| Autor: |
Smolek, Amy, Kapatsinski, Vsevolod |
| Zdroj: |
Morphology; Sep2023, Vol. 33 Issue 3, p287-334, 48p |
| Abstrakt: |
Morphological paradigms are linked together by relations of paradigmatic predictability (implicative relations): when a speaker knows one form of a word, one can predict other forms of the same word. The present paper addresses the question of how implicative relations are learned. We suggest that implicative relations are learned via error-driven predictive learning, a process that crucially relies on instances in which morphologically related forms are close together in time. Previous corpus research has shown that temporal proximity is characteristic of morphologically-related forms. We show that temporal contiguity between such forms, an extreme form of temporal proximity, improves learners' ability to acquire implicative relations between these forms. We then show that a simple predictive error-driven learning model (Rescorla & Wagner, 1972) accounts for the full pattern of results very well. The model is further shown to account for the effects of contiguity only if contiguity highlights (increases the salience of) the differences and similarities between corresponding forms, providing a mechanistic account of how contiguity helps acquire paradigmatic mappings. In contrast, the influence of contiguity on increasing salience of the cues in the base as predictors of the derived form is relatively minor. The proposed model provides a mechanistic account for how implicative paradigmatic relations could be acquired from occasionally experiencing corresponding forms in close proximity. [ABSTRACT FROM AUTHOR] |
| Databáze: |
Complementary Index |
| Externí odkaz: |
|
Full text from SpringerLink