Native English Speakers Wanted for Online Word Relatedness Study

Background

Existing semantic network theories have outlined the process by which various concepts interact in our brains. Spreading Activation Theory as proposed by Collins and Loftus (1975) states that word meaning involves the activation of direct and indirect paths connecting nodes in a semantic network where the nodes represent words and the links represent semantic relations. Empirically, several studies have showed that different types of semantic knowledge such as taxonomic knowledge, thematic knowledge and featural knowledge play a distinct role across different tasks. However, further research is required to uncover the role played by these knowledge types and their relative importance. This study aimed to address this question by using word association data and annotate it with edge labels that map onto a semantic role. This graph allows us to determine the relative contributions of different types of semantic knowledge in a forced-choice triad similarity judgment task.

Research Questions / Hypotheses

This research aimed to investigate how manipulating the instruction of the similarity judgment task to prioritize certain types of semantic knowledge would affect one’s similarity judgment choices. We hypothesized that judgment choices would be modulated by the instructions given. It also aimed to explore how other factors, such as the concreteness and semantic distance of the stimuli, affect the similarity judgments made by participants.

Participants

337 participants participated in the study. Of those, 256 remained for analysis after exclusion criteria were applied.

Methods

The participants were allocated to three conditions based on knowledge type (taxonomic, thematic, and general relatedness) and were given instructions accordingly. They were presented with three words on the screen, one target and two alternative choices. The task was to choose which of the two words related more strongly to the target word. The task took approximately 30 minutes to complete.

Results

Correlation analysis showed that taxonomic relation was the strongest semantic relation across all instruction conditions. A follow-up analysis showed that the relative contribution of semantic roles depends on the concreteness of the stimuli. On the other hand, a significant effect was not observed between instruction manipulation and the relative use of semantic knowledge. Moreover, the relative contribution of different semantic roles did not differ across items that were semantically close or semantically distant.

Implications

The current study allowed us to evaluate existing semantic network theories and contribute to the body of research on the relative importance of different types of semantic relations in similarity judgment. In addition, due to its exploratory nature, it identifies some interesting areas for future research in this field.