The ability to continuously monitor workload in a real-world environment would have important implications for the offline design of human machine interfaces as well as the real-time improvement of interaction between humans and machines. We explored the usefulness of features derived from electroencephalography (EEG) spectra, near infrared spectroscopy (NIRS) hemoglobin concentration, and their combination, under data acquisition and processing conditions that could be applied to real-time usage. We simultaneously recorded from eight EEG and three NIRS channels during different workload conditions of the N-back task (N = 0, 1, 2). EEG and NIRS data were classified independently, and in combination. EEG could be used to reliably classify workload condition for most subjects and NIRS for half of them. NIRS tended to contribute to classification accuracy when combined with EEG in some subjects. We discuss implications and future directions. Copyright 2012 by Human Factors and Ergonomics Society, Inc. All rights reserved.