Table 2.

Points to consider and recommendations to design experimental stimuli and protocols, and tailor it to each particular dog as necessary (middle columns). Suggestions of research questions for future studies that may answer particular methodological 592 challenges are also listed (column on the right). Points to consider and suggestions for future studies (both novel questions or 593 deepening of published questions) are organised by the features more prone to challenges (column on the left)

Features prone to challenges	Points to consider	Recommendations	Examples of future research questions
6.1. Breed, individual differences, and the umwelt of each dog	Diversity of dogs as a species	Sample larger diversity of dog types regarding breed, age, sex, cephalic type, facial morphology, human environment, life history, etc.	How does perception of emotion cues develop and vary over the lifetime of the individual?
	Individual differences within dog types	Consider the umwelt of the dog that may vary within dog types regarding temperament, personality, motivation, mood, etc. Assess preferred rewards (e.g., food vs praise vs play) to ensure optimal motivation and attention during task Use validated psychometric scales and/or behavioural tests to assess individual differences Assess sensitivity to rewards and aversives with e.g., PANAS, Positive and Negative Activation Scale (Sheppard and Mills 2002) to assess each dog’s emotional predispositions and avoid sampling bias or excluding dogs Assess temperament and impulsivity to understand to which dogs the task provides an inherent reward (e.g., play with a human), and which dogs require external rewards (e.g., treat) to increase extrinsic motivation (Deci et al. 1999) Keep motivation and focus high, whilst keeping over habituation and boredom at a minimum (e.g., allowing the dog to leave/stop the experiment at any time for a short break, keep trials/sessions short and stimuli as varied as possible)	How does perception of emotion cues vary between different temperaments?
	Differences between dogs and humans	Consider differences between dogs and humans when adapting experiments developed for humans and regarding sensorial abilities (visual and other) Control for other sensorial contaminants and influxes in the testing environment, which may go undetected by humans but bias dog behaviour (e.g., odour, magnetism, temperature)	To what extent do dogs and humans use similar mechanisms for processing emotion cues of both conspecifics and heterospecifics? Investigate sensory abilities present in dogs but not in humans: Does magnetism or temperature affect dog's visual perceptual mechanisms?
6.2. Experimental design: controlling variables whilst maintaining ecological validity	Presence of the owner	Allow presence of the owner, since 1) this makes the controlled environment of a laboratory more naturalistic and also emotionally equable for different subjects; 2) owners can act as secure bases for dogs in novel environments and when encountering strangers such as the experimenter (Gácsi et al. 2013) But also blind owners (both metaphorically—withhold experiment goal until its end, and literally—use blindfold, earplugs), as owner’s inadvertent cuing must be controlled to avoid Clever Hans effects (Miklösi et al. 1998; Schmidjell et al. 2012)	How does the presence of the owner affect the performance of dogs in tasks of visual perception of emotion cues?
	Collection of dog spontaneous responses and ecological validity	Allow for free full body movement responses to the stimulus (e.g., tail wagging, head turns), since the absence of natural responses may impact perceptual processes Give preference to naturalistic experimental protocol steps (luring/holding lightly vs. extensive training for immobilisation), and avoid conditioned or emotionally primed responses If less naturalistic steps are absolutely needed (e.g., immobilisation in fMRI or to assess eye saccades), discuss how these may have impacted the results (e.g., not moving the head when perceiving emotional cues, which are known to cause head turns in dogs (Siniscalchi et al. 2010, 2013)) But also control for increased random error and risk of correlated systematic error, which can make the correct and precise identification of the influential variable(s) more difficult. Random error effects may mask important effects that would be significant in a more controlled environment, and systematic error associated with other factors may lead to inaccurate associations Discuss how effects found in a highly controlled setting would stand in a real-life scenario. More controlled experiments (e.g., in the lab) facilitate equipment handling and ensure important but small effects are not masked by other variables, but may lose ecological validity Consider the thermal and magnetic properties of the equipment used (e.g., visual display units, fMRI, Fig. 5) with dogs as a potential confound in experiments (since dogs can sense these)	Does immobilisation of the dog affect the perception of emotion cues? How does the wide range of experimental and stimuli properties (Fig. 4, Table 1) influence perception of emotion cues in dogs? For example, real-life demonstrators vs. video, spontaneous vs. posed emotion cues, passive viewing vs. task engaged, trained for remaining immobile vs. allowing movement (e.g., tail or head turns) Does the equipment used in experimental setups with thermal and magnetic emissions impact the stimuli perception or the dog performance?
6.3. Experimental stimuli: spontaneity and validity of stimuli, but with well-defined categories and objectively measured cues	Selection and design of stimuli	Thoroughly define and justify the conceptual basis for the stimuli selection and design (e.g., psychobiological approach, context and triggers used to induce facial expression, quantification of cues that match the dog behavioural repertoire) Avoid face-centric stimuli and include body postures and gestures, which are particularly important for dogs, according to their natural behaviour and recent eye-tracking evidence (Correia-Caeiro et al. 2021)—this will counter the evident face publication bias and consider the umwelt of dogs Avoid the use of the exact same triggers to create stimuli featuring humans and dogs; Give preference to functional equivalent triggers (e.g., adult humans are generally not afraid of thunderstorms (Silverman et al. 2001), while dogs often are, at a clinical level (Lopes Fagundes et al. 2018; McPeake et al. 2017; Overall et al. 2001), hence thunderstorms may be a good trigger for dog fearful behaviour, but not for humans) Avoid (or be particularly cautious with) the use of emotion categories and corresponding behaviours/visual cues common in humans but that may not be found in dogs (or at least not in the same form), including emotion categories and respective cues that are currently still being debated in canine science (e.g., guilt (Ostojić et al. 2015), jealousy (Cook et al. 2018; Karl et al. 2021)) And vice-versa, use emotion categories that are more common/relevant in dogs and have associated emotion cues in dogs but not humans. For example, positive anticipation cues are present in dogs but not humans	What are the differences in potential emotional states and its associated cues between dogs and humans? What contexts and triggers are ideal to collect stimuli for perception of emotion cues in dogs? What potential emotional states may be triggered by sensing magnetism or a distant source of heat and are there any cues displayed during these states?
	Classification and description of stimuli	Avoid emotion labels for stimuli, since these are subjective and too broad Avoid using anthropomorphic/anthropocentric emotion categories, i.e., based solely on human research (e.g., from facial expressions), particularly if there haven’t been fundamental studies demonstrating these to be associated with a particular type of display in dogs Explicitly describe the context in which the stimuli were collected (i.e., dog growling during food competition/territory defence, etc.) Quantify the emotion cues observed in the stimuli (e.g., how many/which/duration of AUs/gestures/postures), by for example using tools such as DogFACS (Waller et al. 2013) or DogBAPS (Huber et al. 2018)	Exactly what AUs/gestures/postures (e.g., facial and bodily expressions) dogs display in response to different emotional triggers? (also very little studied in dogs)
	Validation of stimuli	Give preference to ecological (i.e., pertaining to appropriate environment for the species) and evolutionary (i.e., pertaining to survival value for the species) validity, whilst considering the species natural behaviour, ecology and motivation (Tomasello and Call 2008), in order to ensure laboratory findings can be generalised to the outside world Avoid asking vague “expert opinion” (or a sample of random human observers) to validate the stimuli regarding the emotion as the only validation step—this will likely just incur in circular reasoning and confirm human observers biases Instead validate the stimuli by asking experts to independently quantify cues present in the stimuli and/or describe the context in which the cues were produced (see previous point about classification and description of stimuli)—this will assess agreement on objective and measurable cues instead of subjective impressions Describe in detail in the methods section how the experts validated the stimuli as appropriate for the tested effect (e.g., a “happy dog face” needs to display “relaxed open mouths”, “lip corners retracted”, and absence of “ears backwards”)
	Presentation of stimuli	Present stimuli in an ecological valid way (e.g., facial expressions not at the dog’s eye level) Give preference to video as stimuli (Correia-Caeiro et al. 2021; Karl et al. 2020) and avoid static pictures, since the former includes onset, apex and offset of a visual cue (e.g., facial expression), natural timing, symmetry, intensity, etc. that the dog is more familiar with in their daily life Give preference to spontaneous stimuli and avoid posed stimuli (e.g., real-life demonstrators will vary in their behaviour and will have posed behaviours, spontaneous facial expressions differ from posed facial expressions) But also match the specs of the equipment used for stimuli presentation to the visual abilities/needs of dogs (e.g., screen high refresh rate)	Is there an impact when stimuli are placed at the dog’s eye level when in real-life it is not (e.g., over-inflation of face which may modify eye movements)? Does dynamic information change perception of emotion cues in dogs as it does for humans? How do canine displays performed on command differ from spontaneous displays?