This paper explores areas of reflexive and pronominal noncomplementarity in German Accusativus cum Infinitivo (AcI)-constructions and other configurations that involve embedding of a subclausal (smaller than IP) constituent. The main empirical finding is that there are several non-complementarity-inducing contexts that have not been noticed before. In examples with embedded AcIs or complex DPs, a reflexive may be bound across an intervening subject, and a pronominal may be free even if its antecedent is only a vP-, DP-, or PPboundary away. The proposed account of these facts is based on the claim that vP, DP, and PP are all potentially phase-defining categories (Chomsky 2000). The expected “regular” cases of complementarity are accounted for by postulating as the relevant binding domain for both reflexive and pronominal the minimal phase containing the anaphoric element. Newly discovered cases of non-complementarity that exist at least for some speakers stem from the ability of reflexives, but not pronominals, to covertly raise to the edge of their phase (Safir 2004). Finally, well known cases of non-complementarity in AcIs where the anaphoric element is embedded in a PP can be explained if θ-independent (that is, internally saturated) PPs are adjoined high, to the AcI (vP)-phase-edge. This then achieves a unified account of both the well established binding facts documented in the literature and the surprising long-distance binding of reflexives reported as acceptable for some speakers.My thanks to Judith Aissen, Jorge Hankamer, and especially Jim McCloskey, who have helped me with and greatly contributed to the various drafts of this paper. Thanks also to my JGL reviewers for their helpful feedback. (Any remaining errors are my own.) Finally, I am indebted to all my native speaker consultants.

This article discusses methodological aspects of using Gothic orthography for the purposes of deducing phonological properties of the language. The vowel/glide alternation known as Sievers' law is used as an example. In order for scholars to propose an analysis of this alternation they must form a view of how syllable structure works in Gothic, and in order to do that they must interpret the orthography and determine to what extent it reflects phonology. It is primarily consonant clusters and word divisions that serve as orthographic sources for syllabification, beside alternation of i/j and u/w in related forms. A closer methodological look at some of the arguments that have been brought to bear on the issue of orthography and syllabification reveals that the strongest position that word division consistently reflects phonological syllabification cannot be upheld. Furthermore, it is shown that word division has little bearing on which word-initial clusters are possible in Gothic.I would like to thank Jonas Carlkvist, Rune Palm, Santeri Palviainen, Marc Pierce, and Rudolf Rydstedt for their help and willingness to discuss different aspects of this article. I am also very much indebted to the three anonymous JGL reviewers, whose comments have been of considerable help. All errors are mine.