Online extinction and novelty-triggered recovery of life-long visual memories in navigating ants

Online extinction and novelty-triggered recovery of life-long visual memories in navigating ants | Research Square Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Online extinction and novelty-triggered recovery of life-long visual memories in navigating ants Leo Clement, Cody A Freas, Antoine Wystrach This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9010492/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 9 You are reading this latest preprint version Abstract Insects exhibit flexible memory processes, yet current models often confine these modulations to discrete, reinforced events. During navigation, animals can learn to associate locations with resources or aversive events. However, such events occur sporadically within an otherwise continuous behavioral process involving latent learning. This raises a fundamental question for insect cognition: can insects continuously update visual memories during navigation? We show that navigating ants ( Myrmecia midas ) dynamically revalue visual memories en-route, even in the absence of any motivational change or external reinforcement. Using a trackball system that prevented visual progress, we found that nest-view attraction declined over time, reflecting a gradual, online extinction of these otherwise life-long attractive visual memories. These memory updates were view-specific, independent of path integrator state, and persisted after 24-hours, indicating the involvement of long-term memory processes. However, brief exposure to a novel visual scene was sufficient to partially restore nest-views attraction, showing that extinction is labile, and the original memories are preserved. Unlike classic paradigms, visual memory extinction operates here as a continuous, time-dependent process without reward-omission or prediction error; whereas memory reinstatement is triggered by discrete visual changes. We propose that these dynamics are mediated by opponent-processes in the insect mushroom bodies, where temporal estimates of view exposure drive extinction and novel view detection gates memory recovery. inhibition context-dependent recovery visual navigation mushroom bodies latent learning novelty detection Figures Figure 1 Figure 2 Figure 3 Figure 4 Introduction Learning theory posits that associations between conditioned stimuli (CS) and biologically significant outcomes (US) are not fixed but dynamically shaped by experience. When a CS is repeatedly encountered without its expected US, such as a feeding location that no longer yields reward, the responses gradually weaken, in a process known as extinction learning. Importantly, extinction does not erase the original memory but instead forms a parallel inhibitory memory trace, allowing extinguished responses to recover following time or contextual change. These dual-memory systems enable adaptive flexibility, balancing short-term suppression with long-term retention. Across taxa, including insects, such revaluation is mediated by dopaminergic circuits signalling prediction errors. In vertebrates, memory systems operate online during behaviour, with hippocampal spatial representations and dopaminergic prediction-error signals dynamically revaluing memories (Schultz 2016 ; Stachenfeld et al. 2017 ). In insects, studies show the formation of parallel aversive and appetitive memory traces, mediated by recurrent dopaminergic circuits in the mushroom bodies. Extinction dynamics within this known circuitry are well-established and occur when a previously reinforced stimulus is no longer presented with the reinforcer (Aso et al. 2014 ; Eisenhardt 2014 ; Felsenberg et al. 2018 ). However, such extinction is typically studied in a discrete, trial-based context, where the presence or absence of reinforcement events is clearly structured. Navigation presents unique challenges, where route memories guiding directed movement are formed in a continuous process through latent learning, and without explicit reinforcement (Clement et al. 2024 ). In contrast to associative learning paradigms, there is no clear reward or reward-omission signal during navigation between goals. We therefore asked, can such route-based memories undergo a form of online extinction during continuous, unreinforced experience? Ants rely on sophisticated visual memory systems to establish and adjust their foraging routes, via acquiring multiple view memories which guide movement to goals (Clement et al. 2024 ; Mangan and Webb 2012 ; Schultheiss et al. 2016 ; Wystrach et al. 2020a ; Freas et al. 2019 ). While navigational memories are known to be malleable, their modulation is typically studied as a discrete process tied to reinforcers or motivational changes (Freas et al. 2022 ; Wystrach et al. 2020a ). For example, aversive learning studies show that memories can be altered by negative experiences along the route or failure to reach goals but these changes are usually assessed across multiple trips, not within them (Collett 2014 ; Deeti et al. 2023 ; Freas et al. 2022 ; Le Moël and Wystrach 2020 ; Murray et al. 2020 ; Schwarz et al. 2020 ; Wystrach et al. 2020a , 2019 ). “Rewinding" experiments, where ants are displaced further back along their route, causing them to experience the same views multiple times without reaching the nest, suggest within-trip memory adjustments (Collett 2014 ; Schwarz et al. 2020 ; Wystrach et al. 2019 ). However, such manipulations involve discrete visual changes and capture events, making it difficult to disentangle purely continuous online updating from memory changes driven by experimental interruption or consolidation during transfer. Recent work exploring latent learning in ants has shown that visual memory acquisition in a novel environment (the initial forming of the memory) occurs continuously during route exploration even in the absence of explicit reward or punishment, yet its behavioral expression is gated during the initial acquisition phase (Clement et al. 2024 ). This suggests that the insect brain can support the continuous acquisition of visual memories en-route. Yet a key gap remains; latent-learning memory, by definition, is not used as it is being formed, thus its expression is only revealed during subsequent tests of route knowledge. This makes functional sense for an ant learning a novel scene, because an instantaneous recall of visual memories as they are being built would prevent the insect from distinguishing what is familiar from what is new. However, a more gradual recall, slowly increasing through time, could be viable and potentially beneficial when it comes to ignoring previous memories that turn out to be maladaptive given evidence building through time (Clement et al. 2024 ). What remains unclear is whether such modulation reflects cumulative online updating during navigation, or discrete changes occurring between navigational episodes. To investigate these questions, we used a trackball system (Clement et al. 2025 , 2024 ; Dahmen et al. 2017 ; Murray et al. 2020 ) to test if navigating bull ants ( Myrmecia midas ) can update their view memories enroute when walking in the goal direction is possible but actual visual progress is prevented. Ants could rotate freely on the trackball while on their established route, but translational movement did not result in forward displacement, preventing visual progress. This allowed us to isolate the effects of time and experience on visual memory expression in the absence of external scene changes. M. midas is an ideal model for this question, as it relies heavily on life-long visual route memories and often ignores alternative navigational strategies, including path integration (Freas et al. 2018 , 2017a ) which is typical for ants of this genus (Narendra et al. 2013 ; Narendra and Ramirez-Esquivel 2017 ). We found that ants’ orientation toward nest-aligned views deteriorated over time, indicating an online inhibition of specific view memory attraction that overrides their long-term attractive valence. Critically, this behavioural change could not be explained by the accumulation of the path integrator. Comparisons between ants trained on open versus cluttered routes revealed that these updates were highly view-specific (Supplemental Figure S1 ) rather than reflecting a general reduction in motivation, implicating Kenyon cell–based representations and thus the mushroom bodies as the likely site of modification. Strikingly, these changes persisted for at least 24 h, supporting their integration into long-term memory valence rather than short-term habituation. However, transient exposure to a novel scenery, triggered a partial recovery of the original route-following behaviour when returned to the previously extinguished scene, indicating that this revaluation remains reversible within a single experience. Materials and Methods Study Species Myrmecia midas is a nocturnal bull ant that navigates via terrestrial visual cues along established routes between the nest (typically at the base of a tree; Deeti et al. 2024 ) and foraging trees. While they possess a celestial compass, panorama view memories are the primary navigational cue (Freas et al. 2017a , 2017b ; Freas and Cheng 2018). Study site and collection Experiments were conducted from February to May 2024 on two Myrmecia midas nests on the Macquarie University Campus in Sydney, Australia (33°46 11S, 151°06 40E). For all experiments, inbound foragers were collected in the pre–morning twilight as they descended their foraging tree and individually held with sugar water within phials until testing. At Nest 1 all foragers were collected from a foraging tree 5.8m from the nest and at Nest 2 all foragers were collected from a foraging tree 7.2m from the nest entrance. Testing began each day at sunrise and ran until 14:00h. Tested ants were marked with paint (POSCA™) and released back to the nest. Trackball set–up To record foragers’ paths, individual foragers were tethered upon one of two trackball devices (Dahmen et al. 2017 ; Murray et al. 2020 ). Each device consisted of a polystyrene ball levitated above an aluminium cup by constant air stream provided by a pump. These devices were similar in their construction and recording method but differed in their size; Trackball One: 5cm in diameter; Trackball Two: 10 cm in diameter (Murray et al. 2020 ). Both Trackballs were used during the one-hour extinction test (randomly assigned to individuals across both familiar and unfamiliar conditions), while only Trackball Two, which is permanently housed at Macquarie University, was used to conduct the open/cluttered and re-exposure extinction experiments. Each device had two sensors placed at 90° to the azimuth of their spheres, which allowed the recording of the ball’s movements at 30Hz and allowed translation of this motion into X and Y data to reproduce the forager’s path. Horizontal movement of the ball was restricted by two small wheels. To tether a forager into place atop the polystyrene ball, magnetic paint was applied to the thorax of each ant. The tethering apparatus which held the ant in place consisted of a 3mm magnet which was attached to 5mm of dental thread. This thread was attached to a 0.5mm diameter carbon pin housed within a glass capillary (Clement et al. 2025 ; Clément et al. 2023 ; Dahmen et al. 2017 ). This allowed tethered ants to control both their gaze direction and rotate their body orientation upon the ball, with forward or backwards movement being translated to the ball’s rotation. Importantly this means that while an ant is tethered within the system, translational movement results in no visual feedback or ‘progress’ in reaching the goal location. Experimental Designs Open and Cluttered route experiment To test online visual memory modulation and its view specificity, we examined whether the extinction of specific views at one location influenced visual memory valence along the entire foraging route. Open, uncluttered routes maintain high visual similarity across large distances, whereas cluttered routes contain prominent local objects that make views along the route visually distinct (Zeil 2012 ; Zeil et al. 2003 ). This allows us to assess whether memory extinction is view-specific or represents generalized reductions in homing motivation across the route. At Nest 1, the route was naturally clear of local visual clutter (Supplemental Figure S1 ). In this open environment, foragers were placed upon the track ball 1.5m from the nest and their paths were recorded for 30min. After 30min, a cover was placed over the ant, blocking all visual cues for 1min, and the trackball was randomly assigned and transported to one of three test locations along the foraging route: 1.) Nest , the same location 1.5m from the nest, 2.) Mid , the middle of the foraging route 3m from the nest, 3.) Far , the furthest point (5.8m) from the nest at the base of the foraging tree. Once the trackball was placed at one of these locations, the cover was removed, and paths were collected for 5min. After the Open condition was complete, visual clutter was added to the foraging route. Wooden boards were erected along the route, focused on creating visual contrast between the Nest and Mid sites (Supplemental Figure S1 ). As this represented a large visual change to the route, which M. midas is highly sensitive to (Islam et al. 2020 ), foragers were allowed to become familiar with the new visual route for 28 days. After familiarisation, a separate group of ants were tested on this cluttered route with an identical methodology to open route testing. Foragers were placed upon the trackball 1.5m from the nest for 30min, covered, then tested at one of the three test locations ( Nest, Mid or Far). 24-hour re-exposure experiment To test whether view memory extinction persists over 24 hours, indicating long term changes in memory valence rather than short-term habituation, we tested view-extinct foragers at the same site 24h after extinction. At Nest 2, foragers were randomly assigned to one of two groups: Re-exposure or Control . On the morning of collection, Re-exposure individuals were tethered to the trackball ~ 1.0m from the nest entrance (Supplemental Figure S1 ) and their paths recorded for 15min (initial extinction phase). Following this, ants were held in darkness for 24 h. On the next day, they were placed back on the trackball at the same location for a second 15min recording. Control group individuals were captured and held in darkness for 24h without initial trackball exposure. They were then tested for the first time on the trackball alongside the Re-exposure group’s second test, to ensure motivation and orientation performance were comparable. one-hour view exposure At Nest 1, collected individuals were randomly assigned to one of two conditions: familiar and unfamiliar (for Unfamiliar testing see Supplemental Motivation Control) . Individuals in the familiar condition were tethered upon the trackball at the mid-point of their foraging route, ~3m from the nest and their foraging tree. Tethered individuals were allowed to travel in a chosen direction, with their paths collected for 60min. After this period, a cover was placed over the ant which blocked all visual cues. While covered, the trackball was transported to a distant, unfamiliar location (Unfamiliar Site B; Supplemental Figure S1 ) where the tethered ant was uncovered, and their path was collected at this site for 5min before being re-covered for another 5min while the trackball was transported back to the original familiar location along the route. Once the trackball was placed back at the centre of the foraging route and the cover removed, then the test began with individual paths collected for 5min at this site. Data extraction and analysis All statistical analyses were run using the open-source software R (v 3.6.2. R Core Development Team). For all statistical tests, the p-values were compared to the critical alpha risk at 0.05, with the appropriate correction if needed. The statistical parameter's mean and the associated standard error is given within the text and/or in figures: N represents the number of individuals. As we used two trackball configuration recording at different frequency rates, we down sample path on one of them to match 60 hz. We processed the path by first smoothing the x, y coordinate using a Savitzky-Golay filter (from R “trajr” package) with a window length of 1 seconds. Then, we extracted the angular velocity time series, which was smoothed twice: first with a moving median, then with a moving mean, both with a window length of 0.2s). All parameter estimates from our analysis were obtained by segmenting our data path into non-overlapping windows of either 180 or 10 seconds. Analysis of the ants’ direction of movement To determine extinction learning of specific views, our focus was on the direction of movement, which is a direct measurement of the ants’ ability to use view to direct her path. Within each segmented window we determine the mean direction of movement (mu) as well as the mean circular vector length (r, a measure of dispersion) of each individual. The mean directions (mu) were analyzed using a Rayleigh test (from R package: Circular) that also includes a theoretical direction (analogous to the V-test). To test whether the angular data are distributed uniformly as a null hypothesis or if they are oriented toward the theoretical direction of the nest as indicated by the familiar panorama. As a direct reading of extinction learning, we estimated a path directional error, defined as the absolute angular error between the path orientation and the nest direction. These path directional errors were compared to the 90° threshold (which corresponds to a random direction taken by the population) using a Wilcoxon-Mann-Whitney test. Analysis of the ants’ path integrators effect To control for the effect of the path integrator within the window, we extracted the central value of the X path. This value represents the relative position of the path integrator from the ant's perspective. A negative value indicates that the ant has followed its route and accumulated a negative vector pointing away from the nest. Conversely, a positive value indicates that the ant is heading away from the nest, thus accumulating a vector pointing towards the nest. If these reductions in nest-ward headings were due to an accumulation of a negative vector, we should expect that long negative vector should direct ants' path to the opposite direction of the nest. To test for this effect, we used correlation between error and X path integration value. Statistical models For each experiment, we tested: the effect of the time spent on the path directional error in linear mixed model. As our individuals were recorded across several conditions (across time during extinction and test), the models are mixed models that controlled the individuality effect as a random variable. It should be noted here that in case of deviation of the residuals of these models from normality and or homoscedasticity, the response variables have been transformed. Model were subsequently analyzed via an analysis of variance (Anova), followed by a post-hoc analysis of Tukey's rank comparison. Results Orientation declines during prolonged exposure to fixed scene During the 30min tethering period, homing ants in both Open and Cluttered conditions initially oriented nest-ward (First 3min: V-test at 0°, p < 0.001, Fig. 1 B, dark green). Yet over the 30min tethered on the trackball, ants showed a decline in nest-ward orientation, marked by progressive increases in angular error (ANOVA: First 3min vs Last 3min: F (1,118) = 74.941, p < 0.001), such that by the last three minutes (27 to 30min), their headings were no longer aligned with the nest (Last 3min: V-test at 0°, p = 0.980 Fig. 1 B). This decline of nest-ward orientation was not due to a general loss of motivation caused by the time spent walking on the trackball. Indeed, a control group showed excellent nest-ward orientation when the trackball is placed on the familiar route after spending 1hour walking on the trackball in an unfamiliar environment (see Supplemental Motivation Control). Therefore, the decline of nest-ward orientation results from the sustained exposure to the familiar visual scene in the absence of reward or navigational progress. To disentangle whether this drop in nest-ward orientation reflects a reduction in the use of homing views in general or is specific to the views experienced during the 30min exposure period, we performed two different displacement conditions. After the 30min extinction period, ants, still tethered on the trackball, were covered in darkness for 1min while the apparatus was moved to one of three locations along their familiar route (Fig. 1 A): 1.) The same location near the nest where the 30min exposure happened; 2.) Mid site, 3m from the nest along the route; 3.) Far site, 6m from the nest near the foraging tree, marking the beginning of these ants’ ground-based homing route. Ants tested at the extinction location (1.5m from the nest) remained non-nest oriented (consistent with the end of the view extinction phase; Fig. 1 C, dark green; V-test at 0°, p > 0.05). In contrast, ants tested at the Far site, in both Open and Cluttered conditions, showed a strong reduction in angular error compared to the end of the extinction period (Fig. 1 C, purple; ANOVA Nest vs Far : F (1,38) = 11.213, p = 0.002) and recovered the nest-ward orientation (V-test at 0°, p = 0.002). Thus, moving the ants into a sufficiently different visual scene along their familiar route restored visual homing as these view memories had not been part of the extinction phase. Mid -site testing revealed a difference between open and cluttered routes. In the Open condition, where the scene in Mid remains relatively similar to the exposure site, ants kept high angular errors and exhibited a lack of nest-ward orientation (V-test at 0°, p > 0.05). In contrast, in the Cluttered condition, where the scenery in Mid was highly different from the extinction site (Supplemental Figure S1 , Cluttered ), ants recovered their nest-ward orientation (V-test at 0°, p = 0.007; Fig. 1 D, dark grey) and showed significantly lower angular error than ants in the Open condition (ANOVA Mid Open vs Mid Cluttered : F (1,21) = 6.077; p = 0.023, Fig. 1 D, light grey). Path integration cannot explain these differences, as ants released in Far (both Open and Cluttered ) or Mid ( Cluttered ) oriented nest-ward despite similarly large conflicting path integrator vector accumulated during the exposure period (PI length mean ± SE: Far = 38.61 ± 3.85m; Mid-Open = 38.34 ± 5.99 m; Mid-Cluttered = 53.46 ± 6.63 m). These results demonstrate that the declines in nest-ward homing observed during the extinction period is view-specific rather than a global reduction in homing motivation, a general exhaustion of visual-homing memories, or a conflict with the path integrator. This shows that sustained exposure to familiar views result in a progressive drop in the recall of the memories associated with these views. Ants perform a time-driven, online modulation of the expression of these memories. Loss in view memory expression is long lasting Control and Re-exposed show a clear nest-ward headings during the first 3min on the trackball- (1st Exposure 0h: V-test at 0°, p < 0.001; Control (1st Exposure 24h): V-test at 0°, p < 0.001 ). With comparable orientation during their initial exposure to the trackball (ANOVA post hoc tests, 1st Exposure 0h 0-3min vs Control (1st Exposure 24h), p = 0.684, Fig. 2 A, B, green and light blue). This confirms that the 24-h delay in darkness itself did not reduce homing motivation. As expected, view exposure on Day 1 caused Re-exposed ants to display a gradual decline in orientation over the 15min period, with the final three minutes (12–15min) showing significantly increased angular error (ANOVA post hoc tests 1st Exposure 0h 0–3min vs last 3min, p < 0.001; Fig. 2 A,B, light blue) and a loss of nest-ward orientation (1st Exposure 0h last 3min V-test at 0°, p > 0.05). Crucially, when tested again on Day 2 (2nd Exposure 24h), Re-exposed ants remained non-oriented to the nest (V-test at 0°, p > 0.05; Fig. 2 B). Their angular error was both unchanged relative to the last 3min of Day 1 and significantly higher than that of Day-2 Control ants (ANOVA post hoc tests 2nd Exposure 24h 0-3min vs 1st Exposure 0h last 3min: p = 1.0; 2nd Exposure 24h 0–3min vs C ontrol (1st Exposure 24h): p < 0.001; Fig. 2 A, dark blue). Thus, orientation declines induced during the exposure phase are not a short-lived effect but reflect long-lasting modification of the expression of the visual memories involved. Exposure to a New Scene Triggers Recovery of Visual Memory We next investigated whether this progressive loss in memory expression reflects a transient extinction where the original memory trace remains and can be reinstated, or a forgetting process where the original memory is being erased or overridden for good. The 24-h re-exposure experiment showed that withholding visual input for a full day between sustain exposure period and test did not restore the expression of the view memories. If the original memory trace is preserved, what kind of event could reinstate its expression? Consistent with recent findings in visual latent learning (Clement et al. 2024 ), we hypothesised that experiencing a discrete scene change towards an unfamiliar area may provide such a trigger. We thus tested whether a brief exposure to a novel scene could reinstate memory expression upon return to the familiar site. Identical to our previous experiments, ants initially oriented strongly toward the nest (Fig. 3 A; first 3min: V-test at 0°, p < 0.001) but mean angular error climbed steadily over time, reaching random chance levels (> 90°) by ~ 20 min, when ants were no longer significantly oriented toward the nest (Supplemental Figure S2 ). This significantly increased angular error, coupled with a lack of nest-ward orientation persisted for the remainder of the one-hour recording (Fig. 3 A; first 3min vs. last 3min: post-hoc tests, p < 0.001; V-test at 0°, p = 0.862). Importantly, control experiments with ants tested for the same duration at an unfamiliar site rule out reduced motivation and path-integrator conflict as causes of this effect (see Supplemental Motivation Control, Supplemental Figure S1 , S2, S3; full statistics reported in the supplement). After this one-hour view exposure period, tethered ants were covered and transported in darkness (while remaining on the covered trackball setup) to an unfamiliar site (Site A, Supplemental Figure S1 ). There, the cover was removed, and individuals were let to navigate on the trackball in the novel unfamiliar visual scene for five minutes before being re-covered and returned to the familiar route location for testing. Here, ants showed some evidence of the expression of the original long-term attractive memory of their familiar route: their angular error relative to the nest decreased significantly relative to the end of the exposure period (last 3min vs. test, ANOVA: F (2,26) = 26.111, p < 0.001; post-hoc: p > 0.05), although it remained more elevated than their initial baseline before the sustained exposure (First 3min vs. test: post-hoc p = 0.005). Overall nest-ward orientation was recovered, although this significance was borderline (Fig. 3 , V-test at 0°, p = 0.042). The partial recovery of nest-ward orientations cannot be attributed to the influence of the accumulated path integrator (see Supplemental Figure S2 B,C 1h familiar), nor can the absence of a full recovery be attributed to a general drop of motivation (see Supplemental Figure S3 1h unfamiliar, Motivation Control). This orientation pattern shows that the original memory trace was not erased, and can be, at least partially, reinstated. Therefore, the process at play corresponds to a form of extinction of route memories, whose expression decays steadily with exposure time but can be reinstated after a transient exposure to unfamiliar scenes. Discussion While navigating home, Myrmecia midas ants prevented from making translational progress along their familiar route, and thus exposed to the same visual scene for an extended period, showed a gradual deterioration in their initially accurate, nest-aligned headings. This progressive loss of nest-ward orientation indicates an ongoing modification of the visual memory expression used for homing. Importantly, this effect was view-specific rather than reflecting a general decline in homing motivation or a global disruption of route memory use (Supplemental Figure S2 , S3), and it persisted for at least 24 h. Yet the change remained labile, as a brief exposure to a novel visual scene partially reinstated the original scene’s expression of nest-ward attraction. Together, these findings align with the theoretical framework of memory extinction in several key respects. First, the expression of a memory associated with a specific stimulus (here, the nest-ward heading evoked by a given familiar visual scene) decreased progressively as ants experienced that stimulus without reinforcement (Fig. 1 ). Second, the effect was long lasting (at least 24 h; Fig. 2 ) and may therefore involve long-term consolidation processes that typically require protein synthesis (Eisenhardt 2014 ). Third, rather than reflecting simple forgetting, the original memory expression was recoverable (Fig. 3 ), consistent with the idea that memory extinction is mediated by the formation of a parallel, antagonistic memory trace rather than erasure of the original memory trace (de Bruijn et al. 2021 ). However, our results also differ from canonical extinction paradigms in several noteworthy respects. Most extinction studies rely on Pavlovian conditioning (Dunsmoor et al. 2015 ), where an initially neutral conditioned stimulus (CS) becomes predictive through its association with a reinforcing unconditioned stimulus (US), and extinction is induced by repeated non-reinforced CS presentations. Here, no discrete reinforcement is involved. The visual memory guiding nest-ward orientation is not the product of explicit Pavlovian training but instead arises from latent learning (Clement et al. 2024 ), which operates continuously and without discrete reinforcement events. Likewise, extinction was not induced by repeated discrete trials, but emerged gradually during prolonged exposure to the same scene (Fig. 1 – 3 ; Supplemental Fig. 2). This implies that memory expression is being modified while it is actively controlling behaviour. Finally, recovery of the original memory, which is often spontaneous and time-dependent in classical paradigms (e.g., de Bruijn et al. 2021 ; Dunsmoor et al. 2015 ; Felsenberg et al. 2018 ), could occur abruptly here and was triggered by a discrete event: brief exposure to a novel scene (Fig. 3 ). Below, we discuss these features in turn. Extinction and recall occur simultaneously Most evidence for extinction comes from associative learning paradigms in which behavioural change is assessed across clearly separated learning and test phases (e.g., Bouton 2004 ; Dunsmoor et al. 2015 ). In insect navigation, extinction-like effects have been described in bees visiting no-longer-rewarded feeders, where visit frequency declines across successive trials (Eisenhardt 2014 ). In such cases, extinction is expressed across discrete episodes of reinforcement-based learning. Because behavioural expression is typically measured only after rest periods or after a return to the nest, it remains unclear whether memory valence is modulated online during recall, or is instead updated between trials during consolidation windows. “Rewinding” experiments in ants, where individuals are repeatedly captured and replaced further back along their route, show a decline in route memory expression even with short inter-trial intervals (Collett 2014 ; Schwarz et al. 2020 ; Wystrach et al. 2019 ). However, these manipulations involve abrupt scene changes and repeated handling, which could themselves trigger learning or retrieval processes. This makes it difficult to distinguish continuous, online modulation from trial-like resets imposed by the experimental procedure. In the present study, across all three experiments, the decline in nest-ward orientation emerged gradually while ants remained in the same visual scene and continued to express the associated visual memory (Supplemental Figure S2 A). This online deterioration demonstrates that memory change can occur concurrently with recall. Even if extinction depends on the formation of a parallel inhibitory trace, its behavioural impact must emerge as that trace is being built. This resembles observations in Drosophila larvae aversive conditioning, where avoidance can be expressed while learning is still ongoing (Schleyer et al. 2011 ). In contrast to appetitive conditioning, the aversive US must remain present for behavioural expression (Schleyer et al. 2011 ). Our findings show that similarly “online” dynamics apply to extinction-like processes in a navigational context. A key open question concerns when and how consolidation occurs, given that extinction effects persisted for at least 24 h. Extinction without US omission or prediction error A common mechanistic account of extinction is that after Pavlovian conditioning, presentation of the CS in the absence of the expected US generates a prediction error, which drives formation of the extinction antagonist memory trace (Dunsmoor et al. 2015 ). Consistent with this, extinction in Drosophila has been shown to involve formation of a parallel memory when the CS is presented without the US, potentially driven by prediction-error signalling (Eschbach et al. 2020 ; Felsenberg et al. 2018 ). In insects, parallel associative memories of different valence and timescales can be formed in the mushroom bodies through dopaminergic modulation (conveying the US), producing bidirectional plasticity at Kenyon cell (conveying the CS) to MB output neuron (conveying the CR) synapses (e.g., Aso et al. 2014 ; Cognigni et al. 2018 ; Handler et al. 2019 ). Complex feedback circuits between these actors can support the formation of prediction errors (Eschbach et al. 2020 ; Terao and Mizunami 2017 ; Webb 2024 ). Although much of this framework comes from olfactory conditioning in Drosophila , a similar logic in the mushroom body applies to visual learning in the context of insect navigation (Ardin et al. 2016 ; Buehlmann et al. 2020 ; Kamhi et al. 2020 ; Webb and Wystrach 2016 ; Wystrach et al. 2020b ). Notably, the memory dynamics of extinction may not be identical across insect species. For instance, olfactory memories in ants show remarkable resistance to extinction protocols (Piqueret et al. 2019) whereas our results suggest that visual memories are highly susceptible to time-driven extinction. Such a divergence may reflect the distinct ecological demands of each modality, where the functional necessity for stable olfactory cues contrasts with a visual navigation system that must remain highly plastic to filter out outdated or inaccurate information. However, the extinction observed here cannot be explained by prediction error. The original attractive route memory was acquired through latent learning, where views along the route are learned continuously without the need for explicit reinforcement such as reaching the nest or finding food (Clement et al. 2024 ). Thus, the visual scenes had always been experienced without reinforcement, so that extinction cannot be explained as a prediction-error response to US omission. This creates an apparent paradox: if the original memory forms without a US, and extinction also develops without a US, what determines whether exposure strengthens memory expression or suppresses it? A potential resolution follows directly from mechanistic accounts of latent learning. Clément et al. (2024) proposed that continuous route learning involves both the ongoing appetitive learning of views via continuous dopaminergic input and in parallel, continuous formation of a transient antagonistic trace that suppresses expression of the appetitive memory, potentially via lateral inhibition between MBONs. Such a mechanism would prevent premature recall during initial learning, allowing the insect to distinguish novel from familiar scenes. Retrieval of the latent memory is then triggered by discrete events such as entering the nest or experiencing a sudden change in scenery, which reset the antagonistic trace and unmask the attractive output. Notably, this framework can also account for our extinction results (Fig. 4 ). When ants are exposed to a familiar route scene, the attractive memory is initially expressed as nest-ward orientation (Fig. 4 A). Due to continuous learning, prolonged exposure to the same scene without progress makes the antagonistic trace can gradually rebuild and increasingly suppress expression of the (already learnt) attractive output, yielding the observed decline in nest-ward attraction (Fig. 4 B). This produces extinction-like dynamics without requiring reinforcement or prediction error which are key components of classical paradigms. Sudden, novelty-triggered recovery of an extinguished memory Extinguished responses often show spontaneous recovery over time, consistent with the idea that the parallel extinction memory trace is transient (Bouton et al. 2006 ; Rescorla 2004 ). In Drosophila , recovery can take days (Hirano et al. 2016 ). Here, we found that the extinction memory remained stable after 24h in darkness, it therefore remains possible that additional time would produce spontaneous recovery. Over longer periods without exposure, visual memories in ants can become functionally discounted as outdated, yet this does not imply memory decay, but rather reflects a reweighting of older visual memories as less reliable relative to more recently experienced cues (Freas and Cheng 2017 ). Crucially, we show that recovery can also be triggered rapidly by a discrete event: brief exposure to a novel visual scene (Fig. 3 ). This pattern is predicted by the latent-learning framework described above. If novelty resets the antagonistic trace, then a transient scene change should remove inhibition and immediately allow the original attractive memory to re-emerge (Fig. 4 C). This event-triggered recovery broadens the range of mechanisms underlying extinction-like behavioural dynamics. One possibility is that novelty-dependent modulation acts by resetting the KC→MBON synaptic input of the extinction memory trace. Under this view, darkness may fail to induce recovery simply because it provides no novel visual input. Partial recovery as a consequence of continuous learning Spontaneous recovery after extinction is rarely complete in classical paradigms (Delamater and Westbrook 2014 ; Dunsmoor et al. 2015 ). Consistent with this, recovery of nest-ward orientation in our ants was only partial: headings were biased toward the nest but less accurate than at the beginning of the procedure (Fig. 3 ). This reduction was specific to ants that experienced the extinction procedure and was not simply due to time spent on the trackball (Supplemental Fig. 2, 3). Partial recovery could reflect residual inhibitory influence or a modification of the original memory trace (or both). The latent-learning model predicts the latter. As ants progressively lost nest-ward attraction, they explored alternative headings (Fig. 3 , 4 B). If continuous learning remained active, then views associated with these novel directions would also be incorporated into the attractive memory (alongside formation of an antagonistic trace). When novelty subsequently resets inhibition, the expressed attractive output would then reflect a mixture of the original nest-ward direction and the additional directions sampled during extinction, yielding a weaker and less precise nest-ward bias. Overall, the model proposed for latent learning and event-triggered recall (Clement et al. 2024 ) can account for both extinction and recovery in our data without additional assumptions. These behavioural results therefore provide converging support for that framework. Inhibition versus aversion A dominant view of extinction is that the CS acquires inhibitory properties that suppress the conditioned response, rather than becoming aversive (e.g., Bouton et al. 2006 ; Larrauri and Schmajuk 2008 ). In insect mushroom bodies, different compartments are often described as supporting appetitive or aversive memories that compete for behavioural control, although their interactions can produce more complex forms of regulation (e.g., Aso and Rubin 2016 ; Cognigni et al. 2018 ; Krashes et al. 2009 ; Perisse et al. 2016 ). In Drosophila , extinction of aversive memories can involve formation of competing appetitive traces, although extinction does not actually result in attraction (Felsenberg et al. 2018 ). This raises the question of whether the decline in nest-ward orientation observed here reflects the gradual build-up of an aversive memory opposing attraction, or merely an inhibitory trace that gates attraction without generating repulsion. Aversive learning in ant navigation can be induced by trapping, which leads to active turns and avoidance of the capture zone (Freas et al. 2022 ; Lionetti et al. 2024 ; Wystrach et al. 2019 ). The resulting opponent attraction–repulsion processes have been proposed to support detours, nest pinpointing, and other navigational behaviours (e.g., Le Moël and Wystrach, 2020 ; Murray et al., 2020 ; Schwarz et al., 2020 ). Our results suggest that the extinction observed here reflects inhibition rather than aversion. As ants were walking on a trackball with no change in visual scenery, an aversive memory “winning” over attraction should produce strong turning or avoidance-like behaviour. We did not observe such avoidance, and several ants randomly resampled the nest-ward orientation after having expressed extinction (Supplemental Figure S2 A, S4). Therefore, the extinction memory appears to suppress attraction without driving repulsion to these views. Functionally, this distinction is important as suppression allows ants to disengage from a failing strategy without triggering maladaptive flight responses in a familiar environment. Functional significance These experiments reveal an extinction-like process in the ethological and operant context of navigation, providing insight into both mechanism and function. Ants learn their foraging routes through continuous visual learning, enabling robust shuttling between nest and foraging sites (Amin et al. 2025 ; Clement et al. 2024 ; Freas and Cheng 2025 ; Mangan and Webb 2012 ; Wystrach et al. 2013 ). Yet our results show that prolonged exposure to a familiar scene gradually reduces reliance on that memory and promotes exploration of alternative headings. Such extinction protects ants from becoming trapped in situations where route following is ineffective, even in the absence of overtly punitive signals: for instance, if an obstacle prevents forward progression or in cases where maladaptive looping trajectories are formed (Clement et al. 2025 ). In such conditions, rigid persistence on the ant’s life-long appetitive route memories could be fatal, whereas extinction-like suppression would release behavioural control to alternative strategies such as systematic search and path integration. If successful, these new directions would then be incorporated into the route memory through continued learning, explaining why, on the next trial, the ‘recovery’ is only partial and differs from the original memory expression. Conclusions We describe an extinction-like process in the context of natural navigation, where prolonged exposure to a familiar visual scene without progress steadily suppresses nest-ward attraction. This extinction does not rely on reward omission or prediction error, but instead fits naturally within the dynamics of latent learning: visual memories can be continuously updated while being expressed, and rapidly re-expressed after discrete novelty-driven reset events. Functionally, this mechanism provides an adaptive way to disengage from an ineffective homing strategy and promote exploration. More broadly, our results highlight that “extinction” is not a single canonical process but can be implemented through multiple mechanisms and timescales, and that the mushroom body circuitry may provide the flexibility needed for such task-specific fine-tuned solutions to evolve. Declarations Competing Interests The authors declare no conflicts of interest. No author serves as an Editor/Board Member for Animal Cognition. The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication. Conflict of Interest Statement The authors declare no conflicts of interest. No author serves as an Editor/Board Member for Animal Cognition. The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication. Statement on Animal Ethics No institutional or governmental ethical approval is required for research involving insects in Australia under the Australian Code for the Care and Use of Animals for Scientific Purposes, which does not apply to invertebrates. All procedures were strictly non-invasive and individuals were returned to their nest after testing. Funding Information This project was funded by a Macquarie University Research Fellowship (MQRF0001094) and by the European Research Council (101125881). Author Contribution L.C., C.A.F., and A.W. conceptualized the study, designed the experiments, performed the research, and wrote the manuscript. L.C. performed the statistical analyses and performed data curation. All authors contributed to the interpretation of the results and approved the final version of the manuscript. Acknowledgements We are grateful to Macquarie University for permission to work on site and access to the nests. Data Availability Raw Data and Code will be made available on github: [https://github.com/ClementLe0](https:/github.com/ClementLe0) . References Amin AA, Philippides A, Graham P (2025) Ant visual route navigation: How the fine details of behaviour promote successful route performance and convergence. PLOS Comput Biol 21:e1012798 Ardin P, Peng F, Mangan M, Lagogiannis K, Webb B (2016) Using an Insect Mushroom Body Circuit to Encode Route Memory in Complex Natural Environments. PLOS Comput Biol 12:e1004683 Aso Y, Hattori D, Yu Y, Johnston RM, Iyer NA, Ngo T-T, Dionne H, Abbott L, Axel R, Tanimoto H et al (2014) The neuronal architecture of the mushroom body provides a logic for associative learning ed. L C Griffith eLife 3:e04577 Aso Y, Rubin GM (2016) Dopaminergic neurons write and update memories with cell-type-specific rules ed. L. Luo. eLife 5:e16135 Bouton ME (2004) Context and behavioral processes in extinction. Learn Mem Cold Spring Harb N 11:485–494 Bouton ME, Westbrook RF, Corcoran KA, Maren S (2006) Contextual and Temporal Modulation of Extinction: Behavioral and Biological Mechanisms. Biol Psychiatry 60:352–360 Buehlmann C, Wozniak B, Goulard R, Webb B, Graham P, Niven JE (2020) Mushroom Bodies Are Required for Learned Visual Navigation, but Not for Innate Visual Behavior, in Ants. Curr Biol 30:3438–3443e2 Clement L, Schwarz S, Mahot-Castaing B, Wystrach A (2025) Is this scenery worth exploring? Insight into the visual encoding of navigating ants. J Exp Biol 228:JEB249935 Clément L, Schwarz S, Wystrach A (2023) An intrinsic oscillator underlies visual navigation in ants. Curr Biol 33:411–422 Clement L, Schwarz S, Wystrach A (2024) Latent learning without map-like representation of space in navigating ants. 2024.08.29.610243. https://www.biorxiv.org/content/ 10.1101/2024.08.29.610243v1 (Accessed January 21, 2026) Cognigni P, Felsenberg J, Waddell S (2018) Do the right thing: neural network mechanisms of memory formation, expression and update in Drosophila . Curr Opin Neurobiol 49:51–58 Collett M (2014) A desert ant’s memory of recent visual experience and the control of route guidance. Proc R Soc B Biol Sci 281:20140634 Dahmen H, Wahl VL, Pfeffer SE, Mallot HA, Wittlinger M (2017) Naturalistic path integration of Cataglyphis desert ants on an air-cushioned lightweight spherical treadmill. J Exp Biol 220:634–644 de Bruijn JAC, Vet LEM, Smid HM, de Boer JG (2021) Memory extinction and spontaneous recovery shaping parasitoid foraging behavior. Behav Ecol Off J Int Soc Behav Ecol 32:952–960 Deeti S, Islam M, Freas C, Murray T, Cheng K (2023) Intricacies of running a route without success in night-active bull ants ( Myrmecia midas ). J Exp Psychol Anim Learn Cogn 49:111–126 Deeti S, Tjung I, Freas C, Murray T, Cheng K (2024) Heavy rainfall induced colony fission and nest relocation in nocturnal bull ants ( Myrmecia midas ). Biologia 79(5):1439–1450 Delamater AR, Westbrook RF (2014) Psychological and neural mechanisms of experimental extinction: a selective review. Neurobiol Learn Mem 108:38–51 Dunsmoor JE, Niv Y, Daw N, Phelps EA (2015) Rethinking Extinction Neuron 88:47–63 Eisenhardt D (2014) Molecular mechanisms underlying formation of long-term reward memories and extinction memories in the honeybee (Apis mellifera). Learn Mem 21:534–542 Eschbach C, Fushiki A, Winding M, Schneider-Mizell CM, Shao M, Arruda R, Eichler K, Valdes-Aleman J, Ohyama T, Thum AS et al (2020) Recurrent architecture for adaptive regulation of learning in the insect brain. Nat Neurosci 23:544–555 Felsenberg J, Jacob PF, Walker T, Barnstedt O, Edmondson-Stait AJ, Pleijzier MW, Otto N, Schlegel P, Sharifi N, Perisse E et al (2018) Integration of Parallel Opposing Memories Underlies Memory Extinction. Cell 175:709–722e15 Freas CA, Cheng K (2017) Learning and time-dependent cue choice in the desert ant, Melophorus bagoti. Ethology 123:503–515 Freas CA, Cheng K (2019) Panorama similarity and navigational knowledge in the nocturnal bull ant Myrm e cia midas. J Exp Biol 222(11):jeb193201 Freas CA, Cheng K (2025) Visual learning, route formation and the choreography of looking back in desert ants, Melophorus bagoti. Anim Behav 222:123125 Freas CA, Fleischmann PN, Cheng K (2019) Experimental ethology of learning in desert ants: Becoming expert navigators. Behav Processes 158:181–191 Freas CA, Narendra A, Cheng K (2017a) Compass cues used by a nocturnal bull ant, Myrmecia midas. J Exp Biol 220:1578–1585 Freas CA, Narendra A, Lemesle C, Cheng K (2017b) Polarized light use in the nocturnal bull ant, Myrmecia midas. R Soc Open Sci 4:170598 Freas CA, Wystrach A, Narendra A, Cheng K (2018) The View from the Trees: Nocturnal Bull Ants, Myrmecia midas, Use the Surrounding Panorama While Descending from Trees. Front Psychol 9:16 Freas CA, Wystrach A, Schwarz S, Spetch ML (2022) Aversive view memories and risk perception in navigating ants. Sci Rep 12:2899 Handler A, Graham TGW, Cohn R, Morantte I, Siliciano AF, Zeng J, Li Y, Ruta V (2019) Distinct Dopamine Receptor Pathways Underlie the Temporal Sensitivity of Associative Learning. Cell 178:60–75e19 Hirano Y, Ihara K, Masuda T, Yamamoto T, Iwata I, Takahashi A, Awata H, Nakamura N, Takakura M, Suzuki Y et al (2016) Shifting transcriptional machinery is required for long-term memory maintenance and modification in Drosophila mushroom bodies. Nat Commun 7:13471 Islam M, Freas CA, Cheng K (2020) Effect of large visual changes on the navigation of the nocturnal bull ant, Myrmecia midas | Animal Cognition. Anim Cogn 23:1071–1080 Kamhi JF, Barron AB, Narendra A (2020) Vertical Lobes of the Mushroom Bodies Are Essential for View-Based Navigation in Australian Myrmecia Ants. Curr Biol CB 30:3432–3437e3 Krashes MJ, DasGupta S, Vreede A, White B, Armstrong JD, Waddell S (2009) A neural circuit mechanism integrating motivational state with memory expression in Drosophila. Cell 139:416–427 Larrauri JA, Schmajuk NA (2008) Attentional, associative, and configural mechanisms in extinction. Psychol Rev 115:640–676 Le Moël F, Wystrach A (2020) Towards a multi-level understanding in insect navigation. Curr Opin Insect Sci 42:110–117 Lionetti VAG, Cheng K, Murray T (2024) Effect of repetition of vertical and horizontal routes on navigation performance in Australian bull ants. Learn Behav 52:92–104 Mangan M, Webb B (2012) Spontaneous formation of multiple routes in individual desert ants (Cataglyphis velox). Behav Ecol 23:944–954 Murray T, Kócsi Z, Dahmen H, Narendra A, Le Möel F, Wystrach A, Zeil J (2020) The role of attractive and repellent scene memories in ant homing (Myrmecia croslandi). J Exp Biol 223:jeb210021 Narendra A, Ramirez-Esquivel F (2017) Subtle changes in the landmark panorama disrupt visual navigation in a nocturnal bull ant. Philos Trans R Soc B Biol Sci 372:20160068 Narendra A, Reid SF, Raderschall CA (2013) Navigational Efficiency of Nocturnal Myrmecia Ants Suffers at Low Light Levels. PLoS ONE 8:e58801 Perisse E, Owald D, Barnstedt O, Talbot CB, Huetteroth W, Waddell S (2016) Aversive Learning and Appetitive Motivation Toggle Feed-Forward Inhibition in the Drosophila Mushroom Body. Neuron 90:1086–1099 Rescorla RA (2004) Spontaneous Recovery. Learn Mem 11:501–509 Schleyer M, Saumweber T, Nahrendorf W, Fischer B, von Alpen D, Pauls D, Thum A, Gerber B (2011) A behavior-based circuit model of how outcome expectations organize learned behavior in larval Drosophila. Learn Mem 18:639–653 Schultheiss P, Wystrach A, Schwarz S, Tack A, Delor J, Nooten SS, Bibost A-L, Freas CA, Cheng K (2016) Crucial role of ultraviolet light for desert ants in determining direction from the terrestrial panorama. Anim Behav 115:19–28 Schultz W (2016) Dopamine reward prediction-error signalling: a two-component response. Nat Rev Neurosci 17:183–195 Schwarz S, Mangan M, Webb B, Wystrach A (2020) Route-following ants respond to alterations of the view sequence. J Exp Biol 223:jeb218701 Stachenfeld KL, Botvinick MM, Gershman SJ (2017) The hippocampus as a predictive map. Nat Neurosci 20:1643–1653 Terao K, Mizunami M (2017) Roles of dopamine neurons in mediating the prediction error in aversive learning in insects. Sci Rep 7:14694 Webb B (2024) Beyond prediction error: 25 years of modeling the associations formed in the insect mushroom body. Learn Mem 31:a053824 Webb B, Wystrach A (2016) Neural mechanisms of insect navigation. Curr Opin Insect Sci 15:27–39 Wystrach A, Buehlmann C, Schwarz S, Cheng K, Graham P (2020a) Rapid Aversive and Memory Trace Learning during Route Navigation in Desert Ants. Curr Biol 30:1927–1933e2 Wystrach A, Mangan M, Philippides A, Graham P (2013) Snapshots in ants? New interpretations of paradigmatic experiments. J Exp Biol 216:1766–1770 Wystrach A, Moël FL, Clement L, Schwarz S (2020b) A lateralised design for the interaction of visual memories and heading representations in navigating ants. 2020.08.13.249193. https://www.biorxiv.org/content/ 10.1101/2020.08.13.249193v1 (Accessed August 6, 2025) Wystrach A, Schwarz S, Graham P, Cheng K (2019) Running paths to nowhere: repetition of routes shows how navigating ants modulate online the weights accorded to cues. Anim Cogn 22:213–222 Zeil J (2012) Visual homing: an insect perspective. Curr Opin Neurobiol 22:285–293 Zeil J, Hofmann MI, Chahl JS (2003) Catchment areas of panoramic snapshots in outdoor scenes. JOSA A 20:450–469 Additional Declarations No competing interests reported. Supplementary Files SupplementalMotivationControl.docx SupplementalFigures.docx Cite Share Download PDF Status: Under Review Version 1 posted Editorial decision: Revision requested 29 Apr, 2026 Reviews received at journal 23 Apr, 2026 Reviews received at journal 24 Mar, 2026 Reviewers agreed at journal 13 Mar, 2026 Reviewers agreed at journal 09 Mar, 2026 Reviewers invited by journal 04 Mar, 2026 Editor assigned by journal 04 Mar, 2026 Submission checks completed at journal 03 Mar, 2026 First submitted to journal 02 Mar, 2026 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information

Related documents

• Experimental data on contrast and luminance dependence of visual orientation behaviour in stick insects#965556
• Revisiting absent trails: Spatial memory of trail pheromones in termites.#449706
• Visual adaptation of a biting fly that permanently foregoes flight.#305245
• Graveyards on the Move: The Spatio-Temporal Distribution of Dead Ophiocordyceps-Infected Ants#485648
• Generational Ecological Memory: Towards a Biodiversity Lifelong Learning Framework#276892