Metrics and Optimality Theory: The Case of Russian Verse

Nila Friedberg, University of Toronto

Optimality Theory (OT - Prince and Smolensky 1993) has been recently employed in the analysis of poetic meter by many authors, such as Hayes and MacEachern (1995), Golston and Riad (1995), among others. However, so far there are no OT accounts of Russian meter, which has been traditionally studied from a statistical (Belyi 1910, Taranovski 1953,Tomashevskii 1921), but not a generative point of view. Russian meter presents an especially challenging case for generative approaches to poetry (Halle and Keyser 1971, Kiparsky 1975), since Russian iambic verse has very salient metrical tendencies, but very few strict rules.

In this presentation, I examine the inventory of possible rhythmical forms in the 18th-19th century Russian iambic tetramater, and suggest that the diversity of metrical grammars derives from different rankings of constraints on well-formedness. Thus, the aims of the paper are twofold: on the one hand, I outline a constraint-based analysis of Russian meter, on the other hand, I suggest a modification of OT that allows to model statistical tendencies.

In Russian iambic meter, stress is often ommitted on strong positions, i.e., positions which are ideally expected to be stressed. Stress omission is due to the fact that the majority of Russian words are long. For example, in (1) an unstressed syllable /za-/ occupies a strong position ('S' - strong position, 'W' - weak position; the boundaries of iambic feet are marked by a dash; stressed syllables are capitalized):

Line Kog-DA ne v SHUT-ku za-ne-MOG
Metrical W S / W S / W S/W S

According to Taranovski (1953), 18th century poets ommit stress on the second foot in a line more often than on the first, whereas in the 19th century, the opposite situation holds.

I hypothesize that poets internalize a set of conflicting constraints referring to different structural positions, such as edges of Line or Colon (Colon is a half of a line; in case of a tetrameter line, a colon is a constituent containing two feet). In the 18th century, the first foot was stressed more often than the second since the salience of the left edge of a line was considered more important than the salience of the right edge of a colon. In other words, in the 18th century, the constraint MarkLeft(Line) was ranked higher than MarkRight (Colon), whereas in the 19th century, these constraints were reranked, and poets started stressing the second foot more often than the first.

The problem for the traditional OT approach is that both line types a. and b. in the tableau were possible in the 18th century. In reality, there is no fatal violation of candidate a. - it is simply less statistically preferred than candidate b. In order to account for this fact, I adopt the proposal of Golston and Riad (1995), and suggest the following modification of OT: if a candidate violates a high-ranked violable constraint in the grammar, it is statistically rare; if it violates only a low-ranked constraint, it is statistically frequent. To summarize, the departure of this model from OT is that (i)OT makes no predictions about how frequent the winner is; (ii) In OT, a candidate that emerges as less harmonic than others, is a loser; in our model, such candidate is simply less statistically frequent than others.

Extending the analysis further, we examine the metrical preferences typical of all Russian verse, the metrical preferences typical of only certain periods, and the preferences specifying individual poets. We suggest additional constraints such as Contrast (a line must contain at least one omission of stress), Binary Colon (at least one colon in a line must contain two stressed feet), StressS (do not omit stress anywhere). Employing different rankings of all of the constraints suggested above, we succeed in generating the hierarchies of preferences of all 24 Russian poets mentioned in Belyi (1910).The fact that the line with no stress omissions was never the most preferred type for any poet, is captured by the fixed ranking Contrast >> StressS.

The last issue to be addressed is the status of strict rules. Certain types of stress omission, such as omission of stress on the last foot in a line, are unattested for most 18th-19th century poets (Taranovski 1953). Following Hayes and MacEachern (1995), I suggest that metrical constraints should be divided into violable and inviolable. The constraints responsible for deriving statistical tendencies are VIOLABLE, and always ranked below the inviolable ones.

Thus, the computation of meter may be viewed as follows: first, the inviolable constraints rule out all the impossible line types; afterwards, the remaining well-formed lines are evaluated by the violable constraints and classified into more preferred and less preferred line types.

This approach to meter has the following advantages. First, it shows us that statistical tendencies are not a matter of accident - rather, preferences derive from specific rankings of constraints operating in poets' minds. Second, this model demonstrates that the 'statistical' and the 'generative' approaches to meter (which are generally considered conflicting) can be reconciled, since metrical tendencies constitute an important part of poets' metrical grammars. From the point of view of theoretical linguistics, the model shows a possible way of incorporating statistical tendencies into OT.


Belyj, Andrej. 1910. Simvolizm: kniga statej. Moskva: Musaget.

Golston, Chris and Thomas Riad. 1995. Direct Metrics. Ms., University of Dusseldorf and Stockholm University.

Golston, Chris. 1998. "Iambic Pentamater is Neither." Paper presented at LSA Annual meeting, New York.

Halle, Morris and S.J.Keyser. 1971. English Stress, Its Form, Its Growth and Its Role in Verse. New York: Harper and Row.

Hayes, Bruce and Margaret MacEachern. 1995. Folk Verse Form in English. Rutgers Optimality Archive 119-0000.

Kiparsky, Paul. 1975. "Stress, Syntax and Meter." Language 51.576-617.

Prince, Alan and Paul Smolensky. 1993. "Optimality Theory: Constraint Interaction in Generative Grammar." Ms., Rutgers University and University of Colorado, Boulder.

Taranovski, Kiril. 1953. Ruski dvodelni ritmovi. Beograd.

Tomashevskij, Boris. 1921. Russkoe stixoslozhenie. Munich: Vilhelm Fink Verlag.