Studies of aphasics – people whose language faculty has been impaired due to stroke or other brain trauma – show that there must be a past tense rule that speakers use for regular forms – even very frequent ones – and that irregular forms are stored whole, probably in a different part of the brain. Badecker and Caramazza (1999) describe how we can know this.

Some aphasics display agrammatism; this means that they have difficulty in producing or processing function words in sentences, but can still produce and understand content words. Interestingly, agrammatic aphasics have difficulty producing or processing both regularly inflected forms (like the English past tenses), and also productively derived words (those with suffixes that we use frequently in making up new words – for example, -less as in shoeless or -ly as in darkly), whereas they have far less trouble with irregular forms like sang and flew.

Other aphasics display jargon aphasia; these aphasics produce fluent sentences using function words, but have trouble producing and understanding content words. Instead, they have a tendency to produce nonsense words. Interestingly, jargon aphasics will use regular inflections appropriately on their nonsense words, but they have difficulty processing and producing irregular forms.

We can explain the differential behavior of agrammatical and jargon aphasics if we postulate that we have rules for producing regularly inflected and productively derived forms, and only store irregular forms, and that rules and stored items are located in different parts of the brain. For agrammatic aphasics, the rule is unavailable, presumably because the part of the brain has been damaged that apparently allows us to apply morphological rules, but the irregular forms are still accessible from an undamaged part of the brain. For jargon aphasics, the irregular forms have been lost because the part of the brain that apparently allows access to stored forms has been damaged, but the regular rule is still intact.