Two questions are central to the controversy in phonological theory: (1) are there empirical differences between morphophonemic alterations and allophonic variation, and (2) what are the universal constraints on the ordering of phonological processes within the phonological rule component. This paper illustrates a Natural Generative Phonology (NGP) analysis as opposed to an Abstract Generative Phonology (AGP) and attempts to show that the No Ordering Hypothesis of NGP forces the linguist to posit concrete analyses, i.e. analyses whose phonological rules are transparent from the phonetic level. This theoretical restriction corresponds to cognitive constraints on the language learner. It is assumed that only rules which are completely motivated on the phonetic level are possible, and that once the phonetic motivation for the rule is obscured in any way, underlying forms are changed, and the rule becomes morphologicalized. Data from two cases of language change in Spanish are used to test this hypothesis: (1) stress assignment, where the historic drift towards paradigmatic stress is explained by the morphological analysis required by NGP but could be avoided by AGP, and (2) aspiration and deletion of word-final /s/, where the rule of vowel quality may be morphologicalized, and productive and intrinsically ordered after deletion. (CLK)