Constituents of complex words can themselves be complex words. Some kinds o
f complex constituents appear more often than others. This study presents a
quantitative investigation of this phenomenon. We show that many kinds of
base words are significantly overrepresented or underrepresented. This hold
s not only for constituents of derived words, but also for constituents of
compounds. We furthermore show that the degree of over-representation or un
derrepresentation correlates with word frequency, word length, and degree o
f productivity. We offer a functional explanation of this correlation in te
rms of processing and storage in the mental lexicon.