Fig 1.
Experiment 1 experimental paradigm.
(A) Behavioural testing of each listener’s perceptual bias. A sample tone pair is shown schematically; two complex missing fundamental tones comprised of the 8th-10th and 7th-9th harmonics were played sequentially after a short silent pause. This was repeated once, then subjects were asked to record whether they perceived the second tone to be higher or lower than the first. Tones were constructed such that spectral and fundamental perceptions lead to opposite responses, and a measure of overall perceptual bias was calculated from responses on 20 tone pairs. (B) Stimuli used in ABR testing; MF and FP stimuli differed only in the absence (MF: missing fundamental) or presence (FP: fundamental present) of energy at the fundamental frequency (f0).
Fig 2.
Perceptual bias and relationships to musical experience and age.
(A) Distribution of perceptual bias in our sample (N = 39); this is weighted towards fundamental listeners but nonetheless includes people with a range of responses. (B) Perceptual bias was correlated with cumulative hours of musical training, and was negatively correlated with (C) the age of training onset in subjects who reported musical experience (N = 30). (D) Perceptual bias was not significantly correlated with current age.
Fig 3.
Inter-individual variability in frequency following response to complex tones that contain (black), or lack (red) energy at the fundamental frequency (f0).
(A) Time domain representation of five example subjects (band-pass filtered between 80 and 2000Hz), demonstrating variation in ABR waveform. Grey shading indicates the FFR portion of the signal. (B) Frequency domain representation of the example subjects’ FFR. Variability is evident in the strength of the encoding of the f0 (98Hz) and its harmonics (integer multiples), and in their pattern of relative strength. The relationship between the f0 and mean harmonic PLVs (2nd-5th harmonics), in the (C) missing fundamental and (D) fundamental present conditions demonstrates considerable variability in the strength of each and their relationship, although in general people with a strong representation of the f0 have a strong representation of the harmonics. Histograms are associated with each axis to illustrate the distributions (above: f0 PLVs; right: mean harmonic PLVs).
Fig 4.
(A, B) Fundamental perception bias was correlated with f0 peak magnitude in response to missing fundamental and fundamental present tones in each condition (the effect of age is controlled). (C) Histogram of PLV differences at f0 between conditions. While many subjects showed a statistically significant distinction between conditions as determined by resampling (14/39 subjects), subjects did not show consistency in the condition with greater amplitude; 9 had a greater amplitude in the FP condition and 5 had a greater amplitude in the MF condition. (D, E) The PLV of the f0 in each condition was positively correlated with the mean PLV in the low gamma frequency range in both conditions. Both the relationships between f0 PLV and perceptual bias and between f0 PLV and gamma PLV appear more clearly in the MF condition (red), when the fundamental frequency is not present in the stimulus and must be computed. FP: fundamental present condition; MF: missing fundamental condition.
Fig 5.
In behavioural testing, subjects heard pure tone primes that matched the pitches of a spectral or fundamental perception of subsequently presented complex harmonic melodies, and were asked whether there was a match or mismatch between the two. This was used to determine the harmonic range used for each subject in the ABR recording (see Fig 6), and confirm that subjects could hear in both perceptual modes. During the ABR recording, subjects were primed to hear in a single repeated melody in their specific harmonic range, and were asked to identify occasional mismatches to control for attention and perception.
Fig 6.
Harmonic number range determination.
Using the task illustrated in Fig 5, a range of ambiguous harmonics for each subject was identified: first, MF complex tones with very high harmonics, which encourage spectral perception, were presented. Subjects’ match-mismatch identification accuracy was evaluated, and the harmonics decreased in steps of 4 until a threshold of spectral perception was found (i.e. fewer than 4/5 trials correct) or the range was exhausted. Second, complex tones with low harmonics, which encourage fundamental perception, were presented, and increased according to accuracy until a threshold was found. A range of harmonics based on the midpoint between the thresholds was used for further behavioural testing to confirm that each subject could hear in both modes, and melodies based on the central harmonic were created for the ABR recording.
Fig 7.
Experiment 2 behavioural results: accuracy identifying matches and mismatches in trials of each condition.
Accuracy was significantly above chance in each listening condition. Subjects had significantly better scores in the f0 condition. Mean accuracy across conditions was used for further analysis. Error bars indicate +/-SEM; individual data are superimposed to illustrate the distribution.
Fig 8.
Accuracy on the listening mode task compared with peak f0 magnitude in the FFR.
(A) In the fundamental listening condition, peak f0 magnitude correlated significantly with overall task accuracy, but in the spectral listening condition (B) the trend did not reach significance. The difference between the f0 peak magnitude in each condition (i.e. SP–f0) was also significantly correlated with task performance.
Fig 9.
Neural correlates of perceptual condition to a complex harmonic stimulus at the fundamental frequency.
(A) Phase-locking values averaged across subjects for each condition as a function of neural response frequency (Cz-averaged mastoids electrode channel) show no difference at the group level at f0 (arrow; 160Hz). However, (B) for many individuals (16/29), the f0 peak in each listening condition was statistically distinguishable in either the f0 > SP or SP > f0 direction, as calculated by using a resampling distribution.
Fig 10.
Neural correlates of perceptual condition to a complex harmonic stimulus in the low gamma band (24-48Hz) at an electrode placed over the right hemisphere (C4).
(A) In spectra at electrode C4 (above right hemisphere), an average difference is observed in the gamma range (arrow). (B) Across subjects, gamma band amplitude was significantly greater in the f0 condition and (C) the difference between conditions in the FFR f0 and gamma band was significantly correlated. Relationships between gamma activity and the behavioural measures were not evident.