When we analyzed anonymized aggregate data from more than 54,000 classroom observations across nearly 300 schools during the 2024–25 school year, one finding stood out from the rest. Not because it was surprising — most educators would predict it — but because of how consistent it was. Across districts, across grade levels, across subjects, across the entire school year.
The finding was about questions. Specifically, the cognitive level of the questions teachers ask during instruction. And the data was unambiguous: the vast majority of classroom questions never get past recall.
What the Numbers Show
In more than 38,000 observations where questioning level was recorded, observers categorized the predominant level of questions using a framework aligned with Bloom’s taxonomy — the cognitive hierarchy first described by Benjamin Bloom in 1956 and revised by Anderson and Krathwohl in 2001. The three levels roughly map to:
Level 1 — Knowledge & Comprehension. Recall and basic understanding. "What year did the Civil War start?" "What is the formula for area?" "What happened in chapter three?" The student retrieves information. The cognitive demand is low.
Level 2 — Application & Analysis. Using knowledge in context. "Why do you think the character made that choice?" "How would you solve this if the dimensions changed?" "What pattern do you notice in the data?" The student reasons, compares, or applies.
Level 3 — Synthesis & Evaluation. Creating, judging, designing. "How would you redesign this experiment to test a different variable?" "Which argument is strongest, and why?" "What would happen if this historical event had gone differently?" The student generates something new or defends a position.
Here's the distribution across those 38,000+ observations:
58% were Level 1. 36% were Level 2. 6% were Level 3.
Nearly six in ten observed lessons were dominated by recall-level questions. Barely one in sixteen reached the highest cognitive level. This wasn't a finding from a single school or a single district. It was the pattern across nearly 300 schools, more than 1,100 observers, and an entire school year of data.
Engaged Isn't the Same as Thinking
Here's what makes this finding harder to see from inside the classroom: engagement looks fine.
In the same dataset, 89% of classrooms were rated as having "most students engaged." Students were on task, following directions, raising hands, completing work. By every visible measure, instruction was working.
But when observers asked students directly whether they understood the learning objective, the picture changed. In observations where students were polled, roughly one in three said they didn't understand what they were supposed to be learning. The engagement was real. The comprehension wasn't always there.
This is the gap that questioning level exposes. A classroom full of students answering recall questions correctly looks engaged. They're responding. They're participating. But the cognitive work is shallow — retrieval, not reasoning. Students can be busy and compliant without ever being asked to think deeply about the material. John Hattie's research on visible learning consistently identifies the quality of questioning — not just its presence — as one of the highest-leverage instructional practices available to teachers.
Why It's So Persistent
If questioning level were easy to change, it would have changed. The 58/36/6 distribution isn't a snapshot — it's a year-long constant.
In September, Level 1 questions accounted for 59% of observations. By April, after eight months of instruction, observation, and feedback, the number had dropped to 55%. Level 3 moved from 6.7% to 7.7%. For teachers observed ten or more times over the year, Level 2+ questioning improved by just 2.6 percentage points between their first and last observations. The pattern barely budged.
This isn't because teachers lack skill. It's because lower-level questions are structurally easier in every way that matters during live instruction. They're faster to ask. They produce quicker responses. They're easier to manage — one correct answer, move on, maintain pacing. In a lesson with thirty students and forty-five minutes, a teacher asking higher-order questions has to wait longer for responses, manage ambiguity, and facilitate discussion rather than check answers. The time cost is real.
The data reflects this. In our dataset, 38% of classrooms were rated as teacher-talk-heavy, with an 80/20 talk ratio favoring the teacher. In ELA classrooms, that number reached 54%. When the teacher is doing most of the talking, there's structurally less room for the kind of student reasoning that higher-order questions demand. "Think time and wait time" was the second most commonly observed instructional strategy — noted in nearly 13,000 observations — which suggests that teachers are trying to create space. But the dominant pattern is still ask-answer-move on.
It's also worth noting what observers selected when coding instructional strategies. "Reinforces effort and provides recognition" was the most common strategy observed, at more than 15,000 instances. "Higher-order questioning" was selected fewer than 4,600 times. Teachers are doing many things well. Higher-order questioning just isn't one of the things the data shows happening frequently.
What Observers Can Do
Most walkthrough forms include a questioning-level indicator. It's the kind of item that's easy to check — glance at the questions being asked, select a level, move on. But the most useful observation feedback on questioning isn't about the level. It's about what happens after the question is asked.
During a five-minute walkthrough, an observer can notice: Did the teacher wait after asking the question, or immediately call on the first raised hand? Did students talk to each other before responding? Did the teacher follow up with a probe — "Tell me more," "Why do you think that?" — or accept the first answer and move on? Was there a moment where a student's thinking was visibly stretched?
These are observable in a brief visit. And feedback about one specific question — "I noticed you asked students to compare the two characters, and then gave them thirty seconds to discuss before calling on anyone — that wait time produced much richer answers" — is more actionable than a general note that questioning was at Level 1.
The Gates Foundation's MET Project found that observations are most useful when they generate specific, concrete feedback rather than summary ratings. A questioning-level checkbox captures the pattern. A specific observation about a specific question captures the practice.
What Teachers Can Do
The data doesn't suggest that every question needs to be Level 3. Recall questions have a purpose — checking foundational knowledge, building fluency, establishing shared understanding before going deeper. The issue is proportion: when 58% of instruction stays at recall, students rarely practice the kind of thinking that transfer, critical analysis, and problem-solving require.
The most practical shift isn't to overhaul questioning overnight. It's to plan one higher-order question per lesson. One question that asks students to compare, evaluate, create, or defend — written in advance, not improvised in the moment. Improvised questions default to recall because recall is cognitively easiest for the teacher, too. A planned question can be Level 3 by design.
The data shows that think-pair-share and wait time are already widely practiced — they appeared in nearly 13,000 and 5,600 observations respectively. These are the structures that make higher-order questions work. A Level 3 question asked to a silent room produces anxiety. The same question asked after thirty seconds of partner discussion produces thinking. The infrastructure is already in many classrooms. The missing piece is the question itself.
It also helps to notice which subjects are most affected. In the data, ELA and Language Arts classrooms had the highest rates of teacher-heavy talk ratios — more than 50% at an 80/20 split. These are the classrooms where the content most naturally supports analysis, argument, and interpretation, yet the instructional pattern skews toward teacher-led discussion. A single shift — replacing one teacher-posed comprehension question with a student-debated interpretive question — can change the cognitive demand of an entire lesson segment.
Start With What You Can See
The questioning gap is one of the most consistent patterns in our data, and one of the most addressable. It doesn't require new curriculum, new training, or new technology. It requires attention — from observers who notice the level of questions being asked, and from teachers who plan one question per lesson that pushes students past recall.
Aprenta's pre-built walkthrough forms include questioning-level indicators alongside space for specific, narrative feedback about what the observer saw. The combination — a quick classification plus a concrete note — gives teachers something they can actually use. Try Aprenta free and start making classroom questioning visible.