Research

The Idiographic Dynamics Laboratory studies psychopathology at the level of the individual—and, increasingly, at the level of the discrete moments each individual experiences and the discrete elements and conditions that make up each moment. For fifteen years the lab has pursued one question through successive reformulations: how do we build a clinical science that is at once actionable—capable of yielding decision-relevant information about a particular person—and generalizable, such that its inferences hold beyond the sample in hand? These two demands ordinarily trade against one another. Our work has been a sustained argument that this apparent trade-off is actually an artifact of the unit of analysis, and that selecting the right unit can dissolve the tension altogether.

The ID Lab started in 2013 as the first dedicated idiographic lab in Clinical Psychology. Motivated by the work of Peter Molenaar, we set out to demonstrate that group-level statistics do not, in general, describe the individuals who compose the group—a property known as nonergodicity, which our laboratory helped establish as an empirical cornerstone in clinical psychology (Fisher, Medaglia, & Jeronimus, 2018, PNAS). If the average person is a statistical fiction, then averaging over people is not a safe route to claims about any one of them. The lab's first phase was a direct response to this premise: it built person-specific (i.e. idiographic) modeling into a working paradigm, narrowing the unit of analysis from groups, to individuals, to moments within individuals.

Our current program narrows it once more—to the discrete elements of which those moments are composed—and in doing so changes what is meant to generalize. Person-specific parameters are themselves idiosyncratic; they are unlikely to travel across people. But structure can. Configural psychometrics treats clinical features as present-or-absent elements and asks, using set theory, combinatorics, probability, and information theory, which configurations of elements are minimally sufficient for a clinical outcome—the smallest load-bearing combinations of symptoms sufficient for a target outcome—and which individual elements are necessary across those configurations. At the between-person level, necessity and sufficiency applies to classifications such as diagnosis, distress, and impairment. At the intraindividual level, we examine the necessary and sufficient conditions for moments of acute risk or emotional vulnerability.

The configural psychometric approach is acategorical: it assumes no diagnostic categories in advance and lets the outcome define which configurations matter. And it is, by construction, actionable, because it speaks in the native currency of clinical decisions—a feature is present or it is not; a configuration is sufficient or it is not; one intervenes or one does not.

This inverts the logic of measurement itself. Where latent-variable models define a construct by its internal coherence and then ask, separately, how it relates to the world, the configural approach lets a construct earn its meaning from the work it does—its relation to observable outcomes. Applied to the internalizing disorders of the DSM, a handful of minimally sufficient configurations recovers essentially all of the diagnostic information contained in thousands of possible symptom combinations, reducing that combinatorial complexity by more than 99% while revealing which symptoms—worry, for one—are non-negotiable (Fisher, 2026).

Current directions

Configural psychometrics: theory and methods. The foundational program—sufficiency, necessity, and the combinatorial and information-theoretic machinery behind them. Current work develops formal indices of symptom necessity and dispensability and a resulting taxonomy of keystone, gateway, specialist, and dispensable symptoms. The full analytic pipeline will soon be openly available in version 2.0 of the R package setweaver (Leenaerts & Fisher, 2026) on CRAN.

Set-theoretic and Order-theoretic Taxonomy. In collaboration with Ashley Watts of Vanderbil University and Lorenzo Lorenzo-Luaces of Indiana University, recent work in the lab has replicated the findings from Fisher (2026) in the NESARC wave 1 data, one of the largest psychiatric epidemiological samples in the United States (Fisher et al. under review).

Along with Drs. Watts and Lorenzo-Luaces, current work in the lab is extending the configural approach from the DSM to the HiTOP model. In a recent collaboration with Miri Forbes of Macquarie University (see, Forbes et al. 2024; CPS; Forbes et al. 2026; JPCS), we are applying the configural psychometric framework to the HiTOP model’s internalizing spectrum.

Configural pathways to suicidal ideation and self-injury. The first application of set-theoretic methods to intensive longitudinal (EMA) data utilizes two recently collected data sets from Sarah Victor’s lab at Texas Tech. In this recently-submitted paper, we examined person-specific, time-forward prediction of momentary suicidal and self-injurious risk in two independent samples of recently suicidal Trans and Non-Binary individuals. Predictions were validated out-of-sample and replicated across the independent samples. We see this work as an important foundation for just-in-time adaptive interventions.

Sequential emotion regulation. The most recent study—with data collection planned to continue through the end of December, 2026—involves a purpose-built momentary survey that reconstructs the cascade of emotion-eliciting and emotion-regulatory processes occasion by occasion. An emotion arises, prompts an urge, is acted on, invites an attempt to regulate, is believed regulable, is tried, and succeeds or fails. From this sequence we enumerate, per person, the configurations under which regulation holds or gives way to distress. We again conceive of these models as candidate decision points for precision intervention in real time.

Joining the lab

Prospective graduate students should have a serious interest in idiographic and configural approaches to clinical science, and a genuine appetite for the mathematics and methodology they require: set theory, combinatorics, information theory, and intensive longitudinal design. Fluency in R is an asset; a willingness to build it is essential. The strongest applications will engage with these concepts sincerely, even if they may seem foreign or novel to you. Articulating your personal vision for the future of clinical science is also encouraged.

Selected Publications

Full list available on Google Scholar.

2026

2025

2023

2022

2019

2018

2017

2015