Timbre Processing Research Articles

Spectral versus harmonic information for timbre: Pilot and experimental results

This work summarizes results from six experiments on steady-state timbre perception. The primary variable of interest was the physical similarity of tones within each trial: tones were either similar by virtue of their spectral envelope (i.e., they shared resonance characteristics) or their harmonic envelope (i.e., they shared overtone characteristics). These two continua were chosen as candidates for basic perceptual dimensions because (1) they each contain specific types of information about the sound source/event, and (2) a cursory computational analysis suggests that very different processes are needed to extract information about resonances versus overtones from the auditory input. It was reasoned that if different subsystems subserve the processing of timbre along each of these two dimensions, then performance on perceptual tasks would dissociate under appropriate experimental manipulations. These included (1) attentional focus and response type (timbre similarity judgment, pitch direction judgment, or pitch contour judgment), (2) perceptual priming, and (3) auditory masking. Results demonstrate that in general both dimensions of similarity can support perceptual priming, but also show specific dissociations which inform the understanding of steady-state timbre perception. [Portions of this research were supported by NSF Grant No. DBS-9222358 to Jamshed Bharucha.]

Dissociations entre musique et langage après atteinte cérébrale: Un nouveau cas d'amusie sans aphasie.

We present the neuropsychological study of a patient, I.R., who sustained bilateral damage to the temporal lobes and to the right frontal lobe as a result of successive brain surgeries that occurred ten years earlier. The patient is 40 years old and right-handed; she had no special training in music or in language, representing, therefore, the large majority of listeners. Her performance is compared to that of four neurologically intact subjects who are closely matched in terms of education, sex and age. In the present study, we report I.R.'s performance on various tests aiming at assessing her general cognitive functioning with a particular focus on auditory aspects. The results show that, despite extensive damage to her auditory cortex, I.R.'s speech abilities are essentially intact (see Tables 1 and 2). The only impairments that are detected in the language domain are related to a short-term memory deficit, to some abnormal sensitivity to retroactive interference in long-term memory (see Table 3) and to articulation. These difficulties do not, however, affect linguistic communication, which is obviously undisturbed I.R. is not aphasic). Similarly, I.R. does not experience any difficulty in the recognition and memorization of familiar sounds such as animal cries, traffic noises and the like (see Tables 5 and 7). In contrast, I.R. is severely impaired in most musical abilities: She can no longer discriminate nor identify melodies that were once highly familiar to her; she can no longer discriminate nor memorize novel melodies (see Table 4). Her pattern of musical losses is compatible with a basic and severe perceptual deficit that compromises access to and registration in memory systems. The observation that the auditory impairment affects music and spares language and environmental sounds refers to a neuropsychological condition that is known as music agnosia. I.R. represents, to our knowledge, the fourth case of music agnosia available in the literature (Peretz et al., 1994; Griffiths et al., 1997). The existence of such cases suggests that music processing is not mediated by a general-purpose auditory architecture but by specialized cortical subsystems. Not only does I.R. suffer from music agnosia, but she is also impaired in the discrimination and recognition of musical instruments and of human voices (see Table 5). These latter two deficits probably do not result from the music agnosic condition. Rather, they seem to reflect damage to adjacent brain areas that are specialized in timbre processing (see Peretz. et al., 1994, for the relevant discussion). It is also worth mentioning that I.R. appears to be impaired in musical expressive abilities as well: I.R. can no longer sing a single note. Thus, her losses are rather general in the musical domain, hence justifying the classification of her case as amusia. Cases of amusia without aphasia are relatively frequent in the neuropsychological literature. However, all of these reported cases are anecdotal. Thus, in the present study, special focus is given to the measurement and direct comparison of performance in the language and music domain; in both domains, task characteristics and materials were as similar as possible. To this aim, the lyrics and the tune of the same popular song excerpts were used. The musical and the spoken parts were presented separately in a primed familiarity decision task and in a memory recognition task. In both situations, I.R. performs at or close to chance when she has to deal with music, whereas she recognizes easily and performs normally on the spoken material (see Tables 6 and 7). These results clearly argue for the autonomy of music and language in the processing of auditory information.

Laterality effects in the processing of melody and timbre

Laterality for the processing of melody and timbre was investigated in 64 right-handed non-musicians. In one block of dichotic-listening trials, participants listened for a prespecified target melody, and in a second block they listened for a prespecified target instrument. Females were more accurate on the left ear in the melody task (whereas males tended to show no ear advantage), but there were no significant ear differences in the timbre task for either sex. This supports the idea of a complementary sex-based pattern of lateralization, with males more strongly lateralized for verbal stimuli and females more strongly lateralized for non-verbal stimuli. No relation was observed between lambda measures for the two tasks, suggesting that laterality for melody processing is independent of laterality for timbre processing.

Regular interval stimuli: Are higher-order intervals necessary?

Many stimuli that contain regular temporal intervals are perceived as having a pitch equal to the reciprocal of the interval. Autocorrelation is one method of determining the interval in these regular interval stimuli (RIS). Autocorrelation analysis has been used as a ‘‘model’’ for calculating the pitch and pitch strength of RIS. Autocorrelation is an ‘‘all-interval’’ analysis as opposed to the first-order interval analysis that is often used to describe the temporal properties of neural spike trains. Both actual neural responses (from the AVCN of the chinchilla) and those of a simple neural model were measured to determine the order of intervals contained within RIS. In a psychoacoustic experiment, RIS were generated with no first-order intervals or with only first-order intervals, and listeners discriminated between these stimuli and stimuli with no regular intervals. The results of the experiments and the calculations indicate that the pitch and pitch strength of RIS probably depend on only lower-order intervals (no higher than fourth-order intervals). These results will be discussed in terms of theories of pitch and timbre processing. [Work supported by NIDCD, DC-000293.]

Neural processing of musical timbre by musicians, nonmusicians, and musicians possessing absolute pitch

Cognitive event-related potentials (ERPs) were measured during a timbre discrimination task from three subject groups varying in musical experience. The P3 component of the ERP was recorded from musicians with absolute pitch, musicians without absolute pitch, and nonmusicians during a task comprising timbres of varying difficulty. The three-timbre series, all of which consisted of the same pitch, were (1) string instruments in the same family (cello and viola), (2) flutes made of different materials (silver and wood), and (3) instruments of slightly different size (B-flat versus F tubas). The amplitude and latency of the P3 component varied systematically as a function of musical experience and type of timbre discrimination. The difficult timbre task resulted in mean P3 amplitudes which were larger for musicians relative to nonmusicians, however P3 amplitudes were similar for the two additional timbre series. The mean P3 latencies for musicians were shorter when compared to nonmusicians across all three series. In comparison, the AP subjects displayed the shortest mean P3 latencies, but had smaller P3 amplitudes relative to both musicians and nonmusicians. The implications of these findings suggest that perceptual tasks involving one of the fundamental building blocks of music, namely timbre, does elicit differential brain activity from memory or information processing systems from subjects with varying degrees of musical training.

Signal duration required to distinguish the tone chroma, tone height, and timbre of a sound

Current auditory models suggest that pitch is a primitive feature extracted early on and then used to assist processing of timbre information in music and speech. To test this hypothesis, the number of cycles of a sound required to identify the tone chroma (note name), tone height (octave), or timbre (instrument) of a musical note was determined. There were four instruments (brass, flute, harpsichord, and strings), four octaves centered on C1, C2, C3, and C4 and four notes, C, D, E, and F. The 64 stimuli were presented randomly at a variety of durations, and, in a given session, the listener identified the stimulus within one perceptual category: tone chroma, tone height, or timbre. The results show that performance rises with duration to an asymptotic level. Performance rises fastest to the highest asymptote in the case of instrument identification, and slowest to the lowest asymptote for tone chroma. This suggests that timbre processing can proceed before detailed information about the pitch is available. [Research supported by UK RAE 2239.]

Processing of Timbre and Rhythm in Musicians and Non-Musicians

Expert musicians and non-musicians of similar educational and social class background were compared in two experiments involving perception of timbre and rhythm. In Experiment 1 where dichotic monitoring for the sound of the violin was required, there was a practice effect but no ear or group differences. The rhythm monitoring experiment produced a group by ear interaction with musicians faster on the right ear than the left and faster than non-musicians on the right ear only. Analysis of strategies reported by subjects showed that verbal labelling did not apparently influence laterality. Lack of evidence for individual laterality effects reinforces the claim that with stringent experimental and subject controls there is minimal evidence for musicians non-musician laterality effects.