Basolateral amygdala (BLA) principal neurons form distinct long-range circuits that mediate negative (via projections to the central amygdala) and positive (via projections to the nucleus accumbens) valence (Namburi et al., 2015). In a series of experiments using rats, we investigated the roles of these BLA pathways in Pavlovian fear learning. First, consistent with past work in mice, retrograde tracing showed that the two pathways were largely segregated; but, in contrast to past work in mice, single molecule FISH showed that these two pathways shared similar molecular profiles. Then, we used two-colour fibre photometry to simultaneously record from BLA-Acb and BLA-CeA neurons during Pavlovian fear conditioning under standard (conditioned freezing) or augmented (conditioned suppression of reward seeking) learning conditions. BLA-CeA neurons showed robust excitatory responses to footshock regardless of how fear learning was assessed. In contrast, BLA-Acb neurons only showed excitatory responses to the footshock when fear learning was assessed under augmented conditions. Finally, we silenced each pathway specifically at the time of the footshock during fear learning. BLA-CeA inhibition impaired fear learning regardless of how fear was assessed whereas BLA-Acb inhibition enhanced fear learning only when assessed under augmented conditions. Taken together, these findings show that BLA output pathways can have opposing roles in fear learning that are revealed by certain (environmental) states of learning.