Word Order Is a Compression Code

01The evidence

• Phonological universals can only be tested with cross-language comparisons of phonetic and phonological data, and the field has become explicit that inventory size is a variable to be measured rather than assumed.Ohala and Jaeger
• Language universals are conventionally described as tendencies in ordering, and the explanations offered are functional or historical rather than quantitative.Language universals overview

02The tension

The distribution of basic word orders is heavily skewed toward subject before object, and the functional explanations, such as putting the topic first, do not predict the quantitative shape of the distribution. Historical explanations predict that related languages cluster, which they do, but clustering does not explain why unrelated families converge.

03The theory

A sentence is a code for an event. The speaker pays a cost per morpheme and the listener pays a cost for ambiguity. Word order is the part of the code that resolves which participant did what, and it is only necessary when case marking or agreement is insufficient.

Languages with large phoneme inventories can afford longer, more distinctive morphemes, and can therefore mark roles with morphology. Those languages can afford free order. Languages with small inventories have shorter morphemes, which are more confusable, so they must resolve roles positionally. Order becomes rigid.

That predicts a correlation between inventory size and ordering freedom, and it predicts the direction: larger inventories should show freer order, smaller ones stricter order, with the skew toward subject first emerging because the subject is the most predictable participant and therefore the cheapest to place first.

The claim is quantitative in a way functional accounts are not. It says the exponent relating inventory size to ordering rigidity is measurable and should be similar across unrelated families, because it follows from a channel capacity argument rather than from culture.

04What this is not

05Predictions

• P1Languages with larger phoneme inventories should show measurably freer word order after controlling for family and contact.
• P2A compression model parameterised by inventory size should reproduce the observed distribution of basic orders across a large sample.
• P3The exponent relating inventory size to rigidity should be similar across unrelated families.

06What would refute it

If order is predicted by phylogeny alone, with no residual relation to inventory size, the compression channel is not doing the work.

07What it would change

It would make typology quantitative and connect phonology and syntax through a single information-theoretic parameter.