The results of an analysis of thirteen different computer programs inv
olving 11 participants from 10 countries on eight measured glow curves
originating from LiF(TLD-100) have been compared. The programs use a
shape of a single glow peak which varies from program to program. Gene
rally speaking, the better the approximation to a first order kinetics
(Randall-Wilkins) glow peak shape the better the fit to the main peak
s (Nos 2 to 5) in a measured TLD-100 glow curve. The results deviate a
long with the complexity of the glow curve. Peak areas and activation
energies of peaks 2 to 5 in a TLD-100 glow curve have been determined
to within 3% by the programs with the better approximation to the Rand
all-Wilkins model. Programs with less accurate expressions for the glo
w peak shape show outliers up to 19%. The participants did not handle
the background in a uniform way. At low doses these differences influe
nce the fitting parameters, including peak areas, i.e. the absorbed do
se, strongly. Also for a complex glow curve, where many closely overla
pping peaks of similar intensity are present, the results of the diffe
rent programs vary strongly.