Ling 423/640G: Cognitive Linguistics

Ben Bergen

Meeting 22: Grammar [4]

November 13, 2008

Grammatical constructions appear to play a role in language processing. How about in production?

Structural priming

Hearing or producing a sentence with a particular syntactic structure increases the likelihood that you'll use the same structure in the future.

This is taken as evidence that the overt syntactic structures of sentences are psychologically real - otherwise, how could they be primed?

There are two main models of how this priming works.

• Using one structure activates that structure, and this persistent activation makes the subject more likely to reuse it - since it's already active.
• Language users are constantly reorganizing their grammatical knowledge - strengthening some constructions and weakening others. Using a construction strengthens it.

Previous work only tests whether a construction can be primed by a preceding prime sentence. This work tests whether there are also effects due to the overall proportion of preceding sentences using the given construction.

They did this by contrasting two constructions:

• Double Object [DO]: “Meghan gave her mom a toy”
• Prepositional Object [PO]: “Meghan gave a toy to her mom” [we've called this Dative or Caused-Motion]

Experiment 1

The basic method in both experiments presented subjects with the beginning of a sentence and had them type in a completion for it.

PO-biased: Meghan gave a toy _______

DO-biased: Meghan gave her mom ______

Unbiased: Paul gave ______

In a Priming phase, they saw six pairs of first a biased stem, then an unbiased stem in sequential trials [each pair separated from others by four fillers].

In order to test whether previous experience affected this priming, Ss were first exposed to a long Recent Experience session, in which they produced the same number of PO and DO sentences, or only one type, along with a bunch of fillers. Three conditions:

• Equal Exposure (EE) produced an equal number of DO and PO constructions, with the tokens of each construction were evenly spaced through this part of the experiment.
• Equal Exposure-Block (EE-B) condition also produced an equal number of the target constructions, but the tokens of each construction were blocked such that the tokens of one construction appeared entirely in the first half of the Recent Experience phase, and the tokens of the alternative construction appeared entirely in the second half of this phase.
• Unequal Exposure (UE) condition produced examples of only one construction.

Results showed a priming effect in both the EE and EE-B conditions, but not in the UE condition, where the priming effect is greatly reduced, due to a decrease in priming by the consruction incompatible with the UE construction.

This suggests that the distribution of constructions in recent experience affects structural priming, but the order of presentation does not.

Experiment 2

Fixed two failings of the first experiment - added a 75% recent experience condition, and also increased the size of the EE-B to 10 of each.

Results again showed a priming effect in EE-B, less in UE-75, and none in the UE-100 condition.

Conclusions

These findings show

• The overall proportion of constructions influences structural priming, regardless of order.
• There are local priming effects when the proportions in recent experience are equal
• Structural priming is a long-term effect, due to strengthening of frequent constructions
• For us, what's important is that grammatical structures, or constructions, seem to be cognitively real in language production