*3.4. Procedure*

Signers performed the task in a dyadic setup, in which they were requested to find a total of twelve minor differences between two versions of an altered cartoon illustration of a scene (see Figure 2 as an example) by conversing with their interlocuter. The pictures were presented on laminated sheets, and each signer could only see their own picture and not the picture of their partner. One signer was given the role of circling the differences. This role was alternated between scenes in the task (i.e., different cartoon illustrations). The task was repeated twice with the same interlocutor, each time using a different cartoon illustration.

Participants completed this task in three conditions: (1) with another KQSL-ISL bilingual signer, (2) with a monolingual KQSL signer, and (3) with a monolingual ISL signer. Therefore, there was a total of six picture scene pairs: (1) kitchen (as in Figure 2), (2) field, (3) street, (4) beach, (5) living room, and (6) riverbank. Participants engaged in other tasks in between conditions, to ensure that the task was not repetitious.

#### *3.5. Data Coding and Analysis*

On average, the completion of all three scenes took 15 min, ranging from 8–20 min. Data were coded using ELAN, a video annotation software (Crasborn and Sloetjes 2008). Only the target lexical items, which differ between KQSL and ISL, were coded for this task. Language preference was quantified across individuals and conditions based on the percentage of the lexical signs in each language from the overall produced lexical signs. For example, if the signer produced a total of 80 tokens in their retelling task, of which 63 are KQSL (78.75%), then the preferred language is KQSL.

We carried out multivariate statistical analyses of the data using Rbrul (Johnson 2009). Similar to GoldVarb program developed by Rand and Sankoff (1991), Rbrul can quantitatively evaluate the influence of multiple factors on variation. In addition, Rbrul uses mixed-effects modeling to group individual responses accounting for the effects of individual differences (Baayen et al. 2008; Jaeger 2008).
