Compromise Between Individual Rationality and Collective Rationality in Decision-Making in Schistocerca Gregaria (The Desert Locust)
Received Date: November 01, 2021; Published Date: November 18, 2021
Animals living in groups interact with their environment and base their decision-making on the reliability of information sources such as the “personal information” and the “social information”. As such, we studied the manner in which the information choice is made, within a simple nonsocial animal model, by assessing the compromise between individual and collective rationality in decision-making in the Desert Locust Schistocerca gregaria during the appearance of a predator. We analyzed the relationship between response times (jumping and freeze), the quality of required personal information as a function of group size and we checked the relationship between the choice of use of information (personal vs. social) and arrangement of locusts. Results show that, in small groups, locusts behave by individual rationality and use personal information regardless of its quality. As the group expands, and when personal information is sketchy, locusts tend to share information and behave by collective rationality. However, when personal information is accurate locusts tend again to avoid sharing information adopting an individual rationality. This observed effect is due to the quality of required personal information and access to the information through congestion and distance between conspecifics. Results underline a recurrent individual rationality in locusts in decision-making. Furthermore, the strong adaptation capacity of this insect could lead to a collective rationality in large groups when the required personal information is imprecise, unreliable, or difficult to access, to provide streamlined decision-making.
Keywords: Locusts; Predator appearance; Information quality; Group size; Personal vs social information; Decision-making
Social animals living in groups interact between each other and with their environment by basing their choice of information for decision-making on the relative reliability on two sources of information such as the “personal information” and the “social information” [1,2]. “Personal information” is usually collected through environmental indices whereas “social information” comes from the behavior of conspecifics [1,3] such as advanced “signals” or social “clues” . “Signals” refer to intentional communication while indices refer to information produced incidentally by individuals as behavioral actions stating decisions of other group members or reflecting the indices on which such decisions are based [5-8].
The flexible combination of these distinct sources of information is potentially the basis for adaptive decision-making. Because of the dangerousness of the environment, animals are often forced to make both fast and accurate decisions in their search of food, housing, mating, and moving. Therefore, such adaptability of decisions cannot be based on the selection of an arbitrary combination of personal and social information. King and Cowlishaw  argue that individuals are more likely to make correct decisions when they share personal information with other conspecifics and when information is of good quality. Furthermore, the size of the group may also affect the reliability of the personal and social information.
As the group expands, the probability that social information is correct is likely to decrease in conditions where personal information is poor . Therefore, individuals would likely avoid relying on personal information as well as sharing it. In contrast, in cases of good personal information quality, individuals would promote the sharing of social information, which would properly inform the members of small groups and those less well-informed in major groups with the same probability of reaching a correct decision [9,15]. Through this compromise, animals in larger groups are able to minimize the investment in collecting personal information without affecting their ability to reach correct decisions. In other words, the social context significantly affects the reasoning and the manner in which information is acquired, transmitted and processed by the group members in social animals.
Reasoning in insects is cognitive processes involving inferences and conceptual models . Previous studies on learning, that are close to reasoning (process of information processing as well as reasoning), have examined the ecological settings favoring social over individual learning [11,12,13]. Social learning is the ability to integrate new information coming from congeners. This learning affects all major insect activities including feeding, predator avoidance, sexual behavior, and social interaction . The existence of parental care and overlapping generations in social animals largely contribute to the social learning . However, the way of reasoning and learning in social animals could be different or absent in nonsocial insects.
To further understand the reasoning of nonsocial animal within a society we conducted a study on gregarious Desert Locust Schistocerca gregaria (S. gregaria). It is well known that S. gregaria can be gregarious or solitaries . Through its polymorphism this animal model offers the opportunity to investigate different reasoning ways by keeping individuals living alone or in groups [17- 19]. Such procedure will improve the understanding of rationality, which is the ability to make inferences and conceptualization, for S. gregaria.
To analyze the compromise between individual versus collective rationality in decision-making for this particular animal model, we designed a simple experimental setting in which we introduced a predator and observed the reaction of different locust groups. First of all, we studied how individuals adjust their use of personal versus social information in decision-making in various sized groups. We subsequently attempted to establish how the quality of the required personal information affects the choice of use of information in decision-making.
Material and methods
The study was performed on Desert Locusts Schistocerca Gregaria (Figure 1a). The oothecas originated from a colony maintained at Arbiothech (society of production of Inoculum, Rennes-France). After hatching, locusts were distributed in cages with a volume of 0.1m3. In each cage we placed a population of about thirty locusts. They were subjected to a circadian rhythm of 14h light / 10h dark. The lighting was provided by a 60W lamp placed in each of the cages. The room temperature was maintained around 30 ± 2°C and at 45 ± 3% of humidity. Such conditions are close to the natural environment of the Desert Locusts. Locusts were maintained in these conditions until attaining the imago stage which is the last stage an insect attains during its metamorphosis including growth and development. For our experiments, we used 376 imagos (182 females and 194 males) having approximately the same age (8 weeks old ± 2 days).Experimental setup
To study the reaction of locusts when facing a predator, we built a simple experimental setup in which we placed locusts and we introduced a predator. The experimental device comprises a box having dimensions of 1.20 m length, 0.80 m width and 0.80 m height, and illuminated with a 160 W spotlight (Figure 1b and c). An aperture is made at the front of the box to allow the entrance of the predator (Figure 1b). As predator, we used a remote cat with dimensions of 0.24 m of height and 0.19 m of length and having a speed of 0.27 m/s. We chose this type of predator because locusts are likely to react to large objects in slow motion since they consider them as potential predator . Furthermore, a living animal is not easy to handle because of its unpredictable behavior during the experiments.
To study the effect of group size we created different groups.
The individual number per group (i) was defined according to the
relationship i = 3 N - 2 (with N the group number, 1 ≤ N ≥ 16 and 1
≤ i ≥ 46) allowing to create sixteen various sized individual groups
where the number of male and female is well known (Table 1).
Locusts were fed fresh grass supplemented with wheat bran. Locust
groups used during the experiment were deprived of food on the
previous day. The reason for this is that when we put them in the
experimental device on the day of the experiment, they fed and it
became easier to observe freeze behavior. For each experiment the
grass was replaced, and the walls of the box were cleaned after each
test with another locust group in order to evacuate any remaining
stress signal. After each experiment locust were replaced and never
To collect the useful information, a camera (camera GO PRO high definition, very wide viewing angle “1080 p”) was placed above the device to record the locust reactions (Figure 1c). Responses and behaviors of locusts facing predator’s approach were analyzed based on the recorded videos using the software “Sony Vegas Pro 13.0”.
Table 1: Choice of use of information in each groups in unknown context (UC) and familiar context (FC). F: female; M: male.
MethodsUnknown context vs. familiar context
For each group we established three experiments. As such we defined two different contexts: the unknown context (UC) and the familiar context (FC). The first experiment is considered as UC since the locusts were faced, for the first time, to the predator. To familiarize locusts with the experimental protocol, we removed the predator and waited for five minutes before introducing it again to the same locust group. We sat exposition duration of five minutes including predator entry and time spent with the predator. For the FC, we repeated the experiment two times after the first experience. Based on preliminary experiments and previous studies [21,22], we assumed that locusts can retain information (arousing fear) during an inter-trial interval of ten minutes when a stress factor is presented more than two times.Observation and data collect process
Upon the appearance of the predator, locusts have the choice between two alternatives which are either to jump or freeze (Figure 2). Jump behavior is when a locust moves from point A to point B by jumping or remove by flying. Freeze behavior is noticed by an immobilization of the locust and a flexion of the legs. For each behavior, we measured the response-time, referred to as jump time (JT) and freeze time (FT), for the unknown and the familiar context. Sometimes locusts do not react, neither jump nor freeze and continue to feed without moving. We did not consider this alternative in our study because it was very infrequent and the number of non-reacting locust was negligible comparing to the group size (one to two locusts in the larger groups). The locust decision is based on the choice of use of information picked up in its environment and in its field of view. Morphologically, S. gregaria has faceted eyes with binocular vision in the front, in back, and on the sides allowing horizontal vision field of 360° . However, vertically they have a blind field in the bottom . Locusts could have a limited field of view when there are obstacles or congeners around. To determine the choice of use of information in decisionmaking we analyzed all locusts’ reaction through their responding times chronology for each group.
First of all, we arranged group locusts depending on the
chronology of their response times. Then, in the vision field of each
locust, we noted the number of previous reacted congeners and the
presence or not of the predator. By focusing on the synchronization
of the locust with the different signals present in its vision field, we
deduced if the subject consider or not the behavior of congeners
and the entrance of the predator. By these analyzing procedures,
we defined three potential locusts’ information choices which are
“Personal”, “Social” or combined “Personal/Social”.
The information is “personal” when the subject reacts solely after facing a predator (freeze and jump). None congeners in its vision field reacted before him. We considered the information used in decision-making as “social” if the subject reacted according to conspecifics only. In this case, some congeners in its vision field reacted before it froze or it jumped. The subject did not care about the entrance of the predator, even if he saw it. Finally, if the decision of the locust is based both; on the entrance of the predator (the subject freeze) and behavior of the conspecifics (the same subject jumps after the reaction of congeners in its vision field), we considered that the locust used combined “personal/social” information in decision-making.
To understand the manner in which locusts would optimize their use of personal vs social rationality in decision-making facing the appearance of the predator, we compared the average response times for each group in the unknown and familiar contexts, then between the two contexts (Figure 3). Subsequently, we studied the choice of use of information depending on group size and context (UC and FC). For all groups, values of information choice are expressed as a percentage and reported in Table 1 and plotted in Figure 4.
We estimated the probability of the choice of use of information for Desert Locusts depending on the group size and the quality of personal information. We based all estimations on recorded video’s observations and we accomplished a statistical evaluation using the R open source-software (version 3.0.2, Foundation for Statistical Computing, Vienna, Austria). We performed an analysis of variance using Fisher test applied on regressions between the predefined parameters such as the average jumping times, average freeze times, and the group size, in both unknown and familiar contexts. The analysis was additionally applied on regressions between information choices and group size in both contexts. Furthermore, we applied Welch test , which is a generalization of student t-test (assuming normality) for unequal variance and sample sizes, as in this case, to check any statistical difference in the mean between the different set of parameters. The determination of the p-value allowed us to highlight any significant relationships between the studied parameters, and groups, i.e. test the statistical significance of the findings exhibiting the reliability of our results. We used a significance level of 5 %, i.e. p-value ≤ 0.05, which represents a confidence level of 95 %.
Response-time of the locusts facing predator
For the sixteen studied groups (Table 1), we measured the reaction times (expressed in seconds (s)) in terms of the average freeze times (FT), and the average jumping times (JT) during the entrance of the predator in unknown and familiar context (Figure 3a and b). For the unknown context (Figure 3a), results show that the average FT ranges between 0.40 ± 0.13 s and 0.85 ± 0.13 s with a mean value of 0.55 ± 0.09 s. The average JT ranges between 0.56 ± 0.12 s and 1.10 ± 0.12 s with a mean value of 0.81 ± 0.08 s. The p-values show additionally non-significant difference between the FT (p = 0.22) and the JT (p = 0.16) values. This indicates clearly that the reaction time is not correlated with the average freezing or jumping times, hence, do not exhibit any dependency (Table 2a).
Table 2: Correlation test (a) and means comparison test (b) for freeze and jumping times in an unknown and familiar context.
In a familiar context (Figure 3b), results show slightly higher
response times that in the unknown context and the average
freeze time values is about 0.64 ± 0.11 s (ranging between 0.39 s
and 0.81 s). The average jumping time is about 0.98 ± 0.15 s and
varies between 0.68 s and 1.44 s. The gender does not affect the
reaction time with a p-value of 0.1 for freeze reaction and 0.9 for
jump reaction. Similarly to the unknown context (Figure. 3a),
the reaction times in the familiar context are almost constant,
regardless of the group size (Figure 3b). No significant correlation
could be noticed for the average JT and average FT (p-values higher
than the significance level, Table 2 a). The group size does not seem
to influence the time the locusts take before a decision is made.
However, the reaction time seems to be affected by the gender for
freeze with a p-value of 0.03, but not for the jump reaction showing
a p-value of 0.13.
Welsh’s t-test was additionally performed to compare the average FT in both unknown and familiar context (Figure 3c). Interestingly, the average freeze time in the familiar context (mean average FT value of 0.64 ± 0.11 s) is slightly higher than in the unknown context (mean average FT value of 0.55 ± 0.09 s), however, the null hypothesis of equal means is not rejected as the p-value (0.027) is lower than the 5% significance level (Table 2b). Similar results were obtained when considering the average jumping time, (Figure 3d), showing no difference in the mean between the unknown and familiar context (p-value = 0.0026, Table 2 b).
Information choice and decision-making
We analyzed the behavior of locusts during the entrance of the predator in the two different contexts (Unknown and Familiar). We assessed the choice of use of information for decision-making (Personal, Personal/Social and Social) (Table 1) and plotted the choice of use of information (expressed in percentage) according to the group number (Figure 4). In the unknown context the use of personal information decreases significantly with group size while the use of social information increases significantly with the group size showing a confidence level of 99.9 % (p < 0.001, Table 3 a). For the combined personal/social information use, no important variation was noticed and the percentage is almost constant (Figure. 4a). In small groups 1 to 5 (comprising 1 to 13 locusts, Table 1), locusts significantly prefer the use of personal information compared to the combined personal/social information (p = 0.013) and social information (p = 0.0026, Table 3b). In medium to large groups 6 to 16 (comprising 16 to 46 locusts, Table 1), locusts tend to prefer the use of social information and combined personal/ social information rather than using the personal information before the decision is made (Table 3b). In the unknown context, the gender does not seem to affect the choice of use of information with p-values of 0.69, 0.23, and 0.19 for personal, social, and combined personal/social information, respectively. These values are higher than the 5% (i.e., 0.05) significance level.
Table 3: Correlations test (a) and Means comparison test (b) for the information choice in an unknown context.
Similarly to the unknown context, the use of personal information significantly decreases with group size (p = 0.0036) while the use of social information significantly increases with the group size (p = 0.0017, Table 4a). For the combined personal/social information use, no difference was noticed and the percentage is almost constant (p = 0.33, Table 4a) (Figure 4a). Locusts significantly prefer the use of personal information (P) (p = 0.0094) and combined personal/social information (p < 0.001) compared to social information in groups 1 to 5 (comprising 1 to 13); whereas, in groups 6 to 16 (comprising 16 to 46 locusts) there is no preference for the information used for decision-making (Table 4b and Figure 4b). Nonetheless, we found a significant preference for usage of combined personal/social information compared to social information in groups 6 to 16 (comprising 16 and 46 locusts respectively) (p < 0.001, Table 4b), hence, the group size seems to influence the use of personal and social information. In the familiar context, the gender tends to affect the choice of use of information (p = 0.050) for personal and combined personal and social information (p = 0.045), however, no significant correlation was found for the use of social information (p = 0.12). As main results from this part, we found that locusts seem to preferably use personal information in small groups and social information in large groups. Whatever the group size, locusts use more personal information in familiar than in unknown context. It suggests a significant influence of the context (or quality of personal information required) and the group size on the choice of use of information in decision-making.
Table 4: Correlations test (a) and means comparison test (b) for the information choice in a familiar context.
Effect of group size and collective organization in Decision-making
To study the effect of the group size on the decision-making process in the unknown and the familiar contexts, we analyzed in detail the behavior of locusts in a small-sized group (13 locusts), a medium-sized group (31 locusts), and a large-sized group (46 locusts). The grouping size is based on information choice and decision-making, where in small groups (groups 1 to 5 comprising 1 to 13 locusts) locusts prefer the use of personal information compared to the combined personal/social information and social information. In medium and large groups (groups 6 to 16 comprising 16 to 46 locusts), locusts tend to use the social information and combined personal/social information rather than using the personal information. For the different groups, we assessed the reactions of locusts during the entrance of the predator (Figure 5 and Table 5). We considered solely the first one second in our tests when the information choice was defined. The main reason was that one would expect that locusts may have chaotic behaviors (like jumping several times) after the first reacting individuals and the movements inside the group become random. This may alter our reasoning in analyzing the locust choice of information in decisionmaking.
Table 5: Information choice for small sized group (group five: 13 locusts), medium group (group eleven: 31 locusts), and large group (group sixteen: 56 locusts) in an unknown and familiar context. We consider the reacting locust only during the first second of the entrance of the predator. We noted P: Personal information choice; PS: combined Personal/Social information choice and S: Social information choice.
For the small-sized group (13 locusts), Figures 5a and b show the
locust position and the chronological reaction order in an unknown
and familiar context, respectively. We noticed that for both contexts,
during the entrance of the predator, the first reacting locust is far from the predator and its congeners. Successively reacting locusts
are distant from each other and reacted randomly regardless of
their fellows. Thus, we may conclude that the information on which
a decision is made is personal (Table 5).
For the medium-sized group (31 locusts), the first reacting locusts are distant from each other and are surrounded by few congeners (Figure 5c and d). However, we noted in the unknown context that some locusts have reacted successively at very close distances (Figure 5c). This phenomenon is accentuated in a familiar context (Figure 5d). Observation showed that the first reacting locusts use combined personal/social information in the unknown context and seem to use more social information, picked up from their near neighbors, in a familiar context.
For the large-sized group (46 locusts), locusts have the tendency to be gathered in small groupings (Figure 5e and 5f). The first reacting locusts are located near the predator and close to some congeners. Successively reacting locusts are less distant than in medium groups. Furthermore, there are more locusts in close proximity to each other that are reacting successively. We noticed in both the unknown and familiar contexts that locusts utilize mainly social information and combined personal/social information in decision-making.
The purpose of this study is to analyze how locusts adapt their
use of personal vs social information when facing a predator. We
analyzed several factors that may have an influence on decision
making such as the group size effect and the quality of the required
personal information in unknown and familiar contexts. To
highlight the compromise of information use, we assumed that
every individual decision based on social information is more
accurate when many subjects are involved, hence their personal
information becomes more accurate. Our results provide better
constraints on the strategy of use of information in decision-making
and the collective organization of these nonsocial insects.
Results also suggested that response times are not affected by group size and the average time is almost constant. Surprisingly, the average response times (freezing and jumping times) in a familiar context (Figure 3a), when personal information acquired is accurate, are higher than in an unknown context (Figure 3b), when personal information acquired is poor. The consistency of response times in each groups, and the delay in decision-making between the two contexts are likely due to the fact that locusts often base their response on the variability of information choice. This would reflect an adjustment in the decision-making strategy adopted by locusts (see Figure 4). Similarly to our results, numerous studies conducted on other animal models such as ants , honeybees  and monkeys  show that whatever the group size, the correctness of the decision is almost constant. To maintain this decision accuracy, animals do not use the same decision-making strategy in small and large groups.
Regardless the quality of personal information, locusts from small groups (group 1 to 5 containing 1 to 13 locusts) do not or rarely share the information and are likely to develop an individual rationality through the use of their personal information. In medium to large groups (6 to 16 containing 16 to 46 locusts) locusts tend to share information and, thus, develop a collective rationality through the use of social information. Nevertheless, this penchant for social information is affected by the quality of required personal information. On one hand, when the latter is poor, locusts develop a stronger collective rationality through the preferential use of social information and combined personal/social information (Figure 4a). On the other hand, when the required personal information is accurate, locusts use information without preference (Figure 4b). We note, nonetheless, for large groups a significant preference for the use of combined personal/social information compared to social information (Tables 4b and 5, Figure 4b). Locusts seem to privilege individual rationality in decision-making and develop a collective rationality in a large group when it is necessary.
A conflicting use of social and personal information of varying reliability in decision-making has further been stated on social forager . Van Bergen et al. , reported in a study of sticklebacks that when personal information is reliable, recently acquired and of good quality, sticklebacks ignore social information and based their decisions on personal information. Conversely, when the updated personal information is too old and conflicts with more recent social information or, if personal information is uncertain, sticklebacks lean towards recent social information. This antagonism in choice of information is strongly determined by the number of individuals holding information or present in the group [4,9,5,24,27]. However, according to our observations, other factors, than those cited above, seem to have an influence on the flexibility of use of information.
Moreover, the conflicting use of the information appears to be influenced by the distance between locusts which changes as a function of the group size (Figure 5). Locusts balance the choice of information regardless of the distance separating them from the predator, but depending on the number of individuals around them and the distance between each other. The greater the number of locusts is around and close to the subject, the more the latter interact with neighbors leads to better use of social information in decision-making . This finding has been emphasized by King and Cowlishaw , according to which nonsocial animals are more inclined to despotic decisions in nature by a local communication. This phenomenon has been observed in other animal model such as sticklebacks . In their study, stickleback tends to follow its surrounding neighbors and greatly increases the probability of following when more neighbors engage in a given direction. However, the emergence of sentinels, leader or leader group could arise from stochastic discrimination or predisposition factors.
Our results indicate that the first reacting locusts are distant from each other and are generally separated by several peers. For instance, in a large group size, locusts have the tendency to gather in small groups in which we can tentatively identify some monitoring locusts. These locusts, corresponding to the distant and first reacting individuals during the appearance of the predator, could be considered as “sentinels”. Moreover, Cocroft and Hamel, Couzin and Fernandez et al. [4,30,31] suggested that in predation circumstances, the decision should be quickly reached. Accordingly, a subset of informed individuals (sentinels) regularly updates the members of the group through repeated signals on the level of predation risk, including a decreased risk. This technique allows regulation of the response thresholds of the group members and reduces false alarms that lead to misinformation cascades consequent to a very rapid decision in which a positive feedback dominates. Furthermore, Cocroft and Hamel  suggested that animals belonging to large groups scrutinize their predators less frequently while maintaining their overall detection rate, which allows them more time to feed. King and Cowlishaw , stipulated that in nonsocial animals few individuals have relevant information for a current decision. Ultimately, this means that despotism is common in this animal type and is essential in behavior coordination and decision process. However, other factors appear as actors in these rational patterns and should be checked. Rands & al.  state that in situation of predation, if the danger is weak the subject will base on his own energetic reserves to decision-making. But if the danger is high the subject will base on his energetic reserves and on those of congeners to decision-making . Then, the emergence of collective vs individual rationality is monitoring by the subject energetic reserves and the danger level.
Results from previous studies on other animal models showed
good accordance with our findings which allowed us to propose
testable hypotheses about the relationship between the group
size, the quality of the required information and the streamlined
decision-making. However, the lack of matters to control all signals
received and processed by locust sense organs appears to be a
limitation in this study and allows drawing upon a certain number
of perspectives. It would be interesting to study how locusts perceive their environment, process information, and make such a
rapid and appropriate response called “paradigm of compromise
between speed-accuracy” as reported by Lorenz et al. and Marshall
et al. [26,33].
The gender seems to do not affect the response-time of locust and the choice of information in the unknown context. Whereas in the familiar context, our preliminary results show that the gender may have some effect. Such effect is not yet understood because the gender issues were not the focus of this study because this would complicate the problem. However, it would be interesting to conduct other studies by simplifying the problem and focusing solely on gender.
Desert Locust Schistocerca gregaria are nonsocial insects that can have gregarious or solitary nature. Our detailed study of this interesting animal model allowed us to bring better constraints to the behavior of such nonsocial insects. We, further, propose a simple protocol easily reproduced by any concerned researcher. Our results underline an individual rationality strongly developed and privileged in locusts in decision-making. Furthermore, and if it is necessary, the strong adaptation capacity of locusts could lead to a collective rationality in large groups. When personal information is sketchy or becomes unreliable or inaccessible, locusts search for acquired and low cost social information to improve personal information and to ensure a streamlined decision-making process. The group size and organization of the locust seems to have a direct effect on information choice in decision-making. In small groups, locusts behave by individual rationality and use personal information whatever the quality of personal information. As the group expands and when personal information is sketchy the locusts tend to share information and behave by collective rationality, whereas when personal information is accurate they again avoid information sharing. These observed rational patterns are not only due to the quality of required personal information or group size. They are also due to access to the information through congestion, distance between conspecifics and may be due to other multifactorial effects that it would be interesting to devote our attention to it in future research.
Conflict of Interest
No Conflict of Interest.
- Dall SR, Giraldeau LA, Olsson O, McNamara JM, Stephens DW (2005) Information and its use by animals in evolutionary ecology. Trends in Ecology & Evolution 20: 187-193.
- Van Bergen Y, Coolen I, Laland KN (2004) Nine-spined sticklebacks exploit the most reliable source when public and private information conflict. The Royal Society 271: 957-962.
- Grocott DFH (2003) Maps in mind—How animals get home? The Journal of Navigation vol. 56, pp. 1-14.
- Hamel JA, Cocroft RB (2012) Negative feedback from maternal signals reduces false alarms by collectively signaling offspring. Proceedings of the Royal Society B 279: 3820-3826.
- King AJ, Cowlishaw G (2007) When to use social information: the advantage of large group size in individual decision making. biology letters 3: 137-139.
- Danchin E, Giraldeau LA, Valone TJ, Wagner RH (2004) Public information: from nosy neighbors to cultural evolution. Science 305: 487-491.
- Giraldeau LA, Valone T J, Templeton JJ (2002) Potential disadvantages of using socially acquired information. The Royal Society 357: 1559–1566.
- Valon TJ (1989) Group foraging, public information, and patch estimation. Oikos 56: 357-363.
- Conradt L (2011) Models in animal collective decision-making: information uncertainty and conflicting preferences. Interface Focus 2: 226-40.
- Weber AA, Dyer AG, Combe M, Giurfa M (2012) Simultaneous mastering of two abstract concepts by the miniature brain of bees. Proceedings of the National Academy of Sciences 7481-7486.
- Boyd, R, Richerson PJ (1985) Culture and the Evolutionary Process. Chicago: University of Chicago Press.
- Galef BG, Giraldeau LA (2001) Social influences on foraging in vertebrates: causal mechanisms and adaptive functions. Animal Behaviour 61: 3-15.
- Laland KN (2004) Social learning strategies. Learning & Behavior 32: 4-14.
- Dukas R (2010) Insect social learning. M. Breed, J. Moore (Eds.), Encyclopedia of Animal Behavior, Academic Press, Oxford: 176-179
- Dukas R, Simpson SJ (2009) Locusts show rapid individual learning but no social learning about food. Animal Behaviour 78: 307-311.
- Chapman RF (2013) The insects: Structure and function. Cambridge University Press, Cambridge.
- Rogers SM, Cullen DA, Anstey ML, Burrows M, Despland E, et al. (2014) Rapid behavioural gregarization in the desert locust, Schistocerca gregaria entails synchronous changes in both activity and attraction to conspecifics. Journal of Insect Physiology 65: 9-26.
- Roessing P, Bouaïchi A, Simpson SJ (1998) Effect of sensory stimuli on the behavioural phase state of the desert locust, Schistocerca gregaria. Journal of Insect Physiology 44: 883-893.
- Heifetz Y, Voet H, Appledaum SW (1996) Factor affecting behavioral phase transition in the locust, Schistocerca gregaria (Orthoptera: Acrididae). Journal of chemical ecology 22: 1717-1734.
- Santer RD, Rind FC, Simmons PJ (2012) Predator versus prey: Locust looming-detector neuron and behavioural responses to stimuli representing attacking bird predators. PLOS ONE 7.
- Welch BL (1947) The generalization of "Student's" problem when several different population variances are involved". Biometrika 34 (1-2): 28-35.
- Couzin ID, Krause J, Franks NR, Levin SA (2005) Effective leadership and decision-making in animal groups on the move. Nature 433: 513-516.
- Fraser CP, Ruxton GD, Broom M (2006) Public information and estimation for group foragers: a re-evaluation of patch-quitting strategies in a patchy environment. OIKOS vol. 112: 311-321.
- Lorenz J, Rauhut H, Schweitzer F, Helbing D (2011) How social inﬂuence can undermine the wisdom of crowd effect. PNAS 108.
- Templeton JJ, Giraldeau LA (1996) Vicarious sampling: the use of personal and public information by starlings foraging in a simple patchy environment. Bihavioral Ecology and Sociobiology 38 105-114.
- Buhl J, Sword GA, Simpson SJ (2012) Using field data to test locust migratory band collective movement models. Interface Focus 2: 757-763.
- King, AJ, Cowlishaw G (2009) Leaders, followers, and group decision-making. Communicative & Integrative Biology 147-150.
- Couzin ID (2008) Collective cognition in animal groups. Trends in Cognitive Sciences 13: 36-43.
- Fernandez GJ, Capurro AF, Reboreda JC (2003) Effect of group size on individual and collective vigilance in greater rheas. Ethology 109: 413-425.
- Rands SA, Cowlishaw G, Pettifor RA, Rowcliffe JM, Johnstone RA (2008) The emergence of leaders and followers in foraging pairs when the qualities of individuals differ. BMC Evolutionary Biology.
- Marshall JAR, Dornhaus A, Franks NR, Kovacs T (2005) Noise, cost and speed-accuracy trade-offs: decision-making in a decentralized system. Journal of The Royal Society Interface 3: 243-254.