An analysis of the role that the ontology plays in determining verb meaning and semantic roles is presented. The discussion will contrast the current approaches to semantic role labeling using supervised machine learning methods to semantic approaches that rely on the ontology and verb sub-categorization. The presentation will discuss the issues dealing with the construction of verb meaning and semantic roles for general domains, and for specific domains with their own clearly defined ontologies. It will be indicated that Wordnet can serve as the general ontology on which specific ontologies can be anchored. The semantic-based algorithms will search for concepts beginning with the domain-specific ontology, and then continue up to the WordNet ontology if concepts are not found within the domain dependent ontology. In this way, the domain-specific ontology may override the general ontology.