Learning new words via reading: the influence of context and diversity

Thesis

From mid-childhood onwards, the majority of new words are incidentally acquired via independent reading. Numerous studies have shown that the linguistic contexts in which a new word appears influences the ease with which the word is processed and acquired. Some contextual characteristics (e.g., high topic familiarity) are more beneficial to learning than others. The focus of this thesis is a relatively new contextual characteristic, known as semantic diversity, defined as the degree to which a word’s surrounding contexts overlap in terms of semantic content. The overarching aim of this thesis is to investigate how semantic diversity influences the learning of novel words in young adults. This thesis comprises a total of four word-learning experiments and three computational simulations. They all contrasted two learning conditions: high semantic diversity (multi-themed) and low semantic diversity (single-themed). Both Experiment 1 and Simulation 1 demonstrated that novel words experienced multiple times in one familiar topic (i.e., low semantic diversity) were advantaged in initial learning, a finding attributed to what I refer to as “anchoring”, a process of securing novel words onto long-term memory. On the other hand, Simulation 2 and Experiment 2 showed that novel words that had an anchoring opportunity were better placed to take advantage to high semantic diversity, suggesting that anchoring may benefit initial but not later learning. Simulation 3 used word vectors to mathematically simulate how semantic diversity influences the learning of meanings across Experiments 1 and 2. The simulation results showed that high-diversity words tend to have sparser semantic spaces, potentially indicative of weaker integration into pre-existing knowledge. In order to consolidate the results from Experiment 2, Experiment 3 monitored participants’ eye movement during reading. The results suggest that novel words appearing in high-diversity contexts are encoded more strongly during reading and tend to have more context-independent representations. Finally, while Experiments 2 and 3 showed that experiencing a novel word multiple times in a repeated topic could serve to anchor, Experiment 4 demonstrated that an explicit definition could also achieve the same effect, enabling the novel words to be better placed to take advantage of high semantic diversity. Together, this research demonstrated that initially restricting a novel word’s semantic diversity (i.e., starting small) and capitalizing on learners’ prior knowledge can support anchoring, which can facilitate subsequent learning in more diverse contexts.

Authors: Mak, MHC

• DOI: 10.5287/ora-v0e5bdkdp
• Type of award: DPhil
• Level of award: Doctoral
• Awarding institution: University of Oxford
• Language: English
• Keywords: word learning; anchoring; semantic diversity; contextual variation; reading
• Subjects: Psychology