This paper estimates Taylor rules using real-time inflation forecasts of the Swiss National Bank's (SNB) ARIMA model and real-time model-based internal estimates of the output gap since the onset of the monetary policy concept adopted in 2000. To study how market participants understand the SNB's behavior, we compare these Taylor rules to marketexpected rules using Consensus Economics survey-based measures of expectations. In light of the recent financial crisis, the zero-lower bound period and the subsequent massive Swiss franc appreciation, we analyze potential nonlinearity of the rules using a novel semi-parametric approach. First, the results show that the SNB reacts more strongly to its ARIMA inflation forecasts three and four quarters ahead than to forecasts at shorter horizons. Second, market participants have expected a higher inflation responsiveness of the SNB than found with the central bank's data. Third, the best fitting specification includes a reaction to the nominal effective Swiss franc appreciation. Finally, the semiparametric regressions suggest that the central bank reacts to movements in the output gap and the exchange rate to the extent that they become a concern for price stability and economic activity.