It seems perfectly obvious that both pre-linguistic children and household pets behave with respect to perceived events in ways indicative that they ‘know’ that  DADDY (agent) THREW THE BALL (object), the reverse, that THE BALL (instrument) BROKE THE WINDOW (object) and not the reverse, that THE NEW BALL IS IN THE PAPER BAG (locative), and so on ad infinitum through the busy days of child and puppyhood. But a demonstration of this obvious may be in order. In column (2) of Table 1 are given my assignments of Fillmore’s case relations of all entities to the actions or states in all 32 perceptual events. According to Fillmore (1968), the Agentive case (A) is the animate, responsible source of the action (here, THE MAN) ; the Instrumental case (I) is the inanimate force or object contributing to the action or state (here the various BALLS) ; the Dative case (D) is the animate being affected by the action or state (not applicable here); the Locative case (L) is that which identifies the location or spatial orientation of the action (here THE TABLE, THE PLATE) ; the Objective case (O) is any entity whose rule in the action or state is identified by the action or state (I am not sure I understand this one, but I have coded all ‘ acted upon ’ entities this way); and the Resultive case (R) is the entity resulting from the action (not applicable here).

For the three cases with which I felt I could deal confidently (Agentive, Instrumental and Locative), the entities involved were compared with all entities not assigned this case relation to see if, in the sentences produced, the corresponding noun phrases did display that case relation. For example, in events where the blue ball is the initiator of action (. . .and the blue ball hits the orange ball) it is described in the Instrumental case as compared with events where it is the recipient of action (the black ball hits the blue ball). Table 4 presents the results. We note that in 84 % of the sentences describing events where the man is the actor, we get the man verbed, whereas in only 8 % of the sentences describing events where the man is not the actor does this happen - and these are all instances where the presupposed Osgood is indicated as the unseen agent! Similar results obtain for the Instrumental and Locative cases. The reason for the relatively high frequency of locatives in sentences for events assumed not to include this relation (64 %) is that speakers often included such information gratis (e.g., for #9, the man is holding a blue ball in his hand). The obvious effect of perceptual ‘ case ’ upon sentential case seems to be confirmed.