On their way toward full mastery of the language, children are faced with the complex task of acquiring the target language properties at the distinct levels of linguistic analysis, on the one hand, and, on the other hand, they face the challenge of integrating the knowledge attained and using it appropriately for different purposes. The production of narratives, for example, involves the orchestration of several linguistic devices from distinct levels of analysis, within and across sentences, and bigger discourse units (Berman & Slobin 1994; Morgan 2006). Against this backdrop we turn next to the research undertaken regarding sign language learners’ attainment of those properties that involve the syntax-discourse interface.
Referential establishment and maintenance
Scholars coincide in the observation that gaining mastery of nominal establishment and reference maintenance represents a prolonged process. To a certain extent, this observation comes as no surprise, given the complexity of the acquisition task. It might be useful at this point, to recollect what learners of a sign language like ASL have to accomplish, by drawing on Lillo-Martin’s (1999: 538) succinct description of what the child has to learn to use pronouns and verb agreement appropriately, that is, “(a) to associate a referent with a location (b) to use different locations for different referents (except when a group of referents is being referred to or, in some cases, for possession), (c) to use verb agreement or pronouns for non-present referents, and (d) to remember the association of referents with locations over a stretch of discourse.” Notice that lexical, morphosyn- tactic and discourse knowledge is involved as well as general cognitive abilities, needed, for example, for the memory of several spatial locations.
Interestingly, as indicated previously, there is consensus that it takes some time before children master the appropriate use of referential loci, including the choice of contrastive locations and their consistent use to indicate referential identity during longer discourse stretches. By constrast, the question of whether young children are able to understand the information encoded through referential frameworks remains virtually unexplored (but see Lillo-Martin 1999 for a study indicating that a high level of comprehension was not observed until age 5). We turn next to a summary of the main findings obtained in production studies.
First pronouns. The change from a prelinguistic use of indexical pointing as a gesture to the use of pointing as a linguistic unit (that is, as a pronoun) has been found to occur during the second year (at about age 1;6—1;11 according to Baker et al. 2005). Pronoun reference to the addressee (“you”) is reported to appear at about 2 years (though at times with reversal errors) and reference to a third person at about age 2;5. The following two examples illustrate the use of personal pronouns in a child learning DGS. The pronoun in (91) refers to the addressee, the pronoun in (92) to a 3rd person (the person referred with a locus established toward the door is in the adjacent room) (cf. Hanel 2005: 168, our transl.).
Present vs. non-present referents. During an initial phase, non-present referents are not established productively, which patterns with the development observed for agreement verbs; as we have seen in the previous section, the association of referents with referential loci, for example, in constructions with agreement verbs is only mastered late (4;11, Baker et al. 2005).
We mentioned previously the discrepancy observed by some authors regarding the use of agreement markings with present and with non-present referents. Interestingly, children have been found to express reference to non-present referents through the use of agreement verbs in their citation form together with overtly realised arguments (until the age of about 3;6 in Lillo-Martin 1999: 538). Because a similar “distribution” of target-like and deviant verb forms has not been observed in spoken language acquisition the question arises as to the linguistic and general cognitive factors that might affect the acquisition of agreement with “non-present referents” vis-a-vis its acquisition with “present referents” (Morgan et al. 2006: 27). Morgan et al. (2006: 27), for example, remark that “[i]t is not clear whether this late use of agreement morphology and abstract locations in sign space has to do with linking a word to a non-present referent (a general conceptual issue) or more to do with the particular linguistic devices used to refer to a non-present referent (indexing of abstract locations in sign space).”
Clearly, from a linguistic perspective, the apparent distinction is unexpected once the mechanisms necessary to mark agreement are in place (cf. also Hanel 2005: 210). If agreement morphology is the overt reflection of the establishment of an abstract grammatical relation, children should not make any difference with respect to present and non-present referents. So, most authors seem to agree on the assumption that the observed linguistic behaviour is related to performance limitations pertaining, in particular, to memory for abstract spatial locations (Lillo-Martin & Chen Pichler 2006: 241). What is important about these considerations is that what is assumed to be cognitively demanding for the language learner pertains to the broader level of discourse, as a consistent and appropriate use of referential frameworks involves the orchestration of linguistic devices beyond the sentential level.
Inconsistent use of referential loci. Indeed, the studies available make apparent that prior to the mastery of the agreement system at about the age of 5 (age 4;9 [in Lillo-Martin 1999: 538] [4;11 in Baker et al. 2005]), children do not use referential loci consistently, failing to distinguish different loci for different referents (using one and the same locus, referred to as “stacking”), or not using the same loci for the same referents throughout a narrative. Further, children were also observed to use substitute referents (as for example, the pictures of a picture story, cf. Hanel 2005: 135f.; Lillo-Martin 1999: 538-9). Example (93) illustrates the use of the picture story book as a referent for the object, the subject being null (Lillo-Martin 1991: 158).
Integration of information from different levels of linguistic analysis. Following our differentiation of the information involved in referential establishment and maintenance (section 184.108.40.206), learners’ errors reflect remaining shortcomings at the level of discourse. Note that learners are not only confronted with the tasks of learning (a) the different mechanisms that can be used to establish and maintain reference, and (b) the lexical items that might be involved in these processes (distinction of verbs belonging to different classes), they also need to learn (c) that there is the possibility of shifting the referential framework (Lillo-Martin 1991: 162). What the preceding observations make apparent is that that the powerful “clarity of reference” in sign language production remarked upon previously (Bellugi et al. 1990: 16, cf. section 220.127.116.11) does indeed pose a challenge for language learners. The data discussed in this section suggest that the mastery of this “clarity”, involving the integration of knowledge from distinct levels of linguistic analyses, represents a task that is not mastered in a one-step process. From a more general language learning perspective, this fits well with Leuninger’s (2000: 255) observation that learners do not tackle all tasks at once and that the new tasks they confront background the knowledge attained. This view of language learning processes is certainly well in line with what we know about the complex dynamics that characterises a modularly organised grammar.
-  So-called “reversal errors” refer to the use of the first person pronoun to refer to the addressee and the non-first person pronoun to refer to the first person, a phenomenon that has also beenobserved in spoken language acquisition (Foster-Cohen 1999).