The status of the N1 in compounds In my talk I want to speak to the

The status of the N1 in compounds
In my talk I want to speak to the issue of the accessibility of the first constituent (N1) of
German nominal compounds. Most theories of composition differentiate at least between two
kinds of compounds: root and synthetic compounds. Even if they are assumed to be based
on different structures, all N1s are supposed to be equally inaccessible for anaphora (Postal
1969, Sproat 1985, i.a.). Introspective data tell us that while bound anaphora is infelicitous,
discourse anaphora is possible and exhibits degrees of accessibility, as exemplified by (1).
(1)
a.
Peter hat die Hühnerfütterung völlig vergessen. Deshalb flattern sie noch immer
wie wild im Hof umher.
Peter totally forgot the chicken-PL-feeding-SG. Hence they-PL are still fluttering
wildly in the yard.
Marie hat die Stalltür nicht verschlossen. Nun ist er leer und die Hühner flattern
wild im Hof umher.
Mary did not close the stable-MASC-door-FEM. Now it-MASC is empty and the
chicken are fluttering wildly in the yard.
b.
The N1 of the synthetic compound in (1-a) has a higher degree of accessibility than the
one in the root compound in (1-b). Setting semantic issues aside for the time being, this
difference in accessibility may be traced back to differences in the structural representation
of the compounds. Data from eye tracking during reading as a measure of online processing
will be brought to bear on this hypothesis.
According to Distributed Morphology (DM), every N1 is based on an uncategorized root
(e.g. Alexiadou 2001). The structure of root compounds thus can be described as in (2-a).
Two (uncategorized) roots merge and afterwards merge with a categorizing head to form a
nominal. Considering synthetic compounds to be structured as in Sproat (1985), extended
by assumptions of DM, a (simplified) structure like in (2-b) can be assumed. In this case
the categoryless root GRÜN gets categorized as a verb through the verbalizing affix be-.
The newly formed verbal head yields a theta-role that needs to be assigned. To enable
theta-marking, the demands of the argument structure enforce the categorization of the root
DACH and thus force it to be nominalized (cf. Longobardi 1994). In the next step, it merges
with the verbal head and forms a verbal compound, which itself is the basis for the final
step of nominalization. As a result, the N1 DACH in the case of the synthetic compound
is categorized as a noun before the construction as a whole gets nominalized. In the case of
root compounds the N1 lacks categorial status. We take this difference to predict differences
in anaphoric potential.
(2)
a.
b.
n(P)
√
DACH
(roof)
√
n(P)
n(P)<Ag,Th*,Ev+ >
naf <R+ >
v(P)<Ag,Th*,Ev+ >
GART EN
(garden)
n(P)<R*>
√
DACH
(roof)
v<Ag,Th*,Ev>
vaf
be
(CAUSE)
ung
(ing)
p
GRÜ N
(green)
As a first step towards an experimental investigation of this prediction, we conducted an
eye tracking during reading experiment. The stimuli had three conditions with compounds
including one N1 per item. Condition (a) includes the synthetic compound, (b) the root
1
compound and (c) only the monolexical stem of the N1. An example is given in (3). If our
prediction is on the right track, accessibility should be higher in condition (a) than in condition (b), in comparison to condition (c), respectively. This should result in shorter reading
times for the anaphora region and less regressions to earlier parts of the text. Condition (c)
serves as a baseline in showing the behaviour for regular discourse anaphora. The processing
time for this condition is supposed to be shortest and to have the fewest regressions as there
should be no difficulties in resolution.
(3)
Karl ist ein richtiger Heimwerker geworden.
Seit die Kinder ausgezogen sind,
ist er nur noch am Bauen und Buddeln.
cond. a: Die Dachbegrünung ist wirklich gut gelungen. Es kann
cond. b: Der Dachgarten ist wirklich gut gelungen. Es kann
cond. c: Das Dach ist wirklich gut gestaltet. Es kann
jetzt bei gutem Wetter wunderbar als Rückzugsort genutzt werden.
Als nächstes will sich Karl den Keller vornehmen.
Karl got to be a real home improver.
Since the children moved out
he is always building and digging.
cond. a: The roofi greening worked out pretty well. Iti can
cond. b: The roofi garden worked out pretty well. Iti can
cond. c: The roofi got to be really pretty. Iti can
now perfectly be used as a retreat when the weather is nice.
Up next Karl wants to take the cellar in hand.
The experiment consisted of 18 stimuli and 32 filler items. Semantic factors were controlled
for by keeping the overall content in an item as similar as possible. Thirty participants, all
of them students of our university, were tested separately on an Eye Link 1000 (20 female,
22 right eye dominant, mean age 25.6 (sd 3.6)). Twenty-seven of them went into the final
analysis.
Our analysis revealed that all relevant measures show the
predicted pattern. Total reading times for the anaphora
area (right boxes in (3)) are plotted in Fig. 1. An
omnibus-by-subject ANOVA on raw reading times with
planned contrasts (a)/(c) vs. (b)/(c) showed that these
differences were statistically reliable (p<.001). Surprisingly, even early measures like first pass fixation durations showed this pattern in a statistically reliable way
(p<.001). Also the regression count into the area of the
antecedent (left boxes in (3)) is significant in accordance
with the prediction (p<.05).
We take these results as a first step towards an empirical
foundation for assessing the morphosyntactic status of
N1s in German compounds. Although more (and more
Condition
diverse) evidence is needed, we are confident that further experimental evidence will corroborate the pattern Figure 1: Grand mean of the total
reported here. Accordingly, the morphosyntactic basis of reading time in the anaphora area
a = synthetic compound, b = root comcompounding should be able to reflect these differences. pound, c = monolexical stem
400
●
●
350
TRT
●
●
300
●
●
250
a
b
c
References: Alexiadou, Artemis (2001): Functional Structure in Nominals. Nominalization and ergativity (LA 42). Amsterdam/Philadelphia: Benjamins. • Longobardi, Giuseppe (1994): Reference and Proper Names. A Theory of N-Movement in Syntax
and Logical Form. Linguistic Inquiry 25/4, 609-665. • Postal, Paul M. (1969): Anaphoric Islands. In: Binnick, R. et al. (eds.):
Papers from the Fifth Regional Meeting of the Chicago Linguistic Society. Chicago, April 18-19, 1969. Chicago: Univeristy of
Chicago, 205-239. • Sproat, Richard William (1985): On deriving the lexicon. Phil. Diss. MIT, Cambridge, MA
2