The goal of the current study is to shed some light on a debate in theoretical Linguistics concerning the relationship between the lexicon and syntax, by means of an experimental approach. We start from the assumption that Lexicalist (Pustejovsky, 1991) and Exosqueletal (Borer, 2005) accounts make different predictions regarding the processing of lexical predicates. According to the Lexicalist view, the various contexts in which a verbal predicate may appear are listed in its lexical entry. Accordingly, the more syntactic contexts are compatible with a given predicate, the richer its lexical representation will be, and thus the higher its processing cost. Hence, alternating verbs like molestar (bother) might be interpreted as polysemous, unlike non-alternating verbs like doler (ache). In contrast, the Exosqueletal Hypothesis assumes that lexical units may appear in any syntactic context, and predicts that the fewer the syntactic contexts compatible with a given predicate (as in doler), the greater its idiosyncratic information, which will result in an increase in processing costs. Based on these opposing predictions, we carried out two comprehension experiments with written (Experiment 1) or spoken (Experiment 2) sentences. Experiment 1 used a self-paced reading task, whilst Experiment 2 used a divided attention paradigm in which reaction times to a distractor acoustic signal were recorded during comprehension. The results of both experiments show a greater processing cost for non-alternating verbs, as predicted by the Exosqueletal Hypothesis. Thus, it appears that the more flexible the predicate, the lesser its processing costs.