Rats were trained on a delayed matching to position (DMTP) task that embedded either a differential outcomes procedure (DOP) or a non-differential outcomes procedure (NOP). The DOP, via Pavlovian conditioning (stimulus–outcome associations), results in the use of unique reward expectancies that facilitate learning and memory performance above subjects trained with a NOP that requires subjects to retain cue information for accurate choice behavior (stimulus–response associations). This enhancement in learning and/or memory produced by the DOP is called the differential outcomes effect (DOE). After being trained on the DMTP task, rats were implanted with two cannulae aimed at the basolateral amygdala (BLA) nuclei. Rats trained with the DOP, relative to those trained with the NOP, displayed enhanced short-term memory (STM) performance under vehicle conditions (i.e. the DOE). However, injections of the γ-aminobutyric acid (GABA)A agonist muscimol into the BLA dose-dependently (0.0625 and 0.125 µg) impaired STM performance only in DOP-trained rats. These results support the role of the BLA in the use of established reward expectancies during a short-term working memory task. Despite the fact that extracellular signal-regulated kinase/mitogen-activated protein kinases (ERK/MAPK) have been shown to be necessary for amygdala-dependent long-term potentiation and some forms of long-term and STM, inhibition of the ERK/MAPK signaling cascade by U0126 (2.0 or 4.0 µg) in the BLA was not critical for updating the STM of either spatial information or reward expectation.