Abstract For nearly 60 years, researchers and practitioners have struggled toward agreement on the definition and of self-esteem. Both consensus and precision have proven elusive, and debate about what we are or should be measuring with our instruments continues today. In this article, we offer a clarifying account of the nature of self-esteem as a key aspect of personal identity and examine its legitimacy as a construct. The distinction between implicit and explicit self-esteem is discussed in this context, raising critical questions about the theoretical status of the former. Any discussion of the of a test presupposes a definite quality to be measured. An easy separation of from definition, however, is possible only when the quality reduces to purely operational terms that are rarely contested (e.g., distance, conductivity, blood pressure). Moreover, the routine measurement of even such qualities is intimately tied to supporting networks of theoretical and empirical relations that ultimately blur the distinction between defining the quality and measuring it (Cliff, 1982). This interdependence becomes all the more clear for psychological qualities. Here, the question of whether a test is working as intended usually cannot be answered by referring to a single operationally defined criterion. Rather, one is forced to take up the more fundamental task of trying to determine just what the test is measuring and whether this quality or qualities are consistent with the claims of the designer. This is accomplished by building around the test a of confirmed relations to pertinent constructs and related measures (Cronbach & Meehl, 1955). The of the test is reflected in the extent to which the network is logically consistent with the working definition of the quality that is the target of and the inferences that follow from this working definition. Thus, construct validity supports both the definition of the quality and one's success in measuring it. A test that fails to demonstrate a crucial relation posited in its nomological net may reflect either poor or misunderstanding of the quality being measured. This presents a dilemma for the researcher who is confronted with a critical disconfirmation. How does one proceed? Should the test be improved? The quality redefined? Both? The of self-esteem over the past half-century has been marked by this programmatic dilemma and the dynamics of uncertainty and change are no less prominent today. In the hope of providing a useful starting point for making sense of recent developments in this continuing evolution - especially the current emphasis on implicit versus explicit - we offer a brief conceptual analysis. Self-Esteem as a Hypothetical Variable We begin by specifying the kind of quality that self-esteem represents. A long-standing distinction proposed by MacCorquordale and Meehl (1948) can be applied here. On one hand, there are psychological that amount to nothing more than the quantitative or qualitative relations of more primary observable entities. They contain no surplus content and are merely formal or informal abstractions from empirical observations or relations. These intervening variables are exemplified by dispositional concepts such as plasticity, dominance, and reactivity. Clearly, these are not entities, processes, or events of any kind. They are summary abstractions of behavioural patterns. On the other hand, there are psychological that entail the existence of entities, processes, or events that are not directly observable but are posited to explain that which is observable. Often these hypothetical variables begin as metaphors needed to fill theoretical gaps. Over time, many become reified into self-standing concepts with existential content. Historical examples in psychology are engram, libido, and iconic store. …