Discrimination of ultrasonic vocalizations by CBA/CaJ mice (Mus musculus) is related to spectrotemporal dissimilarity of vocalizations

Abstract

The function of ultrasonic vocalizations (USVs) produced by mice (Mus musculus) is a topic of broad interest to many researchers. These USVs differ widely in spectrotemporal characteristics, suggesting different categories of vocalizations, although this has never been behaviorally demonstrated. Although electrophysiological studies indicate that neurons can discriminate among vocalizations at the level of the auditory midbrain, perceptual acuity for vocalizations has yet to be determined. Here, we trained CBA/CaJ mice using operant conditioning to discriminate between different vocalizations and between a spectrotemporally modified vocalization and its original version. Mice were able to discriminate between vocalization types and between manipulated vocalizations, with performance negatively correlating with spectrotemporal similarity. That is, discrimination performance was higher for dissimilar vocalizations and much lower for similar vocalizations. The behavioral data match previous neurophysiological results in the inferior colliculus (IC), using the same stimuli. These findings suggest that the different vocalizations could carry different meanings for the mice. Furthermore, the finding that behavioral discrimination matched neural discrimination in the IC suggests that the IC plays an important role in the perceptual discrimination of vocalizations.

Figure 1. Vocalization spectrograms.

Spectrograms of two unmodified vocalizations (A = Male Upsweep and F = 40 kHz Harm/2 Jump), and those same two vocalizations doubled in duration (B and G), lowered in frequency by 20% (C and H), with frequency modulation removed (D and I), and reversed in time (E and J).

Figure 2. Discrimination of vocalizations.

Mean discrimination performance across subjects for the five vocalizations types (A-E) against all other vocalizations. The blue horizontal dashed lines represent chance level performance. Error bars are between-subject standard errors. The red horizontal lines connect significantly different bars. The missing bar in each of the graphs is when the stimulus was used as the background.

Figure 3. Multidimensional analysis of vocalization discrimination.

At the top, multidimensional map of mean vocalization discrimination obtained from matrices of discrimination performance for the five vocalization types. Circles connect vocalizations with the two smallest Euclidian distances. At the bottom are the five individual perceptual maps generated for each mouse across the five vocalization types.

Figure 4. Discrimination of manipulated vocalizations.

Mean discrimination performance across subjects and across the five background vocalization types for the eight vocalization manipulations. Error bars are between-subjects standard errors. The blue horizontal dashed lines represent chance level performance.

Figure 5. Correlation between spectrotemporal similarity and discrimination performance.

Percent correct discrimination as a function of spectrotemporal similarity for all vocalization discriminations in the first experiment and all vocalizations versus their manipulations in the second experiment.