Some (but not all) unattested determiners are unlearnable

A striking cross-linguistic generalisation about the semantics of determiners is that they never express nonconservative relations. To account for this one might hypothesise that the mechanisms underlying human language acquisition are not compatible with nonconservative determiner meanings. We stress that the crosslinguistic generalisation alone should not be taken as decisive evidence for this learnability hypothesis, by briefly discussing another determiner meaning that appears to be unattested but nonetheless learnable. We then report results from an experiment with fourand five-year-olds that support the learnability hypothesis more directly: children fail to learn a novel nonconservative determiner but succeed in learning a comparable conservative determiner.