Signals between species can evolve whenever selection favors both the evolution of a signal display by a ‘sender’ species, and a response by a ‘receiver’ species (Bradbury and Vehrencamp, 2011; Maynard Smith and Harper, 2003). However, signal evolution is mediated not only by economics, but also by the psychology of receivers (Endler and Basolo, 1998; Guilford and Dawkins, 1993; Rowe, 2013; Ryan et al., 1990). In ecological communities, animals are faced with a diverse panoply of stimuli. How they categorize stimuli as worth responding to or ignoring will influence when and how signals evolve. Here, we explore how signal evolution is affected by the set of stimuli present in communities of different levels of complexity.

Warning signals are one of the best studied examples of interspecific communication – they advertise prey defenses to potential predators, reducing negative interactions for both predator and prey (Wallace, 1867). Mimicry occurs when warning signals coevolve among multiple prey species. Mimics can vary in their resemblance to models, with low fidelity (‘imperfect’) mimics representing something of a paradox (Cuthill and TD, 1993; Dittrich et al., 1993; Kikuchi and Pfennig, 2013; Sherratt and Peet-Paré, 2017). Understanding variation in the extent of mimicry is a problem that spans evolution, ecology, and cognitive psychology (Guilford and Dawkins, 1993; Mallet, 2001; Rowe, 2013; Ruxton et al., 2018), since selection on mimetic resemblance is mediated by the way that predators categorize prey as profitable or unprofitable (Beatty et al., 2004; Gamberale-Stille et al., 2012; Getty, 1985; Ihalainen et al., 2012; Kazemi et al., 2014; Kikuchi and Sherratt, 2015; Oaten et al., 1975; Sherratt, 2002; Sherratt and Peet-Paré, 2017; Speed and Ruxton, 2010).

The diversity of a community affects predator decisions about prey. For example, in an experiment with artificial prey, diversity affected how predators made decisions in response to warning signals that varied within a single, continuous dimension (Ihalainen et al., 2012). However, warning signals are often multicomponent, that is to say, complex – they consist of many different traits in concert (Bradbury and Vehrencamp, 2011; Hebets and Papaj, 2005; Maynard Smith and Harper, 2003). Indeed, genetic studies of mimicry complexes have revealed discrete variation among multiple mimetic traits (Clarke and Sheppard, 1963; Jiggins, 2017; Kunte, 2009; Dasmahapatra et al., 2012). Consequently, mimetic precision depends not only on how predators generalize within traits, but also on which traits they evaluate, and how they combine them to form higher-level categories. We use the terms ‘categorization’ and ‘generalization’ in the sense that categorization behavior results from using generalizations to make decisions (Seger and Peterson, 2013).

The use of ‘key’ traits is one simple way to classify prey (Balogh and Leimar, 2005; Beatty et al., 2004; Gamberale-Stille et al., 2012) – for example, using the rule ‘avoid yellow prey’ would mean that a predator would have to focus on the key trait of color (Figure 1A). However, the advantage of using key traits or any other form of categorization depends on the community in which these decisions are made (Beatty et al., 2004; Ihalainen et al., 2012; Lindström et al., 2004). In this study we examined the effects of communities on categorization, showing that different components of diversity have critical effects on which decision rules are used, and thus selection on mimetic signals.

Figure 1 with 2 supplements see all

Download asset Open asset

Figure 1—source data 1

Data used to generate Figure 1 and its supplements.

Includes results from the test trials of Experiments 1–5. Please refer to Supplementary file 1 for full description and analysis.

https://doi.org/10.7554/eLife.43965.005

Download elife-43965-fig1-data1-v2.csv

The simplest, most widely studied component of diversity is species richness, the number of species found in a community (Magurran, 1988). It might be difficult for predators to identify and remember the properties of individual prey types in rich communities with a large variety of prey. Predators could be limited by memory capacity (Beatty et al., 2004; MacDougall and Dawkins, 1998), or by the substantial risks of sampling unfamiliar species of prey, some of which might be highly unprofitable to attack (Cohen et al., 2007; Houston et al., 2012; Sherratt and Peet-Paré, 2017). If predators do not remember the characteristics of each discrete prey type but instead use rules (such as avoid yellow), then they could reduce the difficulty of deciding what to eat in a rich community. Therefore, it has been hypothesized that, as richness increases, predators will be more likely to use a key trait to make decisions (Beatty et al., 2004; Wilson et al., 2013). Indeed, Beatty et al. (2004) found that predators could use a key trait to make decisions in diverse communities; however, in their experiment, if predators did not use the key trait, no discrimination was possible at all. To make strong inference that increased richness causes predators to use a key trait for decisions, it helps to include the choice to use either the key trait, or a specific, reliable trait that has more values (by values, we mean unique states or versions). This way, it is possible to determine if predators would actually switch their behavior if they did not have to.

We designed virtual prey communities where predators could either use a completely reliable trait that had many (2-8) different values to perfectly classify prey as 'good' or 'bad', or simplify decision-making by using an unreliable key trait (binary, with only two values) at the price of committing more errors (Figure 1A). In our virtual communities, prey always had two traits (color and shape – which one was reliable and which one was unreliable was randomized). Both traits were discrete, meaning that they could take on different values that did not grade continuously into one another. If the predator learned to identify good prey based on values of the reliable trait R, it could forage without errors. That is to say P(good|$R_i^{+}$) = 1, where + indicates that the value R_i is positively correlated with profitability (for example, circle is always good in Figure 1A). The unreliable trait U was binary, having values of U⁺ and U^- (for example, prey each have one of two colors in Figure 1A). This binary trait only predicted whether an individual was 'good' with a probability of 0.78 (i.e. P(good|U⁺) = 0.78). When only two values of each trait existed in the community (e.g. circle vs. star and blue vs. yellow; Experiment 1 in Figure 1A), the same number of individuals could be classified using either shape or color. However, using the unreliable trait (e.g. color) would carry the cost of committing more errors. Cognitive psychology experiments suggest that in this situation, the reliable trait will be used to the exclusion of the unreliable one due to a phenomenon called the relative validity effect (Hall et al., 1977; Wagner et al., 1968). In the relative validity effect, when an animal can learn to associate two cues with an outcome, it will learn to use the one that is more reliable (valid).

We recruited undergraduate student volunteers to serve as predators on our virtual prey communities. Each subject learned to forage on a grid of 36 prey during a training trial where they were allowed to attack up to 18 of the prey, and received feedback on whether each was ‘good’ or ‘bad’ to eat in the form of a smiley face with a chirp or an X with a gong sound. Their ‘life bar’ would also rise or fall accordingly (subjects lost twice as much life for attacking ‘bad’ prey than they gained for eating ‘good’ prey). After subjects finished the training trial, they took a test trial where they could choose as many prey as they liked, but received no feedback (Figure 2). The test trial served two purposes: 1) it allowed us to measure subjects’ categorization behavior without changing it by providing feedback, and 2) because the test trial was always the same, it allowed us to compare subjects’ categorization behavior after foraging in different training communities. Subjects participated in five experiments presented in random order (Figure 2; Figure 2—figure supplement 1). In Experiment 1, our control to see if the relative validity effect held with our design, we found that the reliable trait was used almost exclusively (Supplementary file 1).

Figure 2 with 1 supplement see all

Download asset Open asset

We tested three mutually exclusive hypotheses for how predators will classify their prey as its phenotypic richness increases. Predators had to choose how much to rely on the key trait U at the price of committing some errors, or the completely reliable trait R at the price of learning about and memorizing multiple values. The first hypothesis was that they should select the former when the price of information (e.g. memory, costs of exploration) limits the profitability of using the reliable trait, so that as richness increases, they should use the unreliable trait to a greater degree (e.g. use color more and begin to ignore shape as the richness of shapes increases; Figure 3A). The second hypothesis was that if the relative validity effect were an invariant aspect of predator psychology, then predators should persist in using the reliable trait across different levels of richness (e.g., learn all of the shapes across Experiments 1–3, always ignoring color; Figure 3B). Indeed, associative learning experiments on the relative validity effect do not show a difference in which trait is used as the number of its values increases – that is, subjects always use the most reliable trait (Baetu et al., 2005; Murphy et al., 2001). However, the number of trait values in these experiments has been low. A third hypothesis is that at high levels of richness, predators may not be able to parse all of the information that they are confronted with and will guess randomly with respect to the reliable and unreliable traits (Figure 3C).

Figure 3

Download asset Open asset

To test these hypotheses, we performed two experiments (2 and 3) that had higher richnesses than Experiment 1. Subjects used the reliable and/or unreliable traits to make decisions in all experiments (Figure 1B), which allowed us to reject the hypothesis that they would not use either trait at high diversities. In Experiment 2, where there were four values of the reliable trait (two associated with profitability, two with unprofitability), subjects decreased their use of the reliable trait significantly (Figure 1C). In Experiment 3, where there were eight values of the reliable trait, subjects again significantly decreased their use of the reliable trait - in fact, they used the unreliable trait more (Figure 1C). These results allow us to reject the hypothesis that the relative validity effect is constant across levels of richness. Instead, they support the hypothesis that species-rich communities carry a high price of information, either in memory constraints or the risks of acquiring information, which increases predators’ tendency to use the simpler (binary) yet unreliable trait in decision-making. This is strong evidence supporting the hypothesis that in rich communities, mimicry could evolve easily on the basis of key features that predators use for identification (Beatty et al., 2004). It also supports important theoretical models of how mimicry evolves that depend upon key features (Balogh et al., 2010; Balogh and Leimar, 2005; Gamberale-Stille et al., 2012). Furthermore, this result is critical to the stability of warning signals that are parasitized by Batesian mimics because it implies that predators will not immediately switch to using more reliable traits simply because they are available. High species richness could still favor broad categorization.

The other component of community diversity is species evenness, or relative abundance (Magurran, 1988; Tuomisto, 2012). In every community, some species are common, while other species are rare. This ‘overrepresentation’ of some species and underrepresentation of others reduces the effective number of species in the community (i.e., lower evenness means effectively fewer species; Jost, 2010). Consequently, it is reasonable to hypothesize that unevenness will decrease predators’ use of the key trait to form categories, reversing the effect of increasing richness. In fact, prior work has shown that both higher frequencies of profitable, non-mimetic prey and lower frequencies of Batesian mimics relaxes selection on mimicry (Finkbeiner et al., 2018; Harper and Pfennig, 2007; Iserbyt et al., 2011; Lindström et al., 2004; Lindström et al., 1997; Pfennig et al., 2001). Here, we ask more generally about the effects of evenness per se, where there is a distribution of relative abundance within both profitable and unprofitable prey.

In uneven communities, rarer prey types will be less important food resources. This led us to test the hypothesis that unevenness will decrease predators’ tendency to categorize prey using an unreliable key trait. It predicts that in uneven communities, predators will use the reliable trait more than in an evenly distributed community with the same phenotypic richness (Figure 3D).

To test this hypothesis, we conducted two experiments with uneven communities. Experiment 4 featured the same eight phenotypes as Experiment 2, and Experiment 5 involved the precisely the same shapes and colors as Experiment 3, but in Experiments 4 and 5, one of the focal ‘good’ values and one of the focal ‘bad’ values of prey were much more abundant than the others. The reliable trait was used to a significantly greater degree in Experiment 5 compared to Experiment 3, as evinced by their placement in different post-hoc groupings (Figure 1C). However, we observed no difference in the use of the reliable trait between Experiments 2 and 4 (Figure 1C).

The hypothesis was rejected; unevenness increased subjects’ use of the reliable trait, but only at high levels of richness. We attribute this outcome to a simple cause. In both Experiments 4 and 5, where communities were uneven, predators mainly distinguished between the most abundant profitable value of the reliable trait, and all others. In other words, predators mainly attacked prey with the most abundant 'good' value, and categorized the other values as not worth attacking. Effectively, the most abundant, good, reliable value became a preferred alternative prey. In support of this interpretation, a model that included two different values for 'good' prey and two different values for 'bad' prey within the reliable trait R fit significantly better than one that lumped 'good' values together, and 'bad' values together (likelihood ratio test, ${\displaystyle {\chi }_{24}^2=70.1}$, P < 0.001). Predators attacked the abundant good prey much more than any other kind of prey in Experiment 5 and exhibited this behavior to a lesser degree in Experiment 4 (Figure 1—figure supplement 1).

Results from the uneven communities contrast with the pattern from Experiments 1–3, which shows increasing reliance on the unreliable trait with increasing richness. We suggest that the unevenness of a prey community will be negatively correlated with predators’ reliance upon key traits to form categories. This may make mimicry less likely to evolve in uneven communities. Furthermore, it connects evenness, a fundamental parameter of community ecology, to the concept of alternative prey from mimicry theory: when one species of profitable, non-mimetic prey is relatively abundant, selection on other, rarer prey to evolve mimicry will be relaxed (Getty, 1985; Holling, 1965; Ihalainen et al., 2012; Kokko et al., 2003; Lindström et al., 2004).

Very few studies from natural systems have collated the data that would be required to measure the relationship between community diversity and signaling systems. Wilson et al. (2013) argued that a negative relationship between mimetic precision and community diversity stems from increased generalization by predators in more diverse communities of velvet ants. Additionally, in experimentally manipulated communities of flowering plants, increased color diversity tended to increase visitation rates by pollinating insects (Fornoff et al., 2017). It would be interesting to know if this occurred because individual pollinators relied on coarser phenotypic categories in richer communities.

Our hypotheses might also be applicable to subsets of communities. Particularly, specialist predators might experience smaller prey communities than generalist predators, and specialist pollinators might visit fewer species of flowers than generalist pollinators. For example, different mimicry rings of Heliconius butterflies are segregated by microhabitat (Elias et al., 2008), and exposed to different suites of predators as a consequence; predation favors precise mimics within their preferred microhabitats (Willmott et al., 2017). Habitat specialization could reduce the size of the community about which a particular bird must learn, allowing them to select for precise mimicry (or none at all) because coarse categorization based on key traits would not occur.

Theoretical models suggest that other ecological conditions than those we explored here can also affect the number of traits that predators use to make decisions. Under some circumstances when the costs of attacking ‘bad’ prey are in a particular balance with the benefits of attacking ‘good’ prey, trusting only the most reliable trait may be most adaptive (Rubi and Stephens, 2016). Yet changing the cost:benefit ratio or underlying frequency of good prey can favor using multiple traits, or using no trait at all (Sherratt and Holen, 2018).

All data for this study are present in the supporting files, and source code to produce the figures from those files is included in the Supplementary RMarkdown file.

1. 1. Alatalo RV
   2. Mappes J

   (1996) Tracking the evolution of warning signals

   Nature 382:708–710.

   https://doi.org/10.1038/382708a0

   • Google Scholar
2. 1. Baetu I
   2. Baker AG
   3. Darredeau C
   4. Murphy RA

   (2005) A comparative approach to cue competition with one and two strong predictors

   Animal Learning & Behavior 33:160–171.

   https://doi.org/10.3758/BF03196060

   • Google Scholar
3. 1. Balogh AC
   2. Gamberale-Stille G
   3. Tullberg BS
   4. Leimar O

   (2010) Feature theory and the two-step hypothesis of müllerian mimicry evolution

   Evolution 64:810–822.

   https://doi.org/10.1111/j.1558-5646.2009.00852.x

   • Google Scholar
4. 1. Balogh ACV
   2. Leimar O

   (2005) Müllerian mimicry: an examination of Fisher's theory of gradual evolutionary change

   Proceedings of the Royal Society B: Biological Sciences 272:2269–2275.

   https://doi.org/10.1098/rspb.2005.3227

   • Google Scholar
5. 1. Beatty CD
   2. Beirinckx K
   3. Sherratt TN

   (2004) The evolution of müllerian mimicry in multispecies communities

   Nature 431:63–66.

   https://doi.org/10.1038/nature02818

   • PubMed
   • Google Scholar
6. 1. Beatty CD
   2. Bain RS
   3. Sherratt TN

   (2005) The evolution of aggregation in profitable and unprofitable prey

   Animal Behaviour 70:199–208.

   https://doi.org/10.1016/j.anbehav.2004.09.023

   • Google Scholar
7. Book

   1. Bolker BM

   (2008)

   Ecological Models and Data in R

   Princeton, NJ: Princeton University Press.

   • Google Scholar
8. Book

   1. Bradbury JW
   2. Vehrencamp SL

   (2011)

   Principles of Animal Communication (2nd ed)

   Sunderland: Sinauer Associates, Inc.

   • Google Scholar
9. 1. Brainard DH

   (1997) The psychophysics toolbox

   Spatial Vision 10:433–436.

   https://doi.org/10.1163/156856897X00357

   • PubMed
   • Google Scholar
10. 1. Caro TM

    (1995) Pursuit-deterrence revisited

    Trends in Ecology & Evolution 10:500–503.

    https://doi.org/10.1016/S0169-5347(00)89207-1

    • PubMed
    • Google Scholar
11. 1. Clarke CA
    2. Sheppard PM

    (1963) Interactions between major genes and polygenes in the determination of the mimetic patterns of Papilio dardanus

    Evolution 17:404–413.

    https://doi.org/10.1111/j.1558-5646.1963.tb03297.x

    • Google Scholar
12. 1. Cohen JD
    2. McClure SM
    3. Yu AJ

    (2007) Should I stay or should I go? how the human brain manages the trade-off between exploitation and exploration

    Philosophical Transactions of the Royal Society B: Biological Sciences 362:933–942.

    https://doi.org/10.1098/rstb.2007.2098

    • Google Scholar
13. 1. Cuthill IC
    2. Stevens M
    3. Sheppard J
    4. Maddocks T
    5. Párraga CA
    6. Troscianko TS

    (2005) Disruptive coloration and background pattern matching

    Nature 434:72–74.

    https://doi.org/10.1038/nature03312

    • PubMed
    • Google Scholar
14. 1. Cuthill IC
    2. TD B

    (1993)

    Mimicry and the eye of the beholder

    Proc R Soc B 253:203–204.

    • Google Scholar
15. 1. Dasmahapatra KK
    2. Walters JR
    3. Briscoe AD
    4. Davey JW
    5. Whibley AC
    6. Nadeau NJ
    7. Zimin A
    8. Hughes DST
    9. Ferguson LC
    10. Martin SH
    11. Salazar C
    12. Lewis JJ
    13. Adler S
    14. Ahn S-J
    15. Baker Da
    16. Baxter SW
    17. Chamberlain NL
    18. Chauhan R
    19. Counterman Ba
    20. Dalmay T
    21. Gilbert LE
    22. Gordon K
    23. Heckel DG
    24. Hines HM
    25. Hoff KJ
    26. Holland PWH
    27. Jacquin-Joly E
    28. Jiggins FM
    29. Jones RT
    30. Kapan DD
    31. Kersey P
    32. Lamas G
    33. Lawson D
    34. Mapleson D
    35. Maroja LS
    36. Martin A
    37. Moxon S
    38. Palmer WJ
    39. Papa R
    40. Papanicolaou A
    41. Pauchet Y
    42. Ray Da
    43. Rosser N
    44. Salzberg SL
    45. Supple Ma
    46. Surridge A
    47. Tenger-Trolander A
    48. Vogel H
    49. Wilkinson Pa
    50. Wilson D
    51. Yorke Ja
    52. Yuan F
    53. Balmuth AL
    54. Eland C
    55. Gharbi K
    56. Thomson M
    57. Gibbs Ra
    58. Han Y
    59. Jayaseelan JC
    60. Kovar C
    61. Mathew T
    62. Muzny DM
    63. Ongeri F
    64. Pu L-L
    65. Qu J
    66. Thornton RL
    67. Worley KC
    68. Wu Y-Q
    69. Linares M
    70. Blaxter ML
    71. Ffrench-Constant RH
    72. Joron M
    73. Kronforst MR
    74. Mullen SP
    75. Reed RD
    76. Scherer SE
    77. Richards S
    78. Mallet J
    79. Owen McMillan W
    80. Jiggins CD
    81. Heliconius Genome Consortium

    (2012) Butterfly genome reveals promiscuous exchange of mimicry adaptations among species

    Nature 487:94–98.

    https://doi.org/10.1038/nature11041

    • PubMed
    • Google Scholar
16. 1. Dittrich W
    2. Gilbert F
    3. Green P
    4. McGregor P
    5. Grewcock D

    (1993)

    Imperfect mimicry: a pigeon’s perspective

    Proc R Soc B 251:195–200.

    • Google Scholar
17. 1. Elias M
    2. Gompert Z
    3. Jiggins C
    4. Willmott K

    (2008) Mutualistic interactions drive ecological niche convergence in a diverse butterfly community

    PLOS Biology 6:e300–e309.

    https://doi.org/10.1371/journal.pbio.0060300

    • Google Scholar
18. 1. Endler JA
    2. Basolo AL

    (1998) Sensory ecology, receiver biases and sexual selection

    Trends in Ecology & Evolution 13:415–420.

    https://doi.org/10.1016/S0169-5347(98)01471-2

    • PubMed
    • Google Scholar
19. 1. Finkbeiner SD
    2. Salazar PA
    3. Nogales S
    4. Rush CE
    5. Briscoe AD
    6. Hill RI
    7. Kronforst MR
    8. Willmott KR
    9. Mullen SP

    (2018) Frequency dependence shapes the adaptive landscape of imperfect batesian mimicry

    Proceedings of the Royal Society B: Biological Sciences 285:20172786.

    https://doi.org/10.1098/rspb.2017.2786

    • Google Scholar
20. 1. Fornoff F
    2. Klein A-M
    3. Hartig F
    4. Benadi G
    5. Venjakob C
    6. Schaefer HM
    7. Ebeling A

    (2017) Functional flower traits and their diversity drive pollinator visitation

    Oikos 126:1020–1030.

    https://doi.org/10.1111/oik.03869

    • Google Scholar
21. Software

    1. Fox J
    2. Weisberg S
    3. Price B
    4. Adler D
    5. Bates D
    6. Baud-bovy G
    7. Bolker B
    8. Ellison S
    9. Firth D
    10. Friendly M
    11. Graves S
    12. Heiberger R
    13. Laboissiere R
    14. Maechler M
    15. Monette G
    16. Murdoch D
    17. Ogle D
    18. Ripley B
    19. Venables W
    20. Walker S
    21. Winsemius D
    22. Zeileis A

    (2018) Companion to Applied Regression, version 3.4.4

    Package 'car'.

    https://cran.r-project.org/web/packages/car/car.pdf
22. 1. Fraser S
    2. Callahan A
    3. Klassen D
    4. Sherratt TN

    (2007) Empirical tests of the role of disruptive coloration in reducing detectability

    Proceedings of the Royal Society B: Biological Sciences 274:1325–1331.

    https://doi.org/10.1098/rspb.2007.0153

    • Google Scholar
23. 1. Gamberale-Stille G
    2. Balogh ACV
    3. Tullberg BS
    4. Leimar O

    (2012) Feature saltation and the evolution of mimicry

    Evolution 66:807–817.

    https://doi.org/10.1111/j.1558-5646.2011.01482.x

    • Google Scholar
24. 1. Getty T

    (1985) Discriminability and the sigmoid functional response: how optimal foragers could stabilize Model-Mimic complexes

    The American Naturalist 125:239–256.

    https://doi.org/10.1086/284339

    • Google Scholar
25. 1. Guilford T
    2. Dawkins MS

    (1993) Receiver psychology and the design of animal signals

    Trends in Neurosciences 16:430–436.

    https://doi.org/10.1016/0166-2236(93)90068-W

    • Google Scholar
26. 1. Hall G
    2. Mackintosh NJ
    3. Goodall G
    4. Martello MD

    (1977) Loss of control by a less valid or by a less salient stimulus compounded with a better predictor of reinforcement

    Learning and Motivation 8:145–158.

    https://doi.org/10.1016/0023-9690(77)90001-7

    • Google Scholar
27. 1. Harper GR
    2. Pfennig DW

    (2007) Mimicry on the edge: why do mimics vary in resemblance to their model in different parts of their geographical range?

    Proceedings of the Royal Society B: Biological Sciences 274:1955–1961.

    https://doi.org/10.1098/rspb.2007.0558

    • Google Scholar
28. 1. Hebets EA
    2. Papaj DR

    (2005) Complex signal function: developing a framework of testable hypotheses

    Behavioral Ecology and Sociobiology 57:197–214.

    https://doi.org/10.1007/s00265-004-0865-7

    • Google Scholar
29. 1. Holling CS

    (1965) The functional response of predators to prey density and its role in mimicry and population regulation

    Memoirs of the Entomological Society of Canada 97:5–60.

    https://doi.org/10.4039/entm9745fv

    • Google Scholar
30. 1. Houston AI
    2. Trimmer PC
    3. Fawcett TW
    4. Higginson AD
    5. Marshall JA
    6. McNamara JM

    (2012) Is optimism optimal? functional causes of apparent behavioural biases

    Behavioural Processes 89:172–178.

    https://doi.org/10.1016/j.beproc.2011.10.015

    • PubMed
    • Google Scholar
31. 1. Ihalainen E
    2. Rowland HM
    3. Speed MP
    4. Ruxton GD
    5. Mappes J

    (2012) Prey community structure affects how predators select for mullerian mimicry

    Proceedings of the Royal Society B: Biological Sciences 279:2099–2105.

    https://doi.org/10.1098/rspb.2011.2360

    • Google Scholar
32. 1. Iserbyt A
    2. Bots J
    3. Van Dongen S
    4. Ting JJ
    5. Van Gossum H
    6. Sherratt TN

    (2011) Frequency-dependent variation in Mimetic fidelity in an intraspecific mimicry system

    Proceedings of the Royal Society B: Biological Sciences 278:3116–3122.

    https://doi.org/10.1098/rspb.2011.0126

    • Google Scholar
33. Book

    1. Jiggins CD

    (2017)

    The Ecology and Evolution of Heliconius Butterflies

    Oxford: Oxford University Press.

    • Google Scholar
34. 1. Jost L

    (2010) The relation between evenness and diversity

    Diversity 2:207–232.

    https://doi.org/10.3390/d2020207

    • Google Scholar
35. 1. Kazemi B
    2. Gamberale-Stille G
    3. Tullberg BS
    4. Leimar O

    (2014) Stimulus salience as an explanation for imperfect mimicry

    Current Biology 24:965–969.

    https://doi.org/10.1016/j.cub.2014.02.061

    • PubMed
    • Google Scholar
36. 1. Kikuchi DW
    2. Pfennig DW

    (2013) Imperfect mimicry and the limits of natural selection

    The Quarterly Review of Biology 88:297–315.

    https://doi.org/10.1086/673758

    • PubMed
    • Google Scholar
37. 1. Kikuchi DW
    2. Sherratt TN

    (2015) Costs of learning and the evolution of mimetic signals

    The American Naturalist 186:321–332.

    https://doi.org/10.1086/682371

    • PubMed
    • Google Scholar
38. 1. Kleiner M
    2. Brainard DH
    3. Pelli DG

    (2007)

    What’s new in Psychtoolbox-3?

    Perception 36:1–16.

    • Google Scholar
39. 1. Kokko H
    2. Mappes J
    3. Lindström L

    (2003) Alternative prey can change model-mimic dynamics between parasitism and mutualism

    Ecology Letters 6:1068–1076.

    https://doi.org/10.1046/j.1461-0248.2003.00532.x

    • Google Scholar
40. 1. Kunte K

    (2009) The diversity and evolution of batesian mimicry in papilio swallowtail butterflies

    Evolution 63:2707–2716.

    https://doi.org/10.1111/j.1558-5646.2009.00752.x

    • Google Scholar
41. 1. Lindström L
    2. Alatalo RV
    3. Mappes J

    (1997) Imperfect batesian mimicry—the effects of the frequency and the distastefulness of the model

    Proceedings of the Royal Society of London. Series B: Biological Sciences 264:149–153.

    https://doi.org/10.1098/rspb.1997.0022

    • Google Scholar
42. 1. Lindström L
    2. Alatalo RV
    3. Lyytinen A
    4. Mappes J

    (2004) The effect of alternative prey on the dynamics of imperfect batesian and müllerian mimicries

    Evolution 58:1294–1302.

    https://doi.org/10.1111/j.0014-3820.2004.tb01708.x

    • Google Scholar
43. 1. MacDougall A
    2. Dawkins MS

    (1998) Predator discrimination error and the benefits of müllerian mimicry

    Animal Behaviour 55:1281–1288.

    https://doi.org/10.1006/anbe.1997.0702

    • PubMed
    • Google Scholar
44. 1. Magrath RD
    2. Haff TM
    3. Fallow PM
    4. Radford AN

    (2015) Eavesdropping on heterospecific alarm calls: from mechanisms to consequences

    Biological Reviews 90:560–586.

    https://doi.org/10.1111/brv.12122

    • PubMed
    • Google Scholar
45. Book

    1. Magurran

    (1988)

    Ecological Diversity and Its Measurement

    Princeton: Princeton University Press.

    • Google Scholar
46. 1. Mallet J

    (2001) Mimicry: an interface between psychology and evolution

    PNAS 98:8928–8930.

    https://doi.org/10.1073/pnas.171326298

    • PubMed
    • Google Scholar
47. Book

    1. Maynard Smith J
    2. Harper D

    (2003)

    Animal Signals

    Oxford: Oxford University Press.

    • Google Scholar
48. 1. Murphy RA
    2. Baker AG
    3. Fouquet N

    (2001) Relative validity effects with either one or two more valid cues in pavlovian and instrumental conditioning

    Journal of Experimental Psychology: Animal Behavior Processes 27:59–67.

    https://doi.org/10.1037/0097-7403.27.1.59

    • Google Scholar
49. 1. Oaten A
    2. Pearce CE
    3. Smyth ME

    (1975) Batesian mimicry and signal detection theory

    Bulletin of Mathematical Biology 37:367–387.

    https://doi.org/10.1007/BF02459520

    • PubMed
    • Google Scholar
50. 1. Ollerton J
    2. Alarcón R
    3. Waser NM
    4. Price MV
    5. Watts S
    6. Cranmer L
    7. Hingston A
    8. Peter CI
    9. Rotenberry J

    (2009) A global test of the pollination syndrome hypothesis

    Annals of Botany 103:1471–1480.

    https://doi.org/10.1093/aob/mcp031

    • PubMed
    • Google Scholar
51. 1. Pfennig DW
    2. Harcombe WR
    3. Pfennig KS

    (2001) Frequency-dependent batesian mimicry

    Nature 410:323.

    https://doi.org/10.1038/35066628

    • PubMed
    • Google Scholar
52. 1. Rowe C

    (2013) Receiver psychology: a receiver's perspective

    Animal Behaviour 85:517–523.

    https://doi.org/10.1016/j.anbehav.2013.01.004

    • Google Scholar
53. 1. Rubi TL
    2. Stephens DW

    (2016) Should receivers follow multiple signal components? an economic perspective

    Behavioral Ecology 27:36–44.

    https://doi.org/10.1093/beheco/arv121

    • Google Scholar
54. Book

    1. Ruxton GD
    2. Allen WL
    3. Sherratt TN
    4. Speed MP

    (2018) Avoiding Attack (2nd ed)

    Oxford: Oxford University Press.

    https://doi.org/10.1093/oso/9780199688678.001.0001

    • Google Scholar
55. 1. Ryan MJ
    2. Fox JH
    3. Wilczynski W
    4. Rand AS

    (1990) Sexual selection for sensory exploitation in the frog physalaemus pustulosus

    Nature 343:66–67.

    https://doi.org/10.1038/343066a0

    • PubMed
    • Google Scholar
56. 1. Schaefer HM
    2. Ruxton GD

    (2009) Deception in plants: mimicry or perceptual exploitation?

    Trends in Ecology & Evolution 24:676–685.

    https://doi.org/10.1016/j.tree.2009.06.006

    • PubMed
    • Google Scholar
57. 1. Schaefer HM
    2. Ruxton GD

    (2010) Communication theory and the form of receiver-mediated selection

    Trends in Ecology & Evolution 25:383–384.

    https://doi.org/10.1016/j.tree.2010.04.003

    • Google Scholar
58. 1. Schielzeth H

    (2010) Simple means to improve the interpretability of regression coefficients

    Methods in Ecology and Evolution 1:103–113.

    https://doi.org/10.1111/j.2041-210X.2010.00012.x

    • Google Scholar
59. 1. Schuman MC
    2. Baldwin IT

    (2016) The layers of plant responses to insect herbivores

    Annual Review of Entomology 61:373–394.

    https://doi.org/10.1146/annurev-ento-010715-023851

    • PubMed
    • Google Scholar
60. 1. Seger CA
    2. Peterson EJ

    (2013) Categorization = decision making + generalization

    Neuroscience & Biobehavioral Reviews 37:1187–1200.

    https://doi.org/10.1016/j.neubiorev.2013.03.015

    • PubMed
    • Google Scholar
61. 1. Sherratt TN

    (2002) The evolution of imperfect mimicry

    Behavioral Ecology 13:821–826.

    https://doi.org/10.1093/beheco/13.6.821

    • Google Scholar
62. 1. Sherratt TN
    2. Whissell E
    3. Webster R
    4. Kikuchi DW

    (2015) Hierarchical overshadowing of stimuli and its role in mimicry evolution

    Animal Behaviour 108:73–79.

    https://doi.org/10.1016/j.anbehav.2015.07.011

    • Google Scholar
63. 1. Sherratt TN
    2. Holen Øistein H

    (2018) When should receivers follow multiple signal components? A closer look at the “flag” model

    Behavioral Ecology 29:e6–e8.

    https://doi.org/10.1093/beheco/ary043

    • Google Scholar
64. 1. Sherratt TN
    2. Peet-Paré CA

    (2017) The perfection of mimicry: an information approach

    Philosophical Transactions of the Royal Society B: Biological Sciences 372:20160340.

    https://doi.org/10.1098/rstb.2016.0340

    • Google Scholar
65. 1. Shrestha M
    2. Dyer AG
    3. Boyd-Gerny S
    4. Wong BB
    5. Burd M

    (2013) Shades of red: bird-pollinated flowers target the specific colour discrimination abilities of avian vision

    New Phytologist 198:301–310.

    https://doi.org/10.1111/nph.12135

    • PubMed
    • Google Scholar
66. 1. Smith JD
    2. Minda JP
    3. Washburn DA

    (2004) Category learning in rhesus monkeys: a study of the shepard, Hovland, and Jenkins (1961) Tasks

    Journal of Experimental Psychology: General 133:398–414.

    https://doi.org/10.1037/0096-3445.133.3.398

    • Google Scholar
67. 1. Smith JD
    2. Coutinho MVC
    3. Couchman JJ

    (2011) The learning of exclusive-or categories by monkeys (Macaca mulatta) and humans (Homo sapiens)

    Journal of Experimental Psychology: Animal Behavior Processes 37:20–29.

    https://doi.org/10.1037/a0019497

    • PubMed
    • Google Scholar
68. 1. Speed MP
    2. Ruxton GD

    (2010) Imperfect batesian mimicry and the conspicuousness costs of mimetic resemblance

    The American Naturalist 176:E1–E14.

    https://doi.org/10.1086/652990

    • PubMed
    • Google Scholar
69. 1. Tuomisto H

    (2012) An updated consumer’s guide to evenness and related indices

    Oikos 121:1203–1218.

    https://doi.org/10.1111/j.1600-0706.2011.19897.x

    • Google Scholar
70. 1. Wagner AR
    2. Logan FA
    3. Haberlandt K
    4. Price T

    (1968) Stimulus selection in animal discrimination learning

    Journal of Experimental Psychology 76:171–180.

    https://doi.org/10.1037/h0025414

    • PubMed
    • Google Scholar
71. 1. Wallace AR

    (1867)

    (Untitled)

    Proceedings of the Entomological Society of London pp. Ixxx–Ixxxi.

    • Google Scholar
72. 1. Willmott KR
    2. Robinson Willmott JC
    3. Elias M
    4. Jiggins CD

    (2017) Maintaining mimicry diversity: optimal warning colour patterns differ among microhabitats in amazonian clearwing butterflies

    Proceedings of the Royal Society B: Biological Sciences 284:20170744.

    https://doi.org/10.1098/rspb.2017.0744

    • Google Scholar
73. 1. Willson MF
    2. Whelan CJ

    (1990) The evolution of fruit color in Fleshy-Fruited plants

    The American Naturalist 136:790–809.

    https://doi.org/10.1086/285132

    • Google Scholar
74. 1. Wilson JS
    2. Jahner JP
    3. Williams KA
    4. Forister ML

    (2013) Ecological and evolutionary processes drive the origin and maintenance of imperfect mimicry

    PLOS ONE 8:e61610–e61617.

    https://doi.org/10.1371/journal.pone.0061610

    • PubMed
    • Google Scholar

Thank you for submitting your article "Community diversity determines predator generalization in mimicry" for consideration by eLife. Your article has been reviewed by three peer reviewers, including Bernhard Schmid as the Reviewing Editor, and the evaluation has been overseen by Ian Baldwin as the Senior Editor.

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

Summary:

This paper asks how generalist predators should choose their prey in species-rich prey communities if prey species are either fully palatable or fully unpalatable. The expectation is that trait generalization by predators should occur even at the expense of making some mistakes in choosing prey, because this increases foraging efficiency. The authors suggest that as a consequence of this expectation less perfect warning signals would be sufficient to deter predators in species-rich as compared with species-poor prey communities. As a further consequence, this might select for imperfect mimicry in diverse prey communities.

The initial expectation is tested with human subjects choosing mock prey species on a computer, of which 50% or species are good and 50% are bad, and all have a reliable (not generalized) and an unreliable (generalized) trait. The results generally confirm that in more diverse prey communities subjects switch from reliable to unreliable traits for identification. They also show that it is effective richness that matters, that is, species-rich but uneven communities still allow subjects to use the reliable trait of the abundant good prey species. Interestingly, subjects were better at learning to choose the good prey than at avoiding the bad prey. As a corollary, the theory may be more appropriate for mutualistic interactions such as between plants and pollinators, where generalization should lead to the evolution of similar flower traits between co-flowering species.

Essential revisions:

The reviewers identified three major areas that should be addressed during revisions. First, the scope of the paper is more focused than suggested by the title and Introduction. Second, the description of the statistical analysis is too sketchy and would not allow one to repeat the analysis. Third, the interpretation of the results should include a comparison with empirical data and discuss potential extensions and alternatives.

1) Scope: it seems misleading to focus on mimicry in the title and Introduction, because the paper is not really about selection of imperfect mimicry. Please choose a title that reflects your main results. Using humans as predators is clever, but still you need more justification (or mention a caveat) about your assumption that psychological decisions translate to natural predator behavior.

2) Statistics: the description of the data analysis is very hard to follow (especially the first, third and last paragraphs in the subsection “Data Analysis”). It is therefore hard to assess the appropriateness of the approach. It is important that you provide data and code so readers can do their own calculations. Generalized linear models with random terms can yield very different results depending on the specific method used. For example it can be dangerous to use readily available functions from R such as glmer, try at least a comparison with glm and constructing the correct approximate F-ratios from the mean deviance changes (make sure scale is estimated and not set to 1) (see e.g. textbook by McCullagh and Nelder). Also, weren't reliable and unreliable traits both present within each experiment? If so, how could there ever be an 'absence' of one of these traits (subsection “Data Analysis”, second paragraph)? There is also some concern why you are removing 'order' from the analyses. Why not partition out its effects, but remain focused on interpreting your main hypotheses? You can still interpret lower-order effects in a model, even if higher-order interactions are present if contrasts are properly coded (for an example, see Schielzeth, 2010). It appears that the statistical analyses deal with the data from the test trials of the different experiments (Figure 2). However, this is not explicitly stated in the manuscript.

3) Results and Discussion: first of all, consider renaming Results as 'Results and Discussion' and renaming Discussion as 'Conclusions'. This seems like a more logical grouping based on how the manuscript is written.

3a) Comparison with empirical data: please discuss to what extent imperfect mimicry is especially prominent in species-rich prey communities. Imperfect mimicry definitely occurs in species-poor prey communities, e.g. at high and low latitudes, whereas striking cases of highly perfected mimicry can be found in very species-rich communities. As an illustration, mimicry was discovered by Bates and Mueller through their perceived need for an explanation of striking similarity in appearance of certain butterfly species in very species-rich communities. Also, please look for publications/empirical data from observational studies and biodiversity experiments (e.g. German Exploratories or Jena Experiment and Cedar Creek biodiversity experiments, respectively) to compare with your predictions.

3b) Potential extensions and alternatives: if you want to keep the setting of co-evolution between prey and predators in diverse communities, you should expand the discussion to other possibilities besides generalists avoiding bad prey based on signal generalization. There is a potentially important aspect missing in your logic, namely that of competition between predators, which you should discuss. Usually, biodiversity experiments and observational studies find that with the number of prey species the number of predator species also increases. This is then tentatively explained by specialization (predator niche differentiation), which becomes possible with high prey diversity. Thus, while prey diversity may select for generalization in generalist predators, it may instead also select for the generalist predators to become more specialized on particular prey, thus making the need to identify bad prey obsolete, as indeed in part observed in your experiments. Provide more background about situations that require predators to negatively select out unpalatable prey as opposed to positively select palatable prey!

Essential revisions:

The reviewers identified three major areas that should be addressed during revisions. First, the scope of the paper is more focused than suggested by the title and Introduction. Second, the description of the statistical analysis is too sketchy and would not allow one to repeat the analysis. Third, the interpretation of the results should include a comparison with empirical data and discuss potential extensions and alternatives.

1) Scope: it seems misleading to focus on mimicry in the title and Introduction, because the paper is not really about selection of imperfect mimicry. Please choose a title that reflects your main results. Using humans as predators is clever, but still you need more justification (or mention a caveat) about your assumption that psychological decisions translate to natural predator behavior.

We have changed our title to read “Signal categorization by foraging animals depends on ecological diversity” which is more in keeping with our main results. We have also reworked the first part of our Introduction to make it more general:

“Signals between species can evolve whenever selection favors both the evolution of a signal display by a “sender” species, and a response by a “receiver” species (Bradbury and Vehrencamp, 2011; Maynard Smith and Harper, 2003). […] Here, we explore how signal evolution is affected by the set of stimuli present in communities of different levels of complexity.”

We have added a paragraph explaining our rationale behind our use of human subjects:

“Human volunteers can readily be recruited to participate in short computer games that are completely harmless and yield large quantities of data. […] Importantly, our experiments did not include XOR tasks, so human behavior is more likely to be representative of non-human species.”

2) Statistics: the description of the data analysis is very hard to follow (especially the first, third and last paragraphs in the subsection “Data Analysis”). It is therefore hard to assess the appropriateness of the approach. It is important that you provide data and code so readers can do their own calculations. Generalized linear models with random terms can yield very different results depending on the specific method used. For example it can be dangerous to use readily available functions from R such as glmer, try at least a comparison with glm and constructing the correct approximate F-ratios from the mean deviance changes (make sure scale is estimated and not set to 1) (see e.g. textbook by McCullagh and Nelder). Also, weren't reliable and unreliable traits both present within each experiment? If so, how could there ever be an 'absence' of one of these traits (subsection “Data Analysis”, second paragraph)? There is also some concern why you are removing 'order' from the analyses. Why not partition out its effects, but remain focused on interpreting your main hypotheses? You can still interpret lower-order effects in a model, even if higher-order interactions are present if contrasts are properly coded (for an example, see Schielzeth 2010). It appears that the statistical analyses deal with the data from the test trials of the different experiments (Figure 2). However, this is not explicitly stated in the manuscript.

We apologise that our code was not visible to the reviewers, although it was uploaded with our submission. In this revision, our analysis is available as an RMarkdown document (Supplementary file 1).

The analyses we have conducted in the revision were done with glm, as glmer failed to converge with the additional interactions requested.

Our new analysis now includes order, using centered variables so we can interpret the estimates of interest even in the presence of a significant interaction with order (please see revised Materials and methods).

We have modified the third paragraph of the subsection “Data Analysis” so that it is clear we mean different factor levels of each trait.

Thank you for referring us to Schielzeth, 2010. We have used centering to repeat our analysis with higher-level interactions, while retaining the ability to analyze lower-level effects.

We have specified that the test trial was analyzed.

3) Results and Discussion: first of all, consider renaming Results as 'Results and Discussion' and renaming Discussion as 'Conclusions'. This seems like a more logical grouping based on how the manuscript is written.

This is very sensible, we have made this change.

3a) Comparison with empirical data: please discuss to what extent imperfect mimicry is especially prominent in species-rich prey communities. Imperfect mimicry definitely occurs in species-poor prey communities, e.g. at high and low latitudes, whereas striking cases of highly perfected mimicry can be found in very species-rich communities. As an illustration, mimicry was discovered by Bates and Mueller through their perceived need for an explanation of striking similarity in appearance of certain butterfly species in very species-rich communities. Also, please look for publications/empirical data from observational studies and biodiversity experiments (e.g. German Exploratories or Jena Experiment and Cedar Creek biodiversity experiments, respectively) to compare with your predictions.

We have explored the literature to find such data, but very little is available. Certainly there are examples of poor mimicry at high latitudes, and precise mimicry at low latitudes, but currently data do not exist to test whether these are outliers, or the norm. We have suggested empirical studies worth more investigation at the end of the Results and Discussion, which include one from the Jena Experiment:

“Very few studies from natural systems have collated the data that would be required to measure the relationship between community diversity and signaling systems. […] It would be interesting to know if this occurred because individual pollinators relied on coarser phenotypic categories in richer communities.”

3b) Potential extensions and alternatives: if you want to keep the setting of co-evolution between prey and predators in diverse communities, you should expand the discussion to other possibilities besides generalists avoiding bad prey based on signal generalization. There is a potentially important aspect missing in your logic, namely that of competition between predators, which you should discuss. Usually, biodiversity experiments and observational studies find that with the number of prey species the number of predator species also increases. This is then tentatively explained by specialization (predator niche differentiation), which becomes possible with high prey diversity. Thus, while prey diversity may select for generalization in generalist predators, it may instead also select for the generalist predators to become more specialized on particular prey, thus making the need to identify bad prey obsolete, as indeed in part observed in your experiments. Provide more background about situations that require predators to negatively select out unpalatable prey as opposed to positively select palatable prey!

We are circumspect about competition between predators driving prey specialization directly; rich communities often have predator species that partition niche space spatially or temporally rather than by particular prey types (e.g. in species-rich Amazonia, some birds follow army ants, and some forage on individual territories). So it is not obvious how predator diversity would change our results. Nevertheless, the reviewers’ point is very valuable and well taken: not all predators will experience the full diversity of a community. We have elaborated on the potential importance of this work to specialist or generalist predators:

“Our hypotheses might also be applicable to subsets of communities. […] Habitat specialization could reduce the size of the community about which a particular bird must learn, allowing them to select for precise mimicry (or none at all) because coarse categorization based on key traits would not occur.”

Author details

1. David William Kikuchi

   Department of Ecology and Evolutionary Biology, University of Arizona, Tucson, United States

   Contribution

   Conceptualization, Software, Investigation, Visualization, Methodology, Writing—original draft, Writing—review and editing

   For correspondence

   dwkikuchi@gmail.com

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-7379-2788

2. Anna Dornhaus

   Department of Ecology and Evolutionary Biology, University of Arizona, Tucson, United States

   Contribution

   Writing—review and editing

   Competing interests

   No competing interests declared

3. Vandana Gopeechund

   Department of Biology, Carleton University, Ottawa, Canada

   Contribution

   Data curation

   Competing interests

   No competing interests declared

4. Thomas N Sherratt

   Department of Biology, Carleton University, Ottawa, Canada

   Contribution

   Conceptualization, Writing—review and editing

   Competing interests

   No competing interests declared

• 998

  Page views

• 121

  Downloads

• 9

  Citations

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