Topographic maps of the late auditory evoked potentials (AEP) were studied in a group of 20 children, aged 9-15 years, with severe language impairment (LI) and an age-matched control (C) group of 20 normal children. The stimulus was a pure tone at 500 Hz with a duration of 100 ms and a rise and fall time of 20 ms. The intensity was 75 dB HL. Six test sequences of 50 stimuli at an interval of 1.0 s were presented to the left and right ear separately. Grand average maps of all the children in the LI and the C group, separately, were calculated and showed a bilateral negativity over the temporal areas, corresponding to the negative peak of the T complex (Tb) at a latency of about 150 ms. The amplitudes were larger contralateral to the stimulated ear in both groups. A difference map between the two grand average maps showed topographic differences at temporal sites. However, the T complex could not be identified in 7 LI children and 1 C child. In the remaining subjects with a T complex the topographic pattern was similar in the two groups but with lower amplitudes and significantly longer latencies in the LI group. The presence and latency of the positive peak of the T complex (Ta) was also examined, showing significant between-group differences. The value of Tb, in diagnosing language impairment, was tested by means of a scoring system and with statistical mapping. The diagnostic sensitivity of Tb latency, amplitude and topography in selecting the LI children was 90% to 40% with a specificity of 80% to 95%. The results indicate slower and deviating processing in the central auditory pathways of LI children. The variation in results between children, with missing components or prolonged latencies in the majority, but not all, of the LI children, may be explained by different pathophysiological causes of their language impairment. The more pronounced deviations of the T complex compared with the vertex-recorded NI may also indicate a specific role of the T complex-related cortical activity in language impairment.