How do multiple meanings affect word learning and remapping?

Many words have multiple meanings. It is currently unclear how a word with more than one meaning is acquired or stored long-term. One possibility is that each word's meaning is less robustly encoded; this, in turn, may make it easier to acquire additional meanings. We investigated this hypothesis across three cross-situational statistical learning experiments (N = 62 (in-lab), N = 61 (online), N = 60 (online)), using the same general procedure: During Learning Phase 1 (LP1), participants acquired both 1:1 (each word has one meaning) and 1:2 (each word has two meanings) mappings. In Learning Phase 2 (LP2), each word received one new meaning. Across experiments, we manipulated the frequency of 1:1 and 1:2 mappings. As expected, 1:2 mappings were always harder to acquire in LP1. However, in the LP2 of Experiment 1, former 1:1 mappings were remapped more easily than former 1:2 ones, while 1:2 mappings were remapped more easily in Experiments 2 and 3. These data suggest that words with multiple meanings are more easily remapped, indicating that each meaning is less robustly associated with the word, which in turn may facilitate the acquisition of an additional meaning. However, the observed inconsistencies in the remapping results between in-lab and online experiments suggest that CSWL may differ across these two settings.