Background: Cervical vagus nerve stimulation (VNS) is an emerging bioelectronic treatment for brain, metabolic, cardiovascular and immune disorders. Its desired and off-target effects are mediated by different nerve fiber populations and knowledge of their engagement could guide calibration and monitoring of VNS therapies. Objective: Stimulus-evoked compound action potentials (eCAPs) directly provide fiber engagement information but are currently not feasible in humans. A method to estimate fiber engagement through common, noninvasive physiological readouts could be used in place of eCAP measurements. Methods: In anesthetized rats, we recorded eCAPs while registering acute physiological response markers to VNS: cervical electromyography (EMG), changes in heart rate (DHR) and breathing interval (DBI). Quantitative models were established to capture the relationship between A-, Band C-fiber type activation and those markers, and to quantitatively estimate fiber activation from physiological markers and stimulation parameters. Results: In bivariate analyses, we found that EMG correlates with A-fiber, DHR with B-fiber and DBI with C-fiber activation, in agreement with known physiological functions of the vagus. We compiled multivariate models for quantitative estimation of fiber engagement from these markers and stimulation parameters. Finally, we compiled frequency gain models that allow estimation of fiber engagement at a wide range of VNS frequencies. Our models, after calibration in humans, could provide noninvasive estimation of fiber engagement in current and future therapeutic applications of VNS.