Final Report Summary - DIALECT EVOLUTION (Principles of dialect evolution in killer whales)
Each pod of killer whales shares a unique repertoire of discrete calls – a vocal dialect, which is thought to pass from mothers to offspring through social learning and therefore appears to be a form of animal culture. Calls change with time, leading to the cultural evolution of dialects, but it was unknown what drives the process of call change. Through this project I aimed to reveal and describe the principles of dialect evolution in killer whales through computer modeling and verification the model outcomes with the real data.
To test the probability of different call change mechanisms, I have created an agent-based model of dialect evolution in MATLAB programming environment. The model represented a population of killer whales with life history and demographic parameters derived from the natural populations. Each whale agent younger than 15 yrs old changed its call every year according to a set of learning rules. I tested ten models with learning rules that differ by template source (mother or matriline), variation type (random errors or innovations) and type of call change (no divergence from kin vs. divergence from kin). Modeling showed that accumulation of copying errors cannot be solely responsible for the dialect evolution: learning from either mother alone or the entire matriline with calls changing by random errors produced a graded distribution of the call phenotype, without the discrete call types observed in nature. Introducing occasional innovation or random error proportional to matriline variance yielded more or less discrete and stable call types. The tendency to diverge from the calls of related matrilines provided fast divergence of loose call clusters. A pattern resembling the dialect diversity observed in the wild arose only when rules were applied in combinations. Similar outputs could arise from different learning rules and their combinations.
To get insights into the structure of the real killer whale repertoires, I have analysed group-specific repertoires of North Pacific killer whales. Analysis of different call types on within-group level revealed existence of group-specific call features that occur in more than one call type. This can develop in two ways: though within-group horizontal transmission of features between calls, or through parallel formation of similar calls in different groups from a certain group-specific template. Analysing distribution of broad call categories (call classes) in the dialects of killer whales on the intra-pod, intra-clan and intra-population level showed that biphonic calls of all studied resident populations of North Pacific killer whales fell into three classes, while transient killer whales produced calls that were different from any of resident classes.
Comparison of the frequency contours of calls of the same types from different groups revealed that the similarity tree across groups was different for different call types. This indicates the deviation from random evolution model that can be caused by horizontal transmission of call features or saturation of structural changes due to constraints in call change. The discrepancy of similarity trees of different calls results in unresolvable contradictions in genealogy analysis. To resolve this problem, I have carried out a phylogenetic analysis of the dialects of the two well-studied populations in the North Pacific: Canadian Northern Resident and Kamchatkan Resident. I used MrBayes software to build Bayesian phylogenetic trees based on the data on presence or absence of a particular call type or subtype in the dialect of each killer whale pod. Bayesian phylogenetic trees were generally similar but not necessarily identical to the acoustic similarity trees. This indicates that acoustic similarity is not always explained by common ancestry, confirming the results of modeling. Besides, I found a correlation between the number of group splitting events and length of the path from the root to the branch tip. This finding is in agreement with the predictions from the theory of punctuated evolution, strengthening the conclusion from our model that innovations may play important role in the evolution of dialects.
I have analysed the library of recordings of the North Atlantic killer whales in the host lab and obtained additional recordings through the collaborative fieldwork in Iceland. I have created catalogs of discrete calls of Norwegian and Icelandic killer whales, measured the calls from the catalogs and compared their parameters to those of the well-studied North Pacific killer whale calls from four resident and two transient populations. I found that both low- and high-frequency components of North Pacific transient killer whale calls had significantly lower frequencies than those of the North Pacific resident and North Atlantic populations. The range of frequency modulation within calls did not show the same pattern, with little consistent variation across the ecotypes. This indicates that though calls cover a similar range of frequencies, the calls produced by transients are shifted to a lower frequency. This finding gives us some insights into cultural macro-evolution of killer whale calls that could be related to ecological specialization or to the phylogenetic history of populations.
The results of my project allow drawing important comparisons with cultural evolution of other socially transmitted signaling systems, including human languages. The fact that evolution of dialects in killer whales requires the similar basic mechanisms that are involved in human language evolution (random errors, innovations, simultaneous divergence and convergence on different scales) suggests that these mechanisms may be universal principles of cultural evolution. Our study strengthens the argument that there is a dramatic difference between cultural and biological evolution. Biological evolution is based almost entirely on the random errors with further natural selection of adaptive genotypes. In cultural evolution, the active involvement of a bearer is possible, leading to innovations as well as divergence to and convergence from other cultural phenotypes.
Besides, our results have important implications to the maintenance and conservation of killer whales in the eastern North Atlantic. The population structure of killer whales in this region is not yet clear: genetic studies do not completely agree with photoidentification results, which may be due to the recent changes in the ranging patterns following the shifts in fish stock distribution. The acoustic repertoires of Icelandic and Norwegian populations described through our project provide an important cue to unraveling this complicated problem, because acoustic repertoires change faster than genetic markers, but slower than social structure and ranging patterns, thus providing an intermediate data layer essential for identifying the population structure. Identifying killer whale populations is a key goal for conservation and management of the species, because any impact must be primarily accessed on the population rather than on species level.
To test the probability of different call change mechanisms, I have created an agent-based model of dialect evolution in MATLAB programming environment. The model represented a population of killer whales with life history and demographic parameters derived from the natural populations. Each whale agent younger than 15 yrs old changed its call every year according to a set of learning rules. I tested ten models with learning rules that differ by template source (mother or matriline), variation type (random errors or innovations) and type of call change (no divergence from kin vs. divergence from kin). Modeling showed that accumulation of copying errors cannot be solely responsible for the dialect evolution: learning from either mother alone or the entire matriline with calls changing by random errors produced a graded distribution of the call phenotype, without the discrete call types observed in nature. Introducing occasional innovation or random error proportional to matriline variance yielded more or less discrete and stable call types. The tendency to diverge from the calls of related matrilines provided fast divergence of loose call clusters. A pattern resembling the dialect diversity observed in the wild arose only when rules were applied in combinations. Similar outputs could arise from different learning rules and their combinations.
To get insights into the structure of the real killer whale repertoires, I have analysed group-specific repertoires of North Pacific killer whales. Analysis of different call types on within-group level revealed existence of group-specific call features that occur in more than one call type. This can develop in two ways: though within-group horizontal transmission of features between calls, or through parallel formation of similar calls in different groups from a certain group-specific template. Analysing distribution of broad call categories (call classes) in the dialects of killer whales on the intra-pod, intra-clan and intra-population level showed that biphonic calls of all studied resident populations of North Pacific killer whales fell into three classes, while transient killer whales produced calls that were different from any of resident classes.
Comparison of the frequency contours of calls of the same types from different groups revealed that the similarity tree across groups was different for different call types. This indicates the deviation from random evolution model that can be caused by horizontal transmission of call features or saturation of structural changes due to constraints in call change. The discrepancy of similarity trees of different calls results in unresolvable contradictions in genealogy analysis. To resolve this problem, I have carried out a phylogenetic analysis of the dialects of the two well-studied populations in the North Pacific: Canadian Northern Resident and Kamchatkan Resident. I used MrBayes software to build Bayesian phylogenetic trees based on the data on presence or absence of a particular call type or subtype in the dialect of each killer whale pod. Bayesian phylogenetic trees were generally similar but not necessarily identical to the acoustic similarity trees. This indicates that acoustic similarity is not always explained by common ancestry, confirming the results of modeling. Besides, I found a correlation between the number of group splitting events and length of the path from the root to the branch tip. This finding is in agreement with the predictions from the theory of punctuated evolution, strengthening the conclusion from our model that innovations may play important role in the evolution of dialects.
I have analysed the library of recordings of the North Atlantic killer whales in the host lab and obtained additional recordings through the collaborative fieldwork in Iceland. I have created catalogs of discrete calls of Norwegian and Icelandic killer whales, measured the calls from the catalogs and compared their parameters to those of the well-studied North Pacific killer whale calls from four resident and two transient populations. I found that both low- and high-frequency components of North Pacific transient killer whale calls had significantly lower frequencies than those of the North Pacific resident and North Atlantic populations. The range of frequency modulation within calls did not show the same pattern, with little consistent variation across the ecotypes. This indicates that though calls cover a similar range of frequencies, the calls produced by transients are shifted to a lower frequency. This finding gives us some insights into cultural macro-evolution of killer whale calls that could be related to ecological specialization or to the phylogenetic history of populations.
The results of my project allow drawing important comparisons with cultural evolution of other socially transmitted signaling systems, including human languages. The fact that evolution of dialects in killer whales requires the similar basic mechanisms that are involved in human language evolution (random errors, innovations, simultaneous divergence and convergence on different scales) suggests that these mechanisms may be universal principles of cultural evolution. Our study strengthens the argument that there is a dramatic difference between cultural and biological evolution. Biological evolution is based almost entirely on the random errors with further natural selection of adaptive genotypes. In cultural evolution, the active involvement of a bearer is possible, leading to innovations as well as divergence to and convergence from other cultural phenotypes.
Besides, our results have important implications to the maintenance and conservation of killer whales in the eastern North Atlantic. The population structure of killer whales in this region is not yet clear: genetic studies do not completely agree with photoidentification results, which may be due to the recent changes in the ranging patterns following the shifts in fish stock distribution. The acoustic repertoires of Icelandic and Norwegian populations described through our project provide an important cue to unraveling this complicated problem, because acoustic repertoires change faster than genetic markers, but slower than social structure and ranging patterns, thus providing an intermediate data layer essential for identifying the population structure. Identifying killer whale populations is a key goal for conservation and management of the species, because any impact must be primarily accessed on the population rather than on species level.