This paper reports some results derived by studying statistical characteristics of irregularities which are produced as a result of plasma mixing by random drifts. It is found that as time progresses, the irregularity amplitude grows in proportion to √t. Characteristics of the irregularity spectrum undergo more violent changes. For example, the spectrum width grows according to a law close to the exponential one. Thus, relatively large-scale random motions (of the size of ~ 10 km, for example) rapidly generate a broad spectrum of small-scale irregularities as small as meter scales. A broadening of the spectrum is stopped by the transverse diffusion or by the longitudinal diffusion and recombination. An assessment is made of the influence of these processes. Equations are obtained, which describe the evolution of the correlation function of plasma density fluctuations for given statistical properties of irregular mixing motions in the case of low-intensity mixing. For the Gaussian spectrum of random drifts, numerical calculations of spectra of irregularities appearing from the initial state, which contain only a smooth large-scale gradient of plasma density, are made. In the initial stage the calculated irregularity spectrum is also close to the Gaussian one and does not involve small scales. A gently-sloping short-wavelength part is then found, where the dependence of spectral power on the wave vector is close to a power law with index - 1 (for a one-dimensional spectrum). Subsequently, the character of the spectrum remains unchanged. The formation mechanism for irregularities investigated in this paper can operate effectively under conditions of middle and high latitudes. © 1990.