Why averages lie
On ergodicity, idiographic research and what they might mean for evidence-based practice.
In 2004 the Dutch methodologist Peter Molenaar published an article with a telling title: A manifesto on psychology as idiographic science. His point was simple and, in our view, far-reaching. Psychology usually studies groups of people and applies the resulting averages to the individual patient. Molenaar showed that this way of working rests on an assumption that psychological systems seldom seem to meet.
That assumption is called ergodicity, a term from statistical physics. A system is called ergodic when the behaviour observed over time in one element shows the same pattern as the behaviour observed at a single moment across many elements. Put differently, the average patient and the patient followed over time then show the same dynamics.
That would be convenient. An RCT showing that treatment X works better on average than treatment Y could then be translated directly into advice for the individual patient: choose X. According to Molenaar, however, this condition is rarely met in psychology; psychological systems are, as a rule, not ergodic.
Typing: fast or accurate
Molenaar used a simple example that many readers have found persuasive. In group studies, typing speed and accuracy are positively related: those who type faster also type more accurately. That is what one would expect, since a good typist can work both quickly and precisely. If, on the other hand, one follows a single person over time and measures her speed and accuracy session after session, the relation within that person often turns out to be negative: when she types faster, her precision suffers. Here the group picture and the individual picture tell opposite stories.
The example does not stand alone. In several psychological domains where this has been studied, such as mood, behaviour, cognition and personality, comparable differences have been found. As a result, group statistics often say little about what happens within one person.
The assumption that the psychological dynamics of an individual correspond to the dynamics of the group average is called the assumption of ergodicity. For most psychological processes that assumption is doubtful.
What this means for evidence-based practice
This is where things become somewhat uncomfortable. A large part of our evidence-based knowledge rests on group studies: RCTs, meta-analyses and treatment guidelines that describe what works on average for the average patient. If those averages cannot simply be translated to the individual, this raises the question of what exactly the guidelines tell us.
The answer is a nuanced one. Guidelines are certainly not worthless, but they say something different from what we sometimes assume. An RCT shows that in this population treatment X works better on average than Y. For which patient X helps, for which Y, and for which neither, remains open. Seen in this light, an RCT yields a hypothesis rather than a prescription.
That distinction carries more weight than it may seem. A hypothesis asks to be tested with this particular patient; a prescription feels like an obligation that stands apart from the patient. The first view leaves room to examine, with a clinical eye, whether the expectation holds for this patient.
The idiographic complement
Non-ergodicity is no reason to abandon group research. It seems more natural to complement it with research at the level at which we also act in the clinic: that of the individual.
Idiographic research, with one person and many measurements over time, shows how a specific patient functions, which variables in his or her system are related, and how those relations change. In this way it forms the counterpart of the group research we already know.
In practice this means that repeated measurements (ESM, diaries, IROC or questionnaires administered several times) are more than a by-product of treatment. From this perspective they are methodologically necessary, because only in such time series does the dynamics that applies to this patient become visible.
The technical means for this kind of research have become more accessible in recent years. R packages such as mlVAR, qgraph and bootnet make it possible to estimate within-person networks, and apps for ESM (PsyMate, Ethica, mEMA) allow data to be collected without a large investment. What was long reserved for researchers is thus coming within reach of clinicians as well, although it does call for some training.
What it can mean in the consulting room
The largest change that idiographic work brings may be a cultural one. We are used to establishing for our patients what is going on, on the basis of diagnoses, of their account and of our experience. Idiographic work invites us to investigate this together with them. The questionnaire a patient fills in twice a day for six weeks is then not paperwork but her own data: a pattern from her own life becoming visible.
Many patients seem to experience this as a relief. Things become concrete, and there is something on the table that represents them, alongside the classification of what they have. Sometimes connections come to light that neither patient nor clinician had seen: a relation between social activity and mood that turns out differently than expected, a day of the week that is consistently difficult, or a pattern of flickering shortly before a relapse.
An honest question
Does this undermine evidence-based thinking? We think not, although it does make our work more demanding. It brings into view a complexity that we could until now partly leave aside. At the same time it offers new means by which interventions can become more targeted. And it offers patient and clinician a form of honesty: what we know, we know mainly about groups; what is happening in this person, we only know once we investigate it together.
Molenaar's manifesto dates from 2004. More than twenty years later it seems to be gradually taking hold, less as an upheaval than as a slow recalibration. That seems rather fitting for a complex system.