Predicting the detectability of tones with unexpected durations

Steep temporal integration for tone detection in a multi-tone, random-frequency masker

In the detection of a tone burst, masking by several tones with random frequencies can produce steep temporal integration. This feature was evaluated for nine normal-hearing adults for 1000-Hz tone bursts presented in a continuous train of four-tone masker bursts. Masker frequencies were randomly selected (250-4000 Hz) for each burst, with the proviso that all tones were separated by ≥0.2 oct. Bursts were 80-ms in duration; when present, signal bursts were gated synchronously with masker bursts. The observed mean temporal-integration function was exceptionally steep-thresholds improved by 26 dB as signal duration increased from 1 to 8 bursts. The results also showed that the individual differences were large, and that the mean psychometric function was exceptionally shallow, spanning a range of 35 dB between 0.6 and 0.9 proportion correct responses, consistent with previous reports. These findings were interpreted in the context of three signal-detection models, one based on the absolute-level cue, and two based on the relative-level cue via template matching; all cues were derived from the excitation patterns of the stimuli. Template-matching models were able to predict the shallow psychometric functions as observed, but all models fall short in the steepness of the observed temporal integration.

How to define thresholds for level and interaural-level-difference discrimination: Insights from scedasticities and distributions

Sensitivity to changes in the stimulus level at one or at both ears and to changes in the interaural level difference (ILD) between the two ears has been studied widely. Several different definitions of threshold and, for one of them, two different ways of averaging single-listener thresholds have been used (i.e., arithmetically and geometrically), but it is unclear which definition and which way of averaging is most suitable. Here, we addressed this issue by examining which of the differently defined thresholds yielded the highest degree of homoscedasticity (homogeneity of the variance). We also examined how closely the differently defined thresholds followed the normal distribution. We measured thresholds from a large number of human listeners as a function of stimulus duration in six experimental conditions, using an adaptive two-alternative forced-choice paradigm. Thresholds defined as the logarithm of the ratio of the intensities or amplitudes of the target and the reference stimulus (i.e., as the difference in their levels or ILDs; the most commonly used definition) were clearly heteroscedastic. Log-transformation of these latter thresholds, as sometimes performed, did not result in homoscedasticity. Thresholds defined as the logarithm of the Weber fraction for stimulus intensity and thresholds defined as the logarithm of the Weber fraction for stimulus amplitude (the most rarely used definition) were consistent with homoscedasticity, but the latter were closer to the ideal case. Thresholds defined as the logarithm of the Weber fraction for stimulus amplitude also followed the normal distribution most closely. The discrimination thresholds should therefore be expressed as the logarithm of the Weber fraction for stimulus amplitude and be averaged arithmetically across listeners. Other implications are discussed, and the obtained differences between the thresholds in different conditions are compared to the literature.

Towards a unifying basis of auditory thresholds: Thresholds for multicomponent stimuli

Sounds consisting of multiple simultaneous or consecutive components can be detected by listeners when the stimulus levels of the components are lower than those needed to detect the individual components alone. The mechanisms underlying such spectral, spectrotemporal, temporal, or across-ear integration are not completely understood. Here, we report threshold measurements from human subjects for multicomponent stimuli (tone complexes, tone sequences, diotic or dichotic tones) and for their individual sinusoidal components in quiet. We examine whether the data are compatible with the detection model developed by Heil, Matysiak, and Neubauer (HMN model) to account for temporal integration (Heil et al. 2017), and we compare its performance to that of the statistical summation model (Green 1958), the model commonly used to account for spectral and spectrotemporal integration. In addition, we compare the performance of both models with respect to previously published thresholds for sequences of identical tones and for diotic tones. The HMN model is similar to the statistical summation model but is based on the assumption that the decision variable is a number of sensory events generated by the components via independent Poisson point processes. The rate of events is low without stimulation and increases with stimulation. The increase is proportional to the time-varying amplitude envelope of the bandpass-filtered component(s) raised to an exponent of 3. For an ideal observer, the decision variable is the sum of the events from all channels carrying information, for as long as they carry information. We find that the HMN model provides a better account of the thresholds for multicomponent stimuli than the statistical summation model, and it offers a unifying account of spectral, spectrotemporal, temporal, and across-ear integration at threshold.

Humans attend to signal duration but not temporal structure for sound detection: Steady-state versus pulse-train signals

Most sounds fluctuate in amplitude, but do listeners attend to the temporal structure of those fluctuations when trying to detect the mere presence of those sounds? This question was addressed by leading listeners to expect a faint sound with a fixed temporal structure (pulse train or steady-state tone) and total duration (300 ms) and measuring their ability to detect equally faint sounds of unexpected temporal structure (pulse train when expecting steady state) and/or total duration (<300 ms). Detection was poorer for sounds with unexpected than with expected total durations, replicating previous outcomes, but was uninfluenced by the temporal structure of the expected sound. The results disagree with computational predictions of the multiple-look model, which posits that listeners attend to both the total duration and temporal structure of the signal, but agree with predictions of the matched-window energy-detector model, which posits that listeners attend to the total duration but not the temporal structure of the signal. Moreover, the matched-window energy-detector model could also account for previous results, including some that were originally interpreted as supporting the multiple-look model. Taken together, at least when detecting faint sounds, listeners appear to attend to the total duration of expected sounds but to ignore their detailed temporal structure.

Comparing and modeling absolute auditory thresholds in an alternative-forced-choice and a yes-no procedure

Detecting sounds in quiet is arguably the simplest task performed by an auditory system, but the underlying mechanisms are still a matter of debate. Threshold stimulus levels depend not only on the physical properties of the sounds to be detected but also on the experimental procedure used to measure them. Here, thresholds of human subjects were measured for sounds consisting of different numbers of bursts using both an alternative-forced-choice and a yes-no procedure in the same experimental sessions. Thresholds measured with the yes-no procedure were typically higher than thresholds measured with the alternative-forced choice procedure. The difference between the two thresholds decreased as stimulus duration increased. It also varied between subjects and varied with the probability of false alarms in the yes-no procedure. It is shown that a previously proposed model of detection (Heil et al., Hear Res 2017) can account for these findings better than other models. It can also account for the shapes of the psychometric functions. The model is consistent with basic concepts of signal detection theory but is based on a decision variable that follows Poisson statistics. It also differs from other models of detection with respect to the transformation of the stimulus into the decision variable. The findings in this study further support the model.

Effect of echolocation behavior-related constant frequency-frequency modulation sound on the frequency tuning of inferior collicular neurons in Hipposideros armiger

In constant frequency-frequency modulation (CF-FM) bats, the CF-FM echolocation signals include both CF and FM components, yet the role of such complex acoustic signals in frequency resolution by bats remains unknown. Using CF and CF-FM echolocation signals as acoustic stimuli, the responses of inferior collicular (IC) neurons of Hipposideros armiger were obtained by extracellular recordings. We tested the effect of preceding CF or CF-FM sounds on the shape of the frequency tuning curves (FTCs) of IC neurons. Results showed that both CF-FM and CF sounds reduced the number of FTCs with tailed lower-frequency-side of IC neurons. However, more IC neurons experienced such conversion after adding CF-FM sound compared with CF sound. We also found that the Q 20 value of the FTC of IC neurons experienced the largest increase with the addition of CF-FM sound. Moreover, only CF-FM sound could cause an increase in the slope of the neurons' FTCs, and such increase occurred mainly in the lower-frequency edge. These results suggested that CF-FM sound could increase the accuracy of frequency analysis of echo and cut-off low-frequency elements from the habitat of bats more than CF sound.

Optimal integration of independent observations from Poisson sources

The optimal integration of information from independent Poisson sources (such as neurons) was analyzed in the context of a two-interval, forced-choice detection task. When the mean count of the Poisson distribution is above 1, the benefit of integration is closely approximated by the predictions based on the square-root law of the Gaussian model. When the mean count falls far below 1, however, the benefit of integration clearly exceeds the predictions based on the square-root law.

Echo frequency selectivity of duration-tuned inferior collicular neurons of the big brown bat, Eptesicus fuscus, determined with pulse-echo pairs

During hunting, insectivorous bats such as Eptesicus fuscus progressively vary the repetition rate, duration, frequency and amplitude of emitted pulses such that analysis of an echo parameter by bats would be inevitably affected by other co-varying echo parameters. The present study is to determine the variation of echo frequency selectivity of duration-tuned inferior collicular neurons during different phases of hunting using pulse-echo (P-E) pairs as stimuli. All collicular neurons discharge maximally to a tone at a particular frequency which is defined as the best frequency (BF). Most collicular neurons also discharge maximally to a BF pulse at a particular duration which is defined as the best duration (BD). A family of echo iso-level frequency tuning curves (iso-level FTC) of these duration-tuned collicular neurons is measured with the number of impulses in response to the echo pulse at selected frequencies when the P-E pairs are presented at varied P-E duration and gap. Our data show that these duration-tuned collicular neurons have narrower echo iso-level FTC when measured with BD than with non-BD echo pulses. Also, IC neurons with low BF and short BD have narrower echo iso-level FTC than IC neurons with high BF and long BD have. The bandwidth of echo iso-level FTC significantly decreases with shortening of P-E duration and P-E gap. These data suggest that duration-tuned collicular neurons not only can facilitate bat's echo recognition but also can enhance echo frequency selectivity for prey feature analysis throughout a target approaching sequence during hunting. These data also support previous behavior studies showing that bats prepare their auditory system to analyze expected returning echoes within a time window to extract target features after pulse emission.

The measurement problem in level discrimination

There is disagreement among theorists over the exact measure to be used to quantify auditory level discrimination. It has been proposed that, for level discrimination tasks, the measure that is most linearly related to the sensitivity index, d', will be the correct measure. The level difference (deltaL) and the Weber fraction (theta) are both candidates, though the latter is sensitive to the physical unit in which it is expressed (e.g., pressure or intensity) while the former is not. Psychometric functions for level discrimination were obtained at a number of pedestal levels for 10-ms sinusoids (either 1000 or 6500 Hz) and broadband noise bursts. These functions were used to assess which of three measures: deltaL, theta = deltap/p, or theta = deltaI/I, is most nearly linearly related to d'. The results suggest that deltap/p is the measure that comes closest to being linearly related to d'.

Comparison of absolute thresholds derived from an adaptive forced-choice procedure and from reaction probabilities and reaction times in a simple reaction time paradigm

An understanding of the auditory system's operation requires knowledge of the mechanisms underlying thresholds. In this work we compare detection thresholds obtained with a three-interval-three-alternative forced-choice paradigm with reaction thresholds extracted from both reaction probabilities (RP) and reaction times (RT) in a simple RT paradigm from the same listeners under otherwise nearly identical experimental conditions. Detection thresholds, RP, and RT to auditory stimuli exhibited substantial variation from session to session. Most of the intersession variation in RP and RT could be accounted for by intersession variation in a listener's absolute sensitivity. The reaction thresholds extracted from RP were very similar, if not identical, to those extracted from RT. On the other hand, reaction thresholds were always higher than detection thresholds. The difference between the two thresholds can be considered as the additional amount of evidence required by each listener to react to a stimulus in an unforced design on top of that necessary for detection in the forced-choice design. This difference is inversely related to the listener's probability of producing false alarms. We found that RT, once corrected for some irreducible minimum RT, reflects the time at which a given stimulus reaches the listener's reaction threshold. This suggests that the relationships between simple RT and loudness (reported in the literature) are probably caused by a tight relationship between temporal summation at threshold and temporal summation of loudness.

Duration selectivity organization in the inferior colliculus of the big brown bat, Eptesicus fuscus

Duration selectivity of auditory neurons plays an important role in sound recognition. Previous studies show that GABA-mediated duration selectivity of neurons in the central nucleus of the inferior colliculus (IC) of many animal species behave as band-, short-, long- and all-pass filters to sound duration. The present study examines the organization of duration selectivity of IC neurons of the big brown bat, Eptesicus fuscus, in relation to graded spatial distribution of GABA(A) receptors, which are mostly distributed in the dorsomedial region of the IC but are sparsely distributed in the ventrolateral region. Duration selectivity of IC neuron is studied before and during iontophoretic application of GABA and its antagonist, bicuculline. Bicuculline application decreases and GABA application increases duration selectivity of IC neurons. Bicuculline application produces more pronounced broadening of the duration tuning curves of neurons at upper IC than at deeper IC but the opposite is observed during GABA application. The best duration of IC neurons progressively lengthens and duration selectivity decreases with recording depth both before and during drug application. As such, low best frequency neurons at upper IC have shorter best duration and sharper duration selectivity than high best frequency neurons in the deeper IC have. These data suggest that duration selectivity of IC neurons systematically varies with GABA(A) receptor distribution gradient within the IC.

Using magnitude estimation to investigate the perceptual components of signal detection theory

Several experiments suggest that the relation between detection, signal energy, and perceived tone intensity is very different from the relation between detection, signal energy, and perceived tone duration. We propose a new task, the magnitude estimation and detection (MED) task, that allows one to assess the relation between the psychological dimensions of a stimulus and detection. In Experiment 1, we used the MED task to assess the relation between perceived tone intensity and detection in a yes/no task. The data show a strong relation between the two. In Experiment 2, we used the MED task to assess the relation between perceived tone duration and detection in a yes/no task. The data show a relatively weak relation between the two. Our data suggest that tone intensity is a less perceptually noisy dimension than tone duration. We present the advantages and disadvantages of the MED task with the hope that this task can aid researchers in better understanding the perceptual and decisional mechanisms underlying various cognitive processes.