Body mass index (BMI) is a common research measure used to allocate the weight of study participants into broad groups, typically underweight, normal weight, overweight and obese categories. Some studies measure the actual height and weight of participants to calculate a measured BMI score but many rely on self-reported height and weight. This paper examines the accuracy of BMI categories based on self-reported height and weight compared with measured data. It critiques the use of mean differences between measured and self-reported BMI scores as a measure of self-reported BMI accuracy. It also demonstrates how reliance on proportion changes in BMI categories as a result of using self-reported BMI scores fails to demonstrate the extent that participants are misclassified into an incorrect BMI category. The paper also examines actual rates of misclassification for BMI categories, using the obesity category as an example. The association of sociodemographic factors with BMI accuracy is also considered, as well as the technical feasibility of using correction factors to improve the accuracy of self-reported BMI. The main finding is that self-reported BMI score and associated BMI category tend to be underestimated. Also, the self-reported BMI misclassification rate appears to increase as measured BMI score increases, with obese study participants in particular being most likely to be misclassified. No correction method has been found that will adjust accurately for misclassification based on self-reported BMI. The analysis suggests that self-reported BMI may have major limitations for use in public health research and policy-making decisions.