When Mistakes Are the Map:

Oliver Ringrose
Mar 29
5 min read

Before I became a dog trainer, I was an automation engineer. I worked with complex control systems—essentially wiring up little artificial brains to run industrial processes. I didn’t always have diagrams, so I’d spend my days fault-finding: visualising how the system should work, tracing signals, watching outputs, and figuring out where it all went a bit “gremlin in the wires.”

Truth be told, I learned the most by making mistakes.

And not just oops mistakes. I mean glorious, confusing, pull-it-all-apart-and-start-again mistakes. The kind of mistakes that force you to think deeper, look closer, and ultimately rewire your own internal systems.

That hands-on, trial-and-error learning didn’t just teach me engineering—it trained my brain to think in processes. Now I find myself doing something eerily similar in dog training… only the systems are fuzzier, messier, and come with significantly more feelings (and fur).

💥 The Study That Sparked All This

A 2024 study from Ruhr University Bochum looked at how animals learn tasks—and it caught my attention like a loose wire catches a sleeve.

Researchers found that mice who made mistakes during learning actually activated more brain regions than mice who didn’t. Makes sense: if everything goes smoothly, the brain can coast. But errors? Errors make the brain work.

Here’s the kicker: The part of the brain that lit up the most wasn’t the “thinking” area. It was the sensory cortex—the regions that process vision, touch, and body awareness.

Translation?

The brain isn’t just thinking harder during mistakes—it’s feeling harder.

It’s reprocessing input, adjusting internal models, and fine-tuning how it interprets the world—not through logic, but through sensation.

And suddenly, the way I see learning—human, dog, or otherwise—shifted entirely.

🧠 Not Just Thinking—Simulating

Temple Grandin writes in Visual Thinking that visual learners don’t always “think in pictures.” Some of us (hi!) think in processes—mental simulations that unfold like animated flowcharts. Not static images, but dynamic models. A kind of brain-based CAD software, but messier and with more coffee.

That was my experience as an engineer: I could “see” the system in motion, even if it was all happening inside my head.

And it’s how I still approach training. I mentally run the behaviour, tweak a variable, rewind, replay.

It turns out this is actually a thing. Neuroscience shows that when people imagine moving, solving, or sensing, their sensory and motor cortices activate—as if they’re doing the thing for real. The brain can’t always tell the difference between doing and rehearsing. This is known as embodied cognition, and it's kind of the brain’s secret rehearsal room.

Which brings me to dogs.

When I see a dog pause mid-training—still body, twitching ears, narrowed eyes—I no longer see that as stuck.

They’re buffering. They’re simulating. They’re replaying sensory data to figure out what comes next. They’re searching for the correct process that will produce the desired outcome.

I think they’re running their own internal macro. And when they move again, nine times out of ten? Nailed it.

And from personal experience—that’s exactly why I don’t like creating formal plans or rigid blueprints in advance. With living beings, it’s almost impossible to do that ethically or effectively. I can build a plan after an experimental process with the dog, once I’ve seen how they think, feel, and respond—but even then, that plan is always subject to change. It has to be.

Because animals, like humans, are unique systems—not templates. They come with their own wiring, thresholds, history, and weird, wonderful variables. And I think training should reflect that.

🧩 Sensory Inputs Are Variable. So Is the Brain.

Here’s the juicy bit my brain’s been chewing on like a rawhide for the past few weeks:

In the real world, sensory inputs are never exactly the same. A cue today isn’t the same cue tomorrow—the lighting is different, the wind’s changed, you’re wearing a different jacket (and yes, they notice). The dog’s output changes too—movement, timing, arousal, motivation.

So what really matters? Not just input or output… but how the data is processed in between.

The brain has to find a stable, repeatable process that works across unstable and variable conditions.

Which is where my engineering brain starts tingling again.

In software terms, I think of this as the brain building multiple macros—pre-packaged sequences of input–processing–output that can be flexibly pulled from depending on what’s coming in.

But here’s the bit I love: even once the brain finds a process that works, it doesn’t stop experimenting.

Why? Because brains are efficiency nerds.

They’re constantly testing, tweaking, and re-running loops to find better, faster, or less effortful ways to solve the same problem.

So even after success, animals continue testing small variations within the same realm of behaviour—something supported by research into motor learning, prediction error, and the exploration vs exploitation trade-off in reinforcement models.

What looks like “being inconsistent” might actually be optimisation in motion.

The dog isn’t confused—they’re refining their algorithm.

🧠💬 A Theory My Brain Is Currently Pondering (and Poking at Relentlessly)

Learning—especially through trial-and-error—isn’t just a cognitive act of remembering. It’s a sensory simulation process where the brain collects data, tests variables, and refines internal macros to solve problems across variable real-world contexts.

It’s messy. It’s dynamic. It’s not plug-and-play.

And that’s why cookie-cutter training plans so often fall flat: they’re based on a single assumed path through the behaviour, rather than recognising that each dog—and each moment—might follow a slightly different macro.

This theory reminds me daily that we aren’t writing scripts for robots.

We’re collaborating with sentient systems that learn by doing, testing, pausing, and adjusting.

It’s slow sometimes. It’s nonlinear. But it’s real learning. And the results are incredible.

🐾 What This Means for Dog Training

I’ve used both zero-error and trial-and-error approaches in training, and here’s where I’ve landed:

✅ Zero-error learning is incredibly useful for anxious, fragile, or inexperienced dogs. It creates safety and avoids stress.

✅ But trial-and-error—done ethically, patiently, and with emotional support—might just create deeper, more flexible, and more robust learning.

Dogs that get to problem-solve become better thinkers.

They become more adaptable.

And they’re more prepared for real-world variability.

And it’s why I work experimentally. Every training plan I make is a living document, written after I’ve observed how the dog processes, simulates, and responds. And even then?

That plan stays flexible, it remains organic.

Because dogs aren’t software—they’re systems. Living, learning, adapting systems. And systems need flow, not force.

My rule of thumb? If a dog fails three times, it’s not a stubborn dog—it’s a training plan that needs rewriting. But I want dogs to be able to experiment, to buffer, to process.

Because error isn’t the enemy. Error is feedback. Error is sensory data. Error is how brains build better macros.

I encourage you to reflect on what learning really is—for dogs, and for us. What are you teaching? And what’s actually being learned?Let’s stay curious. 🧠✨

🧠💻 A Confession (and a Bit About My AI Co-Pilot)

Okay, full transparency: I didn’t write this entirely alone.

I’ve been nerding out with a canine research assistant—a very polite, unfailingly enthusiastic AI trained in behaviour science, neurology, and canine cognition. It’s like talking to Temple Grandin’s brain if it were cross-bred with a neuroscience library and didn’t need to sleep.

They don’t get bored, they don’t roll their eyes at me asking “but what if it’s a motor loop running a sensory simulation with a prediction error in the basal ganglia?” And they never once say, “Can we please talk about something normal?”

Honestly? Peak co-regulation.

📚 References (for my fellow nerds and note-takers)

Ruhr University Bochum (2024). Trial-and-error learning activates the sensory cortex. Neuroscience News.
Grandin, T. (2022). Visual Thinking: The Hidden Gifts of People Who Think in Pictures, Patterns, and Abstractions.
Kosslyn, S.M., et al. (2001). Neural foundations of imagery. Nature Reviews Neuroscience, 2(9), 635–642.
Jeannerod, M. (2006). Motor Cognition: What Actions Tell the Self.
Berns, G.S., et al. (2012). Functional MRI in awake dogs: Reward processing in the canine brain. PLoS ONE, 7(5), e38027.
Cisek, P. & Kalaska, J.F. (2010). Neural mechanisms for interacting with a world full of action choices. Annual Review of Neuroscience, 33, 269–298.