Background: A number of meta-analyses of mindfulness have been performed, but few distinguished between different facets of mindfulness, despite it being known that facets of mindfulness behave differently in different populations; and most studied the outcome of interventions, which tend to involve additional ingredients besides mindfulness. Furthermore, there has recently been some concern regarding possible publication bias in mindfulness research.Objective: Systematic review and meta-analysis of the relationship of different facets of mindfulness with various outcomes, taking into account possible moderators, and controlling for publication bias using a method appropriate given the substantial heterogeneity present.Methods: Random effects meta-analysis with a number of robustness checks and estimation of the possible impact of publication bias on the results. Included are all studies that report correlations of outcomes with all five FFMQ facets, in English, French, German, or Spanish.Study Registration: PROSPERO International prospective register of systematic reviews http://www.crd.york.ac.uk/PROSPERO/display_record.asp?ID=CRD42016041863.Results: For the designated primary measure (SWLS) estimated correlations were: 0.15 [0.07, 0.22] for the Observing facet, 0.31 [0.27, 0.36] for Describing, 0.35 [0.31, 0.38] for Acting-with-Awareness, 0.30 [0.10, 0.47] for Non-judging and 0.28 [0.18, 0.37] for Non-reacting. Grouping all desirable outcomes together, Describing has the highest zero-order (though not partial) correlation; Non-judging the highest effect on avoiding undesirable outcomes. Results seem to be reasonably robust even to severe publication bias.