A study of clouds and variability in brown dwarf atmospheres is presented, with a focus on understanding the mechanism by which condensate clouds are dissipated at the transition between L and T spectral types. The results of a large J-band variability survey of over 60 mid-L to T dwarfs, and related observations are presented. We find statistically significant evidence (at the 95% confidence level) for an increase in large amplitude variability (peak to peak variations larger than 2%) within the L/T transition (L9-T3.5 spectral types), suggesting that the disruption of dust clouds by weather phenomena may contribute to the rapid decline in condensate opacity and J-band brightening observed to occur in this regime. Alternatively, the large amplitude variability observed could be the result of increasing contrast between discrete cloud features and the underlying atmosphere, rather than the development of cloud holes. We highlight the discovery of a T dwarf with 26% variability, and provide a detailed comparison with atmosphere models in order to infer the nature of cloud features responsible. Results are consistent with both the presence of thick storm features or cloud holes in our target’s atmosphere, but are inconsistent with the presence of magnetic spots. Supporting work related to clouds in L dwarf atmospheres, and the L/T transition binary fraction is presented. From studying an unusually blue L-dwarf companion to a nearby M-dwarf star we infer that thin or patchy condensate clouds, rather than low metallicity and/or high surface gravity, are most likely responsible for its peculiar colors. The discovery of significant variability for a different blue L dwarf in our large variability survey provides further evidence for this hypothesis. Finally, combining our own high contrast imaging observations for 8 L/T transition targets with previous results we infer a resolved L/T transition binary fraction of 14+7−6%, consistent with that inferred for the general brown dwarf population. This latter result suggests that the L/T transition sample is not highly contaminated by unresolved multiples.