Valence and concreteness effects in word-learning: Evidence from a language learning app

Abstract

One goal of applied linguistics is to learn languages better and faster. Second language (L2) learners need to acquire large vocabularies to approach native-like proficiency in their targeted language. A number of studies have explored the factors that facilitate and hinder word learning using highly controlled experiments, however, these lack ecological validity and the findings may not generalize to real-world learning. The studies in this thesis respond to this gap in the literature. The studies leverage big data from a popular language learning app called Lingvist to explore how understudied semantic factors such as valence (positivity/negativity) and concreteness impact adult L2 word learning. Chapter 2 explores the shape of valence effects on learning, the interaction between the semantics of the target word and the linguistic context in which the word is learned, and how these effects unfold over multiple exposures to the target word. Users learn both positive and negative words better than neutral ones, and learning improves by 7% when target words appear in emotionally congruent contexts (i.e., positive words in positive sentences, negative words in negative sentences). These effects are strongest on the learner’s second encounter with the word and diminish over subsequent encounters. Chapter 3 examines the interaction between target word valence and concreteness. Increased positivity increased accuracy for concrete words by up to 13%, but had little impact on learning abstract words. On the theoretical front, findings provide support for embodied cognition, the lexical quality hypothesis, and the multimodal induction hypothesis. On the applied front, they indicate that context valence can be manipulated to facilitate learning and identify which words will be most difficult to learn.