Representational similarity analysis (RSA) connects the three branches of systems neuroscience: brain-activity measurement, behavioral measurement, and computational modeling. It addresses the challenge of relating these branches by computing representational dissimilarity matrices (RDMs), which characterize the information carried by a given representation in a brain or model. RDMs are derived by comparing activity patterns across different modalities, subjects, and species, and are tested using randomization and bootstrap techniques. RSA provides a framework for quantitatively relating multi-channel neural activity to computational theory and behavior by comparing RDMs. It has broad potential for experimental design and allows for the integrated analysis of data from all three branches of systems neuroscience. RSA is demonstrated by relating visual object representations measured with fMRI in early visual cortex and fusiform face area to computational models. The RDMs are analyzed using multidimensional scaling and tested for statistical significance. RSA can relate different brain regions, subjects, species, and modalities of brain-activity measurement, and can also relate brain activity to behavior. It is applicable to conventional and novel experimental designs, and can address a wide range of neuroscientific questions. RSA provides a way to abstract from the spatial layout of representations and directly compare brain and model representations. It uses RDMs as signatures of representations and allows for the comparison of different models and brain regions. RSA has been applied to various models, including complex computational models, simple image transformations, and conceptual models. It has been used to analyze the similarity structure of activity patterns in early visual cortex and fusiform face area, and to relate these to models and other brain regions. RSA provides a quantitative way to compare representations across different modalities and can help characterize regional representations. It has been shown to be effective in relating brain activity to behavior and in identifying the best-fitting models for different brain regions. RSA is a promising approach for integrating data from all three branches of systems neuroscience and for advancing our understanding of brain function and behavior.Representational similarity analysis (RSA) connects the three branches of systems neuroscience: brain-activity measurement, behavioral measurement, and computational modeling. It addresses the challenge of relating these branches by computing representational dissimilarity matrices (RDMs), which characterize the information carried by a given representation in a brain or model. RDMs are derived by comparing activity patterns across different modalities, subjects, and species, and are tested using randomization and bootstrap techniques. RSA provides a framework for quantitatively relating multi-channel neural activity to computational theory and behavior by comparing RDMs. It has broad potential for experimental design and allows for the integrated analysis of data from all three branches of systems neuroscience. RSA is demonstrated by relating visual object representations measured with fMRI in early visual cortex and fusiform face area to computational models. The RDMs are analyzed using multidimensional scaling and tested for statistical significance. RSA can relate different brain regions, subjects, species, and modalities of brain-activity measurement, and can also relate brain activity to behavior. It is applicable to conventional and novel experimental designs, and can address a wide range of neuroscientific questions. RSA provides a way to abstract from the spatial layout of representations and directly compare brain and model representations. It uses RDMs as signatures of representations and allows for the comparison of different models and brain regions. RSA has been applied to various models, including complex computational models, simple image transformations, and conceptual models. It has been used to analyze the similarity structure of activity patterns in early visual cortex and fusiform face area, and to relate these to models and other brain regions. RSA provides a quantitative way to compare representations across different modalities and can help characterize regional representations. It has been shown to be effective in relating brain activity to behavior and in identifying the best-fitting models for different brain regions. RSA is a promising approach for integrating data from all three branches of systems neuroscience and for advancing our understanding of brain function and behavior.