Word-internal modification: The case of the Persian comparative marker

1. Introduction

Traditionally, morphemes have been divided into derivational and inflectional classes, in what is sometimes referred to as the Split Morphology Hypothesis (Matthews 1972, Wasow 1977, Anderson 1982, Perlmutter 1988, Beard 1998, Stump 1998).¹ According to this split, derivational morphemes, unlike inflectional ones, change syntactic class, involve idiosyncratic meaning and do not interact with syntactic rules. In addition, it has been suggested that (productive) inflectional morphology should always be peripheral to derivational morphology. Under this view, the grammaticality contrasts in (1) follow from the order between inflectional morphemes (exemplified by plural and comparative marking) and category changing derivational morphemes.

1. (1).

The idea that derivational morphemes change lexical or syntactic category is challenged by the old observation that many affixes involved in deriving new word forms (hence classified as derivational) do not necessarily change the lexical category of the word (Beard 1998). Some well-known English examples are un- in unhappy, re- in rewrite, -ish in greenish, among many others (see Lieber 1980, Selkirk 1982, Newell 2008, a.o.). Unlike other derivational affixes (e.g., -ness, -ment, etc.) which change the lexical category of the base they attach to by projecting their own categorial feature, these affixes do not project; as such, the category feature of the base is maintained. These affixes can therefore be seen as examples of word-internal modification in the domain traditionally known as derivation (see, for example, Newell 2008, who calls them "morphological adjuncts"; also Lieber 1980).²

A sharp split between derivational and inflectional morphology, as suggested by the Split Morphology Hypothesis, has been challenged further in the past few decades, with various linguists presenting cases of morphemes typically categorized as inflectional showing properties associated with derivational morphology (see, for example, Scalise 1984, Booij 1993, Rainer 1996, van Marle 1996, Bobaljik 2005). Most relevant to the present paper is Wiltschko's (2008) account of plural marking in Halkomelem Salish. Plural marking is often presented as a canonical example of inflectional morphology in the nominal domain (Booij 1996, Stump 1998, a.o.). Inflectional morphology is taken to be what is relevant for syntax, an idea captured in the generative tradition of the past few decades by the instantiation of a functional category. Ritter (1992), for example, takes plural to be the instantiation of a functional head associated with number marking. Crucially, Wiltschko (2008) shows that plural marking in Halkomelem Salish has distributional properties quite distinct from those of its inflectional counterpart in a language like English, even though the semantics of plural is quite similar in the two languages. In Halkomelem, plural marking is not obligatory, does not trigger obligatory agreement, and can be found inside compounding and derivation. She concludes that the functional head analysis of plural marking in languages such as English cannot be extended to Halkomelem. Instead, she argues that the distributional properties of Halkomelem plural derive from its distinct syntax, which she analyzes as word-internal modification. Hence, the existence of word-internal modification, already established in the domain of derivation, is extended to inflection.³

This squib investigates the Persian comparative marker in light of the above discussion. It is worth noting that comparatives (and superlatives) are often presented as canonical examples of inflectional morphology across languages (Booij 1996, Stump 1998, among others).⁴ Persian comparatives are also classified as inflectional (see, for example, Mahootian 2002, Kazemian and Hashemi 2014). As we will see below, while the Persian comparative marker has similar semantics to its English counterpart, it behaves in many ways like a non-inflectional suffix; that is, it cannot be treated as a projecting functional head. As a result, the comparative form of an adjective has a distribution very similar to that of a simple adjective. It will be argued here that these properties of the Persian comparative can best be accounted for as word-internal modification, similarly to Wiltschko's proposal for the Halkomelem plural.

2. Persian Comparative: Some Properties

Persian has a highly productive comparative marker –tar, exemplified in (2).

1. (2).

The comparative marker –tar behaves like a derivational suffix in Persian...