Y-mazes enable researchers to determine the relevance of specific stimuli that drive animal behavior, especially isolated chemical cues from a variety of sources. Careful design and planning can yield robust data (e.g., discrimination, degree of exploration, numerous behaviors). This experimental apparatus can provide powerful insight into behavioral and ecological questions.
Reptiles utilize a variety of environmental cues to inform and drive animal behavior such as chemical scent trails produced by food or conspecifics. Decrypting the scent-trailing behavior of vertebrates, particularly invasive species, enables the discovery of cues that induce exploratory behavior and can aid in the development of valuable basic and applied biological tools. However, pinpointing behaviors dominantly driven by chemical cues versus other competing environmental cues can be challenging. Y-mazes are common tools used in animal behavior research that allow quantification of vertebrate chemosensory behavior across a range of taxa. By reducing external stimuli, Y-mazes remove confounding factors and present focal animals with a binary choice. In our Y-maze studies, a scenting animal is restricted to one arm of the maze to leave a scent trail and is removed once scent-laying parameters have been met. Then, depending on the trial type, either the focal animal is allowed into the maze, or a competing scent trail is created. The result is a record of the focal animal's choice and behavior while discriminating between the chemical cues presented. Here, two Y-maze apparatuses tailored to different invasive reptile species: Argentine black and white tegu lizards (Salvator merianae) and Burmese pythons (Python bivittatus) are described, outlining the operation and cleaning of these Y-mazes. Further, the variety of data produced, experimental drawbacks and solutions, and suggested data analysis frameworks have been summarized.
Y-mazes are common, simple tools in studies of animal behavior that allow for a variety of questions to be addressed. In addition to being widely used in laboratory studies, Y-mazes are also functionally compatible with various field environments to study wild animals in relatively remote settings. Researchers have examined the behaviors of wild vertebrates using Y-mazes in a wide variety of taxa across similarly diverse field applications (e.g., lampreys1; cichlid fish2; poison frogs3; lacertid lizards4; garter snakes5).
Many researchers are focused on how and to what degree chemical cues drive animal behaviors in reproductive, spatial, and foraging ecology6. A variety of chemical stimuli can be tested in Y-mazes and at fine scales, such as two chemical trails that only differ slightly in concentration7, or detection ability based on the reproductive status of the target species8. Chemical trails—the principal stimulus used in Y-maze tests—can be naturally created by conspecifics or specifically placed in the environment by a researcher using a defined chemical source1,5. Stimuli can also be tested in unique combinations to determine multimodal influence of cues such as changing contexts of cue presentation (airborne vs. substrate trails9; visual plus chemical cues10). Although there are many other methods for assessing chemosensory responses in reptiles (see discussion section), Y-mazes allow for searching behavior(s) to be assessed and at multiple temporal and spatial scales, which can lead to greater levels of behavioral inference.
Reptiles have been broadly tested for their reliance on chemical cues in reproductive and foraging ecology, and researchers often employ Y-mazes in these studies11,12. The chemical ecology of reptiles continues to be deciphered by studies employing Y-mazes to address a variety of evolutionary and behavioral questions that are valuable to wildlife managers. For example, recent tests with invasive snake and lizard species have revealed that chemical cues alone can influence choice and time allocation within the novel environment of a Y-maze13,14,15.
The use of large Y-mazes for moderately sized focal animals (e.g., large-bodied reptiles) is generally restricted to laboratory settings where the focal animals can be housed easily over the long term, experimental factors (e.g., climate, light, external stimuli) can be controlled, and access to infrastructure (e.g., power, running water) is unlimited. Studies on wild animals, however, are often restricted to specific locations for various reasons (e.g., logistics, permitting). As a result, challenges arise that must be addressed through creative problem solving and methodological adjustments to maintain consistent and comparable results.
Here, two experimental setups have been described using Y-mazes and remote monitoring tools to assess reproductive chemical ecology of invasive squamate reptiles (i.e., snakes and lizards) in different field scenarios: wild-caught, captive Argentine black and white tegu lizards (Salvator merianae) in Gainesville, FL, and wild-caught Burmese pythons (Python bivittatus) in Everglades National Park, FL. As implied by its name, the Y-maze apparatus creates an experimental environment in which an animal enters a main passageway (the base of the Y; "base") which then leads to two divergent passageways (the arms of the Y; "arms"). In these experiments, two types of animals are used for a single trial: scent-laying animals (provide the stimulus scent in a restricted area of the maze) and focal animals (data are collected on this animal as it explores the scent trail).
As an experimental apparatus in chemoecological studies, any Y-maze must be constructed in a way that allows easy removal of the animal within and can be dissembled for thorough cleaning and reset. Also discussed are the constraints inherent to these different testing environments (e.g., diurnal vs. nocturnal animals, infrastructure differences) that prompted methodological adjustments. Although the focus was on tegu lizards and Burmese pythons, these designs can be applied to a wide range of reptile species. In this research on invasive reptiles, Y-mazes benefit the rate and scale of inference because they enable rapid collection of data to inform management goals that shift in-step with the invasion threat posed by a given species. In particular, studying chemoecology of invasive species is critical for the development of effective chemical control tools.
Discrimination is the key observation from empirical tests using Y-mazes where a focal animal chooses between two stimuli and that decision-making process is assessed. A swath of behaviors can also be scored in Y-maze trials during the trial itself (live) or after the trial (video) to expand inferential power. The complexity of the a priori objectives of a given study dictate whether live observation or archived recordings best suit the design. Here, Y-maze methods have been described in detail for addressing chemoecological questions to inform future studies by researchers interested in similar questions on reptile behavior, especially in chemical ecology.
All procedures involving the use of live vertebrates were approved by the Institutional Animal Care and Use Committees of the U.S. Department of Agriculture and the U.S. Geological Survey.
NOTE: Because these studies focus on invasive vertebrates, compliance with containment standards must also be met, which impose specific constraints on the design and execution of experiments. Although many of the methods are similar between the two study locations and diurnal vs. nocturnal study timing, distinct methods have been described in each of the following two sections.
1. Y-maze setup and diurnal protocol for the U.S. Department of Agriculture (USDA) Animal Plant Health Inspection Service (APHIS) Wildlife Services National Wildlife Research Center Florida Field Station: on-site testing of wild-caught, captive tegus
NOTE: Plans for all components of the Y-maze and containment structure are provided in Supplemental File 1.
2. Y-maze setup and crepuscular timing protocol for the U.S. Geological Survey (USGS) trials in collaboration with National Park Service: relatively remote testing of wild-caught Burmese pythons
NOTE:Plans for all components of the Y-maze and containment structure are provided in Supplemental File 2.
Figure 1. Layout of the USGS Y-maze. On the left, a schematic shows the components of the Y-maze with a scale bar for perspective. On the right, a snapshot from the video camera demonstrates the field of view for behavioral recordings. Please click here to view a larger version of this figure.
A multitude of variables can be recorded and/or scored from Y-maze trials. The design of the study should be primarily driven by the desired outcomes/deliverables. Further, if the study is relying on repeated measures (e.g., repeated use of the same focal animals), proper testing and analysis structures are required. For example, as the USDA trials relied on repeated testing of focal tegus, the planning of experimental trials was fully randomized.
Choice data: The majority of studies using Y-mazes report simple binary choice data and analyze the results with parametric statistics such as a binomial test. The chief limitation here is sample size, which directly affects the power of any statistical analysis. In Figure 2, a series of statistical thresholds per study sample size are depicted that demonstrate how many "successes" would need to occur for a given binomial test to yield statistically significant results. These are mathematically derived and therefore generalizable to any Y-maze test. Binomial statistics are easy to generate using online freeware. For calculating probabilities, one-tail distributions are used if an a priori rationale is given; otherwise, the two-tail distribution should be used.
Choice of an arm is often determined by the distance the focal animal moves in a given arm. The simplest way to set this threshold is by establishing a landmark within the maze. For most Y-maze studies, the landmark is the entrance of the arm box. Because reptiles conduct all chemosensory assessment with the chemical-sensing organs in the anterior region of the head, the head is the focal point during a trial. For example, because Burmese pythons are often longer than the entire maze itself, choice is best and most efficiently determined by the movement of the head past a landmark. Other options for determining choice are time spent in an arm and complete movement of the focal animal into a box. Failure is determined by a focal animal not making a choice within a specific period.
More fine resolution analyses can be derived from choice data in the Y-maze. For example, researchers can generate a choice penalty score16. Here, researchers must track the degree to which the focal animal explored the non-target arm of the maze. Non-target can be defined as the arm the researchers determine a priori that the focal animal will not choose based on the alternative hypothesis tested. The simplest example of a non-target arm would be the unscented arm when only one arm contains a target scent. More complex examples would be the choice between two scents from the same source, but presented at different concentrations7. When the experimental design is multi-level and/or the data go from binary to incremental, as with choice penalty, an appropriate statistical approach should be used such as repeated measures analysis of variance (ANOVA) or other methods used with continuous or proportional datasets.
Behaviors: Throughout the duration of an experiment in which focal animals are observed, a variety of individual behaviors can be quantified. This number of variables can either be determined a priori depending on what is known16 or post hoc following preliminary observations on a subset of data14,15. The study objectives and their degree of resolution determine what behavioral assessments should be made within the maze, if any (i.e., in many studies, only choice data are quantified17). Behaviors can be assessed throughout the maze, in sections, or during specific time periods; for instance, behaviors seen only in the base or at the junction of the arms may be prioritized8. Video recordings facilitate behavioral scoring, although the resolution of the video and its length—factors that impose data storage constraints—should be considered before experimentation begins.
Temporal variables: As with behavioral variables, many temporal aspects of animal performance can be quantified during Y-maze trials. For example, researchers can time latency periods (e.g., latency to emerge from the box8). Most temporal variables are associated with exploration of the maze such as total trailing time or time spent in each arm. These variables are usually analyzed in a multi-factor analysis such as multi-way ANOVA.
Observer bias: With any studies involving animal behavior, observer bias significantly influences data collection18. Therefore, observers should be blind to the treatment being tested. The simplest way to do this is to code the video files numerically and then randomly sort them (e.g., random number generator) prior to assigning them to observers. Controlling for observer bias is difficult-to-impossible when live data collection is the only option. In a field setting, this would require two cooperators: an observer blind to the treatment and a coordinator who sets up the trial. Extensive reviews summarize the effects of experimenter bias on data collection and interpretation in behavioral and ecological studies18,19.
Figure 2. Sample sizes and P-values for binomial tests from Y-maze results. Each given sample size represents a set number of trials where a scent is tested in one arm of the Y (target arm) while the other could be a control (non-target). Top number above each bar is the one-tail P-value for that number of target arm choices, bottom is two-tail. Numbers within the top bar represent the maximum number of non-target choices that are still traditionally statistically significant (P < 0.05). Please click here to view a larger version of this figure.
Supplemental File 1. Please click here to download this file.
Supplemental File 2. Please click here to download this file.
While Y-mazes are very powerful tools to investigate chemical ecology in reptiles, their limited design can preclude other avenues of inquiry. However, a diversity of other options is available11,12,20,21,22. For example, tongue-flick assays are simpler to execute and allow simultaneous assessment of behaviors exhibited to an array of chemical stimuli relative to control odors23,24,25,26. Open-field tests are another option where a focal animal freely explores an enclosure until it encounters a source of chemical cues, and its behavioral reactions are subsequently scored27,28. Combinations of these approaches can assess discriminatory capacities of reptiles in varying contexts such as presenting a mix of artificial and natural odors along with refugia29. Y-mazes can also be modified to expose animals to airborne chemical cues alone or in combination with substrate-borne cues16,30, and post hoc inference can be used to redesign data collection if archived video data are available31. Bioassays should be designed to simplify data collection and minimize conflicting stimuli, especially when a specific source of cues is being assessed (e.g., chemical cues21).
Researchers in animal behavior often observe and quantify focal animal responses in novel, artificial environments (e.g., an enclosed maze with a featureless landscape), and care should be taken to assess whether a given animal is exhibiting natural, exploratory behavior versus avoidance, agitation, or similar distressed behavior. Distressed animal behavior in experimental apparatuses is primarily attributed to neophobia: fear of novelty32. An example is escape behavior, where the focal animal pushes against the joints or the edges of the apparatus to achieve egress. Another example is shyness, where the focal animal demonstrates reluctance to enter the maze, the degree of which can be quantified by latency of maze entry. Apparatus (re)design can facilitate engagement of the focal animal to avoid these confounding effects of distress. The most common approach is repeated introduction of the focal animal to the apparatus to remove the novelty of the environment before testing begins, and contemporary statistical models (e.g., generalized linear mixed models) allow for test animals to be used in multiple trials. An important aside relevant to ecological considerations in behavioral testing is that reduced neophobia is associated with the success of invasive species33. Thus, depending on a priori knowledge of the species in question, neophobia may have variable importance as an experimental design consideration.
Acquisition of behavioral data from videos imposes multiple constraints that become major bottlenecks in experimental timelines. For example, the length of a given trial can exponentially increase data extraction time. One workaround is to analyze behavior only until a threshold is met (e.g., total time active). The threshold can be based on the longest video available for a given trial. Alternatively, machine-based observation (e.g., artificial intelligence) can be developed, although this is time- and resource-consuming with considerable effort required for quality control. Another issue is data management: videos must be of sufficient quality to enable behavioral scoring and assessment, resulting in data storage constraints. While cloud storage is now accessible, upload/download rates are often problematic, especially when data acquisition occurs in remote field locations. Additional challenges manifest in the limitations of recording tools that affect the integrity of behavioral observation. Clear viewing of focal animal behavior is always necessary, but visibility is often impeded by uncontrollable factors (e.g., moisture, insects, wind movement). Further, when recordings come from a single perspective (e.g., bird's eye view), behaviors occurring in the vertical plane (e.g., head raises14) are difficult to assess. A solution is to provide multiple camera angles per trial. Lastly, the time of day significantly affects behavioral recording. Nighttime behavioral analysis requires a camera with a nighttime mode and minimal light projection to avoid obstructive glare on the Y-maze surface or attraction of insects that can interrupt the camera feed. Considering the above, foreknowledge of the study site or species biology can inform which constraints are likely to occur with what frequency and thus inform desirable sample sizes.
Behavior is tightly coupled with physiology, and the utility of Y-mazes for evaluation of behavioral endocrinology in a variety of species has been demonstrated. However, this paper emphasizes some variation in the execution of these experiments depending on the target species, research question, and resources available. Therefore, the selection of materials and dimensions of each testing setup should be carefully considered for potential subsequent research expansion. Section 2 describes modifications made to materials outlined in section 1, which were incorporated to accommodate future, more complex behavioral trials with tegus. The increased vertical depth of the Everglades mazes will allow new questions about chemical ecology in wild-caught tegus to be answered without unduly protracting project design and setup, further demonstrating the translatability of this experimental apparatus.
When employing the above-described techniques in a relatively remote setting (see section 2), there are several limiting factors that must be considered, and project planning is paramount. Depending on the statistical power needed for the prescribed treatment experiment and biological timing of the target species (e.g., seasonality), the resources and labor required will be affected. Further, if single or repeated use of focal animals are desired, careful attention to reducing potential stressors is necessary. Each of these factors will either extend the project timeline or require increased labor, space, and materials. For example, section 2 presents the use of wild-caught male pythons as focal animals trailing another group of wild-caught and hormonally manipulated males, all of which require approximately 24 h of quiet acclimation time in holding boxes to minimize stress effects. Although these acclimation periods extended trial times to over two days, stress due to captivity and handling affect wild animal behavior and must be minimized to generate clean datasets34,35.
In summary, Y-mazes are powerful, adaptable tools that can be used to investigate the chemical ecology of diverse wildlife under widely variable conditions, provided there is vigilant a priori planning. Careful consideration must be taken to choose appropriate questions and to properly design the experimental setup for given taxa and conditions. Researchers and managers can significantly benefit from using Y-mazes to better understand animal chemosensory biology as these tools enable flexible experimental designs that provide large volumes of fine-scale behavioral data, especially when combined with remote monitoring tools.
The authors have nothing to disclose.
The development of the first Y-maze was supported by cooperative agreements (15-7412-1155-CA, 16-7412-1269-CA, and 17-7412-1318-CA) between James Madison University (JMU) and the USDA Animal and Plant Health Inspection Service. The development of the Y-maze in Everglades National Park was funded by a cooperative agreement (P18AC00760) between JMU and the National Park Service. We thank T. Dean and B. Falk for their facilitation of this project in Everglades NP and assistance with permitting and funding. We thank W. Kellow for assistance in construction of the USGS Y-maze. C. Romagosa, L. Bonewell, and R. Reed provided administrative and logistical support. We thank the two anonymous reviewers who offered helpful feedback. Funding for the Everglades work and in-kind support was provided by U.S. Geological Survey (USGS) Greater Everglades Priority Ecosystem Science Program, National Park Service (P18PG00352), and USGS Invasive Species Program. Any use of trade, firm, or product names is for descriptive purposes only and does not imply endorsement by the U.S. Government. The findings and conclusions in this publication have not been formally disseminated by the U.S. Department of Agriculture and should not be construed to represent USDA determination or policy.
1" Steel zinc-plated corner brace | Everbilt, The Home Depot | 13619 | See Supplemental File 1, Step 2.1 "90 degree 2.5 cm steel corner brace" |
121.92cm W x 304.8cm L x 1.27cm H white polypropylene Extended Range High-Heat UHMW Sheet | TIVAR | UHMNV SH | See "2.1. Y-maze components and rationale for changes to USDA design " (step 2.1.1. "white polpropylene") |
182.88 cm L x 81.28 cm W x 0.64 cm Thick Clear Acrylic Sheet | Plexiglass | 32032550912090 | See "2.1. Y-maze components and rationale for changes to USDA design " (step 2.1.1.6. "Acrylic pieces") |
2.54 cm W x 2.54 cm H x 243.84 cm L Mill-Finished Aluminum Solid Angle | Steelworks | 11354 | See "2.1. Y-maze components and rationale for changes to USDA design " (step 2.1.1.1. "aluminum angle bracket") |
4.5 kg spool of 5 mm Round Polypropylene Welding Rods | HotAirTools | AS-PP5N10 | See "2.1. Y-maze components and rationale for changes to USDA design " (step 2.1.1. "heat weld") |
5 mm Plain Aluminum Rivets | Arrow | RLA3/16IP | See "2.1. Y-maze components and rationale for changes to USDA design " (step 2.1.1.1. "rivet") |
Aluminum angle, 1.9 cm | Everbilt, The Home Depot | 802527 | See Supplemental File 1, Step 1.2 "aluminum angle (1.9 cm x 1.9 cm x 0.16 cm thick)" |
Aluminum angle, 2.5 cm | Everbilt, The Home Depot | 800057 | See Supplemental File 1, Steps 1.2 and 2.2.2 "aluminum angle (2.5 cm x 2.5 cm x 0.16 cm thick)" |
Aluminum angle, 3.2 cm | Everbilt, The Home Depot | 800037 | See Supplemental File 1, Step 1.2 "aluminum angle (3.2 cm x 3.2 cm x 0.16 cm thick)" |
Aluminum flat bar 1" x 1/8" thick | Everbilt, The Home Depot | 801927 | See Supplemental File 1, Step 3.2.1 "aluminum strap" |
Avigilon 2.0 MP camera | Avigilon, a Motorola Solutions Company | 2.0C-H4SL-BO1-IR | See "1.5 Camera set-up and video acquisition" (step 1.5.1 "Avigilon 2.0 MP") |
Avigilon NVR | Avigilon, a Motorola Solutions Company | HD-NVR3-VAL-6TB-NA | See "1.5 Camera set-up and video acquisition" (step 1.5.3 "NVR") |
Clear acrylic sheet (5.6 mm thick) | United States Plastic Corp. | 44363 | See Supplemental File 1, Step 1.3 "clear acrylic sheet" and step 3.2.1 "clear acrylic door" |
Fillet Weld Nozzle 3/16" x 15/32" / 4.5 x 12 mm | TRIAC | 107.139 | See "2.1. Y-maze components and rationale for changes to USDA design " (step 2.1.1. "heat weld") |
Hanging File Folder Box | Sterilite | 18689004 | See "2.1. Y-maze components and rationale for changes to USDA design " (step 2.1.2.1. "Boxes") |
HardiePanel HZ10 | James Hardie Building Products | 9000525 | See Supplemental File 1, Step 1.1 "fiber cement siding" |
Heat Welding Gun | TRIAC | 141.227 | See "2.1. Y-maze components and rationale for changes to USDA design " (step 2.1.1. "heat weld") |
Kraft Butcher Paper Roll, 24" | Bryco Goods | 24 inch x 175 FT | See "1.2 Protocol for running scent-laying tegus" (step 1.2.1.2 "butcher paper") |
Kraft Butcher Paper Roll, 46 cm wide | Bryco Goods | BGKW2100 | See "2.3. Protocol for running scent-laying pythons" (step 2.3.4. "scenting paper") |
Micro-90 Concentrated Cleaning Solution | International Products Corporation | M-9050-12 | See "1.4 Breakdown and clean-up" (step 1.4.4 "laboratory-grade soap") |
MKV ToolNix – Matroska tools for linux/Unix and Windows | Moritz Bunkus | v.48.0.0 | See "2.2. Camera setup and video acquisition" (step 2.2.4.2. "movie processing software") |
Network Camera | Axis Communications | M3104-LVE | See "2.2. Camera setup and video acquisition" (step 2.2.1. "Project camera") |
Palight ProjectPVC 1/4" | Palram | 159841 | See "2.1. Y-maze components and rationale for changes to USDA design " (step 2.1.2.3. "faceplate") |
Palight ProjectPVC 1/8" | Palram | 156249 | See "2.1. Y-maze components and rationale for changes to USDA design " (step 2.1.2.1. "door") |
Privacy windscreen (green) | MacGregor | Size to fit | See Supplemental File 1, Step 4.2 "green heavy duty shade cloth" |
Protective Glove, Full-Finger | ArmOR Hand | HS1010-RGXL | See "2.3. Protocol for running scent-laying pythons" (step 2.3.11.2. NOTE: "puncture-resistant glove") |
REScue Disinfectant | Virox Animal Health | 44176 | See "1.5. Breakdown and clean-up." (step 1.5.4. NOTE "sanitation solution") |
Reversable PVC trim, 1/2" x 24" | UFP Industries, Veranda products | H120XWS17 | See Supplemental File 1, Step 2.1 "PVC board partition", and step 3.2.1 "thinner PVC trim boards" |
S4S / Veranda HP TRIM | UFP Industries, Veranda products | H190OWS4 | See Supplemental File 1, Steps 1.2, 2.2.2, and 2.2.3 "PVC board" |
S4S / Veranda HP TRIM (1" x 8" Nominal) | UFP Industries, Veranda products | 827000005 | See Supplemental File 1, Steps 3.2.1 "PVC trim board" |
ScotchBlue 24 in. Pre-taped Painter’s Plastic | 3M | PTD2093EL-24-S | See "1.2 Protocol for running scent-laying tegus" (step 1.2.1.3 "plastic sheeting") |
Sterilite 114 L tote box | Sterilite Company | 1919, Steel | See Supplemental File 1, Step 3.2 "arm box" |
Sterilite 189 L tote box | Sterilite Company | 1849, Titanium | See Supplemental File 1, Step 3.2 "Base box" |
Super Max Canopy | ShelterLogic | 25773 | See Supplemental File 1, Step 4.3 "white canopy" |
VLC Media Player | VideoLAN | v.3.0.11 | See "2.2. Camera setup and video acquisition" (step 2.2.4.3. "media file reviewing program") |
White Pavilion Tent | King Canopy | BJ2PC | See Supplimental File 2 "3. Enclosure materials and consideratons" (step 3. "pavilion tent") |