Updated: Dec 18, 2022
Communication is a vital part of all of our lives. When humans communicate with each other using language, we arrange words into specific sequences, which we call sentences. However, only some particular arrangements of words carry meaning, whereas others do not. For example, the sentence ‘frogs are very cool’ carries a precise meaning, whereas the sentence ‘colourless green ideas sleep furiously’ is meaningless. Do other animals also arrange sounds into specific sequences in this manner, or do they simply emit various calls in random order? Do animals use different kinds of sequences in different contexts? In our recent paper (Bhat et al. 2022), we examined whether vocal sequences of frogs varied with the behavioural context in which the frogs are emitting calls (For example, based on whether they were calling to attract mates or engage in territorial disputes). Why frogs? Well, besides the fact that frogs are interesting organisms in their own right, most frogs are nocturnal and communicate with each other primarily through sound. Furthermore, male frogs call for many different reasons, including to attract mates of the opposite sex and to guard their territories against rivals of the same sex. Since calling for a mate and defending one’s territory are very different scenarios for the individual frogs involved, it is important for both the calling frog and the listening individuals that the message that is given out in each of these different behavioural contexts is unambiguous. However, unlike birds, frogs often only have access to a limited set of ‘notes’ out of which to construct their vocalisations (A ‘note’ is jargon for a unit of vocalisation, and in our analogy to the English language, plays the role of words). Thus, using a distinct note for each context is often infeasible. We wished to address whether these frogs arrange their notes into sequences in different ways in different contexts, which in our analogy is asking whether they arrange the same words into different sentences in different contexts.
To do this, we studied frogs of two species — Nyctibatrachus humayuni and Pseudophilautus amboli — in different behavioural contexts in the field across two states over two years. The data collection involved long nights in the field recording in pitch darkness, trying our best not to let the sensitive recording equipment get wet, all the while keeping an eye out for the venomous pit vipers that are common in the region. N. humayuni and P. amboli are only distantly related. We reasoned that if we were to find any patterns that are present in both these species, it is likely to be general and not a simple consequence of shared evolutionary history.
Immediately, we realised that N. humayuni has a relatively small ‘vocabulary’ consisting of two note types, which we called the ‘ascending note’ (AN) and ‘descending note’ (DN). P. amboli, on the other hand, has a large vocabulary consisting of six different note types, which we labelled note types 1-6 [scan QR Code/use link at the end of this article]. To begin with, we counted the number of times different types of ‘notes’ were emitted in different contexts, to check whether these frogs used different notes according to context. This revealed that while N. humayuni appends a variable number of DNs to a single AN according to context, P. amboli uses different note types entirely in different contexts.
We used three different analysis techniques to check whether frogs changed how they arranged their notes based on context. The first analysis was based on a quantity called ‘Shannon entropy’, borrowed from a field of mathematics called information theory. The Shannon entropy quantifies how ‘diverse’, ‘complex’, and ‘unpredictable’ a sequence is based on the notes it contains and how these notes are arranged. For example, if we consider note types to be represented by letters of the alphabet, then ‘aaaaaaaaaaa’ would have low entropy, whereas ‘abracadabra’ would have higher entropy. We found that the calls of N. humayuni had a higher Shannon entropy in the presence of vocal neighbours of the same species, indicating that they used more elaborate, complex sequences in the presence of rival males, perhaps to impress any potential suitors that may be listening to both the male and its rivals. In contrast, the entropy of P. amboli did not vary based on the presence or absence of vocal neighbours, but significantly decreased during territorial fights. This indicated that these frogs used simpler, less diverse sequences during territorial disputes when they were primarily calling out to a rival male and not seeking the attention of a female.
The second analysis relied on a mathematical model called a ‘Markov chain’. Simply put, a Markov chain is a model in which the probability of a frog emitting a particular note type depends only on the note type that directly preceded it.
To draw an example from real life, imagine that we are playing a game where each ‘round’ consists of betting ₹1 on whether a coin toss turns out to be heads or tails.
In this case, regardless of how many rounds we have played, to predict how much money you will have in the next round, I only need to know how much money you currently have — if you have ₹x in this round, you are guaranteed to either have ₹(x+1) or ₹(x-1) in the next round (each with 50% probability, depending on the outcome of the coin toss). Thus, in terms of predictability, it is irrelevant whether you had won previous rounds or not — all that matters is how much money you currently have. This is a simple example of a ‘Markov chain’. When our frogs were modelled in this way, we were able to show that both species modified their vocal sequences according to context. However, Markov chain analysis relies on assumptions that are not always true — it is often much more realistic to assume that the chance of emitting a note type depends on more than just the note type directly preceding it.
To circumvent this issue, we also devised our own novel data analysis technique, which we call ‘co-occurrence analysis’. Co-occurrence analysis measures whether a frog emits certain notes together more often than expected by chance alone. For example, in English, the letter ‘q’ is often followed by the letter ‘u’ (in words such as queue, quack, or quartz), much more often than would happen if letters were arranged at random. Co-occurrence analysis measures whether animals like frogs have similar rules for arranging notes into sequences. This analysis technique that we devised can be applied to all sorts of sequences, and the code used to conduct this analysis is publicly available for use. This analysis too showed that frogs modify the arrangement of their notes according to context, further solidifying our findings.
To summarise, we were able to show using various measures that both species of frogs studied modify the arrangement of their notes according to context. While similar studies have been conducted for ‘complex’ organisms such as birds and mammals in the past, our study is the first to undertake such an analysis in frogs. Frogs have been traditionally neglected as being ‘simple’ creatures but our study suggests that they may be more complex than we think, and may harbour many secrets that have been left uninvestigated. I am confident that further studies on these ancient endemic creatures will teach us much more about how animals communicate — we just need somebody to listen to what they’re trying to say.
Bhat, A. S., Sane, V. A., Seshadri, K. S., & Krishnan, A. (2022). Behavioural context shapes vocal sequences in two anuran species with different repertoire sizes. Animal Behaviour 149, pp 111-129 https://doi.org/10.1016/j.anbehav.2021.12.004
Shikhara (he/him) is currently pursuing a BS-MS degree at the Indian Institute of Science Education and Research (IISER) Pune, India. He is broadly interested in working at the interface of theory and experiment in ecology and evolution.