Aim: Theoretical models of the structure and functioning of adult working memory form the basis for measuring this capacity in research and applied settings. These models conceptualize its structure as varying from a single unitary process to multicomponent structures of up to four components. The latter differentiate between storage and processing capacities and/or between domain (verbal or visuospatial/nonverbal). These models were formulated from young adult (typically student) samples from Western, educated, industrialized, rich and developed (WEIRD), minority world contexts (wealthier regions, constituting a small percentage of the world population). Debates have not considered the cross-cultural applicability of these models and whether they are adequate conceptualizations of working memory functioning in young adults from less WEIRD, majority world contexts (developing countries whose populations form a significant demographic majority). This study sought to explore this issue. Methods: We compared the statistical fit of five leading working memory models using data from a sample of young South African adults (n=162; Mage = 20.41, SDage = 1.82; 38% female) who were multilingual, from predominantly low socioeconomic, rural, contexts. Results: Confirmatory factor analyses showed that a four-component structure composed of domain-specific (verbal, visuospatial) storage and processing provided the best fit for the data. This differs from the prevailing multicomponent model of working memory that dominates research and clinical practice. Conclusions: Research premises, neuropsychological tests and rehabilitation approaches are based on leading models of working memory, and so their cross-cultural appropriateness is important. We discuss how these findings could be incorporated into research and practice.