Much difficulty in the analysis of Latin constructions with dominant participles is caused by a typologically unusual pattern of subject agreement: instead of being assigned case according to their subject function (like, e.g., the accusative subjects of infinitives), subjects of participles appear in the same case as the participle itself. This study presents evidence against deriving dominant participle constructions from participles' attributive uses and argues that they should instead be analyzed as a syntactic nominalization in which the participle is the head that agrees with its subject in case. Our account relies on two separate components: (i) a clausal nominalization rule, introduced at the level of phrase structure, and closely resembling the structure of the English gerund; and (ii) a subject agreement rule, treated as a lexical property of Latin participles, and common to all their uses (not just the dominant use). Our synchronic account is supported by diachronic evidence, which suggests that the dominant participle construction developed in Latin through reanalysis of the absolute participle construction (shared by many ancient Indo-European languages). The change involved the introduction of a clausal nominalization rule, while the subject agreement rule was likely present from early on and not affected by the change.