In this study, the velocity fluctuation near the detonation limits is investigated experimentally. Five explosive mixtures in five different diameter tubes were used and the choice of the mixtures included those considered as "stable" with regular cellular pattern and "unstable" with highly irregular cellular pattern. Photodiodes spaced at regular intervals along the tube were used to measure the detonation velocity. Piezoelectric transducers were also used to record the pressure profiles. Smoked foils were used to register the cellular detonation structure. Away from the limits, the detonation is found to propagate at a steady velocity throughout the length of the tube and the fluctuations of the local velocity are generally small. For stable mixtures with high argon dilution, the onset of the detonation limits is indicated by an abrupt drop in the detonation velocity to about 0.4V(CJ) after a short distance of travel. The detonation may continue to propagate at this low velocity before decaying eventually to a deflagration wave. For deflagrations the optical detector sometimes failed to register a signal due to low luminosity of the front. In unstable mixtures, galloping detonations are observed only in small diameter tubes (e.g., D = 12.7, 3.2 and 1.5 mm). A large number of fairly reproducible cycles of galloping detonations can be observed in very small diameter tubes. In large diameter tubes (e.g., D = 31.7 and 50.8 mm), no galloping detonations are observed in all stable and unstable mixtures. For stable mixtures, no galloping detonations are observed even in small diameter tubes of D = 3.2 and 1.5 mm. Smoked foils records show that the cellular detonation structure changes from multi-headed to single-headed spin as the limit is approached. In a galloping detonation cycle, a decay from multi-headed to single-headed detonation is observed. However, the cellular structure vanishes for further decay of the galloping detonation to the low velocity phase of the galloping cycle. Although galloping detonations could be considered to define the boundary for detonation limits, this definition lacks generality since galloping detonations are not always observed in all mixtures and in all tube diameters. Thus the onset of single-headed spin is perhaps the most appropriate criterion of the detonation limits in tubes. (c) 2014 The Combustion Institute. Published by Elsevier Inc. All rights reserved.