Visuo-spatial compound stimuli discrimination with (Gryllus pennsylvanicus) in two-choices rewarding learning tasks

Abstract

This paper proposes an experimental protocol allowing Gryllus pennsylvanicus to discriminate an A–A and A–B motif pairs of compound visual stimuli. Specifically, this study consists in an operant conditioning procedure including a dichotomous Y-maze, two different pairs of compound visual colored cues and a water reward. Results are conclusive for this visuo-spatial regularities study,(Gryllus pennsylvanicus) were able to significantly discriminate between the two compound visual patterns and learned the association with the reinforcer.

Introduction

Mastering abstract concepts represents a hallmark of intelligence and a prelude to reach higher cognitive skills (see review: Halford et al. 2010). It has been studied in numerous fields of science. One definition of an abstract concept is the representation of relational links between objects or their features. This could be applied to any sensory modality or to other abstract concepts (see review: Zentall et al. 2014). It is widespread in animals (see review: Zentall et al. 2008) and several bases are shared to compare the generality of their mechanisms (see review: Katz et al. 2007).

Same/Different judgement (S/D) represents a canonical example of abstract concepts which consists in comparing two things and a representation of the relation that holds between the two things. As an example of a Same/Different judgement, a two-item image with two circles and another two-item image with a circle and a triangle could be presented, then a reward is given if the Different response is chosen. This procedure is supposed to link the stimulus, the action and the reward. After several repetitions of these learned associations, including other forms of samples (i.e. rectangle and diamond), it is possible then, to further test if the animal could go beyond the specific stimuli used in training with a transfer to new samples. Finally, a transfer test example with novel stimuli (i.e. vertical and horizontal bars) could be used to validate the acquisition of the abstract concept, since it requires the transfer of the learning “rule” with unlearned stimuli (see review: Katz and Wright 2021). Those could be within or across other sensory modalities.

S/D concept was extensively studied with several animals (Blaisdell and Cook 2005; Cook and Wasserman 2007; Truppa et al. 2011; Newport et al. 2015; Magnotti et al. 2017; Fuss et al. 2021) (see also reviews: Chittka and Jensen 2011; Katz et al. 2007; Wright et al. 2021), various sensory modalities (Cook and Brooks 2009) and many experimental protocols (Bodily et al. 2008; Diaz et al. 2021). Astonishingly and despite their tiny brains, invertebrates like ants (see review: Czaczkes 2022), wasps (Weise et al. 2022) and bees (Brown and Sayde 2013) have also been shown to succeed in complex learning tasks (Finke et al. 2023). These insects appear to manipulate the S/D abstract concept (Muszynski and Couvillon 2015) as well as other types, and even more than one abstract concept simultaneously (Avarguès-Weber et al. 2010, 2012a; Giurfa 2021).

Hence, the cricket seems also a suitable animal model to study cognition (see review: Mizunami and Matsumoto 2017a). For this insect, the learning and memory phenomena were deeply studied as reviewed by Mizunami et al. (2017) and see review: Matsumoto (2022). First, simple association tasks and associative learning paradigms were previously shown with different insect species such as crickets (Hollis and Guillette 2015; Mizunami and Matsumoto 2017; Mizunami et al. 2018). This reflects the universality of these lower cognitive features that is more likely conserved throughout evolutionary lineages for different fitness advantages. Specifically, classical conditioning was shown with visual and olfactory modalities, pairing a conditioned neutral stimulus (CS) and an unconditioned vital stimulus (US), which leads to a link that predicts the latter and the conditioned response (CR) (Mizunami 2021). Operant conditioning was also showed with a neutral stimulus triggering an action that predicts a consequence (Matsumoto and Mizunami 2000).

Insect vision is also a whole topic of itself (see reviews: Borst 2009; Honkanen et al. 2017), where pattern learning and visual color processing (Giurfa 2004; Mizunami et al. 2009; Kelber 1996) (see also reviews: Briscoe and Chittka 2001; Dyer et al. 2011; Sancer and Wernet 2021; Schnaitmann et al. 2020) are subtopics that are investigated through different types of associative learning (see reviews: Avarguès-Weber et al. 2012b; Mizunami et al. 2013). Specifically with crickets, vision (Zufall et al. 1989) and discriminative tests with visual conditioning including black/white motifs (Unoki et al. 2006) as well as different color cues (Nakatani et al. 2009) were shown. The cricket vision allows colored tri-chromatic vision, asymmetrically distributed with UV, green and blue receptors in their compound eyes (Zufall et al. 1989; Frolov et al. 2014).

Therefore, this experiment aimed at developing an experimental protocol for crickets, in different incremental levels of visual associative learning tasks. Specifically, it targets the discrimination of compound visuo-spatial stimuli, a logical step just before investigating the S/D abstract concept with crickets. The chosen conditioned neutral stimuli were two pairs of colored small LEGO^® bricks disposed at the entrance of each arm of an Y-maze, one paired for the AA category and the other for AB (AA–AB motif). This learning procedure allows the crickets to discriminate the regularities of the visuo-spatial features in the compound cues. This two-alternative force choice task was followed, depending on the chosen rule, by a rewarding drop of water as the unconditioned vital stimuli. Multiple successful trials of the cricket selecting the “good” arm of the maze (operant conditioning), eventually leads to the association of the desired compound visual cues and rewards.

Materials and methods

Species

Gryllus pennsylvanicus, commonly referred as the fall field cricket, is a species widely spread across southern Ontario, Canada. It is mostly found in fields, forest edges, grassy disturbed areas, and human habitats. They are capable of burrowing in the soil. Gryllus pennsylvanicus is a nocturnal solitary opportunistic omnivore and feeds on vegetation as well as other invertebrates (Carmona et al. 1999; Burgess and Hinks 1987). All crickets used for this experiment originate from a colony at the Carleton University, Ontario, Canada with at least three generations in captivity.

Pre-experimental captivity context conditions

Before the Y-maze experiments, Gryllus pennsylvanicus individual adults were isolated for 48 h without access to water. Since 5 $\mu$L of water is used as a reward, this isolation period ensured that the reward was enticing enough for the individuals to complete the set trials. This time span was used as the standard isolation period to prevent complete desiccation and death, which occurred at a higher frequency when deprived for more than 48 h. Each cricket was isolated in a circular transparent plastic container with a mesh lid. The container had a diameter of 11 cm, a circumference of 37 cm and a depth of 8 cm. Each container included a 1.75 cm high dish of Harlan Teklad’s Rodent Diet (8604 meal food) as well as a piece of egg carton for shelter. The isolation period began at noon and ended at the same time two days later. The containers were placed in the same area of the room where the trials were conducted. Only uninjured individuals were used in the study and came from larger colony bins.

The room used for the isolation period and the subsequent tests was $3.65\times 3.65$ m. It had an average temperature of $24.58\pm 0.45_{}^{\circ }\text{C}$ and an average humidity level of $35.57\pm 6.33\%$, measured with a home weather station at the start of every set of trials. The lighting intensity in the room was set at a constant 82 lux during the tests, measured using the Lux Light Meter application on a Samsung S20 FE 5G. Each test was run midday between 1200 and 1800 h. The cricket colonies were raised in plastic bins (Rubbermaid totes, $45\times 55\times 40$ cm, L$\times$W$\times$H) with multiple Harlan Teklad’s Rodent Diet dishes, water vials, egg cartons and a sand dish for egg-laying. There were approximately 50 crickets per colony. These bins were stored in a room with 12–12 h light and dark cycle under fluorescent, UCF 8-watt bulbs. Before the experiment, none of the crickets used were in contact during their lifetime with colored LEGO^® bricks or any type of experiment.

Experiment setup and protocols

The Y-maze was used alongside with LEGO^® bricks, serving as visual colored cues. Four different colors of LEGO^® bricks were used in this study: blue (RGB: 0, 56, 101), light blue (RGB: 198, 218, 231), yellow (RGB: 255, 205, 00) and green (RGB: 0, 132, 61). The LEGO^® bricks were $1.5\times 0.8\times 0.9$ cm (L$\times$W$\times$H). Two LEGO^® bricks were placed at the entrance of a branch (depending of the experiment), approximately 1 cm deep. 5 $\mu$L of water was placed approximately 3 cm from the end of the rewarding branch. The other branch was empty aside from the presence of LEGO^® bricks at the entrance, depending on the type of test. The Y-maze is made in-house of white plastic with a thin clear plastic cover to prevent the insects from escaping the maze. The Y-maze had a length of 21 cm from the starting point to the end of one of the branches. Each arm has a length of 13 cm from the entry point to the end of the branch. The first part of the maze includes a waiting chamber measuring $5.4\times 4.9$ cm. This starting location was separated from the rest of the maze by a clear plastic divider, allowing the cricket to observe both arms as well as possible visual cues. The maze corridors have a width of 2.5 cm with the width of the end chambers being 5 cm. The maze has a height of 10 cm.

Before the start of an experiment, the cricket was carefully transferred to the starting chamber of the Y-maze by using a rubber straw. Then, the maze was placed in the testing area where it could be recorded from a top-down view using a Jolly Comb Model WGBG-006(H606) webcam and the free Open Broadcaster Software (OBS, https://obsproject.com). The Y-Maze was always positioned with the same orientation regarding the observer and the room (far left corner), the colony was situated at the right side of the maze. The observer always left the room at the beginning of the experiment. Then, the cricket went through a 2-min habituation period in this waiting chamber before the test could begin. The clear plastic divider was lifted and the timer (also being recorded by OBS) was started. Each individual had a maximum of 10 min per trial to select a branch to enter. The first branch that the cricket entered was noted (either left or right) as well as the time needed to enter it. A valid trial is defined when the entire body of the cricket reached the widest portion of the branch, right after the visual cue. Once the individual was completely into the selected branch, the trial was over. For all the tests (excluding Experiment A that contained no reward), the correct branch contained a water reward. If the cricket selected in the incorrect branch, the branch was empty (neutral stimulus). After the branch choice, the cricket was transferred to a new maze and the next trial began for a total of seven trials. If after 10 min, no branch was selected, the trial was deemed incomplete and the next trial began. If the cricket did not select a branch after the first two trials, it was omitted from the results. After each trial, the visual cues were removed (when applicable), then the maze was cleaned with 70% ethanol and left to dry. This prevented possible scent trails left by the individual. The branch containing the water reward and the visual cues alternated between every trial to test spatial memory bias. For the A–B compound condition, the chosen colors’ position remained stable between each trial (i.e Green–Blue but not Blue–Green), as shown in Fig. 1.

Fig. 1.

Y-maze setup: this picture represents the Y-Maze with a cricket inside the waiting chamber, with a clear sliding door separator. As example in this picture, the maze includes at the decision point or arm’s junction, two different pairs of LEGO^® bricks (Blue–Green at the top arm and Blue–Blue at the bottom arm). A tiny 5 $\mu$L drop of water is deposed at one end of an arm maze (not seen)

Before starting the main AA–AB discrimination experiment, several preliminary tests were performed (see Fig. 2), each with eight individuals and for eight trials. Experiment A was done without cues nor reward. This was to verify the presence of branch and room orientation biases. Experiment B included a water drop reward, placed at the end of one branch of the the maze and alternating branches on each trial. This was to test if the water could be seen or sensed by the insect. Experiment C is similar to the previous one; however, the water reward was hidden by a small white LEGO^® brick. It aims at verifying if crickets could detect humidity through other means. Experiment D tested for spatial memory bias as well as negative or positive brick tropism. Twenty individuals were tested with two LEGO^® bricks of the same color as a visual distractor. These were placed on each side at the entrance of one branch of the maze, alternating branches for the bricks while keeping the water drop in the same branch of the maze. All sets of trials start with the reward and the bricks on the same branch of the maze. Experiment E was performed to test for simple learning associations between one pair of compound visual cues and the reward. The compound visuo-spatial stimulus consists in one pair of two LEGO^® bricks of the same color. We used two different colored bricks, blue and green. The blue cues (eight individuals) or green cues (seven individuals) were placed at the entrance of one branch of the maze and alternating the branch of the maze simultaneously for both the bricks and the reward, after each trial. Experiment F tested for color discrimination, including two different pairs of bricks at the entrance of each branch, each pair of a separate color (Blue–Blue and Green–Green or AA–BB configuration). This experiment divided crickets in two groups, where one group (seven individuals) was tested on the green cues while the other group on the blue cues (eight individuals) leading to the rewards.

Fig. 2.

Schemas of the Y-Maze for all experiments: the W represents the waiting chamber, the blue dot at the end of arm’s maze is for the water reward and the colored rectangles represent the LEGO^® bricks. A Empty maze tested for orientation bias. B Maze with water droplet tested for visual bias. C Maze with water droplet hidden with a white LEGO^® brick for humidity sensing bias. D Maze with alternating pair of LEGO bricks of the same color with water droplet remaining on the same side. E Maze with alternating pair of Lego bricks of the same color with water droplet also alternating sides of the maze for simple associative learning test. F Maze with differential color associative learning test in a AA–BB compound stimuli configuration (Blue–Blue and Green–Green: G for Green). G Maze with AA–AB discrimination, composed with set of two different pairs of LEGO^® bricks (Blue–Blue and Green–Blue: (example shown in the figure), other configurations were Green–Green and Green–Blue, Yellow–Yellow and Yellow–Light Blue, Light Blue–Light Blue and Yellow–Light Blue)

Experiment G tested for the AA–AB discrimination consists in one pair of bricks with the same color and one pair of bricks of two different colors, with either pair leading to the reward depending on cricket group. This experiment included 74 crickets, divided in two groups for the AA–AB features: 44 individuals were tested with the reward associated with the compound visual cues of the AA configuration color bricks and 30 individuals were tested with the AB configuration color bricks. These were further sub-divided in four groups of the AA pattern color pairing. In this protocol, one branch of the Y-Maze contains the water reward and the alternate non-rewarding branch was empty as a neutral stimulus. In the AA group, 15 individuals were tested with the Blue–Blue combination, seven individuals were tested with the green–green combination, 16 were tested with the Yellow–Yellow combination and finally six individuals were tested for the Light Blue–Light Blue combination. In the AB group, the crickets were sub-divided in two groups. There were 10 individuals tested with green–blue and 20 individuals tested with Yellow–Light Blue.

Data analysis

The data was analyzed with the Microsoft Excel software. For all experiments, learning was measured as the number of successes per set of trials, transformed into a proportion of success removing trials where either (a) no choice was made and/or (b) until the first success in the series of 7 trials. All experiment results were compared to a random expectation of 50% using a one sample t-test. For Experiments A to C, a two-tailed one sample t-test was used. For Experiments D to G, a one-tailed one sample t-test was used (see Fig. 2). The trials before the first success (where the cricket entered the branch containing the water reward) were omitted since the cricket did not yet have the opportunity to learn the applied rule. Incomplete trials, where the cricket did not make a choice at all within 10 min, were also omitted from the statistics as they can not be classified as either a success or a failure. These data points were omitted from the total results presented (biasing positive results) but were kept in the data while calculating the average number of successful trials per experimental condition (more accurate).

Results

In the empty maze with no cues and no water reward (see Fig. 2 graph A), eight crickets completed eight trials each for a total of 64 trials. In 33 trials, the left branch was chosen and in 31 trials, the right branch was chosen. Results show the crickets did not prefer one of the branches more than the other than expected by chance (t(7) = $-$1.426, p = 0.197) suggesting no spatial bias. In the maze with an alternating water reward at the end of each branch and no visual cues(see Fig. 2 graph B), eight crickets completed eight trials each for a total of 64 trials. Crickets chose the branch containing the water reward and the empty branch equally, 31 trials each. Crickets did not select one of the branches more than expected by chance (t(7) = $-$0.424, p = 0.685) suggesting that crickets cannot see the water droplet. In the test including the water reward hidden with a white LEGO^® brick (see Fig. 2 graph C), eight crickets completed eight trials each for a total of 64 trials. Crickets chose the branch containing the water reward in 29 trials and the empty branch in 33 trials. They did not select one of the branches more than expected by chance (t(7) = 0.357, p = 0.732) suggesting that crickets cannot sense water through other physiological means.

In the experiment with the water reward always in the same branch of the maze throughout all trials but the LEGO^® bricks switching branches (see Fig. 2 graph D), 20 crickets completed seven trials each, making 118 valid trials. Crickets selected the branch containing the water reward during 91 trials and the empty branch during 27 trials. The crickets did select the branch containing the water reward more than expected by chance (mean = $0.73\pm 0.27$ SD, t(19) = 3.886 p < 0.001). The results were also significant for only the right branch (L: mean = $0.59\pm 0.26$ SD, t(9) = 1.093, p = 0.151 and R: mean = $0.88\pm 0.2$ SD, t(9) = 6.084, p < 0.001), as well as for the blue bricks (mean = $0.65\pm 0.3$ SD, t(9) = 4.676, p < 0.001). For the green bricks, the results were not significant (mean = $0.82\pm 0.21$ SD, t(9) = 1.57, p = 0.075).

In the simple associative learning task (see Fig. 2 graph E) with the LEGO^® bricks of the same color at the entrance of the branch containing the water reward, 15 crickets completed seven trials each, making 96 valid trials. Eight crickets were tested using blue LEGO^® bricks and seven were tested using green LEGO^® bricks. Overall, the crickets selected the branch containing the water reward during 53 trials and the branch containing no reward during 37 trials. Individually, cricket did select one branch more than expected by chance (mean = $0.59\pm 0.19$ SD, t(14) = 1.76, p = 0.036). Crickets tested using blue LEGO^® bricks did not select a branch more than expected by chance (mean = 0.57±0.26 SD, t(7) = 0.795, p = 0.226) but the crickets tested using green LEGO^® bricks did select the correct branch more than expected by chance (mean = $0.6\pm 0.08$ SD, t(6) = 3.221, p = 0.009).

In the AA–BB differential color associative learning task (see Fig. 2 graph F), where two different pairs of LEGO^® bricks (Blue–Blue and Green–Green) were placed at the entrance of each branch, 15 crickets completed seven trials each, making a total of 87 valid trials. Overall, the crickets selected the correct branch containing the water reward in 54 trials and the branch with no reward in 33 trials. Individually, cricket did select the branch containing the water reward more often than expected by chance (mean = $0.63\pm 0.19$ SD, t(13) = 2.451, p = 0.029) indicating that they were associating the correct color to the branch with the water reward.

The Yellow–Yellow and Light Blue–Light Blue group included 14 crickets, making a total of 91 valid trials for the experiment. Overall, crickets selected the correct branch containing the water reward in 50 trials and the branch with no reward in 41 trials. Individually, each cricket did not select the branch containing the water reward more often than expected by chance by using an alpha value threshold of 0.05 (mean = $0.62\pm 0.22$ SD, t(13) = 2.03 p = 0.063). This may be due in part to the low sample size in this preliminary experiment.

For the last experiment on the AA–AB discriminative compound color stimuli (see Fig. 2 graph G and Fig. 3), with a pair of AA pattern colored LEGO^® bricks placed at the entrance of one branch and another pair of two different colored Lego blocks at the other branch (AB pattern), we tested 74 crickets for a total of 482 trials. The crickets did select the correct branch more than expected by chance (mean = $0.64\pm 0.18$ SD, t(73) = 6.518, $p=4.03\times 10exp(-9)$).

Fig. 3.

Visual summary of the main AA–AB discriminative compound color stimuli task divided between the AA and AB group. The AA group is subdivided into four groups (Blue–Blue, Green–Green, Yellow–Yellow and Light Blue–Light Blue) and the AB group is subdivided into two groups (Blue–Green, Yellow–Light Blue) for each possible color combination. The p-value for the t-test for each subgroup is indicated. The total p-value at the bottom represents the p-value of the t-test for all the results of AA and AB respectively

Among these, (see Fig. 3) 44 individuals were tested for a total of 281 trials, with the reward associated to the AA configuration color bricks. The crickets did select the corrected branch more than expected by chance (mean = $0.65\pm 0.18$ SD, t(43) = 5.338 $p=1.66\times 10exp(-6)$). This group was further sub-divided in four different color groups. 15 individuals were tested for a total of 94 trials with the Blue-Blue combination. The crickets did select the corrected branch more than expected by chance (mean = $0.62\pm 0.15$ SD, t(14) = 3.313 p = 0.003). 7 individuals were tested with the green–green combination for a total of 45 trials. The crickets did select the corrected branch more than expected by chance (mean = $0.68\pm 0.1$ SD, t(6) = 4.575 p = 0.002). 16 individuals were tested for a total of 102 trials with the Yellow–Yellow combination. The crickets did select the corrected branch more than expected by chance (mean = $0.68\pm 0.24$ SD, t(15) = 3.056, p = 0.004). Finally, 6 individuals were tested for the Light Blue–Light Blue combination for a total of 40 trials. The crickets did not select the corrected branch more than expected by chance (mean = $0.57\pm 0.18$ SD, t(5) = 1.043 p = 0.173).

In the AB group associated with the reward, 30 individuals were tested for a total of 201 trials. The crickets did select the corrected branch more than expected by chance (mean = $0.62\pm 0.17$ SD, t(29) = 3.718, p = 0.00043). This AB group was also sub-divided in two groups. 10 individuals were tested with Green–Blue for a total of 69 trials. The crickets did select the corrected branch more than expected by chance (mean = $0.67\pm 0.06$ SD, t(9) = 9.585 $p=3.0\times 10exp(-6)$). Finally, 20 individuals were tested with Yellow-Light Blue for a total of 125 trials. The crickets did select the corrected branch more than expected by chance (mean = $0.59\pm 0.21$ SD, t(18) = 1.81 p = 0.044).

As different overviews, from experiment D to G (see Fig. 4), the average absolute number of successful trials out of seven for each experiment for all crickets, demonstrates results above 3.5 (half of 7). Also, from experiment A to G (see Fig. 5), the success rate for total trials per experimentation show again results above 50%.

Fig. 4.

This graph represents the average number of successful trials out of seven for each experiment (D to G: excluding trials where the subject did not enter a branch of the maze). A successful trial is when the cricket selects the “correct” branch in the experiment. Experiment D represents the Y-Maze with two LEGO^® bricks of the same color that alternate branch every trial, with the water reward remaining in the same branch. Experiment E is the simple associative learning task with two LEGO^® bricks and the water reward alternating branches every trial. Experiment F is the AA–BB color discrimination task. The experiment G is the AA–AB discriminative compound color stimuli task. The error bars represent standard deviation

Fig. 5.

Graph representing the success rate for total trials per experimentation. The ratios for graph A to C are the absolute numbers and the ratios for graph D to G represent the average ratio (The error bars are the SD). A Empty, B Water drop, C Hidden water drop, D Water drop fixed position but alternate Lego bricks, E Simple associative learning task, F AA–BB color discrimination test, G AA–AB compound discriminative test

Discussion

This paper investigated the visuo-spatial AA–AB compound stimulus discriminative abilities of crickets in an two-forced-choice operant conditioning learning task. The objective of this study was to develop a protocol allowing the observation of crickets when performing association with different pair sets of compound visual colored stimuli. The proposed Y-maze setup was used for dichotomous discrimination cues followed by a water reward as the positive reinforcer. Thus, it allows a temporal associative sequence between the conditioned neutral stimulus, the action and the unconditioned vital stimulus. This capitalized on the intrinsic tendency of these insects to follow walls with fast movements. In all the different experiments of this study, no individual stopped in the middle of an arm when engaged in the arm of the maze. Furthermore, the travel time was almost fixed (1 s) due to the short length of the maze arms. In future works, varying the arm’s length parameter may facilitate the exploration of event timing in the context of learning and memory research.

All experiments prior to the main AA–AB discrimination learning task were done to exclude different possible confounding variables. On every test, alternating the left and right side for the reward seemed sufficient to generate randomness. To completely rule out that the insect did not simply learn the alternation pattern to guide them through the reward, one of these experiments included a constant rewarding side but alternating the visual cues. The results for this experiment (D) suggest that Gryllus pennsylvanicus used a form of spatial memory strategy, returning to the arm of the maze that had the water droplet in the previous trial. It suggests that the the visual cues seem neutral. Also, experiments were done to exclude the potential bias of sensing humidity or seeing the minuscule drop of water. In all these cases, the number of trials were limited to eight. Prior to these, many variations of the number of trials per experiment (4–16 trials) were done to target a “learning threshold”, though data are not presented in the article. The choice values of 8 trials appear a good tradeoff between the expected results and confounding variables such as possible fatigue, lack of curiosity or water nourishment.

In the simple associative test with colored cues (AA–XX) associated with the reward, as well as the color discrimination test (AA–BB), the crickets were successful in both cases (statistically close to significant for Yellow–Yellow and Light Blue–Light Blue), confirming the literature on visual conditioning with crickets (Unoki et al. 2006; Nakatani et al. 2009). However, this present study goes beyond these simple association learning tasks by using more complex compound visual stimuli, composed with an AA–AB configuration.

The obtained results for these tests point out that the crickets may either consider compound visual stimuli with their independent features as a whole, not simply from the addition of features, or that they could use configurational cues or features of the stimuli to either avoid or approach it. In future studies, these visual complex motifs could eventually serve as a basis for other types of conditioning procedures and analogical reasoning learning tasks (Smirnova et al. 2015), as with relational matching-to-sample tasks (Fagot and Thompson 2011), reversal learning tasks (Finke et al. 2023), negative patterning and negative feature discrimination tasks (Durrieu et al. 2020) and above all with transfer tests to test the learning of the concepts of same versus different.

In the main AA–AB discrimination experiment, for each individual, one training set of seven trials was done with two pairs of the compound colored bricks. Several colors were tested with different number of individuals per experiment. Perhaps doing several training sets may improve results. However, in that case, it would require to study the inter-trial period effect on training sets, testing the memory retention of the association as well as the reward habituation (fatigue, curiosity) factors. Since every trial was done during the same day, no data collection was made on these parameters and thus, the inter-trial period effect remains to be explored. In this perspective, it seems that the set size may affect the results (Katz et al. 2002; Lazarowski et al. 2019) and this quantity may vary across species.

The presented results for the AA–AB discrimination experiment were conclusive. This indicates that Gryllus pennsylvanicus could learn to associate one compound visual pattern of a two-item set, an action and a reward. This study did not take into account the individual consistency (Finke et al. 2023) and color preference across all conditioning procedures. Also, other conditioning procedures (i.e. differential rewards, delayed match-to-sample protocol) may improve the learning curve between the visual cue and the reward (Avarguès-Weber et al. 2014). In this study, the “wrong” arm was neutral and did not trigger a consequence upon visiting it. It could have been done differently, like in the case of differential aversive conditioning procedure, using kinins or air puffs, or differential rewarding outcomes (Schmidtke et al. 2010) to study these effects on the data. It would be interesting to observe if the difference is significant between these types of reinforcement paradigms (Awata et al. 2015; Nakatani et al. 2009). However, this Y-maze setup may be suitable with crickets to investigate other visual properties such as numerosity (see review: Bengochea and Hassan 2023) or with other perceptual modalities like olfaction. Another enhancement to the conditioning procedure may consist in using vision but with different spatial arrangements of the LEGO^® bricks of the same color, for 3D spatial perception and transfer tests across another visual domain.

This may allow future experiments in the perspective of studying abstract concept learning with this animal for comparative data with animals with small brains (see review: Wright and Kelly 2017) with their specificities and differences (qualitatively and quantitatively), or even with bio-inspired simulation models (Cyr et al. 2017; Cope et al. 2018; Wessnitzer et al. 2012). The main assumption is that higher cognition is possible with minimal neural circuits.

Conclusion

This study investigated the discriminative abilities of Gryllus pennsylvanicus with AA–AB visuo-spatial colored stimuli. An operant conditioning learning protocol was developed to reach this objective, using a Y-maze with a dichotomous arm choice. Results were significant in these associative experiments, including pairs of compound visual cues and an appetitive reinforcer, a prelude to explore abstract concept learning with these insects.

Author Contributions

Conceptualization: AC; Methodology: AC, JMF; Experimentation and statistical analysis: IM; Writing-Review: AC, IM; Funding acquisition and resources: JMF; Supervision: AC, JMF

Funding

This work was supported by the Natural Science and Engineering Research Concil of Canada through Discovery Grants (2019-06558, NSERC).

Data availability

The complete access to all parameters and result data used to support this study are available from the corresponding author upon request. Videos of the experiment can be viewed at the following link: https://www.dropbox.com/sh/v0qik3kd64jnn5u/AAAkejbXztztBdMm8M4_tL6-a?dl=0

Declarations

Conflict of interest

The authors have no relevant financial or non-financial interests to disclose.

Ethical approval

This study involved crickets dedicated to research for which an approval of an ethical committee is not mandatory. The protocols comply with standard welfare practice and a minimum number of individuals were used to study our scientific question. The animals were not harmed during the experimental procedures.