Phonological patterns closely mirror independently existing phonetic ones. Hence the notion of 'phonologization': the emergence of a phonological process based on phonetic underpinnings. However, there is no agreement among phonologists about precisely how phonetics shapes phonology and what this means for phonology itself. This paper focuses on one facet of the question, the distinction between what I call the 'eliminative' and 'inclusive' views of phonology, which differ in how they answer the question, How much phonetic knowledge is incorporated into the phonology by phonologization? One way to answer this question is to seek empirical territory that reveals specific ways in which phonology does not mirror phonetics. A recent example of this is the work of Moreton, who coins the term 'underphonologization', meaning systematic failures of phonologization despite phonetic underpinnings. In this paper I rather take up instances of what we might call 'overphonologization': phonologization without the phonetic underpinnings. The existence of 'unnatural' phonological rules has been discussed many times before in this context, but here the focus is on a different phenomenon, taking up recent work of Myers and Padgett. Myers and Padgett argue that at least certain kinds of word edge segmental phonology have dubious phonetic underpinnings and are in fact examples of overphonologization. Their source is domain generalization: overgeneralization of effects from utterance edges, where they are phonetically motivated, to word edges, where they are not. The general implication is that at least some word-edge phonological processes are encoded without reference to phonetic underpinnings.