Immediate Semantic Integration vs. Frequency in the Interpretation of German Main Clause Sentences

Falk Huettig¹ and Martin Corley^1,2
falk@holyrood.ed.ac.uk, Martin.Corley@ed.ac.uk
1 Department of Psychology, University of Edinburgh
2 Human Communication Research Centre, University of Edinburgh

We report an eye tracking study investigating an ambiguity in German that has so far not been studied: whether verbs such as haben (to have) are initially interpreted as main verbs or auxiliary markers. In German a sentence such as

(1)

Mein Sohn hat einen alten VW Käfer ...
(My son has an old VW beetle ...)

is potentially ambiguous. On the main verb (MV) reading the second NP serves as the complement of haben and the sentence could end at this point. But on the auxiliary (AUX) reading, haben is an auxiliary, and the sentence would need to continue with a participle:

(2)

Mein Sohn hat einen alten VW Käfer angefahren.
(My son has an old VW beetle run into.)
(My son has run into an old VW beetle.)

Corpus analyses showed that the AUX reading is much more frequent than the MV reading (80% vs. 20%). The question that arises for the present study on syntactic ambiguity resolution is whether German native speakers ever pursue the MV analysis in MV/AUX ambiguities given the strong frequency bias towards the AUX reading. Strong semantic plausibility manipulations were used as a diagnostic tool: if the MV analysis is frequently preferred a plausibility effect should be obtained.

24 sets of items were identified in a norming study that are highly plausible or implausible on the MV reading (F = 77.8, p = .000) but equally plausible on the AUX reading.

(3)

Der Mann hat ein Motorrad mit Seitenwagen verfolgt, ...
(The man has a motorbike with sidecar followed ...)
(The man has followed a motorbike with sidecar ...)
vs.
Der Hund hat ein Motorrad mit Seitenwagen verfolgt, ...
(The dog ...)

The eyetracking experiment showed little evidence that participants ever pursued the MV analysis (either because this analysis was never considered syntactically, or because it was never semantically interpreted), although a trend in first-pass regressions was in the predicted direction. This lack of evidence in the face of a strong plausibility anipulation leads us to tentatively conclude that an early semantic interpretation of "haben" past tense sentences is rarely, if ever, made by native speakers of German, with implications for strong semantic incrementality.