Splitting, in Kernberg's theory, plays a significant role in the development and maintenance of polarized and oscillating representations of self/others characteristic of borderline personality disorder (BPD). However, criticisms have been raised regarding its very existence. To further investigate the true nature of splitting, 60 participants were administered the Splitting Index and indexes of resistance to proactive interference (PI) using a Recent Negative Task with words that reflect negative or positive interactions compared to neutral words. Results demonstrated that the use of splitting is uniquely and significantly predicted by a higher capacity to resist PI and a lower capacity to consistently maintain this resistance when negative words are presented, above and beyond BPD traits, primitive defenses, and the presentation of neutral words. Results showed no evidence of a relationship between splitting and resistance to PI with positive words. Results appear compatible with Kernberg's conceptualization of splitting as an active defensive process.