In this paper, we provide a comparative analysis of three constructions that occur with perception verbs like ver (“see”) in Spanish: infinitives, gerunds, and so‐called pseudorelative clauses. We analyze several phenomena regarding the transparency status of the embedded inflected or uninflected clause, such as the availability of long (i.e., matrix) passive, short (i.e., embedded) passive, the occurrence of modal verbs, embedded negation, and left‐peripheral fronting. Our data from acceptability judgment tasks and a Spanish corpus led us to postulate that, albeit their surface similarities, the three examined constructions have different syntactic structures, reflected by their structural integration and the phasal versus nonphasal status of the embedded clause: We argue that gerunds are secondary predicates, while pseudorelatives and infinitives are complements of the perception verb. With respect to the latter two configurations, pseudorelatives are defective CPs (and thus weak phases) that allow (finite) raising‐to‐object, but not raising‐to‐subject. Infinitives, on the other hand, show interspeaker variation with respect to their structural size as voiceP or TP, accounting for variable acceptability judgments.