Did you know that mice can sing? We can’t hear them, unfortunately – they’re a few octaves out of our range, and in fact even out of the range of many bat calls but they do nonetheless emit noises that bear many similarities to what we call songs. What isn’t known, however, is whether they are born with their songs fully formed in their heads, or whether they learn them from their parents and peers, as we do.
Two papers, published today, have both sought to answer this question, but have come to seemingly different results. In the first paper, Jasmine Grimsley, a post-doc fellow at Northeastern Ohio Universities Colleges of Medicine and Pharmacy, and her colleagues demonstrated changes during development of social vocalizations in mice, suggesting that some learning takes place. In the second paper, Takefumi Kikusui from Azabu University and Kazuo Okanoya from the RIKEN Brain Science Institute, along with their colleagues conducted cross-fostering experiments that suggested that the songs are innate.
Given the two conclusions, we thought it would be interesting to find out a little bit more about the authors, their research and their opinions on each other’s research. Drs. Grimsley, Kikusui, and Okanoya graciously agreed to answer the following questions for us via email.
First, a bit of detail on your scientific background – how did you become interested in animal behavior and evolutionary biology?
Grimsley: I am really interested in hearing research, and with the advent of midbrain auditory implants it has become increasingly important that we understand how higher order areas of the brain process complex signals, such as social vocalizations. For my PhD work at the Institute of Hearing Research in England, I compared how all eight areas of guinea pig auditory cortex respond to conspecific social vocalizations. I try to approach research from a neuroethological perspective; understanding the natural behavior of my animal model is fundamental to this. Currently, I am a postdoctoral researcher in Dr. Wenstrup’s lab at NEOUCOM, and our lab is interested in how the brain codes the meaning of social vocalizations in both mice and bats. We have recently started working with mice because of the opportunities to study genetic variants and disease models.
Kikusui: I started my career with veterinary medicine when I was an undergraduate and belonged to the lab of Veterinary Ethology in the University of Tokyo. I learned a lot of animal behavior including behavioral problems in dogs and cats. To solve the behavioral problems, we must understand the natural behavior of animals, that is, why the animals show species specific behavior (evolutionary biology, or genetic traits) and how the animals adapt to the environments, especially the social experience during the neonatal period of their lives (developmental biology, or acquired traits). Gradually, my research focus has been sifting to the two big questions, one is social cognition in animals and the other is developmental impacts on social brain functions.
Your manuscript includes several different audio recordings of the mice you were studying. Which audio file would you consider the most important and why?
Grimsley: We have compiled a library of the most typical example of each syllable from mice at each of the ages we tested. The “virtual mouse vocal organ” that is downloadable with the paper can produce bouts of mouse song that have patterns that are appropriate for mice of different ages. We hope that this will serve as a useful tool for other scientists interested in mouse communication.
Kikusui: I would recommend the first two audio files; Audio S1 represents C57BL6 song, Audio S2 represents BALB/C song. Because the readers would understand how these two strains emit different songs by these files.
Dr. Grimsley: In your manuscript, you suggest that mice are capable of vocal learning. Dr. Okanoya: Your team suggests that mouse songs do not involve imitative learning. At first glance, your papers seem to conflict. Can you briefly explain the differences between your studies? Could you offer some possible explanations for the different conclusions?
Grimsley: Our results don’t conflict with those of Kikusui and co-workers, because the experiments were different. In fact, both papers suggest that some elements of mouse songs are innate. However, we believe that the design of their experiments did not allow them to conclusively test whether mice are capable of vocal learning. Kakusui et al state that mice would have ample opportunity to hear adult male mating song during the period of exposure to adults (the first 21 days after birth). Since hearing onset occurs when pups are 10 days old, they believe that this allows for approximately 11 days exposure to adult song. In fact, high frequency hearing above 50 kHz (the frequencies in male mating song) does not develop in mice until they are older than 14 days. So the pups may not hear adult male songs until near the end of the 21 day exposure period. In other species, e.g., songbirds, vocal learning takes place during a more extensive period of exposure to adult songs. For example, in white crowned sparrows, where vocal learning is well understood, there is a much longer critical period for juveniles to learn adult song, between 10 and 50 days old. This 40 day period occurs during a time when the birds are able to fly and are considered to be juvenile rather than chicks. A comparable time period in mice may be between weaning (age 21 days) and the onset of sexual maturity (about 40 days). No exposure to adult song was provided during this period.
Okanoya: Showing developmental changes or behavioral complexity alone cannot determine whether or not a behavior is learned. As Grimsley et al also suggested, some types of
sensory manipulations such as deafening or cross-fostering should be necessary. We utilized the cross-fostering procedure to indicate that mice songs do not involve imitative learning. We think it may be necessary to define vocal learning in two types: vocal learning in the weak sense (VLW) and vocal learning in the strong sense (VLS). VLS refers to the involvement of the imitative learning, while VLW only refers to the involvement of the auditory feedback. VLS requires two steps: establishment of the auditory template for the tutor’s vocalizations and the feedback modifications of own vocalizations to match the template. VLW involves only one step: the feedback modifications of own vocalizations to match the “innate” template. While the Grimsley paper may be implying VLW but they never imply VLS. Experimental procedures used by Grimsley, while they are adequate to describe developmental changes, they do not address whether the behavior is learned or not. When we refer to vocal learning, we mean VLS, but not VLW, because human speech learning is obviously of the nature of VLS.
These two sets of experiments are a nice illustration of the complex process of scientific discovery. What have you learned from reading the other paper? Has this influenced the direction you plan to take in your own work?
Grimsley: The paper by Kakusui et al has shown some interesting strain-dependent differences in mouse song. They have also demonstrated that if there is a critical period for mice to learn song it is likely not to be before weaning at 21 days. The strong strain-dependent differences they show mean that it would be interesting to see if, behaviorally, mice are more attracted to the song from mice of their own strain.
Okanoya: As mentioned in the previous question, we feel the necessity to define vocal learning in two types. The complexity in developmental process and behavioral phenotypes as described in Grimsley paper, together with the fact that this behavior does not involve imitative learning, makes the mice vocalization an interesting preparation to study the process of unfolding of the genetically determined complex behavior.
What experiments would really answer this question of whether mouse vocalizations are innate or learned for you?
Grimsley: It would be interesting to test whether the song from a mouse with no auditory experience is different than song from a mouse with song experience. Practically, this could be
tested by deafening young mice prior to hearing onset and raising them with deafened parents. Raising mice with deafened parents would control for any possible factors such as behavioral feedback from the parents that might help shape their song.
Okanoya: We believe, based on our results, mice songs do not involve imitative vocal learning. In this sense, mice songs are innately determined, meaning that the under a normal developmental environment, behavioral phenotype of songs are highly stereotyped depending on the strain. In addition to our study, feedback alternation with an electric device might show the degree of auditory dependence.
What is your next big research project? Or where do you want to go from here?
Grimsley: We are using the ‘virtual mouse vocal organ’ from this study to generate bouts of mouse song to test the responses of the amygdala to social vocalizations. I am also really interested in investigating the effects of social context on the responses to these stimuli, both behaviorally and neurophysiologically.
Okanoya: We would like to know the neural mechanisms by which mice songs gain sequential variability. We suspect the involvement of the basal ganglia and the anterior cingulated cortex for this variability. Also, we would like to know how females respond to the strain differences of the songs. We are in the process of examining these questions.
Both of the manuscripts are freely available to read, comment on and rate.
Grimsley JMS, Monaghan JJM, Wenstrup JJ (2011) Development of Social Vocalizations in Mice. PLoS ONE 6(3): e17460. doi:10.1371/journal.pone.0017460
Kikusui T, Nakanishi K, Nakagawa R, Nagasawa M, Mogi K, et al. (2011) Cross Fostering Experiments Suggest That Mice Songs Are Innate. PLoS ONE 6(3): e17721. doi:10.1371/journal.pone.0017721