Predicting Performance with Letters of Recommendation

Abstract

Previous research has indicated that letters of recommendation are poor predictors of future performance, in part because characteristics of the letter writer and letter reader interfere with the objective analysis of the content of the letter. To help correct this problem, Peres and Garcia (1962) developed a technique for analyzing the content of letters of recommendation by identifying traits mentioned in each letter and placing the traits into one of five categories. It was the purpose of this paper to determine if the Peres and Garcia technique would be a valid method of predicting performance of psychology instructors and graduate students. The results of the two studies indicate that traits from letters of recommendation can be reliably classified into the five Peres and Garcia categories and that these traits are valid predictors of future performance (r's = .32 and .38).