1992). The first 3
trials were random-matrix trials. The 4th trial was the inattention trial, which was a pattern-matrix
trial and was followed by the three questions concerning what the pattern had been (see text for
details). The 5th and 6th trials were again random-matrix trials. The 7th and 8th trials were the
divided-attention and full-attention trials, respectively. They were pattern-matrix trials and were
followed by the same three questions.
centered within the dot-pattern lines, which were in turn centered
within the matrix. The distance between the endpoint of a line
segment located near the converging end of the pattern and the
edge of the nearest black dot was 0.62 ° . The distance between an
formed two converging black lines, as in the Ponzo illusion. On
some trials, the converging end of the pattern pointed toward the
top (Figure 6A); on other trials, it pointed toward the bottom
(Figure 6B). The lines for the length discrimination task were
71
PERCEPTION WITHOUT ATrENTION
347
Finally, the point-biserial correlation provides evidence
that the bias to report top that was observed in the illusion
block was probably not caused by a few participants who
happened to notice the patterns. Participants who got the
forced-choice question fight following the inattention trial
were no more likely to have shown the bias in the illusion
block than those who got the forced-choice question wrong.
On the whole, the results suggest that grouping by similarity did occur without attention but that the patterns could
not be remembered for subsequent report. This may be
because the patterns were forgotten or because they were
never successfully encoded. Dis Exp (1): State main finding
A possible criticism of our results is that although participants seemed to have experienced the Ponzo illusion, it is
possible that the responses were biased for reasons that do Dis Exp (2): Mention possible confounding
not involve grouping by s i ~ t y .
Specifically, the endFigure 8. Typical pattern-matrix display from Experiment 2. All
point of the line segment that appeared near the converging
but eight of the dots were white. Four of the black dots surrounded
the upper line segment by the same distance as the closest dots
end of the pattern on the pattern-matrix trials was closer to
were to the upper line segment in the illusion block of Experiment
a black dot than was the endpoint of the line segment that
1. The remaining four black dots surrounded the lower line segwas nearer to the diverging end of the pattern. It is possible
ment
by the same distance as the closest dots were to the lower line
that the line segment nearer to the converging end appeared
segment in the illusion block of Experiment 1. The two line
longer because the line segmentmwhieh was also blackm
segments for the length discrimination task were identical and
was blurred into the nearby black dot.
were centered within the black dots, which were, in turn, centered
Such blurring might occur for two different reasons. First,
within the matrix.
the line segments were presented parafoveally, where acuity
is not as good as at fixation. The line segment and nearby
black dot may have appeared continuous because the spatial
maximize the chances of blurring, thus providing a conservative
resolution was not sufficient to detect the gap between them.
control for Experiment 1.
Second, participants may have deliberately squinted in orProcedure. Experiment 2 included only the practice block and
der to blur their vision and make one of the line segments
the illusion block (see Figure 4). Other than this difference, the
appear longer than the other on the ambiguous patternprocedure was identical to that of Experiment 1.
matrix trials. Experiment 2 provided a control for these
alternative interpretations of the results of Experiment 1.
Results
The mean percent correct for reporting which line segment was longer on the random-matrix trials in the illusion
block was 88.44% (_+3.46%). For the pattern-maaix trials,
on which the two line segments were the same length, the
mean percentage of trials on which the top one was reported
as appearing longer than the bottom one was 49.06%
(+_4.68%). The 50% value, which was expected by chance,
is contained within this interval.
Experiment 2
Intro Exp (2): Critical manipulation needed to answer w1
The pattern-matrix trials in Experiment 2 included the
black dots that were nearest to the endpoints of the line
segments on the pattern-matrix trials in Experiment 1, but
they did not include the black dots that formed the rest of
the pattern. If the bias in responses on pattern-matrix trials
that was observed in Experiment 1 was caused by a blurring
together of the line segment and the nearby black dot, then
the bias should occur in this experiment as well. If instead
the bias was caused by the Ponzo pattern that was present in
the pattern-matrix trials, then it should not occur in this
experiment.
Discussion
Res Exp (3): Do not list the non-significant
results unless it is critical
The results provide no evidence to suggest that the bias
observed on the pattern-matrix trials in Experiment 1 was
caused by a blurring of the line segment with a nearby black
dot. The same opportunity was present in Experiment 2, but
no bias was observed.
In addition, the lack of bias on the pattern-matrix trials of
this experiment provides further evidence that participants
were not, in general, biased to report top more often than
bottom. If there were such a bias, then it would have
appeared in this experiment as well.
Given the results of Experiment 2, it seems likely that
participants' responses were influenced by the Ponzo illusion in Experiment 1, and therefore that grouping by sinai-
Method
Stimuli. Figure 8 provides an illustration of the pauem-matrix
trials that were used in Experiment 2. The pattern consisted of
eight black dots: four surrounding the endpoints of the top line
segment by the same distance as in Experiment 1 (i.e., 0.60 °) and
four surrounding the endpoints of the bottom line segment by the
same distance as in Experiment 1 (i.e., 2.60°). In Experiment 1, the
line segments were surrounded by only two dots at those exact
distances (see Figure 6A). Four dots were used in Experiment 2 to
74
PERCEPTION WITHOUT ATTENTION
351
requires focused attention. If Gestalt grouping cannot be
References
done under conditions of divided attention, then surely it
Baylis, G. C. (1994). Visual attention and objects: Two-object cost
cannot be done under conditions of inattention. Perhaps the
with equal convexity. Journal of Experimental Psychology: Huinattention method has failed to isolate conditions of
man Perception & Performance, 20, 208-212.
inattention.
Baylis, G.C., & Driver, J. (1992). Visual parsing and response
One reason for the difference in results between our study
competition: The effects of grouping. Perception & Psychophysand that of Ben-Av et al. (1992) may be that the displays
ics, 51, 145-162.
used in the two studies differed in terms of how easily
Baylis, G.C., & Driver, J. (1993). Visual attention and objects:
grouping could occur. The grouping patterns in our study
Evidence for hierarchical coding of location. Journal of Experwere quite salient: Black and white dots on a gray backimental Psychology: Human Perception and Performance, 19,
ground were used for grouping by similarity. In addition,
451-470.
the matrices were perfectly aligned, with symmetrical dots
Beck, J. (1982). Textural segmentation. In J. Beck (Ed.), Organization and representation in perception (pp. 285-317). Hillsforming straight rows and columns. In contrast, the groupdale, NJ: Erlbaum.
ing patterns in the Ben-Av et al. (1992) displays were not as
Ben-Av, M. B., Sagi, D., & Braun, J. (1992). Visual attention and
salient. In some experiments, white plus signs and Ls on a
perceptual grouping. Perception & Psychophysics, 52, 277-294.
black background were used for grouping by similarity. In
Braun,
J., & Sagi, D. (1990). Vision outside the focus of attention.
other experiments, displays with differences in distance as
Perception & Psychophysics, 48, 45-58.
small as 20% were used for grouping by proximity. Finally,
Braun, J., & Sagi, D. (1991). Texture-based tasks are little affected
their stimuli were randomly rotated and were jittered. It is
by a second task which requires peripheral or central attentive
therefore possible that Gestalt grouping does occur preatfixation. Perception, 20, 483--500.
tentively when the patterns are relatively salient but not
Bravo, M., & Blake, R. (1990). Preattentive vision and perceptual
when the patterns are less easily resolved.
groups. Perception, 19, 515-522.
A second reason that Ben-Av et al. (1992) may have
Brown, J. M., Weisstein, N., & May, J. G. (1992). Visual search
for simple volumetric shapes. Perception & Psychophysics, 51,
failed to find evidence of Gestalt grouping is that the
40-48.
method of report may have required that the patterns be
Driver, J., & Baylis, G. C. (1989). Movement and visual attention:
encoded in memory. Participants were asked to concentrate
The spotlight metaphor breaks down. Journal of Experimental
on the primary task. They may therefore have had to rely on
Psychology: Human Perception and Performance, 15,
their memories to make reports concerning the grouping
448 -456.
patterns. If our conclusion is correct---that is, that Gestalt
Driver, J., & Halligan, P. W. (1991). Can visual neglect operate in
grouping patterns cannot be encoded in memory without
object-centered coordinates? An affirmative single-ease study.
attention--then it follows that participants would be poor at
Cognitive Neuroscience, 8, 475-496.
reporting grouping patterns under these conditions.
Duncan, J. (1984). Selective attention and the organization of
A final criticism of our study concerns how our evidence
visual information. Journal of Experimental Psychology: Genfor preattentive Gestalt grouping can be reconciled with
eral 113, 501-517.
previous evidence that grouping occurs after the resolution
Duncan, J., & Humphreys, G. W. (1989). Visual search and stimulus similarity. Psychological Review, 96, 433-458.
of several perceptual constancies (Rock & Brosgole, 1964;
Egly, R., Driver, J., & Rafal, R. D. (1994). Shifting visual attention
Rock, Nijhawan, et al., 1992) and after amodal completion
between objects and locations: Evidence from normal and parihas occurred (Palmer, 1996). There are two possible reasons
etal lesion subjects. Journal of Experimental Psychology: Genfor this apparent conflict. First, it is possible that all of these
eral 123, 161-177.
perceptual phenomena--grouping, constancies, and arnodal
Epstein,
W., & Babler, W. (1989). Perception of slant-in depth is
completion--occur preattentively. Of these three phenomautomatic. Perception & Psychophysics, 45, 31-33.
ena, only grouping has been tested using the inattention
Epstein, W., & Babler, W. (1990). In search of depth. Perception
method. Second, it is possible that grouping occurs at mul& Psychophysics, 48, 68-76.
tiple levels of processing. Neither of these possibilities can
Gibson, B. S. (1994). Visual attention and objects: One versus two
be ruled out by any of the data discussed here.
or convex versus concave? Journal of Experimental Psychology:
In summary, our results indicate that at least one form of
Human Perception and Performance, 20, 203-207.
Gestalt grouping can occur under conditions of inattention,
Grabowecky, M., Robertson, L. C., & Treisman, A. (1993). Preattentive processes guide visual search. Journal of Cognitive
insofar as inattention is defined within Mack et al.'s (1992)
Neuroscience, 5, 288-302.
and Rock, Linnet, et al.'s (1992) method. Although particGrossberg, S., Mingolla, E., & Ross, W.D. (1994). A neural
ipants may be unable to report what grouping patterns
theory of attentive visual search: Interactions of boundary, surappeared on a previous trial, their behavior can be systemface, spatial, and object representations. Psychological Review,
aticaUy influenced by what patterns are present on a current
101, 470-489.
trial. These results, in conjunction with those of Mack et al.
Halligan, P. W., & Marshall, J. C. (1993). When two is one: A case
and Rock, Linnet, et al. suggest that Gestalt grouping does
study of spatial parsing in visual neglect. Perception, 22,
occur preattentively but that encoding the results in memory
309-312.
may require attention. This conclusion is consistent with the
Hochberg, J. (1974). Higher-order stimuli and interresponse contheories of perception that maintain that substantial perceppiing in the perception of the visual world. In R. B. MacLeod &
tual organization precedes the allocation of attention within
H.L. Pick (Eds.), PercePtion: Essays in honor of James J.
a visual scene. Final Gen Dis (6): Write mainly the 4th WGibson (pp. 17-39). Ithaca, NY: Cornell University Press.
78
brief communications
reduced by the 1996 feed ban but remained
constant, I estimate that only two late-stage
infected animals under 30 months old will
be slaughtered for consumption in France.
In summary, robust cohort-based epidemiological analyses should form a suitable framework for re-examination of the
potential risks posed by the consumption
of beef from countries with native-born
BSE cases.
Christl A. Donnelly
Department of Infectious Disease Epidemiology,
Imperial College School of Medicine, St Mary’s
Campus, Norfolk Place, London W2 1PG, UK
e-mail: [email protected]
W3
W2
a
What you see
is what you hear
ision is believed to dominate our
multisensory perception of the world.
Here we overturn this established view
by showing that auditory information can
qualitatively alter the perception of an
unambiguous visual stimulus to create a
striking visual illusion. Our findings indicate that visual perception can be manipulated by other sensory modalities.
We have discovered a visual illusion that
is induced by sound: when a single visual
flash is accompanied by multiple auditory
beeps, the single flash is incorrectly perceived as multiple flashes. These results
were obtained by flashing a uniform white
disk (subtending 2 degrees at 5 degrees
eccentricity) for a variable number of times
(50 milliseconds apart) on a black background. Flashes were accompanied by a
variable number of beeps, each spaced 57
milliseconds apart. Observers were asked to
judge how many visual flashes were presented on each trial. The trials were randomized and each stimulus combination
was run five times on eight naive observers.
Surprisingly, observers consistently and
incorrectly reported seeing multiple flashes
whenever a single flash was accompanied
by more than one beep (Fig. 1a). Control
conditions and catch trials (Fig. 1 legend)
indicate that the illusory flashing phenomenon is a perceptual illusion and is not due
to the difficulty of the task, cognitive bias or
other factors.
Figure 1b shows that observers’ performance was the same, irrespective of
whether a single flash was accompanied by
two beeps, or two flashes by one or no
beeps, suggesting that the illusory double
flash is perceptually equivalent to the physical double flash. This was confirmed by the
testimonies of the observers after the exper-
V
788
4
3
2
1
n=8
1
b
Number of perceived flashes
W1
Illusions
Number of perceived flashes
1. Office International des Epizooties. http://www.oie.int/eng/
info/en_esbmonde.htm (as of 1 December 2000).
2. French Ministry of Agriculture and Fisheries. http://
www.agriculture.gouv.fr/alim/sant/mala/cell-testESB/
Esb/1-1711.pdf (as of 1 December 2000).
3. Donnelly, C. A., Santos, R., Ramos, M., Galo, A. & Simas, J. P.
J. Epidemiol. Biostat. 4, 277–283 (1999).
4. Anderson, R. M. et al. Nature 382, 779–788 (1996).
5. Donnelly, C. A., Ghani, A. C., Ferguson, N. M. & Anderson,
R. M. Nature 389, 903 (1997).
6. Ferguson, N. M., Donnelly, C. A., Woolhouse, M. E. J. &
Anderson, R. M. Phil. Trans. R. Soc. Lond. B 352, 803–838 (1997).
7. Donnelly, C. A. & Ferguson, N. M. Statistical Aspects of BSE
and vCJD: Models for Epidemics (Chapman & Hall/CRC,
London, 1999).
8. Doherr, M. G., Heim, D., Vandevelde, M. & Fatzer, R. Vet. Rec.
145, 155–160 (1999).
9. Butler, D. Nature 382, 5 (1996).
10. Schreuder, B. E. C., Wilesmith, J. W., Ryan, J. B. M. & Straub,
O. C. Vet. Rec. 141, 187–190 (1997).
11. Summary of the Spongiform Encephalopathy Advisory
Committee Meeting on 29 November 1999; News Release,
22 December 1999.
4
2
3
Number of beeps
4
3
W4
2
1
n=8
1
2
3
Number of flashes
4
Figure 1 Illusory flashing. a, Perceived number of visual flashes
by eight observers plotted as a function of the number of auditory
beeps for a single flash. The number of perceived flashes did not
increase linearly with the third and fourth beeps because they fell
outside the optimal window of audiovisual integration, as revealed
by our next experiment. b, Perceived number of flashes by eight
observers plotted as a function of the actual number of flashes
presented for trials with no sound (dashed line), and trials with
single beeps corresponding to catch trials (grey line). Observers
performed the task very well in the absence of sound (dashed
line). The results of the catch trials (grey line) confirm that the
observers’ responses were not determined by their auditory percepts. The curve in a (for a single flash) is superimposed for comparison. Further details can be obtained from L.S.
iment. The performance of non-naive subjects (results not shown) indicated that the
illusion persisted even when subjects were
aware of the fact that the disk was actually
flashed only once.
We next investigated the temporal properties of this illusion by varying the relative
© 2000 Macmillan Magazines Ltd
80
timing of visual and auditory stimuli. The
illusory flashing effect declined from 70
milliseconds separation onwards. However,
illusory flashing occurred as long as the
beep and flash were within approximately
100 milliseconds of each other, which is
consistent with the integration time of polysensory neurons in the brain1,2.
Our results indicate that the illusory
flashing is caused by an alteration of visual
perception by auditory stimuli. The modification of the visual percept by sound, however, was not categorical. It was rather
selective, as sound did not have a fusing
effect when multiple flashes were accompanied by a single beep. We suggest therefore
that the direction of cross-modal interactions is partly dependent on the type of
stimulus. Consistent with previous observations in other modalities3, we propose
that the percept of a continuous stimulus in
one modality is rendered more malleable
by a discontinuous stimulus in another
modality than vice versa.
The influence of auditory cues on visual
perception has been demonstrated in other
settings, in which perceived visual intensity
is affected by the presence of an auditory
stimulus4. This influence, however, is quantitative and does not alter the phenomenological quality of the percept. Others have
shown that the perceived direction of
ambiguous visual motion is influenced by
auditory stimulation5.
Our results extend these previous findings by showing that visual perception can
be qualitatively altered by sound even when
the visual stimulus is not ambiguous. The
conditions under which this alteration
occurs — the stimulus configuration and
the task — are very simple. The illusion is
also surprisingly robust to variation in the
many parameters we manipulated (including disk eccentricity and contrast, spatial
disparity between sound and flash, shape
and texture of the flashing pattern, flash
and beep durations). The simplicity and
robustness of the illusory flashing phenomenon indicate that it reflects a fundamental
and widespread property of polysensory
mechanisms in the brain.
Ladan Shams*, Yukiyasu Kamitani*,
Shinsuke Shimojo*†
W4
*California Institute of Technology,
Division of Biology, MC 139-74, Pasadena,
California 91125, USA
e-mail: [email protected]
†NTT Communication Science Laboratories,
Human and Information Science Laboratory,
Atsugi, Kanagawa 243-0198, Japan
1. Meredith, M. A., Nemitz, J. W. & Stein, B. E. J. Neurosci. 10,
3215–3229 (1987).
2. Wallace, M. T. & Stein, B. E. J. Neurophysiol. 71, 429–432
(1994).
3. Saldaña, H. M. & Rosenblum, L. D. Percept. Psychophys. 54,
406–416 (1993).
4. Stein, B. E., London, N., Wilkinson, L. K. & Price, D. D.
J. Cogn. Neurosci. 8, 497–506 (1996).
5. Sekuler, R., Sekuler, A. B. & Lau, R. Nature 385, 308 (1997).
NATURE | VOL 408 | 14 DECEMBER 2000 | www.nature.com
letters to nature
cortices in one', with twice the number of neurons underneath a
square millimetre of surface than the remainder of neocortex12. The
density of neurons in the rest of neocortex scales similarly as in V1
(ref. 12), and all of the hominoid thalamic nuclei also have neuronal
densities that vary in about the same was as in LGN13±15; altogether,
then, the relations between number of thalamic neurons and
number of neocortical neurons is about the same as the 3/2
power scaling law described above for the primary visual cortex.
The conservation of these scaling relations raises the possibility
that a similar basis for the scaling laws exists for all cortical areas. In
this view, each cortical area would be provided with a map of some
sortÐperhaps one with very abstract quantitiesÐand the job of the
cortex would be to extract some characteristic of the map at each
point that would be represented as a location code by the neurons in
each map `pixel'. Note that the information in the map need not be
supplied by thalamus; this structure would only have to determine
the number of pixels in the map. If n pixels are present in a cortical
region, then the number of neurons per pixel needed to maintain
the same resolution within a pixel as across pixels would vary as n1/2.
M
A 3/2 power relation would result.
12. Rockel, A. J., Hiorns, R. W. & Powell, T. P. The basic uniformity in structure of the neocortex. Brain
103, 221±244 (1980).
13. Armstrong, E. A quantitative comparison of the hominoid thalamus. IV. Posterior association
nucleiÐthe pulvinar and lateral posterior nucleus. Am. J. Phys. Anthropol. 55, 369±383 (1981).
14. Armstrong, E. A quantitative comparison of the hominoid thalamus: II. Limbic nuclei anterior
principalis and lateralis dorsalis. Am. J. Phys. Anthropol. 52, 43±54 (1980).
15. Armstrong, E. Quantitative comparison of the hominoid thalamus. I. Speci®c sensory relay nuclei.
Am. J. Phys. Anthropol. 51, 365±382 (1979).
16. Frahm, H. D., Stephan, H. & Baron, G. Comparison of brain structure volumes in insectivora and
primates. V. Area striata (AS). J. Hirnforsch. 25, 537±557 (1984).
17. Shulz, H.-D. Metrische Untersuchagen an den Schichten des Corpus Geniculatum Laterale tag- und
Nachtaktiven Primaten. Thesis, Johann Wolfgang Goethe-Universitaet Frankfurt (1967)).
18. Fritschy, J. M. & Garey, L. J. Quantitative changes in morphological parameters in the developing
visual cortex of the marmoset monkey. Brain Res. 394, 173±888 (1986).
Acknowledgements
Much of the work described here was done at the Santa Fe Institute and the Aspen Center
for Physics, and I thank those institutions for their support. I also thank T. Albright,
S. Gandhi and I. Brivanlou for comments on an earlier draft.
Correspondence and requests for materials should be addressed to the author
(e-mail: [email protected]).
.................................................................
Methods
To obtain the relation between the number of LGN and V1 neurons, I must make use of
three separately determined scaling relations. First, for haplorhines, the volume V of the
grey matter in V1 is related to the LGN volume v by a power law4,16 (see Fig. 2),
Interocular rivalry revealed in the
human cortical blind-spot
representation
V ˆ Ava
where A = 14.61 6 4.78 and a = 1.125 6 0.057; volumes are measured in cubic millimetres
and refer to both hemispheres.
These volumes can be converted to numbers of neurons, if the neuronal densities in
LGN and V1 are known. The numbers of neurons n in the LGN, as a function of LGN
volume v, conform to a power law17 for 23 haplorhines and 17 strepsirhines,
Frank Tong* & Stephen A. Engel²
* Department of Psychology, Princeton University, Princeton, New Jersey 08544,
USA
² Department of Psychology, University of California Los Angeles, Los Angeles,
California 90095, USA
n ˆ Bv b
with b = 0.659 6 0.06 for haplorhines and b = 0.683 6 0.22 for the strepsirhines, values
that are not signi®cantly different. The scale factor B, however, is different with a value of
0.071 6 0.025 for haplorhines and 0.046 6 0.029 for strepsirhines. This power law is based
on data from ref. 17 combined with data from ref. 4.
The relation between V1 volume V and the number of V1 neurons N also follows the
power law,
..............................................................................................................................................
N ˆ DV d
where D = 0.232 and d = 0.902. This relation is obtained by using the observation12 that
0.195 million neurons are found beneath a square millimetre of V1 surface in primates (the
value is corrected for 18% shrinkage), and the weak power law dependence of V1 thickness
t on cortical surface area S described12,18 by;
t ˆ 0:825S 0:108
When V and v are converted to N and n with the equations above, a power law results
(Fig. 1) with an exponent l:
l ˆ ad=b ˆ 1:54 6 0:072
Received 7 December 2000; accepted 31 January 2001.
1. Finlay, B. L. & Darlington, R. B. Linked regularities in the development and evolution of mammalian
brains. Science 268, 1578±1584 (1995).
2. Barton, R. A. & Harvey, P. H. Mosaic evolution of brain structure in mammals. Nature 405, 1055±
1058 (2000).
3. Andrews, T. J., Halpern, S. D. & Purves, D. Correlated size variations in human visual cortex, lateral
geniculate nucleus, and optic tract. J. Neurosci. 17, 2859±2868 (1997).
4. Stephan, H., Frahm, H. D. & Baron, G. New and revised data on volumes and brain structures in
insectivores and primates. Folia primatol. 35, 1±29 (1981).
5. Schein, S. J. & de Monasterio, F. M. Mapping of retinal and geniculate neurons onto striate cortex of
macaque. J. Neurosci. 7, 996±1009 (1987).
6. Ahmad, A. & Spear, P. D. Effects of aging on the size, density, and number of rhesus monkey lateral
geniculate neurons. J. Comp. Neurol. 334, 631±643 (1993).
7. Hubel, D. H. & Wiesel, T. N. Receptive ®elds and functional architecture of monkey striate cortex.
J. Physiol. (Lond.) 195, 215±243 (1968).
8. Bonhoeffer, T. & Grinvald, A. Iso-orientation domains in cat visual cortex are arranged in pinwheellike patterns. Nature 353, 429±431 (1991).
9. Swindale, N. V. How many maps are there in visual cortex? Cereb. Cortex 10, 633±643 (2000).
10. Issa, N. P., Trepel, C. & Stryker, M. P. Spatial frequency maps in cat visual cortex. J. Neurosci. 20, 8504±
8514 (2000).
11. Frahm, H. D., Stephan, H. & Stephan, M. Comparison of brain structure volumes in Insectivora and
Primates. I. Neocortex. J. Hirnforsch. 23, 375±389 (1982).
W1
NATURE | VOL 411 | 10 MAY 2001 | www.nature.com
81
To understand conscious vision, scientists must elucidate how the
brain selects speci®c visual signals for awareness. When different
monocular patterns are presented to the two eyes, they rival for
conscious expression such that only one monocular image is
perceived at a time1,2. Controversy surrounds whether this binocular rivalry re¯ects neural competition among pattern representations or monocular channels3,4. Here we show that rivalry
arises from interocular competition, using functional magnetic
resonance imaging of activity in a monocular region of primary
visual cortex corresponding to the blind spot. This cortical region
greatly prefers stimulation of the ipsilateral eye to that of the
blind-spot eye. Subjects reported their dominant percept while
viewing rivalrous orthogonal gratings in the visual location
corresponding to the blind spot and its surround. As predicted
by interocular rivalry, the monocular blind-spot representation
was activated when the ipsilateral grating became perceptually
dominant and suppressed when the blind-spot grating became
dominant. These responses were as large as those observed during
actual alternations between the gratings, indicating that rivalry
may be fully resolved in monocular visual cortex. Our ®ndings
provide the ®rst physiological evidence, to our knowledge, that
interocular competition mediates binocular rivalry, and indicate
that V1 may be important in the selection and expression of
conscious visual information.
Despite extensive research, the neural basis of binocular rivalry
has remained highly controversial. Speci®cally, it is debated whether
discrepant monocular patterns rival because of interocular competition or pattern competition. Human psychophysical studies have
provided evidence that rivalry results from interocular competition
among monocular neurons in primary visual cortex (V1)3. However, single-unit recordings in awake, behaving monkeys have
© 2001 Macmillan Magazines Ltd
195
letters to nature
yielded negligible evidence of rivalry-related activity in V1, inconsistent effects in visual areas V4 and MT, and strong effects in
inferotemporal cortex4±6. These neurophysiological ®ndings instead
indicate that rivalry may result from competition among incompatible pattern representations at higher levels of the visual pathway,
well after inputs from the two eyes have converged in V1.
To resolve this issue, we used functional magnetic resonance
imaging (fMRI) to monitor rivalry-related activity in a monocular
region of human V1 corresponding to the blind spot. The blind
spot is a part of the retina that has no photoreceptors; its size is
around 4 ´ 68 and it is about 158 medial to the fovea (Fig. 1a). In
human primary visual cortex, the blind spot is represented as a
relatively large monocular region (around 10 mm ´ 5 mm; J. C.
Horton, personal communication) that receives direct input
solely from the ipsilateral eye and not from the (contralateral)
a
Blind-spot eye
stimulation
Ipsilateral eye
stimulation
Blind
spot
Fovea
Blind-spot eye
Ipsilateral eye
6
6
3
3
0
0
–3
–3
–6
0
15
Time (s)
30
0
15
30
MR signal change (%)
MR signal change (%)
b
Fovea
blind-spot eye. The monocular V1 blind-spot representation is large
enough for functional imaging7.
Functional MRI has suf®cient sensitivity and temporal resolution
to detect rivalry-related responses in stimulus-selective extrastriate
areas8. We predicted that if rivalry arises from interocular competition, then the ipsilateral-responsive neurons in the V1 blind-spot
representation should show increased activity when subjects
perceive a grating pattern presented to the ipsilateral eye and
suppressed activity when subjects perceive a rivalrous grating
presented to the blind-spot eye. Such excitation and suppression
should occur even though both gratings are constantly present.
Furthermore, if rivalry is fully resolved through interocular
competition, then these neural responses during rivalry should be
identical to those evoked by actual stimulus alternations between
the ipsilateral grating and the blind-spot grating.
We performed three types of fMRI scan: V1 blind-spot localization scans, rivalry scans and stimulus alternation scans. Before MRI
scanning, subjects mapped the visual ®eld location of the right eye's
blind spot by manipulating the location and size of a black ¯ickering
circular probe. During V1 blind-spot localization scans, subjects
maintained ®xation while viewing on/off sequences of a ¯ickering
checkerboard pattern using either their left ipsilateral eye or right
blind-spot eye (Fig. 1a). The checkerboard was presented in the
region of visual space corresponding to the blind spot and its
immediate surround with a diameter of 88, almost twice the size
of the blind spot. Although the blind spot could not register the
central portion of the checkerboard, the stimulus was perceptually
®lled in owing to stimulation of the blind spot's surround9.
The V1 blind-spot representation was reliably identi®ed on the
basis of the voxels in the left calcarine sulcus that showed a greater
response to stimulation of the ipsilateral eye than of the blind-spot
eye (see Methods). Figure 1 shows the fMRI response and anatomical locus of this region in one subject. The V1 blind-spot
representation was highly monocular, responding vigorously to
stimulation of the ipsilateral eye (Fig. 1b, left) and negligibly to
stimulation of the blind-spot eye (Fig. 1b, right). In all subjects, this
monocular region was located in the depth of the calcarine sulcus
(Fig. 1c), a location that is always contained within primary visual
cortex10.
During rivalry scans, a red vertical grating was presented to one
eye and a green horizontal grating was presented to the other eye in
the visual location corresponding to the blind spot and its surround
W2
–6
Time (s)
a
c
Rivalry
Stimulus
Percept
b
Stimulus alternation
Yoked
stimulus
Figure 1 Localization of the V1 blind-spot representation. a, Viewing conditions. Subjects
maintained ®xation on a reference point while viewing a ¯ickering checkerboard pattern
(stimulus size 88, check size 18, temporal frequency 7.5 Hz, contrast 100%) with either
the ipsilateral eye or blind-spot eye. The centre of the checkerboard fell on the blind spot
(optic nerve head, size ,48 ´ 68) of the right eye but not the left eye. b, Average fMRI
responses in the V1 blind-spot representation of one subject during stimulation of the
ipsilateral or blind-spot eye. Data represent mean 6 s.e. of eight stimulus periods (on
15 s, off 15 s) and are expressed in per cent signal change relative to the mean MR level
throughout the scan. This region is activated by ipsilateral stimulation only. c, V1
representation of the right eye's blind spot in the left calcarine sulcus (three voxels
highlighted in white, voxel size 3.1 ´ 3.1 ´ 4 mm, slice plane perpendicular to calcarine).
Functional data are superimposed on high-resolution anatomical T2-weighted images.
196
Time (s)
Figure 2 Binocular rivalry and stimulus alternation tasks. a, Rivalrous oscillating sinewave gratings were presented to the blind-spot eye and ipsilateral eye (size 88). When
viewed through red and green ®lter glasses, only the green horizontal grating could be
seen through one eye and only the red vertical grating through the other eye (spatial
frequency 0.67 cycles per degree, speed 2 Hz, direction reversal every 500 ms, contrast
75%, mean luminance through matching ®lter 3.4 cd m-2). Although both gratings were
constantly present, subjects reported alternately perceiving either the red or green
grating. b, On stimulus alternation scans, the physical stimulus alternated between the
red grating and green grating using the sequence of reported alternations from a previous
rivalry scan. Subjects reported when the stimulus changed to the red or green grating.
82
© 2001 Macmillan Magazines Ltd
NATURE | VOL 411 | 10 MAY 2001 | www.nature.com
letters to nature
W3
rivalry versus stimulus alternation (F , 1). Positive responses for
the ipsilateral grating were larger than negative responses for the
blind-spot grating (F = 102, P , 0.005) but rivalry and stimulus
alternation remained equivalent across both types of neural
response.
The equivalence between fMRI responses for rivalry and stimulus
alternation indicates that it is likely that rivalry has been entirely
resolved among monocular neurons in the V1 blind-spot representation, such that neural activity entirely re¯ects the subject's
perceptual state. Thus, functionally equivalent neural responses
are observed when the subject's conscious state alternates between
the ipsilateral grating and blind-spot grating during constant
rivalrous stimulation and when the physical stimulus itself alternates between each grating shown alone.
Our results show that binocular rivalry is resolved in monocular
visual cortex and provide physiological evidence to support interocular competition. These theories predict that left-eye versus righteye inputs are alternately suppressed during rivalry because of lateral
inhibition among monocular V1 neurons3 or feedback inhibition
from V1 to monocular layers of the lateral geniculate nucleus12. In
contrast, theories of pattern competition propose that rivalry occurs
among binocular pattern neurons at much higher levels of the visual
pathway and not among monocular neurons4. Our ®ndings therefore help to resolve the neural basis of binocular rivalry.
Previous studies have found evidence of rivalry-related neural
activity but none has established the involvement of monocular
neurons4±6,8,13±18. In a single-unit study in monkeys, only 3 out of 33
V1 neurons showed signi®cant responses corresponding to conscious perception, suggesting that rivalry takes place at higher levels
of the visual pathway4. However, reanalysis revealed that across this
V1 population, responses during rivalry equalled one-third of the
magnitude of stimulus alternation responses13. These results are
more consistent with an fMRI study showing reliable rivalry
responses in human V1 that were about half the magnitude of
stimulus alternation responses13. (Unfortunately, this study could
not isolate monocular responses.) The authors suggested many
factors that might account for the stronger rivalry effects in human
V1, including interspecies differences, the indirect nature of fMRI in
estimating neural activity, and the effects of eye movements on
single-unit recordings.
In our view, the present ®nding of equally powerful rivalry and
stimulus alternation responses strongly suggests that binocular
rivalry is resolved in monocular visual cortex. Although fMRI
provides an indirect estimate of neural activity, any factors that
might in¯ate response amplitudes during rivalry would also do so
during stimulus alternation. In contrast, certain factors may have
diluted rivalry responses in previous studies. These include suboptimal viewing conditions that lead to frequent perceptual blends,
and variability in the accuracy or timing of subjects' perceptual
report. Isolating factors that weaken rivalry responses is an important direction for future research.
Although competition among binocular pattern neurons alone
cannot account for our ®ndings, it remains possible that feedback
signals from binocular neurons to monocular neurons might yield
the interocular suppression that we observe. However, such a theory
fails to explain why pattern competition should lead to selection at
the monocular level. Furthermore, it remains unclear how feedback
projections from binocular neurons might target a speci®c monocular channel. Given these dif®culties, interocular competition
provides the most compelling explanation for rivalry in monocular
visual cortex.
Our data also address a debate regarding whether common or
separate neural mechanisms underlie binocular rivalry and related
phenomena involving pattern rivalry. For example, two low-contrast
patterns presented to one eye can weakly rival with each other1,19.
Moreover, one of two dichoptic patterns can maintain dominance
even when the patterns are frequently swapped between eyes20,21.
Such rivalry probably involves high-level pattern competition. One
proposal is that pattern competition may generally account for
binocular rivalry20,22. However, our results suggest that a separate
mechanism of interocular competition entirely accounts for the
binocular rivalry in our subjects.
Finally, our ®ndings show that neurons can re¯ect conscious
perception at a much earlier level of the visual pathway than
previously thought4,23. These results have one of two implications.
One possibility is that certain aspects of conscious vision begin to
emerge at the very earliest stage of cortical processing among monocular V1 neurons. Alternatively, our ®ndings may suggest a new role
for V1 as the `gatekeeper' of consciousness, a primary cortical region
that can select which visual signals gain access to awareness. In either
case, our study ®rmly establishes the importance of primary visual
M
cortex in binocular rivalry and conscious vision.
W3
W4
W4
W4
W3
Methods
2
Normalized MR signal change
Ipsilateral grating
Blind-spot grating
Rivalry
Rivalry
Stimulus
alternation
Stimulus
alternation
–1
MRI acquisition
Subjects were scanned at the UCLA Division of Brain Mapping on a 3T General Electric
scanner using a head coil. We collected functional images using 5±6 slices oriented
perpendicular to the calcarine sulcus with the ®rst slice beginning about 10 mm anterior to
the occipital pole (slice thickness 4 mm, inter-slice distance 1 mm, in-plane resolution
3.125 ´ 3.125 mm). We used standard T2*-weighted echoplanar imaging to localize the V1
blind-spot representation (TR = 2.5 s, TE = 45 ms, ¯ip angle 808) and for rivalry and
stimulus alternation scans (TR = 1.0 s, TE = 45 ms, ¯ip angle 458). Functional images were
superimposed on high-resolution T2-weighted anatomical images (in-plane resolution
0.78 ´ 0.78 mm). A bite bar minimized head motion.
V1 blind-spot localization
–2
Figure 4 fMRI response amplitudes for rivalry versus stimulus alternation. Amplitude of
fMRI responses (peak-to-trough difference, time window 0 to +6 s) of each subject (n = 4)
for switches to the ipsilateral grating (left) and blind-spot grating (right) during rivalry
versus stimulus alternation. Responses are normalized relative to each subject's mean
response magnitude across the four conditions. fMRI responses for rivalry versus stimulus
alternation did not reliably differ (F , 1).
198
Subjects
Four healthy right-handed volunteers (three women), aged 21±28, participated. All had
normal or corrected-to-normal visual acuity and normal stereo-depth perception. Two
subjects were right-eye dominant and two were left-eye dominant. Before the fMRI
experiment, subjects received training on how to localize their blind spot in visual space
and to report their online perception during binocular rivalry and stimulus alternation.
1
0
W4
On separate fMRI scans, subjects viewed a counterphasing black/white checkerboard
pattern with either their left ipsilateral eye or right blind-spot eye as shown in Fig. 1a. (In
the actual experiment, a real ®xation point was placed around 158 to the left of the subject's
midline and the stimulus appeared roughly on the midline at a distance that closely
corresponded to the subject's horopter.) Each scan consisted of an initial 30-s rest period
followed by eight cycles of stimulation (15 s) and rest (15 s). We calculated activation maps
and difference maps using multiple regression24. A linear model consisting of sinusoidal
response functions was ®t to the fMRI activity observed during the eight cycles of visual
84
© 2001 Macmillan Magazines Ltd
NATURE | VOL 411 | 10 MAY 2001 | www.nature.com
NeuroImage 19 (2003) 1835–1842
www.elsevier.com/locate/ynimg
Rapid Communication
People thinking about thinking people
The role of the temporo-parietal junction in “theory of mind”
R. Saxea,* and N. Kanwishera,b
a
Department of Brain and Cognitive Sciences, Massachusetts Institute of Technology, Cambridge, MA, USA
b
McGovern Institute for Brain Research, Massachusetts Institute of Technology, Cambridge, MA, USA
Received 15 January 2003; revised 12 March 2003; accepted 14 April 2003
Abstract
Humans powerfully and flexibly interpret the behaviour of other people based on an understanding of their minds: that is, we use a
“theory of mind.” In this study we distinguish theory of mind, which represents another person’s mental states, from a representation of the
simple presence of another person per se. The studies reported here establish for the first time that a region in the human temporo-parietal
junction (here called the TPJ-M) is involved specifically in reasoning about the contents of another person’s mind. First, the TPJ-M was
doubly dissociated from the nearby extrastriate body area (EBA; Downing et al., 2001). Second, the TPJ-M does not respond to false
representations in non-social control stories. Third, the BOLD response in the TPJ-M bilaterally was higher when subjects read stories about
a character’s mental states, compared with stories that described people in physical detail, which did not differ from stories about nonhuman
objects. Thus, the role of the TPJ-M in understanding other people appears to be specific to reasoning about the content of mental states.
© 2003 Elsevier Science (USA). All rights reserved.
W4
Keywords: fMRI; Social cognitive neuroscience; False belief; Mentalising; Superior temporal sulcus; EBA
W1
The remarkable human facility with social cognition
depends on a fundamental ability to reason about other
people. Specifically, we predict and interpret the behaviour
of people based on an understanding of their minds: that is,
we use a “theory of mind.”1 In this study we show that a
region of human temporo-parietal junction is selectively
involved in reasoning about the contents of other people’s
minds.
Brain regions near the temporo-parietal junction (TPJ)
have been implicated in a broad range of social cognition
tasks (Allison et al., 2000; Gallagher and Frith, 2003; Green
and Haidt, 2003). Regions near the TPJ have preferential
responses to human faces (e.g., Hoffman and Haxby, 2000),
bodies (e.g., Downing et al., 2001) and biological motion
(e.g., Grossman et al., 2000). There is also some evidence
that regions within human TPJ are involved in theory of
mind (ToM). A number of studies have reported increased
responses in the TPJ when subjects read verbal stories or see
pictorial cartoons that require inferences about a character’s
(false) beliefs, compared with physical control stimuli
(Fletcher et al., 1995; Brunet et al., 2000; Gallagher et al.,
2000; Castelli et al., 2000; Voegely et al., 2001. A number
of other brain regions have also been implicated in theory of
mind; see reviews by Gallagher and Frith, 2003, and Greene
and Haidt, 2003).
What is the role of the TPJ in these tasks? ToM reasoning
depends upon at least two kinds of representation: a representation of another person per se and a representation of
that other person’s mental states (see Leslie, 1999). While a
representation of a person per se is a likely prerequisite for
* Corresponding author. Department of Brain and Cognitive Sciences,
MIT, NE20-464, 77 Massachussetts Avenue, Cambridge, MA 02139,
USA. Fax: ⫹1-617-258-8654.
E-mail address: [email protected] (R. Saxe).
1
The term “theory of mind” has a more restricted sense, referring to the
suggestion that the structure of knowledge in the mind is analogous to a
scientific theory (e.g., Carey, 1985; Wellman and Gelman, 1992). For
discussions about the so-called theory-theory, see Carruthers and Smith,
1996, and Malle et al., 2001. In this study, we use the term theory of mind
in a broader sense, to refer to any reasoning about another person’s
representational mental states (also called “belief-desire psychology,” e.g.,
Bartsch and Wellman, 1995).
1053-8119/03/$ – see front matter © 2003 Elsevier Science (USA). All rights reserved.
doi:10.1016/S1053-8119(03)00230-1
86
1836
Rapid Communication / NeuroImage 19 (2003) 1835–1842
ToM, achieving a representation of others’ mental states is
the core responsibility of a ToM. Some authors suggest that
the TPJ is involved only in the preliminary stages of social
cognition that “aid” ToM, not in ToM reasoning itself (e.g.,
Gallagher and Frith, 2003). We here provide evidence
against this suggestion, and argue on the contrary that a
region of the TPJ is selectively involved in representation
other peoples’ mental states.
Neuroimaging studies have followed developmental psychology in using “false belief” stories as the prototypical
problem for ToM reasoning (Fletcher et al., 1995; Gallagher
et al., 2000; see also Vogeley et al., 2001). In these scenarios, a character’s action is based on the character’s false
belief (Wimmer and Perner, 1983). False beliefs provide a
useful behavioural test of a ToM, because when the belief is
false, the action predicted by the belief is different from the
action that would be predicted by the true state of affairs
(Dennett, 1978). Note, though, that everyday reasoning
about other minds, by adults and children, depends on
attributions of mostly true beliefs (e.g., Dennett, 1996; Bartsch and Wellman, 1995).
Previous investigations of the neural correlates of ToM
(Fletcher et al., 1995; Gallagher et al., 2000) have compared
false belief (“theory of mind”) stories with two control
conditions: “non-theory of mind stories,” which describe
actions based on the character’s true beliefs, and “control”
stories, consisting of unrelated sentences. These authors
found that the TPJ response was high during theory of mind
stories, but was also high during non-theory of mind stories.
They concluded (see also Gallagher and Frith, 2003) that the
TPJ is not selectively involved in ToM. This conclusion
does not follow. Because the non-theory of mind stories
invite inferences about the character’s (true) beliefs, a region involved in reasoning about other minds should show
a high response to these stories, as well as to the so-called
theory of mind stories. (For an argument against the use of
unrelated sentences as the baseline condition, see Ferstl and
von Cramon, 2002.)
We propose two basic tests for a region selectively involved in ToM reasoning. First, it must show increased
response to tasks/stimuli that invite ToM reasoning (about
true or false beliefs) compared with logically similar nonsocial controls. Second, the region must respond not just
when a person is present in the stimulus, but specifically
when subjects reason about the person’s mental states. Below, we provide evidence that a subregion of the TPJ, here
called the TPJ-M, passes both these criteria for a selective
role in ToM.
his/her false belief. Descriptions of human actions required
analysis of mental causes, in the absence of false beliefs.
We compared these conditions to two non-social control
conditions, (1) mechanical inference control stories, which
required the subject to infer a hidden physical (as opposed
to mental) process, such as melting or rusting (for examples,
see Appendix 1), and (2) descriptions of nonhuman objects.
Unlike previous studies, we did not cue or instruct subjects to attend specifically to mental states. With this design
we were able to look for regions of cortex in individual
subjects that are selectively and spontaneously involved in
understanding the mental (as opposed to physical) causes of
events.
To test whether the response to ToM stories was a response to the presence of a person in the stimulus, we
presented still photographs of people, and nonhuman objects. Downing et al. (2001) reported a bilateral region near
the posterior superior temporal cortex that responds preferentially to the visual appearance of human bodies, compared
with a range of control objects (the extrastriate body area,
EBA). We tested directly the functional and anatomical
relationship between the EBA and the (proposed) TPJ-M.
W2
W2
Methods
Twenty-five healthy right-handed adults (12 women)
volunteered or participated for payment. All subjects had
normal or corrected-to-normal vision and gave informed
consent to participate in the study.
Subjects were scanned in the Siemens 1.5-(9 subjects)
and 3.0-T (16 subjects) scanners at the MGH-NMR center
in Charlestown, MA, using a head coil. Standard echoplanar
imaging procedures were used [TR ⫽ 2 s, TE ⫽ 40 (3 T) or
30 (1.5 T) ms, flip angle 90°]. Twenty 5-mm-thick nearcoronal slices (parallel to the brainstem) covered the occipital lobe and the posterior portion of the temporal and
parietal lobes.
Stimuli consisted of short center-justified stories, presented in 24-point white text on a black background (average number of words ⫽ 36). Stories were constructed to fit
four categories: false belief, mechanical inference, human
action, and nonhuman descriptions (Appendix 1). Each
story was presented for 9500 ms, followed by a 500-ms
interstimulus interval. Each scan lasted 260 s: four 40-s
epochs, each containing four stories (one from each condition), and 20 s of fixation between epochs. The order of
conditions was counterbalanced within and across runs.
Subjects were asked to press a button to indicate when they
had finished reading each story. Subjects read a total of 8 (4
subjects) or 12 (21 subjects) stories per condition.
Fourteen of the subjects from Experiment 1 (7 women)
were also scanned on an EBA localizer in the same scan
session, all at 3.0 T. Stimuli consisted of 20 grayscale
photographs of whole human bodies (including faces) in a
range of postures, standing and sitting, and 20 photographs
W2
Experiment 1
We devised a new version of the false belief stories task
(Fletcher et al., 1995) to compare reasoning about true and
false beliefs to reasoning about non-social control situations. ToM stories described a character’s action caused by
87
1840
Rapid Communication / NeuroImage 19 (2003) 1835–1842
W3
Fig. 4. Experiment 2. Average percent signal change in left and right
TPJ-M, defined in individual subjects (n ⫽ 14) as voxels that respond
significantly more to false belief (FB) than to false photo (FP) stories (P ⬍
0.0001, uncorrected. Response magnitude for these two conditions is illustrative only, since these data were used to determine the region of
interest). In the TPJ-M bilaterally the BOLD response to physical people
stories was significantly lower than to desire stories (P ⬍ 0.05), and not
significantly different from nonhuman description stories (P ⬍ 0.1,
repeated-analysis of variance). Response decreases are commonly observed in the TPJ vicinity during demanding nonsocial tasks (Shulman et
al., 1997; Gusnard and Raichle, 2001).
w2
during either physical people or nonhuman description stories (both paired samples t tests P ⬍ 0.05), which did not
differ from each other (Fig. 4). The left and right TPJ-M did
not differ. Thus, these regions are not involved in the detection of any person in verbal stories, but respond selectively to stories in which describe (or imply) characters’
mental states. Did any regions show the predicted profile of
a response to a person per se? At a lower threshold, a
separate whole brain analysis (P ⬍ 0.001, uncorrected) of
physical people ⬎ nonhuman descriptions revealed regions
of frontal cortex [dorsal medial prefrontal (⫺3 57 39), and
right lateral frontal cortex (39 15 54)].
Discussion
The results of Experiment 2 confirm that the TPJ-M
shows an increased response to stimuli that invite ToM
reasoning compared with logically similar nonsocial controls (false photograph stories). Second, the TPJ-M does not
show an increased response to the mere presence of a person
in the stimulus (physical people stories). The right and left
TPJ-M responses to physical people stories did not differ,
thus resolving the ambiguity of the apparently lateralized
response to photographs of bodies in Experiment 1.
General discussion
stimuli, and is not merely an effect of the difficulty or
logical structure of false belief stories, since the TPJ-M did
not respond to the more difficult and logically similar false
photograph stories.
We asked whether the TPJ-M represents the simple presence of another person (possibly via detecting a human
body and/or biological motion) or is involved specifically in
ToM. We found that the TPJ-M was anatomically and
functionally distinct from the nearby EBA (Downing et al.,
2001), which responded preferentially to the visual appearance
of human bodies, suggesting the presence of at least two
distinct regions involved in social information processing.
A key innovation of this study over previous studies was
the inclusion in Experiment 2 of physical people stories,
which described the physical appearance of human bodies.
Previous studies (Fletcher et al., 1995; Gallagher et al.,
2000) have included “physical” stories describing acting
people, which produced greater activation in the TPJ than a
scrambled sentence control. Our data show that the TPJ-M
response was no greater to stories that described other
people in physical detail than that to stories describing the
physical details of nonhuman objects—and was significantly lower than to stories that did invite a mental state
interpretation (desire stories).
Could the TPJ-M activation reflect mental imagery of the
biological motion or goal-directed action described in the
false belief, human action, and desire stories? We think this
is unlikely. Saxe, R., Xiao, D.K., Kovacs, G., Perrett, D.,
and Kanwisher, N. (unpublished data) found that the TPJ-M
response to a movie of a walking person was much lower
than its response to false belief stories. If the response of the
TPJ-M to verbal stories was merely a consequence of subjects’ imagining biological motion, we would predict the
opposite. Also the TPJ-M was doubly dissociated from its
neighbour, the pSTS-VA (visual analysis of action), which
responded more to the movies than to verbal stories.
In all, our results show that a region of the TPJ2 is
involved in reasoning about other minds, not just in understanding stories involving people per se (Gallagher and
Frith, 2003; p 80). But critically, neighbouring subregions
of cortex have different functional profiles, highlighting the
necessity of careful within-subject comparisons. The
TPJ-M, identified here by responses to (false) belief stories,
W4
2
What is the relationship between the TPJ-M and attention? Selective
attention leads to increases in regions of the TPJ during social perception
tasks (e.g., Narumoto et al., 2001; Winston et al., 2002), and to decreases
in regions of the TPJ during visual attention tasks (Shulman et al., 1997;
Gusnard and Raichle, 2001; Jiang Y., Kanwisher, N., unpublished data).
Downar et al. (2001) proposed “a role for the TPJ in detecting behaviourally relevant events in the sensory environment” (p. 1256) that is
interfered with by demanding visual attention. One possibility is that the
mental states of other people constitute a particular category of such
“behaviourally relevant” stimuli. Alternatively, these results may reflect
functionally and anatomically distinct subregions within the TPJ. Direct
testing of the relationship between the TPJ-M and selective attention is an
important avenue for future work.
W3
In two experiments we found greater BOLD response in
a region within the TPJ bilaterally (here called TPJ-M)
while subjects read stories that describe or imply a character’s goals and beliefs than during stories about nonhuman
objects. This pattern is robust across subjects, tasks, and
91
Rapid Communication / NeuroImage 19 (2003) 1835–1842
may play a broad role in social and even moral cognition
(Moll J et al., 2002; Greene and Haidt, 2003).
1841
False belief (FB) sample story
John told Emily that he had a Porsche.
Actually, his car is a Ford. Emily
doesn’t know anything about cars
though, so she believed John.
—
When Emily sees John’s car she
thinks it is a
porsche
ford
W4
Acknowledgments
This work was funded by grants NEI 13455 and NIHM
66696. Our thanks especially to Yuhong Jiang for comments and conversation, and to Ben Balas, Robb Rutledge,
Miles Shuman, and Amal Dorai for help with data collection
and analysis.
False photograph (FP) sample story
A photograph was taken of an apple hanging
on a tree branch. The film took half an hour to
develop. In the meantime, a strong
wind blew the apple to the ground.
—
The developed photograph shows the apple on the
ground
branch
Appendix
Experiment 1
Instructions: “Read each story silently to your self.
Please make sure you understand what is happening; it is
more important that you understand the story, than that you
go as fast as possible. When you are done reading the story,
press the button.”
Desire sample story
For Susie’s birthday, her parents decided
to have a picnic in the park. They wanted
ponies and games on the lawn. If it rained,
the children would have to play inside.
—
Susie’s parents wanted to have her birthday
inside
outside
Theory of mind (ToM) sample story
A boy is making a paper mache project
for his art class. He spends hours
ripping newspaper into even strips.
Then he goes out to buy flour. His
mother comes home and throws all the
newspaper strips away.
Physical people sample story
Emily was always the tallest kid in her
class. In kindergarten she was already
over 4 feet tall. Now that she is in
college she is 6⬘4⬙. She is a head taller
than the others.
—
In kindergarten Emily was over
4 ft
6 ft
. . .tall
Mechanical inference (MI) sample story
A pot of water was left on low heat
yesterday in case anybody wanted tea.
The pot stayed on the heat all night.
Nobody did drink tea, but this morning,
the water was gone.
Nonhuman description sample story
Human action sample story
Nine planets and their moons, plus various
lumps of debris called asteroids and
comets, make up the sun’s solar system.
The earth is one of four rocky planets
in the inner solar system.
—
The solar system has
four
nine
. . .planets.
Jane is walking to work this
morning through a very industrial
area. In one place the crane is
taking up the whole sidewalk. To
get to her building, she has to
take a detour.
Experiment 2
References
Instructions: “Please read each story carefully. After
each story, you will be given one fill-in-the-blanks question
about the story. Underneath will be two words that could fill
in the blank. Choose the correct word (to make the sentence
true in the story) by pressing the left button to choose the
left-hand word, and the right button to choose the right-hand
word.”
Allison, T., Puce, A., et al., 2000. Social perception from visual cues: role
of the STS region. Trends Cogn. Sci. 4, 267–278.
Bartsch, K., Wellman, H., 1995. Children Talk about the Mind. Oxford
University Press, New York.
Brunet, E., Sarfati, Y., et al., 2000. A PET investigation of the attribution
of intentions with a nonverbal task. Neuroimage 11, 157–166.
92
Neuron, Vol. 23, 747–764, August, 1999, Copyright 1999 by Cell Press
The Generality of Parietal Involvement
in Visual Attention
Ewa Wojciulik*†k and Nancy Kanwisher‡§
* Department of Psychology
University of California, Los Angeles
Los Angeles, California 90095
† Institute of Cognitive Neuroscience
University College London
17 Queen Square
London WC1N 3AR
United Kingdom
‡ Department of Brain and Cognitive Sciences
Massachusetts Institute of Technology
Cambridge, Massachusetts 02139
§ Nuclear Magnetic Resonance Center
Massachusetts General Hospital
Charlestown, Massachusetts 02129
W1
Summary
Functional magnetic resonance imaging (fMRI) was
used to determine whether different kinds of visual
attention rely on a common neural substrate. Within
one session, subjects performed three different attention experiments (each comparing an attentionally demanding task with an easier task using identical stimuli): (1) peripheral shifting, (2) object matching, and (3)
a nonspatial conjunction task. Two areas were activated in all three experiments: one at the junction of
intraparietal and transverse occipital sulci (IPTO), and
another in the anterior intraparietal sulcus (AIPS).
These regions are not simply involved in any effortful
task, because they were not activated in a fourth experiment comparing a difficult language task with an
easier control task. Thus, activity in IPTO and AIPS
generalizes across a wide variety of attention-requiring
tasks, supporting the existence of a common neural
substrate underlying multiple modes of visual selection.
Introduction
Visual attention, which is the ability to selectively process only a subset of the information present in the
retinal image, has been extensively studied over the last
two decades, using a wide variety of experimental tasks
and stimuli (e.g., Posner and Petersen, 1990; Allport,
1993; Desimone and Duncan, 1995). Despite the many
differences in these experimental paradigms, researchers generally refer to the selective process in each of
these situations as visual attention, implying a common
mechanism. However, very little work has been directed
to the crucial question of whether visual attention in fact
consists of a single general-purpose mechanism involved
in all forms of selective visual processing or whether it
instead consists of a heterogeneous set of different mechanisms, each involved in a different kind of selection. In
the present study, we used functional magnetic resonance
W2
W1
k To whom correspondence should be addressed at U.K. address
(e-mail: [email protected]).
94
imaging (fMRI) to address this question, asking whether
there is any region of the human brain that is activated
by each of three very different attention-requiring tasks
yet not activated by a language task that is difficult but
does not place heavy demands on visual attention.
Recent imaging results suggest that parietal areas,
especially the superior parietal lobule (SPL) and the intraparietal sulcus (IPS), participate in many different attention tasks and may subserve a general visual attention function. Parietal activity has been associated with
endogenous and exogenous shifts of spatial attention
(Corbetta et al., 1993; Nobre et al., 1997), maintenance
of attention on peripheral stimuli and divided attention
(Vandenberghe et al., 1997), feature integration (Corbetta et al., 1995), attentive tracking of moving dots
(Culham et al., 1998), nonspatial attention (Coull et al.,
1996; Coull and Nobre, 1998), object-based attention (Fink
et al., 1997), response selection to visually presented
stimuli (Iacoboni et al., 1996), object-oriented action
(Faillenot et al., 1997), and overt and covert attention
shifts (with and without eye movements, respectively;
Corbetta et al., 1998). A few studies have directly compared the effects of one type of attention with another
and found overlapping parietal activations (SPL and/or
IPS) for exogenous and endogenous attention (Corbetta
et al., 1993; Nobre et al., 1997), object-based and spacebased attention (Fink et al., 1997), spatial and temporal
orienting (Coull and Nobre, 1998), overt and covert attention shifts (Corbetta et al., 1998), and attentive tracking
and attention shifts (Culham et al., 1998). The great diversity of tasks used in previous studies, combined with
the similarity of parietal activations across them, suggests that some regions of parietal cortex may play a
very general role in visual attention, rather than supporting any one particular task-specific function. However,
because the different attentional tasks in prior studies
were rarely run on the same subjects, past results are
consistent with the importantly different hypothesis that
nearby but nonoverlapping cortical regions are activated by different attentional tasks, suggesting functional specificity.
To distinguish between these two alternatives, it is
necessary to examine brain activation at a fine grain
within individual subjects and to show that the same
voxels are activated in each of the different attentional
comparisons in a particular subject. In the present study,
each of seven subjects participated in three different
visual attention experiments within a single scanning
session. The tasks and stimuli were very different across
the three experiments (to provide a strong test of generality) yet identical in stimuli and motor requirements
within each experiment (to provide a relatively pure measure of attentional effects). Data were then analyzed
using a voxel-by-voxel analysis testing for Activation
Overlap across Multiple Tasks (AOMT) for each subject’s
data individually (see also Price and Friston, 1997; GrillSpector et al., 1998). Only voxels that showed significant
activation in each of the three different attentional tasks
are candidate loci for a common neural substrate involved in multiple types of visual selection.
Neuron
756
significant differences between the difficult language
task and the visual control task in any region (all twotailed ps . 0.25; note, however, that this analysis was
based on the AOMT regions of each subject individually,
therefore reducing the power to find greater activity for
the easier visual task than for the difficult language task,
as observed in the group analysis; see above).
In summary, there is no evidence that IPTO and AIPS
activity is related to general (not specifically visual) effort. Although the language task produced consistently
higher activity in left temporal cortex than did the easier
visual task (four of four subjects), this was not true in
the AOMT regions, as tested either with the KS test or
with ANOVA. Furthermore, all AOMT regions showed
significant difficulty by experiment interactions, with
strongest responses for the difficult visual task and only
weak responses for the difficult language or easy visual
tasks. It appears that these areas are specifically involved in visually demanding tasks.
Discussion
Figure 6. Interaction of Task Difficulty (Easy/Difficult) by Experiment
(E1c/E2) in the AOMT Regions
Results of the comparisons between the easy (gray bars) and difficult (black bars) tasks of each experiment are shown below the
corresponding bars in each graph (p values are one tailed for E1c).
The y axis plots PSC; error bars indicate standard error. PSC was
extracted for each subject and each AOMT separately.
overlap with the group AOMT voxels (Figure 5B). The
group analysis suggests that there may not have been
sufficient power in the individual subjects’ data to detect
this difference (note that the number of trials in E2 was
about half of that in each of E1a–E1c, since the difficult
language task required a longer interstimulus interval to
allow for the much longer response times; this procedure reduced the power in the individual subjects’
analyses).
Visual versus Nonvisual Effort: Interaction
of Experiment by Difficulty
The apparent interaction in the AOMT regions between
task difficulty and experiment (i.e., significant activations for the difficult versus easy tasks in E1 but not in
E2) was further tested by ANOVA. The ANOVA compared
the percent signal change (PSC, extracted for each subject and each AOMT separately) of easy and difficult
conditions of E1c and E2 (i.e., the two experiments that
all four subjects completed within the same session
as the language control experiment; see Experimental
Procedures). As shown in Figure 6, a significant interaction between task difficulty (easy and difficult) and experiment (E1c and E2) was found for all four AOMT
regions (all ps at least , 0.05). The significant interactions indicate that the AOMT areas responded most
strongly during the visually demanding task and only
weakly during all other tasks (difficult language or two
easy visual tasks). Planned comparisons, comparing difficult with easy conditions for each experiment separately, revealed significantly stronger activity for the
conjunction task than for the feature task in all four
AOMT regions (all one-tailed ps at least , 0.05) but no
W3
fMRI was used to determine whether multiple visual
attention tasks may activate common regions of parietal
cortex, therefore showing a generality of function rather
than functional specificity. Within a single session, subjects performed three different visually demanding tasks
(each with its own control condition) that differed widely
from each other in the kind of attentional selection involved as well as in stimuli and their spatial layout. The
common element across the three experiments was that
each contrasted one condition that placed high demands on visual attention with another condition (using
identical stimuli and matched motor components) in
which attentional requirements were minimal. The AOMT
analysis allowed us to look for exact voxel-by-voxel
overlap in activation within subjects across the three
attention-requiring tasks: peripheral shifting/maintenance
of attention, peripheral object matching, and a nonspatial conjunction task.
Two bilateral regions showed activation overlap
across all three visual attention experiments. One was
found in the posterior fundus of the IPS, at or close to
the junction with the transverse occipital sulcus (IPTO;
Brodmann’s area 19). Although the size of the IPTO overlap in individual subjects was small (averaging about
300 mm3 at the most stringent AOMT threshold), it was
very consistent across subjects (present in six of seven
subjects in the right hemisphere and seven of seven in
the left). The second area of activation overlap was
found in AIPS in five of seven subjects; for all five subjects, this area lay on the medial bank of the IPS (i.e.,
SPL, Brodmann’s area 7). The left AIPS was about the
same size as the more posterior AOMT at IPTO, while
the right AIPS was substantially larger (about 1 cm3).
The group analysis confirmed these results, showing
massive bilateral activation overlap through the entire
extent of the IPS, from its posterior end at IPTO, where
it extended into posterior SPL in the right hemisphere,
to the anterior segment close to the PCS (Figure 2B).
The large size of the activation overlap in the group
data (15.4 and 6.3 cm3 in right and left hemisphere,
respectively) suggests that the individual subjects’ analyses underestimated the extent of the overlap. On the
103
Parietal Involvement in Visual Attention
757
other hand, averaging across subjects produces spatial
blurring due to imperfect coregistration and differences
in anatomical and functional organization across subjects, which can possibly result in an overestimate of
the overlap in the group data. Nonetheless, even when
the group AOMT threshold was increased to p , 10215
(i.e., single comparison threshold of p , 1025), the IPS
overlap remained extensive (7.0 and 2.3 cm3 in right and
left hemisphere, respectively) but began to differentiate
into a posterior focus at IPTO and a more anterior IPS
focus, as observed for individual subjects. Thus, the
group and the individual subject analyses yielded a
highly consistent and robust pattern of activation overlap in at least two bilateral areas in the IPS. (Note, however, that it is possible that future research with higher
spatial resolution may uncover functional specialization
at a finer grain than we can resolve with present techniques.)
The overlapping activations at IPTO and AIPS suggest
that these regions perform a more general attention
function than could be inferred based on any one experiment in this study or on previous studies with single
comparisons. Alternative accounts of the overlapping
activations in terms of either more general processes
(e.g., any effortful task) or more specific ones (e.g., attentional shifting alone) can be discounted. First, while a
difficult language task (as compared with an easier visual task) performed on visually presented words produced strong activations in left superior temporal cortex, it failed to do so in IPTO and AIPS. Indeed, the
group analysis revealed stronger activity for the easier
visual task than for the difficult language task in left and
right IPTO and right AIPS, as would be expected if the
visual task placed higher demands on visual attention
than did the language task. Furthermore, a significant
interaction of experiment (E1c versus E2) by task difficulty was found in all four regions, with strongest responses for the attentionally demanding task of E1c
(conjunction task) and only weak responses for the easy
visual tasks of both experiments or the difficult language
task of E2. Thus, the IPTO and AIPS areas are not simply
activated by any effortful task.
Second, while some of the activations in both E1a
(peripheral attention shifting . central maintenance) and
E1b (peripheral object matching . central color matching) may reflect peripheral space–based selection, and
possibly active suppression of spatial distractors or of
eye movement, these mechanisms should not be active
in E1c (nonspatial conjunction . feature task). In that
comparison, all stimuli were presented at fixation and
without spatial distractors, such that subjects attended
to the same location in both conditions, and there was
no need to plan, make, or suppress eye movements.
These results show that eye movement preparation, execution, or suppression is not necessary to activate
these areas, while focused attention is sufficient for their
activation. Finally, while the stimuli and their spatial layout differed between experiments, emphasizing the general response properties of these areas, the displays
and motor requirements were kept identical within each
experiment, thus excluding any potential confounds due
to differences in retinal stimulation or motor components.
Our results fit well with the heterogeneity of attentional
W4
W4
104
deficits observed in visuospatial neglect, a disorder
characterized by a marked impairment in the ability to
detect or respond to objects in the contralesional space.
Parietal neglect patients show the classic deficits not
only in spatial attention (exogenous and endogenous
orienting as well as feature integration in the contralesional space; e.g., Posner et al., 1984; Eglin et al., 1989)
but also in object-based attention (e.g., Driver and Halligan, 1991; Behrman and Tipper, 1994) and temporal
(nonspatial) attention (Husain et al., 1997). Husain et
al. (1997) showed that when neglect patients have to
identify a target object in a rapid visual stream, their
ability to detect a second object is profoundly impaired,
even though both objects are presented in the same
(foveal) location. Furthermore, the severity of the temporal impairment correlated strongly with the magnitude
of the spatial bias, suggesting a close link between spatial and temporal attention, as observed in our data.
These and our results indicate that parietal function cannot be reduced to one single mode of visual selection;
instead, our data suggest that IPTO and AIPS reflect a
common neural substrate underlying multiple modes of
visual selective processing.
IPTO and AIPS: Location and Function
A comparison of the location of IPTO with respect to
the retinotopic areas of the human brain (Sereno et al.,
1995; Tootell et al., 1997) suggests that it is located in
the vicinity of area V3A, which is transected by the TrOS
but which does not appear to continue in the IPS. In
humans, area V3A has a retinotopic representation of
the whole contralateral visual field and shows high responsiveness to motion (Tootell et al., 1997). However,
our stimuli were stationary. In addition, the overlapping
activation at IPTO often extended beyond the junction
into posterior IPS, a region that is likely to be beyond
the borders of V3A. These considerations suggest that
IPTO may constitute a functional and anatomical region
that is separate from V3A. In a recent paper, Tootell et
al. (1998) found that, among other areas, both V3A and
“V7” (a newly discovered retinotopic visual area adjacent to V3A) were strongly modulated by spatial attention. Based on the anatomical location of V7 and its
susceptibility to attentional manipulations, it is possible
that IPTO corresponds to V7.
Several recent imaging papers reported activations
in an area that showed close correspondence to IPTO
coordinates. Culham et al. (1998) found a posterior intraparietal focus active during both discrete attentional
shifting and continuous attentive tracking. In addition,
Faillenot et al. (1997) reported a similar activation focus
in the right hemisphere in a task involving the matching
of (as opposed to pointing at or grasping) successive
novel shapes, and both Jonides et al. (1993) and Courtney et al. (1996) found that visuospatial memory tasks
also activate a similar right hemisphere region. More
surprisingly, a focus very similar to IPTO was activated
during the viewing of various objects (as compared with
textures) defined by luminance, texture, or motion (GrillSpector et al., 1998). Although this study did not manipulate attention explicitly, one might expect that attention
is more engaged by passive viewing of objects than of
random texture fields. These papers not only extend the
Parietal Involvement in Visual Attention
759
by areas in the IPS). This hypothesis is readily testable
and makes straightforward predictions: if some of the
IPS areas support an inhibitory function, they should be
more active in situations that require distractor suppression; on the other hand, if some of them contribute
to enhancement, they should show stronger activity in
attentionally demanding tasks than in control tasks, independent of the presence of distractors.
Whether or not this hypothesis is confirmed, the important point to note here is that none of the attentional
functions that have been previously attributed to parietal
regions (e.g., shifts of spatial attention and processes
related to eye movement preparation or execution) can
account for activity in IPTO and AIPS. Instead, the broad
range of attentional tasks that activate the same IPS
areas indicate that these regions play a more general
role in visual attention.
shifts of attention but rather generalizes to nonspatial
W4 attention
tasks as well (see also Husain et al., 1997).
Attention Shifts and Nonspatial Attention
in Parietal Cortex
Although the three comparisons generally resulted in
similar regions of activation, some parts of the SPL (lateral, but medial to the AOMT regions) appeared to be
activated uniquely in peripheral shifts of attention, while
activity in lateral IPS (IPL) appeared to be associated
uniquely with nonspatial attention (Figure 4D). The apparent specificity of these activations for the two different types of attention tasks has to be treated with caution
given the many differences between the experiments (in
particular, differences in the attended, peripheral versus
foveal, locations). Nonetheless, the procedure of running multiple attentional tasks in the same subjects
allows us to dissociate the activity in these regions from
visual effort (i.e., all three tasks compared difficult conditions with easy ones, but these two regions showed
activity specific to only one of the comparisons). Our
results thus extend previous findings that have associated spatial shifts of attention with posterior parietal
cortex (e.g., Corbetta et al., 1993; Nobre et al., 1997),
suggesting that parietal involvement in attentional shifting can be dissociated from visual effort.
Conversely, the results of the nonspatial conjunction
task show that spatial shifting of attention is not necessary to activate parietal cortex; extensive regions in the
IPS and SPL also subserve mechanisms involved in nonspatial attention, with lateral IPS/IPL apparently activating uniquely in the nonspatial task. Previous imaging
research produced conflicting results; whereas Rees
et al. (1997) found no parietal activity in a nonspatial
attention task, Coull et al. (1996) and Coull and Nobre
(1998) both reported activations in SPL or IPS. However,
the task of Coull et al. (1996) did not dissociate nonspatial attention from working memory, and Coull and Nobre
(1998) found only one parietal region more active in their
temporal than in their spatial cueing tasks (left IPS) or
in both spatial and temporal tasks as compared with
the neutral baseline (left parietal cortex). The most likely
reason why we found such extensive and highly consistent parietal activations and the previous studies found
none (Rees et al., 1997) or few (Coull and Nobre, 1998)
is that in our experiment, the displays contained multiple
temporal distractors. Our results thus clearly demonstrate that parietal activity is not constrained to spatial
Finally, our results extend the findings of Corbetta et
al. (1995), who reported parietal involvement in a spatial
conjunction task, to also include nonspatial conjunction.
Parietal cortex appears to be involved in feature integration, independent of whether the task depends on spatial (Corbetta et al., 1995) or temporal (E1c) selection,
although this conclusion is not yet definitive given the
differences in difficulty between the conjunction and
feature tasks of both studies.
Conclusion
In summary, we demonstrated that at least two regions
in the dorsal pathway, IPTO and AIPS, were activated
in each of three different visual attention tasks tested,
despite wide variation in task, stimuli, and spatial layout.
Their function cannot be reduced to one single type of
visual attention (e.g., attentional shifting alone) nor can
it be accounted for by effortful processing in general.
These findings provide evidence for the existence and
precise anatomical locus of a common neural substrate
underlying multiple modes of visual selective processing.
W4
W4
106
Experimental Procedures
Subjects
Seven healthy subjects (two men; all under 40 years old) participated
with informed consent in E1a–E1c within one session. Four of these
subjects (all women; three were native speakers of English, and one
was a nonnative but fluent speaker) also participated in E2. Two of
them participated in E2 within the same session as in E1, and two
in another session on a separate day; the latter two also performed
E1c within the second session. Six subjects described themselves
as right-handed, and one as left-handed. One subject’s data from
E1b were used to test attentional modulation of face processing
(Wojciulik et al., 1998). The experimental procedures were approved
by the Harvard University Committee on the Use of Human Subjects
in Research and by the Massachusetts General Hospital Subcommittee on Human Studies.
Experiment 1a: Peripheral Shifting versus Central Maintenance
Stimuli and Experimental Design
The displays (Figure 1A) consisted of one central dot and eight
peripheral dots arranged in a circle whose diameter was about 158.
The dots were light gray and appeared on a darker gray background.
A colored cross, one arm gray, the other half red and half blue, was
superimposed on the central dot. The cross rotated counterclockwise once every 4 s, such that with each rotation, the red (or blue)
part of the cross pointed to the next peripheral dot. The whole
display, except for the cross, blinked on and off every 400 ms. Each
time the display blinked on, the central dot or one of the four dots
indicated by the arms of the cross became smaller (i.e., there was
always one smaller dot in the display). The cross remained in a
particular orientation (e.g., red arm pointing to top right dot) for 4 s
(five blinks of the display) and then rotated counterclockwise, pointing to the next dot, and so on for all eight dots. During the 4 s when
the cross was in a particular orientation, on one of the five blinks
(determined randomly) the attended dot, peripheral or central, became smaller and reverted back to the larger size on the next blink.
The large dots were about 1.38 in size. When the dots became
smaller, they were about 18 in the periphery and 0.88 in the center.
There were 18 epochs in total, 6 epochs of the central (C) maintenance condition, which alternated with six periods of peripheral
shifting; the latter periods were split into 6 left (L) and 6 right (R)
visual field epochs. Each epoch (C, L, or R) lasted 16 s. For four
subjects, the order of the epochs was C-L-R, repeated six times,
and for three subjects, C-R-L, repeated six times. Before the central
Positive affect increases the breadth
of attentional selection
G. Rowe*, J. B. Hirsh*, and A. K. Anderson*†‡
*Department of Psychology, University of Toronto, Toronto, ON, Canada M5S 3G3; and †Rotman Research Institute, Baycrest Centre for Geriatric Care,
Toronto, ON, Canada M6A 2E1
Edited by Edward E. Smith, Columbia University, New York, NY, and approved October 29, 2006 (received for review June 21, 2006)
The present study examined the thesis that positive affect may
serve to broaden the scope of attentional filters, reducing their
selectivity. The effect of positive mood states was measured in two
different cognitive domains: semantic search (remote associates
task) and visual selective attention (Eriksen flanker task). In the
conceptual domain, positive affect enhanced access to remote
associates, suggesting an increase in the scope of semantic access.
In the visuospatial domain, positive affect impaired visual selective
attention by increasing processing of spatially adjacent flanking
distractors, suggesting an increase in the scope of visuospatial
attention. During positive states, individual differences in enhanced semantic access were correlated with the degree of impaired visual selective attention. These findings demonstrate that
positive states, by loosening the reins on inhibitory control, result
in a fundamental change in the breadth of attentional allocation to
both external visual and internal conceptual space.
PSYCHOLOGY
significantly linked with an increased capacity for creativity and
novel thinking.
The present study examined the hypothesis that the increased W1
cognitive flexibility and creative thinking associated with positive mood reflects a fundamental change in selective attention
(23–25). In contrast with the proposed tunnel vision of negative
affective states (9), positive affect may serve the opposite
function: to enhance the scope of attention (6). Preliminary
evidence for this assertion comes from studies examining global
precedence, whereby positive mood is associated with greater
global or holistic processing (i.e., seeing the forest before the
trees) versus local processing (i.e., the trees before the forest)
(23, 24). Under positive mood, individuals indicate a square
made of triangles is more similar to a square than a triangle.
Rather than a genuine change in the manner or breadth of how
attention may be allocated, such cognitive biases have been
interpreted within the affect-as-information framework (26),
whereby happy moods increase access to what is in mind during
the task at hand. In the case of global precedence, positive affect
will accentuate further a bias toward global configurations (27).
Consistent with attention as having a measurable ‘‘breadth’’ or
scope, research has suggested attentional focus can vary in its spatial
extent, which has led to the use of different metaphors in characterizing the nature of attention. For example, attention has been
compared with the beam from a spotlight (28, 29) or to the zoom
lens of a camera (30, 31), suggesting that attention can be either
narrowly focused or more widely distributed to include surrounding
stimuli. Although attention can act in ways unlike a spotlight (e.g.,
refs. 32–34), more than just a metaphor, convergent physiological
evidence from the extent of activation in primary visual cortex
suggests that attention does have a measurable spatial scope (35).
We hypothesize that positive affect may result in a relaxation of
attentional selection, thus increasing the breadth of the proverbial
spotlight of spatial attention.
As selective attention is associated with the inhibitory filtering
of task-irrelevant distraction (36), increased attentional breadth
would be reflected in a decreased capacity to inhibit processing
of spatially adjacent irrelevant information. Our operational
definition of broader attention in the visuospatial domain is thus
an impairment of spatial selective attention, resulting in a more
leaky filtering of unattended information, whereby ignored
information is more fully processed (37, 38). To this end, we used W2
the Eriksen flanker task (39) in which observers are asked to
selectively attend to a central target and ignore irrelevant
flanking distractors. Failure of selective attention is demonstrated when the to-be-ignored flankers influence performance,
indicated by a slowing in response to the central target when
flanked by response-incompatible letters. Also consistent with
attention 兩 emotion 兩 creativity 兩 inhibition 兩 problem solving
Open Gen Intro (1): General & familiar
V
NEUROSCIENCE
iewing the world ‘‘through rose-colored glasses’’ may be less
proverb and more empirical fact. Converging evidence
suggests that affective states are associated with changes in
attention that may affect differentially perception and cognition
(e.g., refs. 1–7). Attentional processes are those aspects of
cognition that allow the control of perception, thought, and
behavior and are generally acknowledged to depend on inhibitory control, such as suppression of irrelevant information and
response inhibition (8). It has been proposed that executive
control (5) and, thereby, the focus of attention may be influenced
by the current affective state of the observer (6). For instance,
it has been long hypothesized that arousal during negative
affective states is associated with a constriction of attentional
focus (7). Evidence for this narrowing of attention (9) sometimes
is referred to as ‘‘weapon focus,’’ where attention is narrowed at
the expense of encoding peripheral details (10). Although the
interaction between negative affect and attention is an active
focus of research, including attentional biases in affective disorders such as anxiety or depression (e.g., ref. 4), much less is
appreciated regarding the role positive affect and well-being may
have on attention.
Within the emerging field of positive psychology, the ‘‘broadenand-build’’ theory suggests that a primary function of positive
emotions is to broaden people’s thought-action repertoires (11, 12),
increasing their flexibility and enhancing their global scope. Consistent with such views, a robust and widely confirmed finding is that
positive affect is linked to a creative and more generative mindset
that results in greater cognitive flexibility across diverse situations,
including medical diagnosis (13, 14), industrial negotiations (15),
intuitive judgments (16), decision making (17), and creative
problem-solving tasks (18, 19). For example, on the remote associates task (RAT; ref. 20), a test of creative problem solving, people
are more likely to solve unusual word associations when they are in
a positive, compared with negative or neutral, mood (16, 18, 19, 21).
Similarly, positive mood generates more solutions to the Duncker
(22) candle task (18, 19), which can be solved only by using the
elements in an unconventional way. Thus, positive affect has been
Author contributions: G.R., J.B.H., and A.K.A. designed research; G.R. and J.B.H. performed
research; G.R. and A.K.A. analyzed data; and G.R. and A.K.A. wrote the paper.
The authors declare no conflict of interest.
This article is a PNAS direct submission.
‡To
Define abbreviation when it first appears in the text
www.pnas.org兾cgi兾doi兾10.1073兾pnas.0605198104
whom correspondence should be addressed. E-mail: [email protected].
© 2006 by The National Academy of Sciences of the USA
112
PNAS 兩 January 2, 2007 兩 vol. 104 兩 no. 1 兩 383–388
Results
8
Mood Induction. Positive and sad music induction increased and
decreased positive affect, respectively (Fig. 1). Mood valence and
overall arousal ratings were submitted to separate one-way
ANOVAs with four levels of the mood induction factor: initial/
preinduction and after neutral, positive, and negative mood
inductions. Valence ratings differed significantly depending on
induction phase [F(3, 66) ⫽ 28.22, P ⬍ 0.001]. The initial mood
of participants before mood induction and task performance was
slightly positive and was nonsignificantly decreased during neutral mood induction, t (66) ⫽ 1.57, P ⬎ 0.12. Positive [t (66) ⫽
4.81, P ⬍ 0.001] and negative [t (66) ⫽ ⫺4.23, P ⬍ 0.001] mood
inductions resulted in a similar magnitude of increased and
decreased positive affect relative to neutral conditions. By
contrast, the overall level of subjective arousal of participants
was consistent with a moderate level of alertness that did not
differ across mood induction phase [F(3, 66) ⬍ 1].
6
5
4
3
2
Fig. 1. Effect of mood manipulations. Participants rated degree of mood
valence after neutral, happy, and sad mood manipulations (from 1 extremely
unpleasant to 9 extremely pleasant). Dotted line, initial mood.
RAT. Consistent with a broader spread of semantic activation,
positive mood was associated with increased access to remote
semantic associations (Fig. 2a). The number of completed
remote associate items was submitted to a repeated measures
ANOVA with mood (happy, sad, and neutral) as a within-subject
variable. A main effect of mood [F(2, 46) ⫽ 3.56, P ⫽ 0.04]
revealed that RAT performance depended on mood. Significantly more RAT problems were correctly solved when participants were in a happy [mean (M) ⫽ 6.08, SD ⫽ 3.16] compared
with sad (M ⫽ 5.17, SD ⫽ 3.40, t ⫽ 2.16, P ⫽ 0.04) or neutral
(M ⫽ 4.75, SD ⫽ 2.64, t ⫽ 2.30, P ⫽ 0.031) mood. The sad vs.
neutral difference was not reliable (P ⫽ 0.50).
Average number correct (± s.e.m.)
a
8
7
6
5
4
3
2
b
60
Flanker incompatibility (ms ± s.e.m.)
the spatially limited scope of attentional focus, previous research
has found that flanker interference decreases with increasing
distance from the central target, despite response competition
demands remaining constant (39, 40). The present study manipulated distractor eccentricity to allow a more fine-grained
analysis of the influence of positive affective state on the scope
of visual selective attention. W2
If positive affect results in a more fundamental broadening of
the scope of attentional selection, then it may have a common
influence on processing of external visual stimulation and internal conceptual representations. Attentional state has been
shown to influence the scope of semantic access (41, 42).
Reduced capacity for attentional selection during positive affect
then may facilitate access to a greater diversity of semantic
W2 information (21). To index increased scope of semantic access,
we used the RAT (20), where participants are asked to override
typical semantic associations to find semantically distant or
remote associations. If positive mood is associated with an
underlying broadening of attention, from perceptual to conceptual processing, then impaired selective attention, as indexed by
the Eriksen flanker task, would be associated with facilitated
access to remote semantic associations, as indexed by performance on the RAT.
55
1
Flanker Task. Response time data were submitted to a repeated
measures ANOVA, with mood (positive, negative, and neutral),
flanker compatibility (compatible vs. incompatible), and spacing
(near, medium, and far levels) as within subject variables. Trials
with response times ⬎1,000 ms were considered incorrect and
excluded from response time analysis. Analysis of response times
was restricted to correct trials only. There was a high level of task
accuracy (93.6%).
A highly reliable main effect of flanker compatibility [F(1,
23) ⫽ 410.91, P ⬍ 0.0001] revealed that responses to incompatible trials (M ⫽ 487.56, SD ⫽ 15.75) were slower than compatible
(M ⫽ 457.68, SD ⫽ 14.62) flanker trials. There was also a main
effect of spacing [F(2, 46) ⫽ 54.75, P ⬍ 0.0001] with response
times at near (M ⫽ 491.25, SD ⫽ 16.12) significantly slower than
c
Happy
50
Sad
45
40
Neutral
35
30
25
20
15
90
80
Flanker incompatibility (ms)
Mood ratings (± s.e.m.)
7
70
60
50
40
30
20
10
10
0
5
-10
0
0
2
4
6
8
RAT (# correct)
10
12
Fig. 2. Effect of mood manipulation on task performance. (a) Correct RAT responses. (b) Magnitude of flanker task incompatibility effects in milliseconds
(incompatible minus compatible). (c) Correlation between RAT (number correctly identified) and flanker compatibility (incompatible minus compatible) under
positive mood.
384 兩 www.pnas.org兾cgi兾doi兾10.1073兾pnas.0605198104
113
Rowe et al.
W3
Response time (ms ± s.e.m.)
a
530
520
Happy
510
Sad
500
Neutral
490
480
470
460
450
440
430
Comp
500
480
460
440
420
520
500
480
460
440
420
Near
Medium
Far
Happy
540
520
500
480
PSYCHOLOGY
Incomp
520
Sad
540
460
440
420
Near
Medium
Far
Near
Medium
Far
Fig. 3. Effect of mood manipulation and flanker distance on compatibility effects. (a) Compatibility collapsed across flanker distances. (b–d) Neutral (b), sad
(c), and happy (d) mood at near, medium, and far flanker eccentricities. Comp, compatible flanking distractors; Incomp, incompatible flanking distractors.
W3
both medium [M ⫽ 462.07, SD ⫽ 13.92, F(1, 23) ⫽ 89.00, P ⬍
0.0001] and far [M ⫽ 464.54, SD ⫽ 15.61, F(1, 23) ⫽ 74.60, P ⬍
0.0001] levels of spacing. The effect of flanker spacing interacted
with flanker compatibility [F(2, 46) ⫽ 4.69, P ⬍ 0.02] such that
compatibility effects were reduced with greater flanker distance,
with near versus far spacing revealing a 45% reduction in the
effect of flanker compatibility.
Consistent with the hypothesized influence of positive affect
on visual selective attention, positive moods resulted in greater
flanker interference relative to both sad and neutral moods (Fig.
2b). There was a marginal overall slowing on positive relative to
both sad and neutral moods [F(2, 23) ⫽ 3.09, P ⬍ 0.06], which
was largely due to an interaction between mood and compatibility [F(2, 46) ⫽ 3.86, P ⬍ 0.03], with a disproportionate slowing
to incompatible flankers under positive mood (Fig. 3a). A
focused test on the interaction revealed that positive moods
resulted in greater incompatibility effects relative to neutral
[F(1, 46) ⫽ 6.56, P ⬍ 0.02] and sad moods [F(1, 46) ⫽ 4.90, P ⬍
0.04]. Sad and neutral moods did not differ (F ⬍ 1). Focused
analyses on incompatible distractor trials demonstrated that
positive moods resulted in pronounced slowing relative to both
sad [F(1, 46) ⫽ 30.37, P ⬍ 0.0001] and neutral mood [F(1, 46) ⫽
39.41, P ⬍ 0.0001]. Sad mood did not result in additional slowing
relative to neutral moods on incompatible trials (F ⬍ 1).
Even as flanker eccentricity increased, positive mood resulted
in pronounced slowing relative to negative and neutral moods
(Fig. 3 b–d). A significant three-way interaction revealed that the
effect of spacing on flanker compatibility was influenced by
mood [F(4, 92) ⫽ 3.43, P ⬍ 0.02]. Critically, under neutral mood,
the effect of flanker compatibility at the far distance was no
longer significant [F(1, 92) ⬍ 1], consistent with abolished
processing of flanker content (Fig. 3b). By contrast, under
positive mood, far incompatible f lankers maintained pronounced and greatest interference relative to compatible flankRowe et al.
ers [F(1, 92) ⫽ 31.12, P ⬍ 0.0001; Fig. 3d]. Although flanker
compatibility effects still were found during negative mood [Fig.
3c; F(1, 92) ⫽ 10.93, P ⬍ 0.002], focused analyses of incompatible
flanker trials demonstrated that positive mood resulted in highly
robust interference relative to both negative [F(1, 46) ⫽ 31.12,
P ⬍ 0.0001] and neutral moods [F(1, 46) ⫽ 40.07, P ⬍ 0.0001].
Further illustrative of the importance of mood relative to flanker
distance, during positive mood, incompatible distractors at far
distances (M ⫽ 498 ms) resulted in statistically equivalent
response times to near distractors during neutral (M ⫽ 491) and
negative (M ⫽ 501) mood (P ⬎ 0.3).
NEUROSCIENCE
Neutral
d 560
560
Response time (ms ± s.e.m.)
Response time (ms ± s.e.m.)
Comp
540
Response time (ms ± s.e.m.)
c
b 560
Incomp
Relation Between RAT and Flanker Performance. We next examined
individual differences in enhanced access to remote associates and
impaired visuospatial selective attention associated with positive
mood. Although these tasks tap putatively different domains of
cognitive function, individual differences in performance during
positive mood may reveal a common underlying influence on
information processing. After removal of 2 response time outliers
(⬎4 SDs), a significant correlation was found between the number
of remote associates correctly identified and slowed response times
associated with flanker incompatibility (incompatible minus compatible) (r ⫽ 0.49, P ⬍ 0.02; Fig. 2c). As such, during positive mood,
individuals with the greatest breadth in semantic access (indexed by
number of remote associates accessed) demonstrated the most
pronounced visuospatial attentional breadth (indexed by increased
flanker incompatibility effect). This finding did not reflect a more
general association between RAT and flanker performance, or
generalization to negative affective states, as indicated by the lack
of significant correlation under neutral (r ⫽ ⫺0.09) and negative
moods (r ⫽ 0.10).
Discussion
Relative to both neutral and sad mood, positive mood was
associated with increased capacity to generate remote associates
114
PNAS 兩 January 2, 2007 兩 vol. 104 兩 no. 1 兩 385
W3
mation (37, 48). In addition, compared with perceptual load
for familiar words (16, 18). This finding is consistent with a
proposed broadening function of positive affect (11, 12). We
(e.g., difficult perceptual discrimination), manipulating cognishow here this broadening in informational access is not retive load (e.g., working memory) results in late relative to early
stricted to its beneficial search for semantically distant associaattentional filtering (37, 38). With regard to performance on the
tions. It extends to a detrimental influence on visuospatial
Flanker task, central target processing difficulty was manipuselective attention. Positive affect impaired the ability to seleclated through perceptual load via perceptual crowding (i.e., near
tively focus on a target and thereby increased processing of
versus far flankers), which is known to impair perceptual enspatially distant flanking distractors, consistent with expanded
coding (49–51). Paralleling the present results, such perceptual
scope of the proverbial attentional ‘‘spotlight.’’ Positive moods
interference effects are suppressed by focal attention (52, 53)
thus facilitate tasks requiring a more global (24) and encomand exacerbated by more diffuse spatial attention (50). It is
passing style of information processing, such as in the RAT, but
important to note that this attentional-load account need not
impair those calling for a narrow, focused style, such as selective
suggest that decreased effort is applied under positive mood.
W4 visual attention. A buoyant mood may represent a fundamental Enhanced RAT performance indicates positive affect did not
shift in the breadth of information processing, the result of which
decrease effort or primary task engagement. Rather, we suggest
would be to cultivate a more open and exploratory mode of
the easing of inhibitory control alters the quality of attention,
attention to both exteroceptive and interoceptive sources of
resulting in a shift from a narrow focused state to a more broad
information. We did not, however, observe that sad mood
and diffuse attentional focus.
resulted in the opposite influence of positive mood (6). This may
In parallel with influences on perceptual resources, positive
reflect that mild melancholic states evoked by music are not
affect may influence the allocation of cognitive resources (5, 21)
exclusively aversive, reflecting mixed-feeling states (43). The
to postsemantic levels of analysis. As such, attentional-load
evocation of anxiety or fear-related states may be necessary for
theory also may account for the effect of positive mood on
the proposed attentional narrowing of negative affect (9, 10).
semantic processing. Similar to studies examining the role of
In addition to the hypothesized increased breadth of attenattention on perceptual encoding, examinations of semantic
tional selection during positive mood, it has been suggested that
memory have demonstrated that attention plays a prominent
affective states selectively modulate task-related neural activity
role in semantic encoding. Attention to specific semantic feathe prefrontal cortex (e.g., ref. 44), paralleling the antetures has been shown to bias semantic retrieval toward the
Mid Genwithin
Dis
(5):
rior hemispheric asymmetries thought to support positive and
attended dimension, inhibiting access to unattended associations
affect (45). Gray et al. (44) found negative states
Relate tonegative
literature
(42). Also, studies examining semantic priming under different
facilitated tasks supported by the right hemisphere, such as
arousal-related attentional states show that focal attention invisual working memory, whereas positive states benefited verbal
creases priming for strong associates, whereas diffuse attentional
working memory, supported by the left hemisphere. Evidence
focus results in increased priming for weak associates (41). With
for enhanced semantic search and impaired visuospatial selecregard to performance on the RAT, greater attention toward a
tive attention shown here may be interpreted within this heminarrow set of semantic features will inhibit retrieval of more
spheric lateralization framework. However, positive affect has
remote/weak semantic associations. For example, biased attenbeen linked with facilitating a broader, more generative mindset
tion to the most strongly associated semantic features of the cue
(11, 12, 18, 46) across diverse situations that include verbal and
‘‘widow’’ (i.e., the death of one’s husband) would inhibit access
visual materials (13, 14). As such, the pattern of facilitation and
to more remote associations (i.e., spider) in common with the
interference in the present results is unlikely related to material
cue’s ‘‘bite’’ and ‘‘monkey.’’ More diffuse attentional focus under
specificity (verbal vs. visual) but rather is indicative of a shift in
positive affect would attenuate such biases, broadening the scope
mode of attentional selection that operates on these visual and
of cognitive resources to include more semantically remote
verbal materials. Mid Gen Dis (4): Discuss scope & limitation
associations on which creative solutions depend.
An analysis of individual differences revealed that Flanker
In summation, the present results suggest a unifying frameinterference during positive mood was correlated with enhanced
work in which to understand the interaction between positive
RAT performance, suggesting a common origin. Such an unaffect, attention, and creativity. Mirroring the broad attentional
derlying increase in breadth of processing, extending from
focus and enhanced creative problem solving during positive
external perceptual to internal conceptual processing, may be
mood, individual differences in creative intelligence have been
derived from a central origin in cognitive or inhibitory control.
associated with decreased attentional filtering. For instance,
Cognitive control is thought to limit the amount of information
creative individuals have been shown to exhibit less latent
entering the focus of attention (8, 36) and, thereby, influences
inhibition, a measurement of attentional decrements to stimuli
the capacity for selective attention (47). Similar to the effect of
deemed irrelevant (54). As such, positive affect may represent a
positive mood, individual differences in cognitive control are
fundamental shift in information processing style, reflecting a
associated with the capacity for selective attention, as indexed by
relaxation of inhibitory control and, thereby, reducing the tensusceptibility to the ‘‘cocktail party effect,’’ where observers
dency to narrowly focus attention across disparate informational
report hearing their own name on an unattended ear during
domains. The result of this altered capacity for attentional
dichotic listening (47). We suggest that positive mood reflects a
selection is a broadening of thought-action routines (11), englobal relaxation of inhibitory control (8, 36), resulting in an
gendering a broad exploratory (11, 18, 55) rather than narrow
altered capacity for selective attention from early perceptual to
vigilant processing mode (25, 56). Donning the proverbial roselater postsemantic levels of analysis.
colored glasses of positive mood then may be less about the color
How altered inhibitory control may alter the extent of visuoand more the expansiveness of the view.
spatial and semantic processing may be best understood with
Final Gen Dis (7): End with conclusion
respect to the ‘‘attentional-load’’ theory of selective attention
Materials and Methods
(37), which attempts to integrate early (perceptual) and late
that fits the first paragraph in Intro
Participants and Design. Twenty-four university students (12 fe(postsemantic) models of attention. Attentional-load theory
male) participated in a multipart experiment for either course
suggests that inhibitory suppression of ignored events is related
credit or monetary remuneration. All participants were tested
to the degree (48) and type (37) of cognitive resources allocated
between 11 a.m. and 1 p.m., in the middle of the peak and
to a primary task. To the extent that attentional resources are
off-peak periods of their circadian cycle (for a review of circamore available, inhibition of ignored events will be decreased,
resulting in leaky and ineffective filtering of unattended infordian arousal patterns, see ref. 57).
386 兩 www.pnas.org兾cgi兾doi兾10.1073兾pnas.0605198104
115
Rowe et al.
Manuscript Triage
Standard definition of medical triage. The process of sorting people based on their need
for immediate medical treatment, traded off against their chance of benefiting from such
care.
Translation for publishing manuscripts. The process of sorting original (or revised) MSs
to journals based on the chance of acceptance, traded off against the quality and desired
audience of the journal.
A good triage process will obtain the greatest gain in outcome (journal quality and
audience) for the invested resources. In the case of MSs, time & energy are the most
valuable resources.
Principle 1 (develop a triage process): A lab would benefit from reaching a consensus
about a systematic process for manuscript triage.
Principle 2 (no more than two weeks of “down time”): A manuscript should not sit idle
on any ones’ desk during the writing, revising, submission, or re-submission process for
more than about two weeks. Thus, except for two week periods of “dead time” in which
nothing is being done, one of the following things always should be happening:
New articles:
(a) Selecting target journal (TJ1) and back-up 2nd tier journal (TJ2);
(b) Reviewing journal policy and instructions to authors;
(c) Writing the new MS;
(d) Peer review of the new MS by lab mates;
(e) Revising a new MS with the lab director (required: two week turn around by
author and lab director);
(f) Papercheck.com (guaranteed three day turn around).
MSs returned from journals (one or more of the following should be happening at all
times):
(a) designing and running additional subjects/new analyses required by reviewers;
(b) following “bottom-up process for writing and reviewing revisions based on
reviewer comments;
(c) peer review of revisions;
(d) Reviewing (completed) revision of a manuscript with lab director (two week turn
around for each person);
(e) Reviewing the decision about which journal to submit to;
(f) Papercheck.com (guaranteed three day turn around).
The following example of a MS triage assumes that before submission you have made in advance
a wise decision about the first tier journal and back-up second tier journal.
Legend: TJ1 = target journal 1; TJ2 = back-up target journal 2;
AE = action editor; RC reviewer comments; EC editor comments;
= follow strategy below.
Initial submission to TJ1
What is the best way
to make this decision?
BOTTOM UP!
Rejected with no RC or EC
Submit to TJ2
Rejected: do not
resubmit.
Rejected with option
to re-submit to TJ1
Strategy 1*: Revise
using any and all
“usable” RC or EC
Strategy 1*: Revise
using any and all
“usable” RC
Submit to TJ2
* You may want to take
only a few comments for
fast turn round, or you may
take most or all for a major
revision
Strategy 2: Revise
using ALL RC
Accepted pending
revisions based on RC
Strategy 3: Revise
using ALL RC:
1) Never change a
word in the MS unless
it directly is required
by an RC
2) Briefest possible
changes;
Submit to TJ2
Re-submit to TJ1
* Put your emotion aside, and think positively
about the improvement of the MS through this
revision process.
3) Letter to AE
contains itemized
response containing
each RC statement to
which you made a
change; followed by
the quote (page,
paragraph, line) of the
change that was made.
Re-submit to TJ1
Return to Checklist for Revision
Bottom-up decision strategy for
“Rejected with option to resubmit to TJ1”
(Revision is stressful; therefore, efficiency is extremely important)
Follow these five bottom-up steps to help reach a decision about whether to:
1) Respond to any and all “usable” RC and submit to TJ2, or;
2) Respond ALL RC and resubmit to TJ1.
Note: in (2), “respond to” means either: (a) to change the MS, or (b) do not
change the MS, but explain why in the letter to the editor. (examples)
Step 1: Separate and quote (“Divide and conquer”)
A reviewer’s comments may contain 5 or 6 separate criticisms within one
paragraph that require a change in the MS. Begin by separating and numbering
the Reviewer’s comments into all the parts that require a revision or a comment to
the editor (and no change to the MS).
Objective: (1) this forces you to analyze very carefully each separate instruction
for a change; and (2) it is the beginning of the letter to the editor. This may be a
very long series of separate comments. If two reviewers make the same comment,
you can list those together in a section at the end, but preferably separate each
reviewer’s comments and response to them in the letter when submitting the
revision.
Return to Checklist for Revision
Example 1
Reviewer 1
Comment 1.1:
“The reference on page 2, paragraph 3, line 12 is incorrect, It should be Smith
(2008)”
Comment 1.2:
“The authors should explain if the instructions were given before or during the
second experimental procedure.”
Comment 1.3:
“The description of the xxx is not clear on page 4, paragraph 2, line 12. This
should be spelled out clearly.”
Comment 1.4:
“The authors only mention one theoretical framework; namely, dual-process
theory. They should mention the other competing frameworks and relate these to
the results of their study.”
Step 2: Eliminate all comments you do not want to respond to
with a change in the MS (“Pull out the weeds”).
Example 2 (highlight the “weeds” in red)
Comment 1.1:
“The reference on page 2, paragraph 3, line 12 is incorrect, It should be Smith
(2008)”
Comment 1.2:
“The authors should explain if the instructions were given before or during the
second experimental procedure.”
The procedure was explained by adding on page 12, paragraph 3, line 12
Comment 1.3:
“The description of the xxx is not clear on page 4, paragraph 2, line 12. This
should be spelled out clearly.”
Comment 1.4:
“The authors only mention one theoretical framework; namely dual-process
theory. They should mention the other competing frameworks and relate these to
the study.” I don’t think we should do this.
Step 3: Revision plan: write notes to yourself about the decisions
for each separate comment. (“Don’t try to eat the whole
meal in one bite.”)
Example 3
Comment 1.1:
“The reference on page 2, paragraph 3, line 12 is incorrect, It should be Smith
(2008)”
Change as suggested
Comment 1.2:
“The authors should explain if the instructions were given before or during the
second experimental procedure.”
We can explain by noting the xxx xxx xxx.
Comment 1.3:
“The description of the xxx is not clear on page 4, paragraph 2, line 12. This
should be spelled out clearly.”
We can mention xxx and cite our earlier paper (ref) here
Comment 1.4:
“The authors only mention one theoretical framework; namely dual-process
theory. They should mention the other competing frameworks and relate these to
the study.” I don’t think we should do this.
Step 4. Consult collaborators and revise the bottom-up revision
plan (“Multiple brains are better then one.”)
Ask the co-authors to study this document. Meet with them and have a discussion
of each point. Then revise the bottom-up revision plan. . If they are not at the
same university, ask them to revise using tracked changes and return.
Step 5: Make the decision (bottom up) and revise the MS (“Decide
and revise.”)
(a) Read together and make re-submission decision (TJ1 of TJ2). Read
through the finished document and decide whether all the decisions, when
taken together, add up to:
1) Respond to any and all “usable” RC and submit to TJ2, or
2) Respond ALL RC and resubmit to TJ1
Two kinds of “responses” are possible under option 2
1) Revise the MS according to all RC.
2) Revise using most of the RC, but provide justification for why several RC
were not followed. In this case, a good statement to make is:
“One reviewer suggested that we should mention the other competing
frameworks (other than dual process theory) and relate these to the study
(Reviewer 1.4). But after carefully considering this suggestion, we ended up
agreeing with the other three reviewers who did not suggest this change.”
(b) Revise the MS in serial order. Once the decision is made (TJ1 or TJ2) assign
the authors to be responsible for revising in response to specific RC that match
their expertise. It is important that each author completes their revision and then
passes that version on the next author – Do not have multiple versions because
it’s a huge waste of time to put them all together!
Important note: As each author revises the MS, also modify this document,
because will become the letter to the editor. Each revision should be quoted
exactly just beneath the numbered RC (unless it’s a huge change to an entire
section of the MS). Send this revised letter along with the revised MS to the next
author who will revise their section. At the end, the entire letter to the editor now
will be completed, except for putting the exact page no.; paragraph no.; and line
no.; of each change.
i) Assign the revision order and revise in serial order passing from one author to
the next using tracked changes and initials after the insertions.
There are several ways to do this. First, relatively simple revisions, you could
assign the revision to the co-authors who are responsible for that portion of the
manuscript (e.g., the analysis or the experimental procedure). That is the
procedure illustrated here and below in the next sections. Second, for more
complex revisions, you could assign the order of revision by section of the
paper (e.g., method, results, discussion, etc) or by the reviewer, or by the
concept being revised. Senior author judgment is required to make these
assignments.
1) Alice will revise first, then pass the MS with her revisions to Brian.
2) Brian will do his parts, then send to Cindy
3) Cindy will complete her parts and reviews the those MS.
Example 4
Comment 1.1:
“The reference on page 2, paragraph 3, line 12 is incorrect, It should be Smith
(2008)”
Change as suggested
Alice is responsible for doing this one
Comment 1.2:
“The authors should explain if the instructions were given before or during the
second experimental procedure.”
We can explain by noting the xxx xxx xxx
Brian is responsible for this one
.
Comment 1.3:
“The description of the xxx is not clear on page 4, paragraph 2, line 12. This
should be spelled out clearly.”
We can mention xxx and cite our earlier paper (ref) here
Cindy will do this one and then review the entire MS
Comment 1.4:
“The authors only mention one theoretical framework; namely dual-process
theory. They should mention the other competing frameworks and relate these to
the study.” I don’t think we should do this.
(The MS is not revised and the nearly finished letter to the editor is on the next
page. If you submit to TJ2 you still may want to sent this letter to show openly
where you are in the revision process for this MS)
ii) Then each author completes the revision in serial order.
Example 5 (Almost finished letter to the editor)
Comment 1.1:
“The reference on page 2, paragraph 3, line 12 is incorrect, It should be Smith
(2008)”
“The first application of this paradigm was by Smith (2008)”
Page xx; paragraph yy; line zz
( by Alice)
Comment 1.2:
“The authors should explain if the instructions were given before or during the
second experimental procedure.”
“The instructions were given while the subjects was responding to the stimuli.”
page xx; paragraph yy; line zz
(by Brian )
.
Comment 1.3:
“The description of the procedure is not clear on page 4, paragraph 2, line 12.
This should be spelled out clearly.”
“The entire procedure begins with the subject receiving 25 training trials in the
same manner described by Lee and Thomas (2008).”
page xx; paragraph yy; line zz
(by Cindy)
Comment 1.4:
“The authors only mention one theoretical framework; namely dual-process
theory. They should mention the other competing frameworks and relate these to
the study.” I don’t think we should do this.
Now the MS is revised and the letter to the editor is complete except for the final
page, paragraph, and line numbers of the changes.
SPECIAL ARTICLE
How to reply to referees’ comments when
submitting manuscripts for publication
Hywel C. Williams, PhD
Nottingham, United Kingdom
Background: The publication of articles in peer-reviewed scientific journals is a fairly complex and stepwise process that involves responding to referees’ comments. Little guidance is available in the biomedical
literature on how to deal with such comments.
Objective: The objective of this article is to provide guidance to novice writers on dealing with peer
review comments in a way that maximizes the chance of subsequent acceptance.
Methods: This will be a literature review and review of the author’s experience as a writer and referee.
Results: Where possible, the author should consider revising and resubmitting rather than sending an
article elsewhere. A structured layout for responding to referees’ comments is suggested that includes the 3
golden rules: (1) respond completely; (2) respond politely; and (3) respond with evidence.
Conclusion: Responding to referees’ comments requires the writer to overcome any feelings of personal
attack, and to instead concentrate on addressing referees’ concerns in a courteous, objective, and evidencebased way. (J Am Acad Dermatol 2004;51:79-83.)
P
British Specialist Registrars in dermatology, and I am
grateful to them for helping me to develop the
learning themes.
I have deliberately not entered into any discussions on the quality of peer review8 or the value
of peer review in publication because it is still hotly
debated if peer review really helps to discriminate
between good and bad research or whether it simply
improves the readability and quality of accepted
articles.9 Instead, I have decided to stick to providing
what I hope is helpful and practical guidance within
the system that already exists.
lenty of guidance is available on conducting
good research,1,2 and Web sites of most scientific journals give clear and helpful instructions
on what is suitable for submission and how to
submit. Yet where does one obtain guidance on
replying to referees’ (peer reviewer) comments once
the manuscript is returned? I could find little in the
literature dealing with this important topic.3-7
This article attempts to address this gap by providing some helpful tips on how to reply to referees’
comments. In the absence of any systematic research
to determine which strategies are best in terms of
acceptance rates, the tips suggested below are based
simply on my personal experience of publishing
approximately 200 articles, refereeing more than 500
manuscripts, and working as an editor for 3 dermatology journals. I have presented some aspects of the
work previously in two workshops with groups of
THAT LETTER ARRIVES FROM THE
JOURNAL
After laboring for many months or years on your
research project and having written many manuscript drafts to send off your final journal submission,
a letter or electronic-mail message from the journal
arrives several weeks later indicating whether the
journal editor is interested in your manuscript. At this
stage, it is every author’s hope that the manuscript is
accepted with no changes, yet such an experience
is incredibly rareeit has happened to me only twice,
and these were both commissioned reviews. More
commonly, one of the following scenarios ensues.
From the Centre of Evidence Based Dermatology, Queen’s Medical
Centre.
Funding sources: None.
Conflicts of interest: None identified.
Reprint requests: Hywel C. Williams, PhD, Centre of Evidence Based
Dermatology, Queen’s Medical Centre, Nottingham NG7 2UH,
United Kingdom. E-mail: [email protected].
0190-9622/$30.00
ª 2004 by the American Academy of Dermatology, Inc.
doi:10.1016/j.jaad.2004.01.049
127
YMJD1792_proof 9 June 2004 3:19 pm
79
80 Williams
J AM ACAD DERMATOL
JULY 2004
Table I. Three golden rules of responding to referees’ comments
constructive suggestions to help you get the manuscript published, it might be a safer bet to follow their
wishes of a complete rewrite. It might be difficult for
the editor to then turn you down if you have done
exactly what was asked of you. If, on the other hand,
the request for a complete rewrite is a cold one, ie,
without suggestions as to exactly what needs to be
done and where, then it might be better to reflect on
the other comments and submit elsewhere.
Referees may recommend splitting a manuscript if
it is part of a large study that tries to cram in too many
different results. Such a request from one of the
referees may appear like a gift to the authoretwo for
the price of one. But a word of warningeif you are
going to redraft the original manuscript into two
related manuscripts, there is no guarantee that both
will be accepted. The best thing under such
circumstances is to have a dialogue with your editor
to test how receptive they would be to having the
manuscript split into two.
Rule 1. Answer completely
Rule 2. Answer politely
Rule 3. Answer with evidence
Accept with minor revision
If you are lucky, the letter will ask for only minor
revisions. In such circumstances, it is probably best to
simply get on with these changes without invoking
too much argument. If you send the revised manuscript back to the editor quickly, it is still likely to be
fresh in his or her mind, and you will probably get
a speedy acceptance.
Major revisions needed
The most common form of letter is one that lists 2
or 3 sets of referees’ comments, some of which are
quite major. In such circumstances, you will need to
work hard at reading and replying to each referee in
turn, following the layout and 3 golden rules (Table
1) that I will develop later in this article. Such
a process can take days to complete, so do not
underestimate the task. Only you can decide whether
such an investment of time is worthwhile. My advice
is always to revise and resubmit to the same journal if
the comments are fair, even if responding to them
takes a lot of time. Some authors go weak at the
knees when requested to do a major revision, and
instead simply send the manuscript elsewhere. This
is understandable, but the authors should still try and
make improvements to the manuscript in light of the
referees’ comments. Authors should also be aware
that in certain fields of research, their work is likely to
end up with the same referee when they send their
manuscript to another major specialty journal. It will
not go down well with that referee if they see that the
authors have completely ignored the referees’ previous comments. So, generally speaking, my advice
is to put in the time needed to make a better
manuscript based on the referees’ comments, and
resubmit along the lines suggested. If you do submit
to another journal, you should consider showing the
latest journal the previous referees’ comments and
how you have improved the article in response to
such commentsesome journal editors feel positively
about such honesty (J. D. Bernhard, MD, written
communication, November 2003).
Unsure as to rejection or possible
resubmission
The wording of some journal response letters can
be difficult to interpret. For example, phrases such as
‘‘we cannot accept your manuscript in its current
form, but if you do decide to resubmit, then we
would only consider a substantial revision,’’ may
sound like a rejection, yet in reality, it may indicate an
opportunity to resubmit. If you are unsure on how to
read between the lines, ask an experienced colleague or, better still, someone who works as a referee for that journal. Failing that, you could simply
just write back to the editor to ask for clarification.
Sometimes, a journal will ask you to resubmit your
article in letter format rather than as an original
manuscript. You then have to decide if the effort
versus reward for resubmission elsewhere is worth it,
or if you are content to accept the bird in the hand
principle and resubmit your original manuscript as
a letter.
The outright rejection
Usually this type of letter is quite short, with very
little in the way of allowing you an opportunity to
resubmit. Outright rejection may be a result of the
manuscript being unsuitable for the journal or because of ‘‘lethal’’ methodologic concerns raised by
the referees that are nonsalvageable. For example,
doing a crossover clinical trial on lentigo maligna
with an intervention such as operation that has a permanent effect on patient outcomes in the first phase
of the crossover study. Sometimes the editors, who
are always pushed for publication space, simply did
not find your article interesting, novel, or important
Journal requests a complete rewrite
Only you can decide if the effort of a complete
rewrite is worth it. If it is clear that the referees and
editor are interested in your manuscript and they are
doing everything they can to make detailed and
128
YMJD1792_proof 9 June 2004 3:19 pm
Williams 81
J AM ACAD DERMATOL
VOLUME 51, NUMBER 1
Table II. Some useful phrases to begin your replies
to critical comments
enough to warrant inclusion. You will just have to
live with that and submit elsewhere.
Dealing with outright rejection of your precious
sweat and toil may not be easy, especially if the
journal has taken ages to get back to you. You have
two main choices at this stage. If you believe that the
referees’ comments are grossly unfair or just plain
wrong, you can write to the editor to appeal the
decision and ask for new referees. The success of
such appeals depends on how confident you are that
their decision was out of order and whether the real
decision for rejection was indeed based on those
comments transferred to you. Appeals such as this
are rarely successfuleI have done it twice with the
BMJ, and failed both times.
The other (better) option is to stop snivelling, pick
yourself up, and resubmit elsewhere. If you do this, it
is important that you read and objectively assess the
referees’ comments from the journal that has turned
down your manuscript. This is for two reasons: (1)
those comments may improve the article; and (2) as
stated earlier, your manuscript may end up with the
same referee even if you send it to another journal. If
you are really convinced that your manuscript is
earth-shattering, then you should not automatically
resubmit to a journal that might offer easier acceptance. It has been my experience that sometimes
a manuscript that was rejected by a medium-ranking
dermatology journal is subsequently accepted by
a higher-ranking oneesuch is the unpredictability of
peer review and journal editor preferences.9
We agree with the referee that ___, but. . .
The referee is right to point out ___, yet. . .
In accordance with the referees’ wishes, we have now
changed this sentence to___.
Although we agree with the referee that. . .
It is true that___, but. . .
We acknowledge that our manuscript might have
been ___, but. . .
We, too, were disappointed by the low response rate.
We agree that this is an important area that requires
further research.
We support the referee’s assertion that ___, although. . .
With all due respect to the reviewer, we believe that this
point is not correct.
separate comments (eg, comment 1.1, 1.2, 1.3), then
answer them in turn. Even if some of the comments
are just compliments, repeat these in your cover
letter followed by a phrase such as ‘‘we thank the
referee for these comments.’’
Rule 2: Answer politely
Remember that nearly all referees have spent at
least an hour of their personal time in refereeing your
manuscript without being paid for it. If you (as a lead
author) receive a huge list of comments, it usually
means that the referee is trying very hard to help you
improve the manuscript to get it accepted. Rejection
statements are usually short, and do not allow you an
open door to resubmit.
It is quite all right to disagree with referees when
replying, but do it in a way that makes your referees
feel valued. Avoid pompous or arrogant remarks.
Although it is only human nature to feel slightly
offended when someone else dares to criticize your
precious work, this must not come across in your
reply. Your reply should be scientific and systematic.
Get someone else to read your responses before
sending them off.
Try to avoid opening phrases such as ‘‘we totally
disagree’’ or ‘‘the referee obviously does not know
this field.’’ Instead, try to identify some common
ground and use phrases starting with words such as
‘‘we agree with the referee, however. . ..’’ A list of
helpful phrases that I have developed over the years
is given in Table 2 for guidance.
THE 3 GOLDEN RULES OF STRUCTURING
YOUR RESPONSE LETTER
Rule 1: Answer completely
It important that all of the referees’ comments are
responded to in sequence, however irritating or
vague they may appear to you. Number them, and
repeat them in your cover letter using the headings
such as ‘‘Reviewer 1,’’ then ‘‘Comment 1,’’ followed
by ‘‘Response.’’ What you are doing here is making
the editor’s and referees’ jobs easy for themethey
will not have to search and cross-reference a lot of
scripts to discover what you have doneeit will all be
there in one clean document.
Typing out or paraphrasing the referees comments as a means of itemizing the points also
achieves two other things: (1) it forces you to listen
to what the referees actually said, rather than what
you thought they might have said when you first read
their comments; and (2) it helps you to understand
how many separate points are being made by the
referee. Quite often, you will just receive a paragraph
with several comments mixed together. In such
a situation, you can split the paragraph into 2 or 3
Rule 3: Answer with evidence
If you disagree with the referee’s comments, don’t
just say, ‘‘we disagree,’’ and move on. Say why you
disagree with a coherent argument or, better still,
back it up with some facts supported by references
that you can cite in your reply. Sometimes those extra
129
YMJD1792_proof 9 June 2004 3:19 pm
82 Williams
J AM ACAD DERMATOL
JULY 2004
references are just to back the point you make in your
cover letter, but occasionally you may add them to
the revised article. Some kind referees go to the
trouble of suggesting missed references or how you
might reword important areas of your document. If
providing the references or rewording makes sense
to you, just go ahead and incorporate them. It is quite
legitimate to use the referee’s comments to add some
extra text and data if their comments require it,
although if this amounts to more than a page, you
would be wise to suggest it as an option to the editor.
Another option is to suggest that the extensive
additions would be better placed in another subsequent article.
Sometimes, if there are no clear published data to
strongly support your methodologic approaches,
you can discuss this with an expert in the field. If
he or she agrees with your approach, then you can
say so in your reply. For example, ‘‘although other
approaches have been used in the past, we have
discussed this statistical methods with Professor Soand-So who agrees that it was the appropriate
analysis.’’
The referee is just plain rude
Anyone who has done clinical research will realize just how difficult it can be, and there is no place
for rudeness from referees. I find it sad that senior
academics can sometimes forget their humble beginnings when they referee other’s work. Nearly all
journals provide clear guidance to their referees to
avoid remarks that they would find hurtful if applied
to their own work, yet some ignore such advice and
delight in rude or sarcastic comments, possibly
because of envy or insecurity. In such circumstances,
all you need to do is complain to the editor and ask
for another nonhostile review.
The dreaded request to reduce the
manuscript by 30%
Such a request typically comes from the editor
who is pushed for space in his or her journal. I have
to confess that, for me, this is the comment that I
dread most of all because it is often accompanied by
3 referees’ comments, the response to which usually
involves making the article longer than the original
submission. A general reduction in text by 30%
basically requires a total rewrite (which is slow and
painful). It is usually easier to make a brave decision
to drop an entire section that adds little to the
manuscript. Ask a colleague who is not involved in
the manuscript to take out their editing knife and
suggest nonessential areas that can go—even though
the process of losing your precious words may seem
very painful to you. Discussion sections are usually
the best place to look for radical excisions of entire
paragraphs. Background sections should be just one
to two paragraphs long—just long enough to say
why the study was done, rather than an exhaustive
review of all previous literature. Please do not skimp
on the methods section unless you are referring to a
technique that can be put on a Web site or referenced.
TIPS ON DEALING WITH OTHER
SCENARIOS
Referees with conflicting viewpoints
At first, this scenario might appear very difficult to
the novice, yet it should be viewed as a gift. You, the
author, have the choice of which viewpoint you
agree with the most (or better still, the one that is
right). Then it is simply a question of playing one
referee against the other in your reply. You can
always appeal to the editor by asking him or her to
make the final decision, but give them your preferred
option with reasons.
The referee is wrong
Referees are not gods, but human beings who
make mistakes. Sometimes they do not read your
manuscript properly, and instead go on at length
about their hobbyhorse whereas, in fact, you have
dealt with their concerns elsewhere in the manuscript. Try to resist the temptation of rubbing their
nose in it with lofty sarcastic phrases such as ‘‘if the
referee had bothered to read our manuscript.’’
Instead, say something like ‘‘we agree that this is an
important point and we have already addressed it on
page A, paragraph B, line C.’’
Sometimes the referee is just plain wrong about
something. If so, it is silly to agree with the referee,
and you are entitled to a good argument. If you are
confident that you are right, then simply argue back
with facts that can be referencedethe editor can then
adjudicate who has the best evidence on their side.
CONCLUSION
Referees are human beings. The secret of a successful resubmission is to make your referees feel
valued without compromising your own standards.
Make your referees’ and editor’s life easy by presenting them with a clear numbered and structured
response letter. Provided you have made a good
attempt at answering all of the referees’ comments in
a reasonable way by following the 3 golden rules,
many referees and editors are too weak at the stage of
resubmission to open another round of arguments
and resubmission. In my experience, I spend up to 90
minutes on the initial refereeing of a manuscript, but
only around 20 minutes on a resubmission. However,
130
YMJD1792_proof 9 June 2004 3:19 pm
Return to Checklist for Revision
Williams 83
J AM ACAD DERMATOL
VOLUME 51, NUMBER 1
2. Altman DG. Practical statistics for medical research. London:
Chapman and Hall 1991.
3. Cummings P, Rivara FP. Responding to reviewers’ comments on
submitted articles. Arch Pediatr Adolesc Med 2002;156:105-7.
4. DeBehnke DJ, Kline JA, Shih RD, Research Committee of the
Society for Academic Emergency Medicine. Research fundamentals: choosing an appropriate journal, manuscript preparation, and interactions with editors. Acad Emerg Med 2001;8:
844-50.
5. Byrne DW. Publishing your medical research paper. Baltimore:
Williams and Wilkins; 1998.
6. Huth EJ. Writing and publishing in medicine. 3rd ed. Baltimore:
Williams and Wilkins; 1999.
7. Rothman KJ. Writing for epidemiology. Epidemiology 1998;9:
333-7.
8. Jefferson T, Wager E, Davidoff F. Measuring the quality of
editorial peer review. JAMA 2002;287:2786-90.
9. Jefferson T, Alderson P, Wager E, Davidoff F. Effects of editorial
peer review: a systematic review. JAMA 2002;287:2784-6.
if you miss some comments completely or your
manuscript changes do not correspond with what
you say you have done in your cover letter, this will
entice your referee to spend hours going through
your manuscript with a fine-tooth comb. If he/she
finds lots of little errors, this leads to a possible
deserved rejection.
Like a good marriage, resubmitting your manuscript in light of your referees’ comments is a process
of give and take.
The author wishes to thank Dr Jeffrey Bernhard for his
constructive comments and for references 5 to 7.
REFERENCES
1. Lowe D. Planning for medical research: a practical guide to
research methods. Cheshire (UK): Astraglobe Ltd; 1993.
131
YMJD1792_proof 9 June 2004 3:19 pm