Perceptual Separation Research Articles

Talker separation and sequential stream segregation in listeners with hearing loss: patterns associated with talker gender.

The purpose of this paper was to examine the relations between the ability to separate simultaneous sentences spoken by talkers of different gender and the ability to separate pitch patterns in a sequential stream segregation task. Simultaneous sentence pairs consisting of 1 sentence spoken by a male talker and 1 sentence spoken by a female talker were presented to 11 listeners with sensorineural hearing loss. Listeners were asked to repeat both sentences and were scored on the number of words repeated correctly. Separate scores were obtained for the male and female sentences. Sequential stream segregation was then measured using series of tones consisting of a fixed frequency (A) and a varying frequency tone (B). Tone series were presented in an ABA_ABA_... pattern starting at a varying frequency either below (ascending pattern) or above (descending pattern) the frequency of the fixed 1000 Hz tone. Fusion thresholds, defined as the frequency separation at which listeners could no longer perceptually separate the tones A and B, were obtained for both ascending and descending patterns. There was no significant difference between ascending and descending fusion thresholds based on the group data, but substantial individual differences were observed. Speech recognition scores for the male talker were strongly related to ascending fusion thresholds, but not descending thresholds. In contrast, speech recognition scores for the female talker were strongly related to the descending thresholds, but not the ascending thresholds. For both the male and female talkers, better recognition scores were associated with lower (nearer to normal) fusion thresholds. Results suggest that the importance of streaming in the perceptual separation of talkers may depend on the nature of the information provided by the changing pitch stream.

Independence and separability of volume and mass in the size-weight illusion.

Numerous size-weight illusion models were classified in the present article according to general recognition theory (Ashby & Townsend, 1986), wherein the illusion results from a lack of perceptual separability, perceptual independence, decisional separability, or a combination of the three. These options were tested in two experiments in which a feature-complete factorial design and multidimensional signal detection analysis were used (Kadlec & Townsend, 1992a, 1992b). With haptic touch alone, the illusion was associated with a lack of perceptual and decisional separability. When the participant viewed the stimulus in his or her hand, the illusion was associated only with a lack of decisional separability. Visual input appeared to improve the discrimination of mass, leaving only the response bias due to expectation.

Accuracy of color scission for spectral transparencies.

When surfaces are overlaid by a transparent filter, color scission refers to the perceptual separation of the colors of the image into the colors of the underlying surface and the color of the overlaying layer. We used filter matching to measure the accuracy of color scission for simulated physical filters and materials. Standard filters were placed on various sets of chromatic materials and match filters on achromatic materials. In the majority of cases, filter matching was close to veridical. The spectral effects of filters are complex, but with respect to the visual system, they can be closely approximated by 3-D affine transformations of cone absorptions or chromaticities. Veridical filter matches can be predicted by neural strategies that match ratios of mean cone absorptions or match mean chromatic contrasts between filtered and exposed regions. However, when the shape of a filter transmittance differed significantly from the shapes of background reflectances, the overlaid region had lower saturation than the surround, and filter matches had broader transmittance spectra than veridical.

Multidimensional Fechnerian Scaling: Perceptual Separability

A new definition of the perceptual separability of stimulus dimensions is given in terms of discrimination probabilities. Omitting technical details, stimulus dimensions are considered separable if the following two conditions are met: (a) the probability of discriminating two sufficiently close stimuli is computable from the probabilities with which one discriminates the projections of these stimuli on the coordinate axes; (b) the psychometric differential for discriminating two sufficiently close stimuli that differ in one coordinate only does not depend on the value of their matched coordinates (the psychometric differential is the difference between the probability of discriminating a comparison stimulus from a reference stimulus and the probability with which the reference is discriminated from itself). Thus defined perceptual separability is analyzed within the framework of the regular variation version of multidimensional Fechnerian scaling. The result of this analysis is that the Fechnerian metric of a stimulus space with perceptually separable dimensions has the structure of a Minkowski power metric with respect to these dimensions. The exponent of this metric equals the psychometric order of the stimulus space, or 1, whichever is greater.

Effects of reverberation on speech segregation

Perceptual separation of speech from interfering noise using binaural cues and fundamental frequency (F0) differences is disrupted by reverberation [Plomp, Acoustica 34, 200–211; Culling et al., Speech Commun. 14, 71–96]. Culling et al. found that the effect of F0 differences on vowel identification was robust in reverberation unless combined with even subtle F0 modulation. In the current study, speech reception thresholds (SRTs) were measured against a single competing voice. Both voices were either monotonized or normally intonated. Each came from recordings of the same voice, but interfering sentences were feminized (F0 increased 80%; vocal-tract length reduced 20%). The voices were presented from either the same or from different locations within anechoic and reverberant virtual rooms of Culling et al. In anechoic conditions, SRTs were lower when the voices were spatially separated and/or intonated, indicating that intonated speech is more intelligible than monotonous speech. In reverberant conditions (T60=400 ms), SRTs were higher, with no differences between the conditions. A follow-up experiment introduced sentences with inverted F0 contours. While acceptable in quiet, these sentences gave higher SRTs in all conditions. It appears that reverberant conditions leave intonated speech intelligible, but make it unseparable, while monotonous speech remains separable but is unintelligible.

The role of perceived spatial separation in the unmasking of speech

Spatial separation of speech and noise in an anechoic space creates a release from masking that often improves speech intelligibility. However, the masking release is severely reduced in reverberant spaces. This study investigated whether the distinct and separate localization of speech and interference provides any perceptual advantage that, due to the precedence effect, is not degraded by reflections. Listeners' identification of nonsense sentences spoken by a female talker was measured in the presence of either speech-spectrum noise or other sentences spoken by a second female talker. Target and interference stimuli were presented in an anechoic chamber from loudspeakers directly in front and 60 degrees to the right in single-source and precedence-effect (lead-lag) conditions. For speech-spectrum noise, the spatial separation advantage for speech recognition (8 dB) was predictable from articulation index computations based on measured release from masking for narrow-band stimuli. The spatial separation advantage was only 1 dB in the lead-lag condition, despite the fact that a large perceptual separation was produced by the precedence effect. For the female talker interference, a much larger advantage occurred, apparently because informational masking was reduced by differences in perceived locations of target and interference.

Assessing sensitivity in a multidimensional space: some problems and a definition of a general d'.

This article provides a formal definition for a sensivity measure, d'g between two multivariate stimuli. In recent attempts to assess perceptual representations using qualitative tests on response probabilities, the concept of a d' between two multidimensional stimuli has played a central role. For example, Kadlec and Townsend (1992a, 1992b) proposed several tests based on multidimensional signal detection theory that allow conclusions concerning the perceptual and/or decisional interactions of stimulus dimensions. One proposition, referred to as the diagonal d' test, relies on specific stimulus subsets of a feature-complete factorial identification task to infer perceptual separability. Also, Ashby and Townsend (1986), in a similar manner, attempted to relate perceptual independence to dimensional orthogonality in Tanner's (1956) model, which also involves d' between two multivariate signals. An analysis of the proposed d'g reveals shortcomings in the diagonal d' test and also demonstrates that the assumptions behind equating perceptual independence to dimensional orthogonality are too weak. This d'g can be related to a common measure of statistical distance, Mahalanobis distance, in the special case of equal covariance matrices.

Auditory stream segregation in dyslexic adults.

Developmental dyslexia is often associated with problems in phonological processing based on, or accompanied by, deficits in the perception of rapid auditory changes. Thirteen dyslexic adults and 18 control subjects were tested on sequences of alternating tones of high (1000 Hz) and low (400 Hz) pitch, which at short stimulus onset asynchronies (SOAs) led to perceptual separation of the sound sequence into high- and low-pitched streams. The control subjects perceived the tone sequence as connected down to SOAs of 130 ms, with segregation of the streams at shorter SOAs; in dyslexic subjects the segregation occurred already at 210 ms. Auditory stream segregation has previously been shown to impair the detection of phoneme order in segments of speech sounds. The observed aberrant segregation of sound streams in dyslexic subjects might thus contribute to their difficulties in achieving awareness of phonemes or phoneme order and in the acquisition of literacy.

Speech and hearing demonstrations in m a t l a b

Education can be greatly enhanced by the provision of tools which allow students to interact with, explore, and manipulate key elements of the phenomena under study. Such facilities are even more important in areas such as the perception, production, and recognition of sounds, whose subject matter is difficult to convey by words or pictures alone. Consequently, a collection of software applications has been developed, each adopting either an ‘‘experiment-replication’’ or a ‘‘white box’’ paradigm. The first category includes demonstrations of audiometric testing, perceptual separation of concurrent vowels, perception of alternating two-tone stimuli, and auditory induction. In the ‘‘white box’’ paradigm, students interact with the parameters of an algorithm or model, such as endpoint detection of speech, cepstral analysis, and models of peripheral auditory function. The applications are written in matlab, which provides a high-level interface to signal processing tools, graphical user interface components, and sound handling functions. As part of an ongoing action research study, the value of this work is being assessed against our learning objectives for final-year undergraduate and masters-level courses in speech and hearing technology. [Work supported by EU ELSNET Language Engineering Training Showcase.]

Perceptual independence of size and weight by dynamic touch.

Historically, the existence of a size-weight illusion has led to the conclusion that perceptions of size and weight are not independent. A dependence of perceived heaviness on physical volume (perceptual separability), however, is different from a dependence on perceived volume (perceptual independence). Three experiments were conducted to evaluate perceptual independence. The relations between perceived size and weight and physical size and mass were evaluated in Experiment 1. Perceived weight, length, and width were structured only by the corresponding physical variables, whereas variations in volume were not separable from variations in mass. F. G. Ashby and J. T. Townsend's (1986) test for perceptual independence was applied in Experiment 2. Perceived weight was independent of perceived length and volume. Experiment 3 used a magnitude estimation paradigm to investigate the extent to which information-perception relations could be related to the observed patterns of separability and independence.

Invariance of perceptual spaces and perceptual separability of stimulus dimensions.

Invariance of perceptual spaces and perceptual separability of stimulus dimensions.

Role of Fundamental Frequency Differences in the Perceptual Separation of Competing Vowel Sounds by Listeners With Normal Hearing and Listeners With Hearing Loss

This study compared the ability of listeners with normal hearing and listeners with moderate to moderately-severe sensorineural hearing loss to use fundamental frequency differences (delta F0) in the identification of monotically presented simultaneous vowels. Two psychophysical procedures, double vowel identification and masked vowel identification, were used to measure identification performance as a function of delta F0 (0 through 8 semitones) between simultaneous vowels. Performance in the double vowel identification task was measured by the percentage of trials in which listeners correctly identified both vowels in a double vowel. The masked vowel identification task yielded thresholds representing signal-to-noise ratios at which listeners could just identify target vowels in the presence of a masking vowel. In the double vowel identification task, both listeners with normal hearing and listeners with hearing loss showed significant delta F0 benefit: Between 0 and 2 semitones, listeners with normal hearing showed an 18.5% average increase in performance; listeners with hearing loss showed a 16.5% average increase. In the masked vowel identification task, both groups showed significant delta F0 benefit. However, the mean benefit associated with delta F0 differences in the masked vowel task was more than twice as large in listeners with normal hearing (9.4 dB) when compared to listeners with hearing loss (4.4 dB), suggesting less delta F0 benefit in listeners with hearing loss. In both tasks, overall performance of listeners with hearing loss was significantly worse than performance of listeners with normal hearing. Possible reasons for reduced delta F0 benefit and decreased overall performance in listeners with hearing loss include reduced audibility of vowel sounds and deficits in spectro-temporal processing.

Auditory grouping

Although our subjective experience of the world is one of discrete sound sources, the individual frequency components that make up these separate sources are spread across the frequency spectrum. Listeners. use various simple cues, including common onset time and harmonicity, to help them achieve this perceptual separation. Our ability to use harmonicity to segregate two simultaneous sound sources is constrained by the frequency resolution of the auditory system, and is much more effective for low-numbered, resolved harmonics than for higher-numbered, unresolved ones. Our ability to use interaural time-differences (ITDs) in perceptual segregation poses a paradox. Although ITDs are the dominant cue for the localization of complex sounds, listeners cannot use ITDs alone to segregate the speech of a single talker from similar simultaneous sounds. Listeners are, however, very good at using ITD to track a particular sound source across time. This difference might reflect two different levels of auditory processing, indicating that listeners attend to grouped auditory objects rather than to those frequencies that share a common ITD.

Separability and Independence of Dimensions within the Same–Different Judgment Task

Ashby, Townsend, and colleagues have developed an important model for the psychophysical identification task, the general recog- nition theory (also called, decision bound theory), based on multi- dimensional signal detection theory, that has been extended to categorization, Garner filtering, similarity judgments, and preference successfully. The virtue of this theory (and its extensions) is the ability to assess the nature of the perceptual representations of stimuli (e.g., their dimensional interactions), within a single framework and across paradigms. This paper adds to this ability by extending the theory, and related multidimensional perceptual models, to the problem of same–different judgments. Three tasks are described for the example of two-dimensional stimuli. In two of these tasks only one dimension is relevant (component same–different). In the third, both dimensions are deemed relevant (standard same–different). Several theorems concerning both response accuracy and response time have been proved relating the property of perceptual separability to certain response time invariances that can be observed in the component same–different tasks. This property, generally speaking, is the constancy of perception of one dimension when another irrelevant dimension changes in the stimulus. Also, under the assumption of within-trial perceptual indepen- dence, response probabilities, and response times of the two–component tasks can be meaningfully related to those same measures in the standard same–different paradigm, provided motor time variance is minimal.

Clutter and Display Conformality: Changes in Cognitive Capture

HUD technology has provided pilots with a variety of performance advantages over traditional head-down cockpit displays. Reduced visual switching times, increased situational awareness, and a variety of flight performance improvements have all been associated with HUDs. Despite these advantages, research has identified some possible HUD shortcomings. Objective and subjective data suggest HUDs increase the difficulty associated with switching one's attention among stimuli within the same visual space, causing pilots to inappropriately fixate on the display, a state termed cognitive capture. Possible factors that might affect cognitive capture include display conformality, clutter, and perceptual separation. The present research used reaction time data to assess the effects of these factors on cognitive capture. Sixty participants completed a simulated ship navigation task using either a conformal or non-conformal HUD. Each of four trials was factorial for display clutter (low, high) and obstacle probability (low, high). Reaction time data showed a performance advantage for conformal displays particularly in the high clutter condition. Subjective data revealed that the conformal display was less distracting, appeared to depict fewer instruments, and required less effort to attend to the environment than the non-conformal display. Findings concerning perceptual separation were inconclusive.

Effect of perceptually separating spectral components on profile discrimination

In spectral-shape discrimination [D. M. Green, Profile Analysis (Oxford U.P., Oxford, UK, 1988)], listeners discriminate a standard spectrum in which all components are equal in level, from a signal-plus-standard spectrum in which the level of one component (the signal component) is higher than the level of other nonsignal components. Because the overall level is randomly selected from a wide range of values on each stimulus presentation, the absolute level of the signal component is not a reliable cue for discrimination. Instead, listeners must make a relative level comparison between the signal and nonsignal components. Under normal conditions of profile experiments, listeners can detect signal increments as small as 1 dB. To be discussed here are special cases where the signal component is perceived as a separate sound from the nonsignal components. Perceptual separation between the signal and nonsignal components was achieved by introducing asynchrony in their onsets and offsets, by modulating them differently, or by presenting them in separate ears. Experimental results showed that such separation made spectral-shape discrimination very difficult. Thus optimum performance requires that the signal and nonsignal components be perceived as belonging to a single auditory object. [Work supported by NIH.]

Perceptual separability, decisional separability, and the identification–speeded classification relationship.

Perceptual separability, decisional separability, and the identification–speeded classification relationship.

Perceptual separability, decisional separability, and the identification-speeded classification relationship.

Three observers participated in several sessions of identification and speeded classification. Predictions from F. G. Ashby and W. T. Maddox's (1994a) response time (RT) theory of speeded classification were tested. A key theoretical concept is decisional separability (DS), that is, that the decision about the level of 1 component does not depend on the perceived value of the other component. The theory predicted that DS would hold in the filtering task but would be violated in the redundancy task, resulting in the redundancy gains. To summarize the results, (a) DS held in the filtering conditions, but not in the redundancy conditions; (b) redundancy gains occurred; (c) despite the redundancy gains, strong evidence was obtained that the stimulus components are perceptually separable; (d) 2 new models that each assumed RT decreases with the distance between the percept and the decision bound provided good accounts of the RT distributions and accuracy rates; and (e) the shift from identification to speeded classification influenced both perceptual and decisional processes.

Multidimensional signal detection analyses (MSDA) for testing separability and independence: A Pascal program

A Pascal (Turbo 6.0) program for testing perceptual separability, decisional separability, and perceptual independence—three types of independence defined in the context of general recognition theory (GRT)—is presented. GRT is essentially a multidimensional extension of signal detectability theory (Green & Swets, 1966) and has been employed in complete-identification experimental paradigms to examine whether information processing of stimulus dimensions is perceptually and/or decisionally separable and/or perceptually independent. The definitions of these different forms of “independence” are provided, the experimental design is described, and the program, MSDA, is described and demonstrated.

Timbre and Auditory Streaming

Listeners rated the subjective difference between pairs of different synthesized orchestral instrumental timbres. They then reported their perception of variations on crossing ascending and descending musical scales played in different timbres. With little or no timbre difference between the scales, auditory streaming by pitch led to the perception of separate high and low halfscales. Increasing timbre difference led to the perception of the complete scales played by each instrument, particularly if the two scales were in distant keys or temporally interleaved rather than simultaneous. If the notes of each scale alternated between two instruments, then perceptual separation by timbre would result in the perception of the "jumping" sequence of notes played by each instrument. This sequence was perceived only if the scales were discordant or temporally interleaved. Multidimensional scaling of the difference ratings led to three dimensions, which were correlated with acoustic parameters resulting from spectral analysis of the sounds. The most important aspects of timbre for perceptual separation were the proportion of energy in the lower partials and, for discordant scales, the duration of the decay. Auditory streaming by timbre thus depends on particular dimensions of timbre and on musical factors such as harmonic relation, simultaneity, and continuity.