This dissertation aims to analyze how the interface between linguistic and visual information occurs based on language comprehension experiments. Active and passive voice sentences, as well as perspective predicates were analyzed using a sentence-image verification technique. The position of the characters representing the subject were manipulated (left-right orientation), and so were their theme roles (agent/source vs. patient/target) by means of attention manipulation. The proposi-tional theory, a relation between the faculty of language in the broad sense (Hauser, Chomsky and Fitch, 2002) and the modularity of mind theory (Fodor, 1983) were presented so as to explain how the interface between linguistic and visual information occurs. Our experiments were carried out with Portuguese speakers, and results indicate that during sentence-image mapping, left-right orientation does not seem to be a relevant factor, except for more complex structures, such as perspec-tive predicates where the subject s theme role is not prototypical. As for theme roles, the attention manipulation on the subject character seems to facilitate lan-guage processing. As for sentence types, active voice sentences seem to be pro-cessed more easily than passive ones. Perspective predicates whose subjects take a passive/target role seem to demand higher processing costs, possibly because of lexical access and representation.