Verbs vary in their tendency to appear in different structures and previous
work shows that listeners learn new verb biases from exposure to new
co-occurrence statistics In 4 experiments, we explore the
context-specificity of this learning. We report novel results that 1)
listeners can learn verb-general, speaker-specific structural preferences
via exposure to speaker-structure co-occurrences; 2) listeners learn
verb-specific, speaker-general structural biases from exposure to 2
speakers with identical biases; 3) listeners do not appear to learn
simultaneous verb- and speaker-specific biases from exposure to 3-way
bindings of speaker, verb, and structure.