Toward a Quantification of Anhedonia: Unified Matching Law and Signal Detection for Clinical Assessment and Drug Development

Abstract

Anhedonia, the loss of pleasure from previously rewarding activities, is a core symptom of several neuropsychiatric conditions, including major depressive disorder (MDD). Despite its transdiagnostic relevance, no effective therapeutics exist to treat anhedonia. This is due, in part, to inconsistent assays across clinical populations and laboratory animals, which hamper treatment development. To bridge this gap, recent work has capitalized on two long-standing research domains dedicated to quantifying responsivity to antecedents and consequences across species: the generalized matching law and signal detection theory. This review traces the integration of these quantitative frameworks, which yielded two empirically derived metrics: response bias (log b) and task discriminability (log d). These metrics serve as primary dependent variables in the Probabilistic Reward Task (PRT). In this computerized task, subjects make visual discriminations and probabilistic contingencies are arranged such that correct responses to one alternative are rewarded more often (rich) than correct responses to the other (lean). Under these conditions, healthy participants consistently develop a response bias in favor of the rich alternative, whereas participants with MDD exhibit blunted biases, which correlate with current and predict future anhedonia. Given the correspondence between anhedonic phenotypes and response bias, the PRT has been reverse-translated for rodents and nonhuman primates. Orderly log b and log d values have been observed across diverse clinical populations and laboratory animals. In addition, pharmacological challenges have produced similar outcomes across species. Taken together, this quantitative framework offers a highly translational approach to assaying reward responsiveness to accelerate treatment development for neuropsychiatric disorders involving anhedonia.

Keywords: anhedonia; drug development; generalized matching law; quantitative models; signal detection theory; translational research.

Fig. 1

Idealized data representing the ratio of responding (ordinate) as a function of the reinforcement obtained on the two alternatives (abscissa). The dotted line represents strict matching and the solid line represents possible variations in sensitivity (top panel) and inherent bias (bottom panel). Figure adapted from Reed and Kaplan (2011)

Fig. 2

The four stimulus and response events in the yes–no signal detection task, where B₁ and B₂ denote responses with respect to the stimuli S₁ and S₂. w and z represent, respectively, the number of hits and correct rejections of S₁ resulting in reinforcement. x and y represent, respectively, the number of misses and false alarms resulting in extinction. Figure adapted from McCarthy (1983)

Fig. 3

Photographs and task schematics for the human (top; Pizzagalli et al., 2005), rat (middle; Kangas et al., 2020), and marmoset (bottom; Wooldridge et al., 2021) PRT

Fig. 4

Upper panel: Mean log b values from select PRT studies. Black bars indicate mean log b values from healthy control subjects across species, including humans (±range of studies referenced below), rats using line length stimuli (Kangas et al., 2020), rats using auditory stimuli (Der-Avakian et al., 2013), and marmosets (Wooldridge et al., 2021). Gray bars indicate mean log b values from human participants with psychiatric disorders, including non-melancholic MDD (Fletcher et al., 2015), bipolar disorder (Pizzagalli, Evins, et al., 2008a), high BDI score (Pizzagalli et al., 2005), MDD outpatients (Pizzagalli et al., 2008c), substance use disorder (Boger et al., 2014), and melancholic MDD (Fletcher et al., 2015). White bars indicate mean log b values following pramipexole administration in humans (Pizzagalli, Goetz, et al., 2008b) and rats (Der-Avakian et al., 2013). Lower panel: Mean log d values for healthy control participants (black bars) and corresponding comparisons (gray bars)