What does reflection mean?
In centroid factor analysis, the columns of correlations are summed. From time to time a column sum is a negative value, meaning that some of the correlations between the sorts are negative to the degree that they yield a negative sum. Since the concepts of negative and positive in factor analysis are arbitrary, when a negative column sum is encountered during factor extraction the program multiplies each correlation in that column (and the coordinate row) by -1. Consider the tables below, with correlations for five sorts. As it happens, before reflection all column sums are negative, so when they are summed the Table Sum is also negative.
