This paper discusses the use of surrogate data testing to determine whether nonlinear techniques are justified for analyzing time series data. It reviews recent efforts to understand the limitations and caveats of surrogate data methods, and introduces new approaches to constrained randomisation. The paper emphasizes the importance of interpreting test results carefully, as the main limitation lies in the interpretability of the test results. It also discusses various examples, including the Southern Oscillation Index, and highlights the challenges of generating appropriate Monte Carlo samples for a given null hypothesis. The paper also addresses issues such as flatness bias in AAFT surrogates, periodicity artefacts, and the need for iterative refinement of surrogates. It concludes by discussing the implementation of these methods in the TISEAN software package and the importance of using appropriate statistical tests for nonlinearity. The paper also discusses the interpretation of test results, the difference between dynamic and non-dynamic nonlinearity, and the challenges of non-stationarity. The paper concludes by emphasizing the importance of testing a hypothesis against surrogates rather than testing against a hypothesis.This paper discusses the use of surrogate data testing to determine whether nonlinear techniques are justified for analyzing time series data. It reviews recent efforts to understand the limitations and caveats of surrogate data methods, and introduces new approaches to constrained randomisation. The paper emphasizes the importance of interpreting test results carefully, as the main limitation lies in the interpretability of the test results. It also discusses various examples, including the Southern Oscillation Index, and highlights the challenges of generating appropriate Monte Carlo samples for a given null hypothesis. The paper also addresses issues such as flatness bias in AAFT surrogates, periodicity artefacts, and the need for iterative refinement of surrogates. It concludes by discussing the implementation of these methods in the TISEAN software package and the importance of using appropriate statistical tests for nonlinearity. The paper also discusses the interpretation of test results, the difference between dynamic and non-dynamic nonlinearity, and the challenges of non-stationarity. The paper concludes by emphasizing the importance of testing a hypothesis against surrogates rather than testing against a hypothesis.