Canadian Raising is a phonological process which raises the nucleus of both the /aɪ/ and /aʊ/ diphthongs above 60Hz (Labov et al. 2005: ANAE, p. 205) before voiceless segments. The /aɪ/ diphthong is raised in much of Canada as well as in many American dialects, including the Inland North, resulting in alternations among a large number of minimal pairs distinguished by their voicing such as /lʌɪf/ ‘life’ ∼ /laɪv/ ‘live’ and /brʌɪt/ ‘bright’ ∼ /braɪd/ ‘bride.' This /aɪ/-raising is a classic example of phonological opacity because it is canonically conditioned not only by surface voiceless segments but also underlyingly voiceless segments, as with the flapped /t/ in /rʌɪɾɚ/ ‘writer.’ However, not all /aɪ/-raising speakers exhibit this opaque pattern: so-called transparent or phonetic /aɪ/-raising speakers only raise before surface voiceless segments as in /rʌɪt/ ‘write' but not /raɪɾɚ/ ‘writer.’ The existence of this latter group has renewed debate about the ultimate origins of the raising patterns and the relationship between transparent and canonical raising. This paper contributes to that discussion with a new model of child language acquisition in variable settings, finding that the presence of transparent /aɪ/-raising as well as its rare attestation and sparse distribution can be accounted for as a contact phenomenon in which some child learners innovate a novel transparent raising pattern when their communities contain the appropriate mix of canonical raising and non-raising speakers.