Metacognitive Skill in Emotion Regulation | Stanford

ELI5/TLDR

Brendan Conway-Smith argues that managing your own emotions is a learnable skill, just like driving a car or playing chess. The first time you try to calm yourself down with a strategy, it’s slow, clumsy, and exhausting. With enough repetition, the strategy moves from “thing you have to think about” to “thing that just happens” — and that handoff frees up your mind to deal with whatever is going on around you. He thinks the standard model of how we get good at physical and mental skills also explains how we get good at not losing it.

The Full Story

Why expertise looks the same in every domain

Conway-Smith starts with a familiar observation: there is a wide gulf between a beginner and an expert at anything, and the gulf has the same shape across wildly different activities. Driving for the first time is slow, jittery, and takes everything you’ve got. Years later you change lanes while holding a conversation. Chess starts as a fight to remember how the knight moves; eventually you “feel” the position. He says emotion regulation works the same way. The first time you try to talk yourself down from a flash of anger, the technique is wooden and you mostly fail. Practice it long enough and it becomes faster, more effective, less effortful, and more flexible.

That last word matters. Flexibility — being able to swap tactics on the fly when the situation changes — is what separates someone who has memorized a technique from someone who actually owns it.

Object level vs meta level

He leans on a framework from Nelson and Narens that splits cognition into two layers. The object level is the part of your mind pointed at the world: drive the car, solve the equation, find the coffee shop. The meta level is the part pointed back at your own mind: steer your attention, calm yourself down, decide whether you actually remember something. Most skill research has studied the first layer. The second is much less mapped, and that is where his work lives.

Two flavors of metacognition

Borrowing the fast/slow split that Kahneman made famous, Conway-Smith says metacognition also comes in two types.

Type one is fast, automatic, driven by feelings. The classic example is the tip-of-the-tongue moment — you have a strong sense you know the actor’s name even though you can’t produce it. John Flavell, who coined the term metacognition, noticed people are surprisingly accurate at predicting whether they’ll recall something. This kind of metacognition is mostly an architectural freebie. You don’t really train it.

Type two is slow, deliberate, conscious, and made of explicit thoughts and instructions. “I feel frustrated” is a type-two fact about yourself. “If I can’t remember a name, sound out the alphabet” is a type-two strategy. “If this conversation gets tense, pause and try to hear the criticism as problem-solving” is a type-two emotion regulation instruction. This is the layer you can actually train.

What “skill” actually means

Skill researchers describe expertise as a high degree of control — the ability to line up your actions with an explicit plan, reliably and flexibly, usually after years of practice. To explain a skill, you have to spell out four things: the goal structure, the action types, the knowledge types, and how the whole thing automatizes.

Goal structure. Cognitive control aims at outward things (win the chess game). Metacognitive control aims at your own mental states as the goal (reduce this anxiety, hold this attention). Subgoals are unavoidable above any level of complexity. Driving breaks down into steering, braking, watching for pedestrians. Reducing anxiety breaks down into picking a strategy, focusing, monitoring whether it’s working, resisting intrusive thoughts.

He also flags work by Craft and colleagues on meta emotion regulation — the layer above the layer. Instead of regulating the emotion directly, you regulate how you regulate.

“Before the meeting, they’d set a rule, ‘Okay, if this discussion gets tense, I will pause, breathe, and try to reinterpret the comment as problem-solving rather than personal criticism.’”

The rule isn’t the calming move itself. It’s the pre-commitment that makes the calming move actually fire when you need it.

Clustered actions. Real skill is never one move. A basketball player chains footwork, dribble, and layup. Someone regulating emotion in real time might be clearing the mind, focusing, and reappraising at the same time — what the literature calls polyregulation.

Knowledge types. This is the most important distinction. Declarative knowledge is the explicit stuff — rules, facts, tactics — and it lives in the temporal medial lobe and working memory. It tells you what to do but doesn’t actually do it. Procedural knowledge is the implicit machinery in the basal ganglia that actually executes the action. You can recite the rules of tennis without being able to return a serve. The two systems work together: declarative knowledge tells you which tactic to pick, procedural knowledge runs it.

Proceduralization — the engine of the whole thing

Here is the load-bearing claim. As you repeatedly practice a slow declarative strategy, it gets gradually replaced by fast procedural operations. This is proceduralization, and it explains how metacognition stops being effortful.

The arc has three stages. Early on, performance is conscious, deliberate, error-prone, and eats your working memory. In the middle, procedural knowledge starts building up — speed goes up, errors come down. In the automatic stage, the action runs largely on procedural knowledge, outside working memory, with very little cognitive demand. The reaction-time curve follows a power law: steep early gains that gradually flatten as you approach a ceiling. He has found this same power-law shape in attentional control and metamemory learning, and he wants to test whether emotion regulation shows it too.

Conway-Smith uses this lens to make sense of three established phenomena. In the attentional training technique, you automatize the move of catching yourself in rumination and stepping away from it, instead of getting dragged along. In the metacognitive threshold, mindfulness practice trains procedural knowledge that detects subtler emotional signals, which lets you intervene earlier. In detached mindfulness, equanimity replaces a reactive procedural habit (getting upset about being upset) with a different procedural habit — noticing the feeling change moment to moment without grabbing onto it.

Cognitive reinvestment

There is a payoff to all this offloading. Once an emotion-regulation move runs without working memory, you get that working memory back. Skill researchers call this cognitive reinvestment — you can spend the freed-up bandwidth on planning, error-monitoring, and adapting to whatever new mess is in front of you. This is also why polyregulation gets easier with practice. Once each individual move is automatic, you can stack several at once: reappraisal, attention control, and mindfulness running in parallel, mutually supporting each other, instead of each fighting for the same scarce conscious attention.

Key Takeaways

Metacognition is your mind’s ability to monitor and control its own states — attention, memory, emotion. It is a separate layer above world-facing cognition.
The framework comes from Nelson and Narens: object level (cognition pointed at the world) vs meta level (cognition pointed at itself).
Type-one metacognition (feelings of knowing, tip-of-the-tongue) is automatic and basically untrainable. Type-two metacognition (explicit strategies and rules) is the trainable layer.
All skill — physical, cognitive, metacognitive — gets faster, more accurate, less effortful, and more flexible with practice. The shared shape is the giveaway that the same mechanism is running underneath.
Declarative knowledge (explicit rules, in working memory and temporal medial lobe) tells you what to do. Procedural knowledge (implicit, in the basal ganglia) actually executes the action. The two are dissociable — you can know the rules without being able to do the move.
Proceduralization is the gradual handoff from declarative to procedural. It is the mechanism that makes hard mental moves eventually become automatic.
Skill learning across domains follows a power law — sharp early improvement that flattens toward a ceiling. Conway-Smith predicts emotion regulation will show the same curve if measured properly.
Cognitive reinvestment: once a regulation tactic runs on autopilot, the working memory it used to consume becomes available for higher-level monitoring, planning, and adapting.
Polyregulation — running multiple regulation strategies at once — only becomes possible after each strategy is automatic enough not to compete for working memory.
Meta emotion regulation (Craft et al.) is the move of pre-committing to how you’ll regulate. The if-then rule set before the meeting, not the calming move during it.
Equanimity is reframed here as a kind of reappraisal — a second-order attitudinal stance toward your own affect, where you watch the feeling change moment to moment instead of grabbing onto it.
Mental health is increasingly framed as a deficit in metacognitive skill rather than a fixed trait, which is why CBT-style interventions emphasize repeated practice of specific tactics like cognitive restructuring and mindfulness.
Childhood emotion regulation can outperform IQ as a predictor of long-run learning and performance outcomes.

Claude’s Take

This is a research talk to a Stanford lab, not a polished podcast, and it shows — Conway-Smith races the clock and pitches his postdoc plans in the last five minutes. The actual content is solid though. The core move — porting motor-skill learning theory (declarative-to-procedural, power law, proceduralization) onto emotion regulation — is the kind of thing that sounds obvious once you hear it but apparently nobody had formally done. That counts for something.

The framework is genuinely useful as a way to think about what mindfulness and CBT are actually doing under the hood. The reframe of equanimity as “training a different procedural habit so the meta-emotion doesn’t fire” is a clean explanation of something that mindfulness teachers usually wave at with metaphor. Same with the cognitive-reinvestment story, which gives a concrete reason why people who practice regulation get better at handling chaos — not because they’re calmer, but because they have working memory free to actually think.

The weak spot is that almost none of this is yet empirically tested in emotion regulation specifically. He is up front that this is the proposed work, not the finished work. The power-law prediction has been confirmed in attentional control and metamemory but is still a hypothesis for emotion. So treat the talk as a well-argued theoretical scaffold, not as established findings. Seven out of ten — clear, useful, slightly under-baked because the science is still ahead of him.