Our study aimed at seeking for a neural correlate of automatic phoneme-category discrimination in children in the condition of vast natural acoustical variation of the male voice. Mismatch negativity, MMN, an auditory event-related response to any discriminable change in an ongoing sound stream, as well as its magnetic counterpart MMNm, and the MMN paradigm as such, have been shown to be extremely useful in studies of speech perception in humans. In practice, the MMN or MMNm can be seen in a difference waveform elicited by subtracting the response to a standard stimulus from the response to a deviant stimulus. (Näätänen and Winkler, 1999). Importantly, MMN is more resistant to the attentional manipulations than other endogenous component, e.g. N2b or P3a, which makes it a very suitable tool for studying auditory discrimination in children (Cheour et al., 2000). A number of earlier studies using MMN and MMNm provided evidence for the existence of language-specific memory traces which guide the categorical perception of speech sounds in the auditory cortex (Näätänen et al., 1997; Dehaene-Lambertz et al., 2000). In adults, a robust MMNm can be recorded in the left hemisphere in response to the change across the phoneme categories despite the vast acoustical variation of stimuli (Shestakova et al., 2002). The present study aimed at probing whether such long-term memory traces could be found in pre-school and early-schoolage children. The magnetoencephalogram (MEG) was recorded in Russian 6-year-old children with a whole-head magnetometer. Stimuli were produced by 450 speakers and randomly distributed across the three phonetic categories of Russian. The MEG data were analyzed using Equivalent Current Dipoles (ECD) and Minimum Current Estimate (MCE) (Hämäläinen et al., 1993). The preliminary results showed that despite the correct behavioral vowel identification no difference between the standard and deviant response could be found in the magnetic responses of the children given that theirs obligatory responses were robust. From our findings we can conlude the following. First, in children neural populations which differ from those in adults might be involved in the preattentive category-change processing. Second, it might be that in 6-ear-old children the vowel-category discrimination is not fully automatic. Therefore it would be absolutely necessary to add a forced-choice task in addition to the main experiment and replicate the magnetoencephalographic experiment using electroencephalogram.