Access to this full-text is provided by Springer Nature.
Content available from Animal Cognition
This content is subject to copyright. Terms and conditions apply.
ORIGINAL PAPER
Animal Cognition (2024) 27:66
https://doi.org/10.1007/s10071-024-01905-2
control of individual variation (e.g., in biomedical research
– see Bateson and Nettle 2015). However, aective state
is necessarily subjective and cannot be directly measured
(reviewed in Ede et al. 2019). Therefore, a variety of behav-
ioural or cognitive tests have been developed to aid interpre-
tation of non-human animal (hereafter “animal”) aective
state, such as the judgement bias test (JBT), a frequently
used paradigm to assess animal’s expectation of reward
(‘optimism’).
Originally adapted for use in rats by Harding et al. (2004),
JBTs are now widely used with procedures varying across
and within species (for reviews see Gygax 2014; Roelofs
et al. 2016). In a JBT, animals are taught to associate one
stimulus (SD; e.g., location) with a reward and another stim-
ulus (SΔ) with a less valuable reward, no reward, or punish-
ment (Save). They are then presented with ambiguous stimuli
(e.g., a location between the SD and SΔ/ave) and their behav-
ioural response (e.g., latency to approach; approached or not
approached) is used to assess the animal’s expectation of
Introduction
Determining the valence of an animal’s emotional state (i.e.,
aective state; how they feel) has value for understanding
animals’ appraisal of their life conditions (reviewed in Fra-
ser et al. 1997; Hemsworth et al. 2015) or interpretation of
an experience (e.g., pharmaceutical intervention, Destrez et
al. 2012; painful procedure, Neave et al. 2013). Understand-
ing aective states can also be important for experimental
Alexandra Protopopova
a.protopopova@ubc.ca
1 Animal Welfare Program, Faculty of Land and Food
Systems, The University of British Columbia, Vancouver,
Canada
2 Department of Immunology, Pathophysiology and Veterinary
Preventive Medicine, Wroclaw University of Environmental
and Life Sciences, Wroclaw, Poland
Abstract
Spatial judgement bias tests (JBTs) can involve teaching animals that a bowl provides a reward in one location but does
not in another. The animal is then presented with the bowl placed between the rewarded and the unrewarded locations
(i.e., ambiguous locations) and their latency to approach reects expectation of reward or ‘optimism’. Some suggest that
greater ‘optimism’ indicates better welfare. Performance in JBTs, however, may also indicate a learning history inde-
pendently from welfare determinants. We hypothesized that dogs’ ‘optimism’ in a follow-up JBT may be impacted by a
learning treatment involving additional trials of a dierent discrimination task. Once enrolled, companion dogs (n = 16)
were required to complete three study phases: (1) a pre-treatment JBT, (2) a learning treatment, and (3) a post-treatment
JBT. During the JBTs, dogs were presented with ve locations: one rewarded, one unrewarded, and three ambiguous
(all unrewarded). Dogs were randomly assigned to a trial-based learning task—a nose-touch to the palm of the hand.
In the Experimental discrimination treatment phase (n = 8), dogs were presented with two hands in each trial and only
rewarded for touching one specic hand. In the Control treatment phase (n = 8), dogs were presented with one hand per
trial in alternating sequence and were yoked to dogs in the Experimental group to receive the same number of rewarded
and unrewarded trials (to control for possible frustration). Using a repeated measures mixed model with JBT repeated
within dog, we found no dierence in the change in approach latency to the ambiguous locations between the dogs across
treatments. ‘Optimism’ as measured in this JBT was not altered by the additional discrimination trials used in our study.
Keywords Cognition · Animal welfare · Canine · Aective state · Discrimination · Learning
Received: 25 April 2024 / Revised: 16 September 2024 / Accepted: 17 September 2024 / Published online: 12 October 2024
© The Author(s) 2024
Eect of pre-session discrimination training on performance in a
judgement bias test in dogs
JosephKrahn1· AminAzadian1· CamilaCavalli1· JuliaMiller1,2· AlexandraProtopopova1
1 3
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
Animal Cognition (2024) 27:66
reward, which is operationally dened as ‘optimism’. In the
case of latency as an outcome measure, slower approaches
to the ambiguous stimuli are considered to represent
more negative (‘pessimistic’) expectations while quicker
approaches represent more positive (‘optimistic’) expecta-
tions (Harding et al. 2004; Mendl et al. 2009). A diverse
array of cue modalities (such as auditory, spatial, olfactory,
tactile, and visual cues), outcome measures (such as latency-
based tasks, choice-based tasks), and task designs have been
employed by previous research in the JBT (Hintze et al.
2018). Among these, spatial discrimination, which involves
the ability to distinguish between locations, has been used in
JBT designs across species (e.g., Hintze et al. 2018).
Outcomes of JBTs have been suggested to be inu-
enced by both state dierences within individuals and traits
between individuals (reviewed in Roelofs et al. 2016, 2019).
For instance, an animal’s ‘optimism’ in a JBT is often attrib-
uted to mood and aective state; within individual changes
in JBTs are often used to demonstrate how individuals’
‘optimism’ changes based on a potentially unpleasant situ-
ation (e.g., separation from mother, disbudding in cattle;
Daros et al., 2014). It has also been suggested by some that
greater ‘optimism’ in a JBT indicates greater welfare (e.g.,
Mendl et al. 2009; Vieira de Castro et al. 2020). However,
some studies, have found no relationship between mood and
the outcome of cognitive bias tests (e.g., Schick et al. 2013;
Iigaya et al. 2016; discussed in Burani et al. 2020). More-
over, the outcomes of JBTs can be inuenced by factors that
may be independent from welfare determinants such as per-
sonality traits (dogs; Barnard et al. 2018), and sensitivity to
reinforcers (rats; Rygula et al. 2012).
One area that has received recent attention is how learn-
ing/training inuences animals’ behavioural responses in a
JBT (e.g., see Mendl et al. 2009; reviewed in Nematipour
et al. 2022).
As noted, the JBT involves an animal displaying a par-
ticular behaviour only when a specic stimulus is presented.
This highlights the role of associative learning within the
JBT, as being able to distinguish between stimuli is a vital
component underlying performance on this test (Hall et al.
2021; Roelofs et al. 2019).
Discrimination is a fundamental way in which animals
learn; a discriminative stimulus (SD) is an antecedent stimu-
lus that signals to the animal that if they were to behave
in a certain way, their behaviour would be reinforced. On
the other hand, an extinction stimulus (SΔ) is an antecedent
stimulus that signals no reinforcement for the target behav-
iour. Discrimination training has been adapted to many
commonly used training procedures in dogs, including the
training of working dogs (Hall et al. 2021) and reducing
undesirable behaviours in companion dogs (e.g., David-
son and Rosales-Ruiz 2022). Dogs are frequently trained to
distinguish between visuospatial cues in the context of agil-
ity and other sports (Gácsi et al. 2009; Marshall-Pescini et
al. 2009). Other examples in which dogs engage in recogni-
tion and discrimination include nding a specic stimulus
from a stimulus array of toys, objects, or locations (e.g.,
Dror et al. 2022).
Whereas discrimination learning is part of everyday life,
researchers have also used formal discrimination training
paradigms to ask animals various questions about their
umwelt (e.g., Kelber et al. 2003). During a discrimination
learning process, animals usually develop a response curve
centered around the SD, generalizing their responses to com-
parable stimuli, including ambiguous stimuli that are more
distant from the SD (Strang and Muth 2023). In line with
this, JBTs rely on animals’ dierentiation of an SD among
similar stimuli, thus generating generalization gradients.
The generalization gradient emerging in the JBT is based on
the concept of associative learning such that stimuli which
are more similar (e.g., closer to SD) will occasion more simi-
lar outcomes/responses (Blough 1969; Jamieson et al. 2012;
reviewed in Roelofs et al. 2016). However, these generaliza-
tion gradients that individuals express are not static and can
be experimentally altered. For example, the introduction of
an SΔ/ave close to the SD can result in a shift of the general-
ization gradient in the direction opposite to the SΔ/ave (i.e.,
peak shift; Spence 1937; reviewed in Ghirlanda and Enquist
2003). Generalization gradients can also be inuenced by
an additional SD (Blough 1969; reviewed in Ghirlanda and
Enquist 2003). Moreover, the width of a generalization
gradient can be narrowed with repeated exposure to a task
or widened with increased duration between tasks (e.g.,
pigeons, Homan and Flescher, 1964; reviewed in Osnes
and Lieblien, 2003). While some researchers have tested
the eect of repeated JBTs on the resulting generalization
gradients, the ambiguous stimuli in these experiments may
lose their ambiguity as animals learn which stimuli are
rewarded/unrewarded with repeated exposures (e.g., Wilson
et al. 2023).
Other studies have tested how learning/training inuence
outcomes of a JBT in dogs. However, they often include
potentially confounding eects as the training methods
selected can inuence both how an animal learns (expressed
through their generalization gradient) and their aective
state. For example, Duranton and Horowitz (2019) sug-
gested that dogs trained in nose-work were more ‘opti-
mistic’, reecting better welfare (assuming that only the
aective component of welfare was altered), compared to
dogs trained for heeling exercises (i.e., teaching dogs to walk
in a heel position). However, engaging in nose-work train-
ing may involve more training for stimulus discrimination
(distinguishing between odours) compared to heeling. Other
studies found that dogs were less ‘optimistic’ if trained using
1 3
66 Page 2 of 10
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
Animal Cognition (2024) 27:66
a greater proportion of aversive techniques (Vieira de Cas-
tro et al. 2020) or two or more aversive techniques (Casey
et al. 2021). Aversive-based training (i.e., those using more
frequent positive punishment/negative reinforcement meth-
ods) may aect mood (i.e., through repeated ‘unpleasant’
experiences), but perhaps also cause higher discrimination
of voice/hand gestures than purely reward-based training
(e.g., Hearst 1962). For example, aversive-based training
can include the introduction of aversive consequences for
responding under stimuli that are similar but not the same as
the trained command (e.g., not moving from a stay position
until a very precise release cue is provided and not when
other similar utterances are made). Therefore, aversive
training introduces a “cost” for the dog to engage in impre-
cise responses. In such cases, it is challenging to disentangle
which factors are inuencing the outcomes in the JBT.
Our study aimed to disentangle the eects of additional
discrimination training from aective states in the results
of a JBT in dogs, as we suspected that this additional train-
ing may alter the performance of dogs. Our objective was
to determine if experiencing training focused on discrimi-
nation could aect the latency to engage with ambiguous
stimuli during a JBT. To this end, dogs received an initial
JBT, then underwent a discrimination training phase or a
control phase, then received a second JBT. We hypothesized
that there would be an interaction eect in which animals
will behave more ‘pessimistically’ following discrimina-
tion training (i.e., longer latencies to approach ambiguous
stimuli) compared to the control.
Materials and methods
Animals
This study was conducted between November 2022 and
March 2023. Privately-owned companion dogs were
recruited for this study from residents of Greater Vancou-
ver, British Columbia, Canada. Dogs were required to be
non-reactive towards humans, have received some form
of nose-to-hand-touch training, and be able to come to the
laboratory on three separate days for the three study phases.
Once enrolled, all dogs were required to complete all three
study phases: (1) a pre-treatment JBT (JBT1), (2) a hand-
touch treatment phase, and (3) a post-treatment JBT (JBT2).
In total, 29 dogs were enrolled, but 13 were excluded for
varying reasons: unable to meet the criteria of the acqui-
sition phase in JBT1 (n = 8), dietary issues with rewards
(n = 1), discontinued following weather-caused delays
(n = 3), and researcher procedural error during the task
(n = 1). We achieved a nal sample size of n = 16; n = 8 in
each treatment (Table 1). To the best of our ability, dogs
in the treatment groups were balanced for sex (Discrimina-
tion, 5 male, 3 female; Control 4 male, 4 female), age (Dis-
crimination, 3.2 ± 1.3 y, mean ± SD; Control, 4.5 ± 3.3 y),
and breed clade (following Parker et al. 2017). Dogs varied
in the intervals between JBT1 and treatment (5.8 ± 5.0 d;
range: 1–22 d) and between treatment and JBT2 (5.7 ± 2.7
d; 1–10 d).
Most dogs in the nal sample (n = 13) were rewarded with
chicken hotdogs (Maplelodge Farms®, Brampton, Ontario,
Canada) and the reward size per trial was approximately
proportional to dog bodyweight (~ 1 g hotdog/10kg dog).
Owners supplied rewards for dogs with dietary restrictions
Table 1 The characteristics of the dogs used in the study. Numbering does not represent order of testing. Dog owner provided breed information
and the clade was assessed following Parker et al. (2017). M, F, L, and R represent male, female, left, and right, respectively
Dog Clade Sex Rewarded
side
Age
(years)
Treatment Days between JBT 1 and treatment Days between treatment and JBT2
1 UK rural M L 2 Discrimination 7 7
2Masti M R 2.3 Discrimination 1 7
3 Pinscher M R 2 Discrimination 1 1
4 Pointer Setter M L 4.5 Discrimination 7 7
5 Drover F R 4.5 Discrimination 3 8
6 Terrier M L 5 Discrimination 7 1
7 Terrier F R 2 Discrimination 7 7
8 Mixed F L 2.9 Discrimination 4 10
9 UK rural F L 0.8 Control 3 7
10 Mixed M R 3.5 Control 5 7
11 UK rural M R 8 Control 8 4
12 Mixed F R 7.2 Control 3 7
13 UK rural F L 8 Control 1 4
14 Terrier M R 0.8 Control 7 1
15 Mixed M L 1.1 Control 22 7
16 Continental Herder F L 6.3 Control 7 6
1 3
Page 3 of 10 66
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
Animal Cognition (2024) 27:66
(P), near-positive (NP), middle (M), near-negative (NN),
and negative (N). Before each trial, The Timer would ver-
bally indicate to The Experimenter the location to place the
bowl (i.e., “positive”, “negative”, “NP”, “M”, “NN”). When
located at position P, the bowl always contained a reward,
while at N or the ambiguous locations (NP, M, NN) the bowl
never contained a reward. When placing the bowl on the
oor, The Experimenter would say “look” while looking at
the bowl. The Experimenter would return to their central
location and say “OK” to indicate to The Handler to release
the dog and allow them to proceed to the bowl. The latency
to engage with the bowl was measured from the time The
Experimenter said “OK” to the time that the dogs’ rst front
paw crossed a 1 m diameter chalk circle surrounding the
bowl. Similar to Duranton and Horowitz (2019), we used a
maximum latency to approach the presented stimuli of 10s
in all three phases (one dog waited for an additional 10s
on two occasions, which was corrected to 10s for analy-
sis). All latencies were measured with a stopwatch, with all
times rounded to the nearest second. The orientation of the
bowl presentation locations (i.e., P on left or right side of
the room) was balanced across treatment groups and con-
sistent within dogs across JBTs. To control for odour, we
placed ve treats between two non-air-tight containers that
were used as the bowl in all trials. The bowl’s opening faced
away from the dogs’ start position, limiting their ability to
(e.g., allergies; n = 3) and the reward size per trial was that
which was normally given by the owner. Rewards were kept
consistent in size and type within each individual across all
trials and phases.
General procedures
All tests involved three researchers hereafter referred to
as “The Experimenter”, “The Timer”, and “The Handler”.
During all study phases The Experimenter, The Timer, The
Handler, and the dog’s owner wore mirrored sunglasses to
limit the eect of inadvertent visual cues to the dog. Water
was available ad libitum during all test procedures. Dogs
were given approximately 5 min to habituate to the room
prior to the beginning of tests in all of the phases.
JBT general procedures
The JBT we used in our study followed the guidelines in
Mendl et al. (2010a, b) with some modications outlined in
the brief description of each test below. The room had the
same orientation for all dogs regardless of rewarded/posi-
tive (P) side (see Fig. 1). During both the acquisition and
testing phase, The Timer had a pre-determined list of the
order of locations where a bowl would be presented. There
were ve locations where the bowl was placed: positive
Fig. 1 Before each experiment, ve circles with a diameter of 1 m were
drawn on the oor in chalk; the centre of each circle was 3.5 m from
the starting position, represented by the dotted grey lines. These circles
were 1.2 m and 20O from the adjacent circle. At its centre, each circle
had a 5 cm piece of adhesive hook and loop tape. The letters P, NP, M,
NN, N indicating the location were drawn on the circle with chalk. The
Timer was seated (next to the stopwatch, coloured black), the dog’s
owner was seated (coloured blue), The Handler was next to the dog,
and The Experimenter was standing behind the middle position
1 3
66 Page 4 of 10
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
Animal Cognition (2024) 27:66
JBT testing
Latency in a JBT test was measured the same as in the pre-
testing. In testing, all ve locations were presented individ-
ually in the order as follows: M, NN, NP, NN, NP, M, NN,
M, NP (Mendl et al. 2010a). This order was maintained for
all tests to ensure each location was presented rst, second,
and third within a block of three trials (i.e., M, NN, NP), in
an eort to minimize bias and improve comparability across
subjects (as replicated from Wilson et al. 2023). Similar to
Vieira de Castro (2020), each ambiguous location trial was
separated from the next with one P and one N trial in ran-
dom order.
Treatment procedures
Treatment room set-up
All procedures during the treatment phase had a consistent
room layout (Fig. 2). The Experimenter was at the opposite
side of the room than in the JBTs in an eort to minimize the
similarity between the JBT and the treatment procedures.
see the contents of the bowl until they crossed the chalk
circle.
JBT acquisition
All dogs were required to complete a pre-training (acquisi-
tion phase) before being tested in the JBT. In the acquisi-
tion phase dogs were presented with the bowl in one of two
probe locations (N or P). All sessions began with two P then
two N trials. Afterwards, each block of four trials following
included two P and two N trials that were pseudo-randomly
ordered such that there was never more than two of the same
location presented consecutively. To complete the acquisi-
tion phase, dogs were required to have had a greater latency
in the most recent three negative trials compared to the most
recent three positive trials (Mendl et al. 2010a, b). Similar to
Barnard et al. (2018), we set a maximum of 50 trials to meet
the acquisition criteria. Dogs were required to complete a
minimum of 15 trials (Mendl et al. 2010a). On one occa-
sion, a dog met the criteria after 14 trials and the trial was
terminated prematurely due to human error; this dog was
included in all analyses.
Fig. 2 The Handler and the dog faced The Experimenter at a distance
of 1.5 m, represented by the grey dotted line. The Timer was seated
(next to the stopwatch, coloured black), the dog’s owner was seated
(colored blue), The Handler walked the dog, and The Experimenter
was seated in front of the dog
1 3
Page 5 of 10 66
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
Animal Cognition (2024) 27:66
trials. This was designed to control for potential dierences
in frustration, as non-rewarded trials are likely frustrating
(e.g., Amsel and Ward 1965; Daly 1971), which may have
resulted a negative emotional state (Amsel 1992). Because
we aimed to disentangle the eects of learning and aective
states, it was important to try to keep the dogs in a similar
aective state (albeit this, of course, could not be fully con-
trolled). Additionally, to ensure that no single hand resulted
in a higher probability of rewarded trials, the number of
unrewarded/rewarded trials was randomized such that there
were no more than two unrewarded trials consecutively for
both or either hand. In cases where there were an odd num-
ber of rewarded and unrewarded trials (n = 1), one side was
rewarded once more than the other. Hands were presented
one at a time in alternating sequence. If the hand that was
pre-determined to be rewarded was not touched within 10s,
Assistance was provided similarly to the Discrimination
treatment.
Data analysis and statistics
Analyses were conducted in R (version 4.0.3, R Foundation
for Statistical Computing, Vienna, Austria) and SAS (ver-
sion 9.04.01; SAS Institute Inc., Cary, NC, USA). We used a
mixed model in SAS (Proc Mixed) with compound symme-
try with JBT repeated within dog. The independent variable
was the interaction between JBT (1 or 2) and the treatment
(Discrimination or Control) and the dependent variable was
the average change in latency to approach the ambiguous
stimuli (combined NP, N, NN) during JBT1 and JBT2. We
pooled the ambiguous locations as a simple approach to sat-
isfy our model’s assumptions (and maintain the most appro-
priate degrees of freedom) because the approaches to the
ambiguous stimuli were considered to be non-independent
within a session. Other more complex nesting approaches
for managing independence are possible in JBTs (see Gygax
2014 for an extended discussion). Due to factors outside of
our control, the number of days between study phases (i.e.,
JBT1 to treatment and treatment to JBT2) varied across
subjects. Therefore, we also included the interactions of
JBT with the number of days from JBT to the treatment
and JBT with the number of days between the treatment to
JBT2. We conducted a second model, similar to the above
model, but with the proportion of times the ambiguous loca-
tions were approached out of all trials at that location as
the dependent variable (i.e., binary outcomes: approached
or not approached within the 10s-time limit). The approach
latency between NP and NN were average across all JBTs
within individuals and compared using a student’s t-test in
SAS (Proc Univariate).
Treatment acquisition
In the treatment acquisition, The Experimenter presented
one hand at a time, alternating left and right. When The
Experimenter would present the hand, The Handler would
say “OK” and release any tension on the leash. When the
dog touched the presented hand, The Experimenter would
say “YES” and the dog received a reward. If a dog did
not touch The Experimenter’s presented hand on one side
ve times in a row, they would be presented with the same
hand twice. If they still did not touch the presented hand, a
treat was placed between The Experimenter’s ngers of the
presented hand so that it was visible to the dog (hereafter
referred to as “Assistance”); the treat was provided when
the hand was touched.
Dogs completed this acquisition phase once they con-
secutively touched four presented hands (two right, two
left) without Assistance in ≤ 3s each. All dogs completed
the acquisition phase with variation in the number of trials
required (10.4 ± 11.4, mean ± SD).
Discrimination treatment
In the Discrimination treatment (n = 8), the goal was to
teach the dog that if they were to respond to one hand, but
not the other, they would receive a reward. The Experi-
menter presented both hands simultaneously. The dog was
only rewarded for touching one of the hands. In an eort
to control for previous learning history across dogs with
nose-to-hand touch in everyday life, the target hand was
selected according to the owner’s non-dominant hand (left
in all cases). If the dog did not touch either hand within 10s
during two sequential presentations, they were presented
with only the rewarded hand. If the presented hand was still
not touched, they were provided with Assistance. After 10
presentations (i.e., one 10-trial block) dogs were given a
30s break. To complete this treatment phase, the dog was
required to touch the rewarded hand 8 times in a 10-trial
block. The 10-trial block was always completed except
by one dog whose trial was prematurely terminated after
8 rewarded-hand touches (experimenter error). All dogs
completed this stage requiring a varied number of trials
(22.3 ± 10.2), resulting in a varied proportion of trials that
were rewarded (70.0 ± 17.3%).
Control treatment
The aim of the Control treatment was to establish a con-
trol condition where no hand is an SD. Dogs in the Con-
trol treatment (n = 8) were randomly match-paired to a
dog in the Discrimination treatment to yoke the number of
rewarded/unrewarded trials to be similar for dogs across
1 3
66 Page 6 of 10
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
Animal Cognition (2024) 27:66
Results
Of the 29 dogs that attempted JBT1, 8 (28%) failed to meet
the acquisition criteria for JBT1. For the dogs that completed
JBT1, the number of trials required to meet the acquisition
criteria was similar between dogs in the two treatments (Dis-
crimination: JBT1, 28.5 ± 5.1, mean ± SE; JBT2, 19.9 ± 2.5;
Control: JBT1, 27.4 ± 3.6, JBT2, 18.0 ± 1.6). Probe location
aected the dogs’ responses across JBTs with dogs having
a greater latency to approach the near negative (5.4 ± 0.1)
than near positive location (2.9 ± 0.2; t15 = 9.1, P < 0.001;
Fig. 3).
We found no dierence in the change in approach latency
to the ambiguous stimuli between the dogs in the Discrimi-
nation and Control treatments (F3,9 = 0.24, P = 0.86; Fig. 4).
The number of days between study phases did not have
a statistically signicant eect on latency (JBT1 to treat-
ment, F2,9 = 0.26, P = 0.78; treatment to JBT2, F2,9 = 2.00,
P = 0.19). In the second model, the proportion of approaches
to the ambiguous stimuli was also not aected by treatment
(F3,5 = 0.13, P = 0.94) or the number of days between study
phases (JBT1 to treatment, F2,5 = 0.39, P = 0.70; treatment
to JBT2, F2,9 = 0.82, P = 0.49).
Fig. 4 Points represent average change in the pooled latency to the
ambiguous locations (NP, M, NN) from JBT1 to JBT2 per dog. Colours
represent match-paired dogs. Black points represent means and error
bars represent standard error of the means. The dashed line represents
no change across JBT1 and JBT2
Fig. 3 Dogs’ average latency to approach the ve locations. Circles and triangles represent JBT1 and JBT2, respectively. Colour represents treat-
ment. Error bars represent standard error of the mean
1 3
Page 7 of 10 66
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
Animal Cognition (2024) 27:66
the location was approached/not approached (similar to a
go/no-go task, where dogs suppress a response to SΔ rather
than engage in a set behaviour procedure; e.g., see Daros et
al., 2014). Although the proportion of trials where the dog
approached/not approached was investigated as an addi-
tional outcome measure, there is still a possibility that this
procedure may have aected the interpretation of our out-
comes (see Gygax 2014).
While suggested by some as a gold standard for assess-
ing animal emotions (e.g., Bateson and Nettle 2015), JBTs
have faced scrutiny for a variety of factors, including low
repeatability within individuals (e.g., Wilson et al. 2023)
and variation across studies in methodology, contributing to
variation in results (discussed in Roelofs et al. 2016; Lag-
isz et al. 2020; Lecorps et al. 2021). Similarly, statistical
approaches in analyzing JBT data vary and are debated,
lacking a unied approach (e.g., Gygax 2014; Bateson
and Nettle 2015; Bethell 2015; Bethell and Koyama 2015;
Burani et al. 2020). We treated the ambiguous locations
as non-independent in our analysis (similar to Neave et
al. 2013), although dierent statistical approaches may be
more or less optimal in managing variability in the data and
adhering to model assumptions.
Although some have considered that JBTs often have
weak acquisition criteria (Roelofs et al. 2016), others have
observed considerable failure rates where dogs do not meet
the minimum acquisition criteria (e.g., 27% in Burani et
al. 2020; 28% in this study) that biases studies to represent
only those animals that are able to complete acquisition.
Our use of a relatively neutral SΔ rather than a punishment
(Save) at location N provided a smaller dierence in pay-
os between locations N and P and possibly contributed to
the considerable failure rates to meet the acquisition criteria
(see Mendl et al. 2009). Similarly, there can be variation in
the number of trials required to meet acquisition criteria that
may not be included in assessing JBT outcomes (see Chan et
al. 2023). We recommend future studies balance treatment
groups by performance in JBT1 when possible; this vari-
able was not included when balancing in our study. More-
over, the eect of hunger/satiation on measured outcomes
(reviewed in Lecorps et al. 2021), and animals learning
that the ambiguous cues are unrewarded (loss of ambigu-
ity) within one JBT (Brilot et al. 2010) or across repeated
JBTs (Doyle et al., 2010; in Roelofs et al. 2016) are com-
mon concerns that were not addressed in our experimen-
tal design. In some cases, methods have been developed to
overcome these challenges. For instance, loss of ambiguity
can be addressed by using partial reinforcement schedules
at rewarded stimuli (e.g., Brilot et al. 2010; Neave et al.
2013) or a secondary reinforcer (e.g., clicker; Keen et al.
2014 in Roelofs et al. 2016). The sequence at which stimuli
are presented can aect the interpretation of those stimuli
Discussion
We set out to examine if a specic training experience focus-
ing on visuospatial discrimination (where dogs were trained
to discriminate between two visual stimuli (hands) based on
their spatial orientation) but not involving stimuli similar to
those involved in the JBT, could alter dogs’ responses to the
ambiguous stimuli presented in a JBT. We hypothesized that
discrimination training would narrow the stimulus response
curve, thus increasing the response latency to ambigu-
ous stimuli, therefore articially creating the ‘pessimistic’
interpretation in the JBT. However, we found no eect of
the training treatment on the dogs’ latency to approach the
ambiguous locations in our JBTs. Our results may suggest
that JBTs provide a robust method to investigate ‘optimism’
and aective state (e.g., mood)—an aspect of welfare—in
dogs. Alternatively, responses in the JBT procedure may not
be sensitive to the type of pre-session discrimination train-
ing used in this study, but could be generally sensitive to
previous learning history (e.g., discussed in Bethell 2015;
Wilson et al. 2023).
Importantly, our small sample size and the notable inter-
individual variation (see Fig. 4) may have limited our ability
to observe signicant dierences. Moreover, the treatment
we used involved a dierent kind of stimulus (hand) com-
pared to a JBT test (bowl), which may involve a dierent
kind of spatial information for the dog, and a dierent pro-
cedure (e.g., dogs in the Discrimination treatment choosing
between two response options in the same location versus
approaching a single response option in dierent locations
in the JBT). Perhaps our learning tasks were too dissimilar
to the JBT procedure, thus limiting the priming eect of the
learning task on the individual within the testing environ-
ment. Similarly, our treatment may have been too short in
duration, did not have strict enough learning criteria, or had
too long intervals between the treatment and retesting in
JBT2 (e.g., the recency eect; Jones et al. 2006).
There were other limitations in our study design. For
example, we rounded all times to the nearest second, reduc-
ing the precision (and potentially the accuracy) of our mea-
surements. Similarly, we were unable to control for many
aspects of the dogs’ lives as they were privately owned. For
example, previous research found that stimulus features
used in training can impact the degree of learning-depen-
dent generalization (Wisniewski et al. 2009) and the stimu-
lus response curve (Strang and Muth 2023). Therefore, the
type of stimuli to which the dog was previously exposed
could further inuence performance on the JBT.
We set a maximum time of 10s in all procedures (simi-
lar to Duranton and Horowitz 2019). However, this meant
that our latency measures were continuous until 10s at
which point the outcome was treated as a test of whether
1 3
66 Page 8 of 10
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
Animal Cognition (2024) 27:66
inherent value for scientists, which may be elevated for early-career
scientists (see Rawat and Meena 2014). The analyses of data and fram-
ing of the research were inherently inuenced by the values and inter-
ests of the authors and should be considered in the interpretation of
this paper.
Competing interests The authors declare no competing interests.
Open Access This article is licensed under a Creative Commons
Attribution 4.0 International License, which permits use, sharing,
adaptation, distribution and reproduction in any medium or format,
as long as you give appropriate credit to the original author(s) and the
source, provide a link to the Creative Commons licence, and indicate
if changes were made. The images or other third party material in this
article are included in the article’s Creative Commons licence, unless
indicated otherwise in a credit line to the material. If material is not
included in the article’s Creative Commons licence and your intended
use is not permitted by statutory regulation or exceeds the permitted
use, you will need to obtain permission directly from the copyright
holder. To view a copy of this licence, visit http://creativecommons.
org/licenses/by/4.0/.
References
Amsel A (1992) Frustration theory: an analysis of dispositional learn-
ing and memory. Cambridge University Press, New York
Amsel A, Ward JS (1965) Frustration and persistence: resistance to
discrimination following prior experience with the discrimi-
nanda. Psychol Monographs: Gen Appl 79(4):1–41
Barnard S, Wells DL, Milligan ADS, Arnott G, Hepper PG (2018) Per-
sonality traits aecting judgement bias task performance in dogs
(Canis familiaris). Sci Rep 8:6660
Bateson M, Nettle D (2015) Development of a cognitive bias meth-
odology for measuring low mood in chimpanzees. PeerJ 3:e998
Bethell EJ (2015) A how-to guide for designing judgment bias stud-
ies to assess captive animal welfare. J Appl Anim Welfare Sci
18:S18–S42
Bethell EJ, Koyama NF (2015) Happy hamsters? Enrichment induces
positive judgement bias for mildly (but not truly) ambiguous cues
to reward and punishment in Mesocricetus auratus. Royal Soc
Open Sci 2(7):140399
Blough D (1969) Generalisation gradient shape and summation in
steady-state tests. J Exp Anal Behav 12:91–104
Brilot BO, Asher L, Bateson M (2010) Stereotyping starlings are more
‘pessimistic’. Anim Cogn 13:721–731
Burani C, Barnard S, Wells D, Pelosi A, Valsecchi P (2020) Using
judgment bias test in pet and shelter dogs (Canis familiaris):
methodological and statistical caveats. PLoS ONE 15:1–23
Casey RA, Naj-Oleari M, Campbell S, Mendl M, Blackwell EJ (2021)
Dogs are more pessimistic if their owners use two or more aver-
sive training methods. Sci Rep 11:1–8
Chan K, Arellano C, Horowitz A (2023) Here Puppy, chew on this:
short-term provision of Toys does not improve Welfare in Com-
panion Dogs. Animals 13(21):3340
Daly HB (1971) Evidence for frustration during discrimination learn-
ing. J Exp Psychol 88(2):205
Davidson MA, Rosales-Ruiz J (2022) Reducing the occurrence of
mouthing and jumping in a dog through conditional discrimina-
tion training. J Exp Anal Behav 118(2):250–260
De Vieira AC, Fuchs D, Morello GM, Pastur S, De Sousa L, Olsson
IAS (2020) Does training method matter? Evidence for the nega-
tive impact of aversive-based methods on companion dog wel-
fare. PLoS ONE 15:1–26
(discussed in Jones et al. 2006; Zentall et al. 2014), a factor
which was not completely controlled for in our experiment.
We encourage future research to continue critically
assessing and rening measures of cognitive/judgement
bias used to estimate aective state. For instance, studies
could investigate further the eect of experimental pro-
cedures (interventions) designed to change this response
gradient around the SD. This can be achieved, for example,
by introducing stimuli outside of the SD or SΔ which may
inuence an animal’s outcomes in a JBT (i.e., peak shift;
suggested by Gygax 2014; Roelofs et al. 2016). We also
encourage continued discussion on statistical approaches in
JBTs and that future studies investigate the implications of
the considerable failure rates in some cognitive tests.
Conclusions
We found no evidence that the learning treatment tasks
aected ‘optimism’ (latency to approach an ambiguous
stimulus) in dogs as measured using a JBT, providing no
evidence against the robustness of the JBT in dogs. The
considerable rate of dogs that failed to meet the acquisition
criteria for the rst JBT in our study may be a concern that
we investigated only particular dogs who were, in some
way, dierently inuenced by our procedures. We encour-
age future investigation into understanding how learning
history can inuence JBT outcomes, which will improve
our understanding of the limitations of JBTs as a means to
predict animals’ expectations and aective state.
Acknowledgements We are thankful for the dogs and their human
companions, who participated in this study. We thank our colleagues
in the Animal Welfare Program at the University of British Colum-
bia for thoughtful discussions, especially Lexis Ly, who also provided
assistance with handling dogs during some experimental trials. We are
thankful for the constructive comments of two anonymous reviewers.
Author contributions Conceptualization, J.K. and A.P.; study design
and data collection, J.K., A.A., C.C., J.M., and A.P.; writing—original
draft, J.K.; writing—review and editing, J.K., A.A., C.C., J.M., and
A.P.; statistical analysis and data visualization, J.K.; supervision and
funding acquisition, A.P.; all authors have read and agreed to the sub-
mitted version of the manuscript.
Data availability The datasets and analysis used in this study are avail-
able at https://doi.org/10.5683/SP3/QJLSKV.
Declarations
Ethics approval All procedures were approved by the University of
British Columbia Animal Care Committee (Protocol A22-0170).
Researcher reexivity statement All authors have worked and/or lived
closely with dogs for many years. This work was conducted in partial
fulllment of J.K.’s master’s degree. Publishing research papers has
1 3
Page 9 of 10 66
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
Animal Cognition (2024) 27:66
Lecorps B, Weary DM, von Keyserlingk MAG (2021) Negative
expectations and vulnerability to stressors in animals. Neurosci
Biobehavioural Reviews 130:240–251
Marshall-Pescini S, Passalacqua C, Barnard S, Valsecchi P, Prato-Pre-
vide E (2009) Agility and search and rescue training dierently
aects pet dogs’ behaviour in socio-cognitive tasks. Behav Pro-
cess 81(3):416–422
Mendl M, Burman OHP, Parker RMA, Paul ES (2009) Cognitive bias
as an indicator of animal emotion and welfare: emerging evidence
and underlying mechanisms. Appl Anim Behav Sci 118:161–181
Mendl M, Brooks J, Basse C, Burman O, Paul E, Blackwell E, Casey
R (2010a) Dogs showing separation-related behaviour exhibit a
‘pessimistic’ cognitive bias. Curr Biol 20(19):R839–R840
Mendl M, Burman OH, Paul ES (2010b) An integrative and functional
framework for the study of animal emotion and mood. Proceed-
ings of the Royal Society B: Biological Sciences. 277(1696),
2895–2904
Neave HW, Daros RR, Costa JH, von Keyserlingk MA, Weary DM
(2014) Pain and pessimism: dairy calves exhibit negative judge-
ment bias following hot-iron disbudding. PLoS ONE 8(12),
e80556
Nematipour B, Bračić M, Krohs U (2022) Cognitive bias in ani-
mal behavior science: a philosophical perspective. Anim Cogn
25(4):975–990
Osnes PG, Lieblein T (2003) An explicit technology of generalization.
Behav Analyst Today 3(4):364–374
Parker HG, Dreger DL, Rimbault M, Davis BW, Mullen AB, Carpin-
tero-Ramirez G, Ostrander EA (2017) Genomic analyses reveal
the inuence of geographic origin, migration, and hybridization
on modern dog breed development. Cell Rep 19(4):697–708
Rawat S, Meena S (2014) Publish or perish: where are we heading?
J Res Med Sciences: Ocial J Isfahan Univ Med Sci 19(2):87
Roelofs S, Boleij H, Nordquist RE, Van der Staay FJ (2016) Mak-
ing decisions under ambiguity: judgment bias tasks for assessing
emotional state in animals. Front Behav Neurosci 10:119
Roelofs S, Alferink FA, Ipema AF, van de Pas T, van der Staay FJ,
Nordquist RE (2019) Discrimination learning and judgment bias
in low birth weight pigs. Anim Cogn 22:657–671
Rygula R, Pluta H, Popik P (2012) Laughing rats are optimistic. PLoS
ONE 7 (12), e51959
Schick A, Wessa M, Vollmayr B, Kuehner C, Kanske P (2013) Indirect
assessment of an interpretation bias in humans: neurophysiologi-
cal and behavioral correlates. Front Hum Neurosci 7:272
Spence K (1937) The dierential response in animals to stimuli vary-
ing within a single dimension. Psychol Rev 44:430–444
Strang C, Muth F (2023) Judgement bias may be explained by shifts
in stimulus response curves. Royal Soc Open Sci 10(4):221322
Wilson C, Hall N, Aviles-Rosa EO, Campbell K, Arnott G, Reeve C
(2023) The eect of repeated testing on judgement bias in domes-
tic dogs (Canis familiaris). Anim Cogn 26(2):477–489
Wisniewski MG, Church BA, Mercado E (2009) Learning-related
shifts in generalization gradients for complex sounds. Learn
Behav 37(4):325–335
Zentall TR, Wasserman EA, Urcuioli PJ (2014) Associative concept
learning in animals. J Exp Anal Behav 101(1):130–151
Publisher’s note Springer Nature remains neutral with regard to juris-
dictional claims in published maps and institutional aliations.
Destrez A, Deiss V, Belzung C, Lee C, Boissy A (2012) Does reduction
of fearfulness tend to reduce pessimistic-like judgment in lambs?
Appl Anim Behav Sci 139(3–4):233–241
Dror S, Sommese A, Miklósi Á, Temesi A, Fugazza C (2022) Mul-
tisensory mental representation of objects in typical and gifted
Word Learner dogs. Anim Cogn 25(6):1557–1566
Duranton C, Horowitz A (2019) Let me sni! Nosework induces posi-
tive judgment bias in pet dogs. Appl Anim Behav Sci 211:61–66
Ede T, Lecorps B, von Keyserlingk MA, Weary DM (2019) Sympo-
sium review: scientic assessment of aective states in dairy
cattle. J Dairy Sci 102(11):10677–10694
Fraser D, Weary DM, Pajor EA, Milligan BN (1997) A scientic con-
ception of animal welfare that reects ethical concerns. Anim
Welf 6(3):187–205
Gácsi M, Kara E, Belényi B, Topál J, Miklósi Á (2009) The eect of
development and individual dierences in pointing comprehen-
sion of dogs. Anim Cogn 12:471–479
Ghirlanda S, Enquist M (2003) A century of generalization. Anim
Behav 66:15–36
Gygax L (2014) The A to Z of statistics for testing cognitive judgement
bias. Animal Behaviour
Hall NJ, Johnston AM, Bray EE, Otto CM, MacLean EL, Udell MA
(2021) Working dog training for the twenty-rst century. Front
Veterinary Sci 8:646022
Harding EJ, Paul ES, Mendl M (2004) Cognitive bias and aective
state. Nature 427(6972):312–312
Hearst E (1962) Concurrent generalization gradients for food-con-
trolled and shock-controlled behavior. J Experimental Anal Anim
Behav 5:19–31
Hemsworth PH, Mellor DJ, Cronin GM, Tilbrook AJ (2015) Scientic
assessment of animal welfare. N Z Vet J 63(1):24–30
Hintze S, Melotti L, Colosio S, Bailoo JD, Boada-Saña M, Würbel H,
Murphy E (2018) A cross-species judgement bias task: integrat-
ing active trial initiation into a spatial Go/No-go task. Sci Rep
8(1):5104
Homan HS, Fleshler M (1964) Stimulus aspects of aversive con-
trols: stimulus generalization of conditioned suppression follow-
ing discrimination training. J Experimental Anal Anim Behav
7(3):233–239
Iigaya K, Jolivald A, Jitkrittum W, Gilchrist ID, Dayan P, Paul E,
Mendl M (2016) Cognitive bias in ambiguity judgements: using
computational models to dissect the eects of mild mood manipu-
lation in humans. PLoS ONE 11(11), e0165840
Jamieson RK, Crump MJC, Hannah SD (2012) An instance theory
of associative learning. Learn Behav 40:61–82. https://doi.
org/10.3758/s13420-011-0046-2
Jones M, Love BC, Maddox WT (2006) Recency eects as a window
to generalization: separating decisional and perceptual sequential
eects in category learning. J Experimental Psychology: Learn
Memory Cognition 32(2):316
Keen HA, Nelson OL, Robbins CT, Evans M, Shepherdson DJ, New-
berry RC (2014) Validation of a novel cognitive bias task based
on dierence in quantity of reinforcement for assessing environ-
mental enrichment. Anim Cogn 17:529–541
Kelber A, Vorobyev M, Osorio D (2003) Animal colour vision–behav-
ioural tests and physiological concepts. Biol Rev 78(1):81–118
Lagisz M, Zidar J, Nakagawa S, Neville V, Sorato E, Paul ES, Bateson
M, Mendl M, Løvlie H (2020) Optimism, pessimism and judge-
ment bias in animals: a systematic review and meta-analysis.
Neurosci Biobehavioural Reviews 118:3–17
1 3
66 Page 10 of 10
Content courtesy of Springer Nature, terms of use apply. Rights reserved.
1.
2.
3.
4.
5.
6.
Terms and Conditions
Springer Nature journal content, brought to you courtesy of Springer Nature Customer Service Center GmbH (“Springer Nature”).
Springer Nature supports a reasonable amount of sharing of research papers by authors, subscribers and authorised users (“Users”), for small-
scale personal, non-commercial use provided that all copyright, trade and service marks and other proprietary notices are maintained. By
accessing, sharing, receiving or otherwise using the Springer Nature journal content you agree to these terms of use (“Terms”). For these
purposes, Springer Nature considers academic use (by researchers and students) to be non-commercial.
These Terms are supplementary and will apply in addition to any applicable website terms and conditions, a relevant site licence or a personal
subscription. These Terms will prevail over any conflict or ambiguity with regards to the relevant terms, a site licence or a personal subscription
(to the extent of the conflict or ambiguity only). For Creative Commons-licensed articles, the terms of the Creative Commons license used will
apply.
We collect and use personal data to provide access to the Springer Nature journal content. We may also use these personal data internally within
ResearchGate and Springer Nature and as agreed share it, in an anonymised way, for purposes of tracking, analysis and reporting. We will not
otherwise disclose your personal data outside the ResearchGate or the Springer Nature group of companies unless we have your permission as
detailed in the Privacy Policy.
While Users may use the Springer Nature journal content for small scale, personal non-commercial use, it is important to note that Users may
not:
use such content for the purpose of providing other users with access on a regular or large scale basis or as a means to circumvent access
control;
use such content where to do so would be considered a criminal or statutory offence in any jurisdiction, or gives rise to civil liability, or is
otherwise unlawful;
falsely or misleadingly imply or suggest endorsement, approval , sponsorship, or association unless explicitly agreed to by Springer Nature in
writing;
use bots or other automated methods to access the content or redirect messages
override any security feature or exclusionary protocol; or
share the content in order to create substitute for Springer Nature products or services or a systematic database of Springer Nature journal
content.
In line with the restriction against commercial use, Springer Nature does not permit the creation of a product or service that creates revenue,
royalties, rent or income from our content or its inclusion as part of a paid for service or for other commercial gain. Springer Nature journal
content cannot be used for inter-library loans and librarians may not upload Springer Nature journal content on a large scale into their, or any
other, institutional repository.
These terms of use are reviewed regularly and may be amended at any time. Springer Nature is not obligated to publish any information or
content on this website and may remove it or features or functionality at our sole discretion, at any time with or without notice. Springer Nature
may revoke this licence to you at any time and remove access to any copies of the Springer Nature journal content which have been saved.
To the fullest extent permitted by law, Springer Nature makes no warranties, representations or guarantees to Users, either express or implied
with respect to the Springer nature journal content and all parties disclaim and waive any implied warranties or warranties imposed by law,
including merchantability or fitness for any particular purpose.
Please note that these rights do not automatically extend to content, data or other material published by Springer Nature that may be licensed
from third parties.
If you would like to use or distribute our Springer Nature journal content to a wider audience or on a regular basis or in any other manner not
expressly permitted by these Terms, please contact Springer Nature at
onlineservice@springernature.com
Available via license: CC BY 4.0
Content may be subject to copyright.