Assessing laboratory animal welfare: the crucial importance of construct validity.

Assessing laboratory animals' welfare - their current and/or past subjective affective states - is essential for ethical and regulatory reasons (and central to biomedical research into, for example, pain, nausea or anxiety). But this is challenging; and in the quest for quantification (and perhaps simplicity), it can be tempting to overlook construct validity. Nevertheless, that our indicators have good construct validity - that is, they accurately reflect the construct or concept of interest - is essential. This is true whether we are interested in short-term emotions like fear, longer-term mood-like states such as malaise, or markers of cumulative stress over a project or even a lifespan. Without it, welfare assessments risk being incorrect: inaccurate and unhelpful for the animals they aim to evaluate and assist. Here (summarising text from a forthcoming edited book), I introduce five validatory tests, as well as highlighting the importance of considering indicators' responsiveness/sensitivity and selectivity/specificity. I also outline how these principles could help improve the construct validation of both humane endpoints and retrospective severity assessments. Careful construct validation can never fully solve the 'Other Minds' problem: that animals' subjective experiences are private (such that we can never measure them, only infer them). However, done well, construct validation would add additional logical rigour to laboratory animal welfare assessment, increase its accuracy, and make benchmarking (e.g. severity classification) more transparent.