This paper introduces a simple three-dimensional (3-D) spectral model for scalar fluctuations with variable anisotropy. This heuristic model is useful in the study of statistical properties of stably stratified flows. It goes a step further than existing analogues by explicitly taking into account variation of anisotropy with the size of the fluctuations. The spectral model is specified by its isovalue surfaces and by its generalized energy spectrum. These surfaces are ellipsoids, (variably) elongated in the vertical direction. The model depends on two arbitrary functions of the spectral parameter. The possibilities of such variable anisotropy models are demonstrated by choosing specific functions. The choice was guided by maximal simplicity requirement and by the tradition of turbulent studies which heavily uses power law functions. These examples show how quantitative predictions are obtained using such 3-D spectral models. They were selected in order to display characteristics that depend explicitly on the variable anisotropy and that could not have been obtained with previous models. The first example is the calculation of the one-dimensional spectrum of temperature for arbitrary direction of measurement and the natural prediction of a change in slope for quasi-horizontal spectra, while the oblique and vertical spectra retain the same slope. The second example is to account for radar echo aspect sensitivity, with angular contrast depending on the radar wavelength. The main advantages of the present model are its great flexibility provided by the use of two arbitrary functions and also the (relative) simplicity of its mathematical formulation.¿ 1997 American Geophysical Union