Assume a language whose verb conjugates for a set of morphological categories, such as voice, tense, aspect, mood, person and number. Then a fully finite verb form is a verb form inflected for all of these conjugation categories. A non-finite verb form is one that is not inflected for all conjugation categories and which, instead, may show declension categories.

you may be praised
(one) to be praised

Taking the verb forms of as introductory examples, the form of #a is conjugated for all of the conjugation categories of the language: tense, mood, voice, person and (verbal) number. The form in #b, instead displays none of these categories and instead declines for gender, (nominal) number and case. The form in #a is finite, the gerundive of #b is non-finite. From the point of view of the language system, finiteness is primarily a property of a morphological construction – like the second person singular present subjunctive passive of any verb – and only derivatively a property of a morphological variant of a particular verb like lauderis.

The category of finiteness exists only in languages whose verb conjugates. In isolating languages like Thai and Vietnamese, this condition is not fulfilled: verbs show no syntactically relevant morphology. Consequently, there is no finiteness contrast, either.

Syntax of non-finiteness

Finiteness is a property of a morphological form, in the first place. A syntactic construction headed by a finite verb is a finite verbal construction; and one headed by a non-finite verb is a non-finite verbal construction. Nevertheless, since non-finite verbal constructions are dependent constructions, they are downgraded in comparison with an independent verbal clause, i.e. one which may make a sentence. Consequently, they have a certain syntactic function in their context. The syntactic context is compatible with a non-finite dependent construction and may even condition it.

The dependent character of a non-finite construction involves two grammatical properties:

  1. In general, the construction is desententialized to a certain extent.
  2. Specifically, its subject – or, in an ergative system, its absolutive function, but henceforth simply “subject” – is affected by the downgrading.

As for #1, the desententialization involves the non-specification of mood, tense, aspect – information that is not coded because it does not matter or because it is carried over from the matrix clause. These are just those verbal categories typically missing from non-finite verb forms.

As for #2, an important subcategory of desententialized constructions has the subject position of its verb reduced. The subject position does not get blocked, as in operations of voice and valency; non-finiteness is not a kind of valency reduction. Instead, the subject position is transformed by empty-place formation. For instance, the subject position of an infinitive may be phorically controlled by an argument of the superordinate verb. A participle has the subject position of the finite verb transformed into a modifying slot.

It is worth recalling that the perspective of desententialization, including the reduction of a finite to a non-finite verb form, has its converse in the expansion perspective. A non-finite verbal construction may be refinitized, as it were, by being made dependent on an auxiliary or even by simply being used as the predicate of an independent clause.

Morphology of non-finiteness

Verbal categories in finiteness

The contrast between finite and non-finite verb forms has been introduced with reference to a set of conjugation categories which was identified by enumeration. The theoretical basis of this distinction is the difference between an independent verbal clause and a dependent verbal clause. The verb form heading an independent clause is by definition finite, and the conjugation categories that it features make a finite verb form in this language by definition. A verb form heading a dependent clause may or may not be finite. It is less than fully finite if it shows less than the verbal categories that are specified in an independent clause.

Note that a non-finite verb form is not defined as one that is not conjugated. That would be erroneous for two reasons:

to praise
to have praised
to be praised

The expression “fully finite” already indicates that finiteness is – at least at the typological level – not a binary category. It may be so in a given language. For instance, in Latin and several other Indo-European languages, it is fruitful to classify verb forms inflected for person as finite and all other verb forms as non-finite. However, at the interlingual level, there are two kinds of complications here:

  1. There are languages which do have conjugation, but person/number is not a conjugation category.
  2. There are languages which not only possess fully finite verb forms, but also forms which feature different subsets of the set of all of the conjugation categories.

The two kinds of phenomena will be illustrated in turn:

The Cabecar finite verb inflects for mood, aspect and voice. For instance, the transitive root butsa- ‘break off’ forms, inter alia, the active perfective butsá ‘broke off (tr.)’ and the middle perfective butsaná̱ ‘broke off (intr.)’. There is no person or number on verb forms. There is, however, an infinitive, viz. butsä ‘break off (tr.)’; and this goes into middle voice, too: butsana̱ ‘break off (intr.)’. The infinitive does not, however, inflect for mood and aspect. It is therefore less finite than the finite forms quoted before. The distinction may be framed as a binary distinction between finite and non-finite forms even though there is no verbal person and number.

The Latin infinitive illustrated by does conjugate for two of the categories of the finite verb, viz. tense and voice. There are other non-finite verb forms which are even less finite in this respect. For instance, the supine lauda-tum (praise-SUP) ‘in order to praise’ shows none of the conjugation categories and instead an ending which, by its morphological shape and by its function, is nominal in nature. Ironically, in this particular language, the infinitive – the would-be epitome of a non-finite form – is more finite than other non-finite forms. More in general, there are gradations of finiteness.

Formally, the transition between fully finite and non-finite forms follows the desententialization continuum operative, in general, in the nominalization, adjectivization and adverbialization of verbal constructions. For instance, the Latin supine is a fully adverbialized verb form of the language.

The privileged position that the category of person (and verbal number) has played in traditional definitions of finiteness is bound up with the form and function of the infinitive. In languages with accusative alignment of the fundamental grammatical relations, person and number of the finite verb bear a phoric relation to its subject. Now as observed above, an infinitival construction is a construction of a dependent clause whose subject has been reduced. Depending on the particular language, the subject position is either left open or the dependent corresponding to the subject of the finite form appears in a different dependency relation, like a direct object or a nominal attribute. Since the verb lacks a subject, it shows no agreement with the subject, either, thus lacking person and number inflection. This, however, is a feature of certain languages which does not generalize over all languages or shape a general concept of finiteness.

Non-finiteness between inflection and derivation

Finally, there is the problem of the status of non-finite verb forms between inflection and derivation. The set of criteria under #1 argues that they are inflected forms, while the set under #2 argues that they are derived forms.

    • Inflection is syntactically relevant morphology. Now there is a set of transformational relationships between finite and non-finite constructions. Since these are paradigmatic relationships which are syntactic in nature, the morphological features conditioned by them are syntactically relevant and, consequently, features of inflection.
    • In many languages and for many non-finite verbal categories, their formation is fully regular and productive. This, again, is a property of inflectional categories not commonly shared by derivational categories.
    • Non-finite forms belong to different syntactic classes. For instance, the infinitive may be a verbal noun, the participle an adjective and the gerund an adverb. The infinitive and the participle may be declined. The notion of a conjugated form which may be declined seems inconsistent.
    • Non-finite forms may be the basis of derivational operations. For instance, the German present participle of eilen ‘hasten’ is eilend ‘hastening’. From it, the adverb eilends ‘in a hurry’ may be derived. Now the basis of a derivational process is a stem, not an inflected form. Consequently, formation of the participle must be a derivational, not an inflectional operation.

The methodological conclusion from these irreconcilable kinds of evidence is the following: Either the distinction between inflection and derivation is not a binary categorial distinction and, instead, a gradual and prototypical distinction. Or else we must, in view of less than full correlation among the various features constituting the distinction between inflection and derivation, select one binary feature to the detriment of all the others in order to found a binary distinction.

The second alternative would run counter to traditional linguistic conceptions of inflection and derivation. Moreover, the distinction between inflection and derivation is just one instantiation of the more general distinction between grammar and lexicon. The latter is fluid, anyway. It suffices to mention grammaticalization and lexicalization, two basic gradual processes of language activity that cross the border between grammar and lexicon. Pending superior insight, it would therefore seem prudent to abide by the ambivalent position of non-finiteness between inflection and derivation.