A collisional drift wave instability driven by electron temperature gradients is described which appears to predict the main features of the precursors of the internal disruptions observed in tokamaks, i.e. the calculated frequency and growth rate are in quantitative agreement with the observed ones, and the predicted profiles are, at least qualitatively, consistent. In these experiments the plasma appears, despite all the uncertainties of the parameters, to be close to the theoretical threshold. It is noted that this instability can appear either if ηe ≡ d ln Te/d ln N > 0 or ηe < 0. It is thus suggested that the theory might also be relevant to the anomalous current penetration observed in all tokamaks. Finally, a collisionless instability also occurs when ηe < 0.