This talk addresses the problem of lexical knowledge, looking at English. The English lexicon consists of three main sectors described below that differ from one-another in the interaction between Morphology (the way words can be decomposed into sub-parts) and Phonology (the way in which sound structure of words can be reduced to general principles). I will focus in particular on the distribution of 'long' vowels (henceforth marked with a colon and generally realized as diphthongs), like those of fi:ni:te, divi:ne, cri:me, na:ture, expla:in. English long vowels have a pervasive tendency to shorten, witness in-finite, divin-ity, cimin-al, natur-al, explan-ation, but such tendency is manifested only in one sector of the lexicon, the one made up of words containing 'Latinate' affixes, like those just shown. The other two sectors are represented respectively by morphologically simple items, like vi:tamin, cra:ne, ni:ghtinga:le, etc., and by items formed with 'Germanic' affixes, like cri:me-less, la:bor-less, na:tion-hood, ta:ble-ful, etc. Both of these sectors are immune to vowel shortening, as the examples show. In addition, within the Latinate affixed lexicon, shortening, while systematic in certain environments, is variable in others, as in sati:re/ satir-ist, versus extre:me/ extre:m-ist. I will propose solutions to these puzzles by way of a general architecture that has three components, whose relative weights or scope varies over the different sectors of the lexicon. These are: (i) Memorization; (ii) Word-to-word association; and (iii) General principles of sound structure (Phonology).