During word learning, many words are associated with many meanings to build a lexicon. A model by McMurray et al. (2012) suggests that vocabulary acquisition may not only depend on building correct associations, but also pruning incorrect ones. Additional evidence for the importance of pruning comes from a word learning analog in pigeons, where the opportunity for pruning incorrect associations between objects and symbols was manipulated during training (Roembke et al., 2016). To investigate pruning in humans, we conducted six supervised word learning experiments. Participants were first trained to link two objects to each word, and subsequently were tested how quickly these were pruned. Across experiments, association strength was measured by using either eye movements to to-be pruned objects, or a post-training accuracy assessment. Learners showed rapid—though potentially not complete—pruning of incorrect associations, but this depended on whether the symbols were auditory words, orthographic words or non-linguistic symbols. Thus, this dissertation provides first evidence that pruning is operative during word learning. We also examined how newly learned words compete against known words for recognition using eye-tracking and found that despite very high accuracy these words were not strong competitors.