Complex classifier constructions
In the acquisition of either a signed or spoken language, the greatest challenges lie in the
integration of elements into larger structured wholes. (Slobin 2008: 22)
Several studies have been concerned with the acquisition of constructions with classifiers, including those with verbs of motion and location. Acquisition tasks commonly distinguished in the literature (Schick 2006: 111; Slobin 2008) concern (a) the selection of handshape to represent a semantic category (entity classifier), a category on how the hand interacts with the object (handle classifier), or a category based on the visual geometric features of the referent (entity classifier, based on SASS properties defined previously), the choice being
related to the thematic structure of the verb, (b) the expression of spatial relations through simultaneous coordination of the hands to represent a moving or located figure with reference to a ground, or description of the ground prior to the articulation of the figure in relationship to the ground, and (c) the selection of handshapes in relation to entities established previously in discourse (for the purpose of creating cohesion).
In general terms, studies on the acquisition of classifier constructions in learners of various sign languages (ASL, BSL, HKSL) coincide in the observation that the target-like mastery is not achieved until well into school age (around the age of 8-9 years, cf. Schick 2006: 111, or later, cf. Slobin et al. 2003). Interestingly, children have been found to master individual components of classifier constructions rather early; however, they continue to have difficulties for some time before they appropriately integrate figure and ground while expressing path or manner through the motion of the verb (Morgan et al. 2008: 5, pace Newport and Meier 1985). The intricate use of linguistic means in the expression of spatial relations is succinctly described by Slobin (2008: 22; cf. also Schick 2006: 112):
The handshapes for figure and ground must be contextually correct and conventionally appropriate; the ground must be indicated as well as the figure, with appropriate timing; the orientations of both figure and ground must be referentially appropriate; the movement must be within signing space and performed with conventional trajectory, rate, and rhythm; and co-occurring features such as path and manner/rate/intensity must be articulated simultaneously.
We turn next to a summary of the major developmental steps identified in the literature.
Whole body depictions. In their longitudinal study of a deaf child aged 1;0-3;0 Morgan et al. (2008: 8f.) observed that before age 2;0 the child used whole body depictions to describe movements such as “falling” or “jumping”. Such whole body depictions impose physical limitation on the simultaneous expression of figure and ground so that learners a this stage express either path or figure.
Real-world substitutes. In a next step, between ages 2;0-2;6, ground, path or manner are expressed through finger tracing, real-world objects or the physical ground itself (Morgan et al. 2008: 8f.). Children express complex movement and manner paths through tracing of an index finger (e.g. zigzagging, pirouetting, overtaking, crossing-over), without, however, combining these descriptions with a handshape classifier for figure. Instead descriptions are preceded by nominal signs (e.g. car, plane, man). In their study, Morgan et al. (2008: 14) observed that “many meaning components were expressed by the child in our study through gesture before he developed conventionalised signs.” At times, the child would use real-world objects (e.g. a toy car) to depict the movement, or use the surface of a table or the floor to depict the path while using a flat palm handshape classifier (vehicle). Notice that Morgan et al. (2008: 7) categorised utterances involving real object manipulation as gestures (and also utterances with whole body pantomime depictions, and directional traces without hand- shape classifiers), because although “these different types of gestures successfully express different semantic aspects of motion and location events (...) [they] do not use the representational sign space in front of the signer.”
Slobin et al. (20 03: 274), based on a study of ASL and NGT acquisition, too, remark on the early ability to productively combine meaning components (including conventional ones and ad hoc gestures) in early learners (2-year old children and hearing parents with 1 year or less of signing experience). Further, although children were found to have problems with the coordination of two handshapes to represent the spatial relation of a figure and a ground at this age, they managed to produce two handed classifier constructions when provided with an adult model; such imitations, as Slobin et al. (2003: 283) remark “provide information about the “growing points” or “zone of proximal development” in their signing. Slobin et al. (2003: 284) also mention that children have been found to initially use objects or the body (their own or the adults’) to circumvent the difficulty of using the two hands in figure-ground classifier constructions (a strategy that is apparently also being used by the signing adults interacting with the children) (Slobin 2003: 284).
In their study of a group of deaf children at the ages of 6-13 years who had been exposed to HKSL at various ages and reached different levels of proficiency, Tang et al. (2007: 297), too, observed an initial stage at which children with little lexical knowledge of HKSL used real-world substitutes (in particular, their own body) for figure or ground. Notice, that the use of real-world objects are also used as “substitutes” by young infants’ to refer to non-present referents (section 22.214.171.124).
Early classifier constructions. According to Morgan et al. (2008: 10) gesture forms disappear from age 2;6 onwards, coinciding with the time the authors identify as the onset for the productivity of classifiers. However, handshape selection
errors continue to occur after this age. The systematic use of motion forms, with different handshapes, was not observed until between 2;6-3;0, neither was the systematic use of the same handshape with different motion and location forms.
Children do not yet produce two-handed classifier constructions to express spatial relations appropriately. According to Tang et al. (2007: 294) two-handed classifier constructions in which “specific thematic information is assigned to the two articulators independently” (cf. example (94)) are acquired later than what they call “typical classifier predicates” (encoding one argument) because they involve a higher degree of morphological complexity.
Later development. Studies on sign language learners’ productions of constructions with two classifiers to encode spatial configurations of entities (transitive motion predicates and locative predicates), such as the one reported in Tang et al. (2007) on the acquisition of HKSL, show that learners are better at encoding figure than ground. Slobin et al. (2003: 289f.) remark that after an initial phase, in which classifiers for ground entities are omitted, NGT learners of late preschool/ early school age often fail to correctly integrate these classifiers. In their study of HKSL learners, Tang et al. (2007: 307) too, observed a frequent drop of ground vis-a-vis figure, which they relate to the roles the figure assumes semantically (i.e. AGENT or THEME) “thus making omission impossible” (Tang et al. 2007: 308).
Further, Tang et al. (2007: 305) point out that the higher frequency of twohanded constructions in the productions of the more advanced (Level 3) learners goes along with a greater number of errors. Mostly, these consist in the wrong choice of orientation of the figure hand against the ground hand. Although the proportion of errors in the expression of ground using the non-dominant hand is lower for these advanced learners, errors remain, which leads Tang et al. (2007: 305) to conclude that the “non-dominant hand for encoding a ground object emerges at a later stage of the acquisition process.” According to Schick (1990: 369) the prolonged development documented for predicates expressing locative relationships is related to the circumstance that this information is “best considered adjunct information” vis-a-vis verb agreement associated with the thematic structure of the verbs (subcategorisation), with the latter type of information being generally acquired earlier than the former.
Integration of information from different levels of linguistic analysis. In
our view, the late mastery of these constructions reflect the challenges imposed by constructions that involve the interface between syntax and discourse. This aspect is reflected in a type of learner error observed by several authors. Tang et al. (2007: 308), for example, remark on the use of classifier predicates without previous production of a lexical antecedent referring to the referent of the classifier (cf. also Morgan 2006; Slobin et al. 2003). The ability to retain a classifier in a narrative was observed in more advanced learners. According to Tang et al. (2007: 310) the results show that learners progressively acquire (a) the knowledge about the morphological composition of classifier constructions, and (b) the referential characteristics of classifiers (relation between classifier and antecedent). The latter dimension, pertaining to the information status of reference forms at the level of discourse, is taken up again in section 126.96.36.199, dedicated to narrative development.
-  As pointed out in a commentary article to Morgan et al.’s (2008) study, Slobin (2008: 22) remarks that “the spontaneous use of roughly representational handshapes and movements maybe a sort of gestural bootstrap into sign language”, but that beyond this “start” children have tolearn the conventionalised handshapes and movement patterns and use the sign space appropriately. The assumption is discussed in more detail in Slobin et al. (2003: 272).
-  Schick (2006: 113) summarising the available evidence (including her own study) remarks onthe early mastery of entity classifiers (at about age 5) whereas SASS and handle classifiers seemto still pose problems at that age. In general, the author remarks on the use of classifiers “thatare more generic than needed”.