Symbolic rules produce past tense forms (procedural memory)
look-up is quicker than rule application
rule application takes more time but is always done ‘on-the-fly’
Time 1: Every form is memorized (irregular and regular)
Time 2: child notices pattern: Verb root+ed = past
Child creates a rule
Child applies rule to all forms: overgeneralization
Time 3: Child realizes that there are irregular and regular forms
creates a dual system: irregular forms are retrieved from memory, regular forms are created by a rule
new and novel verbs will get regular endings in past tense unless exposed to irregular past
the fact that children seem to learn a rule = language must be symbolic
Traditional U-shaped learning predicts children won’t be able to create past-tense forms for novel verbs when they are in the initial stage. This isn’t consistent with data
Overregularization is not common
Cannot account for the presence of two past forms
e.g. dream/dreamed-dreamt, light/lit-lighted
In production experiments
Irregulars produced faster
Frequent irregulars are produced faster than infrequent irregulars (Prasada et al., 1990; Albright & Hayes 2003)
No difference between frequent and infrequent regulars
Maybe because experiments always present the root form? (1) This is a girl who knows how to dance. She did the same
Root presentation might mask differences due to frequency in regulars