The recent empirical evidence documenting the presence of asymmetries in business cycles represents a challenge for the standard equilibrium models of real business cycle. These models successfully explain most first and second moments of the actual time series, but cannot replicate non-linear features of the data, unless a non-linear innovation is introduced. This paper aims at investigating the possible non-linearity in the technology shock, the basic innovation in Real Business Cycle models. In order to measure the unobservable technology shock, we derive some alternative measures of total factor productivity such as revenue-based and cost-based Solow residual and we also control for cyclical factor utilisation. We test for non-linearities and model a nonlinear SETAR model for the productivity shock as a natural extension of the autoregressive linear process, the standard way of representing technology shocks. Our findings suggest that, although the standard Solow residual turns out to be linear, the other measures of technology shock appear non-linear, as soon as non-technological cyclical components are ruled out.