Human Auditory Detection and Discrimination Measured with the Pupil Dilation Response

Abstract

In the standard Hughson-Westlake hearing tests (Carhart and Jerger 1959), patient responses like a button press, raised hand, or verbal response are used to assess detection of brief test signals such as tones of varying pitch and level. Because of its reliance on voluntary responses, Hughson-Westlake audiometry is not suitable for patients who cannot follow instructions reliably, such as pre-lingual infants (Northern and Downs 2002). As an alternative approach, we explored the use of the pupillary dilation response (PDR), a short-latency component of the orienting response evoked by novel stimuli, as an indicator of sound detection. The pupils of 31 adult participants (median age 24 years) were monitored with an infrared video camera during a standard hearing test in which they indicated by button press whether or not they heard narrowband noises centered at 1, 2, 4, and 8 kHz. Tests were conducted in a quiet, carpeted office. Pupil size was summed over the first 1750 ms after stimulus delivery, excluding later dilations linked to expenditure of cognitive effort (Kahneman and Beatty 1966; Kahneman et al. 1969). The PDR yielded thresholds comparable to the standard test at all center frequencies tested, suggesting that the PDR is as sensitive as traditional methods of assessing detection. We also tested the effects of repeating a stimulus on the habituation of the PDR. Results showed that habituation can be minimized by operating at near-threshold stimulus levels. At sound levels well above threshold, the PDR habituated but could be recovered by changing the frequency or sound level, suggesting that the PDR can also be used to test stimulus discrimination. Given these features, the PDR may be useful as an audiometric tool or as a means of assessing auditory discrimination in those who cannot produce a reliable voluntary response.

Keywords: audiometry; auditory detection; auditory discrimination; habituation; human; involuntary response; oddball; oddball paradigm; orienting; orienting reflex; orienting response; pupil dilation; pupillometry; recovery.

Conflict of interest statement

The authors declare that they have no conflict of interest.

Figures

Fig. 1

PDR apparatus. The subject, wearing insert earphones, was seated facing a computer monitor (95 cm), keyboard, IR LED array, and IR sensitive camera. The subject’s head was stabilized by chin and forehead rests. The monitor displayed a fixation point (red circle) which turned into a question mark, prompting the subject to press one key if a sound was detected and a different key if no sound was detected. The subject’s left pupil was monitored by the camera which was interfaced to Eyelink software that provided pupillary areas at a rate of 1000 samples/s

Fig. 2

Sequence of events during a trial. Upon fixating on a small circle on the computer monitor, the subject’s pupil was video imaged (green) and the pupil sizes were readout by Eyelink software. A sound (blue) was presented 1.5 s (+/− 0.5 s) after the onset of video imaging. The sound’s delivery was jittered to prevent pupillary responses in anticipation of a sound. The circle turned into a question mark, which prompted the subject to respond within 2 s (purple) if they detected a sound

Fig. 3

PDR time course and latency. aSolid curve: PDR averaged across 1673 test trials presented at the highest sound level used for four center frequencies (43 dBA; 1, 2, 4, 8 kHz center frequencies), drawn from 84 sessions across 21 subjects. Stimulus onset is at 0 s on the abscissa. Dashed curve: PDR averaged across 4191 catch trials from all 84 sessions across 21 subjects. b First derivative of the PDR trace from (a) shows the average latency of the PDR (arrow) and reveals the two-phase dynamics of pupil dilation in response to sound presentation

Fig. 4

Detectable sounds elicit a larger PDR. a–c Data from 3 typical subjects (AF, AL, AX). Pupil size traces were pooled across 100 sound trials and 50 catch trial presentations in a single session for each subject. Sound trials were pooled across all center frequencies tested (1, 2, 4, and 8 kHz) at the 4 standard sounds levels. The PDRs were later classified according to subjects’ button presses in the DRT conducted while the pupil was monitored: Hit (sound was present, subject pressed “yes” button), Miss (sound was present; subject pressed “no” button), and Correct Rejection (CR; subject pressed the “no” button during a catch trial). A dilation is seen only when the subjects reported a sound (solid line), while sounds that were undetectable yielded a pupil size trace (dashed line) that was similar to responses during catch trials (dotted line). Each trace is the average of all trials of its type in the session (Hits, Misses, CRs respectively: a: 24, 61, 43; b: 30, 68, 50; c: 16, 82, 50). Dilation responses were compared using unpaired t tests (two-tailed; p = 0.05): responses during Hits were significantly different from that during CRs, while responses during CRs and Miss trials were statistically indistinguishable. d PDR averaged across all trials and subjects. As with the individual subjects (a–c), a large dilation was observed only when a sound was detected. Traces are averages of 2233 Hits, 6141 Misses, and 4191 CRs, pooled across all 23 subjects and 85 sessions. Dilation responses in the pooled data were easily discriminable when comparing Hits to Misses (p < 10–97), while responses between Miss and CR trials were indiscriminable (p = 0.32)

Fig. 5

PDR magnitude vs SPL. Average magnitude of the PDR vs SPL across all 85 sessions from 21 subjects for center frequencies of 1 (a), 2 (b), 4 (c), and 8 kHz (d). Each point represents the average PDR sizes expressed as a z-score relative to PDRs obtained in catch trials (n = 4191). Each data point is the average of > 400 trials—trial numbers for each frequency at sound levels of 13, 23, 33, and 43 dB, respectively, are as follows: a: 421, 419, 419, and 414 trials; b: 412, 421, 417, and 419 trials; c: 415, 413, 413, and 415 trials; d: 408, 411, 414, and 415 trials. Asterisks indicate PDRs that are statistically greater (p < 0.05) in magnitude than those obtained in catch trials. Error bars represent the SEM

Fig. 6

Comparison of PDR and DRT in individual subjects. a Average psychometric functions derived from PDR (pupil size; solid lines) and DRT (button press; dashed lines) data. Example curves are shown for 2 kHz (black lines and diamonds) and 4 kHz (red lines and squares). Bars are SEM. b–e Distribution of thresholds obtained with PDR (black bars) and DRT (unfilled bars) for 1, 2, 4, and 8 kHz test stimuli. PDR and DRT data were simultaneously gathered in each subject. Overall PDR thresholds are lower than DRT thresholds but PDR variances are wider. Details of the results are given in Tables 1 and 2. Thresholds extracted from psychometric functions for individual sessions—the averages of which are shown in a—was used to construct histograms in subfigures b–e. Thus, thresholds extracted from pooled data represented by black lines (a) was used to construct histograms in c, and thresholds from pooled data represented by red lines (a) was used to construct the histograms in d

Fig. 7

PDR habituation. Mean PDR magnitude (ordinate) vs trial number (abscissa). Data are based on 36 sessions (100 trials/session) in 12 subjects. After a large dilation on the first stimulus presentation, the dilations rapidly decline, but a small dilation remains for another 32 trials. The later trials (> 32; dashed vertical line) have larger variances that fluctuate around the mean PDR size (z-score = 0) computed across all trials. Comparison of the first 30 and last 70 trials show that the first 30 trials are significantly more positive than the last 70 trials (p < 0.001; t test: Two-Sample Assuming Unequal Variances)

Fig. 8

Distribution of PDR magnitudes during habituation and recovery. a Distribution of PDR magnitudes during habituating trials. (mean = − 0.048; sd = 0.51; n = 561 trials). b Distribution of PDR magnitudes when the SPL is changed. (“SPL oddball”; mean = 0.96; sd = 0.35; n = 63 trials). The PDRs evoked by oddball stimuli (2 kHz gammatone; 38 dBA) are statistically significantly larger (p < 0.005; t test: Two-Sample Assuming Unequal Variances) than those in the habituating trials (1 kHz gammatone; 52 dBA). c Distribution of PDR magnitudes during habituating trials (12 subjects, mean = 0.002; sd = 0.23; n = 1750 trials) and frequency oddball trials. d Distribution of PDR magnitudes when the center frequency is changed (“Freq oddball”; mean = 0.54; sd = 0.30; n = 146 trials). The PDRs evoked by oddball stimuli (2 kHz gammatone; 52 dBA) are statistically significantly larger (p < 0.01; t test: Two-Sample Assuming Unequal Variances) than those from habituating trials (1 kHz gammatone; 52 dBA)

Fig. 9

Recovered PDR to Oddball stimuli. Recovery elicited by oddball sounds at different SPL (a) and center frequency (b). Habituating stimuli, consisted of the 1-kHz gammatone at 52 dBA stimulus in both a and b, while the oddballs differed in sound level (a: 38 vs 52 dB), or in center frequency (b: 2 vs 1 kHz), as detailed in Table 1. The first oddball was presented during trial 49, and succeeding oddballs were separated by 21 ± 3 habituating stimuli. The PDR habituated to the lower-SPL oddball stimuli (a: red line and circles) or different center frequency (b: magenta line and circles) over the course of 6 trials. The magnitude of the PDR to the first presentation of the oddball in both cases was similar to the magnitude of the PDR elicited by the first habituating sound in each session (t test, p = 0.05). However, responses to subsequent presentations of the oddball stimulus were smaller, indicating habituation to the oddball stimuli. By the third repetition, responses to oddball stimuli were significantly smaller than responses to the first habituating trial of each session. While the response to oddball stimuli decreased over the course of the session, it still remained larger than responses to habituating stimuli (blue and black lines in a and b). Further, the recovery was specific to the oddball stimuli: the magnitude of habituating trials flanking the oddballs remained low. The PDR magnitude of habituating trials immediately preceding (black dotted line, upright triangles) and following (blue dashed line, inverted triangles) shows that responses to habituating stimuli were unaffected by the larger response to the oddball. Data in a are based on 63 oddball trials from 6 subjects in 11 sessions. (sessions with subjects AE and AH had 5 and 4 oddball trials respectively; the rest of the subjects had 6 oddballs). Data in b are based on 146 oddball trials from 12 subjects (one session for subject AQ had 5 oddball trials, while one for subject AE had only 3)

Fig. 10

Trial order effect at near-threshold SPL. PDR sizes (ordinate) evoked by five repetitions (abscissa) of a 1-, 2-, 4-, and 8-kHz gammatone, each at one of four SPLs (13, 23, 33, 43 dBA). Each iteration of a sound-level combination was separated from the next one by ≈ 240 s. There is no systematic, statistically significant change in response magnitude (t test, p = 0.05) to later presentations of the sounds at any frequency, in contrast to the evidence of rapid habituation shown in Figs. 7 and 9