Erin L. O'Bryan,1 David J. Townsend2 & Thomas G. Bever1
1 University of Arizona, Tucson, 2 Montclair State University
Post-sentence representations of auditory sentences can be investigated using the voice change monitoring task, in which participants monitor sentences for a syllable spoken in a different voice. "Slips of the ear", i.e., incorrect recall of the voice change location, provide information about how auditory sentences are represented in memory. We report evidence, based on "slips of the ear", that some syntactic information is preserved in the post-sentence representation.
Voice change monitoring has been shown to be sensitive to online and offline effects, such as the reduced relative and filled gap effects (Townsend & Bever, 1991; O'Bryan, Nicol, Townsend, & Bever, 2000). The task provides three kinds of data: online reaction times, online monitoring errors, and post-sentence word identification responses (participants' reports of which word the voice change occurred on.)
Analyses of post-sentence word identification errors by mislocation distance, direction, and syntactic category revealed four types of "slips of the ear". The first is leftward movement. In response to (a) in which the voice change occurred on "flew" (indicated by asterisks), participants reported that it occurred on "hawk" 22.2% of the time.
(a) A hawk *flew* by the radio tower on the hill.
Mislocating the voice change one word leftward accounts for 32.4% of mislocations, more than any other position.
The second type is rightward mislocation when one word to the right is often phonetically reduced, e.g., auxiliaries. Participants responded "have" to (b) 34% of the time, accounting for 79% of errors on this item.
(b) Could the *crew* have known about the malfunctioning safety equipment?
Although auxiliaries are often reduced, they were fully articulated in the stimuli. Across items, 19.8% of mislocations involved choosing the following auxiliary. This suggests that even when auxiliaries are carefully articulated, their reducibility is retrievable from the post-sentence representation.
The third type involved rightward mislocation from a preposition to its object noun phrase. Participants responded "health" in 69.8% of the errors on (c).
(c) The retired women bathed in the large pool *at* the health club.
The high frequency of mislocations from prepositions to their objects suggests that categorical and possibly structural information is intact in the post-sentence representation.
The fourth type of mislocation was incorrectly choosing a verb. In 51.7% of errors on (d), participants chose "infuriated".
(d) The police infuriated the *deaf* man when they handcuffed him.
Incorrectly choosing a verb accounts for 21.5% of mislocations. Verbs were chosen in 60.4% of errors in which there was no voice change. This response bias suggests that verbs have a privileged status in the post-sentence representation.
References
O'Bryan, E. L., Nicol, J. L., Townsend, D. J., & Bever, T. G. (2000). Reduced relatives and WH-gaps in spoken sentence comprehension. CUNY Sentence Processing Conference poster session, Univ. of California, San Diego.
Townsend, D. J., & Bever, T. G. (1991). The use of higher-level constraints in monitoring a change in speaker demonstrates functionally distinct levels of representation in discourse comprehension. Language and Cognitive Processes, 6(1), 49-77.