Article activity feed

  1. Author Response

    Reviewer #1 (Public Review):

    The paper of De Agro et al. proposes a new paradigm to measure wanting (binary choices) and liking (pheromone deposition) in ants in order to test bundling and segregation effects on reward processing.

    By using three different treatments: A) rewards (sugar drops) and costs (runway segments) are segregated; B) rewards are segregated and costs bundled; C) rewards and costs are bundled, the authors observed that the main predictor of pheromone release was the segregation of the runaway segments rather than segregation of the reward. Furthermore, no effect of treatment was observed on preferences for the odor associated with the treatment.

    The authors interpret their finding as a clear demonstration of segregation effects on liking, but not wanting, which was present only for costs but not rewards.

    Strengths: I appreciated the creativity and effort in conducting complex experiments and measurements in insects. Overall, the paper is the first of its kind to propose a method to test reward processing in insects. The design is well thought and the results are straightforward. The analyses seem to be appropriate.

    Weaknesses: My main concern relates to the interpretation of the pheromone release as an index of liking. I am not an expert in the field, but I would probably go for a more parsimonious explanation: the effect could be simply due to the quantity of liquid ingested (and therefore corresponding caloric intake). Did you check whether, in the conditions showing the biggest pheromone release, the ants consumed the biggest quantity?

    First, this could explain for example the puzzling difference observed in the 3 cohorts and the sequence effects.

    Second, a reduced overall caloric intake could also explain why segregated costs seem to drive the results. Digestive processes are possibly kicking in at different times in the segregate all conditions compared to the other two, due to the more time-delayed ingestion of food (i.e. we tend to eat less if we have longer time between meals).

    Finally, this account may also explain the reported difference between wanting and liking, as here the release of pheromone is simply the byproduct of how much sugar has been ingested (and possibly nothing to do with reward processes).

    If pheromones are released proportionally to sugar intake and if sugar intake was different between conditions, is an important point that should be clarified in the manuscript, in order to guarantee interpretability of the results.

    We understand the reviewer’s position, and agree with the need of reserving “high-level” processes as explanation to situations that have no alternative, more parsimonious ones. Indeed, we cannot be certain of what is happening in the ant’s mind, and if its hedonistic experience is indeed separate to its memory evaluation process. To this end, we propose a mechanism that can explain this difference in terms of the memories formed for food quality and path length.

    We have now reduced our claims on the Liking vs Wanting framework. Regarding the origin of pheromone deposition being linked to caloric consumption, we believe this is not the case.

    Reviewer #2 (Public Review):

    Only a few decades ago ants were considered little machines without learning capabilities or personality. Ever since then, we have been able to attribute more and more personality to them. In this study, De Agro et al. have been able to use psychological tricks to manipulate the decision-making process of an ant species. By bundling or segregating costs (distance) and gains (food) they were able to demonstrate that ants, just as humans, experience gains and costs (in most cases) on a logarithmic scale. Moreover, they suggest a quantitative way to disentangle "wanting" and "liking" in ants, allowing for further interesting scientific designs to test theories long applied by behavioral economists on humans.

    The strength of this study clearly lies in the simplicity of its design and its strong foundation on current theories and models. It is clearly written and easily followed even by a non-specialist reader.

    I also particularly liked the exhaustive discussion and the interdisciplinary links it proposes. Including (but not limited to) the potential ecological implications in plant-pollinator interactions, with flowering plants potentially abusing segregation of flower rewards to manipulate the pollinator.

    The weaknesses:

    The statistics seem to lack any control for random factors like individual ant or colony of origin. While the results are quite clear and will likely not change with these additions they could add a little bit more resolution in some cases or help explain certain trends better. Especially since apparently a result with a highly significant p-value of 0.0036 is considered a false positive due to a lack of rational explanations. Individual experience, age, or fitness/size of the colony of origin could all affect the decision-making processes in individuals and should be controlled for (and discussed).

    We thank the reviewer for the comment. All the models we used in the analysis did include a random factor, always specified as ants nested into colony of origin, as appreciable in the R full script available in ESM2. Indeed, we failed to mention it in the paper and discuss it, which the reviewer is correct in requiring. We now added this specification in L261-262 and L271-274.

    Moreover, based also on the comment of another reviewer, we reconsidered our random-effects structure. See point 5 for the full explanation.

    In line with my previous comment, I would also have liked to see a bit more data on individual variation to better appreciate cross-condition comparisons. For instance, the fact that in Figure 4 ants that experienced the "segregated all" effect laid overall more pheromones than the ones that experience "bundled" first is barely acknowledged in the manuscript. These kinds of variations in pheromone deposition rate (not just relative, but in absolute numbers) need to be better discussed.

    Yes, there do indeed seem to be interesting patterns in figure 4, and spent many hours exploring the data in great depth, looking into visit-level pheromone depositions (see updated supplementary figure), to try to understand the patterns we see. We then discussed in detail how to present our findings. The main finding to explain is in condition 3: the overall pheromone deposited in the “good (in this case segregated all) encountered first” is lower, whereas it was higher for all the others conditions.

    We did develop an explanation for the pattern of findings (see below). However, we freely admit to being unsatisfied with it – the explanation is ad-hoc, and there are no strong biological or psychological reasons for it to be true, apart from fitting the data. Ultimately, we decided not to discuss these patterns very extensively, since we felt it added greatly to the length and complexity of the discussion, while not adding a lot of biological insight. Nonetheless, we crafted a manuscript-ready addition outlining our current best guess (and it is a guess) explaining the patterns in absolute pheromone deposition level. The text would be added directly to the end of the paragraph ending at line 428. If the editor and reviewer agree that this is a worthwhile addition, we would be happy to add it.

    Note that there are no significant differences between any of the groups or interactions in condition 1, although a purely visual inspection of the figure might suggest one.

    "In addition, the absolute amount of pheromone deposited (independently from the currently experienced option) varies depending on the first encounter treatment. This effect is puzzling, as it seems to cause a reverse in the absolute pheromone deposition in conditions 2 and 3, where the lower amount is in correspondence with the “bad” option being encountered first in the one, but the “good” option in the other. Indeed, we do not have a fully satisfying explanation for it. Our best guess is that it may be due to an inertia effect in pheromone deposition: when low to no pheromone is deposited for 2 or more visits in a row (the first visit being always low, and the second being of a non-preferred treatment), the ants never subsequently raise their pheromone deposition. This pattern is visible in Condition 3, as both option available are of low value, while in Condition 2, the Reward Segregated option is marginally better, and as such at least some pheromone is deposited even in visit 1 . We provided a visit-by-visit analysis and graphs in the supplement S2. Similar patterns have been reported for this species (Beckers et al., 1993)."

    I would also have liked to see a graph or results focusing on the pheromone deposition rate ONLY at the first experience trial, rather than always in combination with the subsequent trials.

    We would too! Again, we extensively discussed including one, as the first visit is very valuable – it is the only visit in which we can exclude contrast effects, making the results in principle much easier to interpret.

    However, the problem is that these ants deposit a lot less pheromone on their very first visit. This makes biological sense – they may be lost, and don’t know how reliable the food source is yet (see e.g. Czaczkes et al., 2013 figure 5A, where this pattern can be clearly seen). The same is true in the current dataset – which is why data from visit 1 is excluded from the figure (we repeat the analysis with and without visit 1, and find no differences in the results – see supplement).

    As a visual demonstration, we provide two (ugly, sorry) figures below: the pheromone deposition per treatment for only visit 1, and for visits 2-8. Note the massive zero inflation in visit 1.

    Pleasingly, the broad pattern (considering the mean) in just visit 1 follows our expectations. However, any reasonable statistical test on data from just the first visit would find no significant difference.

    In addition, even though the study focuses strongly on differences between "wanting" and "liking" it barely touches upon the data looking at "wanting". A graphical illustration of the Y-maze experiment and the binomial decision would have helped appreciate this result better (even if it is non-significant).

    We thank the reviewer for the comment. We generally try not to overburden our manuscripts with figures, as we aim to maintain the message of the paper focused on what we believe to be the most important finding. For this reason, we believe that a figure for the binomial response would be somewhat wasted, as all it would show are 3 points for each of the 3 conditions, all around 0.5 probability of choosing the predicted option. Below is an example of what such a figure would look like:

    On the other hand, we agree that a graphical illustration of the Y-maze may be of use. We now added Figure 2, showing both the Y maze and the pheromone deposition behavior, as the two main behaviours recorded.

    I also believe that the authors are overstating their claims of showing for the first time that ants prefer closer food sources. The cost of distance has already been demonstrated indirectly in Frank & Linsenmair 2017: "Individual versus collective decision making: optimal foraging in the group-hunting termite specialist Megaponera analis" for instance. While the current study does more directly imply the preference for closer food in a controlled experimental design I would argue that there is sufficient knowledge with indirect observations in natural settings, making the claim of showing it here for the first time unnecessarily hyperbolic.

    We agree with the reviewer. We have now added a reference to Frank and Linsenmair 2017 and weakened our claim. L534-535

    While the results of this study are novel and very interesting to a broad readership, I would suggest including in the discussion and introduction also a newer study on "food wanting is mediated by transient activation of dopaminergic signaling in the honey bee brain" by Huang et al. 2022 in Science and also recommend the accompanying perspective article by Garcia and Dyer on "Why do animals want what they like?".

    Thank you for the comment. We are aware of this new paper but we could not reference it in the earlier version of this manuscript as our submission to eLife happened on the 10th of April, prior to the publication of Huang et al. Considering the suggestions of the other reviewers, we have now reduced our claims about the liking vs wanting framework (see point 1). The reference has now been added in L518-532.

    Reviewer #3 (Public Review):

    This work aims at testing hypotheses derived from the field of behavioral economics (Kahneman's theories), related to subjective value perception in ants foraging for food. The work was conceived to test how ants react to a specific feature which is the segregation or the bundling of food resources. Behavioral economics posits that individuals value more segregated resources than the same amount of resources presented in a bundled way. At the same time, if accessing the segregated resources implies an increase in energetic costs to access them (i.e. longer displacements), then costs would be also perceived as higher in the segregated-resource case than in the bundled-resource case.

    Whether ants conform or not to this model is an interesting question, and irrespective of the results obtained, the experiments presented by the authors have been conceived to address this model as the experimental parameters varied refer to resource separation (drops of sucrose solution with different degrees of spacing between them) and to walking distances.

    Yet, the manuscript suffers from various serious deficits that preclude being enthusiastic with respect to its present form. Various problems are listed below, which reduce the quality of this work. Hopefully, the authors can amend some of these problems to reach a more consistent version.

    1. The inconsistent and unjustified "wrapping" with a "wanting vs liking" framework

    While it is unquestionable that the question raised by the authors revolves around behavioral-economic hypotheses on value perception and is fully addressed by the experiments performed, the "extra wrapping" of the "wanting/liking" framework added, probably to make the manuscript more attractive, is unjustified and excessively speculative. The use of a "wanting vs liking" interpretation framework is inappropriate as neither the experiments were conceived to address this topic, nor the results allow any robust conclusion on this point. These concepts originate in neuroscience analyses of neural-circuit activation in the mammalian brain upon situations that allow distinguishing several components related to reward: 1) the hedonic effect of pleasure itself (liking); 2) motivation to obtain the reward (wanting or incentive salience); and 3) and reward-related learning(1-3). These components refer to different identified neural circuits and brain areas as wanting for reward is generated by a large and distributed dopaminergic brain system including the frontal cortex, while liking is generated by a smaller set of hedonic hot spots within limbic circuitry and which are not dopamine-dependent.

    Clearly, the use of the wanting vs liking terminology requires accuracy and appropriate studies to support it. This is not the case in the present manuscript which was not conceived to tackle this issue. Moreover, inconsistent testing procedures (see below point 3) undermine the use and interpretation of choice data as wanting. The authors have no proof of the involvement of wanting vs. liking systems in their design and even more, cannot disentangle between these components based on their behavioral data. Considering that pheromone deposits after food experience express "liking" can be questioned as it does not dissociate between individual liking and social information transfer (the liking and wanting systems are individually based systems). Moreover, the assignment of a choice in a binary-choice test to a wanting system is also questionable as the experiments cannot disentangle between the eventual individual wanting and the reward-related learning as animals are making choices based on odorant cues they have learned during their previous foraging bouts. In the absence of neurobiological data, the hypotheses of wanting vs. liking remain on a shaky, highly speculative ground.

    Thus, the whole "wanting vs liking interpretation" (which attains alarming speculative levels in the Discussion section) should be omitted entirely from the manuscript if the authors want to provide a solid convincing framework articulated exclusively around the bundling vs. the segregation effects, which is precisely what their experiments tested. The rest is speculation in the absence of analyses supporting the wanting vs liking dissociation. An example of the kind of analysis necessary to go in this direction is provided by a recent work in which a dopamine-based wanting system was shown in honey bees(4), a work that the authors did not consider. We are clearly far from this kind of analysis in the present manuscript. As the authors wrote, "the present study is the first to examine bundling vs. segregation in an animal (line 99)", yet not liking vs. wanting.

    The reviewer makes a very well-argued case for this study not being sufficient evidence for distinct “wanting” and “liking” systems in an insect – a point echoed by the other reviewers. Their comments were helpful and insightful, and we fully agree with them. We have thus omitted the concepts of “wanting” and “liking” from the title, introduction, methods, and results.

    However, we feel that, especially given the results of Huang et al. (which were not published when we submitted the manuscript), the idea that the mismatch between the choice and pheromone data is driven by them acting on two separate systems reasonable: while it is not well supported, it is certainly consistent with this. The discussion seems to us to be the appropriate place to speculate about the meaning of results – especially results we do not fully understand. We would thus like to maintain a short discussion of this hypothesis in the discussion. Perhaps other researchers will be inspired to collect the necessary data to test whether such segregation effects really do affect “liking” but not “wanting” – something which is beyond the capabilities of our strictly behavioral lab.

    1. Some experimental assumptions are not substantiated by data

    The experimental procedure relies on separating or aggregating reward (drops of sucrose solution) and determining the impact of this variation on pheromone deposition while returning to the nest and subsequent choice in a dual test situation in which two of the three treatments designed - distinguished by the odorant experienced en route to reward - were presented. While the "Segregated All Treatment" (Fig. 2A) managed to space the 0.2 µl reward drops by significant 25-cm segments, thus enhancing potentially both reward appreciation (segregated food drops) and cost appreciation (successive segments to be negotiated), the "Segregate Reward Treatment" (Fig. 2B) raises doubts about its validity.

    In this case, three drops were offered at the end of three consecutive 25-cm segments, with the assumption that drops spaced by 5 mm should be perceived as being segregated (two of 0.2 µl and 1 ad libitum). Yet, there is no proof - at least in the manuscript - that spacing two food drops by 0.5 mm induces a segregated perception in ants. The first experience with the first drop may induce both sensitization and a local search that may last until the very close next drop is detected so that for the ant, these drops would be perceived as belonging to the same resource rather than being perceived as segregated resources. The same applies to the vicinity between the 0.2 µl drop and the ad libitum drop.

    This raises the question of the real volume of the ad libitum drop, which is not mentioned (it is just described as beings "large"; line 205). One could argue that if drops separated by 5 mm were bound together, the results would be similar to those of the "Bundled Treatment" (Fig. 2C). Strictly speaking, this is not necessarily true if the volume of the large drop was known. If this were the case, the Bundled Treatment offered a volume that was 0.4 µl smaller than the total food provided in the "Segregate Reward Treatment".

    Overall, further controls are needed to support the assumptions of the different treatments chosen.

    See detailed response to main concern 4 – “The segregation effect”. In brief: we agree that the current experiment cannot distinguish not sensing a difference between a big drop and three little drops from sensing a difference but not responding. However, the inclusion of the “segregated reward” treatment was only added to aid result interpretation in the event of reward segregation fully balancing out cost segregation. Since the response of ants to “all bundled” and “all segregated” treatments were different, the “segregated reward” treatment is in fact not needed to support our claim that segregation affects perceived value in these ants.

    1. Unclear design in the testing procedures

    The authors did not specify in the methods if a reward was provided in the tests in which a Y maze was presented to the ants having experienced a succession of short and long segments. This information was provided later, in the Results section (line 309) and, as expected, no reward was provided during the tests, thus raising the question of the necessity of the three consecutive tests, with no refreshment trials in between. This procedure is puzzling because it induces extinction of the odor-length association - as verified by the authors (see lines 306-309) - and makes the design questionable. Only the results of the very first test should be kept and analyzed in the manuscript.

    The same remark applies to the three tests performed after comparing the experimental treatments, which - one discovers only in the Results Section - were also performed in the absence of refreshment trials. In fact, the absence of coherence in the results of these tests (e.g. lines 328-332) could be precisely due to a change of strategy between the tests following the absence of reward in the first test. This underlines the necessity of focusing exclusively on the first test and dismissing the data of the 2nd and 3rd tests in which performance may have been affected by extinction and strategy change. This again shows why speaking about "wanting" in this inconsistent framework makes no sense at all.

    We thank the reviewer for the comment. Please see point 2 were we provide the full answer. We initially included the subsequent testing in our experimental design so as to gather as much information as possible. A change in preference linked to the absence of the reward is indeed expected. However, the rapidity and direction of change can give valuable information that would be lost if the data were to not be collected. We agree with the reviewer that in this specific experiment the data was not particularly useful, but we believe it would be wrong from us to just not report it. As we wrote above, please note that in ESM2 we report the choice probability of the first choice only, showing the same exact result as when all three choice are considered.

    Reviewer #4 (Public Review):

    The manuscript reports an experiment testing how the distribution of rewards and costs influences perceived reward value in ants. Using a bundling manipulation where rewards and costs were presented either in small separated amounts (segregated) or together in a larger amount (bundled), the results show that ants deposited a greater quantity of pheromones (which was used as an index of "liking") when rewards were segregated and costs bundled compared to when both rewards and costs were bundled (although that difference was statistically significant only in ants experiencing the segregated reward condition first during training) and when both rewards and costs were segregated. By contrast, no evidence was found for a bundling effect in terms of choice behaviour (which was used as an index of "wanting"). The authors suggest that these findings demonstrate a bundling effect and a dissociation between "wanting" and "liking" in ants.

    Overall, the experiment provides a worthy contribution to the study of the biases that affect the perceived value of rewards in a translational perspective from humans to invertebrate animals. The experimental manipulation is clever, and the results clearly indicate that manipulating bundling affected pheromone deposition in ants. However, the data reported do not appear to fully support the conclusions of an increased "liking" of the segregated rewards and bundled costs compared to bundled rewards and costs. In addition, more evidence (along with stronger justifications) would be needed to establish that choice behaviour and pheromone deposition are appropriate and sensitive measures of "wanting" and "liking", respectively. This aspect renders any claim of a dissociation between "wanting" and "liking" in ants somewhat premature and speculative at this stage. I describe these concerns in more detail below.

    1. The main hypothesis tested is that segregated rewards with bundled costs should be the most "liked" option relative to bundled rewards and costs and segregated rewards and costs. The results are interpreted as fully in line with this hypothesis. However, the data reported do not suggest this is the case: The difference between the 'segregated rewards' condition and the 'bundled' condition is not statistically significant when all ants are considered (that difference being statistically significant only for ants that first experienced the 'segregated rewards' condition during training). Although this point is briefly acknowledged in the discussion, more nuance and extra caution are needed in the overall interpretation of the findings, so that this statistically nonsignificant result does not appear as being treated as if it were statistically significant.

    We thank the reviewer for the comment. Indeed, our initial hypothesis was the one described here. However, the results of the segregated rewards vs bundled condition, being not significantly different, forced us to consider an alternative hypothesis. We believe that our current experiment managed only to bundle and segregate costs, not gains. Given this, we would expect segregated rewards vs bundled to perform at chance level, since in both the cost is equally bundled. As such, we are ultimately treating the result as non-significant. We are, however, clearly stating our initial hypothesis, and discussing how the data fits it, as we feel it would be dishonest of us to give the impression we had the second hypothesis from the start, or on the other hand to treat a p-value so near 0.05 as definitely random.

    1. An important requirement to adequately evaluate the findings from the choice behaviour test is to ensure that ants did learn the associations between the reward conditions and the runway scents. Ruling out potential learning confounds is in fact essential to interpret the results as reflecting the operation of motivational processes such as "wanting". Whereas the results from the pilot experiment suggest that ants learned the contingencies between the runway length and its associated scent, the pilot experiment and the main experiment differ in significant ways. Therefore, it is unclear whether the ants learned the contingencies in the main experiment, which could be advanced as an alternative explanation for the lack of preferences between the two scented arms of the Y-maze during the choice test. Another important aspect to consider is that the reward still has to be valued by the organism to appropriately assess "wanting" processes. Indeed, "wanting" is generally conceptualised as conjointly determined by the associative history between the cue or context (scent) and the reward (sucrose solution) on one hand, and the organism's homeostatic or physiological needs such as hunger on the other hand (e.g., Zhang et al., 2009. https://doi.org/10.1371/). In the main experiment, the question arises as to whether reward devaluation could have occurred-resulting in the reward having a diminished value as the ants were able to consume the sucrose solution to satiation multiple times across the experiment. For these reasons, it would be important to provide information showing that (a) the ants learned with which condition the scent was associated and (b) that the reward was still valued during the choice test. These points are key preconditions that need to be fulfilled for ruling out potential confounds that could explain the findings of the choice test as well as for suggesting a dissociation between "wanting" and "liking".

    We thank the reviewer for the comment. As stated in point 1, the “liking” vs “wanting” framework has been greatly reduced in the paper, only being raised as a possible explanation of the observed results, with the lack of learning being raised as a reasonable alternative explanation. We have reason to believe that the ants are actually learning the association presented, as we detail in point 2.1. Of course, we cannot be completely certain, as it is impossible to disentangle preference from learning in such experiments. As such, the possibility is mentioned in the manuscript.

    1. Relatedly, a strong justification needs to be formulated to substantiate that the choice test provides a reliable indicator of "wanting". This is critical to conclude that the results can be interpreted as reflecting a dissociation between "wanting" and "liking". In rodents and humans, "wanting" is typically measured as an increased effort mobilisation during the presentation of a cue associated with a reward (e.g., Pool et al., 2016. https://doi.org/10.1016/j.). It remains however unclear how choice can capture such effects. This questions the extent to which choice represents an adequate operationalisation and measure of "wanting" as described in the incentive salience hypothesis (Berridge & Robinson, 2016. https://doi.org/10.1037/). Moreover, it should be clearly explained and motivated whether, and if so how, choice purely measures "wanting" without being contaminated or influenced by liking-based processes, such as preferences or expected pleasantness for instance.

    We agree with the reviewer. Indeed, our linking of choice with “wanting” and pheromone with “liking” is highly speculative. According to point 1, we strongly reduced our claims and propose the association only as one of several potential explanations.

    1. Little information is provided on how pheromone deposition was measured and on the specificities of this measure, such as its physiological bases, timing properties, and granularity. However, detailed information about this measure is of high relevance to be able to assess if pheromone deposition represents a sensitive measure of "liking". "Liking" is typically measured as hedonic reactions during reward consumption across the rodent and human literature (e.g., Pool et al., 2016. https://doi.org/10.1016/j.). Accordingly, a good index of "liking" should be specifically responsive to reward consumption. By extension, an increased pheromone deposition should be particularly evident after the ants consumed the sucrose drop. As it stands, it is unclear whether this is the case as the pilot experiment showed no statistically significant difference in pheromone deposition between the way towards the sucrose drop or back. If the measure of pheromone deposition allows for distinguishing between pheromones deposited on the way towards the drop and pheromones deposited on the way back in the main experiment, a further test that could be run would be to compare the pheromone deposition on the way towards the drop in the 'segregated all' condition versus the 'segregated rewards' and 'bundled' conditions. A higher pheromone deposition on the way towards the sucrose drop in the 'segregated all' condition could provide converging evidence that pheromone deposition is a sensitive indicator of "liking".

    Unfortunately, in our current setup it was impossible to collect pheromone deposition data on the way towards the drop. Pheromone deposition has to be collected by eye, and the experimenter needed to maintain attention on the delivery of the successive rewards. A camera was not an option either, as the distance and resolution needed to record the whole runway would be insufficient to notice deposition, which instead requires the experimenter to follow the ants and count the individual stereotyped behaviours. We do observe the effect of higher deposition near the drop relative to further down the runway on the way back, which seems to be congruent with the response to consumption. We are however aware this is not sufficient, as it may just be linked with the distance. Regardless, as per point 1, we are decreasing our claims for the liking vs wanting framework.

    Was this evaluation helpful?
  2. Evaluation Summary:

    This manuscript has the potential to be of interest to a broad range of behavioral scientists. It provides insights into how biases can affect value-based behavior in invertebrates, similar to what has been reported in humans. However, there are a number of potential confounders that need to be addressed before drawing robust conclusions from the data reported.

    (This preprint has been reviewed by eLife. We include the public reviews from the reviewers here; the authors also receive private feedback with suggested changes to the manuscript. Reviewer #1, Reviewer #2 and Reviewer #4 agreed to share their name with the authors.)

    Was this evaluation helpful?
  3. Reviewer #1 (Public Review):

    The paper of De Agro et al. proposes a new paradigm to measure wanting (binary choices) and liking (pheromone deposition) in ants in order to test bundling and segregation effects on reward processing.

    By using three different treatments: A) rewards (sugar drops) and costs (runway segments) are segregated; B) rewards are segregated and costs bundled; C) rewards and costs are bundled, the authors observed that the main predictor of pheromone release was the segregation of the runaway segments rather than segregation of the reward. Furthermore, no effect of treatment was observed on preferences for the odor associated with the treatment.

    The authors interpret their finding as a clear demonstration of segregation effects on liking, but not wanting, which was present only for costs but not rewards.

    Strengths: I appreciated the creativity and effort in conducting complex experiments and measurements in insects. Overall, the paper is the first of its kind to propose a method to test reward processing in insects. The design is well thought and the results are straightforward. The analyses seem to be appropriate.

    Weaknesses: My main concern relates to the interpretation of the pheromone release as an index of liking. I am not an expert in the field, but I would probably go for a more parsimonious explanation: the effect could be simply due to the quantity of liquid ingested (and therefore corresponding caloric intake). Did you check whether, in the conditions showing the biggest pheromone release, the ants consumed the biggest quantity?

    First, this could explain for example the puzzling difference observed in the 3 cohorts and the sequence effects.

    Second, a reduced overall caloric intake could also explain why segregated costs seem to drive the results. Digestive processes are possibly kicking in at different times in the segregate all conditions compared to the other two, due to the more time-delayed ingestion of food (i.e. we tend to eat less if we have longer time between meals).

    Finally, this account may also explain the reported difference between wanting and liking, as here the release of pheromone is simply the byproduct of how much sugar has been ingested (and possibly nothing to do with reward processes).

    If pheromones are released proportionally to sugar intake and if sugar intake was different between conditions, is an important point that should be clarified in the manuscript, in order to guarantee interpretability of the results

    Was this evaluation helpful?
  4. Reviewer #2 (Public Review):

    Only a few decades ago ants were considered little machines without learning capabilities or personality. Ever since then, we have been able to attribute more and more personality to them. In this study, De Agro et al. have been able to use psychological tricks to manipulate the decision-making process of an ant species. By bundling or segregating costs (distance) and gains (food) they were able to demonstrate that ants, just as humans, experience gains and costs (in most cases) on a logarithmic scale. Moreover, they suggest a quantitative way to disentangle "wanting" and "liking" in ants, allowing for further interesting scientific designs to test theories long applied by behavioral economists on humans.

    The strength of this study clearly lies in the simplicity of its design and its strong foundation on current theories and models. It is clearly written and easily followed even by a non-specialist reader.

    I also particularly liked the exhaustive discussion and the interdisciplinary links it proposes. Including (but not limited to) the potential ecological implications in plant-pollinator interactions, with flowering plants potentially abusing segregation of flower rewards to manipulate the pollinator.

    The weaknesses:

    The statistics seem to lack any control for random factors like individual ant or colony of origin. While the results are quite clear and will likely not change with these additions they could add a little bit more resolution in some cases or help explain certain trends better. Especially since apparently a result with a highly significant p-value of 0.0036 is considered a false positive due to a lack of rational explanations. Individual experience, age, or fitness/size of the colony of origin could all affect the decision-making processes in individuals and should be controlled for (and discussed).

    In line with my previous comment, I would also have liked to see a bit more data on individual variation to better appreciate cross-condition comparisons. For instance, the fact that in Figure 4 ants that experienced the "segregated all" effect laid overall more pheromones than the ones that experience "bundled" first is barely acknowledged in the manuscript. These kinds of variations in pheromone deposition rate (not just relative, but in absolute numbers) need to be better discussed.

    I would also have liked to see a graph or results focusing on the pheromone deposition rate ONLY at the first experience trial, rather than always in combination with the subsequent trials.

    In addition, even though the study focuses strongly on differences between "wanting" and "liking" it barely touches upon the data looking at "wanting". A graphical illustration of the Y-maze experiment and the binomial decision would have helped appreciate this result better (even if it is non-significant).

    I also believe that the authors are overstating their claims of showing for the first time that ants prefer closer food sources. The cost of distance has already been demonstrated indirectly in Frank & Linsenmair 2017: "Individual versus collective decision making: optimal foraging in the group-hunting termite specialist Megaponera analis" for instance. While the current study does more directly imply the preference for closer food in a controlled experimental design I would argue that there is sufficient knowledge with indirect observations in natural settings, making the claim of showing it here for the first time unnecessarily hyperbolic.

    While the results of this study are novel and very interesting to a broad readership, I would suggest including in the discussion and introduction also a newer study on "food wanting is mediated by transient activation of dopaminergic signaling in the honey bee brain" by Huang et al. 2022 in Science and also recommend the accompanying perspective article by Garcia and Dyer on "Why do animals want what they like?".

    Was this evaluation helpful?
  5. Reviewer #3 (Public Review):

    This work aims at testing hypotheses derived from the field of behavioral economics (Kahneman's theories), related to subjective value perception in ants foraging for food. The work was conceived to test how ants react to a specific feature which is the segregation or the bundling of food resources. Behavioral economics posits that individuals value more segregated resources than the same amount of resources presented in a bundled way. At the same time, if accessing the segregated resources implies an increase in energetic costs to access them (i.e. longer displacements), then costs would be also perceived as higher in the segregated-resource case than in the bundled-resource case.

    Whether ants conform or not to this model is an interesting question, and irrespective of the results obtained, the experiments presented by the authors have been conceived to address this model as the experimental parameters varied refer to resource separation (drops of sucrose solution with different degrees of spacing between them) and to walking distances.

    Yet, the manuscript suffers from various serious deficits that preclude being enthusiastic with respect to its present form. Various problems are listed below, which reduce the quality of this work. Hopefully, the authors can amend some of these problems to reach a more consistent version.

    1. The inconsistent and unjustified "wrapping" with a "wanting vs liking" framework
      While it is unquestionable that the question raised by the authors revolves around behavioral-economic hypotheses on value perception and is fully addressed by the experiments performed, the "extra wrapping" of the "wanting/liking" framework added, probably to make the manuscript more attractive, is unjustified and excessively speculative. The use of a "wanting vs liking" interpretation framework is inappropriate as neither the experiments were conceived to address this topic, nor the results allow any robust conclusion on this point. These concepts originate in neuroscience analyses of neural-circuit activation in the mammalian brain upon situations that allow distinguishing several components related to reward: 1) the hedonic effect of pleasure itself (liking); 2) motivation to obtain the reward (wanting or incentive salience); and 3) and reward-related learning(1-3). These components refer to different identified neural circuits and brain areas as wanting for reward is generated by a large and distributed dopaminergic brain system including the frontal cortex, while liking is generated by a smaller set of hedonic hot spots within limbic circuitry and which are not dopamine-dependent.

    Clearly, the use of the wanting vs liking terminology requires accuracy and appropriate studies to support it. This is not the case in the present manuscript which was not conceived to tackle this issue. Moreover, inconsistent testing procedures (see below point 3) undermine the use and interpretation of choice data as wanting. The authors have no proof of the involvement of wanting vs. liking systems in their design and even more, cannot disentangle between these components based on their behavioral data. Considering that pheromone deposits after food experience express "liking" can be questioned as it does not dissociate between individual liking and social information transfer (the liking and wanting systems are individually based systems). Moreover, the assignment of a choice in a binary-choice test to a wanting system is also questionable as the experiments cannot disentangle between the eventual individual wanting and the reward-related learning as animals are making choices based on odorant cues they have learned during their previous foraging bouts. In the absence of neurobiological data, the hypotheses of wanting vs. liking remain on a shaky, highly speculative ground.

    Thus, the whole "wanting vs liking interpretation" (which attains alarming speculative levels in the Discussion section) should be omitted entirely from the manuscript if the authors want to provide a solid convincing framework articulated exclusively around the bundling vs. the segregation effects, which is precisely what their experiments tested. The rest is speculation in the absence of analyses supporting the wanting vs liking dissociation. An example of the kind of analysis necessary to go in this direction is provided by a recent work in which a dopamine-based wanting system was shown in honey bees(4), a work that the authors did not consider. We are clearly far from this kind of analysis in the present manuscript. As the authors wrote, "the present study is the first to examine bundling vs. segregation in an animal (line 99)", yet not liking vs. wanting.

    1. Some experimental assumptions are not substantiated by data
      The experimental procedure relies on separating or aggregating reward (drops of sucrose solution) and determining the impact of this variation on pheromone deposition while returning to the nest and subsequent choice in a dual test situation in which two of the three treatments designed - distinguished by the odorant experienced en route to reward - were presented. While the "Segregated All Treatment" (Fig. 2A) managed to space the 0.2 µl reward drops by significant 25-cm segments, thus enhancing potentially both reward appreciation (segregated food drops) and cost appreciation (successive segments to be negotiated), the "Segregate Reward Treatment" (Fig. 2B) raises doubts about its validity.
      In this case, three drops were offered at the end of three consecutive 25-cm segments, with the assumption that drops spaced by 5 mm should be perceived as being segregated (two of 0.2 µl and 1 ad libitum). Yet, there is no proof - at least in the manuscript - that spacing two food drops by 0.5 mm induces a segregated perception in ants. The first experience with the first drop may induce both sensitization and a local search that may last until the very close next drop is detected so that for the ant, these drops would be perceived as belonging to the same resource rather than being perceived as segregated resources. The same applies to the vicinity between the 0.2 µl drop and the ad libitum drop.
      This raises the question of the real volume of the ad libitum drop, which is not mentioned (it is just described as beings "large"; line 205). One could argue that if drops separated by 5 mm were bound together, the results would be similar to those of the "Bundled Treatment" (Fig. 2C). Strictly speaking, this is not necessarily true if the volume of the large drop was known. If this were the case, the Bundled Treatment offered a volume that was 0.4 µl smaller than the total food provided in the "Segregate Reward Treatment".
      Overall, further controls are needed to support the assumptions of the different treatments chosen.

    2. Unclear design in the testing procedures
      The authors did not specify in the methods if a reward was provided in the tests in which a Y maze was presented to the ants having experienced a succession of short and long segments. This information was provided later, in the Results section (line 309) and, as expected, no reward was provided during the tests, thus raising the question of the necessity of the three consecutive tests, with no refreshment trials in between. This procedure is puzzling because it induces extinction of the odor-length association - as verified by the authors (see lines 306-309) - and makes the design questionable. Only the results of the very first test should be kept and analyzed in the manuscript.
      The same remark applies to the three tests performed after comparing the experimental treatments, which - one discovers only in the Results Section - were also performed in the absence of refreshment trials. In fact, the absence of coherence in the results of these tests (e.g. lines 328-332) could be precisely due to a change of strategy between the tests following the absence of reward in the first test. This underlines the necessity of focusing exclusively on the first test and dismissing the data of the 2nd and 3rd tests in which performance may have been affected by extinction and strategy change. This again shows why speaking about "wanting" in this inconsistent framework makes no sense at all.

    1 Berridge, K. C. & Robinson, T. E. Am Psychol 71, 670-679. (2016).
    2 Berridge, K. C. & Kringelbach, M. L. Neuron 86, 646-664. (2015).
    3 Berridge, K. C. & Kringelbach, M. L. Curr Opin Neurobiol 23, 294-303. (2013).
    4 Huang, J. et al. Science 376, 508-512. (2022).

    Was this evaluation helpful?
  6. Reviewer #4 (Public Review):

    The manuscript reports an experiment testing how the distribution of rewards and costs influences perceived reward value in ants. Using a bundling manipulation where rewards and costs were presented either in small separated amounts (segregated) or together in a larger amount (bundled), the results show that ants deposited a greater quantity of pheromones (which was used as an index of "liking") when rewards were segregated and costs bundled compared to when both rewards and costs were bundled (although that difference was statistically significant only in ants experiencing the segregated reward condition first during training) and when both rewards and costs were segregated. By contrast, no evidence was found for a bundling effect in terms of choice behaviour (which was used as an index of "wanting"). The authors suggest that these findings demonstrate a bundling effect and a dissociation between "wanting" and "liking" in ants.

    Overall, the experiment provides a worthy contribution to the study of the biases that affect the perceived value of rewards in a translational perspective from humans to invertebrate animals. The experimental manipulation is clever, and the results clearly indicate that manipulating bundling affected pheromone deposition in ants. However, the data reported do not appear to fully support the conclusions of an increased "liking" of the segregated rewards and bundled costs compared to bundled rewards and costs. In addition, more evidence (along with stronger justifications) would be needed to establish that choice behaviour and pheromone deposition are appropriate and sensitive measures of "wanting" and "liking", respectively. This aspect renders any claim of a dissociation between "wanting" and "liking" in ants somewhat premature and speculative at this stage. I describe these concerns in more detail below.

    1. The main hypothesis tested is that segregated rewards with bundled costs should be the most "liked" option relative to bundled rewards and costs and segregated rewards and costs. The results are interpreted as fully in line with this hypothesis. However, the data reported do not suggest this is the case: The difference between the 'segregated rewards' condition and the 'bundled' condition is not statistically significant when all ants are considered (that difference being statistically significant only for ants that first experienced the 'segregated rewards' condition during training). Although this point is briefly acknowledged in the discussion, more nuance and extra caution are needed in the overall interpretation of the findings, so that this statistically nonsignificant result does not appear as being treated as if it were statistically significant.

    2. An important requirement to adequately evaluate the findings from the choice behaviour test is to ensure that ants did learn the associations between the reward conditions and the runway scents. Ruling out potential learning confounds is in fact essential to interpret the results as reflecting the operation of motivational processes such as "wanting". Whereas the results from the pilot experiment suggest that ants learned the contingencies between the runway length and its associated scent, the pilot experiment and the main experiment differ in significant ways. Therefore, it is unclear whether the ants learned the contingencies in the main experiment, which could be advanced as an alternative explanation for the lack of preferences between the two scented arms of the Y-maze during the choice test. Another important aspect to consider is that the reward still has to be valued by the organism to appropriately assess "wanting" processes. Indeed, "wanting" is generally conceptualised as conjointly determined by the associative history between the cue or context (scent) and the reward (sucrose solution) on one hand, and the organism's homeostatic or physiological needs such as hunger on the other hand (e.g., Zhang et al., 2009. https://doi.org/10.1371/journal.pcbi.1000437). In the main experiment, the question arises as to whether reward devaluation could have occurred-resulting in the reward having a diminished value as the ants were able to consume the sucrose solution to satiation multiple times across the experiment. For these reasons, it would be important to provide information showing that (a) the ants learned with which condition the scent was associated and (b) that the reward was still valued during the choice test. These points are key preconditions that need to be fulfilled for ruling out potential confounds that could explain the findings of the choice test as well as for suggesting a dissociation between "wanting" and "liking".

    3. Relatedly, a strong justification needs to be formulated to substantiate that the choice test provides a reliable indicator of "wanting". This is critical to conclude that the results can be interpreted as reflecting a dissociation between "wanting" and "liking". In rodents and humans, "wanting" is typically measured as an increased effort mobilisation during the presentation of a cue associated with a reward (e.g., Pool et al., 2016. https://doi.org/10.1016/j.neubiorev.2016.01.006). It remains however unclear how choice can capture such effects. This questions the extent to which choice represents an adequate operationalisation and measure of "wanting" as described in the incentive salience hypothesis (Berridge & Robinson, 2016. https://doi.org/10.1037/amp0000059). Moreover, it should be clearly explained and motivated whether, and if so how, choice purely measures "wanting" without being contaminated or influenced by liking-based processes, such as preferences or expected pleasantness for instance.

    4. Little information is provided on how pheromone deposition was measured and on the specificities of this measure, such as its physiological bases, timing properties, and granularity. However, detailed information about this measure is of high relevance to be able to assess if pheromone deposition represents a sensitive measure of "liking". "Liking" is typically measured as hedonic reactions during reward consumption across the rodent and human literature (e.g., Pool et al., 2016. https://doi.org/10.1016/j.neubiorev.2016.01.006). Accordingly, a good index of "liking" should be specifically responsive to reward consumption. By extension, an increased pheromone deposition should be particularly evident after the ants consumed the sucrose drop. As it stands, it is unclear whether this is the case as the pilot experiment showed no statistically significant difference in pheromone deposition between the way towards the sucrose drop or back. If the measure of pheromone deposition allows for distinguishing between pheromones deposited on the way towards the drop and pheromones deposited on the way back in the main experiment, a further test that could be run would be to compare the pheromone deposition on the way towards the drop in the 'segregated all' condition versus the 'segregated rewards' and 'bundled' conditions. A higher pheromone deposition on the way towards the sucrose drop in the 'segregated all' condition could provide converging evidence that pheromone deposition is a sensitive indicator of "liking".

    Was this evaluation helpful?
  7. Was this evaluation helpful?