In a series of 120 head-injured patients, recovery rates (rr) were calculated separately on the basis of either Glasgow come scale (GCS) scores or Maryland coma scale (MCS) scores; the latter contains the three variables of the GCS, excludes unevaluable responses, and provides more information concerning the status of brain stem reflexes and motor lateralization. The early (Day 3/4 vs. Day 1) and late (Day 8/14 vs. Day 1) recovery rates from the two scales generally agree (r = 0.76; r = 0.79), but in 39 of 94 patients the MCSrr and GCSrr disagreed by more than 10%. When the MCSrr was greater than the GCSrr, it more accurately reflected a favorable outcome. Graphic representations of clinical courses through serial plots of raw scores were more reliable when unaffected by intubation, sedation, swollen eyelids, casts, etc.; this was more often achieved with serial plots of MCS scores, which are graded as percentages of testable function. Final outcomes (good/disabled vs. vegetative/dead) were well predicted by Day 1 MCS scores above or below 35% (chi 2 = 27.63; p less than 0.001) and Day 1 GCS scores above or below 7 (chi 2 = 23.21; P less than 0.001). However, in 57 very sick patients (Day 1 GCS less than or equal to 7), the GCS did no better than chance (26 good, 31 bad outcomes), whereas 20 of 26 patients with a Day 1 MCS score of less than or equal to 35% had bad outcomes. In patients with severe multiple injuries, the Maryland coma scale may provide a more sensitive index of clinical course; a Day 1 MCS raw score of less than or equal to 35% is of grave prognostic significance.