Readability: A Postscript.
Anthony V. Manzo · 1970
readability research. Recent articles by Coleman (1968), Bormuth (1967, 1968) and Klare (1968) have discussed the breadth and the depth of current readability research. Klare has pointed out that there are hundreds of language variables related to readability. Several of the most highly predictive of these variables, he says, are combined in readability formulae. The common procedure for doing this, Klare explains, is to start with the most predictive variable, then to add the next most predictive, etc., in a regression equation (p. 12). He concedes that beyond the addition of about eight variables, there is a point of diminishing return for the additional work expended. This may seem to subtly suggest that considerable additional work will continue to yield some further predictive value. This is not quite so: if all presently known linguistic variables were combined into a single formula it is dubious whether it would appreciably increase predictability. Furthermore, if such a super formula were developed it would no doubt be cumbersome and time consuming to applyas are many existing formulae.