Modeling cultural evolution. A parallel investigation of changes in bird song and human language

Prof. dr. C.J. ten Cate en Prof. dr. A. Verhagen

Results 2006

5.1 The project has underlined some important commonalities between human language and bird song: two culturally transmitted communication systems. On the linguistic side, the project has explored cultural evolutionary interactions (filter/signal relationships) that have previously been explored in studies of bird song. On the bird song side, the project has explored how the complex hierarchical cultural transmission of song has evolved along a chain of colonization of oceanic islands. We believe that this bolsters the value of bird song as an analogous model of language.
Both subprojects have relied upon both linguistic and biological approaches. The linguistic project was centered on cultural evolutionary simulations, while the bird song project relied upon a linguistic concept of syntactical analysis. We believe that this interchange of methods may be almost as important as the interchange of concepts, since it allows results from natural and social scientists to be more fruitfully compared. In our project, this interchange of ideas was the result of two processes: the linguistic models were constructed as a direct collaboration between one biologist (rfl) and one linguist (fl); both subprojects benefited from structured regular meetings between all four participants.
The bird song subproject is now officially over. Over the next two months however, the major results from this project will be submitted for publication in high-impact journals. The software package Luscinia will continue to be developed for the foreseeable future. For the linguistic subproject: a thesis is planned for this summer, and papers from this will also be submitted in the near future.
a) The aim of this project was to explore how bird song syntax and structure evolved in a well-studied series of isolated chaffinch populations. Our main result over the past year is that we have completed our analysis and demonstrated precisely how chaffinch song has evolved over the course of a colonization chain. The most interesting change from the perspective of our project was that song syntax progressively disappears. In the mainland population, chaffinch song conforms to certain structural rules: certain categories of element appear at certain positions within the song. In the island populations, elements progressively lose their association with particular positions within the song. The fact that this loss of structure occurred progressively, as the chaffinches successively colonized new island groups is important: it suggests that these changes occurred during the process of colonization itself. The project has also discovered a parallel evolution in the process of song learning: whereas mainland chaffinches accurately learn whole song-types, the island populations instead progressively learn song-types less accurately: Azorean chaffinches incorporate syllable recombination; by the Canary Islands, syllable merging and splitting becomes prevalent.
b) We have successfully addressed point 4 in the original proposal.
c) We quickly discovered that pre-existing software analysis packages would not suit our requirements for this project: either they were not well-designed for noisy field recordings, or they did not offer the powerful analysis methods we needed. Luscinia was developed as a response to this. It is now being used in around a dozen research project, and has had an enthusiastic response from other bioacoustics researchers.
d) The progressive loss of syntactic structure in the Atlantic Islands requires further experiments in order to be fully understood.
e) Four papers are being written from this work, and will be submitted this spring. They concern i) the evolution of chaffinch song structure; ii) the evolution of chaffinch song learning; iii) chaffinch song adaptation to environmental conditions; and iv) song structure and syntax in zebra finches.

Subproject B (human language):
a) We have started a study on productivity of so-called adjective-noun (AN) compounds and word phrases in Dutch and German. Both languages can form both types of AN combinations (‘hoogspanning’, ‘donkere kamer’, ‘Dunkelkammer’, ‘Grüne Welle’). However, German has a ‘preference’ for compounds whereas Dutch has a ‘preference’ for phrases.

This phenomenon is linguistically interesting for the following reasons: (1) free variation is often a sign of a transitory phase in which there is a gradual fixation of one of the two variants. (2) although the AN-phrases (‘zwarte markt’) are not words, they seem to be stored in the lexicon instead of being rule-created during production. It has been proposed to consider these lexical phrases to be constructions. (3) in two related languages, there is a different distribution of the two variants.

These aspects are investigated by taking an evolutionary approach. With an agent-based computer simulation, we have studied the factors that cause fixation of one of the two variants, and the creation of the ‘AN-construction’. So far, it seems that there is a selection pressure for the (Dutch situation of a) preference for word phrases, unless the presence of case (as in German) acts as an opposing selection pressure.

A 3-month visit to the Department of Dutch Philology at the Freie Universität Berlin has been carried (april-june 2006) out to study the phenomenon together with Prof. Dr. Matthias Hüning, who has published about AN-combinations, and is starting a major project on the topic in 2007.
The project has lead to a continuing collaboration between the behavioural Biology Group and the Linguistic department at Leiden University. This is witnessed by the fact that the new founded Leiden Institute for Brain and Cognition (LIBC) has awarded a proposal for a joint PhD-project of the linguist Claartje Levelt and Carel ten Cate which will compare birdsong development and speech development in human infants.
Carel ten Cate is currently on a sabbatical leave at the NIAS where he will be working on a comparative project on vocal learning and syntax structure in song and language.