What can (can’t) come next?


Is this the reason for grammatical gender?

… grammatical gender appears superfluous to a lot of people. But, studying Dutch, one of the favourite kinds of questions is fill in the blank sentences. For example:

Hij leest de _____.
He reads the _____.

There are a number of coherent possible answers, but the obvious one in English – book – is ruled out by the article de. Boek has the neuter gender, and cannot have de as its article. Krant – newspaper – can go here nicely though.

McWhorter suggests that features like grammatical gender – among other things – are a sort of cruft that a language accumulates over time – that there is a kind of universal basic language structure that doesn’t have those things, and that creoles all, to some extent, resemble this basic grammar. It is a kind of resurection of Bickerton’s universal language bioprogram. Now, this is simply wrong if one considers languages like Michif creoles – the position that McWhorter takes in his book. Michif has absorbed all the complexities of French nouns and of Cree verbs at the same time, making it by his standards an incredibly complex language. There are ways to get out of this hole, but they all involve seeing Michif as something fundamentally different from a creole. As far as I can tell, Mufwene’s approach still seems to explain the data better.

But the thought that occured to me was that complex and seemingly unnecessary features like grammatical gender can serve to reduce noise in oral communications. Languages that mark grammatical gender in articles, like Dutch or French, sharply reduce the number of words that can follow them given their context, while the amount of information that has to be stored in the lexicon to make this gain in noise reduction is very small. In French with just two genders, it’s just one bit of information per noun. I suspect that by losing grammatical gender, Haitian Creole has had to create other distinctions to compensate. The thing is, given a reasonable corpus, this hypothesis should be empirically confirmable.

Presumably other grammatical features could be evaluated in the same way. Furthermore, instead of simply offering up a more sophisticated version of Martinet’s hypothesis, it seems to me one could recast his core claim: Feature loss in languages should be less likely to occur to the features that most add to noise robustness. This hypothesis can cover a lot of ground – it can explain creolisation not as incomplete acquisition of substrates, but as the result of speakers with different language abilities – some fluent, some marginal – devising mechanisms to increase the noise robustness of their language.

I.e., grammatical gender helps the listener to follow what the speaker is saying by making it easier to anticipate what words can (and can’t) follow a particular — masculine or feminine — article.


Leave a Reply

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

WordPress.com Logo

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

Google+ photo

You are commenting using your Google+ 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 )

Connecting to %s

%d bloggers like this: