Introduction: An early component of the auditory event-related potential (ERP), the mismatch negativity (MMN), has been shown to be sensitive to native phonemic language sound contrasts compared to non-native or allophonic language sound contrasts. So far this has not been attested for different phonetic contexts. In the present study we investigated the mismatch field (MMF) response (the magnetic counterpart of the MMN) of native Danish speakers to the Danish phonetic contrast of [t]-[d] in two different phonetic contexts: One in which the sound contrast was phonemic ([tæ] versus [dæ]), and one in which the sound contrast was allophonic ([æt] versus [æd]), i.e. its phonemic status was neutralized. Methods: The stimuli consisted of the four Danish syllables: [tæ] and [dæ] (meaning ‘take’ and ‘then’, respectively), and [æt] and [æd] (both meaning ‘that’). These were presented in a passive listening MMN paradigm while participants’ MEG was recorded. [tæ] and [æt] acted as standards, and [dæ] and [æd] thus as deviants, respectively. Results: Comparing brain responses to the deviants and the standards, only the phonemic [tæ]-[dæ] contrast showed significant effects (FWE-corrected at p<0.05 at the cluster-level) within the typical MMN time range (100 to 300 ms after deviance onset) and over both hemispheres. Comparing the differences between the two contrasts ([dæ]-[tæ] minus [æd]-[æt]), the phonemic context elicited significantly larger MMF responses than the allophonic context (FWE-corrected at p<0.05 at the cluster-level), again over both hemispheres and within the typical MMN time range. Conclusion: By manipulating the immediate phonetic context in an oddball paradigm, we demonstrate that the human brain’s MMF response to language sounds is highly context-sensitive. This has important impact on the proposed long-term memory traces for native phonological categories. In order to generate different MMF responses to the same language sound contrast depending on the phonetic context, these long-term memory traces must thus be context-sensitive themselves or exist as separate traces for the context-dependent allophones of the phonological categories.