Clonan 2025 contrasting with a previous (non miso) Kumar 2008 paper on sound decomposition , and the use of method in Andermane 2023 showing null effects in a Misophonia study

Clonan et al. (2025) situated their Generalized Perceptual Regression approach against two prior efforts: Kumar et al. (2008), who showed that a similar auditory decomposition could predict average sound unpleasantness from spectrotemporal features in typically hearing listeners, and Andermane et al. (2023), who applied that same model but found no significant difference in predicted discomfort between misophonic and control participants. To address concerns about novelty, the authors argued that Kumar's use of partial least squares (PLS) regression entangles acoustic features with listening preferences and operates on the high-dimensional feature space directly, whereas their own method first reduces dimensionality and then regresses, keeping the acoustic representation independent and interpretable. They further argued that, unlike these prior studies which modeled average group preferences, their framework generates individualized models of each listener's profile, and that it extends earlier work by revealing clusters of acoustic sensitivity across participants, testing out-of-distribution predictions, and demonstrating time-varying prediction in continuous sound mixtures (Clonan et al., 2025).