that,universally, NPs are permittedinpositionswheretheycanbecasemarked.Inorder for thisproposaltobecarried
throughconsistently, it is necessary toclaimthatEnglish toois relentlesslycasemarked, eventhoughitmakes noovert
case distinctions except on pronouns (I/me, she/her, etc.). And even pronouns fail to show an accusative-dative case
distinction, needed to distinguish direct fro mindirect objects. Thus, goes the argu ment, children learning English
acquire the case syste meventhough thereis virtuallynothing in the linguisticinput that tells the mabout it. Therefore,
it is concluded, case marking must be part of UG.
One mightfind it objectionable that this solution attributes a great deal of invisible structure to English, yet gets this
structure in there via innate f-knowledge. In order to answer the argument, though, it is not enough to pronounce it
absurd. Twotasks are incumbentonthecritic.First, an alternativeaccount ofthesyntacticpositioning ofNPs must be
offered, one that is either learnable or else based on other plausible principles of UG. Preferably, this account should
also be shown to be more empirically adequate in dealing with the distribution of NPs cross-linguistically. Second, an
account must be offered of case marking that shows howitis learnable in the languages that have it. My impression is
thatcasesystems (Blake1994) cross-linguisticallyshow patternsreminiscentofthecross-linguisticdistribution ofcolor
names (Berlinand Kay 1969). Thissuggests thatthere are at leastsome innatebiases regarding case systems thatmake
the mco me out as they do. Thus at best the critic might be able to conclude (a) that English speakers do not have an
invisible case system, but that the distribution of NPs is determined by some partly innate linguistic principles more
“natural”than case; and (b) that the acquisition of case in those languages that have it is guided in part by innate case
principles that help the learner structure the primary linguistic input. In other words, perhaps one could arrive at a
more input-driven model of the language learner with respect to these phenomena, but it is unlikely that the
contribution of UG will go away altogether.
4.7 Poverty of the stimulus in word learning
Thediscussion so far has been couchedlargelyinterms oftheacquisitionofgrammar, following Chomsky's emphasis.
But learning the lexicon poses if anything a far vaster problem. Suppose a grammar has as many as a few hundred
rulesinit,and comparethistotheestimatethat an averagespeakerknows ontheorder ofseveraltensofthousands of
words (this includes passive as well as active vocabulary, since after all one must understand as well as speak the
language). By a simple arithmetic calculation, Carey (1978) observes that a child