How do representations of verbs and argument structure link with the real-time processing of verbs in sentences? In recent work, I have been exploring how verbs with different argument structure are processed in the disambiguating context of sentences, unlike much previous work which has studied verb processing in isolated, single-word presentations. In this talk I will focus on two types of argument structure alternations which involve "silent" elements at some level of theoretical representation - English lexical causative verbs (e.g. "The vase broke" vs. "The cat broke the vase") and English activity verbs which permit implicit (unpronounced but interpreted) objects (e.g., "The student read" vs. "The student read the book"). On the surface, the lexical causative verbs have some silent element that distinguishes the causative and non-causative variant, while the activity verbs allow for a silent object. There are many proposals for what the representational differences of each of these are in the theoretical literature (from semantic, morphological, and syntactic perspectives), which will be explored in the context of experimental results on the processing of these types of sentences. Ultimately, the methodological question is what we might be able to learn from behavioral processing studies about verb representation that goes beyond 'performance' factors like memory and lexical frequency, and also what the current limitations of this approach are. More specifically, I will argue that these alternations show different processing effects for the kind of silence that each involves, suggesting that there is a difference in their representation that goes beyond simple transitivity and frequency distinctions. In a series of self-paced reading studies, lexical causative verbs have shown greater processing cost in the transitive condition, while activity verbs show greater cost in intransitive conditions.