Amplitude modulation pattern of rat distress vocalisations during fear conditioning

Abstract

In humans, screams have strong amplitude modulations (AM) at 30 to 150 Hz. These AM correspond to the acoustic correlate of perceptual roughness. In bats, distress calls can carry AMs, which elicit heart rate increases in playback experiments. Whether amplitude modulation occurs in fearful vocalisations of other animal species beyond humans and bats remains unknown. Here we analysed the AM pattern of rats' 22-kHz ultrasonic vocalisations emitted in a fear conditioning task. We found that the number of vocalisations decreases during the presentation of conditioned stimuli. We also observed that AMs do occur in rat 22-kHz vocalisations. AMs are stronger during the presentation of conditioned stimuli, and during escape behaviour compared to freezing. Our results suggest that the presence of AMs in vocalisations emitted could reflect the animal's internal state of fear related to avoidance behaviour.

Figure 1

Example of USV recordings during a trial. (A) Top: USV were recorded during the 4-min trial divided into three periods: pre-odour, odour, and post-shock. The shock was applied during the last second of odour presentation. Bottom, magnified oscillograms of the recording on the time intervals signalled on the traces above. Each number indicates individual vocalizations. (B) Top, oscillograms of two individual vocalisations (zoom-in of (A)). Bottom, spectrograms of the two vocalisations.

Figure 2

Properties of 22-kHz USVs and respiration rate during the fear-conditioning task. (A) Histogram of number of 22-kHz USVs uttered in each period (pre-odour, odour, and post-shock). Chi-squared tests with number of vocalisations in 20 randomly selected 1-s bins. (B–D) Boxplots of the call duration, root mean square (RMS) and peak frequency of the same vocalisations. (E) Boxplot of respiration rate calculated during 1-s bins. Boxes depict the interquartile range and median. Kruskal–Wallis tests for data shown in (B–E) (p values shown on top of each panel; n = 1300, n = 606, n = 1447, for pre-odour, odour and post-shock, respectively). Stars and horizontal black lines indicate between-group comparisons (Bonferroni-corrected rank-sum tests, *p < 0.05, ***p < 0.001).

Figure 3

Analysis of 22-kHz USVs as a function of trial number. Histogram of number of vocalisations (A,F,K) and violin plots of duration (B, G, L), root-mean square (RMS; C,H,M), peak frequency (D,I,N) and respiration rate (E,J,O). White circle indicates the median and grey bar the interquartile range. Insets: statistical significance; number of vocalisations: Holm–Bonferroni corrected chi-squared test; duration, RMS, peak frequency, and respiration rate: Holm–Bonferroni corrected Wilcoxon rank-sum tests. *p < 0.05, **p < 0.01, ***p < 0.001.

Figure 4

(A) Oscillograms (top) and spectrograms (bottom) of two examples of 22-kHz rats’ vocalisations, one with weak (left) and one with strong (right) amplitude modulations (AMs). (B) Mean of the modulation power spectra (MPS) of 25 manually selected vocalisations with strong (A, top) and weak (A, bottom) AMs. Only the first 300 ms of each vocalisation were considered and all were RMS-normalised. (C) Cliff’s delta of the MPS of the vocalisations used in B) comparing strong and weak AMs. Solid black lines designating areas with values > 0.948 (double the threshold of what is considered as large effect size). Dashed black lines show the area within which the mean of the MPS values were calculated for all vocalisations (what is considered as AM scores in this manuscript).

Figure 5

Amplitude modulation score. (A) Boxplot of the AM score in each for the three periods during the first three and last three trials for both groups. Boxes depict the interquartile range and the horizontal line inside each box indicates the median. (B) Violin plots of the amplitude modulation score (see “Methods”) for each period and trial. Insets: p values of Wilcoxon rank-sum tests (Holm–Bonferroni corrected). White circle indicates the median and grey bar the interquartile range. (C) For this panel all vocalisations that are between 0.90 (median) and 1.5 s long were used (1164 in total). The envelope of these vocalisations was calculated, only the first 900 ms was considered and they were z-normalised. Then the spectrograms of the envelopes were calculated. These vocalisations were median split according to their AM score and the mean spectrogram of each group is plotted here. Logarithmic axes. Black lines designate an arbitrary threshold of −6. (D) AM score of the vocalisations emitted right after the shock (10 s post-shock period) and classified into freezing (n = 145) or escape (n = 141) depending on the rat’s behaviour (Wilcoxon rank-sum test). *p < 0.05, **p < 0.01, ***p < 0.001.

Figure 6

Correlations of 22-kHz vocalisations. (A) Peak frequency plotted against number of vocalisations (here it was considered the mean peak frequency and the mean number of USVs emitted during 5-s bins for each rat and trial). (B–F) AM score plotted against the natural logarithm of the root-mean square (RMS), number of vocalisations (calculated as the mean in 5-s bins), peak frequency, duration and respiration rate. G-I) AM score difference plotted against the respiration rate difference [“Post-shock”—“Odour” (G); “Post-shock”—“Pre-odour” (H); “Odour”—“Pre-odour” (I)]. Each dot represents one trial and rat (colour-coded for rat’s identity; see legend). Lines depict the fitted linear regression (see colour legend for the interpretation). Pearson’s correlation coefficient (r) shown for each plot. *** p < 0.001.