This study reports the identification of a type of supervision not currently defined in organizational psychology/behavior literature: ambivalent supervision. Ambivalent supervision is characterized by a person in a supervisory position who displays both supportive and unsupportive behaviors toward his/her subordinates. In current leadership research, destructive leadership and constructive leadership exist on opposite sides of a single continuum. Although this approach appears to be sound, it leaves no room for the co-existence of supportive and unsupportive leadership behaviors. This type of inconsistent leadership behavior has not yet been specifically identified in the leadership literature, nor has it been linked to the social psychology literature on ambivalent personal relationships. Therefore, the first goal of this study is to review the social psychology literature on ambivalent personal relationships and extend it to the workplace by defining ambivalent supervision. The second goal of the study is the creation and validation of a measure of ambivalent supervision. A direct measure of ambivalent supervision will add to the literature by creating a more efficient way to directly identify this type of leadership for research and for practitioner use. The first step for measurement development is item generation. For item generation, I considered content, breadth, and alignment of items with the definition of ambivalent supervision. I then held brainstorming sessions with students and professors to ensure the conceptual meaning of ambivalent supervision was captured. The second step is questionnaire administration. After gathering data from Facebook participants, I examined the new measure's factor structure. This was done using principal axis exploratory factor analysis with varimax rotation. I used item-total correlations to calculate internal consistency. Items with weaker correlations were removed to create a more parsimonious model. I then gathered data from MTurk participants. Confirmatory factor analyses were run to check the goodness-of-fit and unidimensionality of the measure. The final steps included testing the measure for convergent validity, discriminant validity, and criterion-related validity. A parsimonious four-item measure of ambivalent supervision was created and confirmed by strong fit indices. Support was found for convergent, discriminant, and criterion-related validity of the measure. Ambivalent supervision was found to have a positive relationship with stress, counterproductive workplace behavior, anxiety, and somatic complaints. Additionally, a number of regressions were run to confirm that ambivalent supervision leads to negative work outcomes. The interaction of abusive supervision and perceived supervisor support significantly predicted anxiety, somatization, and counterproductive work behavior. When graphed, it was observed that outcomes (anxiety, somatization, and counterproductive work behavior) were equivalent except when both abusive supervision and perceived supervisor support were high (i.e., ambivalent supervision). When abusive supervision and perceived supervisor support were high, negative outcomes for employees were at their highest. This provides preliminary evidence that ambivalent supervision has unique and severe outcomes for employees. In conclusion, in this study, I identified a new form of leadership, created and validated a measure, and linked this measure to expected outcomes. Further research is needed to confirm these results across different samples and possibly identify additional outcomes and interventions for this type of detrimental leadership.