Fifteen years ago now I conducted an experiment with a 2-year-old.1 At the request of a colleague in psychology who wanted to see whether it was possible to replicate a study of elicited imitation published five years earlier (Slobin and Welsh, 1973), I spent a couple of hours playing a game called ‘say what I say’. The subject-my daughter-seemed to find the game great fun, and eventually we wound up with quite a lot of audio tape, which demonstrated beyond any reasonable doubt that although Kate could repeat pretty long utterances-17 to 20-word sentences, for example-virtually verbatim, she could not repeat even close to verbatim any sentence with a relative clause in it. Experimenter: Can you say, ‘The kitty in the water is sad’? Kate: [patiently explaining] I already did. Experimenter: Ohh… I guess you already did. That’s right, [pause] What’s the puppy doing? Kate: Well, he’s reaching those orange. Experimenter: He’s reaching those orange. And what happens to the oranges? Kate: They fell down there. Experimenter: And is the puppy happy? Kate: Yes. Experimenter: [incredulity] He’s happy!?! Kate: But he’s sad there. Experimenter: Can you say, ‘The puppy who spilled the oranges is sad’? Kate: I already did. Experimenter: No, you didn’t. Kate: The puppy, mmm, mmm, the puppy is, mmm, is fall that and, uh, the, uh, the oranges is fell

down and and the basket there spilled all the oranges… [pause 5.5 seconds] Experimenter: [under his breath] I’ll be darned.