Do gender and number features lead to different relative clause attachment preferences in Spanish?

Manuel Carreiras, Moises Betancort and Enrique Meseguer
mcarreir@ull.es
Univeristy of La Laguna. Tenerife, Spain

There is a large amount of evidence showing that the ambiguity of what the relative clause modifies in phrases like (1) tends to be resolved differently in different languages.

(1)

Someone shot the maid of the actress who was divorced
Alguien disparó contra la criada de la actriz que estaba divorciada

After the seminal study of Cuetos & Mitchell, there has been shown that there is a weak but seldom significant tendency for the relative clause to be taken as modifying the second noun phrase, the actress, in English; while in Spanish, a significant preference for the relative clause to modify the first noun phrase, the maid has been found (e.g., Carreiras & Clifton, 1993, Carreiras & Clifton, 1999; though see Fernández, 2000). Readers in other languages showed either an early closure advantage in Dutch (Brysbaert & Mitchell, 1996), French (Zagar, Pynte & Rativeau, 1997) German (Hemforth, Konieczny & Scheepers, 1994), as well as a late closure advantage in Italian (De Vincenzi & Job, 1993; De Vincenzi & Job, 1995; though see Frenck-Mestre & Pynte, 2000).

The fact that the Late Closure strategy fails to apply for these type of structures in some languages is against the universality of late closure and therefore of the garden path model (Frazier & Rayner, 1992). Other hypothesis such as "tuning hypothesis" (Mitchel & Cuetos, 1991); "predicate proximity" (Gibson et al, 1996); "anaphoric binding hypothesis" (Hemforth et al, 1997); "construal hypothesis" (Frazier & Clifton, 1996) have been put forward to explain the previous results.

Most previous studies on the relative clause attachment to either N1 (high attachment) or to N2 (low attachment) that found high attachment preference disambiguated either by gender agreement or pragmatic information. However, Fernández (2000) obtained low attachment preference in Spanish by using number agreement disambiguation. The results obtained by Fernández (2000) imply that high attachment is not a genuine initial general preference in Spanish for this type of ambiguity, but rather a late preference due to the use of gender disambiguation (probably an extra syntactic feature). Carreiras, Betancort and Meseguer (2001) investigated whether gender and number disambiguation lead to different parsing strategies of relative clauses in Spanish at very early stages of processing. They carried out an eyetracking experiment in which gender and number disambiguation was manipulated. Their results showed a high attachment preference for gender disambiguation (total time, second pass and regressions path time), while a low attachment preference for number disambiguation (second pass time). In order to gather more empirical evidence regarding gender and number disambiguation, as well as to trace more carefully the time course of the disambiguation process, a second eyetracking experiment including more items, a completely ambiguous condition, and questions after each sentence was carried out. Preliminary results show an initial low attachment preference for number disambiguation. In contrast, they show a high attachment preference for gender disambiguation at later measures. In addition, at early measures (first pass reading times), the ambiguous condition was faster than the high attachment number condition, but no different from the other three conditions. Theoretical implications of the findings will be discussed regarding the existing models.