[Author Login]
Title:Folk Verse Form in English
Authors:Bruce Hayes, Margaret MacEachern
Abstract: Folk Verse Form in English



Bruce Hayes & Margaret MacEachern




The paper described here is primarily about metrics, but may

be of interest to optimality theorists in general.

We examine the inventory of possible quatrains in the verse

of English folksongs. The data focus is on the distribution of

truncated lines (empty positions at the end of the metrical grid)

within the quatrain. We attempt to explain why only about 25 of

the 625 logical possibilities for quatrain structure are well


Our assumption is that the anonymous composers of folk songs

internalize a set of conflicting well-formed constraints. Some

require empty positions near the end of the line, with the goal

of enhancing the saliency of units like couplets and lines.

Others, part of metrics proper, demand that the metrical grid of

the line be filled with syllables. The diversity of quatrain

types is held to follow from the large number of ways in which

these constraints may be ranked against one another. Thus the

outcome of our analysis is not any particular ranking of the

constraints, but rather the entire factorial typology.

We extend our analysis to account for two further matters:

the greatly differing corpus frequencies of the various quatrain

types, and the existence of quatrains that sound only partially

well formed.

Corpus frequencies are modeled by assigning each constraint

a range of possible strictness values, in arbitrary units. A

simulation program calculates the expected frequencies in a large

hypothetical corpus of quatrains, under the assumption that each

quatrain is generated by a random setting of all the constraints

within their respective strictness ranges. By locating

appropriate strictness ranges, we can model the corpus

frequencies fairly accurately.

These strictness ranges are then used as the basis for a

model of gradient well-formedness judgments: we claim that the

strictness ranges of certain constraints are partitioned into

central and peripheral subranges. Any form that may be derived

only by ranking a constraint within its peripheral subrange will

sound only partially well-formed. In informal terms, a given

form is judged as semi-acceptable when the consultant's

internalized grammar must be slightly "bent out of shape" in

order to generate it. Following this hypothesis, we have been

able to formulate an explicit analysis for the semi-well-

formedness of a number of quatrain types.

At the moment we are fairly enthusiastic about this approach

to semi-well-formedness, since it requires no "fuzziness" within

the constraint system itself, but only within the rankings, which

can easily be treated numerically, hence gradiently. We are

curious if this method could be applied to the many other areas

of linguistics where gradient well-formedness judgments prevail.


Type:Paper/tech report
Article:Version 1