Purpose: The purpose of this paper is to reexamine exploration-exploitation's reciprocality in organizational ambidexterity (OA) research. OA figures prominently in a variety of organization science phenomena. Introduced as a two-stage model for innovation, theory specifies reciprocal reinforcement between the OA processes of exploration (eR) and exploitation (eT). In this study, the authors argue that previous analyses of OA necessarily neglect this reciprocality in favor of conceptualizations that conform to common statistical techniques. Design/methodology/approach: The authors propose joint-variance (JV) as a soluble estimator of exploration-exploitation (eR-eT) reciprocality. An updated systematic literature synthesis yielded K = 50 studies (53 independent samples, N = 11,743) for further testing. Findings: Three primary findings are as follows--JV reduced negative confounding, explaining 45 per cent of between-study variance. JV quantified the positive confounding in separate meta-analytic estimates of eR and eT on performance because of double-counting (37.6 per cent), and substantive application of JV to hypothesis testing supported OA theoretical predictions. Research limitations/implications: The authors discuss practical consideration for eR-eT reciprocality, as well as theoretical contributions for cohering the OA empirical literature. Practical implications: The authors discuss design limitations and JV measurement extensions for the future. Social implications: Learning in OA literature has been neglected or underestimated. Originality/value: Because reciprocality is theorized, yet absent in current models, existing results represent confounded or biased evidence of the OA's effect on firm performance. Subsequently, the authors propose JV as a soluble estimator of eR-eT learning modes.