Small mismatches get blended; large mismatches force a choice
In a virtual-reality heading task, participants turned in the dark with only a brief flash of visual feedback along the way. When the visual feedback was offset only slightly from the true heading, people adjusted their turn as if they were averaging the body-based sense of rotation with the visual cue. When the offset was large, they behaved as if the two cues were competing rather than being merged [1]. The modeling pointed to a hybrid strategy: combine when the gap is small, compete when it is large [1].
This is not a simple case of one cue always winning. The same person could blend cues on one trial and favor one source on another, depending on how far apart the cues were [1]. The practical implication is that conflict resolution is not a fixed hierarchy but a threshold-like rule: agreement invites integration, disagreement invites a contest.
Reliability and prior knowledge shift which cue wins
When landmarks were rotated to conflict with the path walked, adults' use of each cue depended on how precise each cue was and on what they already knew about the environment [4]. That fits a reliability-weighting picture: the cue that seems more trustworthy at that moment gets more influence. In a separate line of work, humans integrating path integration with environmental cues adjusted how much weight they gave each source when the two were put in conflict [5].
The broader principle is that conflict resolution is not just about the size of the disagreement. A landmark that is precise and relevant can pull the estimate toward it even when path integration says otherwise, while a less reliable landmark may be discounted [4][5]. This helps explain why two people can experience the same conflict and resolve it differently.
More landmarks is not always better
In triangle-completion experiments that varied the number of landmarks from an open steppe to a dense forest, homing was best with three landmarks arranged around the goal. Beyond three, people diverged sharply: some were unharmed by the extra landmarks, while others seemed to be impaired in using landmark information [2]. The authors suggest that success in complex settings depends on identifying the correct clearing around the goal—some participants may not be able to see the forest for the trees [2].
This qualifies the simple reliability story. Adding more landmark cues does not automatically improve conflict resolution; in cluttered environments, the challenge may be selecting the right landmarks rather than weighting them. The same conflict between path integration and landmarks can therefore be resolved well or poorly depending on how well the landmark layout is parsed [2].
What this means when you are actually lost
If you are walking and your sense of direction disagrees only slightly with what you see, the evidence suggests you will blend the two and keep moving smoothly [1]. If the disagreement is large—say a landmark appears far from where your internal path says it should be—you are more likely to treat them as rivals and pick one. Which one you pick will depend on how reliable each seems and on your prior knowledge of the place [4][5].
The picture from animal work is consistent with context-dependent integration rather than a single fixed rule: in mice, landmark and self-motion cues interact to shape spatial firing patterns, and the integration can shift with context [3]. That does not mean human and mouse navigation are identical, but it supports the idea that the nervous system can switch between combining and favoring cues depending on the situation [3]. For a curious navigator, the takeaway is that your brain is not blindly averaging signals—it is running a flexible, mismatch-sensitive arbitration between what your body tells you and what the world shows you.
Sources used in this answer
Combination and competition between path integration and landmark navigation in the estimation of heading direction
In a virtual heading task, participants used a hybrid strategy: they combined body-based and visual cues when the mismatch was small but treated them as competing when the mismatch was large, with considerable individual differences in overall reliance on visual feedback.
Not seeing the forest for the trees: combination of path integration and landmark cues in human virtual navigation
In virtual triangle-completion tasks, homing was best with three landmarks around the goal; with more landmarks, individual differences emerged, with some participants impaired in using landmark information, suggesting difficulty identifying the relevant clearing in complex environments.
Principles governing the integration of landmark and self-motion cues in entorhinal cortical codes for navigation
In mice, landmark and self-motion cues interact in a context-dependent way to generate entorhinal cortical firing patterns and behavioral position estimates during navigation.
The Processes of Path Integration and Landmark-based Navigation in Human Adults
In human adults, cue usage during conflicting landmark and path-integration conditions was influenced by the relative precision of the cues and by prior knowledge.
Optimal combination of environmental cues and path integration during navigation
Humans integrating path integration with environmental cues adjusted cue weighting when the cues were placed in conflict, and the size of the conflict influenced homing direction.