Conditioned suppression is a psychology-based task to understand how learned cues interrupt ongoing reward-seeking behavior. In conditioned suppression studies, the presentation of a cue that was previously paired with an aversive outcome is sufficient to suppress a reward response, whether that response is operant (Estes & Skinner, 1941) or consummatory (Leaf & Muller, 1965) by design.

Traditionally, conditioned suppression studies use forward conditioning pairings in which a cue (CS + ) predicts a forthcoming unconditioned aversive stimulus (US) (Arico et al., 2017, Bercum et al., 2021, Bouchekioua et al., 2022, Bouton and Bolles, 1979, Greiner et al., 2019, Piantadosi et al., 2020). Forward conditioning effectively suppresses consummatory behavior during CS+ (Leaf & Muller, 1965). Regardless of the number of forward CS + pairings, a forward CS + remains a conditioned excitor. In the case of conditioned suppression, the forward paired CS + excites the predicted conditioned response associated with the US, typically freezing behavior and reduced reward-related responding (Leaf & Muller, 1965). Backward conditioning, where an aversive stimulus predicts a forthcoming CS+, has similarities and differences to forward conditioning. Like forward conditioning, a few backward pairings result in the CS + becoming an excitatory stimulus (evoking a conditioned response). However, many backward pairings results in the CS + becoming an inhibitory stimulus (not evoking a conditioned response) (Ayres et al., 1987, Barnet and Miller, 1996, Chang et al., 2003, Christianson et al., 2008, Cole and Miller, 1999, Delamater et al., 2003, Heth, 1976, Heth and Rescorla, 1973, Maier et al., 1976, Moscovitch and LoLordo, 1968, Spetch et al., 1981). Rescorla defined a conditioned inhibitor as a CS + that inhibits the expected conditioned response due to the learned association with the absence of the US (Christianson et al., 2012, Rescorla, 1969). Others have used this definition of conditioned inhibition to investigate the learning and memory of “safety” (Ng et al., 2024, Sangha et al., 2013, Sangha et al., 2014). Wagner offered an explanation for the unique shift from conditioned excitation to conditioned inhibition feature of backward conditioning with the sometimes-opponent process theory, which posits a trial-based acquisition of the conditioned response (Wagner, 1981). Others have utilized this feature of backward conditioning to create a “safety signal” with many backward CS + pairings (Christianson et al., 2008). While conditioned response characteristics about forward and backward conditioning are well studied, how the history of backwards conditioning may differ from forward conditioning with respect to conditioned suppression is unknown.

In rodents, freezing is the most readily observed behavior in response to forward conditioning of an aversive stimulus, while locomotor activity is observed in response to backward conditioning of an aversive stimulus (Bouton & Bolles, 1980). Indeed, greater time spent freezing appears to correlate to more conditioned suppression (Bouton and Bolles, 1980, Mast et al., 1982). Despite involving similar behaviors, there is support for the idea that conditioned freezing and conditioned suppression are mutually exclusive behaviors that are mediated by different neural circuits (Amorapanth et al., 1999, Blanchard et al., 2001, Killcross et al., 1997, McDannald, 2010, McDannald and Galarce, 2011).

The bed nucleus of the stria terminalis (BNST) is an extended amygdala structure that is cellularly heterogeneous but predominantly GABAergic (Bota et al., 2012, Siletti et al., 2022, Welch et al., 2019). In comparisons between few pairings of backward and forward conditioning, the BNST is strongly implicated in the processing and conditioned responding to ambiguous threats such as few backward paired CS+ (Goode et al., 2020, Goode et al., 2019, Ressler et al., 2020). Here, we compared conditioned suppression between a few and many pairings of forward and backward CS + that predicted an aversive stimulus and investigated the role of BNST GABA neurons in conditioned suppression evoked by backwards CS + . Consistent with the trial-based model of the sometimes-opponent process theory, the forward CS + suppressed sucrose consumption regardless of the number of pairings, and a few pairings of backward CS + but not many pairings of the backward CS + suppressed sucrose consumption. Diazepam selectively prevented conditioned suppression to few pairings of the backward CS + and not to the forward CS + . Fiber photometry GCaMP recordings showed that BNST GABA neurons signaled sucrose consumption during few pairings of the backward CS + but not to sucrose consumption without the backward CS+, or the onset or offset of the backward CS + itself. Designer receptor activation of BNST GABA neurons, but not designer receptor inhibition of BNST GABA neurons, abolished backward conditioned suppression. The results described here suggest that, while both forward and backward conditioning can produce conditioned suppression, 1) they are distinguishable by their dependency on diazepam and the behaviors that occur during cue presentation during conditioned suppression; and 2) backward conditioned suppression is dependent on both positive allosteric modulation of GABA on GABAA receptors by diazepam and the activation of BNST GABA neurons.