MIT CogNet, The Brain Sciences ConnectionFrom the MIT Press, Link to Online Catalog
SPARC Communities
Subscriber : Stanford University Libraries » LOG IN

space

Powered By Google 
Advanced Search

 

Avoiding ambiguities through constituent ordering

 Jennifer Arnold, Thomas Wasow and Nina Kim
  
 

Abstract:
Dative Alternation (DA) and Heavy-NP-shift (HNPS) are two English constructions that exhibit variation in the order of postverbal constituents, as in (1).

In DA both variants are common, but in HNPS the shifted order is rare and relatively marked. However, ordering in both is influenced by both the relative heaviness and givenness of the constituents (Arnold et al., 1998).

This poster reports the first experiment in a project investigating whether constituent ordering in DA and HNPS is influenced by an ambiguity avoidance strategy. Thirty-two subjects read pairs of sentences and indicated which one was more natural. Both sentences contained two equal-length constituents, and were identical except for the order of the two constituents. Using stimuli like (2), we manipulated two factors: a) construction (HNPS vs. DA) and b) whether the sentence contained a potential temporary ambiguity or not.

The results revealed three effects: a main effect of ambiguity (canonical order preferred more when it was unambiguous) (F's > 15, p<.005), a main effect of construction (canonical order preferred more for HNPS than DA) (F's > 20, p<.005), and an interaction (ambiguity avoidance had a bigger effect for DA than HNPS) (F's > 14, p<.01).

These results suggest that ambiguity avoidance may influence choices in constituent ordering, at least in off-line tasks like writing. However it does not affect both constructions equally: for DA, constituent order provides a good mechanism for avoiding ambiguities. But for HNPS, the shifted construction is so marked that it does not provide a good alternative for the ambiguous construction.

We are currently conducting a study to see if ambiguity avoidance also affects on-line production, and if it affects DA more than HNPS. The results of this experiment will also inform the more general question of whether speakers avoid ambiguities or not. Ferreira and Dell (1998) found that speakers did not avoid attachment ambiguities of a different kind (S-comp/direct object). However, our off-line data suggest that any ambiguity avoidance strategies are likely to interact with other factors. We predict that speakers will use word order to avoid PP-attachment ambiguities more when an unmarked word order is available, as with DA, than when the alternative orders are marked, as with HNPS.

 
 


© 2010 The MIT Press
MIT Logo