The analysis of spectral lag between energy bands, which combines temporal
and spectral analyses, can add strict constraints to gamma-ray burst (GRB)
models. In previous studies, the lag analysis focused on the lags between c
hannel 1 (25-57 keV) and channel 3 (115-320 keV) from the Burst and Transie
nt Source Experiment (BATSE). In this Letter, we analyzed the cross-correla
tion average lags (including approximate uncertainties) between energy band
s for two GRB samples: 19 events detected by Ginga and 109 events detected
by BATSE. We paid special attention to the BATSE GRBs with known redshifts
because there has been a reported connection between lag and luminosity. Th
is extends our knowledge of spectral lags to lower energy (similar to 2 keV
). We found that lags between energy bands are small. The lag between the p
eak of similar to 50 keV photons and that of similar to 200 keV photons is
similar to 0.08 s. The upper limit in the lag between similar to 9 and simi
lar to 90 keV photons is similar to 0.5 s. Thus, there are not large shifts
at low energy. We found that about 20% of GRBs have detectable lags betwee
n energy bands in the Ginga and BATSE samples. From the internal shock mode
l, we found that there are three sources of time structure in GRB pulses: c
ooling, hydrodynamics, and angular effects. We argue that cooling is much t
oo fast to account for our observed lags and that angular effects are indep
endent of energy. Thus, only hydrodynamics can produce these lags. Perhaps
the radiation process varies as the reverse shock moves through the shell.