Finiteness and nonfinite forms

2. FINITENESS AND NONFINITE FORMS
Before we can begin to think about the syntax and semantics of nonfinite verb categories, we need to know what a nonfinite verb category is in contradistinction to a finite verb category; that is, we need a concept of finiteness. The term “finite” derives from the Latin grammatical tradition; its meaning was originally roughly “defined (in respect of person and number features),” thus distinguishing verba finita, which distinguish person and number, from verba infinita, which do not. In Latin, the verba infinita, nonfinite verb categories, are the infinitive, participles, gerund, and supine.⁶
This distinction remains essentially unaltered in the modern pretheoretic use of the term “finite,” except that other verbal features are commonly associated with finiteness, in particular, tense. Although there is very little inflectional morphology in English, it is still possible to use person, number, and tense to define finite verb forms in opposition to nonfinite verb forms: Finite verb forms are those which show an alternation between (usually) third-person singular and other person/number combinations (e.g., he/she eats versus I/they eat) and between simple present and past tenses (e.g., he eats versus he ate), whereas nonfinite verb categories show no person/number or tense marking, as in infinitives [(to) eat], participles (eating, eaten), and gerunds (eating). Modals make no person/number distinction, but syntactically pattern with unambiguously finite auxiliaries that do, and partially distinguish tense.⁷
Even if it were a simple matter to draw such a specific distinction for English or Latin, in a wider crosslinguistic perspective there is no natural categorization of verb forms by person, number, and tense, or indeed by any such feature, which would provide a meaningful crosslinguistic distinction between finite and nonfinite verb categories. For example, Lowe (2015) shows that there is both morphosyntactic and semantic expression of tense in participles in early Sanskrit; even in Latin, participles and infinitives show tense distinctions, and Latin and Sanskrit participles also distinguish number (though not person). Thus, the use of such features to define the finite/nonfinite distinction is undermined before we even look beyond three Indo-European languages. Other properties have also been associated with finiteness, such as mood by Anderson (2007); Bisang (2007) discusses the crosslinguistic relativity of finiteness marking, including, for instance, the correlation between politeness distinctions and finiteness in Japanese.
Alongside a feature-based distinction between finite and nonfinite, syntactic differences are often called upon. Specifically, a finite verb form is one which is capable of functioning as the sole verb form in a sentence, whereas nonfinite verb forms necessarily occur alongside (and in some way subordinate to) another (finite) verb.
However, some languages permit otherwise nonfinite verb forms to function as the only verb in a sentence, while others have verb forms which are restricted to dependent clauses but show no differentiation from matrix clause verbs in terms of (person/number/tense, etc.) marking (e.g., West Greenlandic; Nikolaeva 2007b, p. 3). The difficulty of establishing a crosslinguistically valid definition of finiteness has recently been addressed by authors such as Koptjevskaja-Tamm (1993), Cristofaro (2007), and Nikolaeva (2010, pp. 1177–79).⁸
The only way to preserve a notion of finiteness is to admit its crosslinguistic relativity, at least in specifics, and to approach it as a prototypical notion. Prototypical finite verbs are the heads of prototypically finite clauses; prototypically finite clauses are nondependent, nonmodal, nonnegative, assertive. Within any particular language, the verbal category most strongly associated with these sorts of clauses is the most finite verbal category in the language. Other verbal categories may then deviate from this in one or more ways.
Certain deviations from the language-specific finite verb prototype typically cluster in correlation with the properties of another lexical category in the language. For example, it is quite common for nonfinite verb forms to show typically adjectival properties, in those respects in which they differ from corresponding prototypically finite verb forms. The label “participle” (Lowe 2015) is standardly attached to such forms, although of course the specific properties of a participle in one language may differ significantly from those of a participle in another language, depending on the properties of prototypical finite verbs, and of adjectives, in the respective languages. Similarly, nonfinite verb categories may show typically adverbial, or typically nominal, properties. These are often labeled converbs (Haspelmath 1995) or action nominals/masdars/gerunds (Koptjevskaja-Tamm 1993, 2003, 2006, respectively).
Alongside infinitives (Wurmbrand 2001), which are less clearly associated with any one nonverbal grammatical category, these are the major categories of nonfinite verbs identified in the literature (Ylikoski 2003). However, this four-way distinction hides a much richer variation. As discussed in detail by Lowe (2015, pp. 316–27), nonfinite verb forms may show morphological properties more typical of one category but syntactic properties more typical of a different category. For example, the Classical Arabic converb is functionally adverbial, but shows adjectival agreement features; similarly, the Sanskrit participle is morphologically adjectival, but shows a range of syntactic uses, including both typically adjectival and typically adverbial functions.
Some nonfinite verb forms show nonverbal properties only in certain respects, while in other respects they are more similar to prototypical finite verbs. In several Indo-European languages, including Slavic, Romance, and Indo-Iranian languages, infinitives, which are morphologically adverbs, can be used as main verbs in imperatival (and sometimes other) constructions.² In many languages, what are morphologically participles (i.e., verb forms with morphological features typical of adjectives in the relevant language) can be used as finite verbs. For example, the active participle of Hebrew and Aramaic functions as a main verb, differing from the prototypical finite verbs of the language only in lacking person marking; similarly, in Classical Sanskrit the morphologically adjectival past participle in -tá- is regularly used as a main verb.
Purely morphological deviation from the standard of prototypical finite verbs is not possible, however. If a participle comes to be used exclusively as a main verb, it can no longer be considered as deviating from the finite verb prototype; rather, it constitutes an instance of that prototype, and thereby realigns the properties associated with the finite verb prototype in that language. Thus, in Classical Sanskrit it is still reasonable to consider the past participle a participle, since it can be used adjectivally and hence deviates from the employment of prototypical finite verbs; in many Modern Indo-Aryan languages, the historical reflex of the earlier construction is an exclusively finite verbal construction, meaning that, for instance, agreement in gender is now a prototypical feature of finite verbs in Modern Indo-Aryan, whereas it was a prototypically adjectival/nominal feature in Sanskrit.
Conversely, some languages attest verbal categories which cannot formally be distinguished from prototypical finite verbs, but which can be used as adverbs, adjectives, or nouns. Ebert (2003, 2008, pp. 76–79) discusses verb forms in Kiranti which are morphologically prototypical finite verbs, but which can be used converbally [i.e., as (clausal) adverbs, in chaining constructions]. In Mohawk, the class of “stative verbs” are used as adjectives, despite being morphologically indistinguishable from prototypically finite verb forms (Baker 2003, pp. 249–50, 257–63). Other Iroquoian languages, like Cayuga, attest forms which morphologically correspond to prototypical finite verbs but are used as nouns (Mithun 2000).⁹
Since at least Ross 1972, and especially Hopper & Thompson 1984, it has been recognized that the traditional categories of verb and noun are prototypical concepts, between which there exists a gradient; in any language, groups of words may cluster at different points on this gradient.³ Understanding categories in this way, then, we can analyze nonfiniteness as reflecting degrees of deviation away from the properties of a prototypically finite verb towards properties which are more prototypically nominal; this may involve properties which correlate with clusters of words on the cline approximating to the notions of adjective, adverb, and so forth.
For the purposes of this review, I focus on one or two relatively familiar and unproblematic instances of nonfinite verb forms which can be categorized, using standard terminology, as participles and gerunds. Clearly, these by no means exhaust the wealth of variation possible in terms of the syntax and semantics of nonfinite categories crosslinguistically, but nevertheless the principles of analysis applied to these relatively well-known nonfinite categories can be extended to any type of verb category which can be classified as nonfinite in the terms discussed above.¹⁰
A further preliminary consideration is delimiting the category of verb, including nonfinite verb, from other categories. As discussed above, the distinction between grammatical categories is to some extent a gradient one; indeed, nonfinite verbs are by definition closer to prototypical nouns (or adjectives, or adverbs) than prototypical verbs are. Nevertheless, from an analytical perspective it is important to draw a distinction between word forms or categories which are nonprototypical verb forms/categories and those which are nonverbal forms/categories. One way to understand this is in terms of the notion of inflectional paradigms, and the (gradient) distinction between inflection and derivation. Forms of the same lexeme are related to one another in terms of inflectional morphology, while different lexemes which share a root may be related in terms of derivational morphology. Treating nonfinite verb categories as inflectional categories of a verbal lexeme provides a clear analytical distinction between these and nonverbal categories; for example, the English gerund in -ing can be considered a part of the English verbal paradigm, and hence as a nonfinite verb category, while derived forms such as destruction or delivery are fundamentally nominal, despite standing in a derivational relation with a verbal lexeme.
Of course, the distinction between inflection and derivation is itself a gradient one, and nonfinite verbal categories may well be those which are most likely to be less inflectional than prototypically finite verbal categories. Nevertheless, for analytical purposes I assume that nonfinite verb forms are categorially verb forms, fundamentally at least, related inflectionally (understood in terms of the inflectional paradigm of a lexeme) to a verbal lexeme and hence to prototypically finite verb forms; in contrast, derived nouns (and derived adjectives and adverbs) are categorially nouns (or adjectives or adverbs). The categories I focus on in this review I take to be relatively uncontroversial instances of inflectional verbal categories, assuming that, in at least some instances, it is possible to make a clear distinction. This does not rule out the possibility that there may be more problematic cases, where it is not clear whether a particular form or category is nonfinite, in the sense defined here, or is derivationally related to the verbal system.¹¹