Finiteness a gradient concept or a binary dichotomy? What do subject clitics tell us.
According to the traditional definition of finiteness, non-finite verbs contrast with finite verbs by marking no tense, aspect and mood, by showing no subject-verb agreement, by lacking Nominative Case licensing properties so that they are incompatible with lexical subjects, and finally by the fact that they appear only as dependent verbs. However, the standard view has been recently challenged by the idea that there is no binary distinction between finite and non-finite forms, but rather different degrees of finiteness based on variable features, the values of which give rise to morphologically as well as syntactically different types of non-finite constructions. Evidence in support of this idea comes from infinitival construction that are obvious counter-examples to the claim that infinitives do not have Case properties and do not carry agreement features. These are infinitival structures in Romance in which the verb may occur with a full DP subject (personal infinitives) or show up with inflectional affixes marking subject-verb agreement (inflected infinitives).
Nonetheless, in contrast with this novel interpretation of Finiteness, one finds that Subject Clitics in the Northern Italian Dialects are sensitive to the traditional dichotomy. All examinations of SCLs so far have been based on the undisputed property of these elements to occur with finite verbs. However, shifting the starting point of the inquiry, one observes that SCLs in the NIDs are omitted in all non-finite contexts, as traditionally defined. In addition, SCLs are obligatorily excluded with true imperative forms, complicating in this way an analysis of the distribution of SCLs due to the ambiguous status of true imperatives with respect to Finiteness, and increasing the difficulty involved in characterising Finiteness.
In this paper, I exploit the obligatory omission of SCLs with non-finites and imperatives seeking to establish whether these form share a common property which affects the use of SCLs. The ultimate goal, however, is to explore to what extent a better understanding of the use of SCLs may help for a better characterisation of Finiteness.
The immediate result of my analysis is that if, as suggested, Finiteness corresponds to features [+Agr] [+Tense], SCLs are sensitive to Finiteness in the sense that any alteration of such a feature combination results in obligatory omission of SCLs. At first my analysis reveals that SCLs are sensitive to feature [±Agr]. However, looking at different infinitival constructions and syntactic imperatives also shows that these forms are intrinsically [-Tense]. In this paper, I tentatively suggest that [-Tense] could be made responsible for the obligatory omission of SCLs. In this sense, therefore, SCLs would pinpoint a specific position on the spectrum of Finiteness separating verbal forms specified as [+Tense] from those specified as [-Tense], regardless of the value of Agr.
I interpret this as an alternative way to look at Finiteness. More precisely, if Finiteness is defined in term [±Agr ±Tense], the use of SCLs does mark a binary distinction between two groups of verb forms. Obligatory omission of SCLs may therefore be taken as diagnostic of some property of (non-)Finiteness. The value of Agr might not be the relevant property, but the value of Tense might. However, it is plausible that the notion of Finiteness ought to be characterised in terms other than [±Agr ±Tense], in which case the use of SCLs might turn out to be irrelevant and a gradient concept of Finiteness might be more adequate.
Finiteness and independent infinitives
The paper analyzes constructions with independent infinitives in some
European languages.
These constructions fall into two major classes depending on whether
or not they convey assertion. The assertive constructions have the following
properties: (i) they allow a clause-internal subject; (ii) they have a
time reference; and (iii) they are informationally structured, that is,
have topic and focus. The non-assertive constructions lack all these properties.
Viewed from this perspective, the assertive infinitival constructions
do not differ from regular finite clauses, except that they are based on
infinitives which are not overtly marked for tense and agreement. However,
a number of recent studies have argued that cross-linguistically finiteness
may be independent of categorization by tense and agreement (Vincent 1998
and others). Given that the lack of tense/agreement morphology is not an
obvious diagnostic for finiteness and that the assertive infinitival constructions
occur in non-embedded contexts, they can be considered intrinsically finite.
In line with the proposal in Klein (1998) it is further suggested
that finiteness as a clausal property is a linguistic correlate of assertion.
It is a primitive notion and the finite/non-finite distinction is basically
binary. Assertive utterances, for example statements and questions, are
finite, while non-assertive utterances, for example imperatives and some
exclamatives, are not. The status of a complement clause under this approach
has to do with the matrix predicate. Speech act, propositional attitude
and perception verbs introduce finite clauses, which express independent
assertions and can be paraphrased with it is true that. Other complement-taking
predicates typically introduce non-finite complement clauses, which do
not express independent assertions.
Although the finite/non-finite distinction as outlined above
is assumed to be universal, its morphological realization is language-specific.
There is a strong cross-linguistic tendency to express finite clauses with
verbal forms that take a subject, carry tense and agreement features and
occur in independent clauses. These properties usually come together because
of the nature of assertions. Assertions are typically made about a topic
referent, which tends to be linguistically expressed as a subject; their
truth is evaluated with respect to a certain time, and they can be conveyed
as independent utterances. In contrast, non-finite clauses are normally
realized with verbal forms that do not have subject or tense and mostly
occur in embedded contexts. Yet, this tendency is far from being absolute.
Agreeing forms and forms marked for tense may be used in non-finite clauses
(for example, inflected and tensed imperatives), dependent verbal forms
may be used in finite clauses (assertive independent infinitives), and
for many languages morphological corollaries of finiteness are altogether
non-existent.
Given this, a distinction should be made between "semantic" finiteness
related to assertion, and "morphological" finiteness, which has to do with
the properties of the verbal form. If we assume after Rizzi (1997) that
the (semantic) finiteness is located in the CP component of the clause,
the relationship between these two parameters can be viewed as some kind
of selectional restriction. The complementizer selects the type of clause
(and therefore the verbal form) which it introduces. This largely happens
on a language-specific basis, although we may be able to make some cross-linguistic
generalizations.
David M. Perlmutter, John Moore (University of California, San Diego)
In What Ways Can Finite and Nonfinite Clauses Differ? Evidence from Russian
Russian presents striking contrasts between finite and nonfinite clauses:
certain clause types that occur as finite clauses cannot be nonfinite:
1) unaccusative, passive, and dative experiencer clauses with
the genitive of negation
2) accusative human experiencer clauses
3) dative experiencer clauses with accusative objects
4) clauses with weather predicates
Why these gaps (illustrated here for infinitival clauses)?
These clause types are all impersonal. Thus the generalization
appears to be:
(1) Impersonal clauses cannot be infinitival.
Should (1) be included among the properties of nonfinite clauses in
Russian? And why does (1) hold in Russian (but not in other languages)?
Is there a deeper explanation?
Further research reveals an important exception to (1).
While (1) holds in Russian infinitival root clauses, direct and indirect
questions, purpose clauses, and temporal clauses, it does not hold in infinitival
complements of raising predicates. We need to explain not only why
(1) holds in the cases it does, but also why it fails to hold in infinitival
complements of raising predicates.
Independent evidence from overt NPs and from the "second dative"
phenomenon reveals the key difference between those infinitivals that disallow
impersonals and those that allow them: subjects of the former are
in the dative case, while subjects of the latter are nominative.
The true generalization is thus not (1) but (2):
(2) Impersonal clauses are impossible in environments where the
subject must be dative.
It might appear that the true constraint is not (2) but (3):
(3) Nonfinite clauses (other than raising complements) must have
an overt subject.
If impersonal clauses have no subject, (3) will correctly rule out
nonfinite impersonals. However, (3) cannot be maintained; controlled
purpose and temporal clauses and non-controlled unspecified human subject
clauses are clear counterexamples to it.
An explanation can be found if Russian impersonal clauses, like those
in English, have a dummy or expletive subject. In Russian, as in
Italian and other languages, this expletive subject is silent. Then
(2) can be shown to follow from an independent generalization about Russian:
(4) Silent subject pronouns can only be nominative.
The independent evidence for (4) comes from Subject Pro Drop, which
is possible for nominative but not dative subjects. If impersonals
have a silent dummy subject, (4) explains why they are disallowed in environments
where the subject is dative, but are allowed where the subject is nominative.
We have now explained the distribution of nonfinite impersonals
in Russian without any reference to impersonals in the typology of properties
of finite and nonfinite clauses. Their distribution is explained
through an independent case requirement on nonfinite clauses - a phenomenon
well attested in other languages which must be included among the cross-linguistically
possible properties of nonfinite clauses.
Eric Mathieu (London), Ioanna Sitaridou (Manchester)
‘Against the finite/non-finite distinction’
The present paper challenges the assumption made in transformational
generative grammar that ‘finiteness’ is an elementary notion. The distinction
between finiteness versus non-finiteness is a distinction that the transformational
generative framework has inherited from traditional grammar. Although transformational
generative grammar was keen to dispense with traditional concepts such
as subjects and objects, other normative tenets, such as finite/non-finite,
functional/lexical category, and rich/poor morphology have survived in
current theories. The latter distinction has been tackled by Alexiadou,
Haegeman and Stavrou (2000). We focus herein on the finite/non-finite distinction.
The standard account (e.g. Rizzi 1997) of finite versus non-finite assumes
that: (a) finite forms show mood, tense and agreement distinctions and
allow for a NOM subject; (b) non-finite forms show no mood, tense or agreement
distinctions and do not allow for a NOM subject. However, from a typological
perspective, the facts are much more complicated. Like Portuguese, Galician
and Sardinian, Old Neapolitan (1) shows agreement morphology on a non-finite
verb while a separate NOM subject distinct
from that of the governing verb surfaces:
(1) A che desideramo de volerelo torbare e de voleremone riducere da richeze? (Inflected infinitive)
to what wish-pres1pl C to want-inf-it upset-inf and C to want-inf1pl reduce-inf from riches
‘Why do we wish to upset it and to reduce ourselves to poverty?’
In Spanish, infinitives can also admit NOM subjects, on the condition that the category in which one finds the infinitive is a non L-related position. In this case, however, the infinitive bears no morphological marking for subject-verb AGR:
(2) Después de actuar Caballé, el público estaba expectante. (Personal infinitive)
‘After Caballé performed, the public was expectant.’
(3) Yianis arxise na grafi. (Na-clauses)
(1), (2) and (3) show that a purely morphological definition of finiteness
will not do (cf. Vincent 1998). Our definition of finiteness is multi-dimensional
and based on syntactic rather than morphological criteria. We decompose
‘finite’ and ‘non-finite’ into a +/- binary syntactic feature, which is
a correlate of tense/mood/AGR. The observed patterns are derived from whether
the embedded clause is dependent on the tense and/or mood
and/or AGR of the matrix predicate; in other words on whether or
not the embedded clause introduces a function |(x).
Our research is in line with other attempts which have proposed a feature-based
account of finiteness. However, the advantage of the +/- dependent distinction
we put forward is that it is independently motivated. Anaphors are dependent
on their antecedent whereas R-expressions are not, negative polarity items
are dependent on a Neg operator, whereas Neg quantifiers are not, definite
expressions are dependent on a familiar referent, whereas indefinite expressions
are not, etc. (Koster 1987, Neeleman & de Koot 1999).
| Tense | Mood | Agreement | |
| ‘Inflected infinitives’ clauses | + dependent | - dependent | - dependent |
| - dependent | + dependent | - dependent | |
| ‘Personal infinitives’ clauses | - dependent | + dependent | - dependent |
| Na-clauses | - dependent | - dependent | - dependent |
| + dependent | - dependent | + dependent | |
| Subjunctive clauses | + dependent | - dependent | - dependent |
| English control infinitives | + dependent | + dependent | + dependent |
| - dependent | + dependent | + dependent |
Note: like Horstein (1990), we argue that tense and mood
in control structures can be dependent on their matrix predicate.
References
Denis Creissels (University of Lyon)
Morphological characteristics and syntactic properties of the Setswana gerund
Setswana (a Bantu language spoken in Botswana and South Africa) has
verb
forms (called 'participial' in Cole's reference grammar of Setswana)
whose
syntactic properties entirely coincide with those of prototypical
non-finite forms, but whose morphological characteristics do not show
any
feature of non-finiteness:
- syntactically, these forms are quite similar to the forms typically
recognized as gerunds in traditional European grammar, in the sense
that
they are exclusively used in dependent predications, and typically
express
a temporal relation with the main predication, as in ke-ba-bonye ba-bina
'I
saw them dancing' lit. 'I-them-saw they-dancing';
- morphologically, they include obligatory subject markers (with the
same
distinctions of person, number and gender as in the indicative), and
they
show TAM distinctions exactly identical to those expressed in the forms
of
the indicative.
Finiteness in Nivkh: a Three-way Syntactic Distinction
Basing on syntactic interpretation of finiteness, verb forms in Nivkh (Paleosiberian, isolate) can be divided into three classes: (1) finite forms, (2) nonfinite forms, (3) in-between forms. Finite forms can be used as the only or the matrix predicate, whereas nonfinite forms cannot appear in these syntactic functions. In-between forms share the features of both finites and nonfinites. The syntactic distinctions between given forms, however, do not strictly correlate with their morphological characteristics, i.e. with their ability to agree with the subject in person and number and to be marked for TAM categories.
(1) "Pure" finite verb forms occur as predicates in single-predicate sentences or in matrix clauses of multi-predicate sentences as follows:
a.. Negative indicative forms demonstrate an unusual limited
differentiation of person and number of the subject (2sg and 3sg vs. 1sg,
1pl, 2pl, and 3pl) and are inflected for tense and modality.
b.. Second person preventive forms are specified by number of
the subject but are not marked for TAM
c.. Interrogative and modal indicative (both realis and irrealis)
forms do not take any subject agreement. Interrogatives attach all of the
TAM markers, while modal indicatives are not inflected for tense and aspect.
(2) "Pure" nonfinite verb forms occur as predicates in nonmatrix clauses
of multi-predicate sentences or as nonpredicates in all types of sentences
as follows:
a.. Coordinated forms appear as predicates of coordinated clauses.
Each coordinated form takes the same set of suffixes and agrees with its
own subject in person and number, as negatives. These forms are characterized
by tense, aspect, modality, and mood.
b.. Participles, that function as attributes and are used for
relative-clause formation, display the verbal morphology, having overt
expressions of TAM categories and number of the head nominal.
c.. Supines, that are used for denoting purpose, are not inflected
for any categories.
(3) In-between verb forms embrace the following two types:
(a) In-between forms that are prototypically used as finites and nonprototypically
as nonfinites:
a.. As finites, imperative forms demonstrate a full differentiation
of person and number of the subject and are marked for aspect and modality.
As nonfinites, they indicate concessive meaning.
b.. As finites, indicative forms are optionally inflected for
number of the subject, but do not agree with it in person, and are marked
for all TAM categories. As nonfinites, the indicative forms function as
verbal complements (and can be nominalized).
(b) In-between forms that are prototypically used as nonfinites and
nonprototypically as finites:
a.. As nonfinites, converbs may have (i) the same subject as the finite verb form, (ii) the subject different from that of the finite verb form, or (iii) admit both possibilities. Basically, converbs of types (i- iii) do not take any subject agreement and are characterized by aspect and modality. However, converbs of type (iii) include several converbs that agree with the subject of the finite verb form (but not with its own subject) in person and number according to the agreement pattern typical of negatives and coordinated forms (that agree with their own subjects). In case the subjects of the converb and the finite verb form are different, the converb takes the causative suffix. Moreover, given converbs "respond" to tense / mood of the finite verb forms. As finites, conditional converbs mark the optative meaning.
On defining non-finite verb forms by their word-classes and syntactic functions
Taking most of my examples from the morphologically rich Finno-Permic
(Uralic) languages of northeastern Europe, I intend to discuss some
problems on defining infinitives, participles, converbs, and action
nominalizations (henceforth ANs) on the basis of 1) the "new"
word-classes they can be said to represent and 2) the syntactic
functions in which they prototypically are used.
Cross-linguistically, the infinitive, the participle, the
converb, and the AN seem to be more or less valid subcategories of
non-finite verb forms. Sometimes the categories in question have been
distinguished on the basis of their relation to other (i.e., non-verb)
parts of speech: ANs or verbal nouns are, by definition, nouns as
regards their external syntax (Haspelmath 1996). Correspondingly,
participles are defined as verbal adjectives, and converbs as verbal
adverbs. Another possibility is to define non-finite forms by their
prototypical functions in simple sentences: infinitives occur as
complements and thus in subject and object positions; participles have
attributive functions, and converbs function as free adverbials,
adjuncts.
It is easy to see the proportionality between participles,
(other) adjectives, and the attributive function on the one hand, and
between converbs, (other) adverbs, and the adverbial function on the
other. However, when defining infinitives and ANs, the above
perspectives to non-finite verb forms seem to lead to asymmetry.
Although infinitives are often characterized as nominal, it is strange
to call them verbal nouns if the "verbal noun slot" is, in a sense,
filled by the ANs. Furthermore, it is difficult to assign a new non-verb
word-class to verb forms that function as obligatory arguments only,
without such external syntactic properties as accusative markers when
in
object position. Correspondingly, ANs seem not to fit into the set
of
categories (infinitives, participles, and converbs) that may be defined
functionally: an AN of the Uralic (or Altaic, for that matter) type
functions just like any noun; thus, in a way, its usage covers the
functional area of infinitives and converbs.
In my paper I will discuss the problematic relationships between
ANs, infinitives, and converbs in light of the diachronic development
of
the Uralic AN in *-mV. It has retained its AN use in most Finno-Permic
languages; in some languages, certain case forms of *-mV have fossilized
to variable extents so that there have arisen new infinitive and converb
markers. In addition, grammaticalization processes have also lead to
some typologically less common non-finite forms which might be
characterized as verbal obliques in opposition to converbs (adjuncts)
and infinitives (objects, subjects). When weighing different approaches
to subcategorize non-finite verb forms, I will try to find possible
common denominators for all of them, and in that way I am aiming at
a
better understanding of (non-)finiteness in general.
REFERENCE
Haspelmath, Martin 1996: Word-class-changing inflection and morphological theory. - Geert Booij & Jaap van Marle (ed.), Yearbook of Morphology 1995: 43-66.
The Expression of Finiteness in Swedish: Grammatical and Morphological Features
I will argue in this paper that there is no one-one correlation in Swedish between grammatical TENSE in a clause and the morphological forms which express that TENSE information, and conclude that this supports a constructional or expressive view of morphology, over a morphemic view. The data mostly involve the supine form of the verb, which is used with the auxiliary verb "ha" (`to have') to form the perfective. Of interest is the possibility for `deletion' of the `have' auxiliary when it is in construction with the supine: "ha" is optional before the supine in certain contexts, characterized in the generative literature by a rule of `ha-Deletion'; but it cannot be `deleted' when it is second position in a main clause (see Den Besten 1983, Platzack 1986 and Holmberg 1986):
(1) Han maaste (ha) varit sjuk.
[aa = a with a circle over it]
he must.Pres (have) be.Sup sick
(2) Han *(har) varit sjuk.
he (have.Pres) be.Sup sick
Finite modal auxiliaries in Swedish are clearly recognizable as such, so omission of "ha" in (1) still leaves a finite verb in second position in the clause, while omission of "ha" in (2) leaves the clause without a morphologically-marked finite verb. Semantically, the ("ha")+supine construction carries either a present perfective or a simple past interpretation; without "ha", context biases the perfective or past interpretation. To contrast interpretation and form, I will use TENSE to label to the grammatical information that a clause is PRESENT, PAST, etc., and Finite to label the surface morphological forms (Present, Past, etc.).
The finite form is optional in embedded clauses, which are not subject to the V2 requirement. In the version of (3) without "har", there is no formally finite verb in the embedded clause.
(3) Jag tror att han (har) varit
sjuk.
I think that he (have.Pres) be.Sup sick
Moreover, there are certain main clauses in Swedish where a non-verb marks the clause type, and yet "ha" is optional. The grammaticalized form "kanske" (derived historically from `can happen') can fulfil the V2 requirement by appearing in second position. When such a non-verb is in the crucial position, "har" is optional, as shown in (4):
(4) Sara kanske (har) varit
har.
Sara maybe (have.Pres) be.Sup here
What can we make of this? "ha" must be present if it is in second position to mark a V2 clause, but if it is not in second position, it can be absent, even if it is the only truly Finite verb in the clause. So we can see that second position does not require a Finite verb, and certainly not one expressing TENSE, but rather, it requires any form with a certain morphological feature, shared by all Finite verbs and by "kanske". Let me call this "F2" (feature in second position). Hence, Swedish allows clauses that have grammatical TENSE which lack any morphologically Finite forms, but in a V2 clause, something must be present to express F2. We now see that in many of the examples above, the reason for "har"'s presence may not be to express TENSE at all, but rather to express F2. On a morphemic view, these data would be very odd: TENSE is expressed without any Finite forms to express it, and Finite verbs sometimes appear without expressing TENSE, anyway.
Consequently, I will turn things around, to think of an expressive or constructional model, and propose that Swedish clauses have grammatical information about TENSE and V2, which is expressed by surface Finite and F2 forms, but not in a simple one-one fashion, and I will present a more formal account of how this can work using the framework of Lexical-Functional Grammar (Bresnan 1982, 2001); for related observations about verbal morphology, see Sadler and Spencer (2000).
This view of Swedish makes sense of an example like (5), which has two Finite verbs in it!
(5) Skriver goer han sellan.
[oe = umlauted o]
write.Pres do.Pres he seldom
This is a V2 clause, and on a morphemic view, it is strange that there would be two Finite verbs. On the view advocated here, (5) is not strange at all: the first verb "skriver" expresses TENSE PRES (in virtue of being a Present form), and the second verb, the dummy verb "goer", expresses V2 by being an F2 form.
last updated: 3.5.2001
Thomas Schöneborn