A five-year-old boy whose native language is American English answers his mother and uses two forms of the past tense in consecutive sentences that exemplify productive rules of verb morphology but happen to be wrong, viz. *holded (hold) and *bended (bend; but cf. on bended knee). Eventually, of course, typically after having been corrected, he will learn that the correct preterits are held and bent, respectively, and are part of the class of verbs with irregular past-tense forms (the so-called strong verbs).
The bulk of a language’s morphology conforms to rules that determine the productive sector of its structure, and unproductive rules (like the change of root vowel in English accompanied by the suffixation of a desinence in the preterit) tend to disappear with time. Children understandably apply the productive rules first when learning their native language and only later acquire a mastery of the unproductive sector.
An interesting question of linguistic theory is why unproductive forms (= exceptions) perdure in every language despite the general tendency to whittle away exceptions and replace them by productive ones. Some unproductive sectors of the morphology are large enough (like the English strong verbs) to constitute a distinct class of exceptions with its own localized raison d’être, although it may be difficult to define. Others are isolated enough to drop out of the language with time, although they persist in the speech of those who––perhaps unconsciously––use them to define their linguistic identity in terms of superiority to speakers who are ignorant of the traditional norm. Eventually, of course, the norm changes with the death of those who adhere to it, and the productive rules inevitably triumph, solidifying the new norm.
Even within unproductive sectors of the vocabulary there may occur changes toward the elimination of certain forms. Thus, to continue with the English preterit, not all vowel alternations in strong verbs are being sustained in contemporary speech and writing. Instead of the normative shrank and stank, for instance, one increasingly observes the substitution of the past participle form shrunk and stunk for shrank and stank. In the long run, given the strength of this shift, the traditional forms will wither away and disappear. Meanwhile, those speakers who adhere to the older norm will thereby define themselves linguistically and culturally vis-à-vis the increasingly larger group who use only the newer forms while recognizing the extancy of what used to be their only correct counterparts.