The effects of under-and Overextraction on principal axis factor analysis with varimax rotation were examined in 2 Monte Carlo studies involving 6,420 factor analyses. It was found that (a) when underextraction occurs, the estimated factors are likely to contain considerable error; (b) when Overextraction occurs, the estimated loadings for true factors usually contain substantially less error than in the case of underextraction; and (c) Overextraction can result in factor splitting when a general factor is present and there are no unique variables in the data set. The authors recommend that factor analysts (a) use effective methods to estimate the number of factors; (b) avoid underextraction, even at the risk of Overextraction; and (c) include randomly generated unique variables as "insurance" against factor splitting when a general factor may be present. Researchers conducting exploratory principal axis factor analyses (PFA) seldom know beforehand how many factors underlie their data. This uncertainty creates a practical dilemma: How many factors should be extracted if the true number of factors is unknown? The decision is especially difficult because little is known regarding the effects of underextraction (extracting too few factors) and Overextraction (extracting too many) on the quality of the factor analytic solution.Researchers have dealt with this dilemma in a variety of ways. Many have relied on Kaiser's (1960) eigenvalues-greater-than-1 rule to determine the number of factors. Although several studies have indicated that this rule is unsatisfactory (Hakstian, Rogers, &

No abstract