Monday, 2 April 2007

On counting English grammar

A US correspondent writes to ask how I arrived at the figure of 3500, when I talk (in various places) of 'how much grammar' there is in English.

I never intended this estimate to be taken too seriously, I must say. It was an attempt to give some sort of answer to the question, which I've often been asked by those learning English as a foreign language, 'How many grammatical rules are there in English?' It is possible to give an answer about the number of sounds in English and the number of words (see the estimates discussed in the lexicon section of The Cambridge Encyclopedia of the English Language, or in Words, Words, Words), so why not grammar?

In a way, there is no answer, because it depends on how you analyse grammar. Someone who uses a generative model, for instance, will end up with a very different total from someone who uses a different approach. Some theoretical linguists would say that the question is meaningless, or pointless. But that answer never satisfies the general enquirer. And there is a real sense in which the question does have meaning, as anyone knows who has slogged their way through the sections of a grammar book in learning a foreign language.

What I did was base my count on the biggest reference grammar I know - A Comprehensive Grammar of the English Language, by Quirk, Greenbaum, Leech & Svartvik. It's a 'reference grammar' - that is, an attempt to describe all the features of grammar in English, in much the same way as an unabridged dictionary is a 'reference lexicon'. It didn't deal with all varieties of English - not all regional and occupational varieties, for instance - but as the amount of grammatical variation between varieties is relatively small, this shouldn't matter too much.

I know this grammar very well, because I was responsible for its index. It took a long time to prepare, because in fact there were two indexes - a preliminary one, compiled to help the four authors to standardize their terminology - and the final one, around 100 pages in length, which is printed at the back of the book. (An account of the indexing is provided in an article for The Indexer in 1986, available at, in the section on Indexing.)

All I did was go through that index and count up the number of entries that were making a general point - that is, excluding entries that were just lexical items. The total would be much greater if every irregular noun, verb, adjective, etc were separately counted. I ignored the number of section references to a point, so my count does not distinguish between long entries as well as tiny ones. Sometimes it wasn't clear whether an index reference to a section was making a separate grammatical observation or was an amplification of an earlier one. So the result is necessarily an approximation. Around 3500 'grammatical points'.

Is it useful? I once asked a group of people how many points of grammar there were in English, and answers ranged from 500 to 10,000. So yes, I think an exercise like this one has some value in sharpening our intuitions on the matter. But the figure shouldn't be taken too literally.


Jake said...

As a side question, is there any estimate to the number of phonological rules in English?

And how does the number of phonological and grammatical rules stack up compared with other languages?

DC said...

I don't know of any. It's difficult, because there's no standardized phonological analysis to rely on, but many different methods. Everything depends on what counts as a 'rule'. But the numbers are likely to be large - for instance, there are around 300 consonant-cluster possibilities in English.

I don't know of any comparative data. Here too, it would all depend on which grammatical model was being used. And languages with a complex morphology would present a different range of issues.