This article provides empirical evidence for two hypotheses in the a ordance literature. First, by leveraging a small a ordance change -Twitter increasing its character limit from 140 to 280 on November 7, 2017, employing an instrumental variable approach, and examining 143,771 original tweets published by organizational and leadership accounts half a year before and after the intervention, we showed the direct causal relationship between a ordances and communication behavior on digital media platforms. Second, by exploring what factors could explain the heterogeneity of causal e ects, we showed that previous endogenous perceptions of communication constraint predicted later behavioral changes, despite the same exogenous intervention. These ndings highlight the role of human agencies in the face of technological changes and provide empirical support for the a ordance approach to information communication technologies (ICTs) as a reconciliation between technological determinism and social constructivism.