Bias is said to occur when validity is better for one group than another (e.g., when diagnoses are more valid for male or female clients). This article provides (a) a methodological critique of studies on race bias and gender bias in diagnosis and (b) a narrative review of results from studies with good internal validity. The results suggest that race bias occurs for the diagnosis of conduct disorder, antisocial personality disorder, comorbid substance abuse and mood disorders, eating disorders, posttraumatic stress disorder, and the differential diagnosis of schizophrenia and psychotic affective disorders. Other results suggest that gender bias occurs for the diagnosis of autism spectrum disorder, attention deficit hyperactivity disorder, conduct disorder, and antisocial and histrionic personality disorders. The way that symptoms are expressed (e.g., Black cultural expressions of depression) appears to have a significant effect on diagnoses. It may be possible to decrease bias by expanding the use of (a) mental health screening, (b) self-report measures including some psychological tests, (c) structured interviews, and (d) statistical prediction rules. Finally, evidence exists that (a) the diagnosis of personality disorders should be made using dimensional ratings and (b) training in cultural diversity and debiasing strategies should be provided to mental health professionals.