Sentence structure

In their attainment of the target grammar, DGS learners are confronted with the task of expanding their initial elementary structures by additional structural layers. The availability of complex sentential structures and their associated grammatical features is not only reflected at the level of word order; it also becomes apparent in the target-like marking of grammatical relations between constituents in a clause. Technically, this is expressed in terms of the expansion of the categorial-thematic VP structure by the additional functional projections, namely, the IP and CP.

IP tracking: syntactic arrangements and morphosyntactic landmarks

In our analysis, we have used the diagnostic criteria established in section 3.3 for the assessment of the availability of the IP. At the level of word order, we were interested to establish whether participants adhere to the SOV pattern. The aim here was to find out whether learners correctly set the VP and IP headedness values. Target-deviant orders were scrutinised for a potential impact of language borrowing from German. In order to determine whether the grammatical processes associated with the IP are operative, the data were analysed with respect to person and spatial agreement.

Verb-final structures. One recurrent observation in our scrutiny of the participants’ data with a view to assess their adherence to the target word order constraints is that participants seldom produce SOV structures in which all constituents would be expressed overtly. This finding is in line with previous studies on DGS and other sign languages, in which pro-drop and topic drop were found to occur frequently in spontaneous data (Johnston et al. 2007). While those studies are primarily concerned with the nature (and availability) of the mechanisms necessary to license empty elements (cf. Hanel 2005), the high proportion of subject drop and/or object drop in our data is a critical issue because it reduces the proportion of those utterances that would help us to unambiguously determine whether participants correctly set the VP and IP headedness parameters. Because the participants in this study are acquiring DGS in a bilingual situation in which they also use a manual code of the oral language (that is, LBG) and written German, attention was also paid to potential candidates for language mixing at the level of word order. Recall that DGS and German differ regarding their surface verb placement: while verbs appear clause-finally in DGS across the board, they appear in second position in main clauses and sentence-finally in embedded clauses in German. So, because there is a partial overlap, we decided to pay special attention to the relative position of object complements or other

modifying constituents and verbs in main clauses, since OV and XV patterns are target-like in DGS but deviant in German.

The analysis of the data reveals that where verbs and their modifying complements are expressed in a sequential fashion, participants adhere to the target XV format. XV patterns in the data collected often involve the expression of information about locations, whereby information on the ground precedes the verb in final position. This can be a simple verb as in Hamida’s utterance in (308) or a spatial verb in a complex classifier construction, used to express spatial configurations, as in Simon’s example (147), repeated here in (309). Utterances like (309) not only document the adherence to the target XV order; they also show that processes associated with the IP are operative as the spatial verbs involved appear in their inflected form.

A note is due in this context regarding the observation that the overt expression of object complements or other verbal modifiers occurs often in the context of what appear to be repetitions or semi-repetitions of propositions (as in (308)). Indeed, in several instances, participants recount first the general activity, before they produce a second proposition with the same verb, in which location, manner, patient or even the subject of the activity are further specified (cf. Table 3.46 below for an overview of the different types of repetitions observed in the narratives of the participants in this study). Typically, these sequences appear in the narratives of the third sample, characterised by a more detailed elaboration of the narrative events described. It is interesting to note in this context that the status of (semi-)repetitions in signed and spoken discourse has been addressed from various perspectives to determine what they might reveal regarding word order, language production, discourse organisation, and narrative development.

Repetition in DGS productions has received some attention in experimental studies on sign language production in adult signers. In a study dedicated to monitoring in DGS and German, Leuninger and Waleschkowski (2009: 23, their translation) remark on sequences such as (310)-(312) which they categorise as appropriateness repairs (A-repairs) from a language production perspective. Note that unlike self-repairs that result from the detection of an error (so-called E-repairs), A-repairs follow utterances that are deemed inappropriate to the context, which indicates that speakers do not only monitor their speech for error but also make repairs “to express the same ideas more appropriately” (Levelt 1983: 53). Such repairs might involve a specification of the timing (cf. (310)), the subject (cf. (311)) or the location (cf. (312)) of the activity described.

Particularly the two latter examples are reminiscent of the repetitions we observed in our data. However, our analysis of the data reveals that repetition does not only occur in repairs to appropriateness in the context of hesitations or other types of disfluency. Rather, it appears to reflect a more generalised phenomenon observed in research on ASL and other sign languages that might be ultimately related to the characteristics of discourse in language communities with an oral tradition, signed or spoken. Indeed, the phenomenon was first remarked upon by Fischer and Janis (1989: 281, their emph.) when they observed constructions in which the same verb (or verbs with similar roots) occurred twice, “separated only by the object and/or sentential adjuncts)” (compare example (313)).

Because the repetitions Fischer and Janis (1989: 292) observed involved verb forms that differed regarding the information encoded (in example (313) the second verb form contains aspectual information not expressed in the first instance) the authors invoke “the notion of ‘heaviness’, suggesting that when a verb becomes too “heavy” with attachments, it must split off and do double duty.” According to the authors (1989: 285) what all repeated verb forms in so-called verb sandwiches had in common was that they encoded “d i f f e r e n t information” (ibid., their emphasis). Further, the data obtained in a recent cross-linguistic study on word order in ISL (Irish Sign Language), VGT (Flemish Sign Language), and Auslan (Australian Sign Language) reveals that the phenomenon is quite common across sign languages. Indeed, “verb doubling” was found to occur in 16% of responses in the data collection (Johnston et al. 2007: 192). The few examples mentioned (e.g. boy hug with old^amother hug (ISL)) are strikingly similar to the ones observed by Fischer and Janis (1989) and by ourselves in the present study. So is one of the examples provided in a study of word order in Spanish Sign Language (LSE) (cf. (314) from Morales-Lopez et al. 2010: 20, our free translation based on the LSE glosses).

Beyond the descriptive level, Massone and Curiel (2004) also put forward their assumptions about the origin of variation in sign language production, including phenomena such as repetition, which they observed in their study on Argetine Sign Language (LSA). The authors (2004: 87) identify two main factors, namely, modality of expression, and oral tradition when they state that “[s]uch variations are evidence of a syntactic structure determined mainly by conversational factors that imply redundancy, repetition, focus, deletion of constituents, and syntactic elaboration typical of orally transmitted languages, all of which factors depend on the possibility of providing syntactic information by various articulators, both simultaneously and sequentially.” The relevance attributed to discourse factors is well in line with discourse oriented studies, in which repetition is regarded as a rhetorical feature that contributes to the organisation of discourse, through the provision of additional information and the establishment of links between different parts of the text produced (Tannen 1987). For further illustration we might consider Bavin’s (2004: 20) sketch of the “gradual build-up” style (the description is based on evidence obtained from young Warlpiri users recounting the frog story):

Information is repeated, maybe in a different form or word order. A “build-up” style is often used: information is repeated with some new added. There can be a gradual build-up of information; for example, in telling the frog story a speaker might give the information that someone fell, then someone fell to the water, then someone fell down to the water, and then specify that it was the child and the dog who fell. So not all is revealed at once, and perhaps this is a way of holding the attention of the listeners. Repetition is noted to some extent even in narratives of five-year-olds.

Against this backdrop, and without loosing sight of discourse characteristics related to the visuo-gestural modality of expression, it seems, assumptions about the impact of iconicity on story-telling in sign language, such as those expressed by Taub and Galvan (2001) need to be regarded with caution. Note that, according to Taub and Galvan (2001: 178), “ASL signers consistently incorporate much more conceptual information into their descriptions of motion events than do English speakers” which they argue reflects the “deep influence of iconicity on ASL descriptions of motion events”. As it turns out, the comparison of Bavin’s sketch of the gradual build-up style he observed in oral story-telling and the narrative elaboration documented for some participants in our study makes apparent that there are similarities across modalities that deserve further attention in research on sign language discourse.

From a developmental perspective, what we learn from the literature on narrative skills in spoken language learners is that the functions repetitions serve in narratives change over time. Based on the insights obtained in their broad cross-linguistic investigation, Berman and Slobin (1994: 183) remark that while repetition of nouns in young 3-4 year old children is assumed to reflect problems of lexical retrieval and disfluencies in extended discourse, children aged 5 use repetition as a rhetorical device to express aspectual distinctions, as for example in (315) to express protracted and iterative aspect (examples from Berman & Slobin 1994: 183). Repetition as discourse-based reiteration occurs later in the retelling of the same narrator (cf. (316)). Repetition as a rhetorical device also becomes apparent in (317).

Turning to the evidence obtained in our study we are inclined to interpret examples such as the ones listed in Table 3.46 not only regarding the participants’ command of the target head-final VP and IP (cf. (318)-(320)). These sequences also reflect the progress they make in their attainment of narrative skills, as the repetitions contribute to the creation of cohesion and coherence by providing further specifications on referents, (cf. (321)-(323)), goals (cf. (324) or locations (cf. (325)-(328)).

Table 3.46: Repetitions in the participants DGS narratives.

Table 3.46: continued

* Original numbers of examples discussed in previous sections. ** S=Simon, M=Maria, C=Christa, F=Fuad, H=Hamida

Language contact at the level of word order: verb placement variation. We

indicated previously that the incidence of target-deviant word orders in our corpus is low. However, there is one participant, Fuad, who occasionally produces sequences that do not comply with the target constraints. Consider, for example, the utterances in (212) and (213) provided here in (329) and (330). As we can see in (329), the PP ‘in the forest’ appears after the plain verb form search. Clearly, this sequence is reminiscent of surface SVO constructions in German (and LBG) and therefore represents a candidate for language mixing at the level of word order. The potential status of this utterance as a language contact phenomenon needs to be assessed in the context of the sentential formats produced at the time. Example (330), in which the locative complement appears in preverbal position, shows that Fuad produces other sequences that clearly adhere to the constraints imposed by DGS. Hence, although Fuad produces target-deviant SVO formats, it seems his DGS learner grammar is not a “German” grammar (in the sense that he would have misset the relevant word order parameters). Rather, it seems various parametric options are coexisting at the time, which is reminiscent of the “mobile IP” phenomenon observed in the productions of L1 and L2 learners of German (cf. section 4.3.2 below for further details). The succession of self-repairs in example (214) repeated here in (331) seems to corroborate our assumption about variation in Fuad’s DGS grammar at this stage.

In our analysis of the data we also remarked upon the participants’ target-deviant use of SVO patterns with the auxiliary pam. We will take up this phenomenon below, after our discussion of the main findings concerning the processes related to the IP.

Grammatical processes related to the IP. Thus far we have been concerned with word order and the issue of whether elements in a clause are arranged in a target-like manner. We turn next to the question of whether morphosyntactic processes associated with the IP are operative. Recall that the availability of the IP is

reflected in the target-like inflection of verbs as, by assumption, verbs are raised to the INFL position so that their features be checked. Before we look at what the data reveal regarding verb inflection it is important to acknowledge here that the distinction of plain, agreement and spatial verbs is mastered by all participants: there is, indeed, no evidence of a confusion of verb types as it has been found to occur in the early productions of young infants. That said, it must be noted that the narratives collected in this study vary regarding the range of inflected verbs they contain. Particularly in the narratives produced in the first recording, the range of agreement verbs is very limited, typically including the verbs look- at, wave and take. Although other verbs are used at a more advanced narrative stage, the plot of the frog story itself contributes to a rather restricted selection of agreement verbs. Not surprisingly, verbs of spatial motion and location occur frequently in the recounts of a story that revolves around the protagonists’ search of the run-away frog (we only have to think of the protagonists’ walking around as they search various locations).

Verb inflection: agreement verbs. The analysis reveals that although some participants produce agreement verbs infrequently, in particular in their file 1 narratives, the verbs they use appear in their inflected form without exception. At the same time, we remarked that subject and/or object reference in constructions with these verbs was not always clear, indicating that participants fail to mark referential identity at times. These observations allow for the conclusion that while verb inflection is mastered by all learners at the onset of the study, deficits remain regarding the syntax-discourse interface which models referential maintenance.

For further illustration we might consider the relevant processes as they are sketched in Figure 3.13. In this sketch, utterance (A), produced by Simon in file 1 is represented as an IP structure: by assumption, inflected verb forms are raised to the head-final INFL position, where its features are checked. While the verb form in (A) encodes an object argument by picking out a locus in the sign space (to the right, bottom), referential identity remains generic as the object is neither referred to via an overt lexical element or associated previously with a locus in the sign space. Whether or not the object in (A) and the subject of the previous narrative passage (that is, the dog) are identical cannot be established unambiguously: Simon previously describes the anxiety of the dog running away. He does so in the context of an SRF, with eye gaze directed toward the bottom right, which corresponds with the location of the locus picked up by sting to mark object agreement. However, this type of non-manual marking only represents an optional agreement marker in DGS.

Figure 3.13: Referential ambiguity and the syntax-discourse interface.

From a developmental perspective, the discrepancy observed regarding the mor- phosyntactic and the discourse level patterns well with previous findings on the acquisition of other sign languages (notably, BSL and ASL), indicating that learners take their time before they fully master the integration of the information from different levels of analysis.

Verb inflection: spatial verbs. Typically, participants produce constructions with spatial verbs that inform about the protagonists’ activities as they wonder about the whereabouts of the runaway frog: the boy and the dog walk about several locations, climb up a tree and a stone, or fall down a cliff, which prompts descriptions involving spatial verbs such as fall, go, climb-up. Spatial verbs appear in their target-like inflected form as of the onset of the study (cf. examples (332)-(336)), which corroborates the assumption that processes associated with the IP are operative. However, at times, participants fail to provide specific information on the agent and/or location of the activity, which indicates that deficits remain at the narrative level, in particular at the beginning of this study. For further illustration we might consider Simon’s file 1 example (162) repeated in (332), The utterance involving the verb climb-out is a sophisticated sequence with a complex classifier construction, in which an h2-classifier backgrounds the information about the location the agent climbs out from. However, only the audience acquainted with the frog story might infer that it is the frog climbing out of the jar, because Simon does not provide any further information on the location and fails to reintroduce the referent of this activity. Fuad’s file 1 example (216) repeated in (333), by contrast, is clear because he reintroduces the agent with a full NP and det_art. However, in this case, too, the nature of the location remains unspecified as he only reports previously that the boy and the dog are looking at the frog (he does not mention that the frog is sitting in a jar).

Another spatial verb frequently used in this corpus is the verb go. This verb appears, at times, without any locative specification, as is the case in Christa’s file 1 example in (334), in which we learn that the frog leaves somewhere to the right. Other examples are more sophisticated, with spatial verb forms agreeing with the location established previously. In Christa’s file 1 example (292) repeated in (335), for example, the boy speculates about the frog being in the forest. Notice that the verb go in the subsequent clause correctly agrees with the locus associated previously with the forest.

Participants’ data document the command of the appropriate choice of classifier elements. Constructions with the verb fall, for example, typically involve the classifier element for human beings. Where referents are not specified, in particular when referential shifts succeed each other, it is not always clear who falls, although the information can often be retrieved from the story context. Fuad’s file 1 example (221) provided in (336), marks an exception, as it remains unclear who is actually falling after the dog hits the tree. As it turns out it is likely to be the beehive falling. Yet, because Fuad uses the classifier element for human beings it is not obvious that this interpretation is what he has in mind when he signs the utterance.

det_exist Participants’ use of det_exist to mark spatial and referential agreement adds a piece to the puzzle of determining the status of the structure available to the learners. Clearly, the productive and creative use of det_exist documented in the narratives can be taken as an indicator of the availability of the IP phrase, and where it is used as one of several means to create cohesion it is an indicator of the

mastery of the syntax-discourse interface. We will expand on the latter dimension in section 3.11.2.1.

Personal agreement marker (pam) and language contact. In our sketch of the main characteristics of DGS, we noted that pam is used in constructions with plain verbs to mark object agreement (and case). Hence, constructions with pam provide an additional cue for the availability of the IP. Participants in the present study, however, used this auxiliary fairly infrequently. Indeed, none of the participants used pam in the narratives of the first recording. Neither did they use verbs that would require its use such as the verb like. Because information on the emotions of the protagonists is provided sparingly in the narratives collected at the beginning of this study, it comes as no surprise that this expression is not used at the time (the same holds of predicative constructions with the attributive adjective cross). As for the constructions with pam appearing in the third sample, the analysis reveals that they are target-deviant at the level of word order. Rather than being arranged in accordance with DGS constraints, constituents appear in a sequence that is reminiscent of surface main clause SVO order in German. Note that the target-deviant pattern appears with various types of predicates, namely, (a) in predicative constructions with the attributive adjective cross (cf. (337)), (b) in constructions with the agreement verb sting (cf. (338)), and (c) in constructions with the plain verb like (cf. (339)).

The observation that word order in constructions with pam is target-deviant across the board, raises the question of why participants would choose a sentential format (that is, SVX) they do not use otherwise. By assumption, the use of pam in SVX constructions constitutes a hybrid phenomenon borrowed from LBG. Indeed, the use of manual means to represent spoken utterances in LBG results in hybrid sentential patterns that correspond neither to one or the other language. This is the case of LBG constructions in which the sign auf (‘on’) that corresponds with the DGS sign pam is used to mark the verb complement relation (note that the generalised use of auf to mark object and case agreement has no equivalent in German). Interestingly, there is additional evidence from the written German

samples discussed later in this work indicating that hybrid LBG constructions not only affect the participants’ productions in DGS but also their productions in written German. As we will see in section 4.11.2.4, some of the participants’ written narratives reveal an erroneous generalised use of the preposition auf (‘on’) as an agreement and case marker at a time when the case and agreement paradigms are not yet available in the German learner grammars.

Occasional combinations of the sign have with a predicative adjective, documented in the productions of Simon in his file 3 (compare (175) repeated here in (340), point into the same direction, that is, the influence of LBG resulting in a type of construction that is neither DGS nor German (in German the equivalent of (340) would involve the use of the auxiliary sein, ‘to be’).