Objective: Intracortical electrical stimulation is used to modulate brain activity. However, non-selective neuronal activation limits its efficacy. A biphasic electrical stimulation pulse has cathodic (negative) and anodic (positive) phases; swapping them changes current direction. In computational models, different current directions selectively recruit distinct neuronal populations. Whether this occurs in intracortical stimulation is unknown. We tested whether current direction affects the recruitment of neuronal populations in vivo, and by what mechanism. Approach: We implanted 20 anaesthetized rats with multichannel silicon probes in the visual cortex to simultaneously stimulate and record across all cortical layers. Stimulation used either cathodic-first or anodic-first biphasic pulses. Models also predict that current direction effects strengthen when the second phase is lengthened relative to the first; therefore, in 7 animals, we tested scaled pulses. Results: Cathodic-first pulses evoked stronger responses than anodic-first pulses in input and deep layers. Extending the pulse's second phase transformed cathodic-first magnitude advantage into selective recruitment by collapsing anodic-first neuronal response. This anodic-first deficit was layer- and orientation-dependent. Namely, anodic-first activation spread was weakened specifically in input and deep layers in the horizontal dimension, but not superficially. Conclusion: Although cathodic-first pulses are generally more effective, this advantage is not uniform across cortical depth, indicating current direction-dependent spatial selectivity. This selectivity is consistent with layer-specific fiber architecture: horizontally-passing fibers, more easily activated by cathodic-first pulses, dominate in the input and deep layers, while radial fibers - superficially. Therefore, current direction can be used as a stimulation parameter in cortical prostheses to achieve layer- and orientation-dependent spatial selectivity.