Use of cognitive bias as a welfare tool in poultry

Abstract

In human psychology, the link between cognition and emotions is broadly accepted. However, the idea of using the interaction between cognition and emotions as a tool for a better understanding of animal emotions or for welfare assessment is relatively new. The first avian species used in cognitive bias tests was the European starling followed by the domestic chicken and other species. The most frequently used paradigm is the affect-induced judgment bias. There are many variations of the judgment bias tests in birds. The test itself is preceded by discrimination training. Discrimination tasks vary from visual cue discrimination, discrimination of time intervals to spatial location discrimination. During the discrimination training, birds flip or do not flip the lids of the food dishes, and their latency to approach the cues in a straight alley maze, in a two-choice arena, or different locations in spatial judgment task arena are measured. Alternately, the birds fulfill operant tasks in a Skinner box. Before or after the discrimination training phase, birds are subjected to manipulations that are hypothesized to induce positive or negative emotional states. In the last stage, birds are subjected to judgment bias tests. The assumption is that animals in a negative affective state would more likely respond to ambiguous cues, as if they predict the negative event, than animals in a more positive state. However, the results of some avian studies are inconsistent, particularly those studying the effect of environmental enrichment. In starlings, each of the three studies has supplied conflicting results. In poultry, none of the four studies demonstrated a positive effect of environmental enrichment on emotional states. Only the study using unpredictable stressors in combination with environmental complexity showed that animals kept in a more complex environment are more optimistic. Manipulation of the social environment seems to be more effective in judgment bias induction. Conflicting results could be attributable to the design of the tests, the manner of affect induction, or the data analysis. Further optimization and validation of avian cognitive bias tests could help to avoid problems such as the loss of ambiguity. New methods of attention and memory bias testing are promising. However, regardless of the abovementioned complications, a cognitive bias paradigm is a valuable tool, which can help us better understand avian emotions and assess poultry welfare.

Introduction

Research on animal emotions is mammal- and negative emotions-centric. Studies of emotions in birds are less frequent than those in mammals. However, increasing interest in research on avian emotions is evident from the Web of Science data. While the results of the topic search for avian or bird emotions between 1989 and 1998 yielded only 10 records between 1999 and 2008, and from 2009 to 2018, 44 and as many as 162 records were obtained (hits of the search for topic “[(avian AND emotion) OR (bird AND emotion)] OR (chicken AND emotion)”). From the perspective of methodology, the most frequent methods for the assessment of emotions in poultry are still fear tests, such as the open field test, tonic immobility test, or novel object test (Forkman et al., 2007). However, there are also innovative approaches, such as the use of the infrared thermography (Moe et al., 2012, 2018) or the assessment of facial expressions (Bertin et al., 2018a, 2018b). Measurement of the affect-induced cognitive bias, the subject of this review, is one such innovative method.

Even though the interactions between cognition and emotions are now widely accepted (Luo and Yu, 2015), the idea to use them as a tool for studying animal emotions or for welfare assessment is relatively new (Harding et al., 2004; Boissy et al., 2007; Mendl et al., 2009). After the publication of the first seminal study of cognitive bias in rats (Harding et al., 2004), many variations in the test design in many species followed (Gygax, 2014; Baciadonna and McElligott, 2015; Bethell, 2015; Roelofs et al., 2016). Affective states can influence different cognitive processes, including attention, memory, and judgment. Most of the studies on cognitive bias in animals, including birds, use judgment bias tests, evaluating decision-making under ambiguity. This review focuses on an overview of the studies of cognitive bias in birds and its potential use as a welfare assessment tool in poultry.

Avian Cognition

The question of avian cognition and intelligence is closely linked to the question of the evolution of an avian brain. Ludwig Edinger, the father of comparative neuroanatomy, combined in his theory of brain evolution Darwin’s concept of evolution with Aristotle’s scala naturae idea. Edinger viewed evolution as progressive and linear and mistakenly thought that the largest part of the avian telencephalon represents the palaeoencephalon (striatum), controlling instinctive behaviors, while only a small part of it represents the neoencephalon (pallium, cortex), controlling learned and intelligent behavior (Edinger, 1899, 1908). Only at the beginning of the millennium, an international consortium of avian neuroscientists codified the revision of avian brain nomenclature reflecting true homologies among avian and mammalian brains (Reiner et al., 2004a, 2004b; Jarvis et al., 2005). This new nomenclature reflects the fact that the avian telencephalon has a relatively large and well-developed pallium that performs functions similar to those of the mammalian cortex, although the avian pallium is nuclear, and the mammalian cortex is laminar (Clayton and Emery, 2015; Güntürkün and Bugnyar, 2016; Guy and Staiger, 2017). A new understanding of avian brain organization also corresponds to growing evidence for the complex cognitive abilities of birds (see e.g., Clayton and Emery, 2015; Güntürkün and Bugnyar, 2016 for the review), which have been demonstrated not only in corvids and parrots but also in poultry (Hazel et al., 2015; Freire and Hazel, 2017; Marino, 2017; Garnham and Løvlie, 2018).

Cognitive Bias in Animals

While affect and cognition were historically viewed as independent entities or even opposing forces, recently, they have been regarded as closely interconnected phenomena (Pessoa, 2008; Huntsinger and Schnall, 2013). Cognition has an impact on emotion (Schachter and Singer, 1962; Ochsner and Gross, 2005), while emotional content has an impact on cognitive functions (Mineka and Sutton, 1992; Dolan, 2002; Dolcos et al., 2014). Studies showing that nonpainful stimuli are perceived as painful when participants expect pain (Sawamoto et al., 2000) or voluntary regulation (suppression or enhancement) of emotional responses to unpleasant pictures (Jackson et al., 2000) are just a few examples of the effect of cognition on emotions.

Although the link between cognition and emotion has been extensively studied in humans, cognitive components of emotion are a mostly unexplored source of information about animal emotions (Mendl et al., 2009). Emotional modulation of cognitive processes seems to have adaptive value by helping, for example, a fearful or anxious individual to attend to, memorize, and make judgments about threatening circumstances or stimuli. We can presume that because of the value of such processes for most animal species, their evolution is favored by selective pressures (Paul et al., 2005). Cognitive outputs of emotion are the numerous information processing changes or biases that can be roughly grouped into three main categories: attention biases, memory biases, and judgment biases (Mineka and Sutton, 1992).

Harding et al. (2004) were the first to measure cognitive bias or, more precisely, judgment bias in animals. Rats were trained in a Go/NoGo operant discrimination task to respond by lever pressing to a tone associated with a positive event (food pellet) and to refrain from pressing the lever to avoid a negative event (white noise). Once the rats reached the discrimination criterion, they were allocated either to predictable or to unpredictable housing, inducing a depression-like state. The rationale behind the newly developed test was that the rats were exposed to nonreinforced tones that had intermediate frequencies between the positive and negative tones in the discrimination training to investigate the animals’ anticipation of positive or negative events. These researchers found that rats in unpredictable housing were slower to respond and tended to show fewer responses to ambiguous tones close to the positive tone and to this tone itself.

Negative cognitive bias, the tendency to interpret ambiguous situations pessimistically, is a typical feature of stress-related disorders such as anxiety or depression (Enkel et al., 2010). Studies using neuropsychological tests have shown negative emotional biases and disrupted reward and punishment processing in depressed patients, which may also influence nonaffective cognition. These affective biases are sensitive to antidepressant treatments (Robinson and Roiser, 2015). The main challenge of modeling human psychiatric disorders in animals is to relate symptoms assessed via subjective self-reports of patients to measurements in animals. However, judgment bias tasks seem to offer a translational model of affective biases in depression (Robinson and Roiser, 2015; Robinson, 2018). Several laboratories have focused on pharmacological validation of this model (Enkel et al., 2010; Anderson et al., 2013; Rygula et al., 2014a, 2014b; Hales et al., 2016; Drozd et al., 2019). Cognitive bias in rats represents a stable and enduring behavioral trait. Rats showing such trait pessimism are more prone to stress-induced anhedonia and stress-induced motivational deficits (Rygula et al., 2013; Rygula and Popik, 2016). Rats with congenital learned helplessness (genetic model of depression in rats) show pessimistic response bias in comparison to nonhelpless rats. Environmental enrichment results in more optimistic interpretations of ambiguous cues in both rat strains (Richter et al., 2012).

The paradigm first introduced by Harding et al. (2004) was later tested not only in laboratory rodents but also in many other species (see Gygax, 2014; Baciadonna and McElligott, 2015; Bethell, 2015; Roelofs et al., 2016 for the review).

Overview of the Judgment Bias Tests Used in Birds

Go/NoGo task based on the displacement of lids on food dishes

Discrimination training

The first paper on affect-induced judgment bias in birds was published by Bateson and Matheson (2007) and used wild-caught European starlings as subjects. The visual discrimination training was based on conditioned taste aversion. As claimed by the authors, such associations have the advantage of being acquired quickly, typically within a single trial, and are also more resistant to extinction than learning based on positive reinforcement. The birds were trained to flip the cardboard lid off the Petri dish mounted on a white ceramic tile to gain access to a mealworm hidden underneath (Figure 1a). Starlings were trained to associate white lids with reward (mealworm) and dark gray lids (80% gray) with punishment (unpalatable mealworm injected with quinine sulfate solution).

Figure 1.

Visual discrimination tasks used in starlings. (a) Go/NoGo task (Bateson and Matheson, 2007). The top two images represent negative (80% gray color lid covering a Petri dish with an unpalatable mealworm—punishment) and positive cues (white color lid covering a Petri dish with a palatable mealworm—reward) used in the discrimination training. The small bottom images represent cues used in judgment bias tests (negative 80% gray color lid; ambiguous 20%, 40%, and 60% gray color lids; and positive white color lid). (b) Go/Go task (Brilot et al., 2010). The top two images represent the cues used in the discrimination task; the background color (60% or 0% gray) indicates the type of reward (high or low value, three or one mealworms), and a symbol on the lid (green cross or red triangle) indicates the outcome (rewarded or nonrewarded). The small bottom images represent cues used in the judgment bias tests (high value with 60% gray background; ambiguous 45%, 30%, and 15% gray background; and low value 0% gray background; the symbols indicate the baited Petri dish).

Judgment bias test

The manipulation supposed to induce an affective state was followed by the judgment bias tests. The sessions comprised reinforced trials (white and 80% gray lids) and probe unreinforced trials (20%, 40%, and 60% gray lids) (Figure 1a).

Variations of the original test design

A similar design of the discrimination training and judgment bias tests was used by the same laboratory in two other studies (Bateson et al., 2015; Gott et al., 2019). In the discrimination training, Bateson et al. (2015) used two other shades of gray, 20% and 60% gray lids. The positive cue was rewarded by a mealworm, while the negative cue was punished by the quinine sulfate-injected mealworm. In the judgment bias test, the shades of gray were 20%, 30%, 40%, 50%, and 60%. Within each session, the birds received the positive reinforced trials, negative punished trials, and unreinforced trails with 20%, 30%, 40%, 50%, and 60% gray lids. In the discrimination training, Gott et al. (2019) used the positive cue rewarded by a mealworm, while the negative stimulus was not reinforced, that is, it was an empty dish. In the judgment bias test, each session consisted of reinforced positive trials and unreinforced near-positive, middle, near-negative, and negative trials.

Adaptation of the method for use in other species—Japanese quail

In our laboratory, we attempted to introduce this method using Japanese quail. Nevertheless, due to problems with the discrimination training (although the quails were showing signs of aversion, they continued to consume unpalatable mealworms), we decided to use another approach. Similar problems have, however, also been mentioned by Bateson et al. (2015). The authors stated that unlike in previous experiments, most birds continued to flip the negative lids and some continued to eat the unpalatable mealworms, even after experiencing them multiple times. These findings indicate that the initial assumption of Bateson and Matheson (2007) regarding the advantage of conditioned taste aversion due to the quick acquisition of the association has limited liability.

Adaptation of the method for use in other species—domestic chicken

Ross et al. (2019) adopted the method for laying hens. Inside the testing chamber (Figure 2a), hens were exposed to a plastic food dish covered by a disk-shaped plastic lid. The hens were trained to discriminate between the two cues: a white lid which, if displaced (lids in this test were designed to be displaced by the experimenter, not by the birds themselves), exposed a mealworm located underneath and an 80% gray lid which, if displaced, resulted in punishment via an air puff to the face. In the judgment bias test, the hens were presented with reinforced positive and negative cues and unreinforced ambiguous cues (20%, 40%, and 60% gray).

Figure 2.

Visual discrimination tasks used in poultry. (a) Testing arena (top image) with an opening for insertion of the food dish with the color lid hiding the potential reward and a hose to the air compressor (punishment) above it; cues (bottom images) used in judgment bias tests (negative 80% gray lid associated with punishment; ambiguous 60%, 40%, and 20% gray lids, and positive 0% gray lid associated with reward) (Ross et al., 2019). (b) Two-choice (left-right) arena (top image) and cues used in the judgment bias tests (bottom images). The background color indicates the value (black background—high-value reward, white background—low-value reward), and the symbols indicate the baited feeder (Hernandez et al., 2015; de Haas et al., 2017a, 2017b). (c) Straight alley arena (top image) attached to the conspecifics area separated by plexiglass; cues (bottom images) used during the judgment bias tests (naturally aversive owl silhouette 0c:100o; ambiguous morphed silhouettes 25c:75o, 50c:50o, 75c:25o; and naturally appetitive chick silhouette 100c:0o) (Salmeto et al., 2011; Hymel and Sufka, 2012). (d) Straight alley arena (top image) with a left-right choice of negative (white or black color sign and an empty bowl) or positive (black or white color sign and a bowl containing a reward) cues presented simultaneously during the discrimination training; in the judgment bias tests, the color cues (bottom images) were presented sequentially in the middle of the arena (the partition had been removed)—a negative white or black color sign and empty bowl; ambiguous 25%, 50%, and 75% gray signs and matching color unrewarded bowl; and positive black or white color sign and bowl containing reward (Zidar et al., 2018). All the arenas are presented on the same scale, with the exception of the arena in Figure 2b, which is downscaled to 50%.

Go/Go two-choice visual conditional discrimination task

Discrimination training

This method is also based on flipping off the lids of Petri dishes, and the experimental subjects were again starlings (Brilot et al., 2010). In this task, the background color of the tile (60% or 0% gray) indicated the type of reward (high- or low-value reward, i.e., three or one mealworm), and a symbol on the lid (green cross or red triangle) indicated the outcome (rewarded or unrewarded) (Figure 1b). Half of the birds were trained to associate the red triangle on a dark background with the three mealworms and the green cross on a dark background with no reward in contrast to the red triangle on a white background with no reward and the green cross on a white background with one mealworm. The other half of the birds were trained to the reverse assignment. In each trial, two Petri dishes were presented simultaneously with the same background color and two different symbols on the lid.

Judgment bias test

The test of cognitive bias involved presenting intermediate backgrounds (15%, 30%, and 45% gray) and recording which symbol was selected by the bird (Figure 1b). A choice was recorded as indicative of either an optimistic or a pessimistic bias. In addition, the latency of the decision-making was also recorded.

Adaptation of the method for use in other species—domestic chicken

The experimental design of Brilot et al. (2010) was adopted for use in domestic chicken by Hernandez et al. (2015) and de Haas et al. (2017a, 2017b). The training and testing of hens were performed in a two-choice (left–right) arena (Figure 2b). The active choice was scored as correct (choosing the rewarded side) or incorrect (choosing the unrewarded side). Hens were trained to differentiate the combination of symbols and background of the food container lids for their choice (Figure 2b). For all hens, 100% gray indicated a high-value and 5% gray background indicated a low-value reward. The high-value reward consisted either of four mealworms (Hernandez et al., 2015) or five mealworms (de Haas et al., 2017a, 2017b), and the low-value reward consisted of one mealworm. Two lids with the same background color were presented at any given moment, and only the symbol indicated which food container was baited.

Judgment bias test

This test consisted of trials alternating between high- and low-rewarded trials and three novel ambiguous cue cards with intermediate background colors (75%, 50%, and 25% gray) that were not rewarded (Figure 2b).

Straight alley maze with naturally appetitive and aversive cues

Salmeto et al. (2011) and Hymel and Sufka (2012), in their affect-induced judgment bias tests in young chicks, used ecologically relevant stimuli as the cues. The potential advantage of such stimuli is the reduced requirement for extensive associative training prior to the cognitive bias tests. These researchers utilized silhouettes of a conspecific chick (or mirror) as the appetitive cue, an owl as the aversive cue, and three intermediate ambiguous cues with varying degrees of chick and owl characteristics obtained by morphing as ambiguous cues. To measure approach/avoidant responses, the authors used a straight alley maze (Figure 2b), a paradigm that is commonly used to quantify chick social reinstatement. Latency (s), that is, the time interval between crossing the start line after being released and the goal line, was recorded as a response to ambiguous cues.

Straight alley task with shades of gray as visual cues

Discrimination training

Another design of the judgment bias test based on a straight alley maze was introduced by Zidar et al. (2018) for testing 2 and 5-wk-old chicks. Chicks were individually trained to associate a color cue (black or white) with a reward (one-third of a mealworm). Color signs and similarly colored bowls were placed on both sides of a small partition (Figure 2d).

In the “judgment bias test,” color cues were presented sequentially to the chicks in the middle of the arena (the partition was removed, Figure 2d). In addition to the previously rewarded and unrewarded color cues were introduced three ambiguous color cues, that is, 25%, 50%, and 75% gray. Latency (s) until the chick had approached the color cue was recorded as a response to these ambiguous cues.

Spatial judgment task

The spatial judgment task in which animals are trained to expect food in one location and not another was introduced by Burman et al. (2008). Enriched rats ran faster than unenriched ones to the probe located nearest the unrewarded location, suggesting that they were more likely to anticipate a reward at that specific ambiguous location than the unenriched rats. As expected by the authors, this method has the potential to be adaptable to other animal species. It has actually been adopted for many species, including birds and poultry.

Lindström (2010) and Wichman et al. (2012) were the first to adopt this approach for laying hens and used an identical arena and setup (Figure 3e).

Figure 3.

Spatial judgment task arenas used for (a) broilers (Iyasere et al., 2017), (b) young chickens (Seehuus et al., 2013), (c) canaries (Lalot et al., 2017), (d) Japanese quail (Horváth et al., 2014, 2015), and (e) laying hens (Lindström, 2010; Wichman et al., 2012). The principle used for all the studies was the same. During discrimination training, the birds were trained to discriminate the positive location associated with reward and the negative location associated with punishment (far left or right location balanced between birds). In the next phase, the spatial judgment tests, the three ambiguous locations (near negative, middle, near positive) were also presented. All five locations are shown in the images, but the food dish is only presented in one location during each trial. Of note, Iyasere et al. (2017) used not only the location but also the color of the paper cone as a cue (hiding reward, or punishment—hose from the compressor delivering the air puff, or nonreinforced). The arena of Seehuus et al. (2013) was attached to a pen designed to accommodate the stages of the feed reward cycle (see text for details).

Discrimination training

Hens were trained to associate the position of a food bowl (left or right corner of the testing arena) with a reward (a few pieces of corn), while the position of a bowl in the opposite corner was unrewarded (empty bowl). Half of the birds were rewarded at the left location of the bowl, and the other half were rewarded at the right location. The criterion for completing the discrimination training was that the latency to reach the food bowl in the unrewarded location was at least 5 s longer than when it was located in the rewarded position.

Judgment bias test

Each test series consisted of a food bowl located in the rewarded location, unrewarded location, and three probe trial locations (ambiguous cues) placed between the rewarded and unrewarded ones (Figure 3e). The latency from leaving the start box to starting to peck at the bowl was assessed.

The spatial judgment task is the most frequently used method employed for judgment bias testing in birds (Lindström, 2010; Wichman et al., 2012; Seehuus et al., 2013; Horváth et al., 2014, 2015; Iyasere et al., 2017; Lalot et al., 2017; Adriaense et al., 2019). The size (depending on the size of animals, i.e., breed and age) and shape of the spatial judgment task arenas vary (Figure 3), with canaries and ravens using cages instead of the arena (Figure 3c). Differences are also observed in the reward and punishment used (see Table 1). Most studies used solely the location of the feed dish as a cue (Lindström, 2010; Wichman et al., 2012; Seehuus et al., 2013; Horváth et al., 2014, 2015; Lalot et al., 2017; Adriaense et al., 2019); nevertheless, Iyasere et al. (2017) used not only the location but also the color cue, since the Petri dishes in the different locations also differed by the color of the overlying paper cone (shades of gray).

Table 1.

Comparison of judgment bias test methods used in birds

			Discrimination training									Judgment bias test
	Species	Initial N/N tested	Method	Positive cue	Reward/Large reward	Negative cue	Punishment/Small reward	Training length	Proportion of reaching criterion	Affect induction treatment	Additional factors	Cues	Proportion of unreinforced cues
Bateson and Matheson (2007)	Starling	6/6	Go/NoGo (Figure 1a)	White (0% gray) lid on Petri dish	Palatable mealworm	Dark gray (80% gray) lid on Petri dish	Unpalatable mealworm	n.a.	100.0%	Standard vs. enriched cages	n.a.	0%, 20%, 40%, 60%, and 80% gray lids on Petri dish	71.4% (20/28)
Bateson et al. (2015)	Starling	31/30	Go/NoGo (Figure 1a)	20% or 60% gray lid on Petri dish	Palatable mealworm	60% or 20% gray lid on Petri dish	Unpalatable mealworm	2.03 sessions (16 trials/session)	96.8%	High vs. low early-life competition	High vs. low telomere attrition	20%, 30%, 40%, 50%, and 60% gray lids on Petri dish	55.6% (10/18)
Gott et al. (2019)	Starling	31/31	Go/NoGo (Figure 1a)	20% or 60% gray lid on Petri dish	Mealworm	60% or 20% gray lid on Petri dish	None	3.84 sessions (16 trials/session)	100.0%	High vs. low early-life competition	High vs. low telomere attrition	20%, 30%, 40%, 50%, and 60% gray lids	66.7% (12/18)
Ross et al. (2019)	Chicken	24/23	Go/NoGo (Figure 2a)	White (0 % gray) lid on plastic dish	Mealworm	80% gray lid on plastic dish	Air-puff	19 sessions (20 trials/session)	95.8%	Standard vs. enriched cages	Exploratory vs. nonexploratory individuals	0%, 20%, 40%, 60%, and 80% gray lids	23.1% (6/26)
Brilot et al. (2010)	Starling	8/8	Go/Go (Figure 1b)	Green cross on 60% gray background on Petri dish	3 mealworms	Red triangle on 0% gray background on Petri dish	1 mealworm	13.25 sessions (12 trials/session), i.e., 18.75 trials	100.0%	Standard vs. enriched cages	Stereotyping vs. nonstereotyping individuals	0%, 15%, 30%, 45%, and 60% gray lids with green cross or red triangle	33.3% (5/15)
Hernandez et al. (2015)	Chicken	30/20	Go/Go (Figure 2b)	Green triangle or red cross on a black background on feeding box	4 mealworms	Red cross or green triangle on a white background on feeding box	Food pellets or 1 mealworm	n.a.	66.7%	Nonstressed vs. stressed (5-min isolation)	n.a.	Black, 75%, 50%, and 25% gray, white background with green triangle, and red cross on feeding box	25.0% (3/12)
de Haas et al. (2017a)	Chicken	20/8	Go/Go (Figure 2b)	Green triangle or red cross on black background	5 mealworms	Red cross or green triangle on white background	1 mealworm	14 sessions (10 trials/session)	40.0%	n.a.	n a.	n.a.	n.a.
de Haas et al. (2017b)	Chicken	24/18	Go/Go (Figure 2b)	Green triangle or red cross on black background	5 mealworms	Red cross or green triangle on white background	1 mealworm	28 sessions (10 trials/session)	75.0%	Fearful vs. Nonfearful	n.a.	Green triangle or red cross on 0%, 25%, 50%, 75%, and 100% black background	33.3% (3/9)
Salmeto et al. (2011)	Chicken	n.a.	Go/NoGo (Figure 2c)	Chick image (c)	Naturally appetitive	Owl silhouette (o)	Naturally aversive	n.a.	n.a.	Nonstressed, 5-min isolation (anxiety-like state), 60-min isolation (depression-like state)	n.a.	Morphed images 100c:0o, 75c:25o, 50c:50o, 25c:75o, 0c:100o	n.a.
Hymel and Sufka (2012)	Chicken	n.a.	Go/NoGo (Figure 2c)	Chick image (c)	Naturally appetitive	Owl silhouette (o)	Naturally aversive	n.a.	n.a.	Treatment of 5-min isolated chicks (anxiety-like state) with anxiolytic clonidine and 60-min isolated chicks (depression-like state) with antidepressant imipramine	n.a.	Morphed images 100c:0o, 75c:25o, 50c:50o, 25c:75o, 0c:100o	n.a.
Zidar et al. (2018)	Chicken	96/87 (96/65)	Go/NoGo (Figure 2d)	Black or white sign and bowl	1/3 of a mealworm	White or black sign and bowl	None	n.a.	90.6%	Complex vs. simple environment, cold stress vs. no cold stress, unpredictable stressors	n.a.	Black, 75%, 50%, and 25% gray, white sign and bowl	n.a.
Lindström (2010)	Chicken	13/13	Go/NoGo (Figure 3e)	Positive bowl location	A few pieces of corn	Negative bowl location	None	n.a.	100.0%	Former battery hens 2 m vs. 4 mo after leaving the battery cages	n.a.	Rewarded, near-rewarded, middle, near-unrewarded, and unrewarded location of bowl	23.1% (3/13)
Wichman et al. (2012)	Chicken	27/17	Go/NoGo (Figure 3e)	Positive bowl location	A few pieces of corn	Negative bowl location	None	n.a.	88.9%	Basic vs. enriched pens	n.a.	Rewarded, near-rewarded, middle, near-unrewarded, and unrewarded location of bowl	23.1% (3/13)
Seehuus et al. (2013)	Chicken	25/24	Go/NoGo (Figure 3b)	Positive bowl location	Mealworm	Negative bowl location	Puffed rice soaked in a 2 % quinine sulfate solution	n.a.	96.0%	Denying access to parts of pen representing food reward cycle (litter area “appetitive,” feed area “consummatory,” perches, and dark area “postconsummatory”)	n.a.	Rewarded, near-rewarded, middle, near-unrewarded, unrewarded location of bowl	27.3% to 23.1% (3/11–3/13)
Iyasere et al. (2017)	Chicken	42/14	Go/NoGo (Figure 3a)	Positive location of Petri dish containing white paper cone	Mealworm	Negative location of Petri dish containing black paper cone	Air-puff	8 sessions (6 trials/session)	33.6%	Control vs. corticosterone treated	n.a.	Rewarded, near-rewarded, middle, near-unrewarded, and unrewarded location of Petri dish	33.3% (3/9)
Horváth et al. (2014)	Japanese quail	24/24	Go/NoGo (Figure 3d)	Positive feeder location	Mealworm	Negative feeder location	White noise	11 sessions (4 to 5 trials/session)	n.a.	Quails reared with vs. without adoptive mothers	n.a.	Rewarded, near-rewarded, middle, near-punished, and punished location of feeder	42.8% (3/7)
Horváth et al. (2015)	Japanese quail	24/20	Go/NoGo (Figure 3d)	Positive feeder location	Mealworm	Negative feeder location	White noise	19 sessions (4 to 5 trials/session)	n.a.	Application of dopamine D1 antagonist SCH 23390, D2 antagonist haloperidol, D1 agonist SKF 38393, D2 agonist bromocriptine vs. saline	n.a.	Rewarded, near-rewarded, middle, near-punished, punished location of feeder	42.8% (3/7)
Lalot et al. (2017)	Canary	43/31	Go/NoGo (Figure 3c)	Positive feeder location	Attractive dish	Negative feeder location	Shaking of cardboard above the bird	32 sessions (4 trials/session),	72.1%	Housed in pairs vs. housed singly	Personality traits	Rewarded, near-rewarded, middle, near-punished, and punished location of feeder	100.0% (4/4)
Adriaense et al. (2019)	Raven	8/8	Go/NoGo (Figure 3c)	Positive feeder location	n.a.	Negative feeder location	None	140 trials	100.0%	Positive vs. negative affective state in demonstrator	n.a.	Rewarded, middle, and punished location of feeder	n. a.
Matheson et al. (2008)	Starling	8/6	Go/Go (Figure 4d)	Instant trial (2 s)	Custom starling pellet	Delayed trial (10 s)	30 s time-out	15.83 sessions (54 trials/session)	75.0%	Enriched vs. standard cages	n.a.	Time intervals 2, 3, 4, 5, 6, 7, 8, 9, or 10 s	66.7% (36/54)
Deakin et al. (2016)	Chicken	8/5	Go/NoGo (Figure 4a)	High saturation orange circle on the screen	Mealworm	Low saturation orange circle on the screen	Air-puff	8.2 sessions (40 trials/session)	62.5%	Temperature during testing ~20 °C vs. ~29 °C	n.a.	Orange circle with 50 (low), 100, 150, 200, 250 (high) saturation in MS PowerPoint HSL scale	15.0% (6/40)
Horváth et al. (2016)	Japanese quail	54/45	Go/NoGo (Figure 4c)	0% or 80% gray circle on touchscreen	Mealworm	80% or 0% gray circle on touchscreen	White noise	7–9 sessions (60 trials/session)	83.3%	Cages vs. deep litter pens	n.a.	0%, 20%, 40%, 60%, and 80% gray circles	60.0% (36/60)
Pichová et al. (2016)	Chicken	20/16	Go/NoGo (Figure 4b)	0% or 80% gray circle on touchscreen	Mealworm	80% or 0% gray circle on touchscreen	White noise	9 to 10 sessions (60 trials/session)	80.0%	Enriched cages vs. deep litter pens	n.a.	0%, 20%, 40%, 60%, and 80% gray circles	60.0% (36/60)

Go/Go operant discrimination task: instant vs. delayed food

Discrimination training

This method (Matheson et al., 2008) was the first judgment bias method based on operant conditioning in birds, as was the seminal study published by Harding et al. (2004). However, the design of this method differed significantly, since it was based on temporal discrimination, the discrimination of two time intervals (2 and 10 s), which subsequently served as the positive and negative stimuli. Starlings were trained and tested in the custom Skinner box with three horizontal pecking keys, and a reward (food pellets) was delivered via an external dispenser to a central food hopper (Figure 4d). A trial was started with the center key flashing amber (left panel in Figure 4d). After the bird pecked once at the flashing key, the trial was initiated. The center key stopped flashing and was illuminated continuously for 2 or 10 s before it switched off. The two side keys then started flashing, one red and the other green (right panel in Figure 4d). For each bird, one color was designated as the correct response if the initial interval was 2 s and the other color if it was 10 s. The assignment of colors to intervals was counterbalanced across birds. After the correct response, the chosen key stopped flashing and remained lit for 2 s, after which time the first peck to the key resulted in immediate reinforcement with one pellet. After an incorrect response, both keys switched off followed by a 30-s timeout.

Figure 4.

Operant conditioning tasks. A screen peck task apparatus for laying hens (a) using manual delivery of reward (mealworm through funnel) and punishment (air-puff using a canister) (Deakin et al., 2016), (b) touchscreen Skinner box for laying hens (Pichová and Košťál, 2016), and (c) touchscreen Skinner box for Japanese quail (Horváth et al., 2016). In all three experiments, birds were trained to discriminate the visual cues (orange circles with high or low saturation, dark gray and white circles). Pecking at the color cue was associated with reward (mealworms or granulated food) or punishment (air-puff, white noise). The bottom images in Figure 4a–c represent cues (including three ambiguous ones) used in the judgment bias tests. In each trial, only one color cue was presented. (d) Three horizontal pecking keys of the custom Skinner box used by Matheson et al. (2008) in starlings. During the temporal discrimination, training was initiated a trial with the center key flashing amber (left image). After the bird pecked once at the flashing key, the center key stopped flashing and was illuminated continuously for 2 or 10 s before it switched off and the two side keys started flashing, one red and the other green (right image). For each bird, one color was designated as the correct response if the initial interval was 2 s, and the other color if it was 10 s (see text for details).

Judgment bias test

In the second stage, the delay to reward following a correct response was changed to create differentially valued outcomes. For each bird, correctly discriminating one stimulus (either 2 or 10 s) was followed by a 1-s delay to reward (the instant outcome) and the other by a 15-s delay to reward (the delayed outcome). Second, the probability of reinforcement was reduced by the introduction of unreinforced probe trials in which the initial stimulus had a duration of 2, 3, 4, 5, 6, 7, 8, 9, or 10 s. In a probe trial, the center key was illuminated for the designated interval, the expiration of which was followed by the choice of red and green as described above.

Screen peck task

Although not indicated by the name under which the study was published, the screen peck task described by Deakin et al. (2016) is in a principle manually operated Skinner box; therefore, this method belongs to the group of operant conditioning methods. The authors mention in their paper the potential to completely automate the task using a computerized Skinner box touchscreen apparatus.

Discrimination training

The birds were trained to peck at a high- or low-saturation orange circle (saturation of 250 and 50 on the MS PowerPoint hue, saturation, luminosity [HSL] scale) presented on a computer screen (Figure 4a). For half of the birds, the positive cue was a highly saturated orange circle, and for the other half, it was a circle with low saturation. Pecking at the positive cue was rewarded by a mealworm, while pecking at the negative cue was punished with a 1-s air-puff. Reward and punishment were delivered manually by the experimenter.

Judgment bias test

In this phase, three unrewarded ambiguous cues (saturation of 200, 150, and 100 on the MS PowerPoint HSL scale) were interspersed between the negative and positive cues. The responses to stimuli and latencies to peck them were recorded and analyzed from videos.

Go/NoGo operant discrimination task

The Go/NoGo operant discrimination task was used in a series of experiments in Japanese quail (Horváth et al., 2016) and laying hens (Pichová and Košťál, 2016) in our laboratory. Visual cues used in the discrimination training and judgment bias test (circles in shades of gray) were taken from the original work performed in starlings (Bateson and Matheson, 2007). Cues were presented on a touchscreen monitor of the Skinner box. The Skinner boxes were custom-made operant conditioning chambers with scaled dimensions (including dimensions of cues) for the tested species (Figure 4b and c). The cue presentation, peck responses, and reward and punishment delivery were fully automated and controlled by the Biopsychology-Toolbox (Rose et al., 2008).

Discrimination training

In both Japanese quail and laying hens, the basis of training was a Go/NoGo operant discrimination task. Animals were trained to respond by pecking in response to a positive cue associated with a reward (standard granulated food in quail, mealworm in hens) and to refrain from pecking in response to a negative cue to avoid punishment (white noise). To avoid a possible preference for one color, for half the birds, the positive cue was a white circle, and for the other half, it was an 80% gray circle. The negative cue was the opposite shade of gray.

Judgment bias test

In addition to positive and negative cues, the animals were also exposed to three ambiguous cues (20%, 40%, and 60% gray). The response to positive and negative cues remained reinforced (rewarded, punished) while pecking at ambiguous cues was not reinforced.

Some general remarks regarding the methods

The first stage of every judgment bias test represents the discrimination training, in which animals are trained to associate one stimulus with a positive event and another one with a negative event (or less positive event). This stage usually requires extensive training, which is the drawback of the judgment bias method. Training may have a character of a Go/NoGo or Go/Go task. While the Go/NoGo approach is more frequent, some authors argue that it is not possible to distinguish between No-Go as a response indication and as a response omission (Richter et al., 2012). The Go/Go approach eliminates this problem by requiring an active response to both positive and negative stimuli. Although it is not possible to trace the details of the training success in all the avian studies described above, it seems that birds were more successful in reaching the discrimination criterion in Go/NoGo tasks (Table 1). Lower performance in Go/Go tasks may be caused by the more complex design of tests, such as the combination of background colors and symbols on lids (Brilot et al., 2010; Hernandez et al., 2015; de Haas et al., 2017a, 2017b), or the discrimination of time intervals (Matheson et al., 2008). Of course, a lower number of studies using the Go/Go tasks make it difficult to compare the results with more numerous studies using the Go/NoGo approach.

Two major groups of cues are used in avian judgment bias tests: color or other visual cues (visual discrimination) and spatial location cues (spatial discrimination) (Table 1). High training success is achieved using both cue types, thanks to the biological relevance of both stimuli categories. Vision is the most important sense for birds (Hodos, 2012), and the spatial cognition of birds is also well developed (Reichert et al., 2017). However, the side preference of individuals (side bias) represents a potential obstacle for using tests involving spatial discrimination. Although the side of the reward was randomized, 12 out of 20 hens in the study by de Haas et al. (2017a) did not meet the learning criterion because of the strong left-side preference. In a subsequent study (de Haas et al., 2017b), 75% of the tested animals passed the training stage, but more training sessions were needed to unlearn the side bias.

There are several examples of attempts to reduce the length of discrimination training. For example, Bateson and Matheson (2007) proposed the use of conditioned taste aversion. However, expectations are not always accomplished, as in the case of starlings that continued to eat the unpalatable mealworms even after experiencing them multiple times (see Bateson et al., 2015). Another interesting alternative was applied in studies by Salmeto et al. (2011) and Hymel and Sufka (2012), who used naturally aversive and appetitive visual cues. Unfortunately, none of those studies using morphed silhouettes between chicken and owl provided data about the length of training.

Problems with the number of animals reaching the discrimination criterion demonstrate the importance of using the type of task and stimuli, as well as the type and intensity of reward and punishment, taking into account the given species and its ethology, to maximize the number of individuals that proceed through the discrimination training to the final testing.

Methods of Affect Induction in Avian Cognitive Bias Tests

Environmental enrichment, housing conditions, and housing complexity

As previously mentioned, the cognitive bias paradigm has been applied in many species, including birds. The largest proportion of avian judgment bias experiments used a manipulation of the environment and its complexity for the induction of affective states.

That was also the case of the first experiment testing cognitive bias in birds (Bateson and Matheson, 2007). The authors studied the effect of two environments, standard cages vs. enriched cages, in European starlings (Figure 1a) using a Go/NoGo visual discrimination task based on lid flipping and on conditioned taste aversion. The hypothesis of these researchers was that enriched cages, cages of the same size and shape as the standard ones but containing natural branches of different thickness placed at different heights and angles, a water bath, and a plastic tray of bark chippings allowing foraging behavior would induce a positive affective state that would be reflected by optimistic responses to ambiguous cues. All birds (n = 6) were housed in standard cages during the training. Two days before testing, the birds were transferred into new cages (standard or enriched), allowed them to adapt to the new environment for 2 d and on the following 5 d subjected to judgment bias tests. After the first series of tests, the environment of the birds was exchanged between the groups, and they were tested again under the same time schedule as in the first series of tests. The results showed that the proportion of birds that flipped lids with ambiguous cues was lower if they were housed in standard than in enriched cages but only in the case in which the birds were moved from enriched to standard cages. These results suggest that rather than environmental quality, its recent decline causes a pessimistic bias of starling judgment.

The next experiment from the same laboratory (Matheson et al., 2008) also tested the effect of environmental enrichment on judgment bias in starlings, but this time using a method based on Go/Go operant temporal discrimination (Figure 4d). Moreover, the differences between enriched and unenriched cages were greater than in the first study. Standard cages were half the size of the enriched ones, and access to a water bath and cleaning time (sometimes with birds present) were unpredictable. The daily ration of mealworms in unenriched cages was placed in the bowl with other food. Each of the tested starlings (n = 7) experienced both cage types, with each manipulation lasting 14 d. The probability of classifying ambiguous cues as positive was significantly higher in birds from the enriched cages compared with the standard cages. These conclusions were based on fitting the judgment bias data using the four parameters sigmoid function, since more than the usual three ambiguous probes were used in this experiment (seven intervals between 2 and 10 s with a 1-s increment). Unexpectedly, the performance of the birds differed depending on whether the 2-s time stimulus was associated with a delayed or instant reward. In the group in which the short stimulus was associated with an instant reward, the performance of the birds in the temporal discrimination was considerably less successful. Since the cage manipulation resulted in pessimistic bias in standard cage birds at the negative end of the ambiguous stimulus spectrum, the authors suggested that starlings from standard cages minimized the probability of a negative event.

However, another study from the same laboratory studying the effect of environmental enrichment on starlings (Brilot et al., 2010) used a sequence of environmental changes over the course of 3 wk. In contrast, for the first and third weeks, the birds were subjected to environmentally enriched conditions; these enrichments were removed during the second week (nonenriched conditions). The enriched conditions were similar to those used by Bateson and Matheson (2007). The authors expected a more pessimistic judgment bias, as measured by the Go/Go visual conditional discrimination task based on lid flipping (Figure 1b), under the nonenriched conditions with a more pronounced response to the loss of enrichment. However, the cognitive bias tests did not reveal any changes in proportion or latency of the responses to ambiguous cues. Nevertheless, since part of the subjects developed stereotyped behavior, the authors compared the judgment bias tests of starlings during the first week of environmental manipulation between the stereotyping and nonstereotyping individuals. These researchers found that stereotyping starlings showed significantly more pessimistic responses to the middle ambiguous cue in comparison to nonstereotyping individuals. Since stereotypic behavior is considered an indicator of poor welfare, this result can indicate that poorer welfare is associated with more negative affective states.

Concern about animal welfare is significantly affecting poultry production, which is why the cognitive bias paradigm stimulated a number of research groups to study poultry behavior to test the possibility of using this approach to evaluate the effects of housing conditions on affective states of poultry and their valence.

Lindström (2010) investigated the hypothesis that the environmental change of moving former battery cage hens to litter pens would lead to an optimistic bias. Hens were tested using a spatial judgment task (Figure 3e) 2 and 4 mo after the change of housing conditions. In contrast to the author’s expectations, hens showed longer latencies to reach the middle cue (were more pessimistic) after the longer stay in litter pens, that is, 4 mo after the change of environment, as compared with 2 mo after the change. Since enrichment involves environmental complexity, as well as novelty, it is not clear whether the animals housed in enriched environment remained in the state of positive affect or became used to enrichment after some time (Richter et al., 2012). A similar phenomenon might have also occurred herein, in which an improved environment was stronger shortly after leaving the cages and diminished with time.

Using the same method (Figure 3e), Wichman et al. (2012) studied the effect of environmental enrichment on judgment bias in laying hens, comparing basic vs. enriched pens. Three days before the testing, the environment of a portion of the hens was boosted by the addition of extra perches, nest boxes, wood shavings, and apples or sunflower seeds in the litter. The other group of hens was housed in the basic pen environment, where only the low perch remained, with the minimum amount of wood shavings. However, the differences in judgment of ambiguous cues between environments were not significant. A potential interpretation of these results could be that the extra food, provided as part of the enrichment, caused a reduced motivation of the birds from this environment to work for the food reward in the test.

Ross et al. (2019) housed hens in pens with or without enrichments that are assumed to improve mood and then subjected them to the judgment bias test using visual discrimination based on lid flipping (Figure 2a). Nonenriched pen contained a nest box, a perch, and softwood shavings. Enriched pen provided a larger area bedded with softwood shavings with added features: perches and platforms at various heights, foraging opportunities, and an assortment of enrichments including sand and peat for dustbathing. However, the enriched housing did not consistently cause the hens to respond more “optimistically” in the judgment bias tests.

The work of Horváth et al. (2016), using a visual discrimination task in a touchscreen operant chamber (Figure 4c), provides results for four independent experiments comparing the effects of housing (cage vs. deep litter pen) on judgment bias in Japanese quail. Inspired by the experiments in starlings, birds were exposed to a series of short-term housing changes. After 2 or 3 wk, the housing conditions were changed. Of the 11 judgment bias tests described in the paper, only one demonstrated an optimistic response of quail housed in an enriched environment. The other remaining tests did not show significant differences between the treatments, with the exception of the two test, in which a decline was observed in the response of the enriched group to reward (positive or near-positive stimulus), probably caused by the increased opportunity for foraging in the litter in the enriched group, as well as by the low value of the reward (standard granulated feed).

Similar results were obtained by Pichová and Košťál (2016) using a visual discrimination task in a touchscreen operant chamber (Figure 4b) in laying hens. To avoid problems with reward value, as observed in quail (see above), a custom mealworm dispenser was developed, enabling the computer-driven delivery of mealworms as a preferred food reward for laying hens (Bruce et al., 2003). This work contained two independent experiments. In the first experiment, a series of short-term housing changes (enriched cages vs. deep litter pens) was used, similar to those used by Horváth et al. (2016) in quail. However, of the five judgment bias tests, only one resulted in an optimistic response of laying hens to enriched deep litter housing, one in a pessimistic response (after two changes of the environment), and the other three to no significant differences between housing conditions. In the second experiment, one group of hens was kept in enriched cages, and the second was maintained in deep litter pens, throughout the experiment. Two judgment bias tests conducted with an interim of more than 5 mo did not show significant differences between hens housed in enriched cages and deep litter pens. This study also validated the loss of ambiguity, that is, the decline of responses to unreinforced stimuli (near-positive, near-negative, and middle cues) with repeated testing, a phenomenon described for the first time in sheep by Doyle et al. (2010).

Zidar et al. (2018) contributed to this topic with interesting data. They assigned laying hens immediately after hatching into four treatment groups in a two-by-two setup. Half of the chicks were raised in more complex environmental conditions, whereas the other half were raised in simpler environmental conditions. Two days posthatching, half of the chicks from complex and simple pens were exposed to cold stress over a period of 6 h (18 to 20 °C), while the remaining ones were maintained at an optimal temperature (33 °C). At 4 wk of age, all chicks (both cold stressed and those not stressed) were subjected to a battery of unpredictable stressors. They were exposed to a judgment bias test at 2 wk of age (i.e., before exposure to unpredictable stressors) and to a second judgment bias test at 5 wk of age (i.e., after exposure to unpredictable stressors). The judgment bias test was based on a straight alley with color discrimination (Figure 2d). The cold stress treatment in young chicks did not affect optimism in chicks before or after exposure to unpredictable stressors. Environmental complexity did not affect the latency to reach the color cues before exposure to unpredictable stressors. However, after exposure to unpredictable stressors, there was an optimistic bias in individuals housed in complex conditions. Individuals from the simple environmental conditions showed a significantly prolonged response latency to negative and near-negative cues, suggesting that environmental complexity may reduce the stress-induced reaction by maintaining an optimistic bias.

Although cold stress during early life of chicks in the previous study did not affect their performance in a judgment bias test at 2 or 5 wk of age, the exposure of adult laying hens shortly before judgment bias testing to higher (29 °C) compared to low ambient temperature (20.5 °C) had an effect. Deakin et al. (2016) designed their experiment based on the clear temperature preference of domestic fowl. These researchers hypothesized that exposure to the preferred higher ambient temperature would induce a positive effect. The researchers used a screen peck task to test this hypothesis (Figure 4a). While temperature treatment did not affect the percentage of probe cues pecked, an increased ambient temperature significantly decreased the latency to peck the middle cues.

A different approach to study the effects of environment was used by Seehuus et al. (2013). In the experiment by these researchers, chicks from the day of hatching until 9 wk of age were kept in a special pen (Figure 3b). The chicks were denied access to parts of the pen designed to accommodate the stages of what they called the feed-reward cycle (litter area “appetitive,” feed area “consummatory,” perches and dark area “postconsummatory”) for 4 d by closing off this third of the pen. Judgment bias tests were based on the spatial judgment task (Figure 3b). The only difference between the treatments was the shorter running time to reach the near-negative probe for the closed-litter treatment compared with the two other treatments. This result suggests that the chicks may have been less negatively affected by closing off the litter area than by shutting off the other areas, that is, disturbing the appetitive stage of the feed-reward cycle affected the chicks less than disturbing the consummatory or postconsummatory stages.

Social environment and judgment bias in birds

Our knowledge of the social aspects of animal emotions is still limited (Špinka, 2012). Panksepp et al. (1991) described social isolation as a simple ethological model of depression in young chicks. Salmeto et al. (2011) used this model to induce an anxiety-like state by 5 min of isolation and a depression-like state by 60 min of isolation prior to judgment bias testing. Their judgment bias test evaluating the valence of the induced affect was based on the straight alley (runway) and naturally appetitive and aversive stimuli (Figure 2b). The manipulation affected both the runway start and goal latencies. In nonisolated controls, runway start and goal latencies increased as a function of increasing aversiveness of the cues. Chicks in the anxiety-like state responded with increased latencies to aversive ambiguous cues. Chicks in the depression-like state responded with increased latencies not only to aversive but also to appetitive ambiguous cues.

Bateson et al. (2015) explored the effect of higher nest competition, induced by manipulating brood size, on reward expectation in European starlings. These researchers predicted that starlings from larger broods (brood size of seven vs. two) would display signs of pessimism in the judgment bias test based on color discrimination (Figure 1a and b). The researchers found that starlings with more competitors, specifically those that had experienced more nest competitors larger than themselves, had a longer latencies to respond to ambiguous cues most similar to the cue associated with the reward. Moreover, the birds with greater developmental telomere attrition, a measure of cellular aging associated with increased morbidity and reduced life-expectancy, showed shorter latencies to approach the ambiguous cues, which were interpreted as evidence for a more positive affective state. However, the authors failed to replicate these results in a latter study (Gott et al., 2019).

Horváth et al. (2014) studied another aspect of early social experience, the effect of mothering on judgment bias in Japanese quail. The maternal behavior of this species can be induced by fostering procedures (Richard-Yris, 1994). Brooded and nonbrooded quail chicks were tested in a spatial judgment task (Figure 3d) in which they were subjected to two judgment bias tests on days 19 and 20 after hatching. Although the latencies to reach the near-negative ambiguous location in brooded quail were significantly shorter in the first judgment bias test, indicating optimism, it was not validated by the second judgment bias test (Horváth et al., 2014).

Lalot et al. (2017) studied the influence of social context (birds housed in pairs vs. birds housed singly) on canaries’ emotions in a modified spatial judgment task (Figure 3c). They showed evidence of an optimistic bias (flying faster to the feeder at the central ambiguous location) in birds housed in pairs compared with those housed singly.

Adriaense et al. (2019) experimentally manipulated positive and negative affective states in demonstrator ravens. The emotional valence of another raven observing the demonstrator’s behavior was assessed using a bias paradigm to determine whether emotional contagion, a simple form of empathy, had occurred. Observers showed a pessimistic bias toward the presented ambiguous stimuli after perceiving demonstrators in a negative state, confirming the negative emotional contagion. However, observers did not show any judgment bias after perceiving demonstrators in the positive condition.

Cognitive bias and personality

Several authors have addressed the question of whether optimism or pessimism in cognitive bias tests is linked to birds’ personality.

Cussen and Mench (2014) studied the relationship between personality and cognitive bias in the orange-winged amazon parrot (Amazona amazonica). They developed a subjective personality assessment protocol. The inventory consisted of 36 personality traits and 4 physical traits that were scored from 0 to 7. These researchers found a significant correlation between neuroticism and performance in the attention bias test.

Lalot et al. (2017) studied six personality traits in canaries and found that aggressiveness, neophobia, obstinacy and one sociability index were repeatable, while the other sociability index, boldness and locomotion were not. However, no correlation between the birds’ optimism and any of their personality traits was found.

Another attempt to link personality with a judgment can be found in the report by de Haas et al. (2017a). The authors assessed adult laying hens’ behavior and physiological responses (tonic immobility test, an open field test, a manual restraint test, and plasma corticosterone levels after manual restraint) as indicators of their personality, and they planned to link the results with performance in a judgment bias test based on the color-cue association (Figure 2b). However, success in learning in this experiment was low, and therefore, the judgment bias test was not fulfilled. Nevertheless, success in training was related to personality, with better performance of the hens that showed a reactive personality type by a long latency to walk, struggle, or vocalize during the tests.

Fearfulness represents a part of an animals’ personality: the trait anxiety. de Haas et al. (2017b) tested whether high fearfulness affects discrimination learning and judgment bias in laying hens. Based on their responses to an open field at 5 wk of age, the birds were categorized as fearful or nonfearful. At the adult age, birds that reached the training criteria were subjected to the judgment bias test (Figure 2b). The study showed that fearfulness was associated with differences in discrimination learning ability. However, in contrast to the hypothesis, more fearful birds showed an optimistic-like response to the 25% ambiguous cue in comparison to nonfearful birds.

Ross et al. (2019) classified the personality of laying hens based on the arena and novel object tests as exploratory and nonexploratory. Hens classified as exploratory flipped a higher proportion of ambiguous lids while they were kept in enriched housing than in control housing. In contrast, in nonexploratory hens, the effect of housing was not significant.

Pharmacological treatments

During the processing of affective information, the encoding of reward value and decision-making involves many brain areas and brain systems. However, our knowledge of the underlying mechanisms of cognitive bias is still emerging (Mendl et al., 2009). Pharmacological treatments offer a way to test the physiological, neurobiological, and neuroendocrine mechanisms underlying the judgment bias; nevertheless, very limited data are available in relation to birds.

Chronic unpredictable mild stress was used to examine affect induction and, as a consequence, the induction of cognitive bias in the seminal study in rats (Harding et al., 2004); it was also effective in bias induction in chickens (Zidar et al., 2018). Iyasere et al. (2017) examined whether pharmacological induction of stress by experimentally elevating levels of corticosterone in broiler chickens would affect judgment bias. These researchers demonstrated a pessimistic judgment bias using the spatial judgment task (Figure 3a). Corticosterone treatment caused an increased expectation of punishment under ambiguity.

The brain mesocorticolimbic dopaminergic system is implicated in the valuation of reward-related stimuli and motivation of behavior toward such stimuli, and therefore, it is a candidate for the mechanisms that may underlie the evaluation and response to ambiguous stimuli in a judgment bias task (Mendl et al., 2009). The findings of Zidar et al. (2018) showed that the judgment bias of broiler chickens was related to dopamine turnover rate in the mesencephalon, with higher activity in individuals with a more optimistic response, thus supporting the idea of the involvement of dopaminergic mechanisms. Horváth et al. (2015) tested the effects of intramuscular injections of the dopamine receptor antagonists (15 min before the test) SCH23390 (D1) and haloperidol (D2), and the dopamine receptor agonists (2 h before the judgment bias test) SKF 38393 (D1) and bromocriptine (D2), in Japanese quail hens using the spatial judgment task (Figure 3d). Both antagonists caused a dose-dependent increase in the latencies to approach the food bowl. However, these phenomena applied to all food bowl locations and not just ambiguous ones. Treatment with agonists did not have a significant effect on approach latencies.

Hymel and Sufka (2012) demonstrated pharmacological reversal of cognitive bias in chicks using the anxiety–depression model previously introduced by the same laboratory (Salmeto et al., 2011) (Figure 2c). Judgment bias induced by a 60-min isolation was reversed by 15.0 mg/kg imipramine. However, 0.10 mg/kg clonidine produced modest sedation and thus was ineffective for reversing the bias.

Discussion

The most frequently studied method of affect induction in avian judgment bias studies is environmental enrichment. Nevertheless, of the three studies studying the effect of environmental enrichment on judgment bias in European starlings, one showed optimistic bias as a consequence of enrichment (Matheson et al., 2008), one partially confirmed significant effects of enrichment (only when the birds were moved from enriched to standard cages) (Bateson and Matheson, 2007), and the last one did not show an effect (Brilot et al., 2010). However, all of these studies were conducted with a relatively low number of subjects (Table 1). In poultry, the judgment bias tests did not validate the optimistic bias in birds in an enriched environment. Works that used a simple design with two levels of enrichment in laying hens (Wichman et al., 2012; Pichová and Košťál, 2016; Ross et al., 2019) or Japanese quail (Horváth et al., 2016) did not demonstrate any clear effect of enrichment on judgment bias. There are several possible explanations for these findings. According to Wichman et al. (2012), no significant differences between the two enrichment levels suggest that these environments do not induce large enough differences in the birds’ emotional state to have a significant impact on their behavior in the tests. Ross et al. (2019) argue that such results could perhaps indicate that birds do not have strong judgment bias responses to affective changes, and/or that judgment bias tests are less sensitive to a positive affect than to a negative one. Another possible explanation is that the test employed does not have sufficient sensitivity, which could potentially be increased by optimization (choice of reward and punishment in discrimination training, difficulty of the task, proportion of animals reaching the discrimination criterion, optimization of the number of sessions, proportion of reinforced to nonreinforced cues to prevent loss of ambiguity, proper statistical evaluation of the results, etc.). The effectiveness of mood induction was higher if different types of affect manipulations were combined, as shown by Zidar et al. (2018). Environmental complexity reduced stress-induced negative judgment bias by maintaining an optimistic bias in individuals housed in complex conditions, even after exposure to unpredictable stressors.

Several studies focused on environmental enrichment have suggested that birds respond more to the contrasting effect of the change in environment than to a stable environment. This finding is in line with the idea that mood depends on how recent reward outcomes differ from expectations (Eldar et al., 2016). Raoult et al. (2017) tested the hypothesis of mood as a cumulative expectation mismatch by assessing mismatch in the experimental procedures used to induce mood in published studies. However, based on published data, this researcher did not demonstrate that mood induction mismatch can predict the success of cognitive bias tests.

We have presented several studies related to social influences and the influence of personality traits on cognitive bias in birds. Judgment bias appears to be quite sensitive to the manipulation of the social environment. Nevertheless, studies of this type, especially in poultry, are quite rare. The problem of interpreting the studies linking judgment bias to personality traits is a very wide approach to personality assessment.

Doyle et al. (2010) were the first who pointed out the loss of the ambiguity phenomenon with repeated judgment bias testing. Pichová and Košťál (2016) showed that while the proportion of responses to reinforced positive and negative cues remains stable with repeated judgment bias tests, the proportion of responses to nonreinforced ambiguous cues declines. From this point of view, it is important to maintain a low proportion of nonreinforced cues in the judgment bias test. As shown in Table 1, there is a large variation in this parameter in published avian papers (from 23% to 100%), thereby providing further opportunity for improvement of the tests.

While the judgment of ambiguous stimuli represents the end action in the decision-making process, the attention bias plays a role at the beginning of this process, where it determines which stimulus receives the most attention (Lerner and Keltner, 2000). The advantage of the attention bias approach is that it does not require extensive training. However, until now, only a few attention bias studies have been conducted in birds. The first study was performed by Brilot and Bateson (2012) in starlings. The latency to move, latency to begin feeding, and duration of scanning were measured after the presentation of threatening cues (starling alarm call playback) to animals. The results showed that birds from a less complex environment (denied access to bathing water) interpreted threatening cues as requiring more caution than birds that had access to a water bath. This result is consistent with the higher levels of anxiety in water bath-deprived birds. A similar experimental design was used in laying hens by Campbell et al. (2019), in which the hens were selected based on their range use patterns, that is, “outdoor” birds that ranged daily and “indoor” birds that were never registered to go outside. Indoor-preferring hens showed attention bias toward the surrounding environment following threatening alarm call playback, thus indicating higher anxiety in indoor birds. Cussen and Mench (2014) used the spatial foraging task for attention bias testing, in which orange-winged amazon parrots were trained to choose one of four identical opaque containers to obtain a food reward (a piece of almond). The location of the reward was randomized across trials, and the parrots could not predict which container contained a reward. The study showed that the presence of an unknown passive observer in the room (threat) significantly increased parrots’ latencies to find the reward and number of errors.

Memory bias refers to how the current affective state influences the storage and retrieval of memories related to the stimulus (Bower, 1981; Keen et al., 2014). To our knowledge, there is only one paper describing an animal test of a memory bias, which was published by Burman and Mendl (2018) in rats. The rats were tested in an eight-arm radial maze. Changes in performance appeared to reflect social status and associated affective states. Similar to judgment bias tests, this method also requires training. Despite the necessity for further refinement, this or other newly developed memory bias tests are needed to contribute to the broader picture of the cognitive biases in birds.

Conclusions

The cognitive bias paradigm has inspired many laboratories to study emotions in birds, including poultry. Most laboratories have utilized the judgment bias approach using visual, spatial, or time interval discrimination cues (see Table 1 for an overview). A larger proportion of tests was based on Go/NoGo discrimination, but there are also studies based on Go/Go discrimination. The testing arenas vary from straight alleys, left/right choice arenas to spatial judgment arenas and operant conditioning chambers (see Figures 1–4). This type of test requires extensive training (see the comparison of the number of sessions and trials in Table 1). Some of the discrimination tasks seem to be more demanding, as indicated by the proportion of animals reaching the discrimination criterion, which ranges from 40% to 100%. In some cases, the problem of low success may be related to the low biological relevance of the tasks for the given species, that is, a more ethological approach would be more plausible.

Thus, there is space for improvement not only for the training procedures but also for the judgment bias tests. One known problem is, for example, the exposure to too many unreinforced ambiguous cues that lead to a loss of ambiguity.

Manipulation of the animal affect is a delicate procedure. It seems to be easier to induce negative than positive affective states. Many questions remain associated with the success of affect induction, such as the proper duration of manipulation, contrast between manipulations, or consideration of expectations. The controversial results obtained for some avian judgment bias tests are probably due to differences or imperfections in affect induction.

In conclusion, the cognitive bias paradigm provides new opportunities for a better understanding of animal emotions, including emotions in birds and poultry. Nevertheless, the methods of judgment bias tests must be further optimized and validated. Newly developed methods of attention and memory bias testing must also be further developed and tested.

Conflict of interest statement

The authors declare no real or perceived conflicts of interest.