An administrator once observed one of my classes and told me afterward, not unkindly, that I hadn’t really been doing anything special. Students were focused and working. I was moving between tables, asking questions, mostly staying quiet. Nothing was being explained at the front of the room. From where they sat, it looked like a class running on autopilot. There was nothing outstanding to record on my evaluation.
They were right about what they saw. They were wrong about what it meant. That lesson was the product of months of scaffolding. Some days students arrive and barely acknowledge me; they go get what they need and pick up where they left off last lesson. The day’s objective is on the board, but they already know what it is, because they’ve had agency in my classroom long enough to expect it. Those are often the most boring days for me to teach, my motivational spiel and dad jokes wasted. They’re also the days I know I’ve done my job well. The less visible I was, the better the teaching had been. What looked like nothing happening was the entire point.
This is the structural problem with the single-lesson observation, and it isn’t limited to art, though art makes the problem unusually visible. Most classroom observation models assume good teaching is legible in forty minutes: a clear objective, visible instruction, active engagement, a tidy arc from start to finish. That model works reasonably well for content that’s delivered and checked within a single period. It works badly for anything built on a multi-week arc, where the highest-value work, the sequencing, the scaffolding, the deliberate withholding of help so students build the capacity to work without it, happened in the planning, not in the room, and became invisible by design as students gained independence. I suspect this isn’t unique to art, though I can’t speak to other subjects with the authority thirty years in a studio gives me. But any subject built around sustained projects rather than discrete daily content is vulnerable to the same blind spot, and observation models built for the discrete-content case get used on it anyway, because it’s the model schools have.
The best evaluators I’ve had in my career know this. They didn’t spend the observation talking to me. They walked over to students and asked what they were doing and why. Not “what does the rubric say” or “what’s today’s objective,” just, in their own words, what are you working on and why does it matter. A student who can answer that clearly, without checking in with the teacher first, is showing you something a lesson plan can’t: that the thinking has transferred into their hands, not just been demonstrated in front of them.
That’s the real diagnostic, and it’s a stricter one than it sounds. It’s easy for a teacher to perform engagement for an observer for forty minutes. It’s much harder to fake three years of scaffolding so effective that a fifteen-year-old can explain, unprompted, what they’re doing and why it matters. If a leader wants to know whether an art program, or any long-arc program, is building independent thinking rather than performing it, the fastest test isn’t watching the teacher teach. It’s finding out whether the students still need them in the room.
I don’t think this makes the case for abandoning classroom observations altogether; they still catch things worth catching. But treating a single lesson as a verdict on a program built across months gets the evidence backward. The quietest, least eventful class I ever taught was also, by a wide margin, one of my best. An observation model that can’t tell the difference between that and a class running on autopilot isn’t measuring what it thinks it’s measuring.