When two variables are correlated, if one of the variables forms part of the other, then the size of the correlation will be "inflated". The correlation may be deceptively large due to this sort of score "contamination".

Such inflation / contamination is common in item analysis when the criterion score is the total test score. When item scores are correlated with total test scores, part-whole inflation is to be expected as the total test score includes the item. If this contamination is not corrected, the item discrimination measure (usually a correlation coefficient of some sort) may be artificially large. If the number of items is large the inflation will be minimal; however, for tests with (say) 30 items of less, the inflation effect may make a noticeable difference, potentially making items appear to be more discriminating than they actually are.

Lertap automatically corrects for the effects of part-whole inflation.

In some other programs, such as Iteman, the correction for part-whole inflation is called the correction for "spuriousness".

The effects of part-whole contamination may be investigated by using an external criterion analysis.

See Berk (1978) for more comments and related equations. Lertap uses the Cureton correction formula.