This article provides a detailed description of a novel mouse judgment bias protocol. Evidence of this olfactory digging task’s sensitivity to affective state is also demonstrated and its utility across diverse research fields is discussed.
Judgment biases (JB) are differences in the way that individuals in positive and negative affective/emotional states interpret ambiguous information. This phenomenon has long been observed in humans, with individuals in positive states responding to ambiguity ‘optimistically’ and those in negative states instead showing ‘pessimism’. Researchers aiming to assess animal affect have taken advantage of these differential responses, developing tasks to assess judgment bias as an indicator of affective state. These tasks are becoming increasingly popular across diverse species and fields of research. However, for laboratory mice, the most widely used vertebrates in research and a species heavily relied upon to model affective disorders, only one JB task has been successfully validated as sensitive to changes in affective state. Here, we provide a detailed description of this novel murine JB task, and evidence of its sensitivity to mouse affect. Though refinements are still necessary, assessment of mouse JB opens the door for answering both practical questions regarding mouse welfare, and fundamental questions about the impact of affective state in translational research.
Measuring affectively modulated judgment bias (henceforth JB) has proven to be a useful tool for studying the emotional states of animals. This innovative approach borrows from human psychology since humans experiencing positive or negative affective states (emotions and longer-term moods) reliably demonstrate differences in the way they process information1,2,3. For example, humans experiencing anxiety or depression might interpret neutral facial expressions as negative, or neutral sentences as threatening4,5. It is likely that these biases have an adaptive value and are therefore conserved across species6,7. Researchers aiming to assess animal affect have cleverly taken advantage of this phenomenon, operationalizing optimism as the increased expectation of reward in response to neutral or ambiguous cues, and pessimism as the increased expectation of punishment or reward absence8,9. Thus, in an experimental setting, optimistic and pessimistic responses to ambiguous stimuli can be interpreted as indicators of positive and negative affect, respectively10,11.
Compared to other indicators of animal affect, JB tasks have the potential to be particularly valuable tools since they are capable of detecting both the valence and intensity of affective states10,11. The ability of JB tasks to detect positive states (e.g., Rygula et al.12) is especially useful since most indicators of animal affect are limited to the detection of negative states13. During JB tasks, animals are typically trained to respond to a positive discriminative cue predicting reward (e.g., high-frequency tone) and a negative discriminative cue predicting punishment (e.g., low-frequency tone), before being presented with an ambiguous cue (e.g., intermediate tone)8. If in response to ambiguous cues an animal 'optimistically' performs the trained response for the positive cue (as if expecting reward), this indicates a positive judgment bias. Alternatively, if animals demonstrate the negative trained response to avoid punishment, this is indicative of 'pessimism' or negative judgment bias.
Since the development of the first successful JB task for animals by Harding and colleagues8, several JB tasks have been developed for a wide range of species across diverse research fields7. But despite their increasing popularity, animal JB tasks are often labor-intensive. Moreover, perhaps because they are methodologically different from the human tasks that inspired them, they sometimes produce null or counterintuitive results14 and commonly yield only small treatment effect sizes15. As a result, JB tasks can be challenging to develop and implement. In fact, for laboratory mice, the most widely used vertebrates in research16,17 and a species heavily relied upon to model affective disorders18, only one JB task has been successfully validated as sensitive to changes in affective state19 despite many attempts over the past decade (see supplementary material of Resasco et al.19 for a summary). This article describes the recently validated murine JB task, detailing its biologically relevant design, and highlighting the ways that this humane task can be applied to test important hypotheses relevant to mouse affect. Overall, the protocol can be implemented to investigate the affective effects of any variable of interest on JB in mice. This would include categorical treatment variables as described here (drug or disease effects, environmental conditions, genetic background, etc.), or relationships with continuous variables (physiological changes, home cage behaviors, etc.).
Experiments were approved by the University of Guelph's Animal Care Committee (AUP #3700), conducted in compliance with Canadian Council on Animal Care guidelines, and reported in accordance with ARRIVE (Animal Research: Reporting of In Vivo Experiments)20 requirements.
1. Experiment preparation
2. Digging training: 5 days, two positive trials per day (Table 2)
3. Discrimination training : 10 days, four trials per day
4. Testing
NOTE: Testing duration is 3-5 days (depending on the time taken for each mouse to meet learning criteria), five trials per day for the sessions in which positive and negative test trials are conducted, and three trials per day when the ambiguous test is conducted.
5. Data analysis
NOTE: Exact analyses required will depend on the details of the experimental design. A general overview is outlined here, but researchers are strongly encouraged to refer to Gygax22 when planning analyses for animal JB experiments, and to Gaskill and Garner23 when selecting sample size (since required analyses are often too complex for a priori power analyses).
Figure 1: Diagram of experimental apparatus. The JB apparatus includes a rectangular arena with two arms. Each arm contains a scent dispenser located at the start and a digging pot placed at the end. Reprinted from Resasco et al.19 with permission from Elsevier. Please click here to view a larger version of this figure.
Phase: | Experimental Design | |
All | High Value Reward | Banana chip |
Low Value Reward | Rodent chow | |
DS+ | Mint or Vanilla (counterbalanced) | |
DS- | Mint or Vanilla (counterbalanced) | |
Digging Training | Digging Training Schedule | 5 days: 2 Pos trials/day |
Digging Trial Duration | 5 min | |
Discrimination Training | Discrimination Training Schedule | 10 days: 4 trials/day |
Digging Trial Order | Days 1-5: 4 trials/day | |
Trial 1: Pos | ||
Trial 2: Neg | ||
Trial 3: Pos | ||
Trial 4: Neg | ||
Days 6-10: 4 trials/day* | ||
Discrimination Trial Duration | 5 min | |
Testing | Testing Schedule | 3-5 days (dependent on time to meet learning criteria) |
5 trials/day | ||
Testing Phase Order | Trials 1 and 2: Pos or Neg** | |
Trial 3: test trial | ||
Trials 4 and 5: Pos or Neg** | ||
Test Trial Duration | 2 min | |
* Trials were pseudorandomized so mice always had two Pos and two Neg trials per day | ||
** Trials were pseudorandomized so mice always had one Pos and one Neg trial before and after the test trial |
Table 1: Summary of experimental design and schedule for training and testing. Number and order of trials per day for Digging Training, Discrimination Training, and Testing phases in addition to experimental design details. Reprinted from Resasco et al.19 with permission from Elsevier.
Trial Details | |||||||
Phase | Trial Type | Scented Arm | Unscented Arm | ||||
Odor Cue | Buried Reward | Inaccessible Reward | Odor Cue | Buried reward | Inaccessible Reward | ||
Digging and Discrimination Training | Pos Training | DS+ | Banana | Chow | Water | Chow | Banana |
Neg Training | DS- | No reward | Banana + chow | Water | Chow | Banana | |
Testing | Pos Test | DS+ | No reward | Banana + chow | Water | No reward | Banana + chow |
Neg Test | DS- | No reward | Banana + chow | Water | No reward | Banana + chow | |
Ambiguous Test | Mint/ vanilla mixture |
No reward | Banana + chow | Water | No reward | Banana + chow | |
Learning Criterion | Mice must dig twice as long in the DS+ pot (Pos test) than the DS- pot (Neg test), and dig for a minimum of 3 s |
Table 2. Summary of trial details. Odor cues and rewards presented in each trial type during Digging Training, Discrimination Training, and Testing phases. DS(+): positive discriminative stimulus, DS(-): negative discriminative stimulus, Pos: positive, Neg: negative. Reprinted from Resasco et al.19 with permission from Elsevier. See Supplementary Table S2 in the original article for the expanded table.
Results presented here reflect relevant findings from Experiment 1 of Resasco et al.19. Subjects in this experiment were 18 female C57BL/6NCrl ('C57') and 18 Balb/cAnNCrl ('Balb') mice. Animals arrived at the facility at 3-4 weeks of age and were randomly assigned to environmentally enriched or conventional housing treatments (EH or CH, respectively) in mixed strain quartets25. Each cage contained one C57 and one Balb, in addition to two DBA/2NCrl mice being used in another experiment. Here, the use of female mice allowed group housing of this social species with environmental enrichment, while avoiding the elevated aggression and resource guarding that can occur in male mice26 (although note that the task has also been applied in non-enriched male nude mice19). CH mice were kept in open-top, transparent polyethylene cages (27L x 16W x 12H cm; n = 9) with corn cob bedding, two types of nesting material (crinkled paper strips and cotton nestlets; Figure 2A), and a paper cup 'shelter'. EH cages were opaque plastic, measuring 60L x 60W x 30H cm3 with one red plexiglass window for observations (n = 9; Figure 2B,C). These conditions are known to improve welfare: containing diverse enrichments (as described previously by Nip et al.27) that mice are motivated to access28, and which reduce behaviors indicative of poor welfare (e.g., stereotypic behavior, aggression, and depression-like inactivity27,28,29). Each EH cage also included an attached 'annex' cage (identical to CH cages but containing only bedding), which mice could freely access via a tunnel (Figure 2C). Annex cages allowed for ease of catching and handling EH mice trained to enter this attachment for food rewards when a cup full of sweet oat cereal was shaken. Once a mouse entered the annex cage, the access tunnel could be blocked and mice could be easily removed from cages using cup or tunnel handling21. This approach thus avoided stressful 'chasing' through complex enriched environments30. Food and water were available ad libitum and the colony room was maintained at 21 ± 1 °C and 35%-55% relative humidity, on a reverse 12:12 h light cycle (lights off at 06:00 and on at 18:00). Mice were differentially housed for 5 weeks prior to commencement of digging training in the apparatus (see timeline in Figure 2D).
Mice underwent training and testing in the JB task as outlined in the Protocol above. Latency to dig in the scented arm during the first and full 2 min of positive, negative, and ambiguous test trials was used to test for housing effects on JB. Here, data were analyzed using Generalized Linear Mixed Models, applying transformations where necessary to meet assumptions of normality and homogeneity. See Resasco et al.19 for a detailed description of analyses (e.g., model selection). Briefly, the repeated measures models always included Trial Type, Housing, Strain, Trial Type x Housing, DS+ Odor, Trial Type x Strain, Trial Type x DS+ Odor, Trial Type x Housing x DS+ Odor, Cage (a random effect nested in Housing and DS+ Odor), and Mouse ID (a random effect nested in Cage, Housing, DS+ Odor and Strain). The simple effects of Housing on latency were determined from the Trial Type x Housing when calculating the Least Squares Means31. Note, two-tailed p-values are reported throughout to demonstrate investigation of treatment effects, but the original validation work by Resasco et al.19 used one-tailed p-values where appropriate32 since one specific response was required to validate the task (see Resasco et al.19 for validation discussion).
Before judgment bias can be assessed in any animal task, it is crucial that two technical criteria be met: first, animals must successfully discriminate between positive and negative cues (i.e., meet learning criteria). For animals meeting this criterion, it must then be demonstrated that the ambiguous cue is interpreted as intermediate. If either of these is not met, then inferences cannot be made about judgment bias and corresponding affective states. In this experiment, all but four C57 mice met learning criteria and one C57 was removed before testing for barbering a cagemate (n = 31). In both the first and full 2 min of testing, Trial Type x DS+ Odor was significant (1 min: F2,62 = 5.67, p = 0.006; 2 min: F2,62 = 5.74, p = 0.005), revealing that Mint DS+ mice unexpectedly interpreted the ambiguous odor mixtures as positive (as if 100% mint), while Vanilla DS+ mice treated the same ambiguous odor mixtures as intermediate (Figure 3A,B). This finding indicated that only Vanilla DS+ mice met the technical requirement of treating the scent mixture as intermediate between the DS+ and DS-, and thus Mint DS+ mice were excluded from subsequent JB analyses.
For Vanilla DS+ mice, simple effects of housing were calculated from the Trial Type x Housing term31. Within this group, Housing influenced digging latencies with CH animals being slower than EH to dig in ambiguous trials, but not in positive or negative trials. This was true after 1 min (ambiguous: t = 2.27, d.f. = 92.94, p = 0.014, Cohen's d = 1.148; positive: t = 0.22, d.f. = 92.94, p = 0.414, Cohen's d = 0.110; negative: t = 0.80, d.f. = 92.94, p = 0.214, Cohen's d = 0.404; see Figure 4A) and after the full 2 min (ambiguous: t = 2.14, d.f. = 91.89, p = 0.018, Cohen's d = 1.083; positive: t = 0.39, d.f. = 91.89, p = 0.348, Cohen's d = 0.198; negative: t = 0.61, d.f. = 91.89, p = 0.273, Cohen's d = 0.308; see Figure 4B), even though Trial Type x Housing x DS+ Odor (1 min: F3,65.37 = 0.36, p = 0.7835; 2 min: F3,65.37 = 0.49, p = 0.688) and Trial Type x Housing (1 min: F2,62 = 1.66, p = 0.198; 2 min: F2,62 = 1.41, p = 0.252) did not account for significant variation. These pessimistic interpretations of ambiguous cues by CH mice reflect negative judgement biases indicative of negative affect.
Figure 2: Housing treatments and experiment timeline. (A) CH laboratory cage. (B) Overhead view of EH. (C) Front view of EH with attached 'annex' cage to facilitate mouse catching/handling. (D) Experimental timeline and summary of positive, negative, and ambiguous training and test trials. DS(+): positive discriminative stimulus, DS(-): negative discriminative stimulus, AMB: ambiguous mixture (50% vanilla-50% mint), B: banana chip, C: Rodent diet ('chow'), X: no food rewards. Reprinted from Resasco et al.19 with permission from Elsevier. Please click here to view a larger version of this figure.
Figure 3: Determining whether mice meet requirements for JB assessment. Digging latency least-square means (±standard error) during positive, negative, and ambiguous test trials. (A) 1 min digging latency in mice receiving mint (M, n = 15) or vanilla (V, n = 16) as the positive discriminative stimulus (DS+) (data Box Cox transformed). (B) 2 min digging latency in the same subjects (data logarithmically transformed). During both time periods, Vanilla DS+ mice met the technical requirement of interpreting ambiguous cues as intermediate. Mint DS+ mice failed to do so (instead interpreting the ambiguous cue as positive) and were consequentially eliminated from subsequent JB analyses. Reprinted from Resasco et al.19 with permission from Elsevier. Please click here to view a larger version of this figure.
Figure 4: Impact of housing treatment on affect-modulated JB. Digging latency least-square means (±standard error) during positive, negative and ambiguous test trials. (A) 1 min digging latency in Vanilla DS+ mice from conventional (CH, n = 7) or enriched (EH, n = 9) housing (data Box Cox transformed). (B) 2 min digging latency in the same subjects (data logarithmically transformed). During both time periods, CH animals demonstrated significantly longer latencies to dig during ambiguous trials than EH conspecifics, indicating negative JB. Reprinted from Resasco et al.19 with permission from Elsevier. Please click here to view a larger version of this figure.
The scent-based digging protocol and results outlined here demonstrate the ability of this novel JB task to detect changes in mouse affective state. The task thus presents a valuable tool for diverse fields of research. Similar to any JB task, to assess animal affect it is critical that animals first learn to discriminate between cues (step 4.7.3) and that the ambiguous stimulus is interpreted as intermediate (step 5.3). Though simple, meeting these requirements can be challenging, particularly in laboratory mice for which over 15 past attempts to develop and implement a mouse JB task have failed19. Here, multiple components played an essential role in meeting these technical criteria and contributed to the success and utility of the task.
First, the ethological design of the task promoted successful discrimination learning since both the discriminative cues and required responses were biologically relevant: mice have impressive olfactory abilities, making them capable of rapid learning and considerable memory spans when presented with odor stimuli33, and they are naturally driven to perform digging for general exploration, foraging, and burrow construction24,34. Further, counterbalancing DS+ odors revealed differences in the ways that Vanilla DS+ and Mint DS+ mice interpreted the ambiguous mixture, confirming that the ambiguous cue was interpreted as intermediate for Vanilla DS+ mice, but not for Mint DS+ mice. It is therefore recommended that only Vanilla is used as the DS+ for any future work utilizing the extract brands and mouse strains used here. Importantly, though counterbalancing yielded successful outcomes in the present experiment, we urge researchers to conduct pilot tests to identify appropriate ambiguous mixtures if implementing changes, since counterbalancing can sometimes add considerable noise to the data, increasing the risk of masking treatment effects35.
Even when these essential criteria are met, JB is not always easy to detect, perhaps owing to the small treatment effect sizes that these experiments commonly yield15. Thus, to ensure task sensitivity, a unique Go/Go design is used since this approach has been shown to be more sensitive to changes in animal affect than Go/No-Go designs in other species15. However, the use of an unscented arm containing a low-value reward in all trials differentiates this task from previous failed attempts to validate a Go/Go JB task for mice36,37,38. Here, during positive trials mice choose between a high-value reward in the DS+ arm and a low-value reward in the unscented arm; and in negative trials, they choose between no reward in the DS- arm and a low-value reward in the unscented arm. Although this requires mice to learn a more complex task (i.e., discriminating between cues predicting different values of reward, rather than simply reward presence or absence), it appears that having a consistent option, where a low-value reward was always present, may have made training and testing less stressful for mice and enhanced learning (with 86% of mice meeting learning criteria). While mice are often assumed to be challenging to train or unable to learn difficult tasks18, results here suggest that their abilities should not be underestimated, and that designing low stress, ethological tasks may be a more effective approach for detecting changes in affect than simpler tasks or those with harsher consequences (e.g., involving punishment like air puffs or white light instead of simply reward absence39,40,41).
Finally, to further reduce stressors that might otherwise interfere with treatment effects and introduce unwanted variability, humane handling methods were implemented21. Here, mice were only handled via cup or tunnel methods throughout their lives (including to and from transport cages and the JB apparatus) to avoid the aversive effects of traditional tail handling21. In addition to this, EH animals were trained to voluntarily enter an annex cage for handling, thus avoiding stressful 'chasing' through complex environments. Together, this approach led to the detection of pessimistic judgment biases in CH mice through longer latencies in response to ambiguous cues. Future researchers should similarly consider whether aspects of their treatments of interest, housing, or husbandry practices have the potential to mask treatment effects or induce floor effects (i.e., where all animals show marked pessimism, negating the ability of the task to detect more subtle treatment differences) so that these can be prevented or mitigated.
Further replication and refinement of this promising task are now needed. To date, this JB task has only been applied to animals experiencing long-term, low-arousal negative affect (as a result of restrictive housing or chronic disease19). It is therefore important that future work aims to test the sensitivity of this task to acute stressors and high-arousal negative affective states. Furthermore, maximizing the value of this task would also involve replicate studies investigating test-retest reliability of individuals to the same, or multiple ambiguous probes. Re-testing with the same probes would allow researchers to test hypotheses regarding changes in affect over time, while exposing subjects to a spectrum of ambiguous cues (i.e., near positive, intermediate, and near negative) could potentially allow for the identification of different types of negative states (particularly depression- versus anxiety-like conditions)11,42. Additional validation experiments should also study the value of shorter protocols, as well as potential differences between strains and sexes (though the original publication does address these issues, successfully employing a shorter protocol to assess affect in male mice19). Indeed, this task could potentially be extended to any rodents intrinsically motivated to make burrows43,44 provided appropriate size modifications are made, and validation is confirmed. Such replication and refinements are important since no other valid JB tasks have been developed for mice to date and since JB tasks are sensitive to both the valence and intensity of affective states (as outlined in the introduction), something that most indicators of animal affect fail to do (e.g., hypothalamic-pituitary-adrenal activity can be altered in response to both positive and negative experiences7,45).
Overall, the development of a mouse JB task represents a promising new tool and opens the door for great progress in the assessment of mouse affect. Mice are the most widely used vertebrates in both basic and translational research17, and this task provides a means to answer essential questions about the welfare of these tens of millions of animals used globally, as well as the links between affect and the diseases or conditions they are used to model. Though the use of this task is not recommended for day-to-day welfare assessment, experimental investigation of housing and husbandry practices could help identify refinements that promote mouse welfare and aid in identifying more subtle signs of animal suffering that can be observed cageside. Given the humane and potentially enriching nature of this task, and the low economic cost of implementing the protocol, this novel JB task has great utility.
The authors have nothing to disclose.
The authors are grateful to Miguel Ayala, Lindsey Kitchenham, Dr. Michelle Edwards, Sylvia Lam and Stephanie Dejardin for their contributions to the Reseasco et al.19 validation work which this protocol is based on. We would also like to thank the mice and our wonderful animal care technicians, Michaela Randall and Michelle Cieplak.
Absolute ethanol | Commercial alcohol | P016EAAN | Dilute to 70% with distilled water, for cleaning |
Centrifuge tubes | Fischer | 55395 | 15 mL tubes used to dilute essences here. However, size may be selected based on total volume required for sample size |
Cheerios (original) | Cheerios | N/A | Commercially available. Used as reward to train animals to enter annex cage for handling |
Corncob bedding | Envigo | 7092 | Teklad 1/8 corncob bedding used in digging pots and animal cages |
Cotton pads | Equate | N/A | Commercially available. Modified in lab to fit within tissue cassettes for scent dispensing |
Digging pots | Rubbermaid | N/A | Commercially available. Containers were modified to fit the apparatus in the lab |
Dried, sweetened banana chips | Stock and Barrel | N/A | Commercially available. High value reward in JB task |
JB apparatus | N/A | The apparatus was made in the lab | |
JWatcher event recording software | Animal Behavior Laboratory, Macquarie University | Version 1.0 | Freely available for download online. Used to score digging behavior during JB task |
Mint extract | Fleibor | N/A | Commercially "Menta (Solución)". Discriminative stimulus odor |
Rodent Diet | Envigo | 2914 | Teklad global 14% protein rodent maintenance diets. Low value reward in JB task and regular diet in home cage |
SAS statistical software | SAS | Version 9.4 | Other comparable software programs (e.g. R) are also appropriate |
Vanilla extract | Fleibor | Commercially available "Vainilla (Solución)". Discriminative stimulus odor | |
Video camera | Sony | DCR-SX22 | Sony handycam. |