Electronic Thesis and Dissertation Repository

Thesis Format

Integrated Article


Master of Science




Minda, John Paul


This paper sought to explore cultural preferences for analytic and holistic thinking in classification. Experiment 1 paired the Shepard, Hovland, and Jenkins (SHJ) tasks with the Analysis-Holism scale (AHS) and a demographics questionnaire. Effects of culture on learning rates, alongside the feasibility of online data collection, were assessed. Learning difficulty differences among the six SHJ category sets were observed. Further, as predicted, higher holistic thinking correlated positively with the family resemblance task. Experiment 2 replicated the Norenzayan et al. (2002) task. Unlike in the original study, the effect of instructional condition was not significant across our full sample. Nevertheless, the non-Western sample showed higher holistic thinking in the similarity instruction condition. Moreover, our sample did not show any overwhelming preference for either analytic or holistic thinking strategies. Overall, our results are inconclusive, yet promising, and hint at some effect of culture on classification. This warrants further research in this domain.

Summary for Lay Audience

This paper sought to explore cultural preferences for analytic and holistic thinking in classification. The basic mechanisms of category learning are thought to be universal (Shepard, 1987), however, recent research has found that individuals from Eastern cultures tend to have a holistic processing style, whereas those in Western countries default to an analytic one (Nisbett et al., 2001; Norenzayan et al., 2002). Nevertheless, these results have not always been replicated (Murphy et al., 2017). The broad classification realm primarily includes category learning and categorization paradigms, with two main strategies widely debated: family resemblance and rule-based strategies. The family resemblance strategy uses a judgment of overall similarity, whereas rule-based responding is logical and based on a focal object with necessary and sufficient features. Experiment 1 paired the six classification sets first described by Shepard, Hovland, and Jenkins (SHJ; Shepard et al., 1961) with the Analysis-Holism scale (AHS; Choi et al., 2007) and a demographics questionnaire on an entirely online platform. These six category sets test the reliance on single feature rules, disjunctive rules, and family resemblance. Alongside testing the feasibility of online category learning data collection, the SHJ tasks were used to explore the effects of culture on the learning performance of the different category sets. Learning trends in the expected direction were observed, indicating variable learning difficulty across the SHJ sets. Also, a relationship between family resemblance responding and high holism on the AHS was observed, together with other interesting exploratory findings. Experiment 2 replicated the Norenzayan et al. (2002) task, which tested two instructional conditions. The similarity judgment condition was predicted to elicit higher family resemblance responses. Unlike the original study, our results did not show a significant effect of instructional condition. Further, only the non-Western sample showed a significant effect of condition, depicting higher holistic thinking in the similarity judgment condition as opposed to the classification condition. Moreover, neither thinking style was dominant across our sample. Overall, our results hint at cultural preferences, though the size and direction of the effect remain inconclusive. The findings motivate additional exploration of this domain by addressing limitations and future directions.