This study advances the conceptualization of perceived environmental uncertainty (PEU) and examines the impact of source heterogeneity of PEU on the subsidiary-level divestments of emerging markets multinational enterprises (EMNEs). In Part I, we justify why the existing conceptual and empirical understanding of PEU is incomplete and advance the PEU conceptualization by employing a novel machine learning technique and propose the dynamic dimensionality framework of PEU using textual data from the annual reports regarding the future and uncertainty the firm is perceiving directly. The PEU model focuses on the collective perception of members within an organization on their environment, since PEU is a function of both the perceiver and the environment. Drawing from the PEU model while integrating and upper echelon theory, we apply the dynamic dimensionality framework of PEU in Part II. We argue that when source heterogeneity of PEU is low, the parent EMNE is more likely to undertake foreign subsidiary divestment actions, and a decrease in the source heterogeneity of PEU will lead to an associated increase in the extent of foreign subsidiary divestment actions. Furthermore, as a boundary condition, when the CEO has a technology background, the impact of source heterogeneity of PEU on foreign subsidiary divestment actions will be weaken, for both likelihood and degree. Our final sample include 1,602 firm-year observations of public listed Chinese firms over eleven years (2005-2016), supporting all hypotheses.