The internal compass of the brain, also known as the head-direction (HD) system, forms a crucial component of the neural circuit necessary for effective spatial orientation, and includes regions such as the anterodorsal thalamic nucleus (ADN). It maintains a sense of orientation by integrating self-motion with local cues, which serve as anchors. A recent study in mice showed that network gain, a measure of population activity in the system, decreases during reorientations induced by changes in visual cues. The extent of this reduction in gain influences the dynamics of reorientation, with lower gain leading to faster resets. We hypothesize that gain can serve as a measure of the ‘uncertainty’ in orientation, and it reflects how confident an animal is in its orientation estimate. Thus, manipulating the degree to which a visual cue accurately predicts orientation would influence gain. We used miniscopes to perform Ca2+-recordings in the ADN of freely behaving mice while varying the separation between two identical cues in an attempt to induce uncertainty. We observed that the representation in the ADN switches between two states, which was indicative of the mice switching between anchoring to the two cues. We also observed that the gain remained low during these switch states. By introducing noise in a 1D continuous attractor network (CAN) model, we show that these switches are more prominent when gain is low, and the system maintains a stable representation when gain is high. Thus, we report the presence of unstable equilibria in the HD network, with higher values of gain leading to a robust state. Our findings would help further our understanding of reorientation dynamics in the head-direction system, and how extrinsic cues could impact the same. These findings might have implications on examining how extrinsic cues, network gain and reorientation dynamics could impact spatial navigation strategies.