Research in emotional labor has grown exponentially in the past three decades with the expansion of the service sector in the South Korean economy. Although many studies have been concerned with emotional labor strategies (surface acting and deep acting), the empirical results regarding the consequences on service employees' job stress are still inconclusive. We point out that these mixed results are due to differing perspectives on emotional labor that have been used as focal lens in various research. These perspectives have highlighted quite different aspects of emotional labor and have provided competing predictions on the psychological effects of surface acting and deep acting. In detail, occupational requirement perspectives and emotional display perspectives argue that both surface and deep acting are identical in that both are rule-abiding but forced behaviors in the workplace. From the intra-psychic perspective, however, deep acting is not detrimental to employees' job stress, because it reduces emotional dissonance unlike surface acting. To reconcile the theoretical controversy and more accurately gauge the effects of two emotional labor strategies, this article hypothesizes the non-linear relations between deep acting and two outcome variables: emotional exhaustion and personal accomplishment. Surface acting is hypothesized to have linear relations with outcome variables given that all three perspectives agree on the effects and previous studies have shown relatively consistent results. In contrast, deep acting is expected to show U-shaped relations with emotional exhaustion and inverted U-shaped relations with personal accomplishment based on three lenses of emotional labor. In addition, for further clarification in investigating the effects of deep acting, we examine whether occupational identification moderates the non-linear relations between deep acting and outcome variables. This study used a convenience sample of 211 service employees who work in retail stores in South Korea. We collected the data through weekly surveys over a four-week period. Emotional labor strategies, affective state, emotional exhaustion, personal accomplishment, and work conditions were repeatedly measured as a longitudinal survey design and occupational identification and other demographic variables were measured only once. Due to the characteristic of our data, we needed to utilize the Hierarchical Linear Model (HLM), which was well confirmed by ICC1 values of the outcome variables. The results showed that surface acting were significantly related to only emotional exhaustion, not personal accomplishment, supporting H1a, but rejecting H1b. And H2 predicting non-linear relations between deep acting and job stress was supported. Finally, occupational identification moderated only the inverted U-shaped relations between deep acting and personal accomplishment, supporting H3b. This study contributes to the literature in emotional labor in several ways. First, we pursued a more comprehensive framework, combining three lenses of emotional labor research. Second, we suggested non-linear relations between deep acting and employees' well-being for the first time. Finally, we explored when the effects of deep acting would be strengthened by testing the moderating effect of occupational identification. Despite these contributions, however, this research contains some limitations. First of all, the results should be restrictively applied to other settings because we targeted only salespersons as our sample. Further, we did not test causality between emotional labor strategies and job stress. Even though controlling for the effects of individual differences by using HLM, we still could not effectively rule out all alternative explanations. Hopefully, these limitations will be addressed in future research.