Back

Bats decouple sonar gaze from steering to resolve sensory conflict

Finger, N. M.; Chitnis, S. S.; Capshaw, G.; Kaplanoglu, A.; Krishnan, A.; Moss, C. F.

2026-08-13 neuroscience
10.64898/2026.08.08.743555 bioRxiv
Show abstract

When sensory modalities yield conflicting information, animals must rapidly reassess stimuli to select their actions. We induced auditory-visual conflict in free-flying echolocating Egyptian fruit bats, by fitting animals with prisms that shifted the perceived visual location of a landing perch while echoes returned from its veridical location. Bats that course-corrected within a single goal-directed flight did so by decoupling sonar gaze from steering, to enable rapid reweighting of visual and auditory cues. We designed artificial agents that used Bayesian inference to construct estimates of goal locations in their environment. When competing estimates directed active-sensing behaviors distinctly from steering, agents course-corrected more rapidly. Consistent with this idea, when bats were fit with prisms and earplugs that attenuated auditory localization cues, they were unable to course-correct. Removing prisms produced no systematic after-effects. Our framework suggests that instead of correcting their behavior after failure, animals could efficiently employ active sensing to resolve sensory conflict before failure occurs. O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=178 SRC="FIGDIR/small/743555v1_ufig1.gif" ALT="Figure 1"> View larger version (41K): org.highwire.dtl.DTLVardef@16262edorg.highwire.dtl.DTLVardef@4cd82aorg.highwire.dtl.DTLVardef@103e6fforg.highwire.dtl.DTLVardef@13292bf_HPS_FORMAT_FIGEXP M_FIG O_FLOATNOCover figure.C_FLOATNO Bat wearing helmet with clear-glasses and tracking markers. Photograph (C) 2026 Nikita M. Finger / Moss Laboratory. C_FIG

Matching journals

The top 6 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.