The literature on second position elements (a.k.a. second position clitics) is characterized by a longstanding discussion as to whether these elements are base-generated in second position or whether they are base-generated in first position and only arrive in second position as the result of a subsequent transformation.
In this paper, I will provide two arguments for the transformational approach based on a typological study of second position coordinators (Weisser 2024), i.e. coordinators that appear inside of one of their conjuncts rather than in the semantically transparent peripheral position (e.g. in between the two conjuncts). The first argument will be based on the observation that, although the coordinators usually occupy a second position within their coordinands, they also show positional alternations in some syntactic contexts. These alternations, I argue, can be derived in both transformational and base-generation models. However, we find that the kinds of empirically attested
alternations are systematically restricted in a way that falls out of a transformational approach but that cannot be captured in a base-generation approach. In other words, the base-generation approach massively overgenerates predicting much more alternations than we actually find. The second argument is based on opaque interactions between the second position placement and
other morphosyntactic operations.