Purpose The aims of this study were (1) to assess the test–retest reliability of the primary near infrared spectroscopy (NIRS) variables (i.e., StO2, T[Hb], [HbO2] and [HHb]) during cycling and (2) to investigate potential influences of exercise intensity and adipose tissue thickness (ATT) on this reliability. Methods 21 men and 20 women completed twelve constant work rate tests (6 min) at six different exercise intensities with each intensity performed twice. NIRS variables were measured at the vastus lateralis. The coefficient of variance (CV%), the intraclass correlation coefficient (ICC), mean bias and limits of agreement (LoA) were determined for reliability purposes. Results The reliability of baseline values were acceptable to very good (CV% range: 5.83 – 21.96%). The reliability of end-values (CV% range: 0.02 – 25.02%, ICC range: 0.0 – 0.935) and amplitudes (CV% range: 0.46 – 5099%, ICC range: 0.0 – 0.887) were more variable. In general, the mean biases of end-values and amplitudes showed wide limits of agreement. A homogeneous influence of exercise intensity on reliability could not be established but reliability measures appeared to be lower in people with a lower skinfold thickness. Moreover, the NIRS signals decreased with increasing ATT but stabilized upon reaching a cut-off of 8 mm ATT. In addition, ATT did have a significant influence on [HHb] amplitude. In participants with ATT < 8 mm, higher amplitudes were observed with increasing intensity whereas in participants with ATT > 8 mm, there were no differences between the intensities. Conclusion The study reveals variable results with regards to reliability and there was no consistent influence of exercise intensity on reliability. Participants with a lower skinfold thickness showed stronger reliability. Moreover, NIRS signals decrease when ATT exceeds 8 mm. Careful consideration is necessary when interpreting NIRS signals in such cases.