Combinatorial Structure Emergence

In this study by Tessa Verhoef, Simon Kirby and Carol Padden, the effects of iterated learning of an artificial language indicate that combinatorial structure is not necessarily the result of an expanding meaning space, and in fact may be simply a feature to increase learnability.

This study, presented at the 2011 Annual Conference of the Cognitive Science Society, used a slide whistle to create an artificial whistled language of twelve distinct whistled ‘phrases.’ These phrases were recorded and taught to one test subject over four listen & recall phases. The first test subject was then recorded, and this sequence of the twelve recalled whistles were the whistles that the next test subject listened to. This process was repeated through a chain of ten people: ten ‘generations.’ The ten-generation cycle was then performed on three more groups of ten people for experimental accuracy.

Several prominent linguists have proposed that combinatorial structure emerges as a result of the nature of language structure: the limited number of phonemes from which the language is formed can be combined in an infinite number of ways, but eventually the speaker will discover combinations in this infinitely expanding meaning space that are too close aurally to discriminate accurately between the two. Thus combinatorial structure emerges to differentiate the phoneme combination: if one of the indiscriminable phoneme combinations has a certain structure associated with it – such as a sentence structure that is always paired with the phoneme combination – which the other phoneme combination does not have an association with, the two phoneme sets become differentiated.

What the slide whistle study found was remarkable: even without semantic meaning of any kind associated with the twelve whistles – so no meaning space whatsoever – a combinatorial whistling structure emerged:

In our experiments combinatorial structure emerges long before the signal space is fully exploited.

This suggests that the theory of emergent combinatorial structure outlined above is incorrect.The whistlers in the later generations began to combine the structures of the whistlers they were learning from, such as combining staccato whistles with longer whistles when those two whistles had been taught separately. Additionally, after calculating the errors between the input and output sets in the late phases of the generations, in most of the generational progressions the percent error decreased as the emergent combinations began to occur, suggesting that the combinatorial structure spontaneously forming has more to do with memorization and processing than the expanding meaning space.

I’m including this study in the blog mostly because I think the findings are really fascinating. The discoveries made here seem to suggest that a combinatorial grammar does not need to be established to convey thought, as is mentioned in the introduction of the study, which I have not included here. The structure of language can be far more organic than Chomskyan generative grammar theory suggests, with a structure emerging from learnability and ease of memorization instead of as an internalized rule system.


Verhoef, Tessa, Simon Kirby, and Carol Padden. “Cultural Emergence of Combinatorial Structure in an Artificial Whistled Language.” Expanding the Space of Cognitive Science: Proceedings of the 33rd Annual Conference of the Cognitive Science Society. Boston, Massachusetts, July 20-23, 2011 by L. Carlson & C. Hoelscher & T.F. Shipley. Web. 21 Sept 2015.


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s