Cervical vagus nerve stimulation (VNS) is a neuromodulation therapy for the treatment of several chronic disorders. The effects of VNS are mediated by activation of nerve fibers of different types. In order to maximize the desired and minimize the undesired effects of VNS, assessing activation of vagal fiber types by VNS is essential. Evoked compound nerve action potentials (CNAPs) are commonly used as a method to estimate vagal fiber activation in the context of neurostimulation. However, vagal CNAPs are frequently contaminated by signals from non-neural sources, like electrocardiography (ECG), stimulus artifacts and evoked electromyographic (EMG) activity. In this study, we present a systematic methodology for suppressing non-neural signals in CNAP recordings from the rat vagus. The methodology involves intravenous infusion of vecuronium under ventilation, for suppressing EMG, and digital and analog signal processing, for suppressing ECG and stimulus artifacts, respectively. We compiled A-, B- and C-type fiber activation profiles with and without this methodology and found that our method significantly increased the reliability of CNAPs. We found that the A-component is obscured by the stimulus artifact, whereas the B- and C-components are frequently contaminated by evoked EMG. We extracted CNAPs evoked by square pulses of different polarities and amplitudes and documented effects consistent with well-established biophysical attributes of VNS.