Generative linguistics has long been concerned with the linguistic competence of the “ideal speaker-listener, in a completely homogeneous speech-community, who knows its language perfectly” (Chomsky 1965: 3). Research in formal-linguistics-based second language acquisition takes as its starting point the second language (L2) speaker's underlying mental representation. Here the factors of interest are influence of the learner's native language and, in generative SLA, the operation of innate linguistic mechanisms (Universal Grammar). Similar to methodology in formal syntax, lxSLA adopts techniques such as grammaticality judgment, comprehension and perception tasks supplementing spontaneously produced oral data. While there may be individual differences in oral production, tasks that tap learners' mental representations reveal commonalities across learners from a given native language background with the same amount/ type of exposure and age of initial L2 exposure. When it comes to phonology, age has long been a central factor with numerous comparative studies showing younger learners far outperforming older learners (see Piske et al. 2001). This paper discusses a case of possible non-acquisition by L2 children who had had considerable exposure to the L2. Children's non-acquisition is only apparent, and this allows us to consider the value of lxSLA methodology on the one hand, and and raises issues about what might be lacking in the current socio-SLA paradigm, on the other. We argue that only when we return to the cooperation that marked its birth in the 1960s will we have a comprehensive picture of SLA.