Temporal Gap Detection Thresholds Research Articles

Large cross-sectional study of presbycusis reveals rapid progressive decline in auditory temporal acuity

The auditory system relies on extraordinarily precise timing cues for the accurate perception of speech, music, and object identification. Epidemiological research has documented the age-related progressive decline in hearing sensitivity that is known to be a major health concern for the elderly. Although smaller investigations indicate that auditory temporal processing also declines with age, such measures have not been included in larger studies. Temporal gap detection thresholds (TGDTs; an index of auditory temporal resolution) measured in 1071 listeners (aged 18–98 years) were shown to decline at a minimum rate of 1.05 ms (15%) per decade. Age was a significant predictor of TGDT when controlling for audibility (partial correlation) and when restricting analyses to persons with normal-hearing sensitivity (n = 434). The TGDTs were significantly better for males (3.5 ms; 51%) than females when averaged across the life span. These results highlight the need for indices of temporal processing in diagnostics, as treatment targets, and as factors in models of aging.

Effects of Advancing Age and Hearing Loss on Gaps-in-Noise Test Performance

The present study evaluated the effects of advancing age and sensorineural hearing loss (SNHL) on approximate temporal gap detection thresholds obtained from the Gaps-in-Noise (GIN; Musiek et al., 2005) test. Approximate thresholds from the GIN test were compared among 3 groups of participants (older adults with SNHL, older adults with essentially normal hearing, and young adults with normal hearing). Analyses of variance and regression modeling were used to determine the effect of age and audibility on GIN performance. Comparisons of thresholds obtained from the 4 test lists of the GIN were also conducted. Approximate gap thresholds differed significantly across all groups, with the poorest thresholds found in older adults with SNHL and with the best thresholds found in young adults with normal hearing. The strongest predictor of decreases in GIN performance was increased Audibility Index score, but a significant independent effect of age was also documented. Test list influenced GIN performance only in the group of young adults with normal hearing. The GIN test is a promising clinical measure of temporal resolution. Findings from the present study suggest that age and audibility influence GIN performance and may affect test analysis and interpretation with published normative data.

Behaviorally measured audiograms and gap detection thresholds in CBA/CaJ mice

Tone detection and temporal gap detection thresholds were determined in CBA/CaJ mice using a Go/No-go procedure and the psychophysical method of constant stimuli. In the first experiment, audiograms were constructed for five CBA/CaJ mice. Thresholds were obtained for eight pure tones ranging in frequency from 1 to 42 kHz. Audiograms showed peak sensitivity between 8 and 24 kHz, with higher thresholds at lower and higher frequencies. In the second experiment, thresholds for gap detection in broadband and narrowband noise bursts were measured at several sensation levels. For broadband noise, gap thresholds were between 1 and 2 ms, except at very low sensation levels, where thresholds increased significantly. Gap thresholds also increased significantly for low pass-filtered noise bursts with a cutoff frequency below 18 kHz. Our experiments revised absolute auditory thresholds in the CBA/CaJ mouse strain and demonstrated excellent gap detection ability in the mouse. These results add to the baseline behavioral data from normal-hearing mice which have become increasingly important for assessing auditory abilities in genetically altered mice.

Ear and contralateral masker effects on auditory temporal gap detection thresholds

A temporal processing advantage is thought to underlie the left hemisphere dominance for language. One measure of a temporal processing advantage is temporal acuity or resolution. A standard paradigm for measuring auditory temporal resolution is gap detection in its “within-channel” and “between-channel” forms. Previous experiments investigating a right ear advantage for within-channel gap detection have yielded conflicting results, and between-channel gap detection has not previously been studied for ear differences. In the present study, the two types of gap detection task were employed, under each of three contralateral masking conditions (no noise, continuous noise and interrupted noise). An adaptive tracking procedure was used to measure the minimal detectable gap at each ear (and therefore, the temporal acuity of the contralateral hemisphere). A significant effect of masking noise was observed in both of the gap detection tasks. Within-channel gap threshold durations were longer in the interrupted noise condition for both ears. Between-channel gap threshold durations were shorter in the interrupted noise condition at the left ear, with a trend in the same direction at the right ear. The study found no significant difference between the ears in thresholds in either gap detection task in any of the masking conditions. This suggests that if the left cerebral hemisphere has a temporal processing advantage, then it is not in the form of acuity for temporal gap detection.

Gap detection in Mongolian gerbils ( Meriones unguiculatus)

Hearing thresholds for white-noise stimuli and temporal gap-detection thresholds in six Mongolian gerbils ( Meriones unguiculatus) were determined in a GO/NOGO procedure using the method of constant stimuli. Gerbils were offspring of animals collected in the field and only bred in captivity for four generations or less. Hearing thresholds for 800 ms bursts of white noise ranged between −0.8 and 6.3 dB SPL. The median minimum-detectable gap centered in an 800 ms burst of white noise presented at 50 dB SPL was 2.1 ms. At levels of 40 dB SPL and above, gap-detection thresholds were independent of the sound-pressure level of the stimulus. At 30 dB SPL and below, the minimum-detectable gap increased with decreasing sound-pressure level. Near 5 dB sensation level, gap-detection thresholds ranged from 13.6 to 29.4 ms. The median threshold for the detection of gaps inserted 5 ms after the onset of an 800 ms burst of white noise of 50 dB SPL was 2.3 ms. The individuals’ detection thresholds varied between 2.3 and 3.6 ms for stimuli in which the relative sound-pressure levels of the noise before and after the gap differed by up to 20 dB. The data found in the Mongolian gerbil match results from other mammal and bird species.

Auditory temporal gap detection for noise markers with partially overlapping and non-overlapping spectra

Temporal gap detection thresholds were obtained from six listeners using an adaptive tracking method and constant spectrum-level noises. In separate blocks of trials, the markers bounding the gap were systematically varied in their spectral overlap or separation (expressed in equivalent rectangular bandwidths, ERBs). In the same listeners, gap thresholds were also obtained for noises of the same bandwidths as those constituting the overlap in the overlap conditions (in the presence of a wideband notched noise masker: ‘mask’ conditions). For the spectral overlap/separation conditions, gap thresholds were a systematic, linear function of spectral dissimilarity in four of six listeners. In the mask conditions, gap thresholds were inversely related to bandwidth in all listeners. For the three-, four- and five-ERB conditions, gap thresholds in the same listeners for the spectral overlap conditions were higher than those for mask stimuli with the same available within-channel bandwidth and spectrum levels. These data suggest that the spectral dissimilarity between the markers over-rode the availability of within-channel information in the recovery of the temporal gap.

Spectral integration of auditory temporal information

Measures of temporal acuity reveal a remarkable ability to integrate temporal information across frequency throughout most of the audible spectrum. This apparent trading of spectral bandwidth and temporal resolution was investigated in three different experiments measuring amplitude-modulation detection, temporal gap detection, and intensity discrimination. Each experiment included six spectral conditions: a single band of noise 1600-Hz wide centered at 3000 Hz; a set of four noise bands 400-Hz wide having center frequencies separated by 1200 Hz (1200, 2400, 3600, 4800 Hz); and four conditions composed of only one of the four 400-Hz bands. The form of the modulation transfer functions did not depend upon total bandwidth or frequency region. Sensitivity to modulation was inversely related to total noise bandwidth. Temporal gap detection thresholds and intensity discrimination were strongly dependent upon total noise bandwidth but were independent of the spectral proximity and the center frequency of the noise bands. Analysis of the three experiments revealed that the inverse relation between intensity discrimination and noise bandwidth may be used to predict the variation in amplitude modulation and temporal gap detection thresholds with noise bandwidth.

Temporal gap detection measured with multiple sinusoidal markers: effects of marker number, frequency, and temporal position.

Detection thresholds were measured for silent temporal gaps within combinations of two, three, or four sinusoidal markers (i.e., combinations of one or two pre-gap markers with one or two post-gap markers). The markers were selected from the frequency range 2000-3100 Hz. Sinusoidal frequencies F1 and F4 were used as pre-gap markers, while F2 and F3 served as post-gap markers. Temporal gap detection (TGD) thresholds were measured from sets of three normal-hearing adults who tracked 70.7% correct detection thresholds adaptively across blocks of 50 two-interval, two-alternative, forced-choice trials. For symmetric marker conditions, where pre- and post-gap markers were equivalent in frequency (e.g., F1 = F2 or F1 = F2 and F3 = F4), TGD thresholds were < 10 ms. However, for asymmetric marker frequency alignments across the silent gap, including stimulus configurations where only three markers were presented on a trial (e.g., F1 = F2, F2 not equal to F3, no F4), performance was highly variable and was dramatically disrupted by the presentation of a second post-gap marker. The multiple-marker results reveal that TGD depends greatly on the number of markers presented, both in terms of the marker temporal position before and after the silent gap signal and the marker frequency alignment (symmetry) across the gap. These results, which cannot be predicted from models of the auditory periphery, may reflect perceptual mechanisms that are important in grouping and organizing auditory images.

Evidence for an across-frequency, between-channel process in asymptotic monaural temporal gap detection.

Monaurally measured temporal gap detection (TGD) thresholds characteristically increase as the frequency difference is increased over a range of about half an octave to an octave between two sinusoids that mark the onset and offset of the silent gap. For greater sinusoidal frequency separations, the TGD thresholds often become asymptotic. This pattern probably reflects two different processes. The first process likely reflects within-channel processing within a single auditory filter or channel. The second process is less certain, but may reflect between-channel processing of the silent gap stimulus across two or more independent frequency channels. To evaluate the hypothesis that asymptotic monaural gap detection can be explained by a simple between-channel process, TGD thresholds were measured as a function of frequency separation between a pregap sinusoid presented to the left ear (channel 1) and a postgap sinusoid, of higher frequency, presented to the right ear (channel 2). The rationale for dichotic presentation of the sinusoidal markers and gap signal followed from the fact that the gap detection task must be performed between two independent channels by combining the outputs from each channel (ear) and recovering the gap information centrally. The resulting TGD thresholds for pregap sinusoids from 250 to 4000 Hz were relatively invariant and increased only slightly with increasing marker frequency separation. The average TGD thresholds for four listeners were in the range of 30 to 40 ms, which corresponded closely with their asymptotic TGD thresholds for the same set of stimulus conditions measured monaurally. This correspondence of the two data sets supports an across-frequency, between-channel process for asymptotic monaural gap detection at marker frequency separations greater than about half an octave.

Within-channel gap detection using dissimilar markers in cochlear implant listeners.

Temporal gap detection thresholds were measured between perceptually dissimilar electrical markers in cochlear implant listeners. Both markers were presented to the same electrode pair. The amplitude and pulse rate of the first marker were fixed, and gap thresholds were measured as a function of either the pulse rate or the amplitude of the second marker. In either case, U-shaped functions were obtained, with lowest gap thresholds occurring when the two markers were similar in both amplitude and pulse rate. Because the two markers were presented to the same electrode pair, the data cannot be accounted for on the basis of across-channel interactions. It is hypothesized that when different markers are used, the perceptual discontinuity from the first marker to the second is similar to the sensation of a brief gap, and dominates the gap detection process. Thus, gap threshold functions with electrically dissimilar markers serve more as indicators of perceptual distance between the markers and less as measures of temporal resolution.

Temporal gap detection thresholds measured for conditions that minimize off-frequency listening

Temporal gap detection (TGD) thresholds measured for silent gaps as a function of the frequency separation between a fixed-frequency pre-gap sinusoidal marker (F1) and a variable-frequency post-gap sinusoidal marker (F2) may be confounded by off-frequency listening (Formby et al., Aud. Neurosci. 3, 1–20). Evidence of this confounding role comes from simulation experiments with a single-channel envelope detector model. TGD stimulation results suggest that listeners can improve the signal-to-noise ratio and, in turn, TGD performance by adjusting their optimal auditory filter so that it is centered between F1 and F2 (Forrest and Formby, Aud. Neurosci. 3, 21–33). Off-frequency listening between F1 and F2 may result in underestimates of TGD thresholds that appear as asymmetries in detection performance for F2 markers presented below and above the F1 frequency. To obtain more precise estimates of TGD, unaffected by off-frequency listening, performance has been measured and will be reported for three listeners with markers spaced symmetrically and asymmetrically in linear frequency below and above the center frequencies Fc=500 and 4000 Hz. These new results will be compared with an otherwise comparable set of results that were likely confounded by off-frequency listening (Formby et al., Aud. Neurosci. 3, 1–20). [Research supported by NIH.]

Modified masking period patterns as a function of frequency region and masker bandwidth

The auditory system must encode dynamic aspects of acoustic stimuli at frequencies throughout the audible spectrum. This coding is limited by the temporal resolution of the ear. Recent measures of modulation detection and temporal gap detection thresholds with narrow-band noises of fixed bandwidth indicate that temporal resolution does not vary with noise carrier frequency from 500 to 4400 Hz. The present study investigated temporal processing using a modified masking period pattern technique. Masked thresholds were measured for pure tones of 500, 2000, and 4000 Hz that were 400 ms in duration. Maskers consisted of narrow-band noises centered on the signal frequency that were either unmodulated or sinusoidally amplitude modulated (modulation depth=1.0) at frequencies from 4 to 256 Hz. The noise bandwidth ranged from 200 to 1600 Hz at each center frequency. By comparing masked thresholds for the unmodulated and amplitude modulated noises, one may assess both within-channel (narrowband noises) and across-channel (wideband noises) temporal processing. Preliminary results indicate that thresholds for a given masker bandwidth depend strongly on modulation frequency but vary little with signal frequency over the range tested. The data will be discussed relative to measures of temporal resolution and comodulation masking release.

The effects of sinusoidal amplitude modulation on gap detection in noise

The purpose of this study was to investigate the effects of envelope fluctuation on temporal gap detection. Temporal gap detection thresholds were measured in wideband noisebursts that had been amplitude modulated by a very low-frequency sinusoid. Detection was studied at six modulation depths and in four phase conditions which placed the onset of the silent gap at different amplitudes on the temporal envelope of the waveform. Gap thresholds were obtained in a 3IFC paradigm in four young, normal-hearing subjects. The smallest mean gap threshold, 3.5 ms, was obtained for unmodulated noisebursts. The largest mean gap threshold, 25.7 ms, was obtained at a modulation depth of 100%. Phase did not significantly affect gap detection until modulation depth was increased to 40%. At deeper depths of modulation, gap thresholds were smallest, but also most variable, when the gap onset was placed at a peak in the waveform envelope and largest when the gaps were centered on an envelope minimum or randomly placed. Results of this study are consistent with previous findings and the hypothesis that gap detection is dependent on the similarity of the gap itself to inherent fluctuations in the stimulus envelope, reflecting coherence in terms of phase and depth of envelope fluctuation.

Gap detection thresholds measured for symmetric and asymmetric marker frequency combinations surrounding a silent temporal gap

Temporal gap detection (TGD) thresholds were measured between sinusoids that varied in frequency from 2000 to 3100 Hz for combinations of 2, 3, or 4 markers (e.g., combinations of one or two pre‐gap markers with one or two post‐gap markers). Sinusoidal frequencies F1 and F4 were used as pre‐gap markers, while F2 and F3 served as post‐gap markers. TGD thresholds were measured from three normal‐hearing adults who tracked 70.7% correct detection thresholds adaptively across blocks of 50 2AFC trials. For symmetric marker conditions, where a pair or pairs of pre‐ and post‐gap markers are equivalent in frequency on both sides of the silent gap (e.g., F1=F2 or F1=F2 and F3=F4), TGD thresholds are small (i.e., <10 ms). However, for asymmetric marker conditions, where only three markers are presented on a trial (e.g., F1=F2, F2≠F3, no F4), performance is highly variable. TGD thresholds for asymmetric marker conditions may be very large (e.g., 20–80 ms), and tend to increase as the frequency separation is increased...

An asymmetric roex filter model for describing detection of silent temporal gaps as a function of marker frequency separation

Formby and Forrest [J. Acoust. Soc. Am. 89, 830–837 (1991)] measured temporal gap detection (TGD) thresholds as function of frequency separation between two sinusoids marking a silent temporal gap. Their results were fit well with a roex-filter model. Filter estimates (1) were based on TGD thresholds for post-gap marker frequencies F2≥ pre-gap marker frequencies F1 and (2) assumed symmetry of the roex-filter function for F2≤F1. In a related study, TGD thresholds were measured with narrow-band markers [Formby etal., J. Acoust. Soc. Am. 93, 1023–1027 (1993)]. The resulting thresholds raised doubts about the earlier assumption of filter symmetry. To resolve this issue, TGD thresholds were measured for F2 frequencies presented above and below F1=500 and 4000 Hz for a range of marker and masker presentation levels. Roex functions fitted for conditions of F2≥F1 were typically shallower than those fitted for conditions of F2≤F1. The same qualitative patterns of filter asymmetry have also been found and will be reported for TGD thresholds simulated by single- and multi-channel auditory models [Forrest and Formby; Heinz etal., companion papers at this meeting]. [Research supported by NIH.]

Temporal gap detection thresholds in sinusoidal markers simulated with a single-channel envelope detector model

The temporal envelope detector [N. Viemeister, J. Acoust. Soc. Am. 66, 1364–1380 (1979)] is a good single-channel model for auditory temporal processing. The model has three stages: (1) an initial bandpass filter, followed serially by (2) a nonlinear element, and (3) a lowpass filter. A digital version of the model using a max–min decision rule [Forrest and Green, J. Acoust. Soc. Am. 82, 1933–1943 (1987)] at the model’s output has been used to simulate temporal gap detection (TGD) thresholds for silent temporal gaps in pairs of sinusoidal markers (F1 and F2). Simulated TGD thresholds were measured as a function of marker frequency separation in the same manner as described in psychophysical experiments [Formby etal., companion paper at this meeting]. Simulated TGD thresholds depended directly on marker frequency separation. For a given frequency separation, simulated TGD thresholds were minimized by centering the symmetrical, first-stage roex filter between the two marker frequencies. This single-channel model consistently yielded lower TGD thresholds than did human listeners, but qualitatively simulated an asymmetry in the psychophysical thresholds where the relative change in the TGD threshold was smaller when the frequencies of F2≥F1 and larger for F1≥F2. [Research supported by NIH.]

Temporal gap detection thresholds in sinusoidal markers simulated with a multi-channel, multi-resolution cochlear model

Computer simulations were implemented for estimating detection thresholds for silent temporal gaps in sinusoidal markers that differed in frequency. The stimuli were presented in blocks of 50. 2AFC trials in the same adaptive paradigm used to measure human temporal gap detection (TGD) thresholds [Formby etal., companion paper at this meeting]. The stimuli were preprocessed by a multi-channel, multi-resolution cochlear model. The digital simulation of the cochlear model was implemented by a 30-channel basilar membrane filter bank, followed by 30 hair cell/synapse sections [Liu, Ph.D. dissertation, Johns Hopkins Univ. (1992)]. The output for each channel of the cochlear model was lowpass filtered, and a max–min ratio [Forrest and Green, J. Acoust. Soc. Am. 82, 1933–1943 (1987)] was estimated for each of the 30 channels. A majority vote across the 30 max–min ratio values on each trial was used to select the observation interval containing the silent gap. Results from this majority voting scheme were qualitatively similar to, but consistently poorer than, human psychoacoustic TGD thresholds [Formby etal., companion paper at this meeting]. Other decision rules are being investigated and will be presented for simulating human thresholds. [Research supported by NIH.]

Detection of silent temporal gaps between narrow-band noise makers having second-formantlike properties of voiceless stop/vowel combinations.

Temporal gap detection thresholds were measured in narrow-band noise-burst markers having acoustic characteristics representative of isolated steady-state second-formant (F2) properties for/p,t,k/paired separately with/i,ae,u,o/. The results revealed that gap detection threshold increased systematically as the difference was increased between the simulated stop and vowel F2 frequencies. A strong positive correlation (r = 0.87) between gap detection threshold and linear marker center frequency difference was highly significant (p < 0.001). Differences in other stimulus features had little influence on gap detection performance. Implications for speech perception are discussed.

Detection of silent temporal gaps in narrow‐band noise stimuli having second‐formantlike properties of voiceless stop/vowel combinations

Delattre et al. [J. Acoust. Soc. Am. 27, 769–773 (1955)] reported that good‐quality stop consonant/vowel pairs could be synthesized in two‐formant representation by replacing the initial part of the second‐formant (F2) transition with a silent temporal gap. The best silent gap for each stop varied with the timing and extent of the F2 transition, which we know is (1) dependent on coarticulation between the stop and vowel and (2) represented acoustically by the difference or change in frequency from the stop burst to F2 of the following vowel. Motivated by their report, temporal gap detection thresholds were measured in narrow‐band noise‐burst markers having acoustic characteristics that simulated isolated steady‐state F2 properties of voiceless stop /p, t, k/ and vowel /i, æ, u, ɔ/ combinations. Gap detection thresholds increased systematically as a function of the difference between the simulated stop and vowel F2 frequencies, which accounted for 76% of the variance in the results. Differences in other stimulus features had little influence on detection. [Research supported by NIH.]

Detection of gaps in noise by budgerigars ( Melopsittacus undulatus) and zebra finches ( Poephila guttata)

Temporal gap detection thresholds were obtained for two species of birds, budgerigars and zebra finches, which are known to have different auditory filter bandwidths. Both species showed gap detection thresholds of about 2.5 msec for broadband noise stimuli. Comparing octave bands of noise centered at 1, 3 and 5 kHz, zebra finches showed the smallest gap thresholds for the noise band centered at 5 kHz whereas budgerigars showed the smallest gap detection thresholds for the noise band centered at 3 kHz. The results from zebra finches are generally consistent with filter theories of auditory spectro-temporal perception whereas the result from budgerigars are not. In aggregate, these comparative data suggest the relation between spectral and temporal resolving power in these two species may involve different mechanisms.