I’ve been trying to write a new section of my grammar, and I’m hitting the old problem of whether to organise the data in functional or structural terms. Traditional grammars tend to go for a very structural approach. For example, if you want to know about how a language expresses modality, you might have to piece it together from chapters or sections of different chapters on:

• Verbal morphology and inflectional mood
• Auxiliary verbs
• Serialised or complement taking verbs (if different in behaviour from auxiliaries)
• Adverbs
• And so on

You may even have to resort to the dictionary if the grammar writer feels that their grammar shouldn’t describe lexical items which don’t show any special behaviour worthy of note.

I generally prefer to write chapters organised along functional lines. Of course you need to gather together data about formal systems in one place – it wouldn’t be much good if you could only find out how to inflect a verb by going to five different chapters either. But once you’ve got past a basic discussion of formal structure, I think it’s more helpful (or at least interesting!) to focus chapters in a functionalist-typological way which focuses on how a given task is accomplished in a way that cross-cuts the formal mechanisms involved. The ideal structure of a grammar for me would be:

• Ch 1: Overview of structural area 1 (e.g. phonology)
• Ch 2: Overview of structural area 2 (e.g. word classes)
• …
• Ch N: Overview of functional area 1 (e.g. spatial expressions)
• Ch N+1: Overview of functional area 2 (e.g. temporal expressions)
• …

The second half of the grammar relates the formal structures described in early chapters to the functional needs of any human language and synthesises the individual parts of the language into wholes which serve some kind of purpose.

This approach is common in typology books which are not grammars. To take an example, Subordination by Cristofaro gives a possible definition of subordination in terms of states of affairs (SoAs) as follows:

By subordination is meant a situation of functional asymmetry whereby the profile of one of two linked SoAs is overridden by the other

To illustrate what this means, consider the clause “I saw him go”. It contains two SoAs, an observation and an observed action, but what is described overall is the seeing and not the going. By adopting this functional definition, the book can then describe how different languages accomplish the translational equivalent of complement, adverbial and relative clauses even if these do not appear to be formally distinct from each other or from coordinated clauses in some languages.

But the challenge with writing a grammar for a specific language (instead of a cross-linguistic typology) along broadly functional lines is what underlying typology to use and where to draw the lines. There is always a tension between bringing together clearly related areas into some kind of (hopefully) insightful joined up analysis on the other hand, and the clarity of having a manageable chunk in the other, and between drawing the lines in a way that fit the language vs drawing the lines in a way that meets the reader’s pre-existing expectations for how grammars should be structured.

I was intending to get on to my actual specific problem in this post, but it’s 1am now so I guess I’ll follow this up tomorrow.