On the Diachronic Status of Infinitival Markers: German vs

On the Diachronic Status of Infinitival Markers: German vs. Norwegian
In older stages of German, object control infinitives could select bare infinitives, cf. [1] from OHG
(750-1050), [2] from MHG (1050-1350), and [3] from ENHG for the directive predicate bitten 'ask':
[1]
thoh
bát
er nan (...) thie stéina duan
zi bróte
though ask.3SG.PST he him.ACC the stones do.INF to breads
'Though he asked him to turn stones into bread.' (Otf II, 4: 44)
[2]
diu mich (…) der edeln künste
swaere
that me.ACC
the noble arts.GEN importance.NOM
(iii den rîchen herren
künden
bat
the rich
lords.DAT announce.INF ask.3SG.PST
‘[the story] which importance of noble arts urged me to announce to the rich lords’
(Konrad v. Würzburg: Die Klage der Kunst, 32,8; c. 1250-1287;)
[3]
bat
sye
morgens
herwider gon
ask.3SG.PST her.ACC in.the.morning again
go.INF
'[he] asked her to go [there] in the morning again'
(PFf, p. 194)
According to the well-known view, the absence of the infinitive marker in the examples [1]-[3] is accounted for by the fact that zu as an infinitive marker grammaticalized in German very late, i.e. in the
19th cent. (cf. Abraham 2004). A different scenario has been observed in the history of Norwegian.
Faarlund (2004, 2007) observes that Old Norse (12th - 14th cent.) infinitives can be divided into two
groups: (a) control infinitives representing the CP-size and being introduced by the infinitival complementizer at, and (b) raising infinitives being non-CPs and occurring without the infinitival marker.
[4] and [5] illustrate (a) and (b) respectively:
[4]
þeir ætluðu
at hengja hann
they intended to hang
him
‚They intended to hang him.’
(Faarlund 2007: 59; ex. 1a)
[5]
þótti
honum
hon
vel
hafa gert
seemed him.DAT she.NOM well have done
‚She seemed to him to have done well.’
(Faarlund 2007: 61; ex. 4a)
What is crucial about the infinitival marker at is that it is supposed to occur in all canonical Old Norse
control infinitives, meaning: (i) at functions as functional head already in Old Norse, (ii) there is no
empirical evidence for grammaticalization of at as infinitival marker. Faarlund (2004, 2007) assumes
at to merge in the C position of the embedded clause and accounts for its absence in raising structures
as a consequence of their non-CP-hood. If raising infinitives do not possess the CP layer, there is no
place where the infinitival marker could merge. This situation changed in Early Modern Norwegian
(15th - 19th cent.). The infinitive marker åt does not occupy a C position anymore. Instead it adjoins
to the embedded verb as a proclitic indicating that it can occur in raising infinitives. [6] is a case in
point:
Jtem kiændes
oc
forde
Anund oc
gudri
thus know.PASS also above-mentioned Anund and Gudri
(iii) att hafue opboret xv kiørlagh
to have received 15 cow-values
‘Thus the above-mentioned Anund and Gudri are known to have received the value of 15 cows.’
(Faarlund 2007: 69; ex. 24)
[6]
In Present-day Norwegian the internal position of å, the standard form of the infinitive marker, merges
as a C-head mirroring the structure from Old Norse. Contrasting these two scenarios, we argue that (i)
there is no evidence for grammaticalization of zu as infinitival marker in German, (ii) zu existed already in the OHG period and its possible omission in older stages of German (cf. [1]-[3]) is due to the
lack of the coherence vs. incoherence opposition in the sense claimed by Bech (1955/57), (iii) zu
merges within PP selecting either vPs/VPs or CPs. Several independent arguments support our analysis. First, it is well-known that the (in-)coherence opposition is a very young development in the histo-
ry of German. Demske (2008) illustrates for OHG, Askedal (1998) for MHG and Maché & Abraham
(2011) for ENHG that this opposition has not been established in the language periods under investigation, indicating that it must have become established first in the 19th/20th cent. If this is true, we
should expect two things in older stages of German: (i) violations against the (in-)coherence opposition, (ii) bare infinitive complements embedded under matrix predicates licensing only zu-infinitives
in Modern German. The subject-to-subject raising verb pflegen 'use to', for instance, requires zu 'to'
and does not allow extraposition of the embedded complement in Modern German:
[7]
[7']
dass Fritz seine Krankheiten *(zu)
that Fritz his
diseases
to
'that Fritz keeps ignoring his diseases'
(Sternefeld 2008: 417; ex 32c)
ignorieren pflegt
ignore.INF use.3SG
*dass Fritz pflegt seine Krankheiten zu ignorieren
(Sternefeld 2008: 417; ex 32c)
However, in the 19th cent. pflegen still embeds bare infinitives and allows extraposition of the embedded infinitive clause:
[8]
Die Leitung
solcher Arbeiten
haben
französische
the leadership such
works.GEN have.3PL French
[8] Genieoffiziere, wodurch dieselben weit rascher gefördert
genius.officers by.what the.same far
quicker sponsor.PTCP
[8] werden,
als
sonst dergleichen hier geschehen pflegt
PASS.AUX.INF than usual of.that.kind here happen.INF PF.3SG
'French genius officers are leading such works, whereby they are promoted quicker than it usually
happens.' (KHZ, Mainzer Journal, 13/10/1849)
[9]
wo
sie pflegt
zur
Kirche zu
where she use.INF to.the church to
'where she keeps going to the church'
(J. & W. Grimm (1891): Der Brennberger)
gehen
go.INF
Accordingly, zu, as argued by Haider (1986, 2009), cannot absorb a theta-role. Instead, its presence
only signals that the (in-)coherence opposition has been already established for the matrix predicate in
question. This, in turn, is related to the second argument. If zu would absorb a theta-role, helfen 'help'
would not be supposed to exhibit variation in complementation. Notice, however, that both bare and
zu-infinitives are grammatical:
[10]
Saskia hilft
Oleg, das Holz (zu) spalten
Saskia help.3SG Oleg the wood to maul.INF
'Saskia helps Oleg to maul the wood.'
However, if helfen is accompanied by the correlate dabei, the presence of zu is a must-have:
[10'] Saskia hilft Oleg dabei, das Holz *(zu) spalten
Now, if after helfen a bare infinitive occurs, extraposition is not possible. We are forced to form a verbal cluster:
[11] *weil Saskia Oleg hilft, das Holz spalten
[11'] OKweil Sakia Oleg das Holz spalten hilft
On the other hand, extraposition is possible as soon as zu occurs:
[12]
[12']
OK
OK
weil Sakia Oleg das Holz zu spalten hilft
weil Sakia Oleg hilft, das Holz zu spalten
Finally, we will show that similar facts follow from West-Germanic dialects (e. g. from Zurich German, see Cooper 1990), in which zu may be dropped under certain conditions. This is mainly due to
the fact that these dialects do not exhibit the (in-)coherence opposition.