Abstract
Studies on L1- and L2-acquisition of German and Dutch have shown that the particles too/also and again hamper the realization of finiteness while the particle not promotes it. In this study the authors ask whether adversative but also affects the realization of finiteness. By applying a structural-functional approach to finiteness and its acquisition they investigate the morpho-syntactic development of aber-clauses in the first 12 months of aber-production in four monolingual German children. The analysis reveals that finiteness becomes realized in aber-clauses considerably later than in auch-clauses. Syntactically well-formed aber-clauses appear when finiteness becomes realized in main clauses not containing any particle or connective. However, it takes the children two to four more months to realize finiteness in the aber-clauses. The authors argue that the information-structural properties assumed to cause the delay raised by the particles too/also and again are not fully applicable to but. Particle-specific effects also have to be taken into consideration. In case of but this includes the complexity of information-structural alignment and discourse integration of the twofold contrast established by aber-clauses. Managing this communicative task reduces the capacities for realizing primarily structural conditions which are still not automatized.
Keywords
Introduction
Two decades ago Dietrich and Grommes (1998) reported differences in the development of finite clause structure (henceforth, FCS) depending on the presence vs. absence of the negation particle nicht ‘not’ in L2-acquisition of German: Early clauses containing the negation particle nicht exhibit a finite verb in V2-position more regularly than clauses not containing nicht. Somewhat later, Penner, Tracy, and Weissenborn (2000) reported similar findings for L1-acquisition: Clauses containing nicht (1) show higher proportions of finite verbs in V2-position than clauses containing the additive particle auch ‘also/too’. Early auch-clauses frequently show an infinitive in clause-final position or lack a verb at all (2). Also, elliptical structures are very common and even non-target omission of verb arguments occurs for a longer period (3). The data analysis further indicates that the proportion of finite verbs in V2-position is still higher in nicht-clauses than in particle-free clauses.
(1) a. *CHI: mag nicht nuss (Leo, 2;02.02)
1
like-1SG not nut ‘I don’t like [the] nut’ b. *CHI: sehe gar nicht fische (Leo, 2;02.23) see-1SG not at all fishes ‘I don’t see fishes at all’ (2) a. *CHI: Papa auch gucken (Leo, 2;02.16) Daddy too look-INF ‘Daddy shall look, too’ b. *CHI: hier auch türen (Leo, 2;02.17) here too doors ‘Here [are] doors, too’ (3) a. *CHI: ja, und Gertrud auch (Leo, 2;05.21) yes, and Getrud too ‘Yes, and Gertrud, too’ b. *CHI: auch reparieren (Leo, 2;05.10) too fix-INF ‘[This one] needs to be fixed, too’
Later studies on L1- and L2-acquisition of German and Dutch confirmed these findings (L1: Nederstigt, 2003, 2006; Winkler, 2006, 2009; for a summary see Dimroth, 2009; L2: Dimroth, 2002, 2009; Dimroth, Schimke, & Verhagen, 2009; Schimke, Verhagen, & Dimroth, 2008).
However, a satisfactory explanation for the different effects the particles have on the development of morpho-syntactic clause structure is still missing. Penner et al. (2000) were mainly interested in the syntactic development of scrambling. Concerning the use of verbs, they just stated that auch-clauses remain ‘conservative’. Dimroth (2002, 2009), Dimroth et al. (2009) and Winkler (2009) assume that the effects are caused by information-structural properties of the most typical and initially most frequent uses of the particles. Two aspects are proposed to be crucial in this respect (see Dimroth, 2009 for a summary of this approach): First, the contextual integration of the particle-clause, that is, whether it is anaphoric or not; and, second, the syntactic position of the particle, whether it is placed next to its elements in scope (adjacent henceforth) or not. Regarding contextual integration, it is stated that the negation particle nicht is most typically used non-anaphorically while the additive particle auch is anaphoric. Dimroth (2009) posits that nicht-clauses in natural discourse typically negate a newly introduced assertion. Auch-clauses, on the contrary, add something to a proposition already given in discourse. Most typical in general and most frequent in early production, the added part is the topic-constituent of the auch-clause (2a/3a). Dimroth calls this topic-oriented use the ‘contrastive topic case’ and contrasts it with the ‘neutral topic case’ which is the more frequent use of nicht. The difference of the two particles in topic orientation is mirrored by differences in scope properties. Both particles are typically placed in postverbal position. In the ‘neutral topic case’ (4a), this goes hand in hand with adjacency of the particle to the information units in its scope. In the ‘contrastive topic case’ (4b), however, the particle is non-adjacent to the information units in its scope.
(4) a. ‘neutral topic case’ A: Liesa hat gestern gesungen. ‘Liesa was singing yesterday.’ B: Liesa singt auch heute. ‘Liesa is singing today also.’ b. ‘contrastive topic case’ A: Gestern hat Liesa gesungen. ‘Yesterday, Liesa was singing.’ B: Heute singt Liesa auch. ‘Liesa is singing today also.’
Dimroth (2009) argues that the ‘contrastive topic case’ is challenging for the learner since it exhibits a difference between the scope of the particle and the scope of the assertion that does not exist in the earlier stages of learner grammar. Prior to the acquisition of finiteness as a grammatical property, learners treat affirmative particles like nicht, auch, wieder ‘again’, schon ‘already’, etc. semantically and syntactically as linking-elements (Jordens & Dimroth, 2006). Syntactically positioned between topic and comment, these elements indicate the type of relation between topic and comment. In the transition to FCS, learners have to figure out that, even in the presence of a particle, the assertion has to be marked separately by a verb in V2-position. 2 The ‘neutral topic case’ is assumed to be less challenging since assertion and particle have the same scope here. Contextual integration of nicht-clauses of the ‘neutral topic case’ is also assumed to be less challenging due to their non-anaphoric character. Nicht-clauses of this type do not require information-structural alignment and appropriate encoding of alternatives.
The assumed effect of the two information-structural patterns (‘neutral topic case’ and ‘contrastive topic case’) is, however, only partially confirmed by a study investigating elicited production data from Dutch and German L2-learners (Dimroth et al., 2009). In addition to nicht/niet and auch/ook, this study includes the additive particle wieder/weer ‘again’. The expected difference between the ‘neutral topic case’ (higher rate of FCS) and the ‘contrastive topic case’ (lower rate) is found with the two additive particles but not with nicht/niet (40% FCS in both cases). Also, the analysis reveals a difference between the two additive particles in the ‘neutral topic case’ which is assumed to not hamper the realization of FCS: While about 68% of wieder/weer-clauses are found to be finite, only 55% of auch/ook-clauses are finite. The findings have to be seen in relative terms due to the scant data for the ‘neutral topic case’ and the collapsed analysis of the German and the Dutch data. However, they hint at possible differences between semantically related particles. In a further study on L2-acquisition of German and Dutch, Schimke et al. (2008) compared auch/ook and wieder/weer without differentiating ‘contrastive topic case’ and ‘neutral topic case’. The study confirmed that both particles hamper the realization of finiteness. In this case, however, the effect was stronger with wieder/weer than with auch/ook. This suggests that there are also differences between the two particles in the dominant ‘contrastive topic case’.
Taken together, there is evidence that the two information-structural patterns in particle use have effects on the acquisition of FCS. However, the differences between auch/ook and wieder/weer found by Schimke et al. (2008) and Dimroth et al. (2009) as well as the unexpected finding for nicht in Dimroth et al. (2009) suggest that the information-structural properties do not provide a satisfactory explanation of all observed effects.
So far, only a few particles of the affirmative type have been investigated. The spectrum of particles to take into account for an exhaustive explanation might, however, be broader than this. German children, for instance, start producing a range of other particles between age 2;0 and 2;6, that is, exactly at the time when target marking of finiteness is acquired. Among these particles are additive ja ‘indeed’ and noch ‘even/more’, adversative doch ‘but/yet’, aber ‘but’, wohl ‘indeed/surely’, trotzdem ‘nevertheless/though’, temporal schon ‘already’, noch ‘still’. All these particles (and the adverbial as well as connective uses of some of them) introduce a certain relation between two propositions and require discourse-appropriate alignment of the two propositions including alignment of their information-structural properties.
Given the current state of the art, the adversative particle/connective aber ‘but’ is specifically interesting. First, it belongs to a semantic class of particles that was not involved in the previous studies. Moreover, it shares semantic properties with both nicht and auch, the two particles for which opposite effects on the realization of FCS have been reported. In this article, we present an analysis of the morpho-syntactic development of clauses containing aber in German L1-acquisition that addresses whether aber also has a certain effect on the emergence of FCS and, if so, which factors are responsible for this effect. Theoretically, the emergence of FCS in aber-clauses can resemble that of (i) auch-clauses, (ii) nicht-clauses, or (iii) simple main clauses, or (iv) show an individual path. Figuring out the actual path will provide further insights into the effects information-structural properties have on the realization of morpho-syntactic-clause structure and, eventually, highlight other properties that might play a role.
Before presenting our study in detail, we briefly introduce the functional-structural approach on the acquisition of finiteness proposed by Jordens (2012), which is adopted to determine the emergence of finiteness in early main clauses. Further, we introduce the relevant structural properties of German aber-clauses as well as the findings on commonalities in the semantics of aber with auch and nicht presented by Brauße (1998) and Sæbø (2003).
Acquisition of finiteness
The emergence of FCS is bound to the acquisition of finiteness as a grammatical property. Probably all approaches to L1- and L2-acquisition comprise a specific account of the acquisition of finiteness as the core grammatical feature of clauses. Approaches taking into account V2-languages exhibiting verb inflection consider the emergence of the inflected verb in V2-position as realization of finiteness (cf. Poeppel & Wexler, 1993). That is, an utterance like Papa schläft noch ‘Daddy is still sleeping’ is assumed to be finite. However, it has also been noticed that most of the verbs produced early occur in only one form and only one structural position (Bittner, 2003; Clahsen, 1986; Jordens, 2002). Forms of copula, modal verbs and certain lexical verbs occur in inflected form in second position, whereas lexical verbs of a different type occur in infinitive form in sentence-final position. The finding raises the question of whether the early verb forms displaying target agreement are indeed used as finite verbs, in the sense of assigning the assertion of the proposition (Klein, 1998).
By discussing German and Dutch L1- and L2-data, Jordens (2012) showed that the two inflectional types of early verb use correspond to two types of utterances: infinitive forms occur in utterances expressing action predicates and inflected forms occur in utterances expressing state predicates. While the latter exhibit a theme-subject and an inflected verb in second position (Ball rollt weg. ‘Ball is rolling away’) the former exhibit an agentive-subject and an infinitive form in sentence-final position (Mann Ball werfen. ‘Man ball throw’). As Jordens showed, only when learners come to generalize over these two structures does finiteness as a functional grammatical property emerge in their grammar. That is, the inflected verb in V2-position functions as marking of assertion (Klein, 1998). According to Jordens, learners proceed from a lexically organized grammar to a functionally organized grammar when they come to place verbs of all types in V2-position. Acquisition of finiteness in this sense (functional finiteness henceforth) is a prerequisite for several structural changes. Most prominent among them are the emergence of (i) variable word order (subject inversion), (ii) auxiliary verbs, 3 (iii) verbs expressing action predicates in V2-position and (iv) verbal particles in clause-final position. 4 The presence of these four structures clearly indicates the realization of functional finiteness. We therefore use these four criteria for the determination of functional finiteness in our study.
German aber ‘but’ – syntactic and semantic properties
German aber occurs as a particle and connective. Structurally and even semantically, both uses are hard to disentangle. This is illustrated in (5) where aber is in connective position in (5a) and in particle position in (5b–d). Aber occurs in both main clauses and subordinate clauses and can be placed in front of all syntactic constituents without changing the overall meaning of the proposition (5). In contrast to nicht, auch and wieder, aber cannot be stressed.
(5) Die Sonne scheint am Tag, a. b. der Mond c. der Mond scheint d. der Mond scheint nachts ‘The sun shines in the day, but the moon shines bright at night.’
All studies investigating the semantics of aber agree that aber indicates a certain kind of contrast or polarity 5 to what has been said before or to what can be inferred from preceding discourse or world knowledge. There is, however, a big debate on the nature of this contrast. More recent studies convincingly argue that the contrast is not grounded in the content of the propositions related by aber. Mostly, two propositions can be equally well combined by additive und ‘and’ (6a) or adversative aber (6b).
(6) a. Mia ist gross und Pete ist klein. ‘Mia is tall and Pete is small.’ b. Mia ist gross, aber Pete ist klein. ‘Mia is tall but Pete is small.’
Brauße (1998) argues, that whether additive und or adversative aber is appropriate depends on the actual discourse, more precisely, the implicit question under discussion (the quaestio according to Klein & von Stutterheim, 2002). While in (6) an implicit question like How big are Mia and Pete? logically requires the use of additive und (6a), 6 a quaestio like Are both of them tall? can only be answered logically by using an adversative expression (6b). Brauße (1998) puts the difference between und and aber as follows: und adds information that is compatible with previous information under the given quaestio; aber adds information that is not. Accordingly, Brauße (1998, p. 145) formalizes the difference between und and aber by characterizing aber as negated und.
Sæbø (2003) compares aber with auch and states: ‘aber introduces a presupposition that can be encoded in a meaning definition differing minimally from the one for auch’ (Sæbø, 2003, p. 264). This ‘presupposition should be defined in terms of
(7) a. Max trinkt Bier und Moritz trinkt Bier/Wein. ‘Max drinks beer and Moritz drinks beer/wine.’ b. Max trinkt Bier. Moritz trinkt auch Bier. (*Wein) ‘Max drinks beer. Moritz drinks beer too.’ (*wine) c. Max trinkt Bier, aber Moritz trinkt Wein. (*Bier) ‘Max drinks beer, but Moritz drinks wine.’ (*beer)
Both Brauße (1998) and Sæbø (2003) propose that there is only one underlying form and semantics of aber. This means that its different use types discussed in many other studies (see e.g. Lang & Umbach, 2002; but also Brauße, 1998) emerge from interactions with the linguistic and extra-linguistic context. As Gülzow, Bartlitz, & Kuehnast (2015) showed, all of the core use types of aber are present from early on in acquisition.
Taken together, the reported findings on the syntax and semantics of aber make it hard to derive a specific expectation of the developmental path of aber-clauses towards FCS. The fact that aber includes semantic features of auch and nicht (and und ‘and’) makes both observed paths possible, fast development as with nicht or slow development as with auch. The finding that all core use types of aber are present from early on limits expectations on a decisive impact of anaphoricity. And, finally, the relatively variable syntactic position of aber limits expectations on a decisive impact of scope realization (adjacency). So, all possibilities listed above remain open. However, the higher semantic complexity of aber suggests that if there is an impact of aber, it should slow down the realization of FCS.
The study
Database and coding
The study is based on the longitudinal data of four monolingual German children. Three children (Simone, Cosima, Anna) were recorded for about 90 minutes two to four times per month. The corpus of the fourth child (Leo) is a dense corpus and comprises recordings of about 60 minutes five times a week up to age 2;11 and five times a month from age 3;0 on. The data of Simone, Cosima and Leo are part of the CHILDES database (MacWhinney, 2000). 8 The data of Anna was collected by Dagmar Bittner and transcribed according to the CHILDES transcription guidelines. Using the kwal-command of the CHILDES CLAN tool we selected all aber-productions documented within the 12 months following the first aber-production. 9 With Anna, data collection stopped at age 2;10; in her case, the aber-productions of only 11 months could be selected. Table 1 gives the source of the data and the analyzed age period for each child. 10
Database for the individual children.
The data contained 1154 aber-productions in total (see Table 2). The following criteria lead to exclusion of an aber-production from analyses:
reported from diary (only occurring in the transcripts of Leo);
formulaic expressions (songs, storybooks, daily routines, etc.);
identical and immediate self-repetition;
completely unintelligible utterance;
occurrence in subordinate clause;
occurrence in yes/no-question;
occurrence in an imperative clause.
Number of documented and analyzed aber-productions for the individual children.
Aber-productions in subordinate clauses, yes/no-questions and imperative clauses (n = 19) have been excluded since these clause types do not exhibit a finite verb in V2-position in adult German. 11
As Table 2 shows, 1047 aber-productions remained for morpho-syntactic analysis. These data were coded for the following properties:
syntactic position of aber (clause-initial (see (5a) vs. clause-internal (see (5b–d));
dialogic split of base-clause and aber-clause (yes/no):
yes: base-clause and aber-clause are produced by two different speakers (see (12));
no: both base-clause and aber-clause are produced by the child (e.g. da is(t) so eine Trambahn drauf, aber die fahren nicht mehr. ‘There is such a tram, but these (types of tram) are not going anymore.’ Leo, 2;10.10);
ellipsis (yes/no):
yes: contextually licensed omission of constituents (see (12));
realization of all obligatory constituents in non-elliptic aber-clauses (yes/no; type of missing constituent);
type of inflected verb (copula, modal, auxiliary and lexical verb expressing a state/action predication);
realization of the potentially finite verb (no verb, infinitive, in/correctly inflected or in/correctly placed inflected verb).
In addition, all aber-clauses were coded for the four criteria indicating the realization of functional finiteness (Jordens, 2012):
haben ‘have’ + past participle (yes/no);
verbal particle in clause-final position (yes/no);
word order (SVO, OVS, XVS (X = constituent other than S or O));
type of predication expressed by the aber-clause (state vs. action).
In order to compare the morpho-syntactic structure of aber-clauses on the one hand with main clauses not containing any particle or connective on the other hand, we also controlled the emergence of (g) to (j) in the latter type of clause (simple main clauses henceforth). This was done by checking the transcripts in chronological order until regular production of the four structures (g–j) was attested. To determine the onset of regular production of the four structures, we took into account the differences in both the amount of data per child and the frequency of the four constructions. Typically, the third lexically different instance of each structure (types) was taken as attesting regular production.
Results
Concerning the age of emergence of aber-production as well as the properties listed under (a)–(c), there is considerable individual variation among the four children. At the same time, however, all four children display a common path in the morpho-syntactic development of aber-clauses. In the following, we report on the results indicating this common path in morpho-syntactic development.
First of all, the analyses concerning the regular production of the four structures indicating functional finiteness (properties g–j) reveal a two- to six-month delayed application of functional finiteness in aber-clauses as compared to simple main clauses (Table 3).
Age of onset of regular production of structures indicating functional finiteness.
Further, the order of application of the four structures differs in aber-clauses vs. simple main clauses. While all four structures begin to be regularly produced within one month in simple main clauses, this is a stepwise process in the aber-clauses lasting over three to six months (Table 4).
Onset of regular production of the structures indicating functional finiteness in simple main clauses and in aber-clauses.
The data suggest a typical order of application of the four structures in aber-clauses:
13
(8) step 1: Lexical verbs expressing action predicates in V2-position step 2: OVS word order step 3: haben ‘have’ + past participle step 4: Clause-final position of verbal particles expressing action predicates
In order to relate the delay in the application of functional finiteness in aber-clauses to the findings on auch-clauses reported in the literature (see Introduction), we performed an analysis of auch-clauses focusing solely on the emergence of the auxiliary construction haben ‘have’ + past participle. First, the result confirms the earlier findings: haben ‘have’ + past participle occurs later in the auch-clauses than in the simple main clauses. With respect to the aber-clauses, however, the data show an earlier emergence of haben ‘have’ + past participle in the auch- than in the aber-clauses. 14
The analyses conducted on the morpho-syntactic properties (d)–(f) of the aber-clauses (see below) revealed that qualitative changes towards target clause structure emerge in correspondence to progress in the acquisition of functional finiteness. The first qualitative change in the morpho-syntax of aber-clauses occurs when functional finiteness becomes established in simple main clauses. The second qualitative change occurs when functional finiteness becomes realized within the aber-clauses. Thus, the development of aber-clauses in the first year of aber-production displays three stages distinguished by the emergence of functional finiteness in (i) simple main clauses and (ii) aber-clauses (Table 5).
Realization of functional finiteness (yes/no) in simple main clauses and aber-clauses: Differentiation of Stage I to III of aber-production.
From the perspective of the aber-clauses, Stage I can be called the pre-finiteness stage, Stage II the external-finiteness stage, and Stage III the internal-finiteness stage. Table 6 presents the age of onset of each stage for each child.
Age of onset of Stage I, II and III in the morpho-syntactic development of aber-clauses.
Simone seems to exhibit an alternative strategy for the expression of contrast until age 2;5. See note 15.
Table 7 presents the number of aber-clauses analyzed per child and stage.
Number of analyzed aber-clauses per stage.
In the following, we report the qualitative changes in the morpho-syntactic structure of aber-clauses over the three stages.
As mentioned, Stage I covers the aber-clauses produced prior to the acquisition of functional finiteness in simple main clauses and aber-clauses. The 25 aber-clauses documented for Stage I (Table 7) come from three of the four children. 15 They are characterized by omission of obligatory constituents (n = 14; see Table A1 in the Appendix; please note, tables A1 to A4 are all presented in the Appendix): nine aber-clauses lack a verb (9) (Table A2), five show omission of obligatory arguments, most typically the subject (10), and one aber-clause lacks the verb and an obligatory argument.
(9) a. da aber wasser (Anna, 2;00.29) there but water ‘but there [is] water’ b. hier mund aber (Cosima, 2;05.27) here mouth but ‘but here [is] a mouth’ (10) a. aber andere farbe nehmen (Leo, 2;02.04) but other color take-INF ‘but [we should] take another color’ b. angst habe aber (Anna, 2;00.29) fear have-1SG but ‘but [I] am afraid’
Of the six clauses containing a verb, three contain an infinitive form, two a modal verb and one an inflected lexical verb expressing a state predicate. Only two aber-clauses of Stage I are syntactically complete. One of them is completely well formed (11a). The other one shows the inflected verb stem in clause-final position than in target V2-position (11b).
(11) a. aber ich kann das auch (Cosima, 2;06.10) but I can-1SG this too ‘but I can [do] this, too’ b. ich aber wiederkomme (Anna, 2;00.05) I but again-come ‘but I [will] come back’
There are, further, seven ellipses (12) produced by two children (Leo, Cosima). 16
(12) *MUT: du, ich glaub die trambahn passt nicht drauf auf die eisenbahnschienen, Leo ‘I think the tram does not fit on the rails, Leo’ *CHI: Zug aber. (Leo, 2;02.02) train but ‘but [the] train’
Summing up, aber-clauses at Stage I typically lack the verb (62–73%; Table A4). Verb omission is still high when excluding ellipses and unclear items from analysis (50–67%; Table A3). Omission of obligatory verb arguments is less frequent but also typical.
Stage II in the production of aber comprises the period from the emergence of functional finiteness in simple main clauses to its realization in aber-clauses. For Stage II 123 aber-clauses are documented (Table 7). The main qualitative step from Stage I to Stage II in the morpho-syntactic structure of aber-clauses is the realization of the inflected verb. About 50% of all aber-clauses exhibit a correctly inflected verb now (Table A4). Out of the 29 verbless aber-clauses only five are incomplete due to omission of the verb. Eighteen verbless clauses are well-formed ellipses. When excluding ellipses and unclear items from analysis, omission of the verb occurs in less than 15% of the data now (Figure 1; Table A3).

Omission of inflected verbs in Stage I and II (without ellipses and unclear structures).
The data also reveal a drastic decrease in the omission of obligatory arguments from Stage I to Stage II (Figure 2).

Aber-clauses with incomplete verb-argument-structure in Stage I and II.
Aber-clauses of Stage II further differ from Stage I aber-clauses by the sample and proportion of verb types included (Table A4). Now copula forms also appear and lexical verbs expressing state predicates are the most frequent verb type. Aber-clauses containing non-inflected verb forms predominantly express bans and desires, that is clause types frequently remaining infinite in adult German too.
To conclude, Stage II aber-clauses appear to have a well-formed clause structure but, nevertheless, lack the properties indicating realization of functional finiteness. As a consequence, they render a relatively static SVO pattern.
Stage III starts when the structures indicating functional finiteness (properties g–j) become regularly produced in the aber-clauses. There are 899 data points documented for this Stage. 17 Stage III aber-clauses differ from Stage II aber-clauses by two qualitative changes: now all verb types occur in V2-position and all types of constituents can be placed in clause-initial position (the Vorfeld). That is, lexical verbs expressing action predicates (13) and auxiliary constructions (14) also occur in V2-position, and non-subject constituents occur in the Vorfeld while the subject is realized in the Mittelfeld (14).
(13) a. ich mal aber was schönes (Cosima, 3;00.28) I paint-1SG but something beautiful ‘but I paint something beautiful’ b. aber zuerst bauen wir meine holzeisenbahn auf (Leo, 2;09.24) but first build-1PL we my railway up ‘but first we build my railway’ (14) a. aber dismal haben wir einen ICE genommen (Leo, 2;09.19) but this time have-1PL we an ICE taken ‘but this time we have taken an ICE’ b. jetzt hab ich die aber schon abgewaschen (Anna, 2;05.18) now have-1SG I this but already washed ‘but now I have already washed this up’
Figure 3 illustrates the increase in the use of lexical verbs expressing action predicates and haben + past participle in V2-position in Stage III. Table A4 allows comparison of the sample and proportion of verb types in Stage III and Stage II.

Aux haben ‘have’ + PP and finite verbs expressing action predicates in Stage II and III.
Table 8 gives the number of Stage II and Stage III aber-clauses showing OVS word order.
Number of OVS structures in Stage II and III.
The order of emergence of regular production of the four structures suggests two parts of Stage III: an initial phase in which the structures become regular that contribute to either the communicated content (verbs expressing action predicates in V2-position) or the flow of information in discourse (OVS word order), and a later phase comprising the emergence of the primarily structural operations (haben ‘have’ + past participle and clause-final position of all verbal particles of lexical verbs).
Summary
Our analyses of the morpho-syntactic development of aber-clauses over the first 12 months of aber-production addressed the question whether aber-clauses show a delay in the application of functional finiteness. This question has to be answered with yes. The data reveal a delay of two to six months as compared to clauses not containing any particle or connective. Comparison with clauses containing the particle auch revealed that aber-clauses are even more delayed than auch-clauses. Further, we addressed the question of the steps in which the development of aber-clauses towards FCS proceeds. The analyses indicate a correlation in the morpho-syntactic development of aber-clauses and the acquisition of functional finiteness in general, allowing the identification of three stages: a pre-finiteness stage comprising the aber-productions prior to the acquisition of functional finiteness at all (Stage I), an external-finiteness stage comprising the aber-productions after the emergence of functional finiteness in simple main clauses but prior to its emergence in aber-clauses (Stage II), and an internal-finiteness stage comprising the aber-productions after the realization of functional finiteness in the aber-clauses (Stage III). Comparison of the three stages revealed that well-formed aber-clauses exhibiting an inflected verb in V2-position and all obligatory arguments (Stage II) emerge prior to the application of functional finiteness in aber-clauses (Stage III). Further, the data suggest that regular production of the four structures considered to indicate the presence of functional finiteness (emergence of all types of verbs in V2-position, OVS, haben + past participle, and verbal particles of all types of lexical verbs in clause-final position) in aber-clauses occurs in exactly this order. That is, the more content-related structures become regularly realized before the primarily structural operations.
Discussion
In the following we discuss possible explanations for our two main findings: the considerable delay in the emergence of functional finiteness in aber-clauses and the emergence of well-formed clause structure prior to the realization of functional finiteness. Most interesting in this respect is Stage II of aber-production.
First, it is worth noticing that the late emergence of functional finiteness in aber-clauses is clearly not simply a consequence of the higher semantic complexity of aber (Brauße, 1998; Sæbø, 2003) and its comparatively late acquisition (Evers-Vermeul & Sanders, 2009). Mere semantics of a particle/connective do not necessarily affect the realization of finiteness. This comes to light by the fact that the additive particle auch emerges before the negation particle nicht in children’s speech production, while emergence of finiteness with these particles shows the reversed order (Dimroth, 2009; Penner et al., 2000; Winkler, 2006, 2009). Also, stressed auch is produced prior to unstressed auch (Müller, Höhle, Schmitz, & Weissenborn, 2009; Nederstigt, 2003) but realization of finiteness is less hampered with unstressed auch (Dimroth, 2009).
Obviously, one has to take into account additional factors other than mere semantic complexity of the particle when searching for an explanation of late realizations of functional finiteness. As reported in the introduction, Dimroth (2009), Dimroth et al. (2009) and Schimke, Verhagen, and Turco (2011) propose two factors causing late realization of finiteness in auch- and wieder-clauses: the anaphoric relation of the particle-clause and non-adjacent position of the particle and its constituents in scope (see Introduction). However, as stated before, anaphoricity is less likely to have an impact in the case of aber. As Brauße (1998) and Sæbø (2003) make clear, aber-clauses most typically establish a new assertion by presenting a new topic and a predication that does not hold for the topic of the related proposition. Even if one considers aber-clauses expressing only a new topic and polarity as being anaphoric as in (12) repeated here as (15), there is no evidence that this type of aber-clause becomes finite earlier than types including a completely new comment (16). Both types of aber-clauses are regularly produced from Stage II on.
(15) a. *MOT: du, ich glaub die trambahn passt nicht drauf auf die eisenbahnschienen, Leo. ‘I think the tram does not fit on the rails, Leo’ (Leo, 2;02.02) *CHI: zug aber train but ‘but the train’ (16) a. *CHI: der [=andere stift] geht aber gut.18 (Anna, 2;01.27) this one [=other pen] work-3SG but well ‘but this one works well’
Concerning the second factor, which is non-adjacency, it is worth noticing that non-adjacent positions of aber could easily be avoided given its variable position (5). The data suggest that this is indeed what happened. Specifically, instances of clause-initial aber are exceptional when the topic is contrastive. 19 Taken together, the factors proposed to have a delaying effect on the realization of FCS in auch-clauses and wieder-clauses are seemingly not the decisive factors for the delay in aber-clauses.
In our view, it is worth including the semantic commonality of aber with the negation particle nicht in the explanation of the developmental path of aber-clauses towards FCS. This might be surprising since nicht-clauses are reported to become finite prior to all other clause types investigated so far (Dimroth, 2009; Penner et al., 2000; Winkler, 2006). However, the semantic feature polarity expressed by both nicht and aber corresponds with non-anaphoric use and the necessity of providing an assertion on its own by a syntactically complete proposition. In our view, this constellation explains why both clause types show well-formed clause structure relatively early.
While, however, nicht-clauses are reported to be ahead of simple main clauses in both the realization of verb-argument-structure and finiteness, these are two separate steps in the morpho-syntactic development of aber-clauses. The realization of complete verb-argument-structure in aber-clauses seems to be promoted by the acquisition of functional finiteness in simple main clauses, that is, its emergence in the learner’s grammar at all. By this step, the V2-clause pattern becomes the dominant pattern of main clauses. It is introduced to aber-clauses by the use of verbs which are rote-learned in V2-position, that is, lexical verbs expressing state predicates (fallen ‘to fall’ and haben ‘got’), copula verbs and modal verbs. As the data show, it takes several months until lexical verbs expressing action predicates (rennen ‘to run’) and the auxiliary construction haben ‘have’ + past participle emerge in the V2-position of aber-clauses as well. The same late emergence holds for OVS structures and clause-final position of verbal particles in action predicates.
Our hypothesis is that the children are faced with two challenging tasks in Stage II of aber-production. First, the acquisition of functional finiteness in simple main clauses indicates the onset of a transition from lexical to grammatical organization of clause structure (Jordens, 2012). This means that the children are in the process of reorganizing their grammar by implementing functional properties. The missing application of functional finiteness to clauses of certain types such as, for example, auch- and aber-clauses indicates that this process has not yet reached the level of completely replacing the structures of the earlier grammar (see Winkler, 2009 for a similar argument). This includes the lack of automatization of the required structural operations. Automatization, however, is needed for reducing the effort in the application of structural conditions. In this situation, particularly operations of a primarily structural nature, that is, operations not motivated by semantic or pragmatic requirements, can be hampered by other acquisition tasks. Second, like auch-clauses, aber-clauses require semantic-pragmatic integration in the ongoing discourse. While typical nicht-clauses can be uttered without context (17) this is impossible for typical aber-clauses (18).
(17) Es regnet heute nicht. ‘It is not raining today.’ (18) Aber es regnet heute (nicht). ‘But today it is (not) raining.’
As we have argued above, the integration of aber-clauses in discourse is more complex than that of auch-clauses. While the latter most typically add exactly one new constituent to what has been said before, aber-clauses typically present a twofold or even a complete deviation from previous information. The primary challenge in stage II is the information-structural alignment of the contrasted pieces of information and its coherent integration in the flow of information. The child has to ensure what contrasts with what and in what sense the difference is relevant in the present discourse. Difficulties with this task come to light when analyzing the contrast relations in detail (Bartz & Bittner, forthcoming). In the present analyses the difficulties become visible by the stepwise emergence of the four structures indicating functional finiteness in the aber-clauses (see Table 4 and the results on Stage III): The structures contributing to either the communicated content (verbs expressing action predicates in V2-position) or the flow of information (OVS word order) become realized prior to the structures comprising primarily structural conditions (haben ‘have’ + past participle and clause-final position of all verbal particles of lexical verbs).
To sum up, typical aber- and nicht-clauses share the property of non-anaphoricity. This promotes the realization of syntactically complete and well-formed propositions. Well-formed aber-clauses emerge when the children establish the V2-pattern of main clauses in their grammar. A clear sign of the appearance of this step is the realization of functional finiteness in simple main clauses. Realization of functional finiteness in aber-clauses takes some more months since aber in contrast to nicht requires information-structural alignment and discourse integration of the provided information. In this sense aber is comparable with auch. The even later realization of functional finiteness in aber-clauses as compared to auch-clauses is caused by the higher complexity of new and, in addition, contrastive information in aber-clauses. In Stage II of aber-production, the children are faced with the challenge of information-structural alignment and coherent integration of the contrastive information in the ongoing discourse. In this situation, the children do not have the capacity to realize primarily structural operations which are not yet automatized, like the positioning of verbs used in clause-final position so far (verbs in action predicates) in V2-position or like topicalization by OVS word order. They, instead, rely on structures acquired in earlier stages of grammatical development.
Footnotes
Appendix
Proportion of verb types in aber-clauses in stage I, II, and III.
| No verb | Non-finite forms | Copula | Modal verbs | LexV state predicate | LexV action predicate | Aux: sein ‘be’ + PP | Aux: haben ‘have’ + PP | Unclear | ||
|---|---|---|---|---|---|---|---|---|---|---|
| Stage I | Leo | 8 (73%) | 2 (18%) | 0 | 0 | 0 | 0 | 0 | 0 | 1 (9%) |
| Cosima | 5 (62%) | 0 | 0 | 2 (25%) | 0 | 0 | 0 | 0 | 1 (13%) | |
| Anna | 4 (67%) | 1 (17%) | 0 | 0 | 1 (16%) | 0 | 0 | 0 | 0 | |
| Total | 17 (68%) | 3 (12%) | 0 | 2 (8%) | 1 (4%) | 0 | 0 | 0 | 2 (8%) | |
| Stage II | Leo | 8 (20%) | 3 (8%) | 7 (18%) | 6 (15%) | 10 (26%) | 0 | 0 | 0 | 5 (13%) |
| Cosima | 6 (34%) | 0 | 2 (11%) | 4 (22%) | 4 (22%) | 0 | 0 | 0 | 2 (11%) | |
| Anna | 6 (17%) | 2 (6%) | 5 (14%) | 7 (19%) | 12 (33%) | 0 | 0 | 1 (3%) | 3 (8%) | |
| Simone | 9 (30%) | 5 (17%) | 2 (7%) | 3 (10%) | 4 (13%) | 2 (6%) | 3 (6%) | 0 | 2 (7%) | |
| Total | 29 (24%) | 10 (8%) | 16 (13%) | 20 (16%) | 30 (24%) | 2 (2%) | 3 (2%) | 1 (1%) | 12 (10%) | |
| Stage III | Leo | 74 (16%) | 21 (4%) | 88 (19%) | 52 (11%) | 116 (25%) | 44 (9%) | 17 (4%) | 17 (3%) | 42 (9%) |
| Cosima | 40 (25%) | 4 (2%) | 22 (14%) | 41 (25%) | 15 (9%) | 19 (12%) | 2 (1%) | 8 (5%) | 11 (7%) | |
| Anna | 22 (17%) | 1 (1%) | 17 (13%) | 51 (39%) | 16 (12%) | 17 (13%) | 1 (1%) | 2 (2%) | 3 (2%) | |
| Simone | 26 (19%) | 4 (3%) | 20 (15%) | 39 (29%) | 16 (12%) | 15 (11%) | 0 | 7 (5%) | 9 (6%) | |
| Total | 162 (18%) | 30 (3%) | 147 (16%) | 183 (21%) | 163 (18%) | 95 (11%) | 20 (2%) | 34 (4%) | 65 (7%) |
Acknowledgements
For helpful comments on previous versions of this article, we are thankful to Jeruen Dery, Christine Dimroth, Peter Jordens, and two anonymous reviewers.
Declaration of conflicting interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was funded by: German Research Foundation (DFG), Grant BI 482/4-1 and 4-2 and German Federal Ministry of Education and Research (BMBF), Grant 01UG1411.
