1. Behavioural estimates of auditory filter widths in ferrets using notched-noise maskers
Ana Alves-Pinto, Joseph Sollini, Toby Wells, and Christian J. Sumner
MRC Institute of Hearing Research, University Park, Nottingham, UK
Review by Mark D. Skowronski, June 8, 2016

2. Introduction
The mammalian auditory system segregates sound based on frequency.
Recordings from auditory nerve fibers show that a significant amount of filtering takes place in the auditory periphery.
How does the resolution of the peripheral frequency analysis arise perceptually?

3. Introduction One perceptual correlate of filtering: masking
Masking occurs when sounds within a frequency range (an auditory filter) mask one another.
Therefore, measurements of masking may be used to assess the characteristics of auditory filtering.
One of the most widely used methods for measuring human auditory filter shape and bandwidth: the notched-noise method (Patterson et al., JASA, 1982).

4. Notched-Noise Method

5. Motivations
The ferret is an increasingly popular animal model for hearing research.
Psychophysical filter estimates of ferrets are lacking.
Few non-human species have been characterized using the notched- noise method.

6. Behavioural Experiment
4 adult pigmented ferrets (Mammalia Carnivora Mustelidae Mustela putorius), 2 male, 2 female (also called European polecat, fitch)
Trained in 2 alternative, forced-choice detection task
Positive reinforcement with water
Task: detect 500 ms tone (20 ms ramps) in continuous noise masker
Tone: F0 of 1, 3, 7, 10 kHz
Noise: 2 spectral bands (BW: 0.64, 1.0, 1.5, 3.0 kHz), log-centered on tone frequency
Notch: Up to 50% center frequency at 1 kHz, 30% for higher tone frequencies
“The noise spectrum level was constant for each filter estimate but 2 to 3 levels [20 to 40 dB SPL +/- 3 dB] were tested at each signal frequency.” ???
Tone level varied according to Method of Constant Stimuli, 5 levels used (one level produced 90% detection, one level was below threshold).
Tested in 3-5 sessions, interlacing notch widths, at least 30 trials (signal and no-signal conditions) were recorded per level
Psychometric functions of tone level fit with logistic functions, fixed slopes

7. Auditory Filter Model Rounded exponential (roex):
Signal+masker convolved with room transfer function, scaled to match their levels at threshold, and integrated over roex filters
Threshold taken from filter with best SNR, repeated across several notch widths
Model parameters p and r found by MMSE
Reliability estimated with bootstrap sampling with replacement (500 replicates)

8. Results

9. Results
From Moore and Glasberg, JASA 1983:

10. Discussion
Auditory filter ERB ~ 20% center frequency, similar to guinea pigs (20%), chinchilla (14%), humans (11%)
No effect of level on filter bandwidth (unlike humans)
Compared to critical band method of filter bandwidth estimation (more widely used for non-human animals), notched-noise method BWs are smaller
Notched-noise method BWs are ~20% larger than auditory nerve tuning in the ferret (why the difference?)
Cognitive factors (attending to multiple frequencies)
Detection efficiency, affected by listening strategy
Interfering non-sensory factors