This dissertation reports a new method for estimating mental workload (MWL) using a specific type of eye movement called saccadic intrusions (SIs). These studies demonstrate in laboratory conditions that there is a consistent relationship between MWL and SI eye behavior. The experiments manipulated MWL using the auditory N-back task, which quantitatively varies short-term memory load, thus manipulating MWL. While MWL was being manipulated, participants were engaged in a controlled visual task. Eye movements were recorded during the visual task and analyzed to determine whether participants had increased SIs as a function of MWL. The author developed a novel algorithm to quantify SI eye behavior; this algorithm takes time-series eye movement data as input, automatically searches for the characteristics of SI eye behavior in the data, accumulates occurrences of SI eye behavior (accounting for their amplitudes and dwell time), and quantifies the SI eye behavior into a single value, called the SI measure. Besides the SI measure, pupil diameter was used as a second dependent variable because it was already known to reliably reflect MWL. The analyses revealed that as MWL increased, both pupil diameter and the SI measure increased also. Correlation analyses were conducted across 12 one-minute trials with each participant; an average correlation value across 37 participants was r = 0.57 for the MWL-pupil diameter relationship. The other average correlation value was r = 0.45 for the MWL-SI measure relationship. Multiple linear regression analyses were conducted to predict MWL from the eye movement data. The R2 values from these analyses were 0.54 or higher; both pupil diameter and the SI measure significantly contributed to the prediction of MWL. These results indicate that SI eye behavior reliably reflects MWL. While pupil diameter does not remain accurate in reflecting MWL changes during illumination changes in the environment, SI eye behavior should continue to reflect MWL without being affected by illumination changes. This new measure for estimating MWL may be useful in evaluating a vehicle driver's MWL in real-time despite illumination changes.