Stimulus Generalization Paradigm Research Articles

Perception of degraded speech by chinchillas (Chinchilla laniger): Word-level stimulus generalization.

One characteristic of human speech perception is a remarkable ability to recognize speech when the speech signal is highly degraded. It has been argued that this ability to perceive highly degraded speech reflects speech-specific mechanisms. The present study tested this hypothesis by measuring the ability of chinchillas to recognize noise-vocoded (NV) versions of naturally spoken monosyllabic words using operant conditioning in a stimulus generalization paradigm. Chinchillas do not generalize the vocoded words to be perceptually equivalent to the naturally spoken words. The responses from chinchillas to the vocoded words fall well below their responses to the naturally spoken words. In this case, pitch cues rather than speech cues may be controlling the behavioral responses. To reduce pitch cues, chinchillas were retrained using 64-channel NV words. The responses from chinchillas to the vocoded test words were now similar to those of the 64-channel versions and were similar to those obtained from human listeners. However, responses obtained from chinchillas to time-reversed versions were high and similar to responses obtained to time-normal versions suggesting that the cue controlling behavioral responses was the phonetic structure of the words. These results show that chinchillas used different acoustic cues than human listeners. The ability of chinchillas to recognize NV words as being perceptually equivalent to the naturally spoken versions is inferior compared to that of human listeners. The findings suggest that the ability of human listeners to recognize highly degraded speech is unlikely to be based solely on the general auditory and perceptual mechanisms that are common among mammals. (PsycINFO Database Record (c) 2019 APA, all rights reserved).

Processing pitch in a nonhuman mammal (Chinchilla laniger).

Whether the mechanisms giving rise to pitch reflect spectral or temporal processing has long been debated. Generally, sounds having strong harmonic structures in their spectra have strong periodicities in their temporal structures. We found that when a wideband harmonic tone complex is passed through a noise vocoder, the resulting sound can have a harmonic structure with a large peak-to-valley ratio, but with little or no periodicity in the temporal structure. To test the role of harmonic structure in pitch perception for a nonhuman mammal, we measured behavioral responses to noise-vocoded tone complexes in chinchillas (Chinchilla laniger) using a stimulus generalization paradigm. Chinchillas discriminated either a harmonic tone complex or an iterated rippled noise from a 1-channel vocoded version of the tone complex. When tested with vocoded versions generated with 8, 16, 32, 64, and 128 channels, responses were similar to those of the 1-channel version. Behavioral responses could not be accounted for based on harmonic peak-to-valley ratio as the acoustic cue, but could be accounted for based on temporal properties of the autocorrelation functions such as periodicity strength or the height of the first peak. The results suggest that pitch perception does not arise through spectral processing in nonhuman mammals but rather through temporal processing. The conclusion that spectral processing contributes little to pitch in nonhuman mammals may reflect broader cochlear tuning than that described in humans.

Spectral processing does not give rise to behaviorally relevant cues for “pitch” perception in mammals.

Mechanisms giving rise to pitch reflect spectral or temporal processing is still equivocal, because sounds having strong harmonic structures also have strong temporal structures. When a harmonic tone complex is passed through a noise vocoder, the resulting sound can have a harmonic structure with large peak-to-valley ratios, but little or no temporal structure. To test the role of harmonic structure in mammals, we measured behavioral responses to vocoded tone complexes in chinchillas using a stimulus generalization paradigm. Animals discriminated a harmonic tone complex from a one-channel vocoded version of the complex. When tested with vocoded versions generated with 8–128 channels, animals generalized to the one-channel version and showed no gradient in their behavioral responses. This suggests that spectral structure was not the cue for the behavioral response. To further test this, chinchillas discriminated an iterated rippled noise from the one-channel vocoded tone complex. When tested using vocoded tone complexes having harmonic peak-to-valley ratios that were larger than or similar to the rippled noise, animals again generalized to the 1-channel version rather than the to rippled noise. The results suggest that mammalian “pitch” attributes arise through temporal, not spectral, processing mechanisms. [Work supported by NIDCD R01 DC005596.]

Sound source segregation by goldfish: Two simultaneous tones

The perception of two simultaneous tones was investigated in goldfish using classical respiratory conditioning and a stimulus generalization paradigm. Pairs of tones were used to make up a mixture of 150 Hz and a higher harmonic or a mistuned harmonic. Fish were conditioned to the two-tone mixture and then tested for generalization to several pure tones. The simultaneous tones tended to be segregated in perception, with the generalization gradient for single tones having two peaks corresponding to the frequencies of the tone pairs. There were no consistent differences in the generalization gradients following conditioning to harmonic or inharmonic tone pairs. In addition, experiments were carried out in which the two tones of the pair were heard on alternate trials, always as single tones, followed by generalization tests to single tones. There was more generalization in this experiment, reflecting the fact that conditioning and generalization test stimuli were both single tones. However, the shapes of the generalization gradients were similar to those in which fish were conditioned to two simultaneous tones, indicating that the simultaneity of the tones did not make them harder to segregate. As the frequency separation between the two components narrowed, segregation tended to fail.

Pitch perception in chinchillas (Chinchilla laniger): Stimulus generalization using rippled noise.

Rippled noises evoke the perception of pitch in human listeners. Infinitely iterated rippled noise (IIRN) is generated when wideband noise (WBN) is delayed, attenuated, and added to the original WBN through either a positive (+) or a negative (-) feedback loop. The pitch of IIRN[+] is matched to the reciprocal of the delay, whereas the pitch of IIRN[-] for the same delay is an octave lower. Chinchillas (Chinchilla laniger) were trained to discriminate IIRN[+] with a 4-ms delay from IIRN[+] with a 2-ms delay and then tested in a stimulus generalization paradigm with IIRN[+] at delays between 2 and 4 ms. Systematic gradients in behavioral response occurred along the dimension of delay, suggesting that a perceptual dimension corresponding to pitch exists for IIRN[+]. Behavioral responses to IIRN[-] test stimuli were more variable among chinchillas, suggesting that IIRN[-] did not evoke similar pitches relative to IIRN[+]. Systematic gradients in behavioral response were observed when IIRN[-] test stimuli were presented in the context of other IIRN[-] stimuli. Thus, other perceptual cues such as timbre may dominate the pitch cues when IIRN[-] test stimuli are presented in the context of IIRN[+] stimuli.

Pitch cue learning in chinchillas: The role of spectral region in the training stimulus

Chinchillas were trained to discriminate a cosine-phase harmonic tone complex (COS) from wideband noise (WBN) and tested in a stimulus generalization paradigm with tone complexes in which phase differed between frequency regions. In this split-phase condition, responses to complexes made of random-phase low frequencies, cosine-phase high frequencies were similar to responses to the COS-training stimulus. However, responses to complexes made of cosine-phase low frequencies, random-phase high frequencies were generally lower than their responses to the COS-training stimulus. When tested with sine-phase (SIN) and random-phase (RND) tone complexes, responses were large for SIN, but were small for RND. Chinchillas were then trained to discriminate infinitely-iterated rippled noise (IIRN) from WBN and tested with noises in which the spectral ripple differed between frequency regions. In this split-spectrum condition, responses were large to noises made of rippled-spectrum low frequencies, flat-spectrum high frequencies, whereas responses were generally lower to noises made of flat-spectrum low frequencies, rippled-spectrum high frequencies. The results suggest that chinchillas listen across all frequencies, but attend to high frequencies when discriminating COS from WBN and attend to low frequencies when discriminating IIRN from WBN.

Responses of cochlear nucleus units recorded from chinchillas trained to discriminate iterated rippled noise from flat-spectrum noise

Behavioral responses obtained from chinchillas using a stimulus generalization paradigm show that the perception of pitch strength is dependent on the stimuli used for training. When animals are trained to discriminate a cosine-phase harmonic tone complex (COS) from a flat-spectrum, wideband noise (WBN), generalization gradients indicate that animals rely primarily on temporal information in the envelope. Gradients from animals retrained to discriminate infinitely iterated rippled noise (IIRN) from WBN indicate animals rely more on information in the fine structure. Thus, chinchillas learn to use the information in the fine structure when trained with an appropriate stimulus. In order to begin to examine whether there exists a neural correlate of this learning at the level of the cochlear nucleus, single-unit responses to IIRNs and WBN were recorded from three anesthetized chinchillas that had received considerable behavioral training in discriminating IIRNs from WBN. Preliminary data were obtained for 19 single units consisting of primarylike, chopper, and onset response types. All-order interspike-interval histograms obtained for IIRNs and WBN were qualitatively similar to those obtained from naive, untrained animals for each unit type. Quantitative comparisons among units recorded from trained and untrained animals will be presented. [Work supported by NIDCD R01 DC005596.]

Listening experience with iterated rippled noise alters the perception of ‘pitch’ strength of complex sounds in the chinchilla

Behavioral responses obtained from chinchillas trained to discriminate a cosine-phase harmonic tone complex from wideband noise indicate that the perception of 'pitch' strength in chinchillas is largely influenced by periodicity information in the stimulus envelope. The perception of 'pitch' strength was examined in chinchillas in a stimulus generalization paradigm after animals had been retrained to discriminate infinitely iterated rippled noise from wideband noise. Retrained chinchillas gave larger behavioral responses to test stimuli having strong fine structure periodicity, but weak envelope periodicity. That is, chinchillas learn to use the information in the fine structure and consequently, their perception of 'pitch' strength is altered. Behavioral responses to rippled noises having similar periodicity strengths, but large spectral differences were also tested. Responses to these rippled noises were similar, suggesting a temporal analysis can be used to account for the behavior. Animals were then retested using the cosine-phase harmonic tone complex as the expected signal stimulus. Generalization gradients returned to those obtained originally in the naïve condition, suggesting that chinchillas do not remain "fine structure listeners," but rather revert back to being "envelope listeners" when the periodicity strength in the envelope of the expected stimulus is high.

Repetition ‘‘pitch’’ in chinchillas

Repetition pitches are evoked in human listeners by rippled noises. Infinitely iterated rippled noise (IIRN) is generated when wideband noise is delayed, attenuated, and added to the original wideband noise through positive (+) or negative (−) feedback. In human listeners, the pitch of IIRN(+) is matched to the reciprocal of the delay, whereas the pitch of IIRN(−) is an octave lower. A stimulus generalization paradigm was used to characterize IIRN repetition pitch in chinchillas. Chinchillas were trained to discriminate IIRN(+) with a 4-ms delay from IIRN(+) with a 2-ms delay. In the generalization task, chinchillas were tested with IIRN(+) having delays between 2–4 ms. The pitches evoked by these IIRNs ranged from 500–250 Hz. The delayed noise attenuation for all IIRNs was fixed at −1 dB. A systematic gradient in behavioral response occurred along the dimension of delay for each animal tested, suggesting that a perceptual dimension of pitch exists. Responses to IIRN(−) evoking pitches between 500–250 Hz (delays of 1–2 ms) were also measured. Responses to IIRN(−) were more variable among animals, suggesting that other perceptual cues such as timbre differences may be stronger than the pitch cues. [Work supported by NIDCD R01 DC005596.]

Do goldfish miss the fundamental?

The perception of harmonic complexes was studied in goldfish using classical respiratory conditioning and a stimulus generalization paradigm. Groups of animals were initially conditioned to several harmonic complexes with a fundamental frequency (f0) of 100 Hz. ln some cases the f0 component was present, and in other cases, the f0 component was absent. After conditioning, animals were tested for generalization to novel harmonic complexes having different f0’s, some with f0 present and some with f0 absent. Generalization gradients always peaked at 100 Hz, indicating that the pitch value of the conditioning complexes was consistent with the f0, whether or not f0 was present in the conditioning or test complexes. Thus, goldfish do not miss the fundmental with respect to a pitch-like perceptual dimension. However, generalization gradients tended to have different skirt slopes for the f0-present and f0-absent conditioning and test stimuli. This suggests that goldfish distinguish between f0 present/absent stimuli, probably on the basis of a timbre-like perceptual dimension. These and other results demonstrate that goldfish respond to complex sounds as if they possessed perceptual dimensions similar to pitch and timbre as defined for human and other vertebrate listeners. [Work supported by NIH/NIDCD.]

Perception of the periodicity strength of complex sounds by the chinchilla

The perception of periodicity strength was studied in chinchillas using a stimulus generalization paradigm in an operant-conditioning, positive reinforcement behavioral task. Stimuli consisted of cosine-phase and random-phase harmonic complex tones, infinitely iterated rippled noises, and wideband noise. These stimuli vary in periodicity strength as measured by autocorrelation functions and are known to generate a continuum in the perception of pitch strength in human listeners. Chinchillas were trained to discriminate a cosine-phase harmonic tone complex from wideband noise and tested in the generalization paradigm using random-phase tone complexes and iterated rippled noises as probe stimuli. Chinchillas were tested in three different conditions in which the periods of the fundamental frequencies of the tone complexes were fixed at 2 ms, 4 ms, or 8 ms. Behavioral responses obtained from chinchillas were related to stimulus periodicity strength. For most animals, the behavioral responses to random-phase tone complexes were smaller than those to cosine-phase tone complexes. The behavioral responses were analyzed in terms of the Auditory Image Model of Patterson et al. [Patterson, R.D., Allerhand, M.H., Giguère, C., J. Acoust. Soc. Am. 98 (1995) 1890–1894], and the results suggest that the periodicity information in the stimulus envelope has a large influence in controlling the behavioral response of the chinchilla. Comparison of the generalization data obtained in the present study to magnitude estimation data obtained previously in human subjects suggests a greater influence of stimulus envelope for the perception of periodicity strength in chinchillas than for the perception of pitch strength in human listeners.

Generalization and Transfer: An Interrelation of Paradigms and a Taxonomy of Knowledge Extension Processes

This article integrates work on generalization and transfer into a coherent framework. It analyzes the evolutionary problem with which generalization deals and then outlines a model of responses to a situation requiring generalization and a taxonomy of generalization and transfer processes. Key tenets are that many different generalization processes exist, that there are wide individual differences in which processes may occur, and that generalization of declarative knowledge usually involves concepts. Three experiments tested 2 tenets. One experiment suggested that generalization gradients found in the specialized paradigm used to study human stimulus generalization simply represent failure to perceptually discriminate between stimuli. The other experiments, involving a different paradigm, found step functions along dimensions instead of decremental gradients. They show that the traditional stimulus generalization paradigm is a type of concept learning paradigm. Participants generalize along a continuum by placing stimuli into categories. The experiments also show many different responses to a situation requiring generalization.

Complex sound source determination: Behavioral studies on goldfish

A series of experiments is reported on complex sound perception in goldfish using classical respiratory conditioning and a stimulus generalization paradigm. In general, animals are initially conditioned to a simple or complex sound, and then tested for generalization to novel sounds that differ systematically from the conditioning sound along one or more stimulus dimensions. Generalization gradients to the novel sound set indicate what sounds are perceived as similar or dissimilar, and what features of sounds are analyzed or attended to during conditioning and generalization testing. So far, these kinds of experiments have demonstrated perceptual dimensions in goldfish that resemble the dimensions of pure tone pitch, complex pitch, timbre, and roughness as perceived by humans. In addition, several experiments have revealed auditory stream segregation in goldfish that resembles that demonstrated for human listeners. It is concluded that goldfish share with humans a sense of hearing that is primitive for vertebrates, and is likely shared with most other vertebrate species. At a qualitative level, at least, goldfish know what we know about sounds and their sources. [Work supported by the NIH, NIDCD.]

Perception of two-tone complexes by the goldfish ( Carassius auratus)

Previous experiments on the sense of hearing in goldfish have used a stimulus generalization paradigm to investigate the perceptual dimensions evoked by spectrally and temporally complex sounds. The present experiments investigated the effects on perception of the frequency separation between two tones. In the first set of experiments, six groups of goldfish were classically conditioned to a single tone and then tested for generalization to two-tone complexes having one frequency component equal to the conditioning tone, and the other differing by 2–256 Hz. Generalization declined with increasing frequency differences up to about 32 Hz, and then increased for wider frequency separations. These functions indicate that a restricted range of beat rates produces a perceptual quality that is quite unlike that of a single tone. The generalization function of frequency separation resembles the inverse of the `fluctuation strength' and `roughness' functions for human listeners. The second experiment investigated the effects of spectral location on the perception of a 32 Hz beat rate. Goldfish were conditioned to a two-tone complex (500 and 532 Hz) and then tested for generalization to single tones at various frequencies between 200 and 1200 Hz, and to two-tone complexes having a 32 Hz beat rate but with the lower tone component at various frequencies. For single-tone stimuli, generalization was relatively weak but showed a peak at 500 Hz. For the two-tone stimuli, generalization was more robust, but showed a similarly shaped gradient centered on 500 Hz. Thus, goldfish behaved as if they had acquired information about both temporal modulation and the frequency location of the tone components. These perceptual behaviors appear to be shared with humans and other vertebrates.

Perception of two-tone mixtures by the goldfish

A stimulus generalization paradigm combined with classical respiratory conditioning was used to investigate the perceptual similarities and dissimilarities between single tones and two-tone complexes. Groups of 8 animals were given 40 classical conditioning trials to 6-s pure tones (200, 500, or 700 Hz) and then tested for generalization to two-tone complexes made up of the conditioning tone and a second tone spaced plus or minus 2, 4, 8, 16, 32, 64, 128, 256, 512 Hz from the conditioning tone frequency. Generalization (assumed to be monotonic with judged similarity) reached a minimum for complexes having frequency differences of 36 to 64 Hz, and increased to near 100% for the largest frequency differences (256 or 512 Hz). The dip in generalization at 36–64 Hz suggests that beats at these rates result in a salient perceptual quality that could be similar to roughness or intermittency. The rise in generalization for the larger frequency differences suggests either that roughness gives way to smoothness (a presumed quality of the single conditioning tone), or that the goldfish hears out the conditioning tone in the two-tone complex (analytic listening). These hypotheses will be evaluated in future experiments. [Work supported by the NIH, NIDCD.]

Perception of spectrally and temporally complex sounds by the goldfish ( Carassius auratus)

Behavioral studies on complex sound perception in goldfish were carried out in order to help determine what, if any, differences exist between the sense of hearing of fishes and other vertebrates. A stimulus generalization paradigm was used with classical conditioning in three experiments to determine: (1) the perceptual relations between a pure tone and harmonic complexes having a fundamental frequency equal to that of the tone; (2) the combined effects on perception of pulse repetition rate and spectral envelope; and (3) whether goldfish can be shown to identify a complex source when presented simultaneously with another complex source. Experiment 1 showed that the perceptions of tones and harmonic complexes differ profoundly even for the cases in which they have common periodicities and frequency components. Experiment 2 demonstrated that pulse repetition rate and spectral location simultaneously control behavior, and that repetition rate exerts behavioral control independent of spectral location. Experiment 3 indicates that goldfish did not ‘hear out’ or analyze a complex target source within a mixture of complex sources. In general, goldfish appear to be aware of multiple acoustic dimensions of complex sounds, suggesting both pitch-like and timbre-like perceptual dimensions. These results do not permit a qualitative distinction between the sense of hearing of goldfish and that of other vertebrates.

Perception of temporal acoustic patterns by the goldfish ( Carassius auratus)

The perception of temporal acoustic patterns was studied in the goldfish using classical respiratory conditioning in combination with a stimulus generalization paradigm. Stimuli consisted of a bandpass filtered pulse repeated in various periodic and aperiodic temporal patterns. In each of 14 experiments, animals received 40 conditioning trials to a given stimulus pattern and were then tested for generalization to eight novel stimuli differing only in temporal pattern. In experiments 1–5, animals were conditioned to a periodic pulse train with a particular interpulse interval (IPI) and then tested to novel periodic pulse trains with various IPIs. Generalization gradients were substantially symmetric and monotonic with repetition rate, suggesting a perceptual continuum in goldfish that is similar to periodicity pitch or roughness in human listeners. Several additional experiments indicated that the perceptual qualities of simple and complex temporal patterns are not primarily determined by spectral structure or pulse rate, but rather are determined by the distribution of IPIs. A model for the central analysis of IPIs was successful in accounting for the results of experiments in which animals were conditioned to simple, periodic stimuli. However, the model failed when animals were conditioned to more complex stimuli having aperiodic temporal patterns. These experiments demonstrate the potential usefulness of the stimulus generalization paradigm for investigating aspects of complex sound source perception in non-human animals.

Perception of sinusoid ‘‘ramps’’ and ‘‘damps’’ by the goldfish

A stimulus generalization paradigm combined with classical respiratory conditioning investigated the perceptual tonal qualities of repeated, asymmetrically shaped tone bursts (‘‘ramps’’ and ‘‘damps’’) in the goldfish. Ramps are tone bursts with a slow rise and a relatively fast fall envelope (raised cosine-shaped). Damps are ramps played backward. Human listeners report that ramps sound more tonelike than damps, in spite of their identical spectra. Groups of animals were given 40 classical conditioning trials to 6 s trains of either ramps or damps (400-Hz carrier) repeated at 20/s, and then tested for generalization to eight continuous pure tones above, below, and including 400 Hz. A relatively large generalization response in the region of 400 Hz indicates a perceptual similarity between the ramp or damp conditioning stimulus and a sinusoid. Differences in the magnitude of generalization between groups conditioned to ramps or damps indicate that goldfish perceive ramps as slightly more tonelike than damps. Implications for the mechanisms underlying human perception of ramps and damps are discussed. [Work supported by the NIH, NIDCD.]

Temporal pattern perception in the goldfish: Methods and models

How goldfish perceive temporal patterns of acoustic clicks was investigated using a stimulus generalization paradigm with classical respiratory conditioning. Fish were conditioned to suppress respiration to trains of high-pass-filtered clicks with periodic inter click intervals (ICI). Five groups of animals were conditioned using ICIs between 4 and 25 ms, and then tested for response to eight novel ICIs between 4 and 25 ms. Response magnitude to novel ICIs (generalization) is a monotonic decreasing function of the difference between the conditioning and test ICIs. In further experiments, test stimuli consisted of temporally jittered periodic ICIs, and various pseudorandom ICI distributions. A model predicting the magnitude of generalization to these patterns was developed assuming that generalization is proportional to the number of ICIs common to conditioning and testing stimuli, and that the internal representation of ICIs has a variance predicted from discrimination data. Evidence was obtained that ICIs are analyzed, but that ICI representation is not necessarily congruent with that for pure tones of the same period. [Work supported by a Center Grant from the NIH NIDCD.]

Analytic listening by the goldfish

A stimulus generalization paradigm was used with classical respiratory conditioning to study analytic listeting in the goldfish. Animals were first conditioned to suppress respiration upon the presentation of a long-duration complex sound comprised of two sinusoidal components, 166 and 724 Hz. Conditioned animals were then presented with a set of eight novel test tones with frequencies between 95 and 1514 hz, and including 166 and 724 Hz. Response magnitudes were greatest at the frequencies of the components making up the complex to which the animals were initially conditioned. This is a demonstration that the goldfish had acquired independent information about the frequencies of the individual sinusoidal components making up a complex sound, and thus had listened to the complex analytically. To my knowledge, this is the first demonstration of simultaneous frequency analysis and analytic listening by a nonhuman animal, and suggests that this fundamental aspect of human hearing may be a primitive character shared with the fishes and perhaps with all living vertebrates.