This paper identifies two flaws in a Subset Principle-based learning strategy for learning the raising/control distinction. A logical flaw is found in the presence of verbs that are ambiguous between being raising and control (e.g., begin). An empirical flaw is revealed in the fact that children do not misanalyze raising verbs as control verbs, as the Subset Principle-based strategy would predict. The author offers an alternative learning strategy based on the use of multiple cues from the distribution of raising and control verbs. She proposes that these cues should be used probabilistically to distinguish the two verb classes.