Eric S. Solomon & Neal J. Pearlmutter
Northeastern University
esolomon@neu.edu
This research examines syntactic planning in production. Bock and Cutting (1992) showed that number agreement production fails more often when verbs are separated from subject head nouns by a prepositional phrase (PP) ("The report of the destructive fires...") than by a clause ("The report that they destroyed the fires..."). They proposed that clauses are planned independently, so that local noun number (on "fires") does not interfere with number-marking on the head noun ("report") in such cases. Solomon and Pearlmutter (2001) proposed that number-agreement error patterns in production depend on degree of semantic integration between words and phrases. Speakers completing fragments like "The drawing of/with the flower(s)" produced a larger error effect in the "of" conditions, suggesting that the integration relationship between nouns within a phrase influences planning: More tightly integrated elements (e.g., of-PPs) tend to be planned together, and thus properties of such elements (e.g., grammatical number) will have a greater chance of interfering with each other.
Solomon and Pearlmutter's (2001) experiments confounded semantic integration differences with the argument/adjunct distinction. Two new experiments eliminate this confound by examining semantic integration effects within adjunct PPs. Experiment 1 contrasted for-PPs (1a) with accompaniment-PPs (1b), and Experiment 2 contrasted attribute-PPs (2a) with accompaniment-PPs (2b). In both experiments, larger error effects (plural local noun vs. singular local noun) appeared in the more integrated conditions (for-PPs in Experiment 1; attribute-PPs in Experiment 2), ruling out an explanation solely in terms of the argument/adjunct distinction.
| 1. | a. | The chauffeur for the actor(s) |
| b. | The chauffeur with the actor(s) | |
| 2. | a. | The pizza with the yummy topping(s) |
| b. | The pizza with the tasty beverage(s) |
A third experiment directly contrasts the argument/adjunct distinction with semantic integration, using of-PPs (3a), relative clauses (RCs; 3b), and full sentence complements (SCs; 3c). Of-PPs showed the largest error effect. SCs, which are arguments but are not semantically integrated, showed no error effect; whereas RCs, which are adjuncts but are integrated by coindexation, showed an intermediate effect. This is the reverse of what the argument/adjunct distinction must predict to account for Solomon and Pearlmutter's (2001) results.
| 3. | a. | The report of the nasty auto accident(s) |
| b. | The report that described the auto accident(s) | |
| c. | The report that Megan described the accident(s) |
Finally, we provide data from a semantic integration norming study in which participants rated how "closely linked" head and local nouns were within each phrase from Experiments 1-3. These ratings confirmed the intuitive integration differences and predicted agreement error rates across plural local noun conditions, but they could not account for the Experiment 3 of-PP versus RC error rate difference. This latter result indicates that a factor like Bock and Cutting's (1992) clause-boundedness may be needed in addition to semantic integration.
References
Bock, K., & Cutting, J. C. (1992). Regulating mental energy: Performance units in language production. Journal of Memory and Language, 31, 99-127.
Solomon, E. S., & Pearlmutter, N. J. (2001). Effects of linking prepositions in the production of subject-verb agreement. Paper presented at the 14th CUNY Conference on Human Sentence Processing, Philadelphia, PA.