Proceedings of the Annual Meeting of the Cognitive Science Society

During language acquisition, children must learn when togeneralize a pattern – applying it broadly and to new words(‘add –ed’ in English) – and when to restrict generalization,storing the pattern only with specific lexical items. One effortto quantify the conditions for generalization, the TolerancePrinciple, has been shown to accurately predict children’sgeneralizations in dozens of corpus-based studies. Thisprinciple hypothesizes that a general rule will be formedwhen it is computationally more efficient than storing lexicalforms individually. It is formalized as: a rule R will generalizeif the number of exceptions does not exceed the number ofwords in the category N divided by the natural log of N(N/lnN). Here we test the principle in an artificial language of9 nonsense nouns. As predicted, children exposed to 5 regularforms and 4 exceptions generalized, applying the regular formto 100% of novel test words. Children exposed to 3 regularforms and 6 exceptions did not extend the rule, even thoughthe token frequency of the regular form was still high in thiscondition. The Tolerance Principle thus appears to capture abasic principle of generalization in rule formation.