Constraints Measure Metrical Intuitions

Nila Friedberg
University of Toronto

In this presentation I demonstrate how recent constraint-based theories in phonology and metrics (Golston and Riad 1996, Hayes and MacEachern 1998) can be used to account for speakers' intuitions about complexity of poetic meter.
It is well-known that speakers of a language have intuitions about metrical well-formedness (Kiparsky 1975, Hayes and MacEachern 1998). Moreover, speakers have intuitions that some poets' meter sounds more complex than others. Most Russian speakers agree that Pushkin's meter sounds more elaborate than Nekrasov's, although both employ the same absolute restrictions and deviate from the ideal iambic template just as often. Such intuitions about complexity in meter remain largely unexplained either by literary criticism or by generative approaches to meter (Halle and Keyser 1971, Kiparsky 1975). I propose a theory that measures this complexity and apply it to the omission of stress in Russian iambic tetrameter. Following work in constraint-based theories of metrics (Hayes and MacEachern 1998), I employ stochastic constraints (i.e. constraints responsible for generating frequencies) in order to model metrical preferences of XVIII-XIX century Russian poets. However, I go beyond earlier proposals and demonstrate that metrical complexity can be measured using three explicit parameters - the number and type of constraints actively employed by the poet, and the number of ways a hierarchy of poet's preferences can be achieved. I examine a large statistical database of Russian iambic tetrameter (Taranovski 1953) and formulate weighted constraints, generating hierarchies of metrical preferences for 37 Russian poets at different stages of their lives.
Ideally, weak (W) and strong (S) positions in iambic meter correspond to unstressed and stressed syllables; however, Russian poets often violate the pattern by filling strong positions with unstressed syllables. For example, in (1) the penultimate strong position is filled by an unstressed syllable 'za' (stressed syllables are capitalized)

Metrical positions       W      S  W       S      W   S    W     S             
Actual rhythm           Kog  - DA ne  v  SHUT -  ku  za  - ne - MOG         
                        when            seriously       [he] became ill
                                        Pushkin, Eugene Onegin, 2

(1) shows four sites of potential stress omission. No poets in the database omit stress on the last foot. However, they omit stress on non-final feet; omission on the penultimate foot is the most favoured type. Preferences for omission on first, first and third, second foot, etc., vary among poets. I organize attested omission types into frequency hierarchies, yielding 11 patterns.

I derive these patterns using weighted constraints on well-formedness, some of which aim to make salient edges of metrical constituents (Hayes and MacEachern 1998) and others to govern the number of stresses in each constituent. The frequency of a line type depends on the weight of the appropriate constraint in a poets grammar. Not all constraints are weighted - some are inviolable, such as the requirement that the end of the line be salient.
Looking across poets and times, two types of preference-patterns emerge - SIMPLE patterns (like Nekrasov's) are generated by four constraints, COMPLEX patterns (like Pushkin's) are generated by five. Simple patterns can be derived by weighting constraints in several distinct ways. Complex patterns can be derived by just one solution in constraint ordering, making it harder for a poet to arrive at that result. Both complex and simple patterns actively utilize constraints requiring the saliency of edges, but only complex patterns make use of the constraint Symmetry, requiring paired half-lines to have identical rhythmical structure. The model shows that constraints can be helpful in linguistic research not only in generating desired outputs, but also in measuring speakers intuitions about verse. From the point of view of metrics, this approach objectively measures metrical complexity instead of intuitively stating that the meter of some poet 'sounds more elaborate'.


Golston, Chris, and Tomas Riad. 1995. Direct Metrics. Ms., University of Duesseldorf and Stokholm University.
Halle, Morris, and Samuel J. Keyser. 1971. English stress, its form, its growth, and its role in verse. New York: Harper and Row.
Hayes, Bruce, and Margaret Maceachern. 1998. Folk verse forms in English. Language 74, 473-508.
Kiparsky, Paul. 1975. Stress, Syntax and Meter. Language 51, 576 617.
Taranovski, Kiril F. 1953. Russki Dvodelni Ritmovi. Beograd: Srpska Akademija Nauka.

< Back to List of Abstracts | Back to PLC23 Home Page >

About the PLC23 Committee
Previously held Penn Linguistics Colloquium: PLC22 (1998), PLC21 (1997)

Penn Department of Linguistics
University of Pennsylvania