Conceptualizing and Formulating Finiteness - An Acquisition Perspective
It has been a robust, yet riddling, observation that some two-and three-year-olds, over the period of a few months, use infinitival utterances alongside finite ones. The discussion on this Root Infinitive puzzle, which applies in particular to the Germanic V2 languages, has tried to resolve the following issues in particular (see e.g. Wexler 1994, Wijnen 2000) :
1) Why does it take so long until these children use finite morphemes consistently as required by their respective target languages?
2) Why are there hardly any other consistent errors with respect to verb placement and finite verbal morphology?
My presentation examines this acquisition problem from a language production viewpoint. My conclusion will be that the overt marking (formulation, in Levelt's, 1989, terms) of finiteness involves selecting language-particular concepts, a process which is gradually learned from age of two onwards.
Kuzcaj und Boston (1982) have made the plausible claim that all children can acquire a certain set of temporal concepts (sequence, inclusion) on the basis of experience - despite the fact that temporal concepts differ across cultures. It is then reasonable to assume that children can use these universally available cognitive concepts as a semantic boot-strapping mechanism in order to establish first mappings between finite forms and finite contents. It is then expected that they will realize these mappings first in their linguistic utterances.
However, the world's languages differ in various respects with regard to the marking of finiteness. For one, some languages mark fewer finiteness features than others. This means that speakers must choose relevant conceptual categories, before they can formulate finiteness in their utterances (cf. Slobin 1987). A second typological difference concerns the processes used for marking finiteness. There is a trade-off between morpho-syntactic and lexical marking of finiteness, as I will illustrate. Thirdly, and in sharp contrast to what is generally assumed in the acquisition literature, the expression of finiteness is optional in many adult languages under certain discourse circumstances, though these circumstances differ across languages. Fourth, the same finite content can sometimes be expressed by more than one given morpheme or process. These points are especially relevant from the viewpoint of language production: In order to produce an adequate finite sentence it is necessary to choose from a set of finite concepts those that are relevant in the target language and for the current utterance. The notion of relevance which applies here cannot be motivated by "communicative value", rather it is subject to arbitrary language-specific variation.
If this is an adequate approximation of the acquisition scenario, Root
Infinitives are not eliminable by a one-time learning event, such as maturation
of a rule (Rizzi 1994, Hyams 1996). Rather, their proportions will diminish
and, in languages like German and Dutch, they will never fully disappear,
which is exactly what one finds.
Petra Gretsch (Max-Planck-Institut für Psycholinguistik Nijmegen)
Finiteness and world-referencing in early German child language
This paper compares the interpretation of finite with the interpretation of nonfinite utterances in early first language acquisition. I propose a universal bifurcation of world-referencing within the verbal (and nominal) domain which relates grammatical default settings as 'presence of finiteness' or 'citation case' with a world-referencing pointing to the actual world, whereas nondefault settings as 'absence of finiteness' or 'oblique case' are linked to a world-referencing involving distant worlds, i.e. wish worlds/future worlds. As soon as children mark these default/nondefault settings in their earliest utterances, their usage reflects these proposed differences in interpretation as indicated by their functional potential displayed. So, it has been shown that children make early use of the grammatical finite vs. nonfinite distinction to signal request interpretation and the interpretation of commenting on ongoing events whereas finite structures are prone to describe the actual world.
Within the 'optionality of tense' approach of Wexler (see Wexler 1994 and onwards) the existence of RIs as a phenomenon (primarily of) child grammar receives an explanation but not their bias in interpretation. Following the line of Jordens (1990), Behrens (1993) and Ingram&Thompson (1996) a.o. interpretative features as [+modal] or [-completed] have been assumed to be attached to the infinite morphology of RIs to allow for the encountered interpretational imbalance of finite vs. infinite utterances. But any approach which entertains a direct form-function link is bound to fail in accounting for the full range of empirically established interpretations across different studies, i.e. requests and comments in the case of RIs. Moreover, it is unclear how a child should eventually retract from such a non-target adequate feature association. Having sketched this problem set I will address the following questions:
(a) Which forms allow for which interpretations?
(b) Where does the finiteness element as a default/nondefault separator come into play?
I follow the framework of Klein (1998) where it is assumed that it is the finiteness element that links the (purely descriptive) sentence basis to a time span, about which an assertion is made, i.e. the TopicTime. Granting that the default setting of world reference is the actual world, I will show that in RI the nondefault infiniteness of the verb evades the setting of a TopicTime with corresponding effects on world reference and assertion potential.
As long as children lack other grammatical means to refer to worlds distant from the actual 'here&now', RIs are thus the suitable grammatical structure to convey interpretations where the setting of TopicTime is either undesired or unnecessary. The first case accounts for the modal interpretation by linking the nondefault setting [-finite] with a nondefault world reference, i.e. a wish world. The second case -- the unnecessary TopicTime fixation -- accounts for the commenting on ongoing events interpretation where the TopicTime of an utterance is contextually set via the simultaneously ongoing action.
Coming back to the questions above my results show that (a) the interpretational bias of RIs towards requests and commenting on ongoing events allows nevertheless for the existence of overlaps in form ([+/-finite]) and function (requesting, commenting, describing the actual world, narrating) in the three developmental paths analyzed (monolingual German children). Hence, finite and infinite forms do not constitute a mirror-image of e.o. which was the prediction of approaches involving a direct form-function link, but no substitute form either which was the prediction of the Wexler account. In contrast, the relation between grammatical (non-)default forms and (non-)default world referencing proposed here allows for a flexible modelling of the developmental facts within defined boundaries and it grasps biases of usage without assuming a direct link of interpretative features to specific forms. The first visible expression of this default/nondefault split in development gets manifest in the morphological finiteness distinction, which answers question (b).
The hypothesized default/nondefault bifurcation of world reference is buttressed by parallel findings in the nominal domain, as polish child data on the interpretation of different case marking show. Thus, the claim of the universality of grammatical default forms in relation to default world referencing gets further support with finiteness contituting the center of grammatical default.
Selected References:
Balthasar Bickel (University of California at Berkeley, University of Zürich)
Infinitives, grammatical functions and the Indo-European Integrativity Principle
In the better-known Indo-European languages it has often been observed that the distribution of nominative case correlates with the distribution of finite verb forms indeed this observation is hard-wired into virtually all current avatars of transformational (or 'movement-based') theories of grammar. The correlation typically manifests itself in the requirement that overt subjects of infinitival clauses be in a non-nominative case or marked by an adposition (e.g. Engl. for him to go). In some languages the requirement does not target case-marking but phrase-structural position, i.e. a typological alternative for distinguishing grammatical functions (GFs). This is so in many Romance languages, where subjects of infinitival clauses are typically relegated to postverbal position (Span. por affirmár=me=lo el experto 'since the expert confirms it' vs. *por el experto affirmár=me=lo) (unless the 'infinitival' clause is in fact finite by virtue of subject agreement, as in Portuguese). However, the correlation between specific GF-markers (case or position) and infinitives is by no means universal. There is no such correlation, for instance, in Sino-Tibetan languages; cf. e.g. Belhare han khatma 'you.NOM go.INF', i.e. 'for you to go'.
In this paper I argue that the GF-infinitive correlation is an epiphenomenon of a more fundamental principle that governs some languages but not others, in particular Indo-European but not Sino-Tibetan languages. This principle, called here the Integrativity Principle, consists of the identification of PSAs (privileged syntactic arguments, i.e. constructional pivots and controllers, as defined by Van Valin & LaPolla 1997) with GFs (i.e. elements listed in a predicate’s valence). The effect of this identification is that PSAs such as the controllers of verb agreement are sensitive to GF-markers such as case or position. As a result, the possibility of agreement and therefore of finite verb morphology, depends on the availability of a specific GF-marker. Other effects of the same principle include the following: (i) the absence of a certain case (e.g. nominative) or phrase-structural configuration freezes agreement morphology (cf. e.g. default agreement forms with sentential subjects in German, default agreement forms with non-nominative NPs in Hindi, limited agreement with postverbal nominative subjects in Icelandic); (ii) the thematically highest argument of a verb can serve as the PSA of a construction (such as control, raising, participial relativization, etc.) only if it is associated with a specific GF-marker (nominative and/or ergative, VP-external position). No such effects are generally found in Sino-Tibetan languages.
A survey of Indo-European languages suggests that the Integrativity Principle is by and large stable in this family, but that it is precisely the GF-infinitive correlation that is the weakest effect of the principle: it is this effect, and no other, that is lost in Nepali, a Indo-European language under extremely intense Sino-Tibetan substratum pressure. I propose that this is so because giving up the GF-infinitive correlation assimilates the surface clause constituency of the language directly to what is found in the substratum while all other effects of the Integrativity Principle involve relatively abstract regularities of the syntax/morphology and syntax/semantics interfaces and have thus less immediate impact on basic communicative structures.
Jaklin Kornfilt (Syracuse University)
Turkish at the Intersection of Nominalization and Finiteness Workshop on Finiteness
The notion of "finiteness" is a widely used and familiar one; traditional, descriptive and theoretical studies have relied on it heavily. Still, it is not completely clear what properties of a syntactic domain reflect the notion of finiteness. In most instances, "finiteness" is equated with "tensedness". In this paper, I will discuss "opacity"/"transparency" phenomena in Turkish (and, in passing, in some other Turkic languages) in nominalized subordinate clauses (which are often said to be non-finite), comparing them to their fully tensed, non-nominalized counterparts as well as to each other (given that there are different types of nominal subordinate clauses). I shall focus on the subjects of such domains and study their binding and Case properties. It will turn out that tense of the domain plays a subordinate role in determining the opacity of the clause (with respect to the ability of the subject's being bound and/or Case-marked from outside the domain). Overt morphological agreement between the (nominalized) predicate and the subject is the primary factor, as well as the entire domain's being thematically governed; yet another additional factor, where such thematic government is lacking, is whether the nominalized clause hosts a syntactic operator or not.
I show how all of these considerations make possible a systematic account of opacity facts which, at first glance, seem to imply a fluid, gradual hierarchy of nominalization, but which turn out to be based on discrete and differentiated factors like distinct agreement types and distinct moods. Furthermore, the notion of "parametric" will be shown to play a role in establishing the agreement type that determines syntactic opacity.
Tanmoy Bhattacharya (Universiät Leipzig)
Suspecting Mood
Since Pollock (1994) some studies in a south Asian language Bangla (Bengali) as in Roy (1995), Bhattacharya (1997), and Dasgupta (2000) have often claimed the existence of a Mood (M) head in the clause structure of this language. Furthermore, Amritavalli (2000) has advanced the claim for Kannada that finiteness can be encoded without a supporting tense head in the syntax in the form of such a Mood head. I argue against this rather plausible hypothesis based on evidence from an epistemic modal particle in Bangla (Bengali) which I believe has received no attention in the literature. The arguments presented in this paper are not against an M head in syntax of these languages but rather against the location of finiteness in such a head. Given that modals lack a thematic structure of their own and are possibly not specified for an event variable in their lexical structure (Higginbotham 1985), the tense-semantics in modals is not strong (e.g. the distinction between may and might is rather weak). From this perspective too, finiteness-at-Mood theories are suspect.
Bangla has a robust test in the negative paradigm for finitehood. Whereas
finites show the V+Neg order in the clause structure, non-finites (as in
the imperfective, gerundive, conjunctive and conditional in (2) below)
are always Neg+V:
(1) abdul asbe na FINITE
Abdul come.will not
'Abdul will not come'
(2) a. na khe-te IMPERFECTIVE
Neg eat-IMP/INF
'not to eat'
b. Abdul-er na aSa GERUND
Abdul-gen Neg come.GER
'Abdul's not coming'
c. na khey-e CONJUNCTIVE
Neg eat-PPL
'not (having) eaten'
d. na khe-le CONDITIONAL
Neg eat-COND
'(If s/he) doesn't come'It has been suggested that this distinction in the V-Neg order can be captured by positing an M head for the nonfinites which triggers Neg movement to the left of the verb. However, since there is no obvious modal licensing in these clauses, the implication is that the M head is null. I show that this picture is too simplistic and that the order of Neg and the V are dependent on the finiteness of the clause via tense, not mood. Thus it is shown that finiteness (showing V+Neg order) is independent of the mood of the clause.
If modality is a further qualification on the truth condition of the proposition expressed then naki, the particle that I investigate, has an illocutionary force that comes out as strong modality enforcement since it induces a sense of disbelief towards the content of the sentence from the speaker's point of view. It is a case of epistemic modality expressing the speaker's mildly negative judgement. Accordingly, I call it the suspective particle:
(3) Reboti naki Emerika jacche?
Reboti SUSP America going
'(I heard that) Reboti is going to America, is that true?'Interestingly, in spite of the strong modal characteristic, in the negative context a naki clause is V+Neg.
(4) a. Reboti naki Emerika jabe na?
Reboti SUSP America go.will Neg
'(I heard that) Reboti is not going to America, is that true?'
b. *Reboti naki Emerika na jabe?The Neg-V order can also obtain with a naki-clause but only in the subordinate (here, the conditional and therefore, adjunct) clause:
(5) a. Robin naki [ROnok na khele] ghumobe na
Robin SUSP Ronok Neg eat.COND sleep.will Neg
'Robin, I heard, will not sleep if Ronok does not eat'
b. *Robin naki [ROnok khele na] ghumobe naThe analysis offered here reverts to the traditional analysis of finiteness as located in the Tense head. It is shown that the V-Neg order is accounted for by the simple supposition that the V moves to T to check for finiteness. (I show that HMC is not an issue here.) The reference to verbal features moreover makes sense in terms of how the derivation proceeds by first merging the verb. Once finiteness is detected at this initial stage, this feature is matched with the first available head (here T) carrying a similar feature. Due to a lack of a finiteness feature on the nonfinite form of the verb, movement to T need not take place, no reference to an M head need be made here. The verb encoding modality is also seen in the following:
(6) a. apni naki aSben/*aSun
you(HON) SUS come.will.HON/*come.HON
'(heard) you are coming?'
b. tumi naki aSbe/*eSo
you SUS come.will/*comeThat is, a bare imperative form of the verb is not acceptable with naki. Similarity to the difference between standard Italian and Piedmontese can be noticed in this connection.
However, question arises about naki or other sentential mood denoting particles; how are they licensed? Although Bangla subjunctives and other mood distinctions exhibit tense curtailment, substantiating a weak T, I will argue against theories (such as Manzini 1994) which suggests that subjunctives are like polarity items involving a QR type movement to a sentential operator position. Instead, it is shown that mood licensing is a local phenomenon.
Frans Plank (Universität Konstanz)
What is being lost or acquired when verbs or verb phrases turn into adpositions or adpositional phrases, and especially the other way round?
Whatever verbs and adpositions, and verb phrases (or clauses) and adpositional phrases, may have in common, a finiteness opposition is not usually considered to be among the shared properties. In some sense this is actually surprising: affixal or clitic person-number marking as such IS something which adpositions or adpositional phrases may share with verbs or verb phrases -- and person-number marking has been considered a crucial ingredient of finiteness. What I'd like to look at, therefore, is whether person-number marking on adpositions, in languages which show it, points to some more general structural similarity with verbs that could meaningfully be captured in terms of finiteness. I don't think it does.
Second, I'd like to look at developments of verbs into adpositions and at the attendant losses. These are usually expressed in terms of finiteness: it's always non-finite verbal forms, in non-finite constructions, which turn into adpositions; 'go' and 'concern' in English, for example, cease to be finite when turning into 'ago' and 'concerning'. The losses involved are indeed heavy: it's verbal inflection in its entirety that is gotten rid of, and much syntactic variability too. If finiteness as such is implicated, it's hard to pinpoint what is supposed to be covered by it -- only person-number inflection or inflection for all other verbal categories and much verbal syntax too?
Third, there is a less common kind of development which is probably more telling: the reverse one of adpositions turning into verbs. There are various scenarios for such de-grammaticalisations: lexicalisation of products of conversion (such as German 'ahnen' "to have a vague feeling of something happening or having happened" from 'an' "at"), or reanalyses of local or comitative prepositions as transitive verbs of possession (as best documented in, though not confined to, Afroasiatic languages, especially Maltese and Coptic). Person-number inflection seems to be the easiest acquisition in all such cases; or it does not even need to be acquired, with the former prepositions themselves simply retaining their original person-number markers (as in Maltese), even at the expense of complicating the syntax of agreement considerably. On the other hand, acquiring tense/aspect and mood may pose serious problems, calling for extraordinary solutions, such as the creation of suppletive paradigms. So, if anything profound changes in such developments from adpositions and their containing constructions towards verb(-phrase-)hood, then illocutionary-force categories such as tense and mood would seem to be more important ingredients than person-number of whatever it is (call it 'finiteness') that fundamentally distinguishes such forms-in-constructions from one another.
last updated: 3.5.2001
Thomas Schöneborn