Abstract Rat behaviors in the elevated T-maze (ETM) were evaluated following tectum microinjections of either glycine (GLY, 1, 10, 80 and 120 nmol) or d-serine (D-SER, 160 and 320 nmol), the putative endogenous agonists of GLY-B site at NMDA receptor, or the respective antagonist 7-chloro-kynurenic acid (7CK, 8 nmol). ETM performance was appraised by two validated scores of anxiety, i.e., the inhibitory avoidance duration (AD) and risk assessment behavior, and two scores derived from a newly developed approach to inhibitory avoidance learning curves, i.e., the learning median number of trials (T 50) and avoidance variability (standard deviation of learning curve). Effects on aversive memory consolidation were assessed through changes in the AD measured 48 h after the full-acquisition of inhibitory avoidance. Drug effects were compared to those of vehicle. In most cases, microinjection of GLY-B site agonists into the dorsal periaqueductal gray (dPAG) produced increases in AD, which were compatible with an increase in anxiety. However, neither the intra-periaqueductal injection of 80 nmol GLY, nor that of 160 nmol D-SER, increased the AD. On the other hand, these microinjections invariably produced a parallel left shift in avoidance learning curves, thereby reducing the T 50 but not the variability. Effects of 120 nmol GLY on AD and T 50 were both antagonized by a previous microinjection of 7CK into the dPAG. The inverse relationship of AD and T 50 suggests that increases in the anxiety level reduce the number of trials required for the acquisition of inhibitory avoidance. The above data also suggest the higher consistency and drug sensitivity of T 50 as compared to the AD. In turn, whereas the microinjection of 120 nmol GLY into the superior colliculus (SC) did not affect the T 50, it increased the AD. On the other hand, there was an increase in avoidance variability following the microinjection of either 120 nmol GLY into the SC or 8 nmol 7CK into the dPAG. Therefore, the GLY-B receptors within these structures seem to play opposite roles on avoidance variability. In contrast, neither of these treatments changed T 50. Finally, whereas the risk assessment was solely decreased by the microinjection of GLY into the SC, the aversive memory was only impaired by the microinjection of 7CK into the dPAG. Overall, these data suggest that NMDA/GLY-B receptors of dPAG mediate both anxiety and aversive memory, while those in the SC are most likely involved with attention and visuomotor components of risk assessment behavior.