The article reviews the literature on early warning signals (EWSs) for tipping points in various systems, including climate, ecology, health, social sciences, and physical sciences. Tipping points are characterized by abrupt, rapid, and sometimes irreversible changes in a system's state. EWSs are methods used to identify statistical changes in the underlying behavior of a system that indicate an impending tipping point. The review covers the last 20 years of empirical research, focusing on the metrics used, their success, and the specific systems and tipping points involved. Key findings include: - The majority of early warnings are based on critical slowing down (CSD), which involves detecting changes in the system's response to perturbations. - Non-CSD-based early warnings, such as skewness, flickering, and potential analysis, are also used but are more context-specific. - The use of early warnings has spread from ecology and climate to other disciplines, with ecology leading the way. - Temporal data is the most commonly used type of data for early warnings, followed by spatial data. - The performance of early warnings is generally positive, with 67.8% of studies reporting successful detection of tipping points. - Challenges in detecting early warnings include fast changes, slow responses, stochasticity, multiple drivers, and limited data. - The non-specificity of early warnings, particularly CSD-based ones, means they can also detect smooth and reversible transitions. - Multivariate systems pose additional challenges, and methods like network analysis and dimension-reduction techniques are being explored to address these issues. Overall, the review highlights the progress and limitations of early warning research, emphasizing the need for more system-specific indicators and the importance of understanding the underlying mechanisms of tipping points.The article reviews the literature on early warning signals (EWSs) for tipping points in various systems, including climate, ecology, health, social sciences, and physical sciences. Tipping points are characterized by abrupt, rapid, and sometimes irreversible changes in a system's state. EWSs are methods used to identify statistical changes in the underlying behavior of a system that indicate an impending tipping point. The review covers the last 20 years of empirical research, focusing on the metrics used, their success, and the specific systems and tipping points involved. Key findings include: - The majority of early warnings are based on critical slowing down (CSD), which involves detecting changes in the system's response to perturbations. - Non-CSD-based early warnings, such as skewness, flickering, and potential analysis, are also used but are more context-specific. - The use of early warnings has spread from ecology and climate to other disciplines, with ecology leading the way. - Temporal data is the most commonly used type of data for early warnings, followed by spatial data. - The performance of early warnings is generally positive, with 67.8% of studies reporting successful detection of tipping points. - Challenges in detecting early warnings include fast changes, slow responses, stochasticity, multiple drivers, and limited data. - The non-specificity of early warnings, particularly CSD-based ones, means they can also detect smooth and reversible transitions. - Multivariate systems pose additional challenges, and methods like network analysis and dimension-reduction techniques are being explored to address these issues. Overall, the review highlights the progress and limitations of early warning research, emphasizing the need for more system-specific indicators and the importance of understanding the underlying mechanisms of tipping points.