[PS-3.26] Entrenchment and push chains in learning form-meaning mappings

Harmon, Z. & Kapatsinski, V.

University of Oregon

We examined competition between forms for a meaning in comprehension. 70 adult participants experienced a miniature artificial language. Two suffixes were always mapped onto plural non-diminutive meaning (multiple large creatures), and two other suffixes were mapped onto diminutive singular meaning (one small creature). Plural diminutives (multiple small creatures) were never shown in training. For half the participants, one plural suffix was three times more frequent than other suffixes during training. For the other participants, one diminutive suffix was three times more frequent than the others. At test, participants would hear a suffixed form and click on a meaning. Suffixes were almost never mapped onto the wrong familiar meaning. Infrequent suffixes without frequent synonyms were mapped onto the novel, plural diminutive, meaning as often as onto the correct familiar meaning. However, frequent suffixes were not mapped onto the novel meaning, demonstrating entrenchment. At the same time, semantic competitors of frequent suffixes were preferentially mapped onto the novel meaning, as if a frequent form pushes its competitor out of the shared meaning with which both were paired (a semantic push chain). Using the Naïve Discriminative Learner model (Baayen et al., 2011), we show that these results require mechanisms beyond simple associative learning.