SC- doubly embedded sentences

Lars Konieczny and Tobias Bormann
Centre for Cognitive Science
Institute for Computer Science and Social Research
University of Freiburg

It is a well-known observation that doubly center-embedded sentences, although grammatically well-formed, provoke processing breakdowns, problably due to memory overload. However, not all embeddings behave alike: Gibson and Thomas (1997) found that relative clause within sentence complement (SC/RC) constructions are easier to process than their RC/SC counterparts. This result is predicted by Gibson's (1998) Syntactic Prediction Locality Theory (SPLT). Unfortunately, the ordering factor was confounded with the quality of embedded extraposition variants of their materials.

We conducted an off-line Magnitude Estimation study (Bard, Robertson and Sorace, 1996), using materials similar to Gibson and Thomas', but also varying the position of the inner-most clause (adjacent vs. extraposed).

(1): SC/RC (adjacent) Die Moeglichkeit, dass der Praktikant, den die Krankenschwester unterstuetzte, schlief, nervte den Verwalter ziemlich. (The possibility, that the intern, which the nurse supervised, slept, annoyed the administrator a little.)
(2): SC/RC (extraposed) Die Moeglichkeit, dass der Praktikant schlief, den die Krankenschwester unterstuetzte, nervte den Verwalter ziemlich. (The possibility, that the intern slept, which the nurse supervised, annoyed the administrator a little.)
(3): RC/SC (adjacent) Der Verwalter, den die Moeglichkeit, dass der Praktikant schlief, nervte, unterstuetzte die Krankenschwester ziemlich. (The administrator, who the possibility, that the intern slept, annoyed, supervised the nurse a little.)
(4): RC/SC (extraposed) Der Verwalter, den die Moeglichkeit nervte, dass der Praktikant schlief, unterstuetzte die Krankenschwester ziemlich. (The administrator, who the possibility annoyed, that the intern slept, supervised the nurse a little.)

We were able to replicate Gibson and Thomas' preference for SC/RC embeddings for adjacent variants. While extraposition did not induce a main effect, (p > 0.30 for both analyses), we found significant interaction (for item analysis) between the two factors in that the SC/RC sentences were judged worse when the most embedded clause was extraposed (analysis by items: F(1, 31)=8.68, p<0.007; analysis by subjects: F(1, 23)=3.25, p<0.085). According to theories of syntactic complexity, extraposition should make sentences easier to process. However, Konieczny (2000) reported overall preferences for adjacent vs. extraposed relative clauses quite contrary to SPLT's predictions which led him to hypothesize that locality might apply only to non-predictible items. Since SCs and RCs differ with respect to their predictability (SCs are ontologically necessary arguments, RCs are optional modifiers), the results can be taken to support this claim. We are currently conducting a series of self-paced reading studies to further investigate processing of variants of extraposed clauses.

References

Bard, E. G., Robertson, D., & Sorace, A. (1996). Magnitude estimation of linguistic acceptability. Language, 72(1), 32-68.

Gibson, E. (1998). Linguistic complexity: locality of syntactic dependencies. Cognition, 68,1, 1-76.

Gibson, E., & Thomas, J. (1997). The complexity of nested structures in English: Evidence for the syntactic prediction locality theory of linguistic complexity. MIT manusscript.

Konieczny, L. (2000). Locality and parsing complexity. Journal of Psycholinguistic Research, 29-6.

AMLaP Conference, Saarbrücken, September 2001