On the domain specificity of the human language faculty and the effects of principles of efficient computation phylogeny and ontogeny Anna Maria Di Sciullo Université du Québec à Montréal School of advances studies on Domain specificity and the faculty of language Pontifical Catholic University of Rio de Janeiro and the Federal University of Rio de Janeiro 16th-24th of March 2015 This work is supported in part by funding from the Social Sciences and Humanities Research Council of Canada to the Major Collaborative Research on Interface Asymmetries, grant number 214-2003-1003, and by funding from the FQRSC on Dynamic Interfaces, grant number 137253. www.interfaceasymmetry.uqam.ca. www.biolinguistics.uqam Focus A striking fact in the development of the nominal domain in Indo-European languages is that while pre and post nominal instances of certain morphosyntactic elements are possible in earlier stages of the languages, only one position is generally available in more recent stages. This phenomenon is neither language specific nor category specific, as it can be observed in the development of possessive adjectives from Ancient to Modern Greek and from Latin to the Romance languages (Di Sciullo 2011), in the development of prepositions in the Indo-European languages (Di Sciullo and Nicolis 2012), in the development of the definite determiner in Romanian (Di Sciullo and Somesfalean 2013). This phenomenon also gives rise to novel activity for syntactic categories: heads, clitics, particles, prefixes and suffixes. Sample: Old English Old Germanic is predominantly prepositional and Old English PPs are generally head initial, while personal pronouns display fluctuation, as they are attested both following and preceding P (Alcorn 2011): (1) a. his gebro.ru to him comon when his brethren to him came ‘When his brethren came to him’ (cocathom1,ÆCHom_I,_21:346.24.412) b. ...o..æt se halga gast him to com until the holy spirit him to came ‘... until the holy spirit came to him’ (cocathom1,ÆCHom_I,_21:346.24.4121) c. God cwæd to him God said to him ‘ ‘God said to him’ (cocathom1,ÆCHom_I,_1:181.75.72) d. Pæt wif him cwæd a to the woman him said then to ‘the woman then said to him’ (coaelhom,ÆHom_5:21.690) Sample: Old Italian (13th c) The analysis of Boccaccio's Decameron and of a corpus of 13th century Florentine indicates the following distribution for P elements. P uniformly precedes its complement except in the case of the preposition con, where monosyllabic personal pronouns are cliticized onto the preposition (meco, teco, seco), (2). Instances where con precedes a monosyllabic personal pronoun are also attested, (3). (2) A b. c. d. (3) a. b. …. e per li compagnoni che teco fuggiro , per li dei... (Brunetto, Rettorica) E altre donne, che si fuoro accorte di me per quella che meco piangia , fecer lei partir via... (Dante, Vita Nuova) E sua bieltate è di tanta vertute, che nulla invidia a l' altre ne procede , anzi le face andar seco vestute di gentilezza , d' amore e di fede. (Dante, Vita Nuova) E poco stante venne contra lei un grandissimo cavaliere molto sformato e terribile a vedere , tutto armato d' arme nere , in su 'n un grandissimo destriere ; e avea seco tanta gente , che tutto 'l campo copriano (Bono, Libro Vizi) neiente de lo mondo ; con te le tue , parole voria conte avere... (Rinuccino, Sonetti) Ballata , i' voi che tu ritrovi Amore , e con lui vade a madonna davante , sì che la scusa mia , la qual tu cante , ragioni poi con lei lo mio segnore. (Dante, Vita Nuova) Old Italian (13th-14th c) There are two subclasses for the meco, teco, seco cases. They can in fact either appear on their own, where con appears to function as a postposition, or as post-P complements of the preposition con (e.g. con meco, con teco, con seco). (4) a. E perciò ch' io so bene ch' assai val meglio che tu parli con teco, che né io né altri , sì fo io fine alla mia diceria. (Brunetto, ProLigario) b. ma prendi usanz' a tale che sia con teco iguale ; e s' avanzasse un poco... (Brunetto, Tesoretto) c. Gli altri tenea in pregione , e costui di fuori , con seco , e vestialo nobilemente (Anonimo, Novellino) In this diachronic phase (13th 14th c) meco, teco, seco appear more often without con than with con. It should be the case that later text should display a decreasing number of 'bare' meco, teco, seco and a corresponding increase in con meco, con teco, con seco, as well as the eventual gradual disapperance of -co. Leading to Modern Italian exclusive pronominal use of con, e.g. con me, con te, con se. P-Shell We take the structure of prepositions (P) to be that of the functional shell, as defined in Di Sciullo (1999, 2005). The P-shell consists of two layers of asymmetric projections. The upper and the lower heads are the locus of formal features including valued and unvalued features, including Case features. The DP complement is a sister of F, and can be displaced for feature valuing. The upper layer is the domain of the relational predicate P and the lower layer is the domain of the restrictor of P. Two questions • Why is there an oscillation in the position of the complement of P in earlier stages of the languages under consideration (English, Italian) ? • Why is there no trace of syntactic oscillation in Modern variants ? Biolinguistic hypotheses Evolutionary developmental universals emerge in language historical development and they are formulated in Di Sciullo (2011) in terms of the Directional Asymmetry Principle, according to which language development is symmetry breaking: fluctuating asymmetry precedes directional asymmetry. The Directional Asymmetry principle predicts the occurrence of oscillation between linguistic elements and the gradual elimination of the oscillation in language development. While the computational procedure of the narrow language faculty is reduced to the minimum required by conceptual necessity, complexity may arise from experience (language acquisition, language contact, etc.) giving rise to choice points (symmetry) in functional feature structure, with the consequences of enlarging the set of possible derivations. Principles of parsimony, falling into the third factor, will eliminate the complexity by breaking the symmetry brought about by experience. Principles of parsimony are deep rooted in the human cognitive system and act as soon as possible in language acquisition. Their effects are however gradual in language diachronic development, as environmental dynamics is not deterministically driven by genetic determinism. Outline 1. 2. 3. 4. Language Faculty, language variation and universal constraints Predictions of the Directional Asymmetry Principle Interaction of the computational procedure of the Language Faculty with principles reducing complexity Does ontogeny recapitulates phylogeny? Part 1: Language Faculty, language variation and universal constraints The Language Faculty, defined as a computational procedure, is stable and thus it does not vary through time and space, while language variation requires experience (Chomsky 1995, 2005, 2011). We also assume that linguistic variation follows from the setting of parameters, which are values associated to functional features in the lexicon, as it is the case in the generative grammar tradition stating with Chomsky (1981) and as it has evolved in recent Minimalist Models where feature valuing is the core computational procedure driving variation in linear order externalization. The computational procedure of the narrow language faculty (FLN) do not change in time, the active features of the functional elements change as an effect of language acquisition, languages in contact and individual innovations. We develop the hypothesis that symmetry breaking is part of the principles of efficient computation which apply to the computational procedure of FL and fall into the class of symmetry principles, (symmetry, symmetry breaking, asymmetry). . Language Faculty The Asymmetry of Merge SELECT is asymmetrical, as it is based on the proper subset relation between sets of features. We propose that it is also asymmetrical with respect to functional feature structures. (5) Merge is an operation that applies to a pair of elements in the Numeration whose sets of features are in a proper inclusion relation. (Di Sciullo and Isac 2008: 270) (7 Merge is an operation that applies to a pair of elements in the Numeration whose features are disjunctively valued or unvalued. GB and macro parameters Principles and Parameters Model Parameter : option left open in the principles of UG Macro parameters -Head directionality parameter -Pro-drop parameter -Bounding nodes -Polysynthesis parameter Open questions -Are there restrictions on possible parameters? (limited set?) -Are all parameters binary? How many are there? -Which parameter setting algorithm is correct? -Are there any biological correlates? Phylogenetics, lexicon, syntax Based on lexical properties (Gray and Atkinson 2003; Pagel, Atkinson, and Meade 2007) Language taxonomy based on word cognates. Based on abstract syntactic properties (Longobardi and Guardiano 2009) Parametric syntax serves phylogenetic purposes better than methods based on lexical items. Languages and dialects are more closely related than what it might appear from the classifications based on lexical items. Syntactic data is similar to genetic markers. Parametric comparison method displays the advantages of population genetics: compares values (alleles) of parameters (syntactic polymorphisms) in different grammars, i.e., draws from a finite universal list of discrete biological options. Phylogenetics, stable, developmental Based on stable states Languages with dominant VSO order are always prepositional. With overwhelmingly greater than chance frequency, languages with normal SOV order are postpositional. (Greenberg’s Universals) Based on development Language variation is symmetry breaking (Di Sciullo 2011, 2012) Di Sciullo & Nicolis 2013; Di Sciullo & Somesfalean 2013. (A)symmetry based syntax serves developmental phylogenetic purposes better than stable state methods. Symmetry-breaking comparison method displays the advantages of phase transition dynamics. Languages and dialects are more closely related than what it might appear from the classifications based on stable states. Evo-devo universals Chomsky (2011) refers to factors reducing complexity in terms of natural laws attested in biology and physics. We focus here on developmental constraints affecting the shape of bi-partite organisms and suggest that they play a role in the language diachronic variation. Several works in evolutionary developmental biology provide evidence that variation and change in biology is symmetry breaking (Lewontin 1970, 1974, Graham, Freeman and Emlen 1993; Palmer 2004, a.o.). Symmetry breaking laws, such as ‘selecting for asymmetry’, are part of the natural laws reducing complexity in the natural world. Symmetry breaking may find its origin in the natural laws affecting the evolution of the shape of biological organisms. Seen as a law external to the Language Faculty, it provides a biolinguistic explanation for language variation and development. Evolutionary developmental biology Ø In evolutionary developmental biology, symmetry-breaking is a core aspect of evolution and change in different species (Graham, Freeman and Emlen 1999; Palmer 1996, 2004; Palmer and Lowentin 2004). Phylogenetic patterns of variance in the evolution of bilateral asymmetric species : Symmetry > fluctuating asymmetry > directional asymmetry (Palmer) Development & evolution of claw asymmetry in fiddler crabs (Uca) Phenotype-leads mode of evolution (genetic assimilation) Palmer Science 2004, Current Biology 2009) Symmetry breaking and language development (8) Directional Asymdevelopment is symmetry breaking. Fluctuating asymmetry metry Principle (Di Sciullo 2011) Language precedes directional asymmetry. The P Shell and Feature valuing The constraints in (9) and (10) are specific cases of the Directional Asymmetry Principle in (8), which provides an explanation for the fact that the order of a complement with respect to its head may oscillate at some stage of the historical development of languages, while in a subsequent stage only one of the two options is available. Clearly, this does not imply that there is directionality in language diachronic development. Fluctuating asymmetry refers exlusively to the possibility of left or right positioning of a head in its functional domain, as in (15a,b), where in (15a) the DP complement follows the P head and in (15b) it preceeds it. The linearization of the shell structure at the sensorimotor interface will result in situations where a P head either preceeds its complement, (15a), or follows it, (15b). . Summary of Part 1 Computational procedure of the LFN + experience + Factors reducing complexity The computational procedure of the Language Faculty does not change, whereas language variation requires experience. The oscillation in the position of the complement within respect to its (P) is dependent of grammar internal properties, the gradual elimination of the oscillation can be see as a consequence of factors external to the language faculty. Symmetry breaking as part of the natural laws affecting the evolution of the shape of biological organisms can be seen as a dynamic force external to the Language Faculty, that may provide a biolinguistic explanation for language variation and evolution with respect to the position of a P head and its complement. • Advances and open questions in our understanding of language variation -Models of language variation -Phylogeny and syntax -Developmental constraints: Directional Asymmetry Principle -Valued/Unvalued feature constraint -Head Initial/Final constraint Feature/precedence symmetry tends to be broken diachronically Part 2. Predictions of the Directional Asymmetry Principle Two predictions: A: Stable state / Directional asymmetry should be synchronically widespread. B: Oscillation / Fluctuating asymmetry should characterize older diachronic stages. Prediction A for P Stable state / Directional asymmetry should be synchronically widespread The predictions A and B are validated on a number of languages closely related to Proto-Indo-European Prediction B for P: Fluctuation should characterize older diachronic stages Old Hittite > Middle Hittite > Late Hittite Old Hittite. H and B consider Old Hittite as a strongly (though not exclusively) postpositional language. (17) [LUG]AL URUKussara URU-az katta [p]angaritu[ēt… king Kussara city-ABL down mass-INSTR he-came “ The King of Kussaras come down from the city with great forces” Cases of fluctuating asymmetry with the same adpositional element are attested in Old Hittite. For example, parā (‘towards’) and andan (inside) are attested both prepositionally and post-positionally: (18) a. [UGULA LÙ].MEŠ MUHALDIM marnuandas ispantuziassar LUGAL-I chief cook-PL m.-GEN vessel-N/A king-D/L parā epz[i] “ The chief of the cooks holds out to the king a vessel full of m.” b. [MUŠENhāranann= a p]arā hilam[na] pētumeni eagle-ACC and toward door-DIR we-carry “ and we carry the eagle towards the door” (19) a. kuit kuit E-ri andan harakzi tat sarnikzi what what house-LOC inside is-lost CONN=it replace-3S "he replaces whatever is lost inside the house” b. andan= a E-ri kuit kuit harakzi inside= PTChouse-LOC what what is-lost “But whatever is lost inside the house” Both Middle and Late Hittite are consistently post-postitional. For example, while andan can be used both pre- or postpositionally in Old Hittite (20), it becomes exclusively post-positional in New Hittite (H and B p.96). (20) ÌD-i anda lāhuwai River-DAT/LOC into pour-3SG “she pours into the river” Old Armenian > Late Armenian We exemplify fluctuating asymmetry with the adposition handerj (‘(together) with’). This adposition is attested productively both prepositionally and postpositionally with a 24/13 preponderance of postpositional usage in the Old Armenian Gospels in Künzle (1984:394-395) (quoted in H&B p.168) (21) a. ašakertawkc-n handerj Disciples-INSTR/PL-CL with “with the disciples” b. handerj ašakertawkc-n The language became fully postpositional with ‘few relics of earlier prepositions (most notably aranc=N+DAT “without”, cf. Classical arancc=N+GEN “without”; tcebi=N+ACC “against”, cf. Classical hakar-ak=N+GEN). Notice that = indicates that these relics are actually verbal prefixes, a rather unsurprising fact, on the basis of the discussion of Old Hittite above and of Homeric Greek below Homeric Greek Homeric Greek (1000-800 BC) presents a stage of fluctuating asymmetry in the historical development of Greek. Elements like ἔπι (‘upon’) are used in a variety of contexts (preverb, adposition, adverb) (Bortone 2010, Devine & Stephens 2000, Chantraine 1953). There are both prepositional and adpositional uses; as is typical in the early stages of ancient languages, these early adpositions can govern several different cases, while a progression towards a one-to-one relation is observed diachronically. (22) Homeric Greek Homeric Greek (1000-800 BC) presents a stage of fluctuating asymmetry in the historical development of Greek. Elements like ἔπι (‘upon’) are used in a variety of contexts (preverb, adposition, adverb) (Bortone 2010, Devine & Stephens 2000, Chantraine 1953). There are both prepositional and adpositional uses; as is typical in the early stages of ancient languages, these early adpositions can govern several different cases, while a progression towards a one-to-one relation is observed diachronically. (22) Classical Greek Classical Greek is virtually fully prepositional (directional asymmetry). Luraghi (2003) provides the following important example from Herodotus which shows that ‘incorporated prepositions’ are not syntactically active any longer since they co-occur with omophonous prepositions. The prefix has become a satellite, in Talmy’s terms and case assignment has to be independently carried out by a bona-fide preposition. Latin-Umbrian-Old Italian-Modern Italian Latin is almost fully prepositional, with the exception of enclitic prepositions on personal pronouns (and some indefinites): e.g. mecum, tecum, secum, etc. Umbrian. Fortson (2010) shows that in Umbrian some structures exist where the preposition appears to be cliticized on the noun or on an adjective, as the following cases illustrate, while the language is typically prepositional: (24) nertruco persi(6) tertiama spanti left.at foot third.at dish ‘at (his) left foot’ VIb 37, 39 ‘at the third dish’ IC 2 (25) erer nomneper one’s name.through ‘by/through this one’s name’ (Fortson 2010, p. 123) Fortson reports five adpositions which ‘typically or always’ occur as postpositional clitics: -ař, -a, ‘up to’, -kum -ku “at, with”, -en -em -e “in(to)”, -per -pe “on behalf of”, -ta -tu “from”. Bortone (2010, p.138) suggests that concerning adpositions, ‘In most Classical Indo-European langauges, a choice was gradually made as to whether they should precede or follow the noun phrase.’ Closely related languages like Latin and Umbrian display different choices for the same element. For example while pro is prepositional in Latin, it is a postpositional affix in Umbrian: (26) a. pro populo ‘on behalf of the people’ Latin b. poplu-per Umbrian Old Italian shows that the preposition enclitic on the personal pronoun ceases to be syntactically active, in that it can be accompanied by a separate preposition preceding the DP. (8) shows a clear intermediate stage where the ‘real’ apposition has become the preposition preceding the pronoun. (27) ch’Amor non venga sempre ragionando con meco, et io collui. (FP) (14th Century) that love not comes always talk with me=with and I with him. ‘that love doesn’t always come talk with me, and I with him’ Modern Italian is strongly prepositional. Old Germanic - Old English - Modern English Old Germanic was predominantly prepositional. Old English PPs are generally head initial, while personal pronouns display fluctuation, as they are attested both following and preceding P. (Alcorn 2011) (28) Þa his gebroþru to him comon when his brethren to him came When his brethren came to him’ (cocathom1,ÆCHom_I,_21:346.24.412) (10) God cwæð to him God said to him ‘ ‘God said to him’ (cocathom1,ÆCHom_I,_1:181.75.72) (29) ...oððæt se halga gast him to com until the holy spirit him to came ‘... until the holy spirit came to him’ (cocathom1,ÆCHom_I,_21:346.24.4121) (12) Þæt wif him cwæð þa to the woman him said then to ‘the woman then said to him’ (coaelhom,ÆHom_5:21.690) In Early Modern English (1600), postpositions are often attested. (30) a. God cwæð him þus to God said thus to him b. þæs cyninges þegnas þe him beæftan wærun The king's thegns who were behind him. c. þa gatu him to belocen hæfdon had locked the gates against them d. hiera mægas him mid wæron their relatives were with them This fluctuation is no longer productive in Modern English. 3.5. Section summary We provided cross-linguistic evidence based on English and Estonian that the oscillation of a complement with respect to its prepositional head is Case dependent. The facts indicate that cross-linguistically; the grammatical principles have precedence over principles external to the language Faculty, as argued independently in Di Sciullo and Aguero (2008) on the basis of the difference between Binding and Coreference to account for the so called ‘delay of Condition B effect’ in first language development. In a diachronic perspective, once internal factors allowing for a certain structure vanish, external forces to the language faculty will drive diachronic development. The Directional Asymmetry Principle will reduce the complexity brought about in an instable grammatical system. It follows that P – DP will become the only available option under the Directional Asymmetry Principle for Modern English. 4. Language development in ontogeny and phylogeny We raise the question whether the growth of language in the individual (ontogeny) recapitulates the historical development of languages (phylogeny). In what sense does the growth of language in the individual (ontogeny) recapitulates the historical development of languages (phylogeny)? -In what sense does the growth of language in the individual (ontogeny) recapitulates the historical development of languages (phylogeny)? -Why would there be a bias favoring prepositions instead of postpositions in the diachronic development of P in languages such as English and Italian? -Is there a trace of the syntactic fluctuation in the position of a preposition and its complement in English children? Childes The analysis of Adam, Eve and Sarah from the "Brown" files in Childes has yielded not a single clear case of DPpron - P in these three children speech, while P-DPpron (and P-DP more generally) are abundantly attested since the earliest files. Clear cases of P - DPpron are attested from the earliest files on : Eve At age 1;7 < *** File "eve03.cha": line 2985. < *CHI: oh look at me.< < From file <eve04.cha> < ---------------------------------------< *** File "eve04.cha": line 326. < *CHI: sit by you . [+ IMIT] Sarah Age 2;3.23 < *** File "sarah005.cha": line 978. < *CHI: with you . <---------------------------------------Age 2;4.12 - 2;4.17 < *** File "sarah008.cha": line 153. < *CHI: happy birthday to you . < ---------------------------------------< *** File "sarah009.cha": line 1603. < *CHI: by me . < ---------------------------------------< *** File "sarah009.cha": line 1738. < *CHI: happy (.) to you . Fluctuation in the position of a complement with respect to its P head is not observed in Childes. Ontogeny recapitulates phylogeny? Development is at the core of both biology and language viewed as a biological phenomenon. In this perspective, the principles of language growth are based on genetic endowment, in addition to the existence of learning processes (Chomsky 1968, 2005, 2011, Lenneberg 1967). The existence of specific biological support for language is beyond doubt. Arguments come from the nature of structural properties common to all languages, restrictions on the degree to which languages vary, and linguistic knowledge not attributable to the environment, as well as uniform patterns of normal and abnormal language development--plus the fact that nonhuman mammals with good statistical learning and computational capacities (Gallistel 1990) nevertheless do not develop language. The principles of language growth are not learned on the basis of quantitative data a child is exposed to in a given span of time. If these principles were genetically determined, it would be unlikely that language ontogeny would recapitulate language phylogeny. Based on our targeted analysis, we find no evidence supporting the recapitulation theory on the grounds of surface data, whereas we find reasons to relate principles of language growth to developmental universals. Principles of language growth and Developmental Universals According to the Unique Checking Constraint (UCC) (Wexler 1998), the genetic system development has the property that more than one checking is dispreferred at young ages, (49) . In the clausal domain, AGReement and Tense cannot be both checked, via the UCC, and thus one is eliminated giving rise to the Optional Infinitive stage. Similarly, in the nominal domain, postpositional structures would be more complex, and thus would be dispreferred by the genetic system. Principles of Language growth : Unique Checking Constraint (UCC) The genetic system development has the property that more than one checking is dispreferred at young ages. In the clausal domain, AGReement and Tense cannot be both checked, via the UCC, and thus one is eliminated giving rise to the Optional Infinitive stage. (Wexler 1998) In the nominal domain, checking the same feature twice, on P and on F, would be more complex, and thus would be dispreferred by the genetic system, the unvalued feature of P rather than the unvalued feature of F would be eliminated given principles of efficient computation, giving rise to prepositions, (50). (50 ) [ … P [ F [ D [uD] uD] [ … P [ F [D [uD] Principles of Language development: Featural symmetry tends to be disprefered. Section summary Development is at the core of both biology and language viewed as a biological phenomenon. In this perspective, the principles of language growth are based on genetic endowment, in addition to the existence of learning processes (Chomsky 1968, 2005, 2011, Lenneberg 1967). The principles of language growth are not learned on the basis of quantitative data a child is exposed to in a given span of time. If these principles were genetically determined, it would be unlikely that language ontogeny would recapitulate language phylogeny. Based on our targeted analysis, we find no evidence supporting the recapitulation theory, whereas we find reasons to relate principles of language growth to Evolutionary developmental universals. Conclusion We develop the view that language variation is the result of the conjunction of the operations of the Language Faculty and factors reducing complexity. It follows from this model that language phenotype is limited by the Language Faculty, the language genotype. We proposed in previous work that symmetry breaking is one of the laws reducing complexity in language development (Di Sciullo 2011). Symmetry introduces choice-points, thus instability, in a system that seeks to eliminate symmetry in order to reinstate an asymmetrical stable state. We proposed a developmental universal relying on symmetry breaking and tested its empirical predictions. As predicted by the V/UFC and the HI/FC, symmetry is available in earlier stages of language such as English and Italian and is gradually reduced. The surviving derivations are shorter than the derivation that was eliminated in languages such as English and Italian. This brings further justification to the hypothesis that symmetry breaking is part of the factors reducing complexity. We argued further that symmetry breaking reduces the choice between a valued and an unvalued variant of a functional feature associated with a functional head. As predicted by our hypothesis this choice, which was available in earlier stages of languages such as English and Italian, is gradually reduced. The surviving derivation is shorter than the derivation that was eliminated. This brings further justification to the hypothesis that symmetry breaking is part of the factors reducing complexity.
© Copyright 2024