Structural frequency as a variable in sentence processing: training alternative structures in relative clause disambiguation

Javier García-Orza, Jose Miguel Rodríguez Santos, Maria Jose Blanca
jgorza@uma.es, jmiguel.rodriguez@uma.es, Blamen@uma.es
University of Malaga (Spain)

The role of experience at structural levels in the sentence disambiguation process is a controversial issue. Although some models (constraint-based models, tuning) defend the relevance of structural experience, this variable is neglected or ignored by a great number of other models (garden-path, construal, prosodic phrasing, parameterized head-attachment). Currently some proponents of the role of experience in sentence processing have expressed their doubts about the role of experience with syntactic structures (Gibson and Pearlmutter, 1998). These doubts increase with the finding of new data from corpus studies and production studies which seem to show that structural exposure is not able to explain attachment preferences (Mitchell and Brysbaert, 1998; Pickering, Traxler and Crocker, 2000).

In this research we defend the role of experience with syntactic structures in sentence processing, through the presentation of several experiments carried out in Spanish. In these studies structural frequency is manipulated, so one experimental group is exposed to high attachment relative clause sentences (1a), and the other experimental group to low attachment sentences (1b).

(1)

a.

Los vecinos encontraron al inquilino (masc) de la casera (fem) que pagaba disgustada (fem) las facturas (low attachment)
The neighbors found the tenant (masc) of the lanlady (fem) who paid angry (fem) the bills

b.

Los vecinos encontraron al inquilino (masc) de la casera (fem) que pagaba disgustado (masc) las facturas (high attachment)
The neighbors found the tenant (masc) of the landlady (fem) who paid angry (masc) the bills

Data show an asymetric effect in both off-line and on-line experiments. The initial high attachment preferences, habitually found in Spanish, do not change in the group exposed to high attachment. However, the group exposed to low attachment sentences, those sentences that are less frequent in Spanish and are not initially preferred by people, do change their structural preferences.

These findings are discussed in relation to models of sentence disambiguation, and several plausible explanations are analysed. The need to consider structural frequency as a variable in sentence processing is argued for.

We also claim that we need to consider structural frequency as a variable together with others in the processing of sentences, and not as the only one, as is claimed by the tuning hypothesis.

Several mechanisms through wich this variable might work are considered. At issue is wheter structural frequency is reduced to sentence disambiguation or whether it takes part in nonambiguous sentence processing too.

References

Gibson, E. & Pearlmutter, N.J. (1998). Constraints in sentence comprehension. Trends in Cognitive Sciences, 2, 262-268

Mitchell, D.C.; Brysbaert, M. (1998). Challenges to recent theories of crosslinguistic variation in parsing: Evidence from Dutch. In D. Hillert (Ed.), Sentence processing: A cross-linguistic prespective (pp. 313-335). NY: Academic Press.

Pickering, M.J.; Traxler, M.J. & Crocker, M.W. (2000). Ambiguity resolution in sentence processing: Evidence against frequency-based accounts. Journal of Memory and Language, 43, 447-475.