Non-native processing of relative clause attachment ambiguity

Maren Heydel, University of Abertay Dundee
Wayne Murray, University of Dundee

Attachment of a relative clause to a complex noun phrase constitutes a sentence processing phenomenon for which cross-linguistic differences have been reported for English on one hand (attaching low) and a number of other languages attaching high. These differences, which have been fairly consistent, have so far defied a conclusive explanation.

Gricean principles have been suggested as one possible explanation, as in Construal (Frazier and Clifton, 1996). To examine the possibility of a link between Gricean principles and non-native speakers' attachment preferences in the foreign language (English) we tested a population of speakers of English as a Foreign Language for whose native languages high attachment of the relative clause has been reported (French, German, Greek, Spanish). Participants were asked to complete a production task which required them to read a short passage that was compatible with either high or low attachment and then complete a paraphrase as indicated (ex. 1 and 2). A week later their offline attachment was tested using the same type of items (ex. 3).

As the alternative structure to express high attachment (an Anglo-Saxon genitive) becomes more functionally available, participants should be more inclined to use it in high attachment situations. If the availability (rather than the existence) of alternative structures drives participants' attachment preferences, participants should then be more likely to favour low attachment.

We found that overall participants favoured high attachment (as they would in their native languages) while, at the same time, there was considerable variation between participants congruent with their level of English language proficiency attested by independent proficiency tests. However, there was a significant positive correlation between the relative
availability of the Anglo-Saxon genitive for the description of high attachment situations and of the Norman genitive for the description of low attachment situations in production on one hand and the probability of low attachment in comprehension on the other hand. We interpret these results in terms of Gricean principles being operative rather than language independent processing in non-native speakers using only one set of parsing routines (i.e. that of their native language).

(ex. 1) High Attachment situation:

There is an actress, and she has a manager. The manager was on the balcony, and somebody shot him.

Somebody
shot.............................................................who was on the balcony

(ex. 2) Low attachment situation:

There is an actress, and she has a manager. Somebody shot the manager. The actress was on the balcony.

Somebody
shot.............................................................who was on the balcony

(ex. 3) Comprehension:

Someone shot the manager of the actress who was on the balcony. Who was on the balcony?

References

Frazier, L., & Clifton, C. Jr. (1996). Construal. MIT Press.