to choose A and to reject B at C-colored background and vice versa at D-colored background (at constant location). It is the first evidence of bees’ ability to make decisions depending on background color and the first evidence of wasps’ ability to perform situational learning. The described behaviors resemble “conditioned switching”, which is well known in vertebrates. Statistically significant individual differences between conspecifics have been recorded.
Резюме. В полевых экспериментах насекомое обучали выбирать одну из двух визуально различных фигур – A или B – в зависимости от ситуации (ситуационное обучение). Показано, что пчелы и осы способны: 1) выбирать A и отвергать B, если фигуры предъявляют на одном локальном участке фуражировочной территории, и, наоборот, выбирать B и отвергать A на другом участке, отстоящем от первого на 1–8 м; 2) выбирать A и отвергать B на фоне цвета C, а на фоне цвета D – наоборот. Способность пчел принимать решение в зависимости от цвета фона и способность ос к ситуационному обучению доказаны впервые. Описанные поведенческие события похожи на условнорефлекторное переключение, хорошо известное у позвоночных. Отмечены статистически достоверные индивидуальные различия среди изученных особей.
Introduction
This issue is dedicated to the memory of Professor
G.M. Dlussky, who is well known as a distinguished myrmecologist. However, his scientific interests covered widespread field of problems. He dealt not only with ants but also with many Hymenoptera and other insects including their ecology and behavior, especially foraging behavior of anthophilous insects [e.g., Dlussky, 2002, 2013]. The last PhD thesis study (of Sergey Lysenkov) supervised by Dlussky was aimed at insect pollinating activity investigation (V. Kartsev had the pleasure to act as opponent at the defence of this interesting work; the main results are published in: Lysenkov, [2009a, b]). It is foraging that requires the most complicated behavior and decision-making to obtain food under irregular conditions. During foraging, many animals demonstrate well developed individual learning and even intelligence – ability to perform abstract mental operations. Regarding non-human animals, now it is called “cognition”. Professor
G.A. Mazokhin-Porshnyakov provided the first evidence
of insects’ intelligent behavior about 50 years ago; namely “generalization of visual stimuli” (discrimination of rewarded and unrewarded feeders different in special abstract characters) [Mazokhin-Porshnyakov, 1968; Mazochin, 1969]. Since that moment, the honeybee has been serving as a model for investigation of cognition in insects. We also added paper wasps as an object of cognitive experiments.
Current research of insect cognition is a rapidly developing field and includes some scientific directions [for
92 V.M. Kartsev, O.V. Ryzhkova, Ya.A. Terehov
reviews see Reznikova, 2007; Shrinivasan, 2010; Menzel, 2012; Kartsev, 2014). One of the fundamental problems is the so-called contextual learning – the learning according to particular situation, or context. We pick up two kinds of contexts.
The first one is concerned with different behavioral activities (motivations). Thus, bees and wasps did not use individual foraging experience directly when searching for nest entrance and did not use the experience of searching for nest entrance when foraging [Kartsev, 1996]. Then the results were confirmed for bumble bees [Collborn et al., 1999; Fauria et al., 2002; Worden et al., 2005; Kartsev et al., 2005]. However, it turned out, contextual isolation is not absolute; some indirect interrelationships between learning under different activities were found. Contextual isolation is not concerned with cognition, but belongs rather to physiological phenomena.
The second kind of context occurs within one behavioral activity, in our case within searching for food in different situations/contexts. We call it situational choices (not to confuse with the first context) [Kartsev, 2014] or situational learning. Therefore, decision-making depends on particular situation; context serves as obligatory additional condition indicating how to reach a goal. For example, an animal had to choose object A and reject object B in the pair at the first location and vice versa at the second one. Obviously, an ability to form mental association between decision-making and additional condition is an evidence of cognition. Can insects do it? In this article, we give new evidence that they can.
Collett and Kelber [1988] were perhaps the first authors who demonstrated bees’ ability to learn tasks in a context- specific way. They used square constellations of 4 cylinders (2 blue cylinders on one side, and 2 yellow cylinders on the other) in two locations 33 m apart. At the first location, a feeder was placed between yellow cylinders, and at the second location between blue ones. Much later, in fruitful maze experiments concerned with contextual learning, one of the most fascinating bee abilities was discovered: the ability to master abstract inter-relationships, such as “sameness” and “difference” [Giurfa et al., 2001]. In a “delayed symbolic match-to-sample” task, experimental bees had to use the identity of a sample stimulus (which can be A or B) to choose between two other comparison stimuli (C and D) that were presented simultaneously at some distance from the first stimulus [Zhang et al., 1999, 2005]. In another series of maze experiments, left and right turns were indicated by different color cues placed on back wall of each box of the maze in which turning decision had to be made. For example, if a blue cue was presented, then “turn left!” and if yellow mark was presented, then “turn right!” The results revealed that bees learn this task well, just as well as the task of simply following an indicator [Zhang et al., 1996].
It should be noted as well that not only abstract paradigm of test task, but details of training procedure play very important if not primary role in cognitive experiments. Besides, each behavioral act depends on innate predispositions facilitating or inhibiting learning. Therefore, two logically similar tasks may be quite different in their neurological basis [Kartsev, 2014].
In the original experiments described in this article, situational learning was studied in a different way from that which was used in cited literature. Instead of landmarks in a maze or at a foraging area, two flat figures (rewarded and unrewarded) distinguished by color or by shape were presented at a small horizontal training table. As additional condition there was location of training table (distance between two locations varied from one to several meters) or color of background (for example, at the yellow background stimulus A was rewarded and B was unrewarded, while at the blue one it was vice versa).
Methods
Common methods of free flying insect training in field experiments were used [Mazokhin-Porshnyakov, 1968; Mazochin, 1969; Kartsev, 2014]. These methods are based on those developed by Karl von Frisch when studying color vision in bees [Frisch, 1914].
First of all an insect was attracted to a feeder positioned on a small table, 25 × 25 cm, covered with glass. Then the animal was marked by acrylic dye. In some cases, the insect learned two feeding places (locations). When it began to perform regular foraging trips to the table, a pair of visually different figures was offered. A small cup with 50% sugar water was placed at the center of rewarded figure. The unrewarded one contained concentrated sodium chloride solution (which served as a stimulus of aversive conditioning and facilitates learning – Prof. Zh. Reznikova, unpublished data, personal communication). These chemicals are known to be undistinguishable for bees and wasps from a distance. However, in the beginning of every new season odor control with unmarked cups was performed to exclude any odors absorbed by sugar or by salt used in experiments. After every visit, test figures were rearranged randomly in four positions. Positions 1–2 were situated in the direction “farther – nearer” in reference to the watcher (for bees, it was approximately the north to south direction); positions 3–4 were situated in the “left – right” (west to east) direction; if the position of the rewarded figure was 1, the other figure was placed in the opposite position 2 and so on.
Each cup was placed inside a small cylinder 1.5 cm high. Therefore, in order to taste searching object, the insect had to penetrate into the cylinder. It ensured that the insect had really chosen given figure. A choice was considered to be made after landing at the feeder and putting at least the head into the cylinder (Color plate 11: 1–2). Only the first choice in each foraging trip was taken into consideration. The choice was “correct” or “incorrect” depending on which figure, rewarded or unrewarded, was chosen. Proportions of correct and incorrect choices were analyzed statistically by (chi-square) test and by the modified Fisher’s test – φ-test [Plokhinsky, 1970].
In each experimental trial, only one subject took part, all others (recruited nest mates) were caged for the time of the experiment.
Colors or shapes of figures served as discriminative stimuli. Colors, undoubtedly distinguishable for bees, looked from the human point of view like violet (ca. 450 nm), blue (470 nm), green (500 nm), yellow (580 nm),
Abilities of honey bees Apis mellifera Linnaeus, 1758 and paper wasps Vespula spp. to situational learning Color plate 11.
_19_Kartsev_Ryzhkova_Terehov/Image_001.jpg)
Figs 1–2. Training table with rewarded and unrewarded feeders different in their colors (1) and the bee in the moment of choosing the given feeder (2).
Рис. 1–2. Дрессировочный столик с подкрепляемой и неподкрепляемой кормушками, различающимися по цвету (1), и момент выбора пчелой определенной кормушки (2).
_19_Kartsev_Ryzhkova_Terehov/Image_002.jpg)
Figs 3–4. Designs of experiments.
– the design of experiment 1: feeding location is the indicator of reward. After every insect visit, the training table was randomly placed at location 1 or location 2, and feeders were randomly rearranged. At location 1 pink was rewarded (+) and orange was unrewarded (–); at location 2 vice versa: orange was rewarded (+) and pink was unrewarded (–).
– the design of experiment 2: background color is the indicator of reward. After every insect visit, one of the two background colors was randomly presented and feeders were randomly rearranged. On blue background, orange was rewarded (+) and pink was unrewarded (–); on yellow background vice versa: pink was rewarded (+) and orange was unrewarded (–).
Рис. 3–4. Схемы экспериментов.
1 – схема эксперимента 1: подкрепление зависит от места кормления. После каждого прилета насекомого дрессировочный столик случайным образом ставили на место 1 или на место 2, а кормушки случайным образом переставляли. На месте 1 подкрепляли розовый (+) и не подкрепляли оранжевый (–), а на месте 2, наоборот, подкрепляли оранжевый (+) и не подкрепляли розовый (–).
4 – схема эксперимента 2: подкрепление зависит от цвета фона. После каждого прилета насекомого случайным образом выбирали один из двух возможных цветов фона, а кормушки случайным образом переставляли. На голубом фоне подкрепляли оранжевый (+) и не подкрепляли розовый (–); на желтом фоне, наоборот, подкрепляли розовый (+) и не подкрепляли оранжевый (–).
Table 1. Bees. Proportions of correct (+) and incorrect (–) choices of colors at two different locations.
Таблица 1. Пчелы. Распределения правильных (+) и ошибочных (–) выборов цветов на двух разных местах.
_19_Kartsev_Ryzhkova_Terehov/Image_003.jpg)
Notes. Signs (+) and (–) symbolize rewarded and unrewarded figures as well as correct and incorrect choices (choices of rewarded and unrewarded figure, respectively); P – statistical significance of differences: P1 – significance of differences between empirical and theoretical (+) : (–) proportions at the first location; null hypothesis ratio is 1 : 1; P2 – the same at the second location; P3 – significance of differences between proportions of choices “stimulus A
: stimulus B” (in present case, “pink : orange”) at different locations; if P3 < 0.05 ÷ 0.001, it means that this individual significantly preferred rewarded pink at the first location, while at the second one it preferred rewarded orange; ns – not significant.
* Labels of insects are taken from original experimental protocol, the extension after hyphen indicates year (-08 is 2008).
** Format of date: DD.MM.YY, 22.06.07 means 22 June 2007.
*** Different superscript-superscript letters in a line indicate cases when proportions (+) : (–) differ statistically significantly between feeding locations (small letters correspond to P < 0.05, capital letters correspond to P < 0.01).
**** For this bee, blue and yellow colors were used instead of pink and orange.
Примечания. Знаки (+) и (–) символизируют подкрепляемые и неподкрепляемые фигуры, как и правильные и ошибочные выборы (выборы подкрепляемой и неподкрепляемой фигур соответственно); P – статистическая значимость различий: P1 – достоверность различий между факти- ческим и теоретическим распределениями (+) : (–) на первом месте; нулевая гипотеза – распределение (+) : (–) соответствует распределению в от- ношении 1 : 1; P2 – то же на втором месте; P3 – достоверность различия распределений выборов «стимул A : стимул B» (в данном случае «розовый
: оранжевый») на разных местах, если P3 < 0.05 ÷ 0.001, это значит, что данная особь достоверно предпочитала подкрепляемый розовый на одном месте, в то время как на другом месте она предпочитала подкрепляемый оранжевый; ns – недостоверно.
* Номера насекомых даны по оригинальному экспериментальному протоколу, цифры после дефиса указывают год (-08 означает 2008).
** Формат даты: день – месяц – год, 22.06.07 означает 22 июня 2007 года.
*** Разные надстрочные буквенные индексы в пределах строки указывают на достоверность различий распределений (+) : (–) на разных местах (строчные буквы соответствуют достоверности P < 0.05, заглавные буквы соответствуют P < 0.01).
**** Для данной пчелы использовались синий и желтый цвета вместо розового и оранжевого.
orange (605 nm), pink (620 nm; really it was not diluted red invisible for bees as a color), and white (Whatman paper). These colors were different in their intensity as well; however, in the frame of our experimental paradigm it is not important which specific visual character the insect used when discriminating the figures. In the color experiments, the figures were circles 5 cm in diameter.
In the shape experiments, two figures were a circle (3.5 cm, in some variants 6 cm in diameter) and a cross as if it was constructed of five squares 2.5 × 2.5 cm. These figures were black on white background.
Results
Feeding location as an indicator of reward.
After each consecutive visit to the sweet lure, the training table was randomly positioned at one of two preliminary learned locations. At location 1 figure A in pair AB was rewarded, while at location 2 figure B was rewarded. Thus, two contrary tasks had to be solved in parallel depending on location of decision-making (Color plate 11: 3). The experimental insect, at least in the beginning of the experiment, often returned to previous
feeding location; however, if training table was returned (absent), quickly moved to the second location. Flying insects do this easily. The distance between locations varied from 1 to 8 m. This distance does not exceed individual foraging territory.
1-1. Bees. Bee learning abilities and other behavioral features are presented in Table 1.
Let us examine column “P3”. It indicates whether proportion of choices of colors “pink : orange” at the first location (where pink was (+) and orange was (–)) differs significantly from that at the second one (where, inversely, pink was (–) and orange was (+)). In other words, it indicates whether bees are able to categorize two contrary tasks in reference to additional condition – feeding location. Therefore, we consider that an individual solved a task (successfully learned), if P3 value is statistically significant. For example, for bee 2-07 on 24.06.07 (Table 1) proportion pink (+) : orange (–) at the first location is 36 : 17, while proportion pink (–) : orange (+) at the second location is 14 : 41. It is easy to calculate that the differences are highly significant (P3 < 0.001).
After a variable training period, four individuals out of
six solved the task. Two individuals did it during 60 visits
Table 2. Wasps. Proportions of correct (+) and incorrect (–) choices of colors at two different locations (distanced from each other by 8 m). Таблица 2. Осы. Распределения правильных (+) и ошибочных (–) выборов цветов на двух разных местах (отстоящих друг от друга на 8 м).
_19_Kartsev_Ryzhkova_Terehov/Image_004.jpg)
Notation as in Table 1. Примечания как в таблице 1.
in the first day and two others succeeded in the second or on the third day of learning. Thus, bees are able to solve our task making the decision depending on the location of discriminative objects.
One more conclusion is that 1 m distance is enough to serve as an additional condition (Table 1: bees 2-07 and 4-07 solved the task at one-meter distance). Nevertheless, distance between two locations seems to be important. None of three individuals which took part in one-meter distance variant solved the task during the first day of training (about 60 visits), but both individuals in 8 m distance variant did it. However, additional experiments are needed to form a well-grounded conclusion regarding the role of distance between feeding places.
Dynamic of learning is rather sophisticated. For example, for bee 2-07 correct choice portion increased significantly (P < 0.05) on the third training day in comparison with the second day. Therefore, this insect remembered its previous day experience. On the contrary, for bee 4-07 no pair of days differed significantly, and no long-term learning dynamics was revealed. Nevertheless, this bee solved the task statistically significantly during the second day of training and during all four training days in sum (total proportion pink (+) : orange (–) at the first location is 71 : 80, proportion pink (–) : orange (+) at the second location is 32 : 118; the differences are statistically significant, P < 0.001). Thus, as it is known to be in many other tasks, bees demonstrated individual features, and the speed of learning varied.
It is interesting to note that in some cases proportions (+) : (–) differed significantly between different locations. Superscript-Superscript letters in corresponding lines in Table 1 mark these cases. It has happened for four out
of six individuals. We suppose that it is associated with an inborn bee tendency to follow one preferred color: a bee learned the rewarded color at one location and then continued to choose it at the other one, where this color was unrewarded. This tendency is more evident in the beginning of training. For example, bee 4-07 on 27 June did not fly properly and performed only 30 visits preferring orange at both locations; at the first location this led to prevailing of incorrect choices. If training period was longer, this problem would be overcome through learning. 1-2. Wasps. The same experiment was performed also
on paper wasps Vespula spp. (= Paravespula = Paravespa). The only difference was that more contrasting colors, violet and orange, were used as discriminative stimuli, because color vision in wasps is not as perfect as it is in bees. The results are presented in Table 2.
Three of the four experimental individuals solved the task during 70–80 visits, the level of statistical significance being the highest (P < 0.001). This is the first evidence of wasps’ ability to situational learning.
In contrast to the bees, in the wasps studied, the percentage of correct choices was independent of the location of the training table (unlike to Table 1, in Table 2 there are no superscript letters in lines indicating such cases). It could be suggested that the tendency to follow a single color is stronger in bees than in wasps.
Background color as an indicator of reward.
The training procedure was performed at a constant feeding location; however, the background colors of the training tables were changeable. During consecutive visits, the insect unpredictably met the training table of color C or color D. In the pair of discriminative stimuli A and B, at the background C stimulus A was rewarded while at
Table 3. Bees. Proportions of correct (+) and incorrect (–) choices of colors at yellow and white training tables.
Таблица 3. Пчелы. Распределения правильных (+) и ошибочных (–) выборов цветов на желтом и белом дрессировочных столиках.
_19_Kartsev_Ryzhkova_Terehov/Image_005.jpg)
Notation as in Table 1. Примечания как в таблице 1.
Table 4. Bees. Proportions of correct (+) and incorrect (–) choices of colors at yellow and blue training tables.
Таблица 4. Пчелы. Распределения правильных (+) и ошибочных (–) выборов цветов на желтом и синем дрессировочных столиках.
_19_Kartsev_Ryzhkova_Terehov/Image_006.jpg)
Notation as in Table 1. Примечания как в таблице 1.
background D stimulus B was rewarded. Thus, instead of the feeding location, the condition affecting decision- making was background color (Color plate 11: 4).
2-1. Discrimination of colors in bees. For bees, two pairs of training tables were used in different variants of experiments: yellow/white and yellow/blue. Pink and orange circles were used as discriminative stimuli (like in location experiments).
The results are shown in Tables 3 and 4.
Fourteen individuals out of 21 tested (Tables 3 and 4 in sum) solved the task during about 80 visits on the first day and 4 individuals did it on the second day of the training period. In sum, 18 individuals among 21 demonstrated significant learning. Thus, the main conclusion is that bees are able to choose one or the other color in a pair depending on the background color. As far as we know, this is the first evidence of such an ability.
There is one more interesting fact to note. Our bees learned better at yellow and white training tables (Table 3) than at yellow and blue ones (Table 4). In the first variant, all 5 individuals solved the task during 50–60 visits. In the second variant only nine out of sixteen did so. Total proportions of (+) : (–) (calculated from data shown in the Tables) are 227 : 43 (84% +) and 729 : 377 (66% +). Differences between these proportions are statistically
significant (P < 0.001). At least three hypothetical reasons could explain this fact: 1) the complexity of each task depends on particular background colors; 2) differences between experimental variants are associated with individual differences between insects (colonies) studied or
with weather fluctuations. Each of these possibilities is worth investigating.
_19_Kartsev_Ryzhkova_Terehov/Image_007.jpg)
_19_Kartsev_Ryzhkova_Terehov/Image_008.jpg)