Circadian rhythm of locomotor activity of the desert beetle T.gigas usually has two narrow peaks: morning (M) and evening (E). While entrained with diurnal (Tz = 24 hr) full or skeleton photoperiods, the M peak is precedes light, while the E peak coincides with light. In a variety of natural and laboratory conditions both peaks tend to maintain a stable mutual phase relationship, about 12 hr apart. The phase responses of the M and E peaks were studied using 6-hr, 30 lx green LED-light pulses applied around ct3, ?t12 and ct18. The PRC for the E peak, plotted versus ct0 (extrapolated moment of light-on) as abscissa, had the same position, as the PRC for the M peak. Both PRCs were asymmetric, but in an opposite way: for the M peak the area of phase advances was bigger, than the area of phase delays, while for the E peak, vice versa. The transient PRCs on day 1, 2 etc. did not differ from the steady state PRC, i.e, the phase response was accomplished virtually in one cycle. Period changes were almost all positive (period became longer after a light pulse). The only "dead zone" in the period response curve (decrease of Dt down to zero) was around subjective evening – early night. Here again, the M peak appeared more "eager" to phase advances than the E peak. Our data support the hypothesis that M and E peaks are controlled by putative separate oscillators. These oscillators seem to have different properties, tend to phase shift to a different extent, and are extremely strongly mutually coupled with phases locked at approximately 180°. The asymmetry of properties of the M and E oscillators has a clear adaptive significance.