In recent years there has been a growing realization that geological media vary over a broad scale range. As such heterogeneity cannot be described in a deterministic way, a parallel growth in stochastic analyses of geological/geophysical data has emerged. The stochastic description of these media is usually through some form of correlation function, of which the von Karman is the most widely employed. Using this form, media can be described in terms of a characteristic scale size (or correlation length), L and a coloured scaling regime, with scaling described by H, the Hurst exponent. Beyond the correlation length, the material follows a white noise spectrum where material average properties dominate, below the correlation length local heterogeneity dominates. Hence, the correlation length is a fundamental parameter in a range of geodynamical problems. In situ information about stochastic properties of deep crustal rocks can be obtained from the statistical analysis of reflection seismic data. Typical correlation distances within the crust are found to be several hundred metres. Here we show that correlation distances derived from reflection seismic data are strongly influenced by the spectral content of the source. In particular we conclude that there is no reliable evidence for hundred metre scale correlation lengths for crustal heterogeneity.