In this paper, we propose Stochastic sketching for global optimization based on a simulation of human behaviour. Stochastic sketching tries to do things simply in the human way without too much interpretation instead of modeling the thought and strategies of human beings and applying an artificial model to problems. We introduce and discuss concepts and components essential to stochastic sketching in detail, including sampling guide, zooming controller, sketching model, precision threshold, and satisfaction probability. Experimental results of stochastic sketching on several test functions and a set of recommended parameter settings are given, as well as preliminary comparisons between stochastic sketching and related evolutionary algorithms including evolution strategies, evolutionary programming, and genetic algorithms.