Cardinals vs. Articles in Dutch Relative Clause Attachment

Marielle van Wilgenburg, Frank Wijnen and Hugo Quené

Utrecht University

The well-known relative clause attachment ambiguity exemplified by Cuetos & Mitchell's (1988) Alguien disparó contra la criada de la actriz que estaba en el balcón has become notorious for its capacity to elicit highly variable resolution preferences both within and across languages. Fodor has proposed in various places (e.g. Fodor 2002) that prosody, either explicit or implicit, is a primary factor in determing syntactic ambiguity resolution, particularly in this structure. Schafer et al. (1996) showed that when there is a difference in prosodic prominence between the two attachment sites,
perceivers attach the RC to the most prominent one, in an auditory task. Our study set out to replicate this effect in silent reading. We constructed 28 sentences containing RC ambiguities, contrasting (a) the syntactic number of the critical nominal heads (pl-sing vs. sing-pl), and (b) type of determiner in the plural NP (article vs. cardinal): De jongen vernielde de/twee brommer(s) naast de/twee fiets(en) die tegen het hek gezet waren. ('The boy wrecked the/two moped(s) next to two/the bike(s) that were put against the fence.') In a pretest, 6 subjects read aloud those items in which a determiner contrasted with a cardinal, without preparation or rehearsal. Perceptual analysis showed that NPs with a cardinal were accentuated significantly more often than NPs with an article. The main experiment consisted of a paper-and-pencil sentence completion task, in which subjects (n=32) had to fill in a missing auxiliary (such as waren 'were' in the above
example), so that number marking would indicate the intended RC attachment. The overall preference for NP2 attachment turned out to be stronger than has been reported before for Dutch (60.8%). NP2 preference was stronger when the NP2 was plural than when it was singular (66.9% vs. 54.7%), and this preference was stronger for cardinals than for determiners (70.9% vs. 62.9%). Given the pretest data, the latter result can be explained as an effect of implicit prosody. However, it may also be related to the different semantics of definite articles and cardinals. We are currently looking into this issue. Another question is why an effect of the cardinal-article contrast does not show up in the NP1 attachments. We think this may be due to the thematic prepositions we used to connect the two NPs. Also, prepositions were accentuated quite frequently in the pretest, and this may be indicative of packaging the Prep-NP2 substructure as a separate unit.

References

Cuetos, F. & Mitchell, D.C. (1988). Cross-linguistic differences in parsing: Restrictions on the use of the Late Closure strategy in Spanish. Cognition 30, 73-105.

Fodor, J.D. (2002). Prosodic disambiguation in silent reading. Proceedings of NELS 32.

Schafer, A., Carter, J., Clifton, C. & Frazier, L. (1996). Focus in relative clause construal. Language and Cognitive Processes 11, 135-163.