A study was conducted to determine whether children give evidence of using any of six comprehension strategies and whether children of same and different ages use different strategies. It was studied how comprehension performance can best be predicted by other facts about the child, including his language and his language input. The six comprehension strategies are: child as agent, probable event, animate noun as agent, relative object size, position bias, and word order. Of particular interest is whether the "animate agent" strategy can be shown to exist rather than the competing explanations of "relative object" size preference or, for some sentences, a judgment that the animate toy as agent is simply the more "probable event." Twenty-five children, eight aged 24-25 months, 10 aged 30-32 months, and seven aged 36-44 months, were studied in two to five 30-minute testing sessions in a small playroom with the mother present. Free speech samples for mother and child during free play, and the child's comprehension of the noun vocabulary to be used in a test with 24 toys were evaluated. The child's comprehension of 36 semantically reversible sentences was then tested by telling him "do what I say" and presenting the appropriate two toys as a sentence was spoken. The study did not find evidence for the "child as agent" strategy or the "relative object size" strategy. Word order strategies emerged as the first linguistic generalizations to be systematically applied across all the sentences of the study. (SW)