Pitch continuity and speech source attribution.

Darwin, C. J.; Bethell-Fox, C. E.

doi:10.1037/0096-1523.3.4.665

Cited by 55 publications

(39 citation statements)

References 21 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The tone precursors maintainedacoustic compatibility with the FOof the test word, without having the incompatible formant spacing of some of Diehl et al 's stimuli. The finding of no difference in rate effects induced by the tone precursors and the plain precursors is consistent with the importance that has been attributed to FO and its harmonics in isolating one speaker's voice from many (Darwin & Bethell-Fox, 1977), and ir determining what acoustic energy contributes to phonetic perception Darwin & Gardner, 1986).…”

Section: Discussionsupporting

confidence: 71%

Induction of rate-dependent processing by coarse-grained aspects of speech

Gordon

1988

Perception & Psychophysics

View full text Add to dashboard Cite

This study investigated the idea that human speech recognition can involve analyzing the speech signal at multiple levels of resolution, using the information obtained from relatively coarse levels of analysis as a context for interpreting detailed acoustic cues to segment identity. Three experiments examined the effectiveness of coarse-grained aspects of speech in inducing rate-dependent processing of closure duration as a cue to phonological voicing in a medial stop consonant (specifically, rabid vs. rapid). Experiment 1 showed that the rate of articulation of a severely filtered precursor phrase influenced voicing judgments about a segment in an unfiltered test word. Experiment 2 showed a similar effect when the amplitude envelope of the precursive speech was filled with a constant-frequency sine wave set at the fundamental of the test word. The contextual effects of these coarse-grained aspects of speech did not differ from those of acoustically detailed precursive speech. Experiment 3 showed that no context-dependent processing occurred when the amplitude envelope of the precursive speech was filled with white noise, indicating that the precursive sounds must have some acoustic continuity with the test word for integration to take place. The results support the idea that context-dependent processing can be based on coarse aspects of the speech signal.

show abstract

Section: Discussionsupporting

confidence: 71%

Induction of rate-dependent processing by coarse-grained aspects of speech

Gordon

1988

Perception & Psychophysics

View full text Add to dashboard Cite

show abstract

“…Darwin and Bethell-Fox (1977) showed that, in synthesized speech, a sudden change in fundamental frequency created a boundary across which phonetic interpretation processes would not integrate spectral information. Bregman, Ahad, Kim, and Melnerich (1994) and showed that the suddenness of changes in amplitude affected the tendency of a particular frequency to stand out from a background of other pitches.…”

Section: Discussionmentioning

confidence: 99%

Cumulation of the tendency to segregate auditory streams: Resetting by changes in location and loudness

Rogers

Bregman

1998

Perception & Psychophysics

108

View full text Add to dashboard Cite

In four experiments, the accumulation, over time, of a tendency to hear separate high and low streams in a sequence of high (H) and low (L)tones, presented in a galloping rhythm (HLH-HLH-...), was studied. Each trial was composed of two parts, an induction sequence, then a test sequence, with no break between them. The test sequence was always heard at the far left. When the induction sequence and the test sequence were identical, the presence of the induction sequence increased the tendency for the test sequence to split into two streams. However, when the sequences differed in location (cued by differences in interaural timing or intensity over headphones and by loudspeaker placement in a free field) or when they differed in loudness, the accumulation of the segregative tendency was reset, and the test sequence sounded more integrated. When the induction sequence changed in location or loudness in gradual steps toward the value of the test sequence, resetting was much less. It appears that the accumulation of information about streams in different frequency regions is sensitive to sudden changes in parameters, even when they affect the frequency regions equally. This prevents the system from accumulating data across unrelated events.It has been known for some time that the tendency for a listener to segregate a sequence of alternating high and low tones into separate high and low streams increases as the sequence continues (Anstis & Saida, 1985;Bregman, 1978). This cumulative effect and the stream segregation effect itselfhave been explained by some authors as being due to the fact that repeated high-low and low-high transitions habituate neural processes that detect frequency transitions (Anstis & Saida, 1985;van Noorden, 1975). Anstis and Saida also reported that a sudden shift of an alternating sequence ofhigh and low tones from one side of the head to the other led to the temporary loss of separate high and low streams. It appeared that the segregative tendency had been reset and had to build up anew at the new location. Their explanation was that the buildup of segregation was specific to a particular pool of neurons that detected frequency transitions and that those stimulated by the left-and right-ear signals were different. The same resetting of segregation was found when the alternating high-low sequence was changed abruptly to a new frequency region; this supported the idea that each region had its own frequency transition detectors.Rogers and Bregman (1993) used a new paradigm to study the buildup of the strength of the tendency to segregate higher and lower tones into separate streams. They used an alternation of a high (H) and a low (L) tone in a galloping rhythm, which can be schematized as follows: HLH-HLH-HLH-... , where a dash represents a silence equal in length to a tone (100 msec). Each trial was divided into two parts, a long induction sequence and a short test sequence. The test sequence (1.2 sec long) followed the induction sequence (4.8-8 sec long) with no delay. All the test sequences ...

show abstract

“…In addition, a formant (frequencies with greater energy that correspond to vowel identity) of a single vowel may become segregated when the formant has a differing F 0 under certain conditions (Broadbent and Ladefoged, 1957;Gardner, Gaskill, and Darwin, 1989). Finally, speech stimuli with discontinuous pitch contours tend to segregate at the discontinuities (Darwin and Bethell-Fox, 1977).…”

Section: Harmonicity and Pitchmentioning

confidence: 99%