We examine different approaches to the analysis of noise in amorphous hydrogenated silicon associated with trapping and generation – recombination processes, which appear to predict very different noise spectra. In one approach the broad noise spectrum observed is assumed to be composed of a distribution of Lorentzian noise spectra, each associated with traps at a given energy depth, with appropriate weighting according to the energy distribution of characteristic time constants. This latter weighting is taken to mirror the energy distribution of states in the gap. This represents a linear superposition of the (weighted) contribution from individual trapping levels, each with its own characteristic time constant. This approach thus assumes that each trap level is an independent source of fluctuation in free carrier number, unaffected by the presence of other traps in the material. At first sight this assertion seems plausible, since in the multi-trapping situation envisaged, cross-correlation effects must be very small. However, the presence of several groups of traps, or, in the limit, a continuum, results in a distribution of characteristic time constants, which is not a simple linear superposition of the time constants for each level. Thus the assertion that a flat density of states, or a region which is flat, such as the top of a broadened level, results in a region of 1/f slope in the noise spectrum, may not be valid. We present an alternative model in which the distribution of time constants is appropriately incorporated, and compare the predictions of this model with the ‘superposition’ approach, using computed noise spectra.