The self-organization of phonological patterns  

Pierre-Yves Oudeyer 

Sony CSL Paris, e-mail: py@csl.sony.fr

Human vocalizations have a complex organization.  They are are digital
and compositional: they are built through the combination of units, and
these units are systematically re-used from one vocalization to the
other.  These units appear at multiple levels (e.g.the gestures, the
coordination of gestures, the phonemes, the morphemes).  While for
example the articulatory space that defines the physically possible
gestures is continuous, each language only uses a discrete set of
gestures.  While there is a wide diversity of the repertoires of these
units in the world languages, there are also very strong regularities
(for example, the high frequency of the 5 vowel system /e,i,o,a,u/). 
The way the units are combined is also very particular : 1) not all
sequences of phonemes are allowed in a given language (this is its
phonotactics), 2) the set of allowed phoneme combinations is organized
into patterns.  This organization into patterns means that for example,
one can summarize the allowed phonemes of Japanes by the pattern "CV": a
syllable must be composed of two slots, and in the first slot only the
phonemes belonging to a group that we call "consonnant" are allowed,
while in the second slot, only the phonemes belonging to the group that
we call "vowels" are allowed. 

It is then obvious to ask where this organization comes from.  There are
two complementary kinds of answers that must be given (Oudeyer, 2003). 
The first kind is a functional answer stating which is the function of
systems of speech sounds, and then showing that systems having the
organization that we described are efficient for achieving this
function.  This has for example been proposed by (Lindblom, 1992) who
showed that digitality and statistical regularities can be predicted by
searching for the most efficient vocalization systems.  This kind of
answer is necessary, but not sufficient : it does not say how evolution
(genetic or cultural) might have found this optimal structure.  In
particular, naive darwinian search with random mutations (i.e.  plain
natural selection) might not be sufficient to explain the formation of
this kind of complex structures : the search space is just too large
(Ball, 2003).  This is why there needs a second kind of answer stating
how evolution might have found these structures.  In particular, this
amounts to show how self-organization might have constrained the search
space and helped natural selection.  This can be done by showing that a
much simpler system spontaneously self-organizes into the more complex
structure that we want to explain. 

(Oudeyer, to appear) has shown how a system of this kind, based on the
coupling of generic neural devices which were innately randomly wired
and implanted in the head of artificial agents, could self-organize so
that the agents develop a shared vocalization system with digitality,
compositionality and statistical regularities.  We present now an
extension of this system which gives an account of the formation of
phonotactics and of the formation of patterns in the allowed phoneme
combinations.  The extension is based on the addition of a map of
neurons with temporal receptive fields.  These are intially randomly
pre-wired, and control the sequencial programming of vocalizations. 
They evolve with local adaptive synaptic dynamics. 

The system provides a necessary complement to the functionalist
explanation.  Interestingly, it does not require the presence of a
functional pressure for efficient communication.  It does not require
any social pressure and agents have no social skills at all in fact. 
While modern speech codes are obviously influenced by the function of
communication, the simplicity of the system allows to put forward a new
hypothesis for the initial invention of shared organized vocalization
systems: they might be a self-organized side effect of certain brain
structures evolved for other functions than communication.  We will
develop this hypothesis by explaining which are these brain structures
and what was their initial function. 


Ball P. (2001) The self-made tapestry, Pattern formation in nature,
Oxford University Press.

Lindblom, B.  (1992) Phonological Units as Adaptive Emergents of Lexical
Development, in Ferguson, Menn, Stoel-Gammon (eds.) Phonological
Development: Models, Research, Implications, York Press, Timonnium, MD,
pp.  565-604. 

Oudeyer P-Y.  (to appear) From Analogous to Digital Vocalization, to
appear in the volume "Evolutionary Pre-Requisites for Language", ed. 
Tallerman M., Oxford University Press. 

Oudeyer P-Y.  (2003) L'auto-organisation de la parole, Thèse de Doctorat
de L'Université Paris VI.