Near-infrared spectroscopy (NIRS) monitors changes in the regional cerebral oxygenation (rSO) and has been used to study cerebral physiologic functions in normal states and during epileptic seizures. Yet, the limitations and pitfalls of the technique are not fully understood. The authors evaluated NIRS changes over the frontal lobes during language tasks known to be associated with the integrity of the dominant frontal lobe in 17 normal adults (handedness: 14 right, 3 left). Recording protocol involved a baseline (3 minutes) with the subject relaxed and thinking of a blue sky and a second baseline during voluntary mouth movements mimicking speech. Two standardized neuropsychological word-generation tasks (controlled word-association tests: CFL and Animals) were then administered (4 minutes total) followed again by the two baseline procedures. Mouth movement without verbalization increased rSO values in excess of 2 SD of baseline fluctuations in 70% of the subjects. A t-test comparison of these baseline measurements was statistically significant (P < 0.0001). A general linear model repeated-measures procedure was then used to statistically examine NIRS increments during the cognitive tasks above and beyond the contribution produced by mouth movements during the second baseline. Compared to the second baseline, rSO further increased significantly during the word generation tasks (left frontal F value = 21.4, P < 0.0001; right frontal F value = 15.2, P < 0.001), confirming the utility of the technique. There was no apparent difference related to handedness. These findings corroborate the usefulness of NIRS to demonstrate focal cerebral activation during an "executive" language task. However, interpretation of changes can be confounded by extracerebral factors such as mouth movement, a finding of particular relevance in NIRS clinical applications.