The surface interference EMG signal provides some information on the neural drive to muscles. However, the association between neural drive to muscle and muscle activation has long been debated with controversial indications due to the unavailability of motor unit population data. In this study, we clarify the potential and limitations of interference EMG analysis to infer motor unit recruitment strategies with an experimental investigation of several concurrently active motor units and of the associated features of the surface EMG. For this purpose, we recorded high-density surface EMG signals during linearly increasing force contractions of the tibialis anterior muscle, up to 70% of maximal force. The recruitment threshold (RT), conduction velocity (MUCV), median frequency (MDFMU) and amplitude (RMSMU) of action potentials of 587 motor units from 13 individuals were assessed and associated to features of the interference EMG. MUCV was positively associated with RT (R2 = 0.64 ± 0.14) whereas MDFMU and RMSMU showed a weaker relation with RT (R2 = 0.11 ± 0.11; 0.39 ± 0.24, respectively). Moreover, the changes in average conduction velocity estimated from the interference EMG predicted well the changes in MUCV (R2 = 0.71), with a strong association to ankle dorsi-flexion force (R2 = 0.81 ± 0.12). Conversely, both the average EMG MDF and RMS were poorly associated to motor unit recruitment. These results clarify the limitations of EMG spectral and amplitude analysis in inferring the neural strategies of muscle control and indicate that, conversely, the average conduction velocity could provide relevant information on these strategies.