Sometime ago Ananthakrishna and coworkers had predicted the existence of chaos in jerky flow based on a nonlinear dynamical model consisting of the time evolution equations for three types of dislocations and an equation for the evolution of the stress. Our main focus here is to report the verification of this prediction by analysing the stress signals obtained from samples of AlCu alloys subjected to a constant strain rate test. The analysis of the stress signals is carried out by using several methods. The analysis shows the existence of a finite correlation dimension and a positive Lyapunov exponent. We also carry out a surrogate analysis of the time series to ascertain that the signals are not from a power law stochastic process. From the analysis we find that the minimum number of variables required for a dynamical description of the jerky flow appears to be four or five, consistent with the number of degrees of freedom envisaged in the model.