In German, finiteness is instantiated by the verb overtly through inflectional morphology and its appearance in second position (V2) in complementizer-less clauses. In contrast, non-finite verbs, including verbal particles, infinitives and participles, invariably occur in clause-final, also known as verb-final (VF) position. This is exemplified in (1) by the finite verb form macht (bolded) in V2 and the particle auf in VF.

The requirement on finite verb placement is imposed by the so-called verb-second rule, which is widely attested in Germanic languages. The standard generative analysis of V2 in German postulates head movement of the verb from its base position within VP to the head of CP (cf. Thiersch 1978; den Besten 1983). Convincing evidence for the V-to-C movement of finite verbs is provided by den Besten (1983), who observed that in German, complementizers and V2 are in complementary distribution. As shown in (2), the V2 movement, which must apply in main clauses (cf. (1)), is blocked in the presence of a lexical complementizer (2a–b). In certain embedded clauses where the complementizer can be omitted, the relevant clause must appear in the V2 word order (2c–d).1

The clear complementary distribution of lexical complementizers and V2 is adduced as evidence for the argument that finite verbs in second position occupy the same position as lexical complementizers, namely the head of CP. Thus, in addition to the requisite morphology (morphological finiteness), children acquiring finiteness in German must also acquire a syntactic structure with the appropriate functional layers for V2. It is the acquisition of functional finiteness, i.e. the requirement that the finite verb appears in the head of a functional projection (FP), that this paper will focus on.

Several empirical studies have investigated when and how the transition from a lexically organized grammar to the (adult) functionally organized grammar takes place in the learner system, and have proposed a set of lexical and syntactic properties that can be viewed as manifestations of functional finiteness in the learner grammar (Dimroth et al. 2003; Jordens 2012; Jordens & Bittner 2022, among others).

The first property concerns children’s increased use of finite action verbs in V2 position (Jordens 2012). As has repeatedly been pointed out in the L1 acquisition literature on German, in the early stages of finiteness acquisition, child utterances exhibit a correlation between the syntactic distribution of finite and non-finite verb forms and the semantic properties of the verb (Clahsen 1986; Jordens 1990; Meisel 1994; Ingram & Thompson 1996; Jordens 2012). Meisel (1994) and Jordens (1990; 2012) report that finite verbs that appear in non-final position in child utterances generally denote states (e.g. lassen ‘let’, heißen ‘be called’, passen ‘fit’) or changes of state (e.g. werden ‘become’, gehen ‘go’, fahren ‘drive’), as shown in (3a–b). Non-finite verbs that appear in clause-final position, by contrast, frequently denote actions (e.g. putzen ‘clean’, spielen ‘play’, werfen ‘throw’), as illustrated by rausräumen ‘take out’ in (3c). In other words, verbs that appear in non-final and final positions are in complementary distribution and do not overlap.

Based on these data, Jordens (2012) proposes that in early child language, the distributional properties of verb forms with finite and non-finite morphology are determined by semantic factors.

Once a functional projection is available in the system, however, the functional head reserved for finite verbs is no longer associated with specific semantic interpretations and is now accessible to verbs of all semantic types (Jordens 2012). This has the consequence that utterances produced with an underlying functional clause structure will more frequently show action verbs in their finite forms in V2 position as compared to earlier stages.

The second piece of evidence for the existence of a functional projection in the child grammar comes from children’s productive use of auxiliary verbs such as haben ‘have’, sein ‘be’ and werden ‘become’ in German passive or perfect constructions (Dimroth et al. 2003; Jordens 2012). As semantically vacuous verbs, auxiliaries suggest to children that finite verbs in the second position need not be filled by lexical items with descriptive content. Children will hence realize that verb movement to V2 is not motivated by the need to deliver propositional contents, but to host finiteness inflection to express more abstract grammatical information such as tense, agreement and the illocutionary force of the sentence. These abstract grammatical features associated with this position will contribute to the projection of a functional head in the child grammar.

The third indicator is the greater variety of word orders (Dimroth et al. 2003; Jordens 2012). Crucially, the frequent occurrence of utterances with OVS word order indicates that children now distinguish between the subject and the topic and have different positions for them. Since the topic and the subject each require a specifier position, non-subject-initial word orders provide evidence that the learner grammar at this stage consists of two projections, a lexical VP and a functional FP, which accommodate the subject in SpecVP and the topic in SpecFP, respectively.

The final property that can be viewed as a manifestation of functional finiteness in the learner grammar, suggested by Bittner & Bartz (2018), is the target-conforming use of action particle verbs in child utterances, i.e. with the verbal particle in clause-final position and the finite simple verb in V2 position, as shown in (1) above. The discontinuous placement of the particle and base verb indicates that the learner grammar has two distinct positions for verbal elements, which further hints at the existence of a functional layer above the original lexical layer.

These four key properties summarized in (4), which previous studies have proposed as indicators of functional finiteness, will be adopted in the present study as criteria for determining the acquisition of functional finiteness in the child data.

The word doch in German can perform distinct syntactic functions, occurring as a coordinating connective (5a), a response particle (5b), an adverb (5c), or a discourse particle (6). Small caps are used to indicate stress.

The use of doch as a discourse particle can be differentiated from other uses of the word by a number of defining characteristics (Thurmair 1989; Abraham 1991; Meibauer 1994; Karagjosova 2004). Phonologically, it is unstressed, unlike when it functions as a response particle or a sentence adverb, as shown in (5b) and (5c).3 Syntactically, it consistently appears in the clause-internal position (unlike the connective in (5a)) and cannot stand alone as a response to questions (in contrast to its usage as a response particle (5b)). Semantically, it does not influence the truth value of its host sentence (in contrast to the response particle use) and obligatorily scopes over sentential negation (cf. (6b) and (6c)).4

As a discourse particle, doch can appear in all non-interrogative clause types in German including declaratives (6b), exclamatives (7a) and imperatives (7b).5

The semantic contribution of doch as a discourse particle is twofold: for one, it establishes a coherence relation between the doch-utterance p and the pre-context by anaphorically referring to a proposition q that is salient in the discourse; for the other, it introduces new information into the context by expressing a contrast between p and q (Karagjosova 2001; 2004; Grosz 2010; Egg & Zimmermann 2012).

Spelling out the inherent anaphoric and contrastive nature of doch in greater detail, we observe that a crucial prerequisite for a felicitous use of doch is that there must be a contextually salient proposition q evoked in the discourse (Egg & Zimmermann 2012; Grosz 2016), which can be (i) explicitly expressed by discourse participants (8a), or (ii) implicitly presupposed and inferred from the relevant utterances (8b), or (iii) provided by the non-linguistic context (8c). As a result, doch is banned in out-of-the-blue expressions without any contextual embedding.

The second aspect of the semantic denotation of doch is that its complement proposition p in the utterance doch p expresses a contrastive and conflicting meaning to q, which in this sense makes doch resemble the adversative connective aber in German and but in English. The proposition p must be a proposition that is regarded by the speaker as uncontroversial knowledge. This triggers what is known as the uncontroversiality presupposition in Grosz’s (2010) terminology, which crucially excludes doch in contexts where p is completely new information. This point can be clearly illustrated in B’s response in (9) saying that the news about Max’s illness is not yet shared with the hearer. Note that the status of given/new information is consistently judged from the perspective of the speaker, be it real with respect to the knowledge state of the hearer or inferred by the speaker from the actions of his conversational partners.

Furthermore, p and the contextually salient q must be in a conflicting relation, with p being a correction of q (the correction presupposition, in Grosz’s (2010) terminology). The ps and qs presupposed by the doch-utterances in (8a–c) are given in (10a–c), respectively. It is evident that in all these cases, the speaker’s proposition p is opposite to his assumption about the belief of his partner q. Again, the discourse familiarity status of the information is speaker-oriented.

According to Grosz (2010), in contexts where no correction is intended, the occurrence of doch cannot be licensed (11).

Compared with particles previously examined for their influence on finiteness acquisition, such as nicht ‘no’, auch ‘also’, and aber ‘but’, doch is most similar to aber. Both particles necessitate contextual background information prior to making the utterance, and additionally introduce new, contrasting information to an existing salient proposition. This complexity requires that children not only identify the proposition referred to by the particle’s host utterance but also understand the informational status of the aber/doch-proposition within the context and coherently integrate this contrasting information into the ongoing discourse (Bittner & Bartz 2018).

The discourse particle doch, however, differs from aber in two aspects. First, doch, when used as a discourse particle, is constrained to the middle field of the clause, specifically to the right of the finite verb in V2 and to the left of the non-finite verbal elements in VF. In contrast, aber, when used as a connective, exhibits greater positional flexibility, being able to appear either clause-initially or clause-internally, as illustrated in (12).

Second, when examining B’s responses containing aber and doch to A’s statement in (13), subtle semantic differences emerge: while aber conveys a relatively mild denial (13a), doch signifies a more direct correction (13b). This contrast reflects underlying differences in how the two particles interact with presuppositional common ground: (13b) constitutes a more direct correction, as the use of doch triggers the uncontroversiality presupposition and thus implies that Speaker A’s utterance conflicts with what Speaker B assumes to be shared common ground. By contrast, since aber does not trigger such a presupposition, (13a) constitutes a comparatively weaker denial.6

Given the significant similarities between doch and aber, with the latter reported to slow down the process of finiteness acquisition in German (Bittner & Bartz 2018), a tentative prediction at this stage is that doch may behave in a similar way, delaying the emergence of finiteness marking in the host utterances. However, whether doch- and aber-clauses differ in their impact on finiteness development remains to be tested.

The aim of this study is to explore the role of the discourse particle doch in the development of finiteness in early German, in particular as it affects the development of functional finiteness, i.e. the emergence of a functional projection hosting the finiteness marker. Specifically, the study sets out to answer three research questions:

A corpus study designed to address these research questions will be presented in the next section.

The present study draws on spontaneous production data of three monolingual German children (Leo, Simone and Cosima), which are available as part of the CHILDES corpus (MacWhinney 2000). All the data used in this study were presented in transcribed text format. The three corpora were chosen for the following reasons. First, all of them have a relatively high sampling density, which can present a comprehensive picture of the developmental trajectory of children’s language acquisition. Second, the corpora were examined in a number of previous studies with respect to the relation between finiteness and various kinds of particles in early German acquisition (Dimroth 2009; Winkler 2009; Bittner & Bartz 2018). Therefore, using the same data can make the results from the present study directly comparable with previously reported findings.

The time span that is of particular interest for the present investigation covers the first two years since the first occurrence of doch as a discourse particle in child productions, which happens at age 2;00 for both Leo and Simone, and at age 2;03 for Cosima. Doch-utterances produced by children during this period of time were extracted from the corresponding corpora using the kwal-program in CHILDES, along with ten lines of context before and after each doch-utterance.7 Some general information about the three children under investigation and their doch-productions is presented in Table 1.8

In the next step, the utterances were manually sifted according to the exclusion criteria listed in (14).

Criteria in (14i–v) were adopted from Bittner & Bartz (2018), while (14vi) was added mainly to sift out occurrences of doch as a response particle.

Consequently, from the initial selection of 1422 utterances, only 604 doch-productions, comprised primarily of matrix declarative sentences with only a few wh-questions, remained for further analysis. Table 2 provides the number of utterances extracted and analyzed for each child.

The 604 total utterances were subsequently coded for a number of morphological and syntactic features listed in (15), which were adopted from Bittner & Bartz (2018).

The last three features were of particular relevance to the typical manifestations of functional finiteness introduced in (4) in Section 2.1, which are repeated for convenience in (16).

The effect of the discourse particle doch on the functional finiteness realization in child utterances was assessed by comparing the ages at which children regularly produce these constructions in doch-clauses with the ages when these structures appear regularly in clauses without any particles.11 The regular production time of each structure was determined following the third-use criterion introduced by Bloom et al. (1982): in general, the third occurrence of the relevant construction marks the onset of its regular production.

Despite individual variation, all three children start using doch as a discourse particle two to four months prior to their regular production of finite simple main clauses (cf. Table 3). The data concerning finiteness in simple main clauses in Table 3 are taken from Bittner & Bartz (2018).

The early occurrence of doch, however, does not contribute to an early realization of functional finiteness in doch-clauses. On the contrary, doch-clauses exhibit a general five- to sixteen-month delay in terms of functional finiteness marking as compared to simple main clauses. The delay observed in doch-clauses is fairly robust in that it is observed no matter whether the regular production time of finite doch-clauses is compared with that of simple main clauses as a whole or on each of the four finiteness-indicative constructions. Furthermore, when taking aber-clauses into consideration, which, according to Bittner & Bartz (2018), demonstrate the most delayed realization of functional finiteness compared to simple clauses and other particle-containing clauses studied to date, it emerges that the delay in doch-clauses is even larger than that in aber-clauses. Table 4 compares the onset age of regular production of the four constructions in simple main clauses, aber-clauses and doch-clauses. Here again, the data pertaining to finiteness in both simple main clauses and aber-clauses are borrowed from Bittner & Bartz (2018), where the onset of regular production of each structure is also determined according to the third- use criterion.12

The data in Table 4 unequivocally show a delayed realization of finiteness in doch-clauses as compared to simple main clauses. Putting this together with results from previous studies, the temporal order of functional finiteness marking in particle-containing clauses can be formulated in (17) (the symbol < stands for ‘earlier than’).

To conduct a more in-depth analysis of the development path of functional finiteness in doch-clauses, the data were divided into three stages (as outlined in Table 5 below), defined by whether functional (importantly, not morphological) finiteness is present overall, external to doch-clauses (i.e. in simple main clauses), or internal to children’s doch-productions, respectively, according to the criteria listed in (16). Names of the stages are in keeping with the terminology used in Bittner & Bartz (2018).

The pre-finiteness stage (Stage 1) covers the time period in which neither simple main clauses nor doch-clauses show finiteness as a functional grammatical property in child grammar. In the external-finiteness stage (Stage 2), properties associated with functional syntactic structure are observed in simple main clauses but remain largely absent within doch-clauses themselves. Data from the external-finiteness stage will be taken as crucial evidence for the delay in functional finiteness realization in doch-clauses. In the final, internal-finiteness stage, properties indicating functional finiteness are finally consistently realized within doch-clauses, marking children’s regular application of functional finiteness in this specific syntactic context. The onset age of Stage 3 is based on the first regular occurrence of the four indicators of functional finiteness in doch-clauses. Table 6 details the onset age and the number of doch-utterances identified at each developmental stage and draws a comparison between the data of doch-clauses and aber-clauses in Bittner & Bartz (2018).

It can be observed that doch-clauses constitute a relatively small proportion of children’s utterances in the corpus, especially in comparison to simple main clauses. Therefore, there exists the possibility that the observed delayed realization of functional finiteness in doch-clauses results from their lower frequency in child speech. But Table 6 rules out the possibility that the delayed appearance of functional finiteness in doch-clauses as compared to aber-clauses is purely due to a limited number or low frequency of the relevant utterances in the child data. In fact, doch-clauses are produced more frequently than aber-clauses even before functional finiteness is assumed to be established in both clause types (i.e. the onset of Stage 3). Despite this higher frequency, the third occurrence of doch-clauses showing functional finiteness properties still appears one and two months later for Leo and Simone, and ten months later for Cosima. This suggests that while frequency might play a role, it cannot fully account for the delay observed in doch-clauses.

In the following, I introduce the main quantitative and qualitative characteristics of doch-utterances at each stage and summarize the implications the data have for the general developmental pattern of functional finiteness in doch-clauses.

Regarding the third research question, namely the explanation for the impact of doch on the acquisition of functional finiteness, I argue that the delayed application of functional finiteness in doch-clauses is a consequence of the complexity in the semantics of doch and its significant similarities to aber ‘but’.

In Stage 1, children frequently use doch in a target-deviant way. In erroneous patterns, doch typically appears in contexts where no contradiction is intended, violating the correction presupposition. An example is given below in (39).

In this context, the interpretation that the candlestick holder remains in the bag, as the child’s use of doch might suggest, is inconsistent with the actual scenario, as indicated by the mother’s response. Assuming the child understands that the bag is indeed empty, the use of doch in this context appears inappropriate. It is not until Stage 2 that children gradually develop a target-like mastery of doch with almost no improper uses.

With respect to the primary finding of the present study, which says that doch-clauses exhibit a delayed acquisition of functional finiteness as compared to other particle-containing utterances, I argue that the reason lies in the considerable similarities between doch and the earlier acquired connective aber ‘but’. This resemblance necessitates a longer period for children to not only understand the intricate semantic features of doch but also to discern its distinctions from aber in the context of expressing contrastive meaning. The additional time required for these cognitive processes accounts for the slower developmental pace in the acquisition of functional finiteness in doch-clauses.

In order to account for the variations between individual particles, their semantic features and the comprehension difficulties they potentially impose on children must be analyzed. According to previous studies, the earlier occurrence of functional finiteness in nicht-clauses is promoted by its non-anaphoric use, whereby nicht simply negates the truth of its host utterance without referring to any previous utterances (Dimroth 2009; Winkler 2009). In contrast, particles that possess the opposite property, i.e. those that are used anaphorically, such as auch ‘also’, require children to comprehend the content of the utterance at hand as well as to look back in the pre-context and trace antecedents there. The requirement of looking backward, which is inherent in the anaphoric particles, is argued to increase the information-processing load for children and therefore results in a lag in finiteness realization (Dimroth 2009).

The particles auch and aber also contain more complex semantic meanings than nicht. The focus particle auch, besides its anaphoric use, conveys an additive meaning, which contributes to the introduction of new information on the topic. As a concrete example, auch in (40) on the one hand shows that the current statement is a comment on the previously introduced topic Maria comes, and on the other hand adds the time today to the set of situations where Maria may appear.

According to Dimroth (2009), the semantic properties of auch prove more difficult for children, who have to learn to consider the previous discourse, identify the antecedent of the auch-clause and integrate the newly added information into their previous discourse-related knowledge. These complications associated with the use of auch account for the prolonged period of time when the particle stays in the V2 position (between the topic and the predicate) before children master its appropriate use, which is argued to be exactly what impedes the realization of finiteness in auch-clauses.

With respect to aber ‘but’, the particle requires some previous information in the discourse and is excluded in out-of-the-blue contexts. This contextual-embedding property is also observable in utterances containing auch, but does not apply to those with the negative particle nicht: as illustrated in (41), among these three particles, only nicht is acceptable in an out-of-the-blue utterance.

Furthermore, similar to auch, aber also expresses an additive meaning and introduces new information to the discourse.20 But aber crucially differs from auch in that the new information it introduces is the opposite to what has been uttered in the context, therefore giving rise to an adversative meaning. Therefore, similar to utterances containing auch, those with aber require children to look back to the previous information in the discourse. But aber-utterances extend the cognitive demands placed on children beyond that: they additionally require children to comprehend the status of information of the aber-proposition in relation to the context and also critically assess the consequences of adding a piece of contrastive information to a set of assumptions that have been processed (Sperber & Wilson 1986). The difficulties children show in this contrastive information-structural alignment process are proposed by Bittner & Bartz (2018) to be the major cause for the delay of functional finiteness in aber-clauses.

Based on these proposals, the late development of functional finiteness in doch-clauses can be explained by invoking the semantic properties that doch shares with aber. Doch resembles aber in that both entail some prior contextual information in the emergent linguistic context and ask children to precisely identify the referred antecedent in the previous context, increasing the processing load on information management. Furthermore, both aber and doch have intrinsic additive and contrastive denotations, which require a proper information-structural alignment of the hosting utterance into the discourse.21

In examining the reasons for the delayed development of functional finiteness in doch-clauses compared to aber-clauses, I suggest the following possibilities. The first potential explanation emerges from the considerable similarities between aber and doch, which may lead to confusion among children in discerning how these two particles differ in conveying contrastive meaning. Given that aber is found to be acquired earlier, the subsequent acquisition of doch, which appears to share semantic properties with aber, might compel children to critically evaluate the nuanced distinctions between these two words. The complex semantics of doch, combined with the necessity to discern fine distinctions between it and the earlier acquired aber, intensify the cognitive demands placed on children. Consequently, the increased processing load is likely a contributing factor to the more protracted development of functional finiteness in doch-clauses.22

Furthermore, while doch as a discourse particle is in most cases unstressed, it can also be realized in stressed form, which, as argued by Dimroth et al. (2018), has the same function as the finite verb in establishing a relation between the comment and the topic, specifying whether the comment holds for the topic or not.23 As a concrete example, both (42b–c) can be used to express a contradiction to what is described in (42a), with the only difference that the stress is put on the finite verb in (42b) but on the particle doch in (42c), as indicated by small caps (Dimroth et al. 2018: 269).

If stressed doch overlaps functionally with verb finiteness, children may initially prefer lexical devices over verbal inflection for the abstract function of linking predicate and topic. Evidence for this tendency has been reported by Dimroth (2009), who found that children at early stages often resort to lexical devices, including negation nicht and particles such as auch ‘also’ and noch ‘still, yet’, as linking elements that validate relations between the predicate and the topic. Moreover, the use of such particles has been shown to interfere with the marking of finiteness on the verb (Penner et al. 1999; Nederstigt 2002; Tracy 2002). Nederstigt (2002), for instance, reports a conflict between the child Caroline’s use of auch and finite verbs. On this basis, stressed doch could likewise hinder both morphological finiteness marking on the verb and the expression of functional finiteness. Unlike doch, however, the connective aber is always unstressed, and thus unaffected by this factor. The additional complexity introduced by the stressability of doch may therefore contribute to the comparatively slower development of functional finiteness in doch-clauses relative to aber-clauses. Unfortunately, the phonetic realization of doch in child speech was not analyzed due to the unavailability of audio data. Future research could draw on auditory data and carry out a more in-depth analysis of children’s stressed and unstressed uses of doch, their interrelations with the sentence topic or predicate, and their potential influence on functional finiteness realization in the relevant utterances.

Finally, as mentioned earlier, doch-clauses were produced less frequently by children than simple main clauses, which may contribute to the later observed establishment of functional finiteness in doch-clauses. However, as Table 6 shows, frequency alone cannot fully explain this delay. This paper has therefore mainly focused on additional factors beyond frequency that may hinder the realization of finiteness in doch-clauses.

To summarize, the comparatively slow development of functional finiteness in doch-clauses was attributed to the relatively longer time children require to understand the semantic concepts involved in this particle, and knowing how to use it in the target-like manner. Moreover, the substantial semantic overlap between aber and doch presents an additional challenge, as it may potentially lead to confusion among children in differentiating their roles in expressing contrastive meaning. This semantic overlap results in a heavier cognitive load and a more complex learning task for German children when it comes to doch, which prolongs the acquisition process of the particle and impedes the development of the morphosyntactic and functional expressions of finiteness in subsequent stages.