The influence of soil properties on microbiological processes is often examined by comparing the behaviour of taxonomically disparate soils. One of the limitations of this approach is that the results can be confounded by the unmeasured properties which vary between soils of different type or between soils which have had different climatic and management histories. This study tested the hypothesis that the heterogeneity between 100 small contiguous undisturbed soil cubes (about 1·7 cm3), sampled from the surface of a very small field plot (14 by 14 cm), was sufficiently large to use for the exploration of how soil properties influence biological processes. After incubation of the soil for 35 days, the coefficients of variation for nitrate (NO3), ammonium (NH4), gravimetric water content (θg), bulk density (BD), pH buffering capacity (pHBC), and pH were 28, 39, 27, 10, 13, and 2%, respectively. A multiple regression equation predicting nitrate concentration had an r2 value of 0·89 and significantly included 4 predictor variables, with only pH being non-significant. These analyses confirmed the hypothesis. When the values of measured soil properties of adjoining soil cubes were meaned to estimate values for larger soil volumes, the multiple regression equations for predicting NO3 concentration explained more of the variation (r2 values as high as 0·99). However, information concerning the influence of certain soil properties on N mineralisation and nitrification was lost, with only pHBC and BD remaining significant in the regression model. It was concluded that at a given physical scale of investigation, the structure of the spatial variability may determine whether or not a relationship between 2 variables is observed. Smaller samples are more likely to identify functional relationships which may exist between measured variables at the microscale.