The reliability and validity of the Attachment Q Sort (AQS; Waters & Deane, 1985) was tested in a series of meta-analyses on 139 studies with 13,835 children. The observer AQS security score showed convergent validity with Strange Situation procedure (SSP) security (r=.31) and excellent predictive validity with sensitivity measures (r=.39). Its association with temperament was weaker (r=.16), which supports the discriminant validity of the observer AQS. Studies on the stability of the observer AQS are still relatively scarce but they have yielded promising results (mean r=.28; k=4, n=162). It is concluded that the observer AQS, but not the self-reported AQS, is a valid measure of attachment.