Abstract This paper presents a reevaluation of the choice between the two analyses for German partial fronting phenomena proposed in the literature, remnant movement and reanalysis. We show that the empirical arguments which were presented in favor of an extraction analysis are not convincing, and we provide empirical evidence supporting a reanalysis-like approach. Turning to a detailed data discussion, we compare three different kinds of partial constituents: verbal, adjectival, and nominal ones. Adjectival complements pattern with coherently selected verbal complements whereas nominal complements turn out to be less restricted. On the theoretical side, we show that a reanalysis-like theory can be given a formally precise rendering in the HPSG architecture in terms of a lexical argument-raising specification, which is already widely employed in HPSG analyses of coherence in Germanic and restructuring verbs in Romance languages. The account we propose generalizes previous HPSG approaches to partial complements of different categories and correctly predicts the interaction of (partial) VP topicalization with embedded partial NPs or APs.