Early warning signals (critical slowing down)
The precursors of transition — and what we can do with them in the consulting room.
One of the — to our minds — most promising findings of the past fifteen years in complexity research is the discovery that many systems appear to announce themselves before they tip. Just before a turning point, they show characteristic signals — subtle, but possibly measurable.
This phenomenon is often called critical slowing down, and it is generic enough to appear in very different systems: in shallow lakes about to turn turbid, in ecosystems tending towards collapse, in financial markets before a crash — and, as it now seems, also in psychological systems about to undergo a phase transition.
What happens just before a tipping point?
When a system finds itself closer to a tipping point, it becomes less resilient. A perturbation from which the system first recovered quickly now takes longer to fade. This has several measurable consequences:
- Critical slowing down: the system recovers more slowly from perturbations
- Increased autocorrelation: the value at moment t resembles that at t-1 more and more, because the system no longer resets quickly
- Increased variance: the fluctuations become larger; the system swings more strongly
- Changing skewness: the distribution of values shifts, with more extreme outliers in one direction
- Flickering: the system makes brief excursions into the future state, before falling into it definitively
The beauty of these signals is that they are in principle measurable — if one has enough measurements to analyse the time series.
Early warning signals in mental health care
In recent years a Dutch line of research — among others Merlijn Olthof, Fred Hasselman, Marieke Wichers — has begun to examine whether these principles also work in psychotherapeutic processes. The preliminary findings are exciting: in time series of patients with mood disorders they indeed find statistical signals that precede both sudden gains and sudden losses in treatment.
"A marked drop in entropy, or a marked increase in variance, may in time series of patients be a precursor of an approaching tipping — in whichever direction."
This is no longer science fiction. In Schiepek's Synergetic Navigation System, and in emerging ESM-based clinical-feedback systems, we are beginning to actually use these signals to inform therapeutic decisions. Not to predict the future — the system is too complex for that — but to develop sensitivity to moments at which intervention has greater or lesser chance of success.
What it does in the consulting room
Three things change when one learns to look at early warning signals.
First, your reliance on intuition gains a methodological backbone. Many experienced clinicians already feel when something is about to happen with a patient. Sometimes that intuition turns out to be based on precisely the signals that are now quantifiable: increased variability in mood, oscillations between extremes, delayed recovery after sessions. The method confirms and refines what clinicians already know.
Second, you learn to look more patiently at time series. One measurement says little. A trend says more. But the change in variability around the trend may be more crucial than the trend itself.
Third, and perhaps most importantly, you learn to respect tipping moments. A patient about to tip — towards recovery or relapse — is in an unstable state. Small interventions there can have great effects. But large interventions may send the system into chaos. That is no reason to do nothing; it is reason to act with attention.