This study re-evaluates the representation of phonological length in Finnish in the light of Optimality Theory. The Obligatory Contour Principle (OCP) prohibits sequences of identical elements. It follows that long segments cannot be represented as sequences of two short segments, but must be interpreted monosegmentally (i.e. non-linearly). In Finnish, long segments have been traditionally represented bisegmentally, consequently violating the OCP. There are, however, phenomena in Finnish, such as dialectal epenthesis, which pattern long consonants with short ones rather than with bisegmental consonant clusters. This study provides an Optimality Theoretic analysis for dialectal epenthesis, and shows that the analysis predicts that monosegmental representation is the optimal way of representing long consonants in Finnish as well.

This paper shows that Cognitive Grammar is valuable for typological studies. The Norwegian and Russian reflexive-middle-passive systems are analysed comprehensively and compared. Whereas Cognitive Grammar is compatible with the typological tradition (represented by Kemmer), it needs amendment in certain important respects. The main theoretical contribution of the paper is that the Cognitive Grammar notions of “instantiation”, “prototype”, “entrenchment” and “extension” are useful; without these notions, interesting generalizations about the Norwegian and Russian systems are inexpressible. The main empirical contribution of the paper is that a full-fledged analysis of the Russian system is presented. In addition, Kemmer's analysis of Norwegian is improved upon, partly because findings from the generative tradition are incorporated.

This article presents an explanation for the lack of some neuter gender forms in an otherwise productive paradigm for the formation of neuter gender adjectives. The explanation draws on a proper understanding of (exemplar-based) analogy under some conditions. One condition is that the problematic form resembles a free morpheme by the loss of a morpheme/syllable boundary in its derivation. The second condition is that the problematic forms tend to share some property (for example, referring to a nonverifiable property such as internal (mental) states). Problematic forms might thus mutually support other such forms by analogy and similarity, and thus create a paradigmatic gap. It is argued that exemplars (with context), and analogy between similar exemplars, necessarily restrict and licence the use of derived forms.