Data from the phonological assimilation of borrowings have typically been employed in linguistic literature in an ad hoc manner for the justification of various proposals concerning the present or past states of the source or target languages. These data are sometimes found in orthographic form, transcribed phonetically or transliterated into native orthography either by speakers of the target language or by linguists describing the language. Whenever borrowings are utilized as evidence for a particular linguistic claim, it is assumed that there exists some well-defined theory of loan-word assimilation which elevates the data beyond question or doubt. Until recently few such theories have been formulated, and the reasons for this are reasonably clear. The nativization of borrowings involves perception, analysis and production in terms of the target system, and unless we already know exactly what the units and processes comprising this system are, it is difficult to propose intelligent hypotheses to account for the loan-word data. In other words, any account of loan–word assimilation is necessarily predicated on an adequate theory of language, or in the absence of such a theory, can be used as evidence in its construction. Our objective here is not to discuss the pros and cons of those few theories of phonological adaptation which have been proposed, or their interaction with phonological theories in general, but rather to caution against the indiscriminate use of orthographic loan–words in (a) the formulation of theories of borrowing, (b) the empirical justification of general theories of language or (c) making specific claims about the phonologies of the source or target languages.