This systematic review examines the use of 'doubly multivariate' methods, specifically canonical correlation analysis (CCA) and partial least squares (PLS), to investigate brain-behaviour associations in psychiatric disorders. The review identifies 39 studies across four diagnostic groups: ADHD, ASD, MDD, and PSD, as well as a transdiagnostic group. Most studies (67%) used CCA, focusing on the association between brain morphology, resting-state functional connectivity, or fractional anisotropy and symptoms/cognition. Key findings include the shared link between clinical/cognitive symptoms and frontal morphology/brain activity and white matter association fibers. Additionally, less investigated behavioural variables like physical health and clinical history were identified as important features. However, most studies were at risk of bias due to low sample size/feature ratio and/or in-sample testing only. The review highlights the importance of mitigating these biases and discusses the generalizability and stability of brain-behaviour associations, emphasizing the need for rigorous out-of-sample testing. The interpretation of multivariate models requires valid model weights, and the choice between CCA and PLS depends on the specific characteristics of the data. Overall, 'doubly multivariate' methods are promising for understanding how brain and behaviour interact in psychiatric disorders, but challenges remain in ensuring the reliability and validity of these associations.This systematic review examines the use of 'doubly multivariate' methods, specifically canonical correlation analysis (CCA) and partial least squares (PLS), to investigate brain-behaviour associations in psychiatric disorders. The review identifies 39 studies across four diagnostic groups: ADHD, ASD, MDD, and PSD, as well as a transdiagnostic group. Most studies (67%) used CCA, focusing on the association between brain morphology, resting-state functional connectivity, or fractional anisotropy and symptoms/cognition. Key findings include the shared link between clinical/cognitive symptoms and frontal morphology/brain activity and white matter association fibers. Additionally, less investigated behavioural variables like physical health and clinical history were identified as important features. However, most studies were at risk of bias due to low sample size/feature ratio and/or in-sample testing only. The review highlights the importance of mitigating these biases and discusses the generalizability and stability of brain-behaviour associations, emphasizing the need for rigorous out-of-sample testing. The interpretation of multivariate models requires valid model weights, and the choice between CCA and PLS depends on the specific characteristics of the data. Overall, 'doubly multivariate' methods are promising for understanding how brain and behaviour interact in psychiatric disorders, but challenges remain in ensuring the reliability and validity of these associations.