Systematic Review Finds Dog Judgment Bias Tests Methodologically Fragile

A comprehensive systematic review of thirty-two studies has concluded that judgment bias tasks, while capable of detecting certain emotional shifts in domestic dogs, currently suffer from significant validity and reliability limitations that prevent their use as standardized field measures for animal welfare. The research, published in Frontiers in Veterinary Science, represents the first systematic evaluation of the canine evidence base for this cognitive assessment tool, which aims to measure optimistic or pessimistic responses to ambiguous stimuli as proxies for underlying affective states. Despite growing adoption in veterinary behavior science, the authors found that the methodological heterogeneity across studies precluded meta-analysis, necessitating a narrative synthesis that highlights pervasive concerns regarding test consistency and interpretive clarity.

The review identified five distinct paradigm types within the literature, yet twenty-eight of the thirty-two included studies, representing eighty-eight percent of the corpus, relied on the same spatial discrimination task. In this dominant paradigm, dogs are trained to associate specific locations with positive or negative outcomes and then tested on their approach latency to intermediate, ambiguous locations. The remaining four studies each utilized a different approach, including visual grayscale discrimination, auditory tone frequencies, Go/No-Go touchscreen tasks, and attention bias measures. This concentration on a single spatial method has created a practical impediment to comparing findings across studies, as variations in the number of locations, training protocols, and latency scoring procedures introduce substantial noise into the data.

Validity concerns are particularly acute due to the confounding influence of spatial learning ability on task performance. One key study demonstrated that dogs with stronger spatial working memory showed significantly shorter latencies to ambiguous probes, suggesting that the task may measure cognitive competence rather than emotional valence. Furthermore, the review highlighted poor test-retest reliability, with one longitudinal study reporting intraclass correlation coefficients as low as 0.009 across repeated sessions. In the absence of any affective manipulation, dogs in that study became progressively slower to approach ambiguous locations, a drift toward pessimism that complicates the interpretation of longitudinal welfare monitoring. These findings indicate that the current spatial paradigm does not yet meet the psychometric thresholds required for stable clinical application.

Despite these limitations, the review found that judgment bias tasks are sensitive to specific experimental manipulations, particularly in pharmacological and training-related domains. A randomized controlled trial involving sixty-four dogs found that intranasal administration of oxytocin produced a significant shift toward optimistic bias, with a very large effect size of d = 2.16 in communicative contexts. Similarly, observational and quasi-experimental studies reported associations between aversive training methods and pessimistic cognitive bias. One matched-pair case-control study of one hundred dogs showed that those trained with two or more aversive methods were significantly slower to approach ambiguous locations, a finding that aligns with elevated cortisol levels and increased stress behaviors observed in similar populations.

Environmental enrichment interventions also yielded mixed but directionally consistent results. A randomized trial found that nosework training induced a positive judgment bias in pet dogs, while olfactory enrichment with specific oil blends reduced latency to ambiguous locations in shelter dogs. However, not all enrichment strategies produced measurable effects; a study testing the provision of simple and complex toys over twelve days found no significant differences in cognitive bias or activity levels.

The review also examined the convergent validity of judgment bias tasks with independent welfare indicators, finding only partial alignment. While some studies demonstrated coherent patterns where cognitive bias shifts coincided with changes in heart rate, cortisol, or behavioral assessments, others revealed significant dissociations. For instance, one study found that conspecific separation in shelter dogs increased activity and immune markers but did not alter cognitive bias, suggesting that the JBT and physiological measures may operate on different timescales or capture distinct aspects of the animal’s internal state. This lack of universal convergence underscores the need for multi-method approaches in welfare assessment, relying on a combination of cognitive, behavioral, and physiological data rather than a single metric.

Risk of bias assessment using the SYRCLE tool revealed that most experimental studies in the corpus had high or unclear risk in domains related to randomization, allocation concealment, and blinding. Only two studies achieved low-risk ratings across the majority of domains, and no study was uniformly low-risk across all ten criteria. The authors noted that the small sample sizes in many studies, combined with the high degree of methodological variability, create a landscape where selective reporting and small-study effects may inflate apparent effect sizes.

The authors recommend several priority areas for methodological improvement to advance the field. Standardization of spatial paradigm protocols is urgently needed, including agreed-upon compartment configurations, training criteria, and procedures for detecting spatial memory confounds. They also suggest that researchers should explicitly measure and report potential confounds and employ statistical controls where appropriate. Expanding the investigation of non-spatial paradigms, such as visual or auditory tasks, may provide less confounded alternatives for measuring affective states. Additionally, the review calls for systematic investigation of test-retest reliability across representative dog populations and the development of pre-specified analysis protocols to reduce the risk of selective outcome reporting.

In conclusion, while judgment bias tasks offer a promising avenue for assessing the emotional experiences of domestic dogs, the current evidence base does not support their use as fully valid and reliable tools for routine welfare assessment. The methodological fragility of the dominant spatial paradigm, characterized by spatial memory confounds and poor temporal stability, requires substantial improvement before the JBT can be confidently applied in clinical or shelter settings. Future research must prioritize standardization, mechanistic clarity, and multimodal validation to ensure that cognitive bias measures accurately reflect the affective states of dogs rather than their cognitive abilities or response to testing conditions. Until these improvements are implemented, practitioners are advised to interpret JBT results with caution and to integrate them with other behavioral and physiological indicators for a comprehensive welfare evaluation.

Subject of Research: Systematic review of judgment bias tasks in domestic dogs

Article Title: Assessing affective states in domestic dogs through judgment bias tasks: a systematic review of methods, influencing factors, and welfare applications

Article References: Gazzano, V., Ciurli, L., Curadi, M. C., Baragli, P., Gazzano, A., & Ogi, A. (2026). Assessing affective states in domestic dogs through judgment bias tasks: a systematic review of methods, influencing factors, and welfare applications. Frontiers in Veterinary Science, 13, Article 1926433. https://doi.org/10.3389/fvets.2026.1926433

Image Credits: AI Generated

DOI: 10.3389/fvets.2026.1926433

Keywords: animal welfare, canine behavior, cognitive bias, judgment bias, systematic review, veterinary science, affective states, methodological validity, Assessing, affective, states, domestic