Broadly speaking, when an animal must choose between options, it can employ one of three different strategies, characterized by different levels of precision: random choice; following a heuristic or rule of thumb; or comparison of outcome value and choice for the highest. Traditional economic theory, exemplified by expected utility theory, assumes (human) decision-makers are rational and perform strict value-based choice (Mankiw, 2011). The now well-established field of behavioural economics has vigorously pushed back against this idea, demonstrating that humans often make decisions which are not fully logical, economically rational, or ‘optimal’ (Camerer et al., 2011; Tversky and Kahneman, 1974), even when actively comparing options.

A major insight lying at the heart of behavioural economics is that value is perceived. This can lead directly to deviations from optimality: between the acquisition of the information and the evaluation of possible outcomes, something gets lost in translation. As has been well established for over a century by the study of pyschophysics, perception is non-linear, usually on a logarithmic scale (Gescheider, 1997). Value perception for humans is likewise non-linear, as famously stated by Kahneman and Tversky, and extensively demonstrated thereafter (Camerer, 2004; Kahneman and Tversky, 1979). Moreover, humans weigh losses more strongly than gains. Finally, value is relative, usually to an expectation or some sort of anchor, see Figure 1.

Figure 1

Download asset Open asset

The non-linear nature of perceived value results in many behavioural biases, a very prominent one being the bundling vs. segregation effect. Crucial to the current experiment, the perceived value of a compound item or option can be changed by presenting it either as one option (bundling) or as multiple small parts (segregation). Due to diminishing returns, bundling results in a weaker total sensation than segregation – either lower value for positive value options or lower cost for negative value options (see Figure 1). This is because more of the sensation occurs on the shallow part of the curve. This fact is regularly exploited by consumer psychologists and marketing experts, for example by bundling option when selling new cars: when spending €50,000 on a new car, spending €51,000 for the model with included sound system might not be experienced as painfully costly, even if, considered by itself, €1000 might be more than most people are willing to spend on a sound system for a car. Bundling and segregation have been extensively studied by consumer psychologists (Johnson et al., 1999; Naylor and Frank, 2001; Noone and Mattila, 2009).

As in the study of human economics, non-human animals have been treated and modelled as rational economic agents, leading to deep insights into animal behaviour via the optimal foraging theory framework (Davies, 2012; Emlen, 1966; MacArthur and Pianka, 1966; Pyke et al., 1977). However, often inspired directly by behavioural economic research on humans, predictable deviations from optimality and rationality have been described (Zentall, 2015). For example, pigeons, rats, and ants all show a preference for high-effort over low-effort associated options (Clement et al., 2000; Czaczkes et al., 2018a; Lydall et al., 2010), much as humans do (Norton et al., 2012). Similarly, much as in humans, the addition of an irrelevant option in an option set (a ‘decoy’) can change the preference structure in many animals, including birds, cats, bees, and ants (Bateson et al., 2002; Bateson et al., 2003; Parrish et al., 2015a; Sasaki and Pratt, 2011; Scarpi, 2011; Schuck-Paim et al., 2004; Shafir et al., 2002).

Ants and bees also show relative value perception, changing the perceived value of an option depending on their expectations (Bitterman, 1976; Couvillon and Bitterman, 1984; Wendt et al., 2019; Wendt and Czaczkes, 2020). An ant which is trained to expect a very sweet reward, for example, will be more likely to reject a moderate reward than an ant which was expecting a moderate one – a negative contrast. Likewise, ants expecting a very mild reward are more likely to accept a moderate reward than ants which were expecting that quality – a positive contrast (Wendt et al., 2019). These changes are also mirrored in the ants’ deposition of recruitment pheromone: ants deposit more pheromone to resources they perceive as higher quality (Beckers et al., 1993; Jackson and Châline, 2007), and indeed ants deposit more pheromone for moderate rewards if they had been expected poor quality, and less pheromone if they were expecting high quality. This demonstrates that a key aspect of prospect theory – relative value perception – is present in insects as well as in humans.

Crucially for the current experiment, there is evidence that social insects, especially ants, perceive value logarithmically. This is perhaps not surprising, given than logarithmic perception is a key assumption of the well-established Weber-Fechner law (Fechner, 1860; Weber, 1834), which describes the psychophysics of perception. Wendt et al., 2019, demonstrated a faster rise in food acceptance in the lower range of acceptance food qualities (e.g. from 0.1 to 0.3 molar) than in the higher range (e.g. from 0.5 to 1.5 molar). Recently, De Agrò et al., 2021, demonstrated that Lasius niger ants have a strong aversion to risky food sources (i.e. with fluctuating quality), which could be fully explained by logarithmic value perception, as predicted by prospect theory. Ants prefer a certain food source offering 0.55 M sucrose to one which fluctuates between 0.1 and 1.0, but if the options are logarithmically balanced (0.3 M vs. 0.1 or 0.9), ants are completely indifferent.

While the bundling vs. segregation effect is extremely well studied in humans, surprisingly, to our knowledge no attempt has been made to examine it in animals. In a related study, chimps were shown to prefer whole rewards (potato chips) to rewards which were broken into smaller pieces, even when the broken rewards had an higher absolute quantity (Parrish et al., 2015b). However, consumption of the whole and broken rewards took the same amount of time, and rewards were chosen before consumption could begin, so it is likely that both broken and whole rewards were considered as one unitary bundle.

In order to study bundling and segregation in ants, we segregated rewards spatially, using small sucrose drops along a linear runway. However, while this segregates rewards, it also segregates a potential cost – the walking distance to the reward. Thus, we first demonstrate that ants do indeed prefer closer to farther food sources, and thus that food distance is considered a cost. We then train individual ants to two of three treatments: rewards and costs bundled (‘bundled’), rewards and costs segregated (‘segregated all’), and rewards segregated but costs bundled (‘segregated reward’). We record pheromone depositions on the ant’s return from each treatment type, and then ask ants to choose between the pair of treatments they were trained on. We predicted that the ‘segregated reward’ treatment would be the preferred option, as bundling of costs should minimize their impact, and segregation of rewards should boost theirs. As we did not know the relative strength of the rewards and costs, we had no strong a priori predictions about the relative perception of ‘bundled’ and ‘segregated all’. However, as prospect theory predicts that losses are weighted more strongly than gains (Kahneman and Tversky, 1979), we had a weak expectation that ‘bundled’ would be perceived as slightly better than ‘all segregated’.

Bundling and segregation treatment strongly affected the ants’ pheromone deposition, but not their choices. No consistent preference was found in the binary choices of the ants from the main study. While ants significantly preferred the segregated reward over the bundled reward in their second of three trials, we can conceive of no plausible biological or psychological reason for this to be so, and interpret this as a false positive. Thus, there is no evidence that bundling or segregation affect choice.

However, the pattern of pheromone deposition, which correlates strongly with perceived food quality in these and other ants (Beckers et al., 1993; Czaczkes et al., 2018b; Jackson and Châline, 2007; Wendt et al., 2019), was partially in line with our predictions: ants deposited most pheromone when the costs were bundled and rewards segregated, and less in the other two treatments (Figure 4). Our weaker prediction of ‘bundled’ presenting a higher deposition than ‘segregated all’ was also supported. Thus, pheromone deposition seems to be in line with predictions from classical behavioural economic and perception research (Camerer, 2004; Johnson et al., 1999; Kahneman and Tversky, 1979; Naylor and Frank, 2001; Noone and Mattila, 2009). Specifically, we observed more deposition for ‘bundled’ over ‘segregated all’, and for ‘segregated reward’ over ‘segregated all’. By contrast, the preference for ‘segregated reward’ over ‘bundled’ is not significant. This seems counter-intuitive, as the ‘bundled’ and ‘segregated reward’ options only differ in the boosting of a gain. Moreover, the contrast of ‘segregated all’ vs. ‘bundled’ – for which we had no strong a priori prediction as we had no idea how gains and losses are weighted – shows a very clear difference.

The observed result suggests two, non-mutually exclusive, interpretations: Either ants experience a strong segregation effect for losses, and a very mild one, if any, for gains; or our treatment failed to segregate rewards effectively. If the first interpretation is true, the ‘bundled’ and the ‘segregated reward’ treatments will be perceived as almost identical. In the contrast between ‘segregated all’ and ‘bundled’, the higher number of depositions for bundled makes sense, as the segregation of losses drives the preference completely. The higher deposition for ‘segregated reward’ over ‘segregated all’ remains equally clear, as the two options only differ in the realm of losses. If our second interpretation is correct, the three sucrose solution drops, especially in the ‘segregated reward’ condition, are still considered by the animal as a single (i.e. bundled) reward. In the ‘segregated all’ condition the separation between rewards seems more believable, however the results observed strongly suggest that the reward segregation have a very low effect on the perceived value, being overridden by the segregation of losses. It is impossible to disentangle these two options from the current data.

However, the pattern is somewhat more complicated, as we saw large differences in pheromone deposition for the same treatment, depending on the treatments they were paired with. For example, while ants encountering the ‘segregated rewards’ deposit twice as much pheromone as ‘segregated all’ when they are paired, for ‘segregated rewards’ they deposit much less than when paired with the ‘bundled’ treatment (see Figure 4). A similar pattern is seen with ‘bundled’, where more pheromone is deposited when paired with ‘segregated all’ and less when paired with ‘segregated reward’. One possibility is that these differences arise due to a series of contrast effects. This pattern is further complicated by a strong effect of first encountered treatment (see Figure 5). In every condition the effects strengthen when the preferred option is presented first, and drops to chance level when presented second. This suggest that the bias generated through the bundling process has a similar strength to the bias for the first experienced odour (Oberhauser, 2019), and thus they appear to counterbalance each other.

Bundling and segregation of options to modify perceived value in insects may have ecological implications, especially in plant-pollinator interactions. If, by segregation, the perceived value of a reward can be increased, plants may be selected to split rewards amongst multiple smaller flowers in an inflorescence, or flowers on a plant. Similarly, they may be selected to attempt to bundle costs, favouring multi-flower inflorescences, allowing insects to walk between flowers, over multiple small, separate flowers.

The finding that bundling and segregation affect ant value perception adds this behavioural economic effect to several others which have also been shown to affect perceived value in insects, including decoy effects (Sasaki and Pratt, 2011; Shafir et al., 2002; Tan et al., 2015), invested effort increasing perceived value (Czaczkes et al., 2018a), relative value perception (Bitterman, 1976; Couvillon and Bitterman, 1984; Wendt et al., 2019), and labelling effects (Hemingway and Muth, 2022; Wendt and Czaczkes, 2020). It is becoming clear that the underlying patterns driving value perception in insects are in many ways parallel to those of humans. This implies either an extremely early evolutionary origin of shared perceptual mechanisms resulting in shared psychophysical laws or convergent evolution on similarly effective systems. The finding of a bundling and segregation effect, alongside the way in which insects respond to differences between experienced and expected rewards, strongly implies that insects share the same broad value function shape as humans (see Figure 1). Thus, we would predict that any other behavioural economic effects in humans, which arise from this value function, to also be present in insects. Still missing is a demonstration of an inflection point – that is, like in humans, losses loom larger than gains for insects. That our results imply bundling to more strongly influence losses than gain (see above) suggests this is the case, but formal testing will be required.

The absence of a significant preference in the Y-maze test, contrasted with the clear effect appreciable in the pheromone deposition, is puzzling. We have to consider the possibility that processes often assumed to be tightly linked to each other are instead separate. In this specific case, choice in the Y-maze depends on memory formation, as the choice is made by mentally comparing two options previously experienced. Pheromone deposition instead happens immediately after experience, and does not require any mental comparison to be expressed.

It is possible that ants have simply failed to associate each odour with the reward treatment, and thus chooses randomly in the Y-maze. This seems highly unlikely, as many examples of similar experiments on the same species are available in the literature, where ants are demonstrated to be extremely capable learners, learning even complex multimodal associations with fewer exposures than used in the current study (Czaczkes and Kumar, 2020; De Agrò et al., 2020).

The most likely explanation to us is that while a difference in the value of options is perceived, the same difference is not recorded during memory formation. Scalar utility theory (Kacelnik and Brito e Abreu, 1998; Rosenström et al., 2016) is an influential framework that has been developed to describe decision making under uncertainty. This theory postulates that encoding neurons, responsible for the internal representation of values (i.e. quantity, quality, delay, etc.), have logarithmically spaced sensitivities and specificities. Prospect theory (see Introduction) also assumes a process based on a logarithmic function. Thus, both make very similar behavioural predictions, even though one concentrates on value perception, and the other on memory acquisition. However, they do not have to be identical. Most experiments recording both pheromone deposition and subsequent choice use food quality as the measured unit (De Agrò et al., 2021; Oberhauser and Czaczkes, 2018; Wendt et al., 2019). Logarithmic perception of quality is probably the reason a mismatch between choice and instantaneous value perception has eluded discovery. We believe that the mismatch produced in this experiment has to do with the type of cost chosen as treatment: distance travelled. As previously mentioned, our treatment seemed unable to produce a segregation effect for the reward, while showing a strong segregation effect for losses, that is, the travelled distance. It is reasonable to assume that ants perceive value logarithmically, in accordance with prospect theory (De Agrò et al., 2021). In other words, the overall perception of the experience is affected by the artefacts of a log curve, segregation effect included (see Figure 1). Being produced instantaneously, pheromone deposition is likely linked to this information stream. However, when registering distances in memory ants need to be extremely precise. They possess an internal step counter and a visual odometer, which allows them to judge distances and return successfully to the nest (Narendra, 2007; Wittlinger et al., 2006). Encoding steps logarithmically would be absolutely insufficient for accurate homing, given the need to already cope with errors introduced in the path integration process (Merkle et al., 2006; Merkle and Wehner, 2010; Müller and Wehner, 1988; Schwarz et al., 2011). For this, a linear, 1:1 correspondence is required. It is possible that distance is memorized linearly and in a separated stream from the value perception. When asked to choose, the ant can only compare the two memories of distances, which due to their linear nature are immune to the segregation effect (Figure 6). If the ant odometer truly is a rare example of linear perception, it would be an invaluable system for investigating information processing in insects, as it allows the roles of perception and post-perceptual processing to be disentangled.

Figure 6

Download asset Open asset

The dissociation between immediate reaction and subsequent choice could also be explained by a separation between ‘liking’ and ‘wanting’. The duality between ‘liking’ and ‘wanting’, first described in rats (Berridge et al., 1989) and then confirmed extensively in humans (Brauer and De Wit, 1997; Leyton et al., 2002; Pool et al., 2015), distinguishes two aspects of reward. ‘Liking’ refers to acutely perceived hedonic reactions, while ‘wanting’ refers to motivation and desire for a goal. While ‘liking’ and ‘wanting’ usually co-vary, these are neurologically separate processes, and can, in vertebrates, be separately inactivated (Berridge et al., 1989) or enhanced (Berridge and Valenstein, 1991; Leyton et al., 2002; Treit and Berridge, 1990). A dedicated network for ‘wanting’ has been recently discovered in insects as well (Garcia and Dyer, 2022; Huang et al., 2022), making it very likely to be present in L. niger ants as well. Under this perspective, the immediate reaction to experiencing the reward (i.e. the pheromone deposited) may correspond to how much the animal ‘liked’ each option. The choice of which reward to head for in the Y-maze task may reflect how much ‘wants’ either option. However, neurological evidence would be required to support this speculation. Moreover, further experiments should test whether pheromone deposition and binary choice are indeed appropriate proxies of ‘liking’ and ‘wanting’, respectively.

Our work also demonstrates empirically that individual ants prefer near to distant food sources. Such a preference have been previously observed in nature (Frank and Linsenmair, 2017; Nyamukondiwa and Addison, 2014) but never directly tested. Ants also recruited significantly more to closer food source, as reported in this and other ant species as well (Devigne and Detrain, 2006; Fewell et al., 1992).

This study adds the bundling and segregation effect to the collection of value-distorting effects from behavioural economics which also affect animal value perception. Critically, by finding this effect in an invertebrate, we demonstrate that the complexities of the vertebrate brain are not required for these effects to manifest. Animal behaviour, much like human behaviour, can often be modelled as a value-maximizing system. However, deviations for strict economic rationality, such as the one demonstrated here, may play important roles in the animals’ ecology, especially in their foraging behaviour and biotic interactions.

Raw data collected in the experiments and used for the analysis are available in Source data 1.

1. 1. Bateson M
   2. Healy SD
   3. Hurly TA

   (2002) Irrational choices in hummingbird foraging behaviour

   Animal Behaviour 63:587–596.

   https://doi.org/10.1006/anbe.2001.1925

   • Google Scholar
2. 1. Bateson M
   2. Healy SD
   3. Hurly TA

   (2003) Context-Dependent foraging decisions in rufous hummingbirds

   Proceedings. Biological Sciences 270:1271–1276.

   https://doi.org/10.1098/rspb.2003.2365

   • PubMed
   • Google Scholar
3. 1. Beckers R
   2. Deneubourg JL
   3. Goss S

   (1992) Trails and U-turns in the selection of a path by the ant Lasius niger

   Journal of Theoretical Biology 159:397–415.

   https://doi.org/10.1016/S0022-5193(05)80686-1

   • Google Scholar
4. 1. Beckers R
   2. Deneubourg JL
   3. Goss S

   (1993) Modulation of TRAIL laying in the antlasius niger (Hymenoptera: Formicidae) and its role in the collective selection of a food source

   Journal of Insect Behavior 6:751–759.

   https://doi.org/10.1007/BF01201674

   • Google Scholar
5. 1. Ben-Shachar MS
   2. Lüdecke D
   3. Makowski D

   (2020) Effectsize: estimation of effect size indices and standardized parameters

   Journal of Open Source Software 5:2815.

   https://doi.org/10.21105/joss.02815

   • Google Scholar
6. 1. Berridge KC
   2. Venier IL
   3. Robinson TE

   (1989) Taste reactivity analysis of 6-hydroxydopamine-induced aphagia: implications for arousal and anhedonia hypotheses of dopamine function

   Behavioral Neuroscience 103:36–45.

   https://doi.org/10.1037//0735-7044.103.1.36

   • PubMed
   • Google Scholar
7. 1. Berridge KC
   2. Valenstein ES

   (1991) What psychological process mediates feeding evoked by electrical stimulation of the lateral hypothalamus?

   Behavioral Neuroscience 105:3–14.

   https://doi.org/10.1037//0735-7044.105.1.3

   • PubMed
   • Google Scholar
8. 1. Bitterman ME

   (1976) Incentive contrast in honey bees

   Science 192:380–382.

   https://doi.org/10.1126/science.1257773

   • PubMed
   • Google Scholar
9. 1. Brauer LH
   2. De Wit H

   (1997) High dose pimozide does not block amphetamine-induced euphoria in normal volunteers

   Pharmacology, Biochemistry, and Behavior 56:265–272.

   https://doi.org/10.1016/s0091-3057(96)00240-7

   • PubMed
   • Google Scholar
10. 1. Brooks M
    2. Kristensen K
    3. Benthem K
    4. Magnusson A
    5. Berg C
    6. Nielsen A
    7. Skaug H
    8. Mächler M
    9. Bolker B

    (2017) GlmmTMB balances speed and flexibility among packages for zero-inflated generalized linear mixed modeling

    The R Journal 9:378.

    https://doi.org/10.32614/RJ-2017-066

    • Google Scholar
11. 1. Camerer CF

    (2004)

    Advances in Behavioral Economics, The Roundtable Series in Behavioral Economics

    148–161, Prospect theory in the wild: evidence from the field, Advances in Behavioral Economics, The Roundtable Series in Behavioral Economics, New York, Princeton University Press.

    • Google Scholar
12. Book

    1. Camerer CF
    2. Loewenstein G
    3. Rabin M

    (2011) Advances in Behavioral Economics

    Princeton University Press.

    https://doi.org/10.2307/j.ctvcm4j8j

    • Google Scholar
13. 1. Clement TS
    2. Feltus JR
    3. Kaiser DH
    4. Zentall TR

    (2000) “ work ethic ” in pigeons: reward value is directly related to the effort or time required to obtain the reward

    Psychonomic Bulletin & Review 7:100–106.

    https://doi.org/10.3758/bf03210727

    • PubMed
    • Google Scholar
14. 1. Couvillon PA
    2. Bitterman ME

    (1984)

    The overlearning-extinction effect and successive negative contrast in honeybees (apis mellifera)

    Journal of Comparative Psychology 98:100–109.

    • PubMed
    • Google Scholar
15. 1. Czaczkes TJ
    2. Brandstetter B
    3. di Stefano I
    4. Heinze J

    (2018a) Greater effort increases perceived value in an invertebrate

    Journal of Comparative Psychology 132:200–209.

    https://doi.org/10.1037/com0000109

    • PubMed
    • Google Scholar
16. 1. Czaczkes TJ
    2. Koch A
    3. Fröber K
    4. Dreisbach G

    (2018b) Voluntary switching in an invertebrate: the effect of cue and reward change

    Journal of Experimental Psychology. Animal Learning and Cognition 44:247–257.

    https://doi.org/10.1037/xan0000171

    • PubMed
    • Google Scholar
17. 1. Czaczkes TJ

    (2018c) Using T- and Y-mazes in myrmecology and elsewhere: a practical guide

    Insectes Sociaux 65:213–224.

    https://doi.org/10.1007/s00040-018-0621-z

    • Google Scholar
18. 1. Czaczkes TJ
    2. Beckwith JJ
    3. Horsch AL
    4. Hartig F

    (2019) The multi-dimensional nature of information drives prioritization of private over social information in ants

    Proceedings. Biological Sciences 286:20191136.

    https://doi.org/10.1098/rspb.2019.1136

    • PubMed
    • Google Scholar
19. 1. Czaczkes T.J
    2. Kumar P

    (2020) Very rapid multi-odour discrimination learning in the ant Lasius niger

    Insectes Sociaux 67:541–545.

    https://doi.org/10.1007/s00040-020-00787-0

    • Google Scholar
20. Book

    1. Davies NB

    (2012)

    An Introduction to Behavioural Ecology

    Oxford; Hoboken, NJ: Wiley-Blackwell.

    • Google Scholar
21. 1. De Agrò M
    2. Oberhauser FB
    3. Loconsole M
    4. Galli G
    5. Dal Cin F
    6. Moretto E
    7. Regolin L

    (2020) Multi-Modal cue integration in the black garden ant

    Animal Cognition 23:1119–1127.

    https://doi.org/10.1007/s10071-020-01360-9

    • PubMed
    • Google Scholar
22. 1. De Agrò M
    2. Grimwade D
    3. Bach R
    4. Czaczkes TJ

    (2021) Irrational risk aversion in an ant

    Animal Cognition 24:1237–1245.

    https://doi.org/10.1007/s10071-021-01516-1

    • PubMed
    • Google Scholar
23. 1. Devigne C
    2. Detrain C

    (2006) How does food distance influence foraging in the ant Lasius niger: the importance of home-range marking

    Insectes Sociaux 53:46–55.

    https://doi.org/10.1007/s00040-005-0834-9

    • Google Scholar
24. 1. Emlen JM

    (1966) The role of time and energy in food preference

    The American Naturalist 100:611–617.

    https://doi.org/10.1086/282455

    • Google Scholar
25. 1. Evison SEF
    2. Petchey OL
    3. Beckerman AP
    4. Ratnieks FLW

    (2008) Combined use of pheromone trails and visual landmarks by the common garden ant Lasius niger

    Behavioral Ecology and Sociobiology 63:261–267.

    https://doi.org/10.1007/s00265-008-0657-6

    • Google Scholar
26. Book

    1. Fechner GT.

    (1860)

    Elemente der psychophysik

    Leipzig: Breitkopf und Härtel.

    • Google Scholar
27. 1. Fewell JH
    2. Harrison JF
    3. Stiller TM
    4. Breed MD

    (1992) Distance effects on resource profitability and recruitment in the giant tropical ant, paraponera clavata

    Oecologia 92:542–547.

    https://doi.org/10.1007/BF00317846

    • PubMed
    • Google Scholar
28. Book

    1. Fox J
    2. Weisberg S.

    (2011)

    An R Companion to Applied Regression (Second. ed)

    Thousand Oaks CA: Sage.

    • Google Scholar
29. 1. Frank ET
    2. Linsenmair KE

    (2017) Individual versus collective decision making: optimal foraging in the group-hunting termite specialist megaponera analis

    Animal Behaviour 130:27–35.

    https://doi.org/10.1016/j.anbehav.2017.06.010

    • Google Scholar
30. 1. Garcia JE
    2. Dyer AG

    (2022) Why do animals want what they like?

    Science 376:456–457.

    https://doi.org/10.1126/science.abp8609

    • PubMed
    • Google Scholar
31. Book

    1. Gescheider GA.

    (1997)

    Psychophysics: The Fundamentals, Auflage (Revised. ed)

    Mahwah, N.J: Lawrence Erlbaum Assoc Inc.

    • Google Scholar
32. Website

    1. Hartig F

    (2018) DHARMa: Residual Diagnostics for Hierarchical (Multi-Level / Mixed) Regression Models

    Accessed September 8, 2022.

    https://cran.r-project.org/web/packages/DHARMa/vignettes/DHARMa.html
33. 1. Hemingway CT
    2. Muth F

    (2022) Label-Based expectations affect incentive contrast effects in bumblebees

    Biology Letters 18:20210549.

    https://doi.org/10.1098/rsbl.2021.0549

    • PubMed
    • Google Scholar
34. 1. Huang J
    2. Zhang Z
    3. Feng W
    4. Zhao Y
    5. Aldanondo A
    6. de Brito Sanchez MG
    7. Paoli M
    8. Rolland A
    9. Li Z
    10. Nie H
    11. Lin Y
    12. Zhang S
    13. Giurfa M
    14. Su S

    (2022) Food wanting is mediated by transient activation of dopaminergic signaling in the honey bee brain

    Science 376:508–512.

    https://doi.org/10.1126/science.abn9920

    • PubMed
    • Google Scholar
35. 1. Hunter JD

    (2007) Matplotlib: a 2D graphics environment

    Computing in Science & Engineering 9:90–95.

    https://doi.org/10.1109/MCSE.2007.55

    • Google Scholar
36. 1. Jackson DE
    2. Châline N

    (2007) Modulation of pheromone TRAIL strength with food quality in Pharaoh’s ant, monomorium pharaonis

    Animal Behaviour 74:463–470.

    https://doi.org/10.1016/j.anbehav.2006.11.027

    • Google Scholar
37. 1. Johnson MD
    2. Herrmann A
    3. Bauer HH

    (1999) The effects of price bundling on consumer evaluations of product offerings

    International Journal of Research in Marketing 16:129–142.

    https://doi.org/10.1016/S0167-8116(99)00004-X

    • Google Scholar
38. 1. Kacelnik A
    2. Brito e Abreu F

    (1998) Risky choice and Weber’s law

    Journal of Theoretical Biology 194:289–298.

    https://doi.org/10.1006/jtbi.1998.0763

    • PubMed
    • Google Scholar
39. 1. Kahneman D
    2. Tversky A

    (1979) Prospect theory: an analysis of decision under risk

    Econometrica 47:263.

    https://doi.org/10.2307/1914185

    • Google Scholar
40. 1. Lakens D

    (2022) Sample size justification

    Collabra 8:33267.

    https://doi.org/10.1525/collabra.33267

    • Google Scholar
41. 1. Lakshminarayanan VR
    2. Chen MK
    3. Santos LR

    (2011) The evolution of decision-making under risk: framing effects in monkey risk preferences

    Journal of Experimental Social Psychology 47:689–693.

    https://doi.org/10.1016/j.jesp.2010.12.011

    • Google Scholar
42. 1. Lenth C
    2. Ctistis G
    3. Plachetka U
    4. Möller M
    5. Wackerbarth H

    (2020) Emmeans: estimated marginal means, aka least-squares means

    Scientific Reports 10:8260.

    https://doi.org/10.1038/s41598-020-65181-z

    • PubMed
    • Google Scholar
43. 1. Leyton M
    2. Boileau I
    3. Benkelfat C
    4. Diksic M
    5. Baker G
    6. Dagher A

    (2002) Amphetamine-Induced increases in extracellular dopamine, drug wanting, and novelty seeking: a PET/ [ 11C ] raclopride study in healthy men

    Neuropsychopharmacology 27:1027–1035.

    https://doi.org/10.1016/S0893-133X(02)00366-4

    • PubMed
    • Google Scholar
44. 1. Lydall ES
    2. Gilmour G
    3. Dwyer DM

    (2010) Rats place greater value on rewards produced by high effort: an animal analogue of the “ effort Justification ” effect

    Journal of Experimental Social Psychology 46:1134–1137.

    https://doi.org/10.1016/j.jesp.2010.05.011

    • Google Scholar
45. 1. MacArthur RH
    2. Pianka ER

    (1966) On optimal use of a patchy environment

    The American Naturalist 100:603–609.

    https://doi.org/10.1086/282454

    • Google Scholar
46. Software

    1. Magnusson A
    2. Skaug H
    3. Nielsen A
    4. Berg C
    5. Kristensen K
    6. Maechler M
    7. van BK
    8. Bolker B
    9. Sadat N
    10. Lüdecke D
    11. Lenth R
    12. O’Brien J
    13. Brooks M

    (2020)

    Generalized linear mixed models using template model builder, version V 2.0.2

    GlmmTMB.
47. 1. Mailleux AC
    2. Deneubourg JL
    3. Detrain C

    (2000) How do ants assess food volume?

    Animal Behaviour 59:1061–1069.

    https://doi.org/10.1006/anbe.2000.1396

    • PubMed
    • Google Scholar
48. 1. Mailleux AC
    2. Detrain C
    3. Deneubourg JL

    (2005) Triggering and persistence of trail-laying in foragers of the ant Lasius niger

    Journal of Insect Physiology 51:297–304.

    https://doi.org/10.1016/j.jinsphys.2004.12.001

    • PubMed
    • Google Scholar
49. 1. Mailleux A-C
    2. Detrain C
    3. Deneubourg J-L

    (2006) Starvation drives a threshold triggering communication

    The Journal of Experimental Biology 209:4224–4229.

    https://doi.org/10.1242/jeb.02461

    • PubMed
    • Google Scholar
50. Book

    1. Mankiw NG.

    (2011)

    Principles of Economics (International ed of 6th revised ed edition. ed)

    Mason, OH: South-Western.

    • Google Scholar
51. 1. Merkle T
    2. Knaden M
    3. Wehner R

    (2006) Uncertainty about nest position influences systematic search strategies in desert ants

    The Journal of Experimental Biology 209:3545–3549.

    https://doi.org/10.1242/jeb.02395

    • PubMed
    • Google Scholar
52. 1. Merkle T
    2. Wehner R

    (2010) Desert ants use foraging distance to adapt the nest search to the uncertainty of the path integrator

    Behavioral Ecology 21:349–355.

    https://doi.org/10.1093/beheco/arp197

    • Google Scholar
53. 1. Müller M
    2. Wehner R

    (1988) Path integration in desert ants, cataglyphis fortis

    PNAS 85:5287–5290.

    https://doi.org/10.1073/pnas.85.14.5287

    • PubMed
    • Google Scholar
54. 1. Narendra A

    (2007) Homing strategies of the Australian desert ant melophorus bagoti. I. proportional path-integration takes the ant half-way home

    The Journal of Experimental Biology 210:1798–1803.

    https://doi.org/10.1242/jeb.02768

    • PubMed
    • Google Scholar
55. 1. Naylor G
    2. Frank KE

    (2001) The effect of price bundling on consumer perceptions of value

    Journal of Services Marketing 15:270–281.

    https://doi.org/10.1108/EUM0000000005506

    • Google Scholar
56. 1. Noone BM
    2. Mattila AS

    (2009) Hotel revenue management and the Internet: the effect of price presentation strategies on customers’ willingness to book

    International Journal of Hospitality Management 28:272–279.

    https://doi.org/10.1016/j.ijhm.2008.09.004

    • Google Scholar
57. 1. Norton MI
    2. Mochon D
    3. Ariely D

    (2012) The IKEA effect: when labor leads to love

    Journal of Consumer Psychology 22:453–460.

    https://doi.org/10.1016/j.jcps.2011.08.002

    • Google Scholar
58. 1. Nyamukondiwa C
    2. Addison P

    (2014) Food preference and foraging activity of ants: recommendations for field applications of low-toxicity baits

    Journal of Insect Science 14:48.

    https://doi.org/10.1093/jis/14.1.48

    • PubMed
    • Google Scholar
59. 1. Oberhauser FB
    2. Czaczkes TJ

    (2018) Tasting the unexpected: disconfirmation of expectations leads to lower perceived food value in an invertebrate

    Biology Letters 14:20180440.

    https://doi.org/10.1098/rsbl.2018.0440

    • PubMed
    • Google Scholar
60. 1. Oberhauser FB
    2. Koch A
    3. Czaczkes TJ

    (2018) Small differences in learning speed for different food qualities can drive efficient collective foraging in ant colonies

    Behavioral Ecology and Sociobiology 72:1096.

    https://doi.org/10.1007/s00265-018-2583-6

    • Google Scholar
61. Thesis

    1. Oberhauser FB

    (2019)

    Individual cognition and collective behaviour in ants

    Springer.

    • Google Scholar
62. Book

    1. Oliphant TE.

    (2006)

    A guide to NumPy

    Trelgol Publishing USA.

    • Google Scholar
63. 1. Parrish AE
    2. Evans TA
    3. Beran MJ

    (2015a) Defining value through quantity and quality-chimpanzees (pan troglodytes) undervalue food quantities when items are broken

    Behavioural Processes 111:118–126.

    https://doi.org/10.1016/j.beproc.2014.11.004

    • PubMed
    • Google Scholar
64. 1. Parrish AE
    2. Evans TA
    3. Beran MJ

    (2015b) Rhesus macaques (macaca mulatta) exhibit the decoy effect in a perceptual discrimination task

    Attention, Perception & Psychophysics 77:1715–1725.

    https://doi.org/10.3758/s13414-015-0885-6

    • PubMed
    • Google Scholar
65. 1. Pool E
    2. Brosch T
    3. Delplanque S
    4. Sander D

    (2015) Stress increases cue-triggered “ wanting ” for sweet reward in humans

    Journal of Experimental Psychology. Animal Learning and Cognition 41:128–136.

    https://doi.org/10.1037/xan0000052

    • PubMed
    • Google Scholar
66. 1. Pyke GH
    2. Pulliam HR
    3. Charnov EL

    (1977) Optimal foraging: a selective review of theory and tests

    The Quarterly Review of Biology 52:137–154.

    https://doi.org/10.1086/409852

    • Google Scholar
67. Software

    1. R Development Core Team

    (2020) R: A language and environment for statistical computing, version 1.1.1

    R Foundation for Statistical Computing, Vienna, Austria.

    http://www.r-project.org
68. Software

    1. Reback J
    2. McKinney W
    3. jbrockmendel JDB
    4. Augspurger T
    5. Cloud P
    6. gfyoung S
    7. Klein A
    8. Roeschke M
    9. Hawkins S
    10. Tratner J
    11. She C
    12. Ayd W
    13. Petersen T
    14. Garcia M
    15. Schendel J
    16. Hayden A
    17. MomIsBestFriend VJ

    (2020) Pandas-dev/pandas, version 1.5.1

    Github.

    https://github.com/pandas-dev/pandas
69. 1. Rosenström T
    2. Wiesner K
    3. Houston AI

    (2016) Scalar utility theory and proportional processing: what does it actually imply?

    Journal of Theoretical Biology 404:222–235.

    https://doi.org/10.1016/j.jtbi.2016.06.003

    • PubMed
    • Google Scholar
70. 1. Sasaki T
    2. Pratt SC

    (2011) Emergence of group rationality from irrational individuals

    Behavioral Ecology 22:276–281.

    https://doi.org/10.1093/beheco/arq198

    • Google Scholar
71. 1. Scarpi D

    (2011) The impact of phantom decoys on choices in cats

    Animal Cognition 14:127–136.

    https://doi.org/10.1007/s10071-010-0350-9

    • PubMed
    • Google Scholar
72. 1. Schuck-Paim C
    2. Pompilio L
    3. Kacelnik A

    (2004) State-dependent decisions cause apparent violations of rationality in animal choice

    PLOS Biology 2:e402.

    https://doi.org/10.1371/journal.pbio.0020402

    • PubMed
    • Google Scholar
73. Software

    1. Schutten GJ
    2. Chan C
    3. Leeper TJ
    4. Foster J

    (2020)

    Read and write ODS files., contributors and other, version 1.1.1

    ReadODS.
74. 1. Schwarz S
    2. Wystrach A
    3. Cheng K

    (2011) A new navigational mechanism mediated by ant ocelli

    Biology Letters 7:856–858.

    https://doi.org/10.1098/rsbl.2011.0489

    • PubMed
    • Google Scholar
75. 1. Shafir S
    2. Waite TA
    3. Smith BH

    (2002) Context-Dependent violations of rational choice in honeybees (Apis mellifera) and gray jays (perisoreus canadensis)

    Behavioral Ecology and Sociobiology 51:180–187.

    https://doi.org/10.1007/s00265-001-0420-8

    • Google Scholar
76. 1. Tan K
    2. Dong S
    3. Liu X
    4. Chen W
    5. Wang Y
    6. Oldroyd BP
    7. Latty T

    (2015) Phantom alternatives influence food preferences in the eastern honeybee Apis cerana

    The Journal of Animal Ecology 84:509–517.

    https://doi.org/10.1111/1365-2656.12288

    • PubMed
    • Google Scholar
77. 1. Treit D
    2. Berridge KC

    (1990) A comparison of benzodiazepine, serotonin, and dopamine agents in the taste-reactivity paradigm

    Pharmacology, Biochemistry, and Behavior 37:451–456.

    https://doi.org/10.1016/0091-3057(90)90011-6

    • PubMed
    • Google Scholar
78. 1. Tversky A
    2. Kahneman D

    (1974) Judgment under uncertainty: heuristics and biases

    Science 185:1124–1131.

    https://doi.org/10.1126/science.185.4157.1124

    • PubMed
    • Google Scholar
79. 1. Tversky A
    2. Kahneman D

    (1981) The framing of decisions and the psychology of choice

    Science 211:453–458.

    https://doi.org/10.1126/science.7455683

    • PubMed
    • Google Scholar
80. Software

    1. Ushey K
    2. Allaire JJ
    3. Tang Y

    (2021) Reticulate: interface to “Python, version 2.0

    Github.

    https://rstudio.github.io/reticulate/
81. 1. van der Walt S
    2. Colbert SC
    3. Varoquaux G

    (2011) The numpy array: a structure for efficient numerical computation

    Computing in Science & Engineering 13:22–30.

    https://doi.org/10.1109/MCSE.2011.37

    • Google Scholar
82. Book

    1. Van Rossum G
    2. Drake FL.

    (2009)

    Python 3 Reference Manual

    Scotts Valley, CA: CreateSpace.

    • Google Scholar
83. Software

    1. Waskom M
    2. Botvinnik O
    3. O’Kane D
    4. Hobson P
    5. Lukauskas S
    6. Gemperline DC
    7. Augspurger T
    8. Halchenko Y
    9. Cole JB
    10. Warmenhoven J
    11. Pye C
    12. Hoyer S
    13. Vanderplas J
    14. Villalba S
    15. Kunter G
    16. Quintero E
    17. Bachant P
    18. Martin M
    19. Meyer K
    20. Miles A
    21. Ram Y
    22. Yarkoni T
    23. Williams ML
    24. Evans C
    25. Fitzgerald C
    26. Brian FC
    27. Lee A
    28. Qalieh A

    (2017) Mwaskom/seaborn, version v0.8.1

    Zenodo.

    https://doi.org/10.5281/zenodo.883859
84. Book

    1. Weber

    (1834)

    De Pulsu Resorptione Auditu et Tactu. Annotationes Anatomicae et Physiologicae

    C.F. Koehler, Lipsiae.

    • Google Scholar
85. 1. Wendt S
    2. Strunk KS
    3. Heinze J
    4. Roider A
    5. Czaczkes TJ

    (2019) Positive and negative incentive contrasts lead to relative value perception in ants

    eLife 8:e45450.

    https://doi.org/10.7554/eLife.45450

    • PubMed
    • Google Scholar
86. 1. Wendt S
    2. Czaczkes TJ

    (2020) Labeling effect in insects: cue associations influence perceived food value in ants (lasius niger)

    Journal of Comparative Psychology 134:280–292.

    https://doi.org/10.1037/com0000212

    • PubMed
    • Google Scholar
87. 1. Wickham H

    (2007) Reshaping data with the reshape package

    Journal of Statistical Software 21:1–20.

    https://doi.org/10.18637/jss.v021.i12

    • Google Scholar
88. 1. Wittlinger M
    2. Wehner R
    3. Wolf H

    (2006) The ant odometer: stepping on stilts and stumps

    Science 312:1965–1967.

    https://doi.org/10.1126/science.1126912

    • PubMed
    • Google Scholar
89. 1. Zentall TR

    (2015) When animals misbehave: analogs of human biases and suboptimal choice

    Behavioural Processes 112:3–13.

    https://doi.org/10.1016/j.beproc.2014.08.001

    • PubMed
    • Google Scholar

Decision letter after peer review:

Thank you for submitting your article "Bundling and segregation affects "liking", but not "wanting", in an insect" for consideration by eLife. Your article has been reviewed by 4 peer reviewers, and the evaluation has been overseen by a Reviewing Editor and Christian Rutz as the Senior Editor. The following individuals involved in the review of your submission have agreed to reveal their identity: Giorgia Silani (Reviewer #1); Erik Frank (Reviewer #2); Yoann Stussi (Reviewer #4).

The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted this to help you prepare a revised submission.

Essential revisions:

Overall, the reviewers agreed that the manuscript is of broad interest for many behavioral scientists in the fields of behavioural economics, decision-making, and reward processing. We were also positively impressed by the open practices in openly sharing data and analysis code in a very clear and detailed manner. However, the manuscript requires substantial revisions of the interpretation and discussion of the data as well as supplementary data to rule out some important alternative explanations. While there seems to be a consensus concerning the value of the demonstrated bundling versus segregation effects, we also agree that the interpretation in terms of "wanting versus liking" appears too speculative and questionable given the data. Therefore, we strongly suggest removing the interpretation in terms of motivational and hedonic processes and staying closer to the actual bundling effects which are innovative by themselves. In terms of alternative interpretations of the data, there are some major confounds that we would like to see addressed with supplementary data or supplementary analysis.

1) Concerning the results from the Y maze behavior: For interpretation of the results of the Y maze in terms of choice behaviour it is necessary to demonstrate that the associations between the reward conditions and the runway scents were successfully learned. The pilot experiment differs in many ways from the main experiment, therefore one cannot rule out that the lack of preference in the Y maze simply reflects a lack of learning. Also related to the interpretation of the behavior in the Y maze, the choice test is administered under extinction, an analysis of the first choice only would allow to rule out that the absence of preference reflects extinction processes.

2) Concerning the results from the pheromone deposit: A possible alternative interpretation of the pheromone deposit measure would be that it simply reflects the caloric intake rather than food value, it is worth controlling that across the different reward conditions the total amount of food consumed does not systematically vary and that the pheromone release did not vary according to the quantity of food that was consumed. More generally, the manuscript needs to provide more information about how pheromone deposition was measured and on the specificities of this measure, such as its physiological bases, timing properties, and granularity. Importantly, the results from the pheromone deposit measure are interpreted as if they support the hypothesis that segregated rewards with bundled costs should be the most "liked" option relative to bundled rewards and costs and segregated rewards and costs. But the data reported does not fully support this: The difference between the 'segregated rewards' condition and the 'bundled' condition is not statistically significant when all ants are considered. More nuance is needed so that this statistically nonsignificant result does not appear as being treated as statistically significant.

3) Concerning main experimental manipulation: the experimental design relies on the assumption that ants can perceive drops spaced by 5 mm as being segregated (two of 0.2 µl and 1 ad libitum). Since this is the key manipulation on which the experimental manipulation relies, we would like to see data demonstrating that spacing two food drops by 0.5 mm induces a segregated perception in ants.

4) Concerning the statistical approach: The choice of the random-effects structure for the GLMM analyses should be clearly motivated and justified. In particular, random intercepts were modelled (despite repeated measurements) and an appropriate specification of the random-effects structure is pivotal to reaching correct inferences when using mixed-effects models.

Reviewer #4 (Recommendations for the authors):

1) Concerning the ethics statement, it is mentioned 'No' next to animal subjects. Does this mean that no ethical approval was required or submitted for this study? If so, is this a typical procedure for studying with ants?

2) Across the paper, there is a lot of emphasis that the study is the first to report a bundling effect and a "wanting" / "liking" dissociation in ants. However, I don't think such an emphasis is necessary or even warranted as it doesn't provide helpful information to evaluate the importance or robustness of the results. I feel that the manuscript would benefit from removing these mentions or at least keeping them to a minimum.

3) It would be beneficial to report how the sample size for the main experiment was established or provide some considerations about the power of the study (e.g., a power sensitivity analysis to indicate the smallest effect size in terms of the odds ratio that could be detected with a certain power threshold given the sample size).

4) The degrees of freedom of the t-tests associated with the effects from the generalised linear mixed models (GLMM) should be consistently reported.

5) The choice of the random-effects structure for the GLMM analyses should be clearly motivated and justified. This is particularly important because only random intercepts were modelled (despite repeated measurements) and an appropriate specification of the random-effects structure is pivotal to reaching correct inferences when using mixed-effects models (see Barr et al., 2013. https://doi.org/10.1016/j.jml.2012.11.001; Bates et al., 2015. https://arxiv.org/abs/1506.04967; Frossard & Renaud, 2019. https://arxiv.org/abs/1903.10766).

6) Complementing the results with confidence intervals (e.g., 95% for statistical tests reporting a t statistics) around the effect sizes of interest could provide valuable information and be a nice addition to the manuscript. A useful R package to calculate confidence intervals for effect sizes is effectsize (https://cran.r-project.org/web/packages/effectsize/index.html; https://easystats.github.io/effectsize/).

7) I would like to commend the authors for openly sharing their data and analysis code in a clear and detailed manner.

Essential revisions:

Overall, the reviewers agreed that the manuscript is of broad interest for many behavioral scientists in the fields of behavioural economics, decision-making, and reward processing. We were also positively impressed by the open practices in openly sharing data and analysis code in a very clear and detailed manner. However, the manuscript requires substantial revisions of the interpretation and discussion of the data as well as supplementary data to rule out some important alternative explanations. While there seems to be a consensus concerning the value of the demonstrated bundling versus segregation effects, we also agree that the interpretation in terms of "wanting versus liking" appears too speculative and questionable given the data. Therefore, we strongly suggest removing the interpretation in terms of motivational and hedonic processes and staying closer to the actual bundling effects which are innovative by themselves. In terms of alternative interpretations of the data, there are some major confounds that we would like to see addressed with supplementary data or supplementary analysis.

1) Concerning the results from the Y maze behavior: For interpretation of the results of the Y maze in terms of choice behaviour it is necessary to demonstrate that the associations between the reward conditions and the runway scents were successfully learned. The pilot experiment differs in many ways from the main experiment, therefore one cannot rule out that the lack of preference in the Y maze simply reflects a lack of learning. Also related to the interpretation of the behavior in the Y maze, the choice test is administered under extinction, an analysis of the first choice only would allow to rule out that the absence of preference reflects extinction processes.

2) Concerning the results from the pheromone deposit: A possible alternative interpretation of the pheromone deposit measure would be that it simply reflects the caloric intake rather than food value, it is worth controlling that across the different reward conditions the total amount of food consumed does not systematically vary and that the pheromone release did not vary according to the quantity of food that was consumed. More generally, the manuscript needs to provide more information about how pheromone deposition was measured and on the specificities of this measure, such as its physiological bases, timing properties, and granularity. Importantly, the results from the pheromone deposit measure are interpreted as if they support the hypothesis that segregated rewards with bundled costs should be the most "liked" option relative to bundled rewards and costs and segregated rewards and costs. But the data reported does not fully support this: The difference between the 'segregated rewards' condition and the 'bundled' condition is not statistically significant when all ants are considered. More nuance is needed so that this statistically nonsignificant result does not appear as being treated as statistically significant.

3) Concerning main experimental manipulation: the experimental design relies on the assumption that ants can perceive drops spaced by 5 mm as being segregated (two of 0.2 µl and 1 ad libitum). Since this is the key manipulation on which the experimental manipulation relies, we would like to see data demonstrating that spacing two food drops by 0.5 mm induces a segregated perception in ants.

4) Concerning the statistical approach: The choice of the random-effects structure for the GLMM analyses should be clearly motivated and justified. In particular, random intercepts were modelled (despite repeated measurements) and an appropriate specification of the random-effects structure is pivotal to reaching correct inferences when using mixed-effects models.

We note that many of the comments were shared by multiple reviewers, and the editor effectively summarized the main, shared positions in the points below. To make our response easy to navigate, we added a numbering system to the different points made. We will include our main arguments as answers to the editors comments, and we will refer to those, while adding further specific comments, in the answers to reviewers below when these points are raised.

1. The liking vs wanting framework

While there seems to be a consensus concerning the value of the demonstrated bundling versus segregation effects, we also agree that the interpretation in terms of "wanting versus liking" appears too speculative and questionable given the data. Therefore, we strongly suggest removing the interpretation in terms of motivational and hedonic processes and staying closer to the actual bundling effects which are innovative by themselves.

Having seen how this is a shared opinion of multiple reviewers, we agree with greatly reducing our claims about liking vs wanting as an explanation to our results. Indeed, the experiments were not designed to test this specific segregation, and as such cannot make broad claims on its presence in our experimental subjects.

However, we still believe we can’t ignore the unusual split result between the binomial choice test and the pheromone deposition. This was unexpected for us, as in virtually all studies we ever carried out using this odor learning procedure we found a clear response in the Y-maze choice tests. Occasionally this was not mirrored in the pheromone data, but never the other way around. Note that it was never our intention to suggest that our paper was designed to test liking vs wanting. Indeed in the introduction we are very clear in saying that this reasoning came after observing the results (L99-106).

We however agree that our inclusion of this argument was too wide for this manuscript, given the ideas’ speculative nature. We agree that this paper does not demonstrate the presence of a separation in liking vs wanting in ants, but we feel we should at least mention it as a possible explanation to the difference in the effect between pheromone deposition and binomial choice. This position we believe is also sustained by the recently published paper by Huang, J. et al. (2022), that came out just a month after out submission to eLife.

Accordingly, we changed the title of the paper to “Bundling and segregation affects pheromone deposition, but not choice, in an ant” and removed every mention of the liking vs wanting literature from the introduction, focusing solely on economic theories and bundling vs segregation. In the discussion, we reduced greatly its contribution, but we mention liking vs wanting as a possible, yet speculative, explanation to our observation. We limit this discussion to one paragraph and refer to the recent Huang et al. ⁽²⁰²²⁾ paper as a demonstration of the existence of this dichotomy in insects, and our observation as a possible effect. We also make very clear that this is speculation, note statements of fact, by ending the paragraph with “However, neurological evidence would be required to support this speculation.” The mismatch between memory and perception is maintained as an alternative explanation, indicating also how this mismatch may cause a failure in memory itself, as was suggested by some reviewers.

In terms of alternative interpretations of the data, there are some major confounds that we would like to see addressed with supplementary data or supplementary analysis.

Regarding some of the requested additional data, intended to exclude alternative explanations, we believe we have sufficient evidence, stemming either from the literature or from other pilot experiments performed by us. We will present our argument below when brought up. We are of course open to discussing collecting more data if our position was to be found unsatisfactory.

2. The Y-maze results

2.1. Presence of learning

Concerning the results from the Y maze behavior: For interpretation of the results of the Y maze in terms of choice behaviour it is necessary to demonstrate that the associations between the reward conditions and the runway scents were successfully learned. The pilot experiment differs in many ways from the main experiment, therefore one cannot rule out that the lack of preference in the Y maze simply reflects a lack of learning.

Yes, this is an issue which affects all ‘negative results’ in behavior-based learning assays: it is impossible to distinguish lack of learning from learning but without a resultant behavioural response. Thus, as far as we can tell, is impossible to demonstrate learning under this exact bundling vs segregation setup as separate from preference.

However, we have good reasons to expect learning to have taken place, given their successful learning in other contexts using an identical odour-association choice assay. These ants learn such associations very well in a multi-reward context. In 2018 we performed a control experiment required for a different experiment. In that experiment, as in the experiment we report here, ants had to encounter multiple sucrose solution drops over the course of one visit, and associate all of those to an odor. We designed the pilot to demonstrate the ability of ants to associate a smell with a multi-reward visit. Each ant (n=20) would be allowed onto a 20cm long scented runway. At the end of it, the animal would encounter a small 0.2ul drop of sucrose solution. The solution could be either 1.5M or 0.25M, depending on the condition. After the ant had fully consumed the drop, a second one would be presented, and after that a third one, that the ant consumed until satiated. All the three drops had the same molarity. The ant was then allowed back to the nest, unloaded, and allowed to make a second visit. Here the ant would experience a differently scented runway, and presented with a different molarity drop (0.25M if 1.5M for the first visit and vice-versa). This procedure was repeated 8 times, for a total of 4 visits for each odor. In essence, this experiment is almost identical to the “Segregated reward” condition as presented in this manuscript, where we also presented a succession of three identical drops. At test, 90% of the ants chose the scent associated with high molarity for the first choice, 75% did so for the second choice, 70% for the third. This experiment demonstrates the ability of ants to associate an odor with a multi-reward experience. We are attaching the raw data for this pilot experiment to the response. If the reviewers and editor were to find this data crucial to the discussion of the current experiment, we can add the full pilot as a supplement.

Regardless, in the new version of the manuscript we give more space to the perception vs memory explanation (Figure 6) as an alternative explanation to the data. We raise the possibility that the ants simply didn’t learn the task, but the bundling vs segregation effect had an effect or weighting on their perception. We find this explanation unlikely, given the corroborating evidence we presented, and thus lean more towards an equal weight in memory but not in perception.

2.2. Results under extinction

Also related to the interpretation of the behavior in the Y maze, the choice test is administered under extinction, an analysis of the first choice only would allow to rule out that the absence of preference reflects extinction processes.

Regarding the test being administered under extinction, note that visit number was always included in the models in order to observe whether the subsequent testing had a decreasing effect of choice. Even when no difference was found between subsequent testing visits, we included the relevant post-hoc analyses too, regardless of the preference in the very first testing visit.

When looking only at the first choice, for condition 1, the ants chose the segregated option 51.6% of the times (p=1). For condition 2, they chose the segregated reward option 49.7% of the times (p=1). Lastly, in condition 3 they chose the segregated option 34.7% of the times (p=1). This data is available in the Supplementary file 1 (before supplement ESM2).

3. Pheromone deposition

3.1. The origin of pheromone deposited

Concerning the results from the pheromone deposit: A possible alternative interpretation of the pheromone deposit measure would be that it simply reflects the caloric intake rather than food value, it is worth controlling that across the different reward conditions the total amount of food consumed does not systematically vary and that the pheromone release did not vary according to the quantity of food that was consumed. More generally, the manuscript needs to provide more information about how pheromone deposition was measured and on the specificities of this measure, such as its physiological bases, timing properties, and granularity.

We feel fairly confident in excluding the possibility that pheromone deposited is linked to caloric consumption. First of all, ants participating in this experiment do not consume the liquid. The reward gets stored in the social stomach and not assimilated, and gets then passed on other members of the colony. As such, metabolic processes should not take place during the experiment.

More importantly, however, good evidence exists that, in this species, pheromone deposition is released after a specific volume threshold is consumed (A. C. Mailleux et al., 2000; A.-C. Mailleux et al., 2003, 2005). Critically, after passing this limit, ants which consume the same quality of food in the same manner deposit the same amount of pheromone, regardless of the specific volume drunk. Ants will not freely return to the nest before reaching this threshold, if allowed to feed further, and the gaster only becomes very visibly distended after this threshold. Thus, we are confident that all the ants in our experiment consumed more than this threshold food volume, and so even if the different treatments cause small changes in volume consumption, this should not affect their pheromone deposition.

Finally, regardless of the reported literature, we feel that even if the ants were to deposit more pheromone only as a byproduct of different caloric intake, the results would still support our interpretation. The only difference between the treatments is the segregation of rewards and costs itself. All the drops presented are of the same sizes across treatments, and they all carry the same molarity. As such, if the ants were to consume the food at a different rate or in a different amount, this variance would have to be caused by the segregation effect itself. Some evidence is available demonstrating that ants tend consume more overall volume for moderately high sucrose concentrations ^{(Sola & Josens, 2016)}, which would indicate that the segregation has an effect on the perceived value. In other words, even if pheromone was linked to caloric consumption, the deposition would act as a proxy of consumption rate, just another modulated measure.

We also now provide, when first introducing the pheromone deposition measurement, some more detail on the behavior, and how it correlates with the (perceived) value of a resource.

3.2. The presence of not statistically significant results

Importantly, the results from the pheromone deposit measure are interpreted as if they support the hypothesis that segregated rewards with bundled costs should be the most "liked" option relative to bundled rewards and costs and segregated rewards and costs. But the data reported does not fully support this: The difference between the 'segregated rewards' condition and the 'bundled' condition is not statistically significant when all ants are considered. More nuance is needed so that this statistically nonsignificant result does not appear as being treated as statistically significant.

Apologies, we were not very clear in our discussion here, and didn’t sufficiently distinguish out original hypotheses from our interpretation of the results. Our interpretation was that Bundling vs Segregation has an effect only on costs, but not on gains (L428-435, original manuscript version). This idea is derived from the data as is, considering the Bundling vs Segregated reward as identical, the bundling as preferred to the segregated all, the segregated reward being preferred to segregated all. Under this interpretation, the segregated rewards condition should NOT result different from the bundled condition, as in both the costs are bundled, and they only differ in the gains (L435-439, original manuscript version). For the same reason, Segregated reward (that presents bundled cost) is preferred to segregated all (that presents segregated costs); bundled (bundled cost) is preferred to segregated all (segregated costs).

We understand that the confusion originates from us repeating our a priori hypothesis in the discussion, which instead expected to see a difference in segregated reward vs bundles. We feel that is more honest of us to declare our a priori hypothesis, and then specify what the data suggest, rather than to give the impression to having had the second hypothesis all along. We address this issue by explicitly distinguishing our a priori hypothesis from our interpretation. L391, L394, L398, L403-404

Another reason for confusion is that we discussed at length the nuance of the results. We felt that declaring a result to be definitely at chance level when the p-value is 0.059 would be careless, especially considering the pattern of preference when the initial encounter is included.

We have now reworded the discussion and made clearer the expectation for each condition depending on the working hypothesis.

4. The segregation effect

Concerning main experimental manipulation: the experimental design relies on the assumption that ants can perceive drops spaced by 5 mm as being segregated (two of 0.2 µl and 1 ad libitum). Since this is the key manipulation on which the experimental manipulation relies, we would like to see data demonstrating that spacing two food drops by 0.5 mm induces a segregated perception in ants.

This issue is related to point 2 – how can we distinguish a lack of response to a perceived stimulus from a stimulus that is not perceived? We cannot. Indeed, we were very careful not to make strong claims that ants do not respond to reward segregation. We now state explicitly “…either because the ants do not segregate gains, or because our manipulation failed to trigger segregation”.

Happily, for our main claim that bundling and segregation influences the behavior of ants, it does not matter: this treatment was added in case reward segregation affected the ants in an equal and opposite way to cost segregation, since they both co-occur in the “all segregated” treatment. Has the response to “all segregated” and “all bundled” been indistinguishable, we would have required the “reward segregated” treatment to interpret the result: a lower reward perception of “segregated all” compared to “segregated all” would have been taken as evidence that reward and cost segregation, when co-occurring, balance each other out. As it happens, however, there was a difference between “all segregated” and “all bundled”, showing that costs are experienced differently if bundled or segregated. Unclear is only whether rewards are perceived as segregated, but have a weaker effect that segregated costs, or whether the ants only perceived the costs, but not the rewards, as segregated.

In summary: we agree completely that we cannot know whether the reward segregation treatment was really perceived as segregated by the ants. However, given the results, it does not matter for our main message.

To make this clear, we have now added these considerations to the discussion. L410-416

5. The choice of random effect structure

Concerning the statistical approach: The choice of the random-effects structure for the GLMM analyses should be clearly motivated and justified. In particular, random intercepts were modelled (despite repeated measurements) and an appropriate specification of the random-effects structure is pivotal to reaching correct inferences when using mixed-effects models.

Our choice of random effect structure was informed by our knowledge of the study species. Different colonies can have an overall different tendency to deposit pheromone. In each colony, different individuals can have a different tendency to deposit overall more or less pheromone. For this reason, the individual ants nested in colonies are included as a random intercept in the model: the overall pheromone deposited may vary per each individual independently from the treatment. However, we do not have the same expectation for across trials changes: an over-performer ant will consistently over-perform in terms of pheromone deposited, with no expectation for it to differentially change behaviour (Beckers et al., 1992). However, the evidence is sparse, and we are mostly relying on our own experience. Given this, prompted by one reviewer comments, we tried including visit number as a random slope for models looking at pheromone deposition (Binomial choice models only presented 3 subsequent visits, all being of equal treatment. As such we only included a random intercept). This turned out to be not feasible all the time, causing a failure in model convergence. We believe that the reason is the low number of repetitions per condition per subject: each performs only 4 visits to each segregation option. Moreover, visit 1 almost always results in no pheromone deposition, due to the existence of pre-training and generally low deposition on the first 1-2 visits to a food source, leaving probably too few repetitions to estimate the random slope. Following the suggestions of the reviewer, we tried simplifying the random effect structure in those cases, including Visit number as random intercept. This sometimes fixed the convergence failure, in which case we maintained it in the model, while in others the model kept failing to converge. In those cases we were forced to remove it from the final model.

Reviewer #4 (Recommendations for the authors):

1) Concerning the ethics statement, it is mentioned 'No' next to animal subjects. Does this mean that no ethical approval was required or submitted for this study? If so, is this a typical procedure for studying with ants?

Indeed, ants and all invertebrates (notably excluding cephalopods, and recently some crustaceans) are not protected by European law (or by any other country as far as we know) in research. As such, no ethical approval is required to perform experiments on these animals.

2) Across the paper, there is a lot of emphasis that the study is the first to report a bundling effect and a "wanting" / "liking" dissociation in ants. However, I don't think such an emphasis is necessary or even warranted as it doesn't provide helpful information to evaluate the importance or robustness of the results. I feel that the manuscript would benefit from removing these mentions or at least keeping them to a minimum.

We agree with the reviewer. See point 1.

3) It would be beneficial to report how the sample size for the main experiment was established or provide some considerations about the power of the study (e.g., a power sensitivity analysis to indicate the smallest effect size in terms of the odds ratio that could be detected with a certain power threshold given the sample size).

Sample size was set a-priori, but based on pragmatic considerations: we knew how much research time we could allocate to data collection, and could make a good estimate of how many ants could be tested in this time. This was then used to set the sample size. Using power analyses can be tricky when we have poor a-priori estimates for the effect size. We consider our pragmatic a priori sample size setting approach to be functional while avoid the possibility of p-hacking.

4) The degrees of freedom of the t-tests associated with the effects from the generalised linear mixed models (GLMM) should be consistently reported.

DF are now reported in the text, thank you.

5) The choice of the random-effects structure for the GLMM analyses should be clearly motivated and justified. This is particularly important because only random intercepts were modelled (despite repeated measurements) and an appropriate specification of the random-effects structure is pivotal to reaching correct inferences when using mixed-effects models (see Barr et al., 2013. https://doi.org/10.1016/j.jml.2012.11.001; Bates et al., 2015. https://arxiv.org/abs/1506.04967; Frossard & Renaud, 2019. https://arxiv.org/abs/1903.10766).

We would like to thank the reviewer greatly for the clarification provided via email. It helped us immensely in navigating the development of a random effect structure. We answered fully to the comment in the general response, and the changes are now appreciable in the supplement.

6) Complementing the results with confidence intervals (e.g., 95% for statistical tests reporting a t statistics) around the effect sizes of interest could provide valuable information and be a nice addition to the manuscript. A useful R package to calculate confidence intervals for effect sizes is effectsize (https://cran.r-project.org/web/packages/effectsize/index.html; https://easystats.github.io/effectsize/).

Thank you for suggesting the package! We did not know it and we found it quite useful. We added a calculation of the effect size in the analysis provided in the supplement. We however did not add it in the main text, to maintain readability. Indeed, effect size calculated by the effectsize package is directly derived from DF and t value, and as such we felt it would be redundant.

7) I would like to commend the authors for openly sharing their data and analysis code in a clear and detailed manner.

Many thanks! Open science is very important to us, and we work hard on it. It is pleasing to see that it is sometimes noticed.

Author details

1. Massimo De Agrò

   1. Animal Comparative Economics laboratory, Department of Zoology, University of Regensburg, Regensburg, Germany
   2. Department of Biology, University of Florence, Florence, Italy

   Contribution

   Conceptualization, Data curation, Software, Formal analysis, Supervision, Visualization, Methodology, Writing – original draft, Writing – review and editing

   For correspondence

   massimo.deagro@gmail.com

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0001-9284-5964

2. Chiara Matschunas

   Animal Comparative Economics laboratory, Department of Zoology, University of Regensburg, Regensburg, Germany

   Contribution

   Investigation, Methodology, Writing – review and editing

   Competing interests

   No competing interests declared

3. Tomer J Czaczkes

   Animal Comparative Economics laboratory, Department of Zoology, University of Regensburg, Regensburg, Germany

   Contribution

   Conceptualization, Supervision, Funding acquisition, Investigation, Writing – original draft, Project administration, Writing – review and editing

   Competing interests

   No competing interests declared

• 748

  Page views

• 55

  Downloads

• 0

  Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.