Despite the relatively extensive research on pay levels and the consequences of income disparities, little is known about which reference groups people choose for comparative evaluation of personal income and why different selection patterns emerge. The aim of this paper is to dig deeper for answers to the following three questions: (1) What are the most important reference groups for income comparisons? (2) Who tends to use which type of reference group? (3) Which reference groupsare most detrimental to life satisfaction? The analysis is based on data from the 2008 and 2009 pretest modules of the Socio-Economic Panel Study (SOEP). The results show the working sphere (colleagues and members of the same profession) to be the most important point of reference for income comparisons, whereas neighbors are the least important. No clear-cut picture emerges for the differential selection of reference groups. Structural characteristics - e.g., level of income, education, and type of employment - are of only minor importance in the selection of reference groups for income comparisons. The results also suggest that individuals are likely to select those reference groups whose income is closest to their own. Therefore, the level of income relative to a reference group is related to the relevance of that group in income comparisons. The consequences of such comparisons for life satisfaction prove to be negative: the more importance an individual attaches to income comparisons with reference groups, the lower his or her life satisfaction. Income relative to neighbors and colleagues only affected life satisfaction when the respondent perceived such reference groups to be relevant in income comparisons. These results challenge previous research suggesting that people are unconscious of the true impact of comparison processes. Nevertheless, the results point to various difficulties in the measurement of social comparison processes andshow personal predispositions to be a major factor influencing such comparisons. The results strongly indicate the need for advanced instruments to measure the cognitive processes underlying social comparisons.