Generative metrical theory is a branch of generative linguistics that seeks to describe the rules of poetic meter by analogy with the rules of syntax. The guiding questions behind generative metrical theory are:

  • What are the minimum (i.e. necessary and sufficient) conditions for a line to be "metrical"? (These conditions should be independent of performance practice, which is necessarily variable; i.e., whether or not a line is metrical should not depend on how it's read aloud).
  • What makes one metrical practice more "weird" than another? (Why is it weirder to have a "trochaic substitution" in the middle of a line than at the beginning?)

Traditional descriptions of poetic meter have been precisely that, descriptions, without any theoretical backing. To take an example of iambic pentameter from the Poetic Meters node,* the following line by Elizabeth Barrett Browning is declared to be iambic pentameter because "it consists of alternating unstressed and stressed syllables, starting with an unstressed syllable, and ending with a stressed one"; that is, that it has five rising duple feet (iambs):

How do I love thee? Let me count the ways.

Since it is a line of monosyllables, in which all stress must be determined by lexicality and phrasal structure, this is not a particularly good example of iambic pentameter. Depending on context, stress could go all over the place:

How do I love thee? Let me count the ways!

But ideally, a theory of meter would not depend on performance or context. One definite constraint on stress is in polysyllabic lexical words: you can say "parody," but not "parody." We know this because English words have certain acceptable stress patterns, the rules of which we may not be able to spell out, but which, as native speakers (assuming we are native speakers), we can intuit.

But let's look at another line, which is also universally scanned as iambic pentameter:

Shall I compare thee to a summer's day?

Here we have some definite constraints on stress: compare and summer must be stressed. And if we lay out the "ideal" template for an iambic pentameter line in terms of weak and strong positions:

Shall I compare thee to a summer's day? 
| W S |W S | W S|W S |W S|

We can see that the second syllable of "compare" and the first syllable of "summer," that is, the two syllables that are definitely stressed, both fall into strong positions. But what about the other strong positions? If you were reading the line out loud, would you really stress "to"? Hopefully not, and Shakespeare's line is a counterexample to the claim that an iambic pentameter line is really a line of ten alternating unstressed and stressed syllables. In fact, almost no line of iambic pentameter has such a structure, and if it did, it would be profoundly annoying. According to generative metrical theory, iambic pentameter is a template of ten alternating weak and strong positions with certain rules about how actual lines can map into those positions ("correspondence rules"). Roughly, the rules are that the primary stress of a polysyllabic, lexical word cannot fall in a weak position, but any kind of syllable can appear in a strong position.

There are about a million refinements to this description of iambic pentameter involving the nuances of phrasal stress, word groupings, phonological structure, foot and colon (or half-line) structure, and syllable count. For many Anglophone poets, including Shakespeare, for instance, the maximum content of a position is not a syllable, but a moraic trochee, a phonological unit of weight. These are the sorts of questions about meter that generative metrical theory seeks to flesh out.

This description has been based on the work of Kristin Hanson, who bases her work on research by Halle and Keyser, Liberman and Prince, and Kiparsky, among others.

*(This is not, of course, to trash the Poetic Meters node, which is useful, but to offer comparison to traditional descriptions of meter to elucidate what generative metrical theory is trying to do differently.)