Printer Friendly

What does finiteness mean to children? A cross-linguistic perspective on root infinitives *.


The discussion on root infinitives has mainly centered around their supposed modal usage. This article aims at modelling the form-function relation of the root infinitive phenomenon by taking into account the full range of interpretational facets encountered cross-linguistically and interindividually. Following the idea of a subsequent "cell partitioning" in the emergence of form-function correlations, I claim that it is the major fission between [+-finite] which is central to express temporal reference different from the default here&now in tense-oriented languages. In aspectual-oriented languages, a similar opposition is mastered with the marking of early aspectual forms.

It is observed that in tense-oriented languages like Dutch and German, the progression of functions associated with the infinitival form proceeds from nonmodal to modal, whereas the reverse progression holds for the Russian infinitive. Based on this crucial observation, a model of acquisition is proposed which allows for a flexible and systematic relationship between morphological forms and their respective interpretational biases dependent on their developmental context. As for early child language, I argue that children entertain only two temporal parameters: one parameter is fixed to the here&now point in time, and a second parameter relates to the time talked about, the topic time; this latter time overlaps the situation time as long as no empirical evidence exists to support the emergence of a proper distinction between tense and aspect.

I. Introduction

The phenomenon of "root infinitives" or "optional infinitives" has received a great deal of attention in the generative literature over the last ten years (see, e.g., the articles in Linguistics 40-4 [2002]). The term "root infinitive" relates to the fact that early child utterances show no finiteness marking within their structure although they are used as if they were root sentences, that is, as nonembedded structures. This is in contrast to the regularities of the respective target languages where it is assumed that root sentences are required to be marked for finiteness (but see Lasser 1997 for a critical review of German adult language and finiteness marking). The term "optional infinitive" focuses on the alleged optionality of infinitival or tense-marked verbal structures which is supposed to be the main characteristic of an early stage of acquisition, the optional infinitive stage (Wexler 1994). The following examples from Dutch, German, Greek, Swedish, Icelandic, Russian, and English have been discussed under one of the two headers "root infinitive/optional infinitive" (henceforth shortened to RI) in the literature cited below:
(1) Maarten 1;11 (Gillis corpus cited from Kramer 1993)
 (Context: Maarten's mother tells him he has to eat a slice of
 Steven also sits down at the table to eat. Maarten remarks:)
 Steven ook boke eten
 Steven also slice-of-bread eat-INF
 'Steven should eat a slice of bread, too.'

(2) Max 2;9 (Tubingen corpus)
 (Context: Max turns to the tape recorder.)
 M: des n n man n sanne anhorn, den anmachen
 that (n once n?) (su)sanne listen-INF, that onmake-INF
 den anmachen.
 that onmake-INF
 'I want to listen to that, switch it on, switch it on.'

(3) Janna 1;11 (from Stephany's data, CHILDES database; cited
 from Varlokosta et al. 1996: 822)
 Nitsi tola! [= (na) aniks-is tola]
 open-PERF-SUBJ-3SG now
 'You shall open (it) now!' (lit. 'S/he shall-open now.')

(4) Sara 2;0 (Josefsson 1999: 106)
 S: Tova har
 sleep-INF here
 A: vem ska sova dar uppe? 'Who will sleep up there?'

(5) Birna 2;6 (Sigurjonsdottir 1999: 632)
 Eg fara til Dodiar
 I go-INF to Dodi-GEN
 'I am going to Dodi.'

(6) Zhenya G. 1;7 (Brun et al. 1999: 123)
 Mama maslo kupit'
 Mummy butter buy-PERF-INF
 'Mommy has bought butter.'
 This sentence was uttered in reference to his mother's putting
 groceries (including butter) away.

(7) Nina 2;3 (Schutze and Wexler 1996: 670)
 Him fall down

These varieties of cross-linguistic data are very heterogeneous in form and concomitantly in interpretation. Thus one might wonder what justifies a common label like "root infinitive" for the child structures above. It is the aim of this article to scrutinize the formal and functional, that is, interpretational properties of root infinitives from an integrative perspective, which takes cross-linguistic and interindividual variation into account. So, the term "functional" is understood here as the pragmatic category like request or comment in which a specific utterance structure is put to use. A dynamic developmental model is proposed which interrelates formal and these functional properties of early infinitival and finiteness-marked structures in child language. The focus of this enterprise is given by the question as to how children encode temporal information by morphological means, and temporality, in turn, is crucially linked to finiteness. The data compiled in this article demonstrate in this respect that early form-function relations in language development are too heterogeneous to follow a rigid general pattern, as, for example, the scheme "lexical Aktionsart before aspect before tense before modus" that might be suggested by scopal characteristics as an enlargement of syntactic structuring. If the forms children use are exclusively analyzed from a target perspective, the illusion of acquired grammatical categories arises--an illusion which might help us as adult interlocutors in interpreting children's utterances in the first place, but which need not mirror the grammatical system of the child. How can we distinguish between tense and aspect in early child data if the utterances in question do not exhibit any properties specific to tense or aspect, not to speak of combinations of these categories, that is, utterances with more than one event talked about?

The developmental model proposed here interprets the dynamics of form-function relations in child language according to cell divisions: the first two basic cell separations emerge out of interpretational biases, which are associated with the two most salient morphological oppositions. They imply a time-related fission into a "default fixation," which links the topic time to the here&now, and a "default fixation," which links the topic time to a time different from the here&now. The topic time is the time about which something is said (Klein 1994, 1998, forthcoming), and a restricted set of corresponding prototypical illocutionary forces is attached to the two possible fixations of the topic time. I claim that in tense-oriented languages like Dutch and German, it is the discovery of the distinction between infinitival and finite marking which is responsible for this first fission, whereas in aspectual-oriented languages like Russian, it is the perfective/imperfective opposition which is used to convey the same basic distinction in temporal reference BY IMPLICATION (then to be complemented by the finiteness distinction in a further step). The term "by implication" is meant to indicate the main difference between the finiteness-bound fission and the aspectual fission in forms: the temporal reference of the set of verbs used with a perfective verb stem point to the fulfilment part of a wish ("future/modality effect") or the target state of a two-state verb ("resultative/completion with pastness effect")--a fission which is accidentally operative due to the characteristics of the different sets of verbs involved without exhibiting an oppositional distribution in forms. For Greek however, early productive aspectual oppositions and their clustering with finite (tense) morphology are attested.

These morphological based fissions provide the first step towards the creation of an abstract grammatical tense/aspect/mood system (henceforth TAM system); they form the bootstraps into grammar a child can rely on. Summing up, the effects of the first fission are cross-linguistically the same: reference to the here&now or reference to a time point different from it as the very first step towards a TAM system.

The individual markings chosen by a learner depend on the repertoire offered by the target language and their saliency hierarchy, in addition to the developmental context (first cell fission vs. further divisions). And there may also be individual preferences. It is only later that the early conglomerate of temporal reference gets formally and functionally expanded and subcategorized into our familiar categories of tense, aspect, and mood. It should be noted that the morphological finiteness marking as such does not provide a fully fledged functional category "finiteness" as, for example, discussed under the header "IP" for "inflectional phrase" in a generative framework (for an account of the acquisition of "finiteness" in Dutch, see Jordens 2002). Crucial characteristics of IP-headed structures are still absent at the RI stage as, for example, the V2 phenomenon and the topicalization of non-nominative NPs. But the early morphological differentiation into [+/-finite] and its functional counterpart for temporal interpretation serves as a stepping stone for the emergence of a TAM-related first functional segregation.

The next section gives an overview of the phenomena subsumed under the label RI. In Section 3, two German case studies present a window on emerging form-function relations. Section 4 then introduces the cell-partitioning model for the acquisition of early form-function relations, and in Section 5, its consequences are discussed.

2. Cross-linguistic overview on root infinitive structures

As illustrated by the examples above, RIs occur in varying forms and convey varying interpretations. Let us first consider the range of variation across languages in more detail, with particular focus on the role of morphological markings of finiteness and its range of interpretation. I will concentrate on the languages cited above, although they surely represent only a small fraction of the language families in the world.

The syntax-oriented generative discussion about root infinitives in recent literature is primarily concerned with their imputed tree structure, especially with the (under)specification of higher-order functional nodes (Rizzi 1994; Wexler 1994; Hyams 1996); varying views of the functional role of root infinitives, that is, their modality, are used as an argument with respect to the much disputed issue of optionality in the child's grammar, see Jordens (1990), Ingram and Thompson (1996). The following empirical section tries to avoid any a priori analysis pro or contra a specific tree structure. The empirical facts will pose the empirical challenge, which the model to be presented later must meet.

In each of the following subsections, we will first present some examples in context (as far as available), such as to allow the reader a small scale first-hand judgement of the data. Then we sketch the developing form-function relations, which means here the relation between a specific morphological marking on the verb and its preferred or exclusive interpretational array.

2.1. Dutch root infinitives
(8) Matthijs 2;5 (Wijnen 1997: 14)
 eendje zien
 duck-DIM see-INF
 'I want to see the ducky.'

(9) Peter 2;1 (Blom 1999: 16)
 Peter Bal pakken
 Peter ball get-INF
 (Context: Peter wants to get the ball.)

(10) Josse 2;8 (Blom 1999: 16)
 op Kist zitten
 on box sit-INF
 (Context: Josse wants his mother to sit on the box.)

(11) Abel 2;5 (Blom 1999: 16)
 ah, mij bril vallen [= mijn bril valt]
 ah my glasses fall-INF
 (Context: his glasses are falling.)

(12) Laura 2;4 (Blom 1999: 17)
 boot svaje [= varen]
 boat sail-INF
 (Context: refers to a picture with a sailing boat.)

On the form side, it can be noted that the morphological marking -INF (orthographic -en) does at least not violate the assumption the INF-marked verbs occur in clause-final position (due to the occurrence of bare isolated infinitives) which gives the child two different formal factors at hand. The term -INF marking is used to separate the infinitival -en marking from its counterpart -en in the finite paradigm for 1st and 3rd plural forms. Overt subjects might surface as in (9), (11), and (12), or stay implicit as in (8) and (10). With respect to the associated functions, modality-related interpretations, that is, wishes, as in (8) through (10) are complemented by nonmodal, that is, early "declarative" ones (11) and (12), which refer to ongoing events and picture descriptions.

In Blom (2002) it is shown that the children Abel, Daan, Josse, Laura, and Matthijs produce approximately the doubled amount of RI with modal interpretation compared to the amount of RI with noumodal interpretation in one-word utterances (see Blom 2002: 117ff.). Only the child Peter does not adhere to this picture, his nonmodal RIs outnumber their modal counterparts.

By far, most root infinitives are eventive verbs (between 93% and 100% in the four child corpora in Wijnen 1997), whereas finite structures show a more or less equal distribution of eventive and noneventive verbs; they range from 36% in Peter's corpus to 57% in Matthijs' corpus. Root infinitives prefer a temporal reference "future" in 86% of all the analyzed cases, for noneventive root infinitives this tendency is even stronger (compare also Jordens 1990). But finite verbs occur more or less exclusively with a temporal "present" reference ranging from 87% in Peter to 99% in Josse (all percentages from Wijnen 1997). The complex relationship between temporal reference and modality is looked at in more detail in a later section. Regarding the morphological status of the infinitive-marked verbs, it has been argued that particle verbs occur more often in nonfinite form than verbs without a particle (Jordens 2000: 23).

Explicit infinitive marking is reported to emerge earlier than finiteness marking and to outnumber its finite counterpart for a time-span up to several months (cf. Jordens 1990; Wijnen 1997). The class of root infinitives can be split up again into modal and noumodal used root infinitives to the effect that the nonmodal ones occur prior to the modal ones (Blom 1999). But there is considerable interindividual variation. According to Blom's study of six Dutch children, "(t)he numbers of [-realized] (= modal use, PG) and [+perfective] (= nonmodal, amounts primarily to commenting on ongoing action, PG) RIs differ significantly from each other, the children differ significantly from each other and the use of [-realized] and [+perfective] RIs differs significantly from child to child" (Blom 1999: 18f.).

2.2. German root infinitives
(13) Max 2;9 (Tubingen corpus (1))
 (Context: Max puts the toy toilet in the bathroom of the
 playhouse; he closes the cover of the toilet whilst uttering.)
 hier\ des zumachen here this close-make-INF
 'I am closing this'

(14) Katrin 1;5 (Ingram and Thompson 1996)
 M: Was mochtest du haben? 'What do you want to have?'
 K: Stift haben?
 Crayon have
 M: Ach, du mochtest einen Stift haben. 'Yes you want to have a
 (the mother gives Katrin a pencil)

(15) Max 2;9 (Tubingen corpus)
 (Context: playing with Lego; the adult interlocutor
 S. holds a piece
 in her hands.)
 M: n des auch n dach sein ? that-one also ?a roof be-INF
 'That one should be/is a roof, too.'

Not surprisingly, German and Dutch are quite similar in their form-function relation of root infinitives: just as in Dutch, the -en marking is used as the corresponding INF-marking, and the -en marking is shared by the 1st and 3rd finite paradigm forms. The interpretational range of German root infinitives covers commenting on ongoing events (13), as well as clear modal cases (14) and lesser clear instances as, for example, (15), which could involve deontic modality or plain existential predication.

Behrens (1993) reports that root infinitives can refer temporally to past, present, and future and that--at least for the child Simone--root infinitives which refer to the present will diminish as the finite present tense paradigm gets more productive. Ingram and Thompson (1996: 113) found in their German data a modality bias of a mean percentage of 79% in root infinitives. If ambiguous so-called possible modal interpretations are excluded, there still remains a fraction of 55% strictly modal root infinitives. Concerning the morphological status of the infinitive-marked verbs, they noted that 72% of them are complex predicates, whereas finite marked verbs show only 4%.

2.3. Greek root infinitives
(16) Spiros 1;9 (from Stephany's data, CHILDES database;
 RI-interpretation from Varlokosta et al. 1996)
 S: tuto seli
 this want-IMPF-NONPAST-3SG
 'I want this.' (lit. 'He wants this.')
 U: afto thelis?
 this want-IMPF-NONPAST-2SG
 'Do you want this?'
 S: afto -- seli.
 this want-IMPF-NONPAST-3SG
 'I want this.' (lit. 'He wants this.')

(17) Mairi 1;10 (Stephany 1985: 79)
 kiki for: "na kimi[??]-i"
 'He shall (go to) sleep.'

The target language Greek has no morphologically marked infinitive. Nevertheless, Greek child data have been discussed under the explicit header RI following a usage-based definition (see Varlokosta et al. 1996, 1998). So, tense-marked structures, that is, the form with the suffix -i (perfect participle form or third person singular agreement marking), is supposed to figure as a root infinitive in Greek as in (16).

This analysis opposes the thorough description of Greek language acquisition by Stephany (e.g. 1985, 1997) where it is shown that the na-subjunctive construction and its predecessors forms the closest correlate to the Germanic RI phenomenon in expressing nonfactuality (compare [17]). Whereas Varlokosta et al. (1996) argue that an exclusion of this na-construction as RI equivalent is justified by the fact that this form does not constitute the earliest and the most common sentence form, Stephany's data analysis exhibits the opposite. So, the subjunctive form constitutes together with the present indicative form the most frequent verb form used by children from 1;10 to 2;10 (Stephany 1997: 202f.). Note that the question of the functional potential of Greek root infinitives has been reduced a priori to a modal interpretation in both approaches due to the specifics of Greek grammar.

Concerning the form of the subjunctive in Greek, it falls together with the future marking except for their different preceding particles which mark this opposition in interpretation: the future interpretation requires [??]a marking; the modal interpretation requires na marking. As indicated by the term predecessor, this distinction between interpretations is blurred in early child language due to the omission of the particle (as in example [20]) or the typical occurrence of the underspecified marker a. In general, the acquisition pattern for the subjunctive/future marker follows the order: subjunctive form without particle < subjunctive form with optional global particle na/a in less than 50% of tokens < particle [??]a/a and na/a optionally used in more than 50% < particles used in more than 80% of tokens, with particle types distinguished in at least 50% of these (after Stephany 1997: 248).

2.4. Russian root infinitives
(18) Sasha P. 1;8 (Brun et al. 1999)
 (Context: This utterance was produced after the child
 had put on
 his pants.)
 '(He) has put (the pants) on'

(19) Sasha J. 2;4 (mental age: 1;6) (Brun et al. 1999)
 (Context: The boy is describing the actions of his sister,
 who is playing with her toy stroller in the same room.)
 kacat' kolyasocku
 swing-IMPERF-INF stroller-DIM
 '(She) is swinging the stroller.'

(20) Varvara 1;7 (Brun et al. 1999)
 M: zacem ty zvonis' (po telephonu)? 'Why are you calling
 on the phone?'
 V: Pozdravlyat' babusku
 Congratulate-IMPERF-INF grandma
 '(I will/want) to contratulate grandma.'

(21) Zhenya 1;9 (Brun et al. 1999)
 (Context: The boy wants to clean up the mess by himself.)
 sam, sam ubrat'
 self self clean-up-PERF-INF
 '(I will/want) to clean up by myself.'

The aspectual markings of perfective (PERF) and imperfective (IMPERF) are the first TAM-related markings in verbs to appear in Russian.

This is due to the aspectual information encoding in the verb stem itself. So, this earliest distinction may not be productive at the outset, but it has repercussions on the correlated interpretations of the individual verb forms from early on. Concerning the morphological infinitive marker in Russian, this is unambiguously expressed by the suffix -t'. The peak of Russian root infinitives is reached around 1;8, which is distinctly earlier than that of the Dutch and the German children from the corpora cited in the preceeding subsections.

The examples above illustrate the broad potential of temporal reference which Russian root infinitives display in combination with the mandatory aspectual marking. It includes reference to past, as in (6) and (18), reference to present, as in (19), and reference to future/expressing wishes, as in (20) and (21). In general, past reference is restricted to verbs marked as perfective, and reference to the present is limited to verbs marked as imperfective. Only future reference, which is closely intertwined with the functionality of wishes and requests, is possible with both aspects in child language. Brun et al. (1999: 112) give the following percentages across the pooled root infinitival utterances from three children: past reference (26.1%), present reference (48.3%), and future reference (25.6%). These findings suggest that Russian root infinitives do not carry a specific interpretation as, for example, a deontic/requesting function which may also be due to the pooled data analysis. With respect to a more detailed path of acquisition, it has been reported in Bar-Shalom and Snyder (1999) that Russian children start with an imperative or modal interpretation of root infinitives before they produce declarative root infinitives. Thus, the Russian infinitive marking correlates in the early stages with prospective/ modal interpretation associated with requests and wishes as a developmental trend.

2.5. English root infinitives

It is difficult to justify the label "infinitive" for English root infinitives, since finite and infinite forms are identical except for third person singular (and some idiosyncratic forms such as are). The major fission in the early system is between bare or infinite forms, on the one, and finite forms (s-marked forms or -ed marked forms), on the other. Neither form has an interpretational bias comparable to the Dutch and German root "infinitives" tend structures. According to Ud Deen (1997), English root "infinitives" tend towards a nonmodal interpretation. A modal interpretation, in contrast, is indicated by lexicalized boulemaic elements such as want or wanna in early child English (see also Hoekstra and Hyams 1999). A typical example is the utterance want more apple from Erik (2;1), given in Hamann (1996: 155). This is in line with the fact that in adult English, the bare infinitive is marked for the feature [+perfective], which renders it incompatible with progressive interpretation (see Giorgi and Pianesi 1997). Consequently, it has been suggested that it is the element to which conveys modality, rather than the bare verb (Blom 1999). But in contrast to the Greek case of na, the particle to plays no role in signalling early modal or deontic interpretation.

According to the cross-sectional study of de Villiers and de Villiers (1973), the acquisition order of morphemes in American English is -ing < irregular past tense markings <-ed <-s. A correlation between the -ing marked forms and atelicity vs. -ed marked forms and telicity has been observed (cf. Bloom et al. 1980). None of the above-mentioned forms conveys a typical root infinitive functionality. Whereas the early -ing forms are initially restricted to comments on ongoing events and only later encode the abstract aspectual information of imperfectivity, -ed marked verbs signal completion of events.

2.6. Summary

The following table sums up the main cross-linguistic form-function relations including their developmental progression.

This description might suggest that an unambiguous identification of the morphological infinitive marking is always possible. But as already mentioned above, this is not necessary the case. In the languages cited above, the infinitival suffix is often homophone with another ending in the verb paradigm. In German, infinitive, first person plural, and third person plural are equally marked by -en; in Greek, the finite present subjunctive form and the nonfinite perfect tense form are homophone. Note that the developmental progression of root infinitives in Dutch and German reverts the development of Russian. This fact is not only a problem for modality-centered accounts of root infinitives, such as, for example, Hoekstra and Hyams (1999); it is also a challenge for any cross-linguistically homogeneous conception of the development of form-function correlations.

As for temporal reference, the range of inter- and intralanguage interpretations for infinitive forms suggest a rather indirect and flexible association of temporal information with these forms. In addition, the categories for the assessment of the correlating interpretation of an individual verb form are heterogeneous across the various analyses. So, temporal, aspectual, mood-related, or "Verbkategorie"-related labels have been considered to form the relevant interpretational distinction if one had to be modeled.

As the developmental progression indicates, the form-function relation of root infinitives is sensitive to changes in the child's grammatical system. Otherwise the target state could not be reached at all. Note, furthermore, that children vary considerably in the way in which they approach the same target language, as has been shown for the case of German by, for example, Fritzenschaft et al. (1990), Tracy (1991), d'Avis and Gretsch (1994), Gretsch (2000). The observed heterogeneity can be smoothed by using more fine-grained stage divisions, but we will never get rid of a certain amount of demarcating individual or at least strategybound characteristics. I assume that a similar diversity would become apparent in the other languages, too, as soon as the pooled data analysis gets split up interindividually.

A discussion of root infinitives would not be comprehensive without touching upon the issue of overt vs. dropped subjects. Kramer (1993) has shown that the occurrence of subjects correlates with the finiteness status of the verb: root infinitives prefer dropped subjects, whereas finite structures prefer overt ones. A possible explanation for this correlation has been found in the mechanism of (nominative) case assignment such that only structures with a raised finite verb in [I.sup.0] allow for a nominative case marking of the subject. But the array of root infinitive examples above faces us again with a cross-linguistically and language-specifically heterogeneous picture that renders the case-bound explanation suspiciously inflexible. Apart from the subject issue the occurrence of wh-elements (Rizzi 1994; Haegeman 1995) and the occurrence of focus particles (Penner et al. 1999) have been shown to have an effect on the (in)finiteness marking of the verb.

To sum up, the cross-linguistic overview faces us with a whole range of phenomena which are lumped together under the label "root infinitive." These phenomena share some properties in form and function. But at the same time, there are also many differences with regard to form and interpretation, which leave purely syntax-oriented approaches with many problems. This will be even clearer in the following two German case studies.

3. Two case studies of German root infinitives

My own empirical study on two German children revealed a complex picture of emerging form-function relations. For presentational purposes, three different developmental points roughly within the "optional infinitive stage" were selected. For the child Max, a total of 1101 utterances were analyzed, the three data points comprise the corpora from 2;06.29 and 2;09.00 and 2;11.18. For Benny, the total of utterances sums up to 1108 at the age points 2;02.15 and 2;10.28 and 3;00.18, which are roughly comparable in development since both learners, but Benny in particular, were slow learners.

3.1. Assessment of form and function

Concerning the form side, not only finite and infinitival structures have been taken into account, but also utterances without a verb, such as plain DPs, utterances exhibiting only a verb-associated particle, and other truly nonfinite structures. The reason for their inclusion is that if infinitival structures and nonfinite structures were to have the same interpretational potential, the interpretational bias could hardly be localized at the infinitival morphology exclusively. Examples (22) to (24) illustrate three nonfinite structures with requesting, descriptive, and commenting interpretation, respectively:
(22) Max 2;09.00 (Tubingen corpus)
 (Context: Max is looking around for a Lego window.)
 M: gro[??]e Fenster big window
 S: Ein gro[??]es Fenster soll da noch rein? (A big window should
 be in there?)
(23) Max 2;09.00 (Tubingen corpus)
 (Context: Max is fetching the bag with toys.)
 M: tasche bag
 S: Und da ist die Tasche. (And there is the bag.)
(24) Max 2;11.18 (Tubingen corpus)
 (Context: Max throws the tractor over.)
 M: UMfal\ (2)
 S: Oh, oh, jetzt hat se sich den Kopf angehauen. (Oh, oh, now
 she has bumped her head.)

Whereas the category of verbless structures is easy to determine, it is less easier with the category of finite vs. infinitival structures. The criteria taken into account for the latter distinction are the morphological ending, the positioning of the verbal element in the utterance, and the positioning of verbal prefixes and verbal particles in relation to the rest of the verbal part, if existent. For the verb ending to count as finite or infinitival, agreement and its intraindividual reliability were also checked. The positioning factor expected infinitival verbs and verbal particles in verb-end position and finite verbs roughly in verb-second position allowing for some flexibility within that topological frame. This distribution correlates with the target system of German. The flexibility alluded to possible verb-first, verb-second, or in Benny's case, also verb-third positioning of the finite verb since the grammatical sentence frame is still under construction (for the emergence of this sentence frame from an ordering determined by a conceptual organization of elements, see Dimroth et al. (forthcoming). One-word utterances were categorized according to their morphology only. The placement of verbal particles played no role if the agreement marking was unambiguous. Thus, my classification of form is very conservative, which was necessary to achieve a maximal independence of interpretational factors.

Regarding the interpretational partition, the above examples illustrate under (22) a request function, under (23) an existential/descriptive function, and under (24) a comment on ongoing events. These were the central categories regarding the function side. In addition, the category of personal narratives is interesting because of its clear but rare past time reference. But concerning the frequency of narratives in early child language it plays only a minor role. This lead to the following four-way classification system of interpretation: (i) requests/wishes, (ii) existentials/ stative descriptions, (iii) comments on ongoing events, and (iv) personal narratives of past events. The criteria for classification included cotext and context information under the assumption of a coherent baseline of conversation and action. With respect to the partitioning of the child's speech into utterances, the intonational factor was the main guideline (the transcripts have been checked against the audio data by a second person who was not involved in the taping sessions).

Input repetitions, unclear cases, yes-no answers, and onomatopoeic utterances were omitted altogether from the analysis. Moreover, questions were also excluded. The category of answers in question-answer pairs has to be treated separately, too, since infinitival structures that occur turn adjacent to a question could be likewise interpreted as a target-adequate grammatical ellipsis. In the case of grammatical ellipsis, the utterance was to be counted as finite. As expected, the development of these answers in misleadingly "root infinitival" form follows a steady rise and thus reflects the growth of the [+finite] marked verbs which will eventually constitute the syntactic default of verbal markings according to the target. For Max, the percentage of these cases of possible adjacency ellipsis starts out with 0% by the first data point and climbs to 75% at the third data point, that is, three-quarters of his speech are answers to questions of the interlocutor. This rise also documents the growing mutuality of adult-child conversation and its more and more interlocking character. The respective numbers for Benny are 25% to 37%. Thus, this category of answers has to be set apart from the analysis of the development of form-function correlations centering on ROOT infinitivals.

This functional frame of the four above-mentioned pragmatic and text-oriented categories enables a descriptive and systematic approach which avoids a too narrow and too restricted classification.

3.2. The data

The overall development on the form side shows a steady growth of the finite structures, a rise and following decline of infinitival structures, and a steady decline of the verbless utterance organization. This holds for Max' development (Figure 1) as well as for Benny's development (Figure 2). The numbers in the columns represent the absolute numbers, on which the proportions in the figures are based.

The developmental of Benny and Max is roughly in line with the findings of Lasser (1997). For Dutch, a similar trend is reported in the cross-sectional study of Wijnen and Bol (1993), although the individual differences in time-course and magnitude of the phenomenon are enormous (see also Lasser 2002). Note again that questions, answers from turn-adjacent Q-A pairs, input repetitions, yes-no answers, onomatopoeic utterances, and unclear cases are excluded, which explains the overall low numbers. Especially the Q-A pairs usually constitute the bulk of turn pairs in adult-child communication of the typically studied sort, where the adult interlocutor is notoriously concerned with making the child talk.

For examples of the category verbless see (22) to (24). Infinitival structures are illustrated in (2), (13), and (15) from earlier sections. More infinitival structures and their finite counterparts are presented below (restricted here to one infinitival and one finite example for the first and the last data point of each child).

The child Max (data points from 2;06.29 and 2;09.00 and 2;11.18):
(25) Max 2;06.29 (Tubingen corpus)
 (Context: Max puts a seat into the Lego car. S. points to the
 steering wheel, which is located in a way that the Lego man can
 not sit in the car.)
 S: Aber jetzt ist doch das Lenkrad viel weiter unten.
 (But now the steering wheel is much further down.)
 M: n fehlt was
 there? is-missing FIN something
 'There is something missing.'

(26) Max 2;06.29 (Tubingen corpus)
 (Context: Max fetches a book about space flights.)
 M: des des hier auch lesen
 that that here also read INF
 'I want you to read that, too.'
(27) Max 2;11.18 (Tubingen corpus)
 (Context: Max puts Mickey Mouse on a toy trailer.)
 M: so Mickey Mouse fahr mit
 so/now drive FIN with
 'Now, Mickey Mouse come with us.'
(28) Max 2;11.18 (Tubingen corpus)
 (Context (picture book): Max looks at a fireman who is attaching
 a waterhose.)
 M: derda Wasser so anmachen
 this-one water so on-turn INF
 'This one is turning the water on.'
 S: Der macht das Wasser an, hmhm. (He turns the water on,

The child Benny (data points from 2;02.15 and 2;10.28 and 3;00.18):
(29) Benny 2;02.15 (Tubingen corpus)
 (Context: Benny is drinking milk.)
 B: Milch trink ich
 milk drink FIN I
 'I am drinking milk.'

(30) Benny 2;02.15 (Tubingen corpus)
 (Context: Benny wants to sit in his highchair.)
 B: naisitze Bub
 in-sit INF boy
 'I (= Benny refers sometimes to himself as 'Bub') want to sit
 in there.'
(31) Benny 3;00.18 (Tubingen corpus)
 (Context: Benny and I. are playing road barricade; Benny wants
 to get through.)
 B: aufmachen wieder
 open INF again
 'Open the barricade again!'
 I: (I. pushes B. away.) Brrm. Umleitung, tschuB! (Brrm. Bypass,
 (32) Benny 3;00.18 (Tubingen corpus)
 (Context: following immediately after (31))
 B: ach aber gleich geh ich wieder rum
 oh but soon drive FIN I again round
 'Oh, but I will soon come back again!'

A requesting character of the child's utterance is present in (26), (27), (30), and (31), whereas the examples under (25), (28), (29), and (32) show a nonrequesting, descriptive, or commenting function. This split is not correlated with the structural infinitival vs. finite division as the small example section already indicates.

The following figures visualize the distribution of the four interpretational categories (i) requests, (ii) descriptions, (iii) comments, and (iv) narratives across the different structure types. The first diagram presents the overall distribution of interpretations across the cleared utterance total (Figure 3). This distribution represents the background of our analysis in modeling the proportions of the interpretational categories at the three data points. If we compare the interpretational distribution of the utterance total with the structural type of the verbless utterances in Figure 4, the first data point hardly shows any difference; the second data point excludes comments from the verbless structures, but the rest of the proportions is interchangeable again. Regarding the last data point, the pattern emerges that narratives are excluded from being expressed by verbless utterances; moreover, the category of comments returns as a relevant segment for the verbless type, but to a reduced amount. As can be depicted from Figure 1 at the beginning of the data section, the type of verbless utterances constitutes three-quarters of the structure types for Max at the first data point and reduces to one-quarter by data point three, which explains the similarity of the two bars at the first data point and their slight separation with respect to the other two data points. The structural type of verbless utterances shows no bias towards a specific interpretational category, but it excludes narratives in the case of Max. In general, the type of verbless utterances diverges more and more from mirroring the proportions of the utterance total without losing much of its interpretational versatility.


This development is reversed if we compare the interpretational distribution of finite structures against the background of the utterance total (Figure 5). Here, the finite structure type adjusts more and more towards the distribution of the utterance total. Finite structures are getting more and more versatile and supercede the whole spectrum of interpretational categories eventually qualitatively as well as quantitatively (see Figure 1 for the quantitative distribution, where the amount of finite structures rises up to 70% at data point three). The descriptive and the commenting function constitute the core categories within the spectrum of the finite structures at the outset. The interpretational bias is then given up for the benefit of a strictly grammar-driven use of finiteness, as is required by the target language.


In contrast to the development of finite structures, infinitival structures, that is, structures which entail an infinitival verb and no finite verb, show a raising tendency towards a requesting function, but no sharp exclusion of interpretation in earlier stages (Figure 6). With the raising integration of finite structures into the overall system, the requesting function of the type of infinitival structures gets stronger. Moreover, the commenting interpretation, which poses the second strongest functional segment within the infinitival structures, gets concomitantly weaker or even diminishes at all.


To sum up the development of Max (i) verbless structures show no bias towards a specific interpretational category and they exclude narratives; (ii) finite structures start out with descriptive and commenting functions and allow eventually to represent all functions, that is, to get independent of interpretational biases due to their grammaticization; (iii) infinitival structures show a bias towards a requesting function and a commenting function with a decreasing weight of the latter. Thus, a comparison between verbless structures and infinitival structures reveals differences in what they are used for and justifies their separation into two different structural categories. Moreover, the segment of finite structures vs. infinitival structures displays a reversed development: from multifunctionalism to monofunctionalism in the case of infinitival structures, and the opposite direction towards diversification, from bifunctionalism to multifunctionalism, in the case of finite structures.

Benny's development shows commonalities, but at the same time, individual differences. His overall distribution of interpretations across the utterance total includes narratives from data point one on, and the interpretational categories are more evenly distributed across all data points (Figure 7). The verbless structures display no interpretational bias towards any of the functional categories (Figure 8). The quantitative drop of the verbless structures (see Figure 2) is as drastic as in Max' case, falling from three-quarters of all structures down to less than a quarter; nevertheless, the verbless structures still mirror the distribution of the utterance total closely. This is in slight contrast to the developmental trend in Max, where verbless structures diverged from mirroring the utterance total.


With regard to the distribution of finite structures in Benny's data (Figure 9), their type adjusts more and more towards the distribution of the utterance total, as was the case with Max. Although the range of interpretations covered by Benny's finite utterances is already wide at data points one and two, the best proportional fit towards the utterance total is reached at data point three. The core categories of interpretation at the outset are the requesting function and the commenting function. Again, any interpretational bias disappears and is replaced by a grammar-driven use of finiteness, as depicted by the similarity of each last data point of finite structures and of the utterance total.


The development of infinitival structures is not as homogeneous as in Max' case but exhibits similarities: the requesting function is again the strongest, and it never drops beyond the 50% boundary for all three data points (Figure 10). The commenting function is the second strongest as in Max' data, but the developmental direction of functional reduction is not shared to the same extent. Moreover, the segment of narrative interpretations plays for Benny a noteworthy role in infinitival structures, whereas Max excluded them from being expressed by infinitival structures.


Summing up, the development of Benny displays that (i) verbless structures display no interpretational bias; (ii) finite structures start with a bias towards requesting and commenting function and get close to the proportional fit of the utterance total at the latest at data point three; (iii) infinitival structures display a strong bond with the requesting function and, to a lesser extent, with the commenting function.

The differences between Max and Benny consist of differences in pace and qualitative differences in the development of finite vs. infinitival form-function relations. Whereas verbless structures do not attract interpretational biases, infinitival structures are prone to do so. But these biases can differ in their individual balancing, although they nevertheless eventually converge towards the same target system. Thus, infinitival structures prominently take requests or wishes as their interpretational relatum. Finite structures are generally less biased. Here, it depends on the individual division of labor between the structure types, whether the structural segregation of requests from comments and descriptive functions is more or less sharp. This also determines the particular developmental directions from quasi monofunctionalism to multifunctionalism or the other way around, depending on the individual state of the system. These findings are compatible with the range of individual characteristics of early form-function relations found in Blom (1999, 2002).

In general, the shifts between form-function correlations are subtle, but also systematic. At the first point in time, the form-function correlations of the utterances in total are mirrored by the verbless category. There is a steady shift towards the finite category such that, in the end, the overall correlations are reflected in the finite utterance organization. The raising integration of finite-marked verbs and, thus, finite structures leads to a strengthening of the form-function relation of infinitival structures and their requesting function. Moreover, the commenting interpretation, which poses the second strongest functional segment within the infinitival structures, gets concomitantly weaker. Overall, the category of infinitival utterances always stays distinct with respect to their functional segments associated across its internal development. Initially, mainly requests and comments are associated with the infinitives; later, this form is narrowed down to cover requests and grammatical ellipsis (in answers). Why should it be that the formal segment of infinitival utterance organization is separated early on? How can individual differences be accounted for within an abstract developmental system? A different view on early form-function relations aims at an answer and will be presented in the next section.

4. Early cell partitioning

4.1. The model

The metaphor of cell partitioning allows us to capture early form-function relations and their dynamics in an elegant and parsimonious way and with some predictive power. Following Klein (1994, 1998), I assume that three time variables are relevant for the description of temporal relations: the time of utterance (TU), the time talked about, the so-called topic time (TT), and the time of the situation itself (TSit). The notion of "topic time" can be seen as an interpretation of Reichenbach's often used but completely undefined notion of "reference time." This account allows for a natural link between aspect and tense, both of which are treated as time-relational notions. Tense is viewed as the relation between topic time and the time of utterance. There are three basic choices: in the present tense, TT overlaps with TU (in brief, TT at TU), in the past tense, TT precedes TU (TT < TU), and in the future tense, TU precedes TT (TU < TT). Aspect relates the topic time to the time of the situation. There are two basic choices: in the imperfective aspect, TT is included in TSit, which, metaphorically speaking, "views the situation from its inside, as ongoing;" in the perfective aspect, TT includes TSit, which amounts to the "outside view" in which the situation is shown as "completed." There are other possibilities, for example, prospective aspect (TT < TSit), as well as various subdivisions, not to be discussed here. This definition relates to "grammatical aspect." As to lexical aspect, that is, the inherent temporal characteristics of verbs (and verb phrases), this framework distinguishes 0-state, 1-state, and 2-state verbs; 0-state verbs are "atemporal" (and hardly play a role in early acquisition), 1-state verbs roughly correspond to telic verbs, and 2-state verbs correspond to telic verbs (for details, see Klein 1994).

A target tense/aspect system thus requires the interrelation of three different temporal parameters (restricted by the inherent temporal structuring of the involved verbs). I assume with Smith (1980) and Fantuzzi (1996) that in the early stages of language development, children only manage two of these parameters. In my approach, this amounts to a fixed here&now, which collapses with the TU. Topic time and time of situation are equated; we shall note this as TT/TSit. Thus, children have only TU = here&now and TT/TSit at their disposal; this severely constrains the range of temporal information conveyable by grammatical means.

Additionally, the factor mood crosscuts the tense/aspect system and is closely related to temporal interpretation. For the cell-system, as developed in the following, it is relevant that modal interpretations as requests and wishes comprise a state of affairs which should come about in the future. It is implicit in the request function that this state of affairs does not take place at the here&now, although the wish, the intention is felt and expressed at the here&now. This is labelled here as the future/ irrealis-effect of modal interpretations.

The following stages are claimed to represent a universal evolution of interpretational conglomerates, as reflected in the acquisitional sequence of grammatical choices of form. This scheme is restricted to the onset of morphological oppositions. Therefore, a label such as "first stage" does not exclude earlier stages, which exclusively entail an (extended) here&now frame. Similarly, "Stage 3" does not necessarily form the endpoint of development. The intention is to concentrate on these functionally universal stages in acquisition which are central for a possible encoding into a root infinitival form during the so-called optional infinitive stage.

First fission. The first stage which gets associated with a morphological opposition comprises only two temporal parameters: TU and TT/TSit. TT/TSit can be fixed to a point at the here&now (default interpretation), or can be fixed to a preceding or following point with respect to the here&now (nondefault interpretation). Thus the nondefault interpretation subsumes references with "pastness-effect" and with "future/irrealis-effect."


Second fission. At the second stage, there are still only two temporal parameters involved (TU and TT/TSit); but now, there are more fine-grained relations between them, thus superimposing a "temporal directionality" to the three-way partitioning of the default interpretation vs. a perfective/completion/"pastness" interpretation and a progressive/ "futureness"/irrealis interpretation. Languages differ as to whether they exhibit a basic realis/irrealis split in their temporal system (such as Inuktitut) or whether they are based on a past "tense"/nonpast "tense" split, as, for example, German. The different dimensions associated with those two basic partitionings might not be analyzed as such in the child system.


Third fission. The third stage is characterized by profound changes initiated by the emergence of the third temporal parameter. With the separation of TT and TSit, the stage is open for a productive combination of the tense-bound and aspect-bound properties. (The distinction between perfect aspect and perfective aspect has to be treated as a separate issue and is not directly relevant for the exposition of the model.)

This model of development roughly sums up the expansion and partitioning of functional categories that are associated with a language-particular (and, to a lesser extent, a person-specific) morphological form, for example an infinitival form, a participle form, or an aspectual form from the target languages. The morphological markings point thereby to functional conglomerates and only stepwise unfold the multidimensionality of the system (cf. also Bowerman (1985), who conforms the sensitivity of children to their respective target language with respect to early temporal oppositions). A closely related view on the development of temporal relations has already been proposed in Guillaume (1929), where he writes:

Aspect, mode, temps ne se referent pas, comme l'enseigne la grammaire traditionnelle, a des phenomenes de nature differente, mais aux phases internes d'un phenomene de nature unique: la chronogenese; en un mot, l'aspect, le mode, le temps representent une seule et meme chose consideree en des moments differents de sa propre caracterisation (Guillaume 1929: 11).

Coming back to root infinitives, I claim that depending on the language-particular saliency hierarchy of morphological forms, the infinitival form as a formal expression of a functional conglomerate enters the above scenario earlier or later. This explains the varying interpretations and differing developmental progressions across languages. I will show this in the following sections, where tense-oriented languages get contrasted with aspectual-oriented languages.

4.2. Tense-oriented languages

In tense-oriented languages, the first fission between default and nondefault interpretation is expressed by finiteness marking: the default here&now interpretation during the first stage is associated with the grammatical default marking [+finite], whereas the nondefault interpretation "different from the here&now" is related to the nondefault marking [-finite]. Note that the marking for [-finite] has to be distinguished from the absence of finiteness marking at all [[??]finite], which is not grammatically associated with any interpretational bias (as also indicated by the empirical findings).

In the Dutch and German data, root infinitives are primarily interpreted as boulemaic modals (wishes/requests) and, secondary, as comments on ongoing events. The boulemaic modal segment nicely fits the "different from the here&now" interpretation, due to the future/irrealis-effect. The wish-world reference of [-finite] is corroborated by the use of infinitival markings in counterfactuals (see also Wijnen 1997 and for a cross-linguistic overview of future and modality marking in the Germanic languages, Abraham 1989). (3)

Comments on ongoing events at first seem to run against the above model, since their most obvious relation to the here&now should favor a finite marking. But considerations of economy suggest the following commonalities between both functional categories. The explicit marking [-finite], as indicated by infinitival marking in Dutch or German, is either motivated by reference to a non-here&now time/world (wish world/ modality) that the learner is not yet able to express by target-like grammatical means, or by reference to a parallel ongoing action; in the latter case, the shared focus of attention of speaker and hearer makes an explicit marking of the topic time superfluous. Here, the prominent nonverbal context overrides any other interpretational bias. In the same way, the grammatical marking [[??]finite] is expected to likewise constitute a perfect conveyor of commenting ongoing events, a fact which is confirmed by the verbless data. But in contrast to the commenting function, the grammatical nonmarked form [[??]finite] gives way within the functional category of boulemaic modals as soon as the infinitival marking emerges. The following table gives an overview of the possible form-function correlations after the first fission within a tense-oriented system.

During the following development, verbless structures and infinitival structures narrow down more and more to answers in turn-adjacent question-answer pairs. This means that the finiteness marking becomes obligatory, and verbless structures and infinitival structures are restricted to cases of grammatical ellipsis.

The eventivity constraint (4) on root infinitives suggested by Hoekstra and Hyams (1999) can be interpreted as a consequence of the specific topic time setting of root infinitives that prefers an action-orientation in the comment segment and in the modal segment. Hoekstra and Hyams (1999: 246) also claim that "the modal interpretation of children's RIs (= root infinitives, PG) (and adult RIs) is determined by the inherent quality of infinitives as being marked [-realized]." This feature association might hold for later stages in acquisition, but definitely not for the optional infinitive stage. This is deafly demonstrated by the many non-modal root infinitives from Section 2. At the two first points in time, the root infinitives of Max and Benny show only a slight bias towards the modals in relation to comments: 61% modals vs. 39% comments, if all other functions (some root infinitives convey picture descriptions, some convey narratives) are ignored. If these are included, the percentage even drops to 45% for modals. Their share then increases at the cost of comments, just as the share of true grammatical ellipsis (answer category) increases. At the third point in time, we have 41% modals, 46% grammatical ellipsis, and only 13% comments across the two children. This development reflects the progression from nonmodal towards modal in tense-oriented languages as noted in Table 1. In Hoekstra and Hyams (1999), the feature [-realized]--whatever its grammatical status is--is directly associated with the infinitival morphology. The present proposal states an indirect relationship between a "different from here&now" interpretation and the infinitival form mediated by a topic time setting. The only feature directly attached to the infinitival morphology is [-finite = grammatical topic time set to nondefault]. This makes any considerations of how children should "unlearn" a fixed link between infinitival form and [-realized] as a general "modalizer" obsolete.

In a further fission of the [-finite] category, the form of the past participle (ge- verb stem -t) is introduced to express that the target state of 2-state verbs obtains (resultative or completion effect in cases like runterge-fallen 'fallen-down'). This implies that the event which led to the target state has taken place at a time that is "different from here&now," which refers to the common header of infinitives and past participles. Since root infinitives show a preference for particle verbs and since these particles are already used before in order to express a "result state of causation" (Jordens 2000), the functional category of resultatives is already familiar to Dutch (and German) children. But only at that distinctly later stage, marked by the emergence of participle forms, this category is related to a morphological productive opposition.

The following figure gives an overview on the cell development of tensed-based languages like Dutch and German. (5)


The stepwise unfolding of structural and functional categories and their relations is captured in the subsequent cell divisions. This account models the empirical data as described in Section 3 in a detailed but, at the same time, flexible system. Since verbless structures were shown not to attract interpretational biases, structures with [[empty set]finite] play no role regarding the form-FUNCTION mapping and its development. Infinitival structures are prone to take prominently requests or wishes as their interpretational relatum due to the boulemaic effect of the [-finite] form as argued above. This early fixation of a form-function link explains why the category of infinitival utterances always remains distinct with respect to their functional segments associated across its internal development. This is in contrast to the finite and verbless structures as the figures from Max' and Benny's development indicate. The early mapping of [-finite] structures and requests can still be detected in the target language as a fallback mechanism. Through the whole chain of linguistic developments towards the target, the roots of the system are never forgotten and can be reactivated. Thus, every cell division already encodes a small piece of the target system without revealing the complete picture in advance.

The different individual interpretations of a specific structural form are accounted for by the internal flexibility concerning the functional division of labor between the cells. For Max, infinitival structures eventually exclude the functional segment of comments, whereas for Benny, the "economy-driven" interpretation crosscuts all form types through all data points. Note also that the generally small segment of narratives relates primarily with infinitival structures in Benny's case. This nicely illustrates the [-finite as +past participle] cell from the figure above.

Having made the first step of a [+/-finite] split, there is still a long way to go towards a target-like grammatical finiteness marking. Within a stage-model of acquisition as laid out in Dimroth et al. (forthcoming), the first cell divisions can be interpreted as a stepping stone towards a correlation of formal (including topological information) features and interpretative features, which is necessary to enter the central cognitive ordering stage. Summing up, the formal segment of infinitival utterance organization is separated from early on due to the default/nondefault opposition of forms and their related interpretation in terms of time/ world continua in acquisition.

4.3. Aspectual-oriented languages

In contrast to tense-oriented languages, aspectual-oriented languages enter the system of form-function relations from a different angle. Here, the first fission is associated with the salient formal properties of an aspectual opposition--judged from the target perspective. It is only at the next stage that the [+/-finite] split becomes relevant. The temporal information conveyed by a finite- or infinite-marked verb is nevertheless the same as in tense-oriented languages--but put on top of the already existing aspectual distinction: [+finite] is associated with the default here&now interpretation and [-finite] is linked to the nondefault "different from the here&now" interpretation. But since this morphological opposition occurs in another developmental context, the interpretation-related effects differ. This mirrors the modeling of the interindividual differences encountered earlier on a wider cross-linguistic range.

The temporal information conveyed by the prior aspectual split shows an entangled "Aktionsart" /aspect effect which is correlated with a default interpretation for the imperfective verb stems (ongoing activity) vs. a nondefault interpretation of perfective verb stems (the fulfilment part of a wish "future/modality-effect or the target state of a 2-state verb "resultative/completion with pastness-effect"). This is the reason why the first fission in morphological forms cannot be universally interpreted as a realis/irrealis bifurcation. In aspectual-oriented languages, it is the second split which introduces "temporal directionality" and thus a future/ irrealis-bound pointer.

In the case of Russian, we not only find the formal aspectual opposition of [+imperfective] and [+perfective], but also an early [+imperative] marking. Equipped with this basic repertoire of morphological markings, the Russian child can already cover the corresponding basic functions--commenting ongoing events and descriptions for the default-bound imperfective form, resultative or completion interpretations of 2-state verbs for the nondefault-bound perfective form, and requests/wishes for the imperative verb form. The finiteness opposition becomes operative at a second stage, with the consequence that the [+finite/indicative] marking allows again for the default interpretation (here&now) in contrast to [-finite] marked verbs with the nondefault interpretation (preferably an irrealis interpretation). Now we have a coexistence of two forms for one function: imperative form as well as the infinitival form relate to the expression of requests or wishes. (5) This coexistence pattern is nicely exemplified by two examples taken from Bar-Shalom and Snyder (1999):
(33) Varya from Transcript 2 (Protassova's corpus, Bar-Shalom and
 Snyder 1999: 59)
 Dajti- dat', mama, mama, davaj davaj.
 give-IMP Give-INF Mama, Mama, give-IMP, give-IMP
 Give, give, Mama, Mama, give give.
(34) Varya from Transcript 4 (Protassova's corpus, Bar-Shalom and
 Snyder 1999: 59)
 Na, prichesat', pricheshi.
 here, comb-INF comb-IMP
 Here, comb, comb.

The pressure towards economy and nonambiguity leads to a substantial reorganization of this system and changes the functionality of the infinitival marking. During this reorganization, the nondefault character of the infinitival marking is freed from the narrow interpretational functionality and thus gives way for a more syntactically-oriented function. This change in grammatical status is reflected in the temporal unboundedness of the Russian root infinitive data which occur in a subsequent stage of development.

At that point in development, the combination of aspectual marking and infinitival marking leads to a pattern with [+perf/-fin] expressing past events, [+imperf/-fin] expressing ongoing events, and [+perf/-fin] or [+imperf/-fin] expressing future/irrealis-bound events in addition to their finite indicative and imperative counterparts. Hence, the reversal in the progression line of root infinitives (tense-based languages going from nonmodal to modal and aspect-based languages going from modal to nonmodal) finds an explanation in terms of differences in the developmental dynamics within an identical universal basic functionality, that is, expressing nondefault time/world continua as associated with "distant from here&now" reference.

The entangled development of aspectual and finiteness bound cell divisions actually calls for a three-dimensional representation which is not possible here. Thus, the second fission should be folded on top of the first one, since aspectual information is already encoded in the verb stem. According to an anonymous reviewer, perfective verbs cluster with past tense inflection and imperfective verbs cluster with present tense inflection --this is in line with the idea that a default form like [+imperfective] correlates with a default temporal interpretation as "here&now" which forms the core of grammatical present tense. Moreover, it also correlates with the default value on the finiteness axis, which is [+finite]. The clustering of perfective verbs with past tense inflection follows from the respective reasoning for the nondefault values (with the exclusion of the finiteness value). The combination of two opposing forms (nondefault form [+perfective] plus default form [+finite]) eventually allows for the grammar-oriented reanalysis of form-function relations towards a greater target like interpretational flexibility (or even independence, as in the case of finiteness). An explanation for how the substantial reorganization process above could be also structurally initiated is given in Tracy (2002). In this work, the coexistence of competing structures is reinterpreted as a stepping stone towards the inference of new system properties. This mechanism could likewise be adapted to the emerging form-function relations as encountered here.


4.4. Aspectual-oriented languages without infinitival marking

How do children express early modals, that is, the wish-world concept, if no infinitival marking is available, as is the case in Greek? According to Stephany (1997), it is the form expressing subjunctive mood which relates to a future/modal conglomerate on the functional side. Varlokosta et al. (1996, 1998), however, argue for the perfect participle form as the Greek root infinitive equivalent. Both accounts draw on a shared data source--the children Spiros, Janna, and Mairi, collected by U. Stephany (plus two additional children in the work of Stephany). The aspectual orientation in Greek manifests itself in an obligatory distinction between the perfectire and imperfective aspect in the future tense and in the past tense. The present tense paradigm is exclusively based on the imperfective verb stem.

It could well be that individual differences among the children studied and intraindividual divergencies across development are responsible for the two opposing accounts. According to an anonymous reviewer, however, it is less plausible to analyze the perfect participle form as root infinitive equivalent since this form corresponds to the traditionally called aparemfato, which is practically not present in the input speech of mothers. Moreover, finite forms abound, and it can be shown that young children at the age of 1;11 already have a productive command of verbal inflections, including person and number marking. Therefore, a view following Stephany's (1985, 1997) analysis has to be strongly preferred.

What matters for the present purpose is the possibility of subjunctive (and, perhaps, perfect participle marking) to be related to the future/ modal segment. The subjunctive marking fits easily into the above model, since it embodies the nondefault "different from the here&now" interpretation as reflected in its target-adequate irrealis meaning proper (thus contrasting with the indicative nonpast forms). The interpretation of the implausible but also suggested perfect participle choice is less obvious. But here exists an inherent logic, too, if we take the notion "different from the here&now" seriously. Under the assumption that the directionality of the distance is not yet formally encoded, the perfect participle form fits in smoothly, although it is pointing forward on the temporal axis instead of backwards from a target perspective. What matters for the child is only the difference from the here&now. Due to the homophony of the perfect participle suffix -i with the third person singular suffix -i, it can be argued that an enforced saliency is given compared with participles from other languages that could not as easily get promoted to root infinitive equivalents.

Consider now the diverging progression lines of the -i marked "root infinitives" and the subjunctive marked "root infinitives. Here, the split between subjunctive and indicative forms is prior to the separation of the perfect participle, or better, aparemfato, marking. Thus, the functional segment of modal/future interpretation can, in principle, already be covered by the subjunctive marking when the perfect participle/aparemfato marking emerges if it plays a role at all (see the discussion about the two opposing accounts above). This restricts the possibilities of the perfect participle/aparemfato marking to a smaller functional array than the earlier partitioning, which is reflected in the progression line differences.

The figure below is partly based on the developmental description from Stephany (1997).

Let us sum up. In the absence of empirical data that could support a three-variable system for the encoding of temporal relations in early child language, I assume that children have only two separable time variables available, the time of utterance (TU) and a time span which does not differentiate between topic time and situation time (TT/TSit). The introduction of the third temporal parameter leads to a substantial reorganization of the child's grammatical system, and opens the space for feature combinations by "grammaticalization" of previously purely interpretationally used forms.

The ongoing discussion in the generative literature about whether aspect or tense comes first in child language reveals a theoretically driven preconditioning in the sense that the height in the syntactic tree of a functional node determines which functional category takes over the other's functionality. So, Brun et al. (1999: 125) propose "that the mechanism of tense encoding be realized within the aspectual system" if the target language is aspect dominant. Accordingly, the equivalent formulation for tensed-based languages should claim that it is the finiteness marking which encodes early temporal information. Therefore, a more abstract model has to be favored to capture the first respective form-function correlations in child language. Additionally, the role of modality --for young children restricted to boulemaic modality--has to be taken into account. As long as the interpretation of the tense category future can not be sharply divided from the boulemaic modality of wishes and requests across respective form-function correlations in child speech, an unanalyzed conglomerate category "tense+aspect+modality" represents this specific functional segment best.


4.5. A note on the case of English

If we compare the languages and their developmental properties characterized above with the acquisition of English, it becomes clear why English is a particularly problematic case for the study of the default/ nondefault split: in the absence of a sufficiently salient encoding of the [+/-finite] distinction, the English child must resort to lexical means if he or she wants to convey the irrealis-bound modal interpretation. There exists no morphologically visible infinitival marker. This explains the early emergence of the wanna/want to construction, whereas German children integrate their respective lexical counterparts of boulemaic modality comparably late into their system.

5. Consequences of the model

This view of development as a partitioning into more and more fine-grained chunks of formal and functional information is not new. Apart from the classical work by Roman Jakobson in phonology, similar partitioning accounts have been put forward for various areas of acquisition (for the area of modality, cf. Stephany 1995). What does the cell division account buy us in the case of root infinitives?

Firstly, it allows for smooth transitions from one stage to another, where a critical number of grammatical elements has to be analyzed and categorized before a cell division can take place. This parallels the development of the nominal and the verbal domain in the lexicon. During the transition phases, protoforms can occur which mimic a target-like behavior. They are not properly analyzed but pace the way or the next major split (see Jordens 2000). Secondly, the lesser the number of cells, the higher the contextual dependency of their individual interpretation and, thus, their potential multifunctionality. This effect has been, thirdly, grasped within an indirect modelling of form-function relations, where the setting of topic time gets associated with a default or nondefault form of finiteness marking. This does not a priori exclude other pragmatic interpretations. Thus, the functional segment "commenting ongoing events" is also compatible with the infinitival marking. This is due to the fact that an explicit topic time setting is not necessary in this case, which is expressed by a grammatically not set topic time as also embodied in the infinitive, as argued above. Fourthly, this indirect modelling of the form-function relation allows for a cross-linguistic and interindividually flexible but systematic frame for developmental description, encompassing periods of smooth transition and of substantial reorganization. These periods already comprise target-adequate settings into default and nondefault categories but do not yet show fully target-consistent form-function relations with respect to aspect, tense, or modality.

5.1. Modality

Stephany (1995) interprets the early requestive-indicative split in speech acts as a first indication of emerging nonepistemic (dynamic/deontic) and later epistemic modality, where indicative acts might act as a precursor to the descriptive function of language:

At approximately 1;6, when the inflectional stage begins, verb forms split into modal and non-modal ones, a distinction that develops prior to different tense-aspect forms within the category of non-modal verb forms (...) (Stephany 1995: 110-116).

She concludes that the lexical signalling of modal means does not generally precede their grammaticalized expression in first language acquisition. This is in agreement with my own observations, though we differ in terms of our descriptive categories. German children do not split their verb forms into modal and nonmodal ones, but introduce a formal opposition ([+/-finit]), which conveys an interpretation as nondefault = "different from the here&now" that is compatible with a modal interpretation. English children, on the other hand, have only lexical means at their disposal and resort to the particular verb form of want.

The functionality of referring to a wish world itself is acquired even earlier. Thus, plain object-labelling allows for requesting for an object. On the other hand, the infinitival form itself can not get directly attached to a modal interpretation, if we take into account the other major function of root infinitives, which is commenting on ongoing events. What is encoded in the infinitival marking is exclusively the--already target-adequate--"different from the here&now" interpretation. Thus, the beginning of the inflectional stage opens the developmental space for new form-function correlations without misleading the child into target-inadequate fixations of features.

The boundaries between modal and other interpretations, such as future tense, are often blurred, as becomes clear from the different punctuation marks and labels attached to various child utterances in the transcripts. Blom (1999), for example, subsumes prospective root infinitives with an intentional interpretation under "nonmodal" if they do not explicitly express necessity or possibility. I know of no data which would show that children of about two years can already distinguish between pure intentionality and formal necessity. Therefore, I argue for a combined category intentional/modal/future to capture the range of functionality of wishes and requests. Only late, these structures are separated into target-adequate chunks of grammatical information.

Why should a request or wish conveyed by a root infinitive imply a topic time setting different from the here&now if the wish or the obligation already holds at the time of utterance? At that stage, topic time and situation time cannot be kept apart; the child does not talk about the time right now and about the world as it is right now--in which he or she may have certain wishes or in which there are certain obligations--but about a time and a world in which the situation described by the root infinitive clause obtains; and this time and this world are different from what is the case at the here&now.

5.2. The first fission

The first fission states that under the nondefault interpretation, TT/TSit is different from the here&now. It leaves open whether this time is before or after TU. Thus, the model assumes a stage where children do not bother with a linguistically relevant distinction between pastness and futureness (nor about other features, such as the degree of remoteness from the here&now). This seems to be at odds with the major past-nonpast split which characterizes many of the languages studied in the present context, such as German or Dutch. But in other languages, such as Inuktitut, the major opposition is between future and nonfuture (embedded within a more general realis/irrealis opposition, see Swift 2000). Whereas irrealis/future is obligatorily marked in this language, the realis/nonfuture forms is either zero-marked (as with telic and atelic verbal expressions indicating completion and noncompletion, respectively) or it is marked for remote past. Such cross-linguistic differences lead to interesting consequences for the acquisition process of the temporal system. Thus, Swift (2000) notes:

The early emergence of future time markers in Inuktitut child speech, which is in striking contrast to the findings reported for languages previously studied, can be motivated by the future/nonfuture opposition of temporal encoding, in which future is the marked member of the opposition (Swift 2000: 150).

If we now compare Dutch/German with Russian and with Inuktitut, the common base of the first fission is: reference to the here&now vs. reference to a point in time different from the here&now. This opposition is then either directly (tense-oriented languages) or indirectly (aspectual-oriented languages) associated with the realis/irrealis opposition. Thus, irrealis is expressed from early on by a specific form from a morphological opposition pair, although the target-adequate full set of formal modality distinctions is found among the latest stages in acquisition. The observation, that the wish/request functionality is so frequent--and also so variably expressed (verbless utterances as well as root infinitives)--in early child language might lead to the impression that, from the child's perspective, it is the default. This is not the adult's view; we are used to taking the indicative form/declarative function as the default. Nevertheless, true positive assertion markers, which are associated with the [+finite] marking (cf. Klein 1998), occur later in the acquisition process than the [-finite] marking (see also Dimroth et al., forthcoming)--a hint that points again towards a prior irrealis/wish world functionality. Despite all this, it is uniformly the "(proto)-indicative"/ "(proto)-declarative" category which bears the default in the linguistic system of the child, too. This shows a strong preference, if not a constraint, on child language in the sense that this sort of default characterization of form-function relations is either imprinted in the system from early on or potentially deduced from other factors. Among these factors, the temporal here&now-centricity might constitute the most prominent one. In addition, the default issue shows that frequency accounts for determining default categories have to be supplemented by further factors in order to be reliable as a linguistic relevant characterization.

5.3. The subject issue

As reported in Kramer (1993), the occurrence of dropped subjects is significantly correlated with root infinitives in Dutch and German. She also claims that "(i)n the low number of instances in which subjects do occur with infinitives, their occurrence can be explained by assuming that the children project a phonetically null modal in Infl or Tense" (Kramer 1993: 206). And she concludes therefore that "the Case Filter is present from the earliest stage of language acquisition" (Kramer 1993: 206). The fact that root infinitives often function as "comment on ongoing events" rules out a pure null-modal explanation of the subject-drop. What makes the modal interpretation and the commenting interpretation particular prone to the dropping of subjects is their specific context and context reliance, as reflected in the grammaticalized inherent addressee in imperatives (taking the wish to obligation) and in the here&now focussing of comments on ongoing events. Here, the subject is a superfluous syntactic topic element, and its omission is an instance of topic-drop (also common in these contexts in adult language). Therefore, I argue for a dependency of the topic-drop frequency of root infinitives on their usage: the higher the percentage of modals and comments on ongoing events, the higher the percentage of topic-drop. Note that the category of answers in root infinitive form has to be excluded from any topic-drop analysis in general, if the turn-adjacent question gives a possible mould for grammatical ellipsis.

The reverse progression of the usage in Dutch/German and Russian root infinitives (the former from nonmodal to modal, the latter from modal to nonmodal), noted in Section 2, predicts a rise of the percentage of topic-drop for the Dutch/German case and a fall of topic-drop for Russian, if we set aside other intervening factors as growing utterance length, overall decrease of the request/wish function, detachment from egocentricity, etc. This remains to be tested cross-linguistically. Nevertheless, the proportions of RI-structures to other structures in German vs. Russian acquisition indicates that it is mainly in tense-oriented languages where RIs play a major role in development judged by their magnitude of occurrence. It is noteworthy that Russian/Greek and German differ sharply concerning their grammatical status of topic-drop in general. One of the major advantages of this pragmatic-functional view on topic-drop in root infinitives is that the interindividually and cross-linguistically varying amount of overt subjects/topics can be accounted for by alluding to differences in usage. It is claimed here that these usage-based differences may eventually lead to different settings of the so-called topic-drop parameter across languages, and not the other way around.

For rigid syntactic approaches to the RI phenomenon as, for example, the case filter account, cases without topic-drop (or here better subject-drop) pose a problem. Compare the following two German examples from Lasser (1997: 203):
(35) Simone 2;0
 Mone auch'n Loffel habe(n)
 Mone also-a spoon have-INF
(36) Simone 2;11
 Ich erstma das Buch angucken
 I first the book look-at-INF

In these examples Mone and ich ('I') represent the nominative subject of the root infinitive structure. If nominative case depends on the existence of a finite marked element of a functional node like, for example, INFL as has been proposed within the generative framework, the above structures should not pass the case filter. The assumed reason for this is that the nominative case cannot be checked for.

Since English bare verbs are not associated with the default/nondefault split typical for the finiteness distinction, the difference in preference for overt subjects in English and Dutch/German (cf. Phillips 1996) falls out naturally. In the same vein, I argue for the absence of wh-elements in root infinitives on interpretational grounds. The difference in correlation between Dutch/German and English in this respect also shows the different status of English bare verbs in not being marked for nondefault interpretation. If the topic-drop analysis is correct, it also explains the absence of finite subject DPs, which subsume plural DPs or definites, as discussed in Hoekstra and Hyams (1998), since backgrounded topical elements are typically nonplural and pronominalized.

5.4. The early "tensepect" relation

The prevailing target-oriented perspective in the analysis of child data has seduced many researchers to operate with the notions of aspect and tense as if used in the adult language. Faced with an incomplete proposition that possibly describes a single event, a temporal analysis based on aspectual and tense terms can hardly be justified as long as there is no separation of topic time and situation time. This fact has already been emphasized in Smith (1980), there framed in Reichenbach's terminology:

It would be of utmost interest if a shift in grammars were demonstrated--one that constituted reorganisation rather than a more gentle change. Such a shift would show that child grammars are not necessarily part of adult grammars. If this is the case, the structure of the former does not allow inference about the structure of the latter. (...). I suggest that children's early temporal ordering system differs from the adult system in two ways: only two times are involved, and orientation is fixed at ST (= speech time, PG) (Smith 1980: 264-265).

If a child makes use of a particular morphological opposition to express interpretational differences, this does not imply that this form-function correlation is necessarily a target-adequate pairing. Although the core form-function relations are generally pointing towards a target-adequate direction, the cell size is hardly fine-grained enough to model the categories of the target language properly. The model sketched above relates forms to functions via their default or nondefault interpretation of topic time reference instead of attaching interpretative features in a direct manner to a specific form. This indirect mechanism of form-function relation allows for a target-adequate feature setting (topic time set to default or nondefault) combined with developing interpretational associations. These associations may differ from the target associations but still be comprehensible for adult native speakers. This is an important point in the framework proposed here: no resetting of features is necessary. What has to be learned is the complex calculation of individual parameters within a multidimensional temporal matrix on top of the more and more fine-grained cell differentiation. The adult native interlocutor is able to strip away the involved complexity, thus concentrating on the basic topic time setting in order to understand and communicate with a smaller child.

Central to this account is the question as to how children discover the third temporal parameter, that is, the separation of topic time and situation time. The above scenarios for aspectual-oriented languages suggest an answer to this question in terms of inhomogeneous form-function relations: if two diverse forms end up being associated with the same segment in interpretation, the child will seek for a solution to this puzzle. The solution embodies the unfolding of formerly compressed dimensions, as demonstrated in the unfolding of the "tensepect relation" to the tense relation and the aspect relation, as soon as one of the parameters allows for the full combinatorics of the system. Moreover, this model allows us to capture the diversity of forms encountered cross-linguistically in early child language within a more or less similar space of functions to be covered.

5.5. The role of finiteness

Finiteness has been shown to be crucial to temporal interpretation in child language. It should be emphasized that finiteness is not just morphological inflection. It is a more abstract grammatical category, whose presence or absence has numerous consequences in morphology, syntax (in particular word order), and semantics. It is closely connected to some other categories; but it cannot be reduced to, for example tense marking (compare the case of Ancient Greek), person marking (compare the Pacific language family), the category verb in general (compare the language Paez (6)), or morphological finiteness (Chinese). For an overview on the variety of finiteness expressions across languages, see Maas (1997).

It is evident from the data above that the morphological form of the aspectual distinction is not able to convey a default vs. nondefault interpretation of topic time, as the finiteness distinction does. Although the form of imperfective and perfective marked verbs develops prior to the finiteness split in aspectual-oriented languages, it does not associate with the basic functional modal vs. nonmodal opposition. Hence, the finiteness distinction is the decisive one to enable temporal reference different from the default here&now. For tense-oriented languages, this prominent split occurs first with the effect that the point of substantial reorganization of the system of morphologically-encoded temporal reference comes late in development. The necessity to express wishes and requests is already satisfied by the interpretational range of the infinitival morphology. In contrast to this, aspectual-oriented languages lack this interpretational segment as long as only the form opposition of aspectual markings is available. This fact forces the early integration of the imperative form into the system, which covers the functional pressure for differentiation but misses a formal opposing category "in-imperative." It is only the emergence of the finiteness split that opens up a SYSTEMATIC opposition in terms of default vs. nondefault temporal reference. In the case of Greek, this is achieved by the opposition between the indicative and the subjunctive form, which relates to a similar systematic opposition into a default and nondefault verb form. This predicts a developmental advantage for those Greek children who opt for the subjunctive form to convey "root infinitival" interpretation instead of the perfect participle form, the aparemfato.

5.6. The notion of default

The notion of default used here relates to the temporal reference (and, therefore, world reference) of interpretation. In the default case, time and world talked about are given by the here&now. In the literature on root infinitives, many other notions of default are used, for example in terms of morphophonological material--the more material is added, the lesser the default characterization. Another default relates to the amount of structure--as measured by the number of nodes involved, the specification of heads, and associations between nodes and their interpretation. A further default characterization is based on functional grounds. Here, differences between the child and adult perspective come into play. Thus, nonepistemic interpretations are prior to epistemic ones in development, hence favoring boulemaic and deontic functions (Stephany 1995), whereas the adult perspective takes the opposite view and assumes (for structural and interpretational reasons) the epistemic/declarative function as basic. And last but not least, the notion of default might be based on frequency considerations such that the overuse of a form leads to the stipulation of its being basic--see the defining features of root infinitives in Vadokosta et al. (1996).

The default notion used here is backed up by cross-linguistic evidence. Cognitively and perceptually, the here&now is obviously the first thing at hand. But there are also many linguistic data which point to the default character of the here&now reference. Consider, for example, zero-tense marking in languages like African-American English, where the grammatical system does not include an overt morphological marking for reference to the here&now. Dechaine (1995: 68) even claims that not only African-American English but also Standard American English lacks "NOW" as a semantic tense operator, since it is not necessary as the default. But whereas African AE has chosen the morphological zero tense option (abbreviated TO), Standard AE entertains a morphological form which correlates with subject agreement. Dechaine (1995) says:

There are four possibilities: [TO] could be temporally vague, [TO] could be consistent with a situation which follows the utterance situation ([s.sub.u] < [TO]), precedes it ([TO] < [s.sub.u]), or overlaps with it ([TO] [??] [s.sub.u]). I consider these four possibilities, and conclude that only one is tenable: [TO] overlaps with the utterance situation (Dechaine 1995: 71).

This is precisely as in early child language. The here&now focus in early child language might be altered and expanded; but it survives into the adult system as a default setting of temporal reference. This demonstrates that we never get completely rid of that egocentricity of the early years. Whether it is the form of the morphological present tense which expresses this default, or whether another form is taken to be basic for a semantic tense logics does not matter here (for a formal semantic approach of German tenses which assumes the infinitive to be the base case, see Bauerle and von Stechow 1980).

6. Conclusion

A cross-linguistic comparison rapidly showed that root infinitives are neither formally nor functionally a homogeneous phenomenon. A fresh look at German data in particular demonstrated the need to include more subtle shifts and tendencies into the overall picture of emerging form-function relations. Strict feature associations, for example, claiming that the infinitival morphology is fixedly correlated with a modal interpretation, do not do justice to the empirical findings. Thus, German root infinitives are initially used alike for comments on ongoing events and requests/wishes (with some interspersed descriptions). Later, they narrow down towards the category of requests/wishes, on the one hand, and as target-adequate adjacency ellipsis in question-answer pairs, on the other. In these latter cases, their infinitival status is arguable. A similar picture has been described for the development of Dutch. In contrast to these tense-oriented languages, aspectual-oriented languages exhibit a different behavior. In Russian, root infinitives start out as modality marker indicating requests/wishes. Later, they additionally assume functions such as comments on ongoing events and narratives with reference to past events. In Greek, which is aspectual-oriented as well, a formal category infinitive is not available in the first place. Nevertheless, it has been claimed that root infinitives also exist in Greek child language. This claim is based on two substitute forms with primarily modal function, namely, the subjunctive and possibly also the perfect participle form.

These facts call for a more abstract model. The idea advanced here is the model of a successive "cell partitioning." Its first fission is between a default interpretation (here&now) and a nondefault interpretation (different from the here&now), which is associated with the most salient morphological opposition available in the target language. In a later step, temporal directionality is introduced, allowing the child to mark forward and backward reference with respect to the time parameter talked about. Up to that point, only two temporal parameters are involved: the time of utterance, which is set to the here&now, and a yet unanalyzed conglomerate of time talked about, the "topic time," and the time at which the situation obtains, the "time of situation." The separation of topic time and situation time and, thus, the availability of three temporal parameters leads to a major reorganization of the temporal system. It is reflected in a partly radical shift in form-function correlations. It is only after the separation of topic time and situation time that the target-specific categories of tense and aspect are applicable to child language. This is mirrored in the data, when aspectual morphology and tense-bound morphology are productively combined. Several further characteristics, which have been discussed in relation to root infinitives--for example, the role of (c)overt subjects or default settings--naturally concur with the assumptions of this model.
Table 1. Cross-linguistic form-function correlations

 Form Interpretation Developmental

Dutch -en (INF marking) modal (esp. wishes) from nonmodal to
 verb placement and nonmodal modal
 preferred temporal
 reference: future

German -en (INF marking) modal (esp. wishes) from nonmodal to
 verb placement and nonmodal modal
 (commenting on
 ongoing events)

Greek subjunctive form imperative force for subjunctive
 (Stephany) by a priori marking: from
 definition future/modal to
 [??]a for future
 and na for modal

 -i as participle for -i marking:
 marking, modal only
 (Varlokosta et al.)

Russian -t' (INF marking) open temporal from modal to
 reference nonmodal

English bare stem no special modal vs. non-
 interpretation modal decoupled
 from morphology

Table 2. Combinatorics of the tense/aspect system

Tense/Asp. system Imperfective aspect Perfective aspect

Past tense [+past/+imperf] [+past/+perf]
Present tense [+present/+imperf] [+present/+perf]
Future tense [+future/+imperf] [+future/+perf]

Tense/Asp. system Prospective aspect

Past tense ([+past/+prosp])
Present tense ([+present/+prosp])
Future tense ([+future/+prosp])

Table 3. Form-function correlations after the first fission within a
tense-oriented system

Structure No marking Infinitival marking
 (verbless structures)

Feature no marking [[??] marked as [-finite]
 finite] plain DP, biased to thematic
 structures with verbal 2-state verbs
 particle only, etc.

Associated no grammatical TT TT grammatically
topic time involved set to nondefault
setting (different from
 here&now in CL)

Functional no functional (boulemaic) modals
category specialization (TT at a wish
associated world/time)

 comments on
 ongoing events
 (TT spec. obso-
 lete due to an
 shared focus
 of attention)

Effects on complete context restricted context
context dependency dependency in the
dep. case of comments

Structure Finite marking

Feature marked as [+finite]
 mainly copula

Associated TT grammatically
topic time set to default
setting (here&now in CL)

Functional existensials/
category descriptions

Effects on grammatically
context "context independent"
dep. but pragmatically

Figure 1. Dvelopment of structure types in Max

 verbless infinitival finite
 structures structures structures

Max I (2;6.29) 111 67 39
Max II (2;9.0) 24 29 8
Max III (2;11.18) 12 43 113

Note: Table made from bar graph.

Figure 2. Dvelopment of structure types in Benny

 verbless infinitival finite
 structures structures structures

Benny I (2;2.15) 81 55 26
Benny II (2;10.28) 6 33 23
Benny III (3;0.18) 8 64 104

Note: Table made from bar graph.


* Correspondence address: Max Planck Institute for Psycholinguistics, P.O. Box 310, 6500 AH Nijmegen, The Netherlands. E-mail:

(1.) The data labelled "Tubingen corpus" stem from the DFG-project "Erwerb der komplexen Syntax" (Acquisition of complex syntax) at the University of Tubingen. I am very grateful to R. Tracy, who gave me permission to publish them. I should note that I was involved in the collection, transcription, and further steps of data processing of the project. For further information about the corpus and related work see Fritzenschaft et al. (1990); Traey (1991); d'Avis and Gretsch (1994); Gretsch (2000); Tracy (2002).

(2.) At first sight umfal could be interpreted as a verb particle (um-) plus verb stem (fall-), but the usage of umfal in other utterances indicates already a nominal status as in der macht auch nen umfal.

(3.) In the words of Wijnen (1997):

Importantly, the temporal interpretation of these non-eventive root is restricted to future. Perhaps this observation, and the predominance of future interpretations in eventive RIs can be explained by assuming that non-tense marked verbs in general have a default interpretation corresponding to 'irrealis'. This might also explain the legitimacy of infinitives in, e.g., counterfactual exclamations in the adult language (Wijnen 1997: 17-18).

(4.) The eventivity constraint claims that "RIs (= root infinitives, PG) are restricted to event-denoting predicates" (Hoekstra and Hyams 1999: 241). It should be emphasized that around 2-10% exceptions occur to this strong correlation, which shows that the restriction above should not imply a fatal clash of semantic features; compare the following example which is typical for pretend-play talk:
Max04 2;09.00
playing with lego; S. holds a piece in her hand
n des auch n dach sein? that-one also a (?) roof be-INF
That one should be/is a roof, too.

(5.) Regarding the differences between imperatives and infinitivals in request function, an anonymous reviewer pointed out the fact that whether the child is the agent of the desirable action or whether he or she only asks the interlocutor to perform the action determines the verb form chosen in Russian child language.

(6.) For an overview on the peculiarities of the language Paez, see Jung (1984), where it is shown that the person-marking morpheme combines unexpectedly with the first accentuated syllable in this language.


Abraham, Werner (1989). Futur-Typologie in den germanischen Sprachen. In Tempus-Aspekt-Modus. Die lexikalischen und grammatischen Formen in den germanischen Sprachen, Werner Abraham and Theo Janssen (eds.), 345-389. Tubingen: Niemeyer.

Bar-Shalom, Eva; and Snyder, William (1999). On the relationship between root infinitives and imperatives in early child Russian. In Proceedings of the 23rd Annual Boston University Conference on Language Development, Annabel Greenhill, Heather Littlefield, and Cheryl Tano (eds.), 56-67. Somerville, MA: Cascadilla.

Bauerle, Rainer; and von Stechow, Arnim (1980). Finite and non-finite temporal constructions in German. In Time, Tense, and Quantifiers. Proceedings of the Stuttgart Conference on the Logic of Tense and Quantification, Christian Rohrer (ed.), 375-421. Tubingen: Niemeyer.

Behrens, Heike (1993). Temporal Reference in German Child Language: Form and Function of Early Verb Use. Zutphen: Koninglijke Wohrmann b.v.

Blom, Elma (1999). On the use and interpretation of root infinitives in early child language. Unpublished manuscript, Utrecht University.

-- (2002). On the use and interpretation of root infinitives in early child Dutch. In Modality and its Interaction with the Verbal System, Sjef Barbiers, Frits Beukema, and Wim van der Wurff (eds.), 103-132. Amsterdam: John Benjamins.

Bloom, Lois; Lifter, Karin; and Hafitz, Jeremie (1980). Semantics of verbs and the development of verb inflection in child language. Language 56, 386-412.

Bowerman, Melissa (1985). What shapes children's grammars? In The Cross-linguistic Study of Language Acquisition, vol. 2, Dan I. Slobin (ed.), 1257-1319. Hillsdale, NJ: Lawrence Erlbaum.

Brun, Dina; Avrutin, Sergey; and Babyonyshev, Maria (1999). Aspect and its temporal interpretation during the optional infinitive stage in Russian. In Proceedings of the 23rd Annual Boston University Conference on Language Development, Annabel Greenhill, Heather Littlefield, and Cheryl Tano (eds.), 120-131. Somerville, MA: Cascadilla.

d'Avis, Franz-Josef; and Gretsch, Petra (1994). Variations on "variation": on the acquisition of complementizers in German. In How Tolerant is Universal Grammar? Rosemarie Tracy and Elsa Lattey (eds.), 59-109. Tubingen: Niemeyer.

Dechaine, Rose-Marie (1995). Zero tense in Standard and African American English. In Proceedings of NELS 25, Jill N. Beckman (ed.), 63-77. Graduate Linguistic Students' Assoc., Umass.

de Villiers, Jill G.; and de Villiers, Peter A. (1973). A cross-sectional study of the acquisition of grammatical morphemes. Journal of Psycholinguistic Research 2, 267-278.

Dimroth, Christine; Gretsch, Petra; Perdue, Clive; Jordens, Peter; and Starren, Marianne (forthcoming). Finiteness in Germanic languages: a stage model for first and second language development. In Information Structure, Linguistic Structure, and the Dynamics of Language Acquisition, Christine Dimroth and Marianne Starren (eds.). Amsterdam: John Benjamins.

Fantuzzi, Cheryl (1996). The acquisition of tense and temporal reference. In Proceedings of the 20th Annual Boston University Conference on Language Development, Andy Stringfellow, Dalia Dahana-Amitay, Elizabeth Hughes, and Andrea Zukowski (eds.), 201-212. Somerville, MA: Cascadilla.

Fritzenschaft, Agnes; Gawlitzek-Maiwald, Ira; Tracy, Rosemarie; and Winkler, Susanne (1990). Wege zur komplexen Syntax. Zeitschrift fur Sprachwissenschaft 9, 52-134.

Giorgi, Alassandra; and Pianesi, Fabio (1997). Tense and Aspect: From Semantics to Morphosyntax. New York: Oxford University Press.

Gretsch, Petra (2000). Fokale Ellipsen in Erwachsenen- und Kindersprache. Tubingen: Niemeyer.

Guillaume, Gustave (1929). Temps Et Verbe. Theorie des aspects, des modes et des temps. La societe de linguistique de Paris, XXVII. Paris: Libraire ancienne honore Champion.

Haegeman, Liliane M. V. (1995). Root infinitives, tense and truncated structures. Language Acquisition 4, 205-255.

Hamann, Cornelia (1996). Null arguments in German child language. Language Acquisition 5, 155-208.

Hoekstra, Teun; and Hyams, Nina (1998). Agreement and the finiteness of V2: evidence from child language. In Proceedings of the 22nd Annual Boston University Conference on Language Development, Annabel Greenhill, Mary Hughes, Heather Littlefield, and Hugh Walsh (eds.), 360-373. Somerville, MA: Cascadilla.

--; and Hyams, Nina (1999). The eventivity constraint and modal reference effects in root infinitives. In Proceedings of the 23rd Annual Boston University Conference on Language Development, Annabel Greenhill, Heather Littlefield, and Cheryl Tano (eds.), 240-252. Somerville, MA: Cascadilla.

Hyams, Nina (1996). The underspecification of functional categories in early grammar. In Generative Perspectives on Language Acquisition, Harald Clahsen (ed.), 91-128. Amsterdam: John Benjamins.

Ingram, David; and Thompson, William (1996). Early syntactic acquisition in German: evidence for the modal hypothesis. Language 72, 97-120.

Jordens, Peter (1990). The acquisition of verb placement in Dutch and German. Linguistics 28, 1407-1448.

-- (2000). Particles in child Dutch. Unpublished manuscript, Vrije Universiteit Amsterdam.

-- (2002). Finiteness in early child Dutch. Linguistics 40(4), 687-765.

Josefsson, Gunlog (1999). Non-finite root clauses in Swedish child language. Working Papers in Scandinavian Syntax 63, 105-150. University of Lund.

Jung, Ingrid (1984). Grammatik des Paez. Ein Abriss. Unpublished doctoral dissertation, University of Osnabruck.

Klein, Wolfgang (1994). Time in Language. London: Routledge.

-- (1998). Assertion and Finiteness. In Issues in the Theory of Language Acquisition. Essays in Honor of Jurgen Weissenborn, Norbert Dittmar and Zvi Penner (eds.), 225-245. Bern: Lang.

-- (forthcoming). On finiteness. In Semantics Meets Acquisition, Veerle Van Geenhoven (ed.). Dordrecht: Kluwer.

Kramer, Irene (1993). The licensing of subjects in early child language. In Papers on Case & Agreement II, Colin Phillips (ed.), 197-212. MIT Working Papers in Linguistics 19. Cambridge, MA: MIT.

Lasser, Ingeborg (1997). Finiteness in Adult and Child German. MPI Series in Psycholinguistics 8. Nijmegen: Max-Planck-Institute for Psycholinguistics.

-- (2002). The roots of root infinitives. Linguistics 40(4), 767-796.

Maas, Utz. (ed.) (1997). Infinitive. Sprachtypologische Studien. Osnabruck: Secolo Verlag.

Penner, Zvi; Tracy, Rosemarie; and Wymann, Karin (1999). Die Rolle der Fokuspartikel AUCH im fruben kindlichen Lexikon. In Das Lexikon im Spracherwerb, Jorg Meibauer and Monika Rothweiler (eds.), 229-251. Tubingen: Francke Verlag.

Phillips, Colin (1996). Root infinitives are finite. In Proceedings of the 20th Annual Boston University Conference on Language Development, Andy Stringfellow, Dalia Dahana-Amitay, Elizabeth Hughes, and Andrea Zukowski (eds.), 588-599. Somerville, MA: Cascadilla.

Rizzi, Luigi (1994). Some notes on the linguistic theory and language development: the case of root infinitives. Language Acquisition 3, 371-393.

Schutze, Carson T.; and Wexler, Kenneth (1996). Subject case licensing and English root infinitives. In Proceedings of the 20th Annual Boston University Conference on Language Development, Andy Stringfellow, Dalia Dahana-Amitay, Elizabeth Hughes, and Andrea Zukowski (eds.), 670-681. Somerville, MA: Cascadilla.

Sigurjonsdottir, Sigridur (1999). Root infinitives and null subjects in early Icelandic. In Proceedings of the 23rd Annual Boston University Conference on Language Development, Annabel Greenhill, Heather Littlefield, and Cheryl Tano (eds.), 630-641. Somerville, MA: Cascadilla.

Smith, Carlota S. (1980). The acquisition of time talk: relations between child and adult grammars. Journal of Child Language 7, 263-278.

Stephany, Ursula (1985). Aspekt, Tempus und Modalitat: Eine Studie zur Entwicklung der Verbalgrammatik in der neugriechischen Kindersprache. Tubingen: Narr.

-- (1995). Function and form of modality in first and second language acquisition. In From Pragmatics to Syntax. Modality in Second Language Acquisition, Anna Giacalone Ramat and Grazia Crocco Galeas (eds.), 105-120. Tubingen: Narr.

-- (1997). The acquisition of Greek. In The Cross-linguistic Study of Language Acquisition, vol. 4, Dan I. Slobin (ed.), 183-333. Mahwah, NJ: Lawrence Erlbaum.

Swift, Mary (2000). The development of temporal reference in Inuktitut child language. Unpublished doctoral dissertation, University of Texas, Austin.

Tracy, Rosemarie (1991). Sprachliche Strukturentwicklung: Linguistische and kognitions-psychologische Aspekte einer Theorie des Erstspracherwerbs. Tubingen: Narr.

-- (2002). Growing (clausal) roots: all children start out (and may remain) multilingual. Linguistics 40(4), 653-686.

Ud Deen, Kamil (1997). The interpretation of root infinitives in English: is eventivity a factor? Unpublished manuscript, UCLA.

Varlokosta, Spyridoula; Vainikka, Anne; and Rohrbacher, Bernhard (1996). Root infinitives without infinitives. In Proceedings of the 20th Annual Boston University Conference on Language Development, Andy Stringfellow, Dalia Dahana-Amitay, Elizabeth Hughes, and Andrea Zukowski (eds.), 816-827. Somerville, MA: Cascadilla.

--; Vainikka, Anne; and Rohrbacher, Bernhard (1998). Functional projections, markedness and "root infinitives" in early child Greek. Linguistic Review 15, 187-207.

Wexler, Kenneth (1994). Finiteness and head movement in early child grammars. In Verb Movement, David Lightfoot and Norbert Hornstein (eds.), 305-350. Cambridge: Cambridge University Press.

Wijnen, Frank (1997). Temporal reference and eventivity in root infinitivals. In The Interpretation of Root Infinitives and Bare Nouns in Child Language, Jeannette C. Schaeffer (ed.), 1-25. MIT Working Papers in Linguistics 12, Cambridge, MA: MIT.

--; and Bol, Gerard (1993). The escape from the optional infinitive stage. In Language and Cognition 3. Yearbook of the Research Group for Theoretical and Experimental Linguistics of the University of Groningen, Ale de Boer, Jelly de Jong, and Rita Landeweerd (eds.). Groningen: Groningen University Press.

Received 23 February 2001

Revised version received

6 February 2003

Max Planck Institute for Psycholinguistics, Nijmegen
COPYRIGHT 2004 Walter de Gruyter GmbH & Co. KG
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2004 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Gretsch, Petra
Publication:Linguistics: an interdisciplinary journal of the language sciences
Date:Mar 1, 2004
Previous Article:Aristotle's kinesis/energeia-test and the semantics of the Greek perfect (1).
Next Article:Susanne Wurmbrand: Infinitives. Restructuring and Clause Structure.

Related Articles
Preface finite options: how L1 and L2 learners cope with the acquisition of finiteness (1).
Finiteness in early child Dutch *.
The roots of root infinitives: remarks on infinitival main clauses in adult and child language *.
Finiteness and children with specific language impairment: an exploratory study *.
The acquisition of past tense in preschool children with specific language impairment and unaffected controls: regular and irregular forms *.
Aktionsart, aspect, and the aquisition of finiteness in early child grammar *.
Signs of a developing grammar: subject drop and inflection in early child Dutch *.
"Finite" and "nonfinite" from a typological perspective *.
The English gerund-participle; a comparison with the infinitive.
Finiteness, subjunctives, and negation in English.

Terms of use | Privacy policy | Copyright © 2018 Farlex, Inc. | Feedback | For webmasters