This study investigated the formulation of verb argument structure in Catalan- and Spanish-speaking children with specific language impairment (SLI) and typically developing age-matched controls. We compared how language production can be guided by conceptual factors, such as the organization of the entities participating in an event and knowledge regarding argument structure. Eleven children with SLI (aged 3;8 to 6;6) and eleven control children participated in an eye-tracking experiment in which participants had to describe events with different argument structure in the presence of visual scenes. Picture descriptions, latency time and eye movements were recorded and analyzed. The picture description results showed that the percentage of responses in which children with SLI substituted a non-target verb for the target verb was significantly different from that for the control group. Children with SLI made more omissions of obligatory arguments, especially of themes, as the verb argument complexity increased. Moreover, when the number of arguments of the verb increased, the children took more time to begin their descriptions, but no differences between groups were found. For verb type latency, all children were significantly faster to start describing one-argument events than two- and three-argument events. No differences in latency time were found between two- and three-argument events. There were no significant differences between the groups. Eye-movement showed that children with SLI looked less at the event zone than the age-matched controls during the first two seconds. These differences between the groups were significant for three-argument verbs, and only marginally significant for one- and two-argument verbs. Children with SLI also spent significantly less time looking at the theme zones than their age-matched controls. We suggest that both processing limitations and deficits in the semantic representation of verbs may play a role in these difficulties.