Procedure

Scope and procedure of the investigation were developed by the author in close collaboration with the research team concomitant to the Berlin bilingual education programme.

Timing and researchers involved. Initially, sessions were scheduled every 5-6 months for the period of 3 years. Later it was decided to collect additional data once a year in years 4 and 5 after the beginning of the study. Signed and written narratives were elicited on separate dates to avoid potential cross-linguistic effects. DGS narratives were recorded first, with a time span of about 2 weeks before the written data were collected.

Because linguistic skills of investigators might affect the data elicited,^[1] to the extent that participants assess their language skills, which might influence their language choice (and language mixing), we decided to split the collection of the data by person: a deaf signer, member of the research team, would record the narratives elicited in DGS, using this language also in interactions with the participants; the written productions, in turn, were recorded and collected by the conductor of the concomitant research team and the author (both hearing, second language learners of DGS).

Procedure. Participants received a paper copy of the frog story book pictures. During the sign language elicitation sessions these copies were put on a board situated at the left or right side of the participant. During the written German sessions participants had their picture copies on their table. Participants were also given word lists that had been prepared by the research team conductor. Furthermore they were allowed to ask for words during their writing (upon request, information about their spelling was provided on the paper or via fingerspelling).

Only for the first recording we decided to shorten the rather long frog story (the pages not provided are indicated in Table 2.7 below).

DGS productions were recorded on an individual basis, whereas participants were grouped for the writing sessions. The participants were videotaped during their narration of the frog story in DGS and in written German. In the DGS recordings, the video camera was placed next to the addressee. Video-recordings of the written German sessions were made for the purpose of capturing the participants’ writing process, their reviewing of the written productions, and potential interactions between the participants and the researcher. The video camera during these sessions was placed in front of the group. The analysis of these clips, however, is not part of this study.

Data coding: DGS. All DGS data were transcribed and coded in ELAN,^{[2] [3]} using annotation conventions established specifically for the purpose of this analy- sis.15 In a first step, the signed narratives were translated into German glosses by deaf individuals native in DGS. In a second step, the transcriptions were counter-checked by the deaf colleague in charge of the collection of the DGS data and the author (independent inter-coder reliability).

The ELAN files created for each recording contain several annotation tiers (cf. Table 2.6), apart from the two tiers created for glosses of the signs produced with the dominant and the non-dominant hand respectively. Separate tiers were created for specific measures of (a) syntax (syntactic categories, word order patterns, clause types), (b) morphosyntax (verb inflection, with information on verb type, arguments encoded, locus selection), (c) referential establishment and maintenance (loci selection, coindexation), (d) reference forms and referential functions served, (e) referential shifts (signalling and marking of referential frameworks), and (f) error types. Separate tiers, except for the line created for annotators’ remarks, were aligned to allow for the identification of simultaneously occurring phenomena. Subsequently, the data were entered into a data base that permits analyses of frequency and distribution of specific phenomena.

In this context, a note is due on the caution imposed on word order analysis in the interpretation of data for which it is difficult, at times, to establish clause boundaries. This issue is not exclusive to the present study but has also been raised in research on the grammatical properties of sign languages (cf. Johnston et al. 2007: 189), where it has been pointed out that “[t]he identification of clause boundaries is difficult, and differences in the analysis can lead to differences in putative constituent orders attributed to an utterance.” We might add to this that the intricacy of an appropriate interpretation is increased where the analysis pertains to productions of bilinguals (or bilingual learners) because, in this case, errors are often readily interpreted as candidates for language borrowing.

Data coding: written German. The handwritten narratives were transcribed and entered into a data base using annotation conventions established specifically for the purpose of this analysis. Several coding lines were created (cf. Table 2.6) for specific measures of (a) syntax (syntactic categories, word order patterns, clause types), and (b) morphosyntax (verb inflection). The data base was created to also allow for analyses of error frequency and distribution (including deviances at the lexical, morphological and syntactic levels).

Table 2.6: DGS and written German data bases: information coded.

• [1] The phenomenon is well known in the area of language acquisition research where it is referred to as “researcher paradoxon”. Basically, the problem refers to the circumstance that theparticipants’ behaviour without the presence of the researcher is impossible to determine. Theissue is particularly critical in studies on language choice and language contact in bilinguals. Incase studies with investigators involved as participant-observers (as it often occurs in child L1case studies) the researcher turns into an independent variable (Tracy 1994/5: 194). As for thepresent study, the investigators did not engage in conversational interactions with the participants. The procedure was explained to them in either language prior to the recording. Exchangeswith the participants occurred on an occasional basis during written language productions forthe purpose of vocabulary requests only.
• [2] ELAN is a professional tool for the creation of complex annotations on video and audio resources. It is available from http://tla.mpi.nl/tools/tla-tools/elan/.
• [3] In the choice of our own specific conventions we have been inspired by those used in common transcription systems such as BTS, CHILDES or ECHO.