Summary: | The purpose of this study was to determine if a listener is able to
recognize sentential syntactic type on the basis of prosodic or segmental cues
when the availability of the other type of cue is severely reduced. Through a
process known as "spectral inversion" (Blesser, 1972), prosodic cues (temporal
and waveform amplitude cues) were maintained, while segmental cues were
reduced.
Two experiments were conducted, Experiments 1 and 2, in which
participants listened to digitized recordings. In Experiment 1, it was
demonstrated that when speech was spectrally inverted, listeners were able to
use syllabicity to some extent to identify words or word combinations from a
closed-set, but that overall, word recognition ability was severely reduced.
In the main experiment, Experiment 2, fifteen participants (21 to 29 years)
listened to three lists of sentences, each containing five exemplars of nine
different syntactic types of sentences that ranged in complexity. Each list was
presented in a separate condition. Condition 1 consisted of non-altered
sentences, condition 2 of spectrally-inverted sentences, and condition 3 of
concatenated sentences with reduced prosodic cues. Participants indicated the
type heard using a closed-set, forced-choice, nine-alternative response
paradigm. Participants had near-perfect accuracy in recognizing the syntactic
type in conditions 1 and 3, but were less accurate in condition 2. Examination of
response errors, however, indicated that poor performance in condition 2 was
primarily attributable to decreased ability to recognize only two of the syntactic
types, and did not necessarily reflect overall poorer performance. Additionally, it
was shown by the error patterns in condition 2, that the number of syllables in
the sentence served as an important cue. It was concluded that recognition of
syntactic type is possible with prosodic cues (temporal and waveform amplitude
cues) even when segmental cues are severely reduced.
|