top of page

Beyond Saito's Threshold: The Conceptual Evolution of Hikikomori from Culture-bound Syndrome to Digitally-Mediated Withdrawal

Yuma Tamura
21/11/26

Hikikomori was first described by Saito (1998) as severe social withdrawal from education, employment, and relationships lasting more than six months. Hikikomori has developed from a phenomenon thought to be specific to Japan into an internationally recognised and contested concept. This review follows that development, looking at how hikikomori is defined, how it is classified diagnostically, its psychological characteristics, its causes, how it is treated, and how it has changed in a post-COVID, digitally mediated world. I drew on 38 sources to synthesise the literature thematically. Prevalence is estimated at around 1.2% in Japan (Koyama et al., 2010), and similar patterns have been identified in Hong Kong, South Korea and parts of Europe and North America (Kato et al., 2012; Wong et al., 2015; Eckardt, 2023). However, the literature surrounding hikikomori is fragmented. The six month criterion has little empirical justification and the distinction between primary and secondary hikikomori is weak due to widespread subthreshold distress. This supports dimensional models rather than categorical ones (Kato, Kanba & Teo, 2019). The causes of hikikomori are multifaceted, arising from the interaction of family dynamics, sociocultural expectations, structural economic conditions and individual vulnerability. No single treatment has proved to be consistently effective, but multidisciplinary approaches seem to be the most promising (Kato, Kanba & Teo, 2019). Recent research suggests that the internet acts as a contextual amplifier of withdrawal, not a direct cause of it (Esposito & Addeo, 2025; Calaresi et al., 2025). Ideas of “semi-withdrawal” also challenge the already-established diagnostic frameworks. This review concludes that hikikomori is best understood as a prolonged state of social non-participation. Future research should aim to reach a diagnostic consensus, develop assessment methods less reliant on self-report systems, and use dynamic models suited to the digital world.

bottom of page