Predicting the Direction of Derivation in English Conversion

Conversion is a word formation operation that changes the grammatical category of a word in the absence of overt morphology. Conversion is extremely productive in English (e.g., tunnel, talk). This paper investigates whether distributional information can be used to predict the diachronic direction of conversion for homophonous noun–verb pairs. We aim to predict, for example, that tunnel was used as a noun prior to its use as a verb. We test two hypotheses: (1) that derived forms are less frequent than their bases, and (2) that derived forms are more semantically specific than their bases, as approximated by information theoretic measures. We find that hypothesis (1) holds for N-to-V conversion, while hypothesis (2) holds for V-to-N conversion. We achieve the best overall account of the historical data by taking both frequency and semantic specificity into account. These results provide a new perspective on linguistic theories regarding the semantic specificity of derivational morphemes, and on the morphosyntactic status of conversion. 1 The Morphology of Conversion Word formation operations that change the grammatical category of a word in the absence of overt morphology pose interesting linguistic challenges. Such operations are highly productive in English, especially between the categories of noun and verb (consider examples such as tunnel or walk, also more recent examples such as email). This phenomenon is also observed in other languages, for example German (Vogel, 1996), Dutch (Don, 1993), Hungarian (Kiefer, 2005), and Bulgarian (Manova and Dressler, 2005). We call these cases “conversion”, without any theoretical commitment. Conversion in English, especially N-to-V conversion, has been extensively studied within morphology and syntax. Historically, accounts for this phenomenon involved (a) conversion (proper), a category-changing word-formation operation assumed to be different from other types of derivational morphology (Koziol, 1937; Bauer, 1983; Plag, 1999), or (b) zero-derivation, involving a derivational affix akin to -ize, -ify, -er, etc., but which happens to be phonologically null (Bloomfield, 1933; Marchand, 1969; Kiparsky, 1982). Most current syntactic theories of conversion, though, are based on underspecification. This is the idea that meaning-sound correspondences are stored in the lexicon as uncategorized roots and obtain a grammatical category when combined with syntactic heads, which may be phonologically null or overt (Halle and Marantz, 1993; Pesetsky, 1995; Hale and Keyser, 2002; Borer, 2005a; Arad, 2005). The range of meanings resulting from conversion has been a key source of evidence for various theoretical approaches. Verbs derived via N-to-V conversion can have a wide variety of meanings, seemingly constrained only by the template ‘action having to do with the noun’, such as in the phrase celluloid the door open, meaning ‘use a credit card to spring the lock open’ (Clark and Clark, 1979). V-to-N conversion has a narrower semantic range. It is likely to result in a noun referring to the event described by the verb or to its result, e.g. talk (Grimshaw, 1990). This paper presents a computational study of conversion: we study which factors are able to account for diachronic precedence in cases of English V-to-N and N-to-V conversion. The goal is to predict, e.g., that tunnel was originally used as a noun and walk as a verb. Historical precedence provides a theory-neutral ground truth which we treat as a proxy for the actual direction of conversion.