Cross-linguistic universals (features present in all languages) can often be informative about learning biases. The most well-known universal in the domain of meaning is the generalization that all determiners (words like “every”, “some”, “no”, "most", “the”, “a”, and the like) have ‘conservative’ meanings. Intuitively, the observation is that in a sentence like "every/some/no/the fish swims", one doesn't need to look beyond the fish to determine whether the sentence is true. Put another way, those sentences are about the fish (and their properties); the larger class of swimming things is irrelevant. This might seem obvious, but it rules out many hypothetical 'non-conservative' determiners. For instance, no language has a word "equi" such that "equi fish swims" would mean “the fish and the swimmers are numerically equivalent” (here, both fish and swimming things matter, so "equi" fails to be conservative). This robust cross-linguistic generalization has long been thought to reflect a fact about the architecture of the language faculty, as opposed to general cognitive constraints, communicative pressures, or historical accident. If true, then non-conservative determiners are predicted to be unlearnable: human minds should be ill-equipped to pair non-conservative meanings with members of the syntactic category determiner. But evidence bearing out this bold prediction has proven elusive. With this in mind, I'll present a series of word learning experiments showing that adult participants fail to learn novel non-conservative meanings for nonce determiners, even when explicitly taught, but succeed at learning their conservative counterparts. And since conservativity is a property tied to a specific syntactic category, this effect disappears, as predicted, when the same non-conservative meanings are paired with verbal syntax instead of determiner syntax. I have recently begun to extend these findings to children. This body of results suggests that the conservativity universal is related to learnability, and supports theories on which this generalization reflects a deep fact about the human language faculty.