Although liking or affect was a central construct during the 1980's and 1990's in supervisor-subordinate dyadic research, a review of the literature suggests that since the turn of the century leader-member exchange (LMX) has emerged as the primary construct to assess relational quality. A question exists whether liking, as a reflection of interpersonal attraction, should continue to be included in organization research that focuses on dyadic relations and whether liking is redundant to LMX. To answer this question, we examined the incremental validity and construct redundancy of liking in relation to LMX based on meta-analyses. Results indicated that liking and LMX have similar patterns with common correlates with respect to effect sizes and significance. Moderator analysis revealed that reporting source of liking and LMX result in significantly higher correlations with antecedent and outcome variables when common reporting source data is used. Incremental variance analysis results demonstrated that liking can potentially explain additional variance over LMX for consequences. Overall, our results suggest that liking is an important and distinct construct that facilitates the development of LMX and therefore should not be abandoned in organizational research that examine supervisor-subordinate relational quality.