Lee and Schwarz (L&S) review an extensive body of research on cleansing behavior in healthy adults. Here, we discuss implications for the field of clinical psychology. As psychiatric disorders can be considered to be cognitive states at the end of a continuum (Keyes, Reference Keyes2002), grounded procedures should reproduce in individuals with psychiatric disorders. Disorders most associated with moral conflicts – important triggers of cleansing behavior according to L&S – are obsessive-compulsive disorders (OCD, Chiang, Purdon, & Radomsky, Reference Chiang, Purdon and Radomsky2016). In the washing subtype of OCD, individuals respond to unwanted thoughts and ideas (obsessions) with extensive cleaning rituals (compulsions). Even though extensive cleaning leads to short-term relief in individuals suffering from OCD, it is also a major factor that contributes to the disorder's maintenance. The initial feeling of relief as well as the patient's assumption that extensive cleaning prevented a feared outcome can lead to an extensive and repeated execution of compulsions. Accordingly, exposure to obsessions combined with a prevention of compulsion is considered first-line treatment for OCD (Koran, Hanna, Hollander, Nestadt, & Simpson, Reference Koran, Hanna, Hollander, Nestadt and Simpson2007).

L&S's grounded-cognition account predicts that cleansing effects should be accentuated in individuals with psychiatric disorders. In a moral context, compared to a healthy control group, grounded separation effects should be stronger in disorders associated with hypermorality (e.g., OCD) and weaker in those associated with hypomorality (e.g., antisocial personality disorder; Braun, Léveillé, & Guimond, Reference Braun, Léveillé and Guimond2008). In a non-moral context, the influence of cleansing behavior on social threat – as proposed by L&S – might be investigated by comparing individuals with different levels of social anxiety (e.g., social phobia vs. low social anxiety). Similarly, we should find evidence for grounded separation in disorders associated with ruminations, such as depression. Potentially, cleansing may separate the self from these negative experiences or thoughts. If this were the case, a grounded cognition view could provide a new vantage point for understanding cognitive processes related to specific disorders, and might even suggest improvements in therapeutic interventions. Indeed, Reuven, Liberman, and Dar (Reference Reuven, Liberman and Dar2013; see also D'Olimpio & Mancini, Reference D'Olimpio and Mancini2014) showed that effects of grounded procedures were stronger in OCD patients compared to controls (replication of Zhong & Liljenquist, Reference Zhong and Liljenquist2006; but see Siev, Zuckerman, & Siev, Reference Siev, Zuckerman and Siev2018). In light of this initial evidence, it is surprising that experiments testing grounded procedures in clinical disorders are still scarce.

There are comprehensive requirements to be met before moving into the clinical field (involving preregistration, eligibility criteria, power analysis, sample stratification, etc.; Schulz, Altman, Moher, & the CONSORT Group, Reference Schulz, Altman and Moher2010). Additionally, clinical researchers are not merely interested in statistical, but clinical significance, that is, the practical importance of a treatment effect. Are the grounded-cognition effects of separation stable enough to meet such criteria? L&S discuss the replicability of central findings in the field, and their meta-analysis comes to the conclusion that there is a valid but small effect underlying the many studies. Given the notoriously low power of much of psychological research (especially if it rests on group comparisons), it is little wonder that some of the effects reproduce in some studies but fail to replicate in others – indeed, a chequered replication performance is a hallmark of a small genuine effect tested with many underpowered studies.

However, replication is not only relevant across studies, but also within studies. This is a matter of experimental design. From our inspection of the literature reviewed by L&S, the dominant experimental design in the field seems to be very small. The typical experiment is a 2 × 2 design where a cleansing manipulation (e.g., handwashing or no handwashing) is crossed with a manipulation that elicits the effect of interest (e.g., cognitive dissonance before and after a choice; e.g., Lee & Schwarz, Reference Lee and Schwarz2010a). Sometimes the design only consists of an experimental and a control group (e.g., Schnall, Benton, & Harvey, Reference Schnall, Benton and Harvey2008). Even though designs generating such solitary outcomes are common in many areas of psychology, they are only able to demonstrate the effect of interest exactly once, giving them only minimal internal validity. In that case, statistical significance is far from guaranteeing replicability – remember that out of all solitary outcomes that are significant at a p value of exactly 0.05, only about half would be expected to replicate significantly and in the same direction.

Even more importantly, small designs often fail to include sufficient control conditions. For example, Lee and Schwarz (Reference Lee and Schwarz2010a) observed reduced cognitive dissonance for a choice of a music CD after handwashing with soap, compared to mere inspection of the soap (see de los Reyes, Aldao, Kundey, Lee, & Molina, Reference de Los Reyes, Aldao, Kundey, Lee and Molina2012, for a replication). But because this is a solitary effect, we do not know whether it is specific to cleansing. Maybe the critical difference is simply between a completed action and a passive waiting period? What if potting a plant, which actually soils the hands, had reduced dissonance by the same amount as washing them? There is a second experiment in the paper to validate the effect, but it just substitutes jam for music, never examines the actual cleansing manipulation, and is therefore open to the same questions. This is suggestive of an overly confirmative research strategy (Firestone & Scholl, Reference Firestone and Scholl2016) and future studies should put emphasis on eliminating alternative explanations.

We suggest introducing parametric manipulations within single experiments to see whether the theoretically critical manipulation actually drives the magnitude of the effect – in other words, parametric experiments exploring a dose–response relationship. For instance, five ordered levels of handwashing (no washing, dry cloth, clear water, plus soap, plus disinfectant) would predict exactly one ordering of conditions with respect to the dependent variable, out of a possible 120 orderings. This is a powerful permutation test, which entails up to four internal replications of the effect, tells us the range of possible effect sizes, and immediately rejects most alternative explanations. Comparing parametric variations between clinical groups further increases combinatorial power – not just linearly, but exponentially. To conclude, we believe that using these design options and also exploring clinical populations will spawn studies that have the potential to solidify the account of grounded procedures and improve understanding of cognitive processes in clinical disorders.