Regional Differences in Affective Expression on Social Media: A Comparative Study of Urban and Rural Twitter Users in Japan

<title>Abstract</title> <p>Affective expression in social-media text is known to vary with the topic being discussed, and, separately, a range of unrelated behaviors are known to vary by region within Japan; whether affective expression itself differs regionally, and whether any such difference depends on topic, has not previously been examined together. The present study compared posts from 100 Twitter users in central Tokyo (the 23 special wards) with 100 users in Tokushima City, within two topic domains chosen for their contrasting nationwide popularity: anime (high interest) and national politics (low interest). Users' topic domains were identified with a field-association-word technique, and affective expressions were extracted from their posts using a 61-category, three-axis (affective, emotional, evaluative) sensibility dictionary combined with morphological and dependency analysis. Affective-expression intensity was higher among Tokyo users than Tokushima users in both the anime domain (median, Mdn = 0.253 vs. 0.188), Wilcoxon rank-sum test, p = .00035, r = .25, and the politics domain (Mdn = 0.192 vs. 0.085), p = .0001, r = .28. The balance of positive and negative expressions also differed by region in both domains, with Tokyo users producing proportionally more negative affect. A parallel comparison showed that Tokyo users used standard Japanese more, and dialect less, than Tokushima users, consistent with earlier sociolinguistic work, but the two regions did not differ in general topical preference or in smiling-emoticon use. Taken together, the results indicate that urban–rural differences in social-media affect track topic domain rather than holding uniformly across topics.</p>