Tag: differentiation

Scaffolding Strategies for Grades 6–12: Eight That Work, and When to Take Them Away

Scaffolding is temporary support that lets a student do a task they could not yet do alone, and that is removed as they get better at it. Fading is part of the definition, not an optional extra. The scaffolding strategies that work in grades 6–12 are cheap: modeling, cue cards, checklists, worked models, and warning students about likely errors.

Key Takeaways

A support that never comes off is not a scaffold. Reviewers of the research name three characteristics: contingency, fading, and transfer of responsibility. A sentence starter handed out every Monday for a year meets the first and fails the other two.
Scaffolding and differentiation are not the same thing. A scaffold is temporary and aimed at a task. Differentiation adjusts for a difference that is still going to be there next month. Mixing them up is how a scaffold quietly becomes a lower standard.
Rosenshine names eight concrete scaffolds, and none of them costs money. Modeling, thinking aloud, cue cards, checklists, a completed model to compare against, reading prompts, an organising prompt, and anticipating likely errors.
Helping too fast can make high-quality support worse than low-quality support. In a study of 768 students aged 12 to 15, when teachers intervened frequently, less contingent support produced higher achievement and task effort than highly contingent support.
A scaffold that helped in September can hurt in November. The expertise reversal effect is the finding that guidance which helps novices becomes redundant load for learners who already have the schema, and gradual fading beats an abrupt switch.

Walk into most secondary classrooms and you will find scaffolds. Sentence starters taped inside a folder, a graphic organiser for every essay, a word bank on the board since August. What you will rarely find is a plan for taking any of it away. That is the part of scaffolding that almost nobody runs, and it is in the definition.

This guide covers what scaffolding actually means in the research, how it differs from differentiation, eight specific strategies for grades 6–12, what the evidence does and does not show, and the harder question of when to remove a support. The strategies and sequences are compiled from the sources at the end and from the structures those sources describe. Read them as recommendations, not as a record of any one classroom.

What Is Scaffolding in Education?

Scaffolding is a temporary support a teacher provides so a learner can complete a task that is currently beyond them, withdrawn gradually as the learner becomes competent. Barak Rosenshine defines a scaffold plainly as “a temporary support that is used to assist a learner,” and says scaffolds are “gradually withdrawn as learners become more competent.”

The word comes from a 1976 tutoring study by David Wood, Jerome Bruner and Gail Ross. It is worth knowing what that study actually was, because it gets cited as though it were a classroom trial. According to a research briefing summarising the paper, it involved 30 children aged three, four and five, ten per age band, each working one-to-one with a single tutor to build a wooden block pyramid. Direct hands-on help dropped from a median of 12 instances per child at age three to 3 at age five, and constructions completed with no help rose from 64.5 percent to 87.5 percent.

That is a descriptive study of preschoolers with no comparison group. It named something real and it cannot tell you anything about a ninth grader writing a thesis statement. The briefing says so itself. The origin story is not the evidence base.

The three characteristics that define a scaffold

Janneke van de Pol, Monique Volman and Jos Beishuizen reviewed 66 articles published between 1998 and 2009 in primary and secondary education, and pulled out three characteristics that recur across the definitions.

Graphic defining the three characteristics that make a support a scaffold - contingency, fading and transfer of responsibility - each with a sign that it is missing.

Characteristic	What it means	How you can tell it is missing
Contingency	Support is adapted to the student’s current level of performance — at that level or slightly above	Everyone gets the same handout regardless of what they can already do
Fading	The level or amount of support decreases over time	The support looks identical in week two and week twenty
Transfer of responsibility	The student takes increasing control of the task, cognitively and metacognitively	Students still wait to be told which support to use

The three characteristics as defined in van de Pol, Volman and Beishuizen (2010). The right-hand column is this guide’s practical restatement, not part of their framework.

Contingency is worth one extra note, because it has a prerequisite nobody mentions. Support can only be adapted to a student’s current level if you know what that level is this period, which makes checking for understanding the thing that makes scaffolding possible at all. A scaffold chosen from last month’s data is not contingent on anything.

The reviewers’ own finding about these three is the uncomfortable one. Contingency was operationalised far more often than fading or transfer of responsibility. Many studies named all three in their definition and then measured only the first. Establishing that fading happened, they note, requires several observations over time and ideally across different tasks. So the research literature has the same blind spot the average classroom does: everybody agrees support should come off, and almost nobody checks whether it did.

What Is the Difference Between Scaffolding and Differentiation?

A scaffold is temporary and attached to a task; differentiation is an ongoing adjustment for a difference between students. Scaffolding assumes the support comes off. Differentiation does not. Treating a permanent accommodation as a scaffold is how the support stops being support and becomes a different, lower standard.

	Scaffolding	Differentiation
What it responds to	A task the student cannot yet do alone	A difference in readiness, interest or profile
Time horizon	Temporary, by definition	Ongoing for as long as the difference persists
Success looks like	The student no longer needs it	The student can access the same standard
The standard itself	Unchanged	Unchanged
Main failure mode	It never gets removed	It becomes a permanent lower track

A definitional comparison drawn from the scaffolding characteristics in van de Pol and colleagues’ review. The distinction is conceptual, not a research finding.

Both are legitimate and they overlap in practice. An accommodation written into an IEP or a 504 plan is not a scaffold and is not a teacher’s to fade. It is a documented entitlement, changed through the team that wrote it, and “they seem to be doing better” is not the process. The problem this article is about is the reverse case: a temporary support that was never meant to be permanent, quietly becoming permanent because removing it would be awkward. If what you need is the full picture of adjusting instruction for a mixed class — including the honest question of how you do any of this with 150 students — differentiated instruction strategies is the page for that. This one is about the supports that are supposed to disappear.

Eight Scaffolding Strategies That Work in Grades 6–12

Rosenshine’s “provide scaffolds for difficult tasks” principle names eight specific supports. All eight are things a teacher can build in a planning period, and all eight are designed to be removed.

Graphic listing eight scaffolding strategies named by Rosenshine: modeling the steps, thinking aloud, cue cards, checklists, a completed model to compare against, who-why-how reading prompts, a step-by-step organising prompt, and anticipating likely errors.

Model the steps yourself. Do the task in front of them, in the order you want it done. Not a description of the task — the task. This is the cheapest scaffold there is and the one most often replaced by instructions.
Think aloud while you solve it. Narrate the decisions, including the dead ends. Students who have only ever seen finished work assume competent people do it right first time, which is the single most discouraging belief in a classroom.
Cue cards that complete part of the task. Rosenshine’s framing is important: a cue card does some of the work, so the student can spend their attention on the part you are teaching. A card listing transition words while you teach argument structure is doing its job.
Checklists students use to judge their own work. Rosenshine’s version has the teacher model using the checklist first. An unmodelled checklist gets ticked without being read.
A model of the completed task to compare against. Not an exemplar to copy — a finished piece students hold their own work next to. This is the scaffold that makes self-assessment possible at all, because a student cannot evaluate against a standard they have never seen.
Prompts for generating questions while reading. Rosenshine names “who,” “why” and “how” as the prompts. Three words on the board, and the comprehension task changes from reading to interrogating.
A step-by-step prompt for organising material. His example is drawing a central box and skimming for main ideas. The scaffold is the procedure, not the graphic organiser you photocopied.
Anticipate the errors and warn them first. Tell students the mistake most people make here, or show it. This is the one teachers skip because it feels like putting the idea in their heads, and it is the one that saves the most re-teaching.

Rosenshine groups all of this under cognitive apprenticeship, where a master models, coaches and supports until the learner can work alone, and he is explicit that providing scaffolds is a form of guided practice rather than a substitute for it. He also notes something honest: students may still reach for a scaffold on an especially hard problem even after they have outgrown it on ordinary ones. That is not regression. It is what a scaffold is for.

Two adjacent problems have their own pages. If the students you are scaffolding cannot start a task because they cannot plan one, that is a different lever — executive functioning strategies for students covers it. If they do not know what success looks like in the first place, start with learning targets, because a scaffold toward an unclear goal is just extra paper.

Does Scaffolding Actually Improve Achievement?

The evidence is positive for thinking and learning activities and genuinely mixed for achievement. Reviewers found only eight effectiveness studies in a decade of research, most of them one-to-one tutoring with simple tasks. The one large secondary-classroom experiment found scaffolding is “not unequivocally effective.”

Van de Pol and colleagues’ review of a decade of scaffolding research concluded that results on students’ cognitive and metacognitive activities indicate scaffolding is effective, while results on affect were less consistent. Their 66 articles break down into 27 theoretical, 26 descriptive and only eight effectiveness studies, some with very small samples — nine students in one case. Their own recommendation was for more research in naturalistic classrooms with a wider range of tasks.

Five years later the same lead author ran something much closer to a real classroom, and the result is the most useful finding in this article. In a study published in Instructional Science, 30 social studies teachers and 768 students aged 12 to 15 worked through a five-lesson project on the European Union. The researchers measured how contingent each teacher’s support was and how long students worked independently before the teacher intervened.

Graphic summarising a study of 768 students aged 12 to 15, showing that when teachers helped often, low contingent support produced higher achievement than high contingent support, and the reverse when students worked independently for longer.

What they found runs against the theory. When independent working time was short — that is, when teachers helped frequently — low contingent support produced higher achievement and task effort than high contingent support. Only when students worked independently for longer did high contingent support come out ahead on achievement, after correcting for task effort. Students appreciated highly contingent support more either way. The authors’ conclusion is that scaffolding is not unequivocally effective and that its value depends on independent working time and student effort.

Read carefully, that is not an argument against scaffolding. It is an argument that when you intervene matters at least as much as how well. Carefully tuned help delivered constantly appears to do less good than blunter help delivered after students have struggled for a while. One study in one subject with one age band cannot settle that, and this is pre-vocational social studies in a five-lesson project, not your course. But it is the closest thing in this literature to a secondary classroom, and it points the same way as the fading research below. The piece on productive struggle in grades 6–12 walks through that timing question on its own terms.

When Should You Take a Scaffold Away?

As soon as the student can do the part the scaffold was doing, and gradually rather than all at once. The expertise reversal effect is the finding that instructional support which helps a novice can stop helping, or start hurting, once the learner has built the relevant knowledge.

Slava Kalyuga, Paul Ayres, Paul Chandler and John Sweller set this out in “The expertise reversal effect” in Educational Psychologist. The mechanism is working memory. A novice has no schema, so explicit guidance helps them build one. A learner who already has the schema has to spend working memory cross-referencing your support against what they already know — load with no return.

The studies they summarise are mostly adult and vocational: electrical trainees, mechanical trade apprentices, industrial machinery operators, plus one high school biology study where added text coherence helped low-knowledge readers and the original, less coherent text served high-knowledge readers better. The paper reports no effect sizes and most of its samples are not secondary students. The pattern is consistent across eight separate reversals, which is what makes it worth taking seriously, but nobody should present it as a finding about teenagers.

Their practical recommendations are three, and they are the whole fading argument in a paragraph. Match the support to the learner’s experience, because a design that ignores experience produces inconsistent results and can backfire. Reduce guidance as knowledge grows, since experienced learners generally do better with minimal guidance. And fade gradually rather than switching abruptly — they cite Renkl and colleagues finding that a gradual move from worked examples to independent problems outperformed an abrupt one.

How Do You Fade a Scaffold Without Losing Half the Class?

Fade one component at a time, on a schedule you set in advance, and keep the scaffold available for the hardest version of the task. Deciding in advance is the part that matters, because a scaffold with no removal date is a scaffold that stays.

Write down what the scaffold is doing for the student. One sentence. “This organiser is holding the structure so they can concentrate on evidence.” If you cannot write it, you cannot tell when it is no longer needed.
Name the evidence that it has done its job. Not a feeling. A specific thing a student produces — a paragraph with the structure in it and no organiser on the desk.
Remove one component, not the whole support. Drop the sentence starters and keep the headings. Keep the checklist and stop modelling it. Abrupt removal is the version the research says works least well.
Lengthen independent working time before you lengthen the task. This is the lever the 768-student study points at. Let students work longer before you intervene, and intervene well when you do.
Make the scaffold available rather than issued. On a shelf, in a folder, on request. A student choosing to pick up a cue card is transfer of responsibility; the same card on every desk is not. This also solves the fairness problem that contingency creates: if support is available to everyone and chosen by each student, nobody has been publicly handed the easier version.
Keep it for the hard version. Rosenshine’s point stands: reaching for a scaffold on an unusually difficult problem is appropriate use, not relapse. Fade it for the routine task and leave it in reach for the stretch one.

The step students notice most is the fifth. Moving a support from issued to available changes who is deciding, and that decision is the thing you were trying to teach the whole time.

Where Does Scaffolding Go Wrong?

Three failure modes cover nearly all of it: the support never comes off, the support quietly replaces the standard, and the teacher intervenes before the student has done any thinking worth supporting.

The permanent scaffold. Nobody decided to keep it forever; no date was ever set. Fix: set the removal condition at the same time you build the scaffold, in writing.
Scaffolding the standard instead of the task. The organiser gets so complete that filling it in is the assignment. Fix: check whether a student could score well on your rubric without doing the thinking the standard names. If they could, the scaffold has eaten the task.
Helping too fast. The teacher circulates, sees a stalled student, and solves it. Fix: the fourth step above. The secondary-classroom study found frequent intervention undercut even well-tuned support.
The same scaffold for everyone. Which fails the contingency characteristic outright. Fix: two versions, or one version offered rather than issued.
No way to tell whether fading happened. The review’s own blind spot. Fix: one dated note per unit on what came off. Establishing fading takes observations over time, and a memory is not an observation.

What to Do Next Week

Pick one scaffold already in use, write down what it is doing and what evidence would show it is no longer needed, then remove one component of it. That is the whole intervention, and it costs about ten minutes.

Clay’s take, as a grades 6–12 teacher: the honest reading of this research is that scaffolding is not the high-leverage move it gets sold as. Eight effectiveness studies in a decade, mostly one-to-one tutoring, and the one large secondary experiment finding that helping frequently made good support worse — that is not a reason to stop scaffolding. It is a reason to stop treating a stack of graphic organisers as evidence of good teaching.

The part that matters for how students are treated is the fading. A support you never remove tells a sixteen-year-old you do not expect them to get past needing it. Taking it away on a plan, telling them that is the plan, and leaving it in reach for the hard version tells them the opposite. Holding a high standard and handing over the support are the same act, done in the right order. That is the argument for high expectations, applied to paper.

And if a student is far enough behind that no scaffold is reaching them, the conversation is a different one. How to talk to a student who is failing covers that, and it is a better use of a planning period than a sixth organiser.

Sources

Van de Pol, J., Volman, M., & Beishuizen, J. (2010). Scaffolding in teacher–student interaction: A decade of research. Educational Psychology Review, 22, 271–296. Review of 66 articles from 1998 to 2009 in primary and secondary education; only 8 were effectiveness studies, most one-to-one tutoring with simple tasks, one with 9 students. The authors state there is no agreed definition of scaffolding and no instrument covering all three characteristics. Read in full, via the author’s thesis chapter.
Van de Pol, J., Volman, M., Oort, F., & Beishuizen, J. (2015). The effects of scaffolding in the classroom: Support contingency and student independent working time in relation to student achievement, task effort and appreciation of support. Instructional Science, 43(5), 615–641. 30 social studies teachers, 768 students aged 12–15, pre-vocational education, a five-lesson project. The authors state scaffolding is “not unequivocally effective.” Abstract and findings read; the full paper is paywalled and was not read.
Kalyuga, S., Ayres, P., Chandler, P., & Sweller, J. (2003). The expertise reversal effect. Educational Psychologist, 38(1), 23–31. A review article summarising eight reversals across other researchers’ studies. It reports no effect sizes, and most samples are adult trainees and apprentices rather than school students; one high school biology study is named. Read in full.
Rosenshine, B. (2012). Principles of instruction: Research-based strategies that all teachers should know. American Educator, 36(1), 12–19, 39. Source of the scaffold definition and all eight named scaffolds. A practitioner synthesis of three research areas, not a study; adapted from the author’s 2010 report for the International Academy of Education. Read in full.
Wood, D., Bruner, J. S., & Ross, G. (1976). The role of tutoring in problem solving. Journal of Child Psychology and Psychiatry, 17(2), 89–100. I read a summary, not the paper: Support that fades, Structural Learning research briefing. The study involved 30 children aged three to five, one-to-one, with no comparison group; the briefing states it describes a pattern rather than showing that fading caused the gains. The journal article itself was not reachable.

What is scaffolding in simple terms?

Temporary help that lets a student do something they cannot yet do alone, which you then take away as they improve. Rosenshine’s definition is a temporary support used to assist a learner, gradually withdrawn as learners become more competent. If the help is never withdrawn, it stopped being a scaffold and became a permanent feature of the task.

What is the difference between scaffolding and differentiation?

Scaffolding is temporary and aimed at a specific task the student cannot do yet. Differentiation is an ongoing adjustment for a difference in readiness, interest or profile that is likely to still be there next month. Both keep the standard the same. The difference is whether the support is supposed to disappear.

What are some examples of scaffolding strategies for secondary students?

Rosenshine names eight: modeling the steps, thinking aloud while solving, cue cards that do part of the task, checklists students use on their own work, a completed model to compare against, who-why-how prompts for reading, a step-by-step prompt for organising material, and warning students about likely errors in advance.

Does scaffolding actually improve student achievement?

The evidence is mixed. A review of a decade of research found only eight effectiveness studies, mostly one-to-one tutoring, and concluded scaffolding helps cognitive and metacognitive activities. A study of 768 students aged 12 to 15 found it is not unequivocally effective, because its value depends on how long students work independently first.

When should you remove a scaffold?

When the student can do the part the scaffold was doing, and in stages rather than all at once. The expertise reversal effect shows that guidance helping a novice can become redundant working-memory load for a learner who already has the knowledge. Research cited in that paper found gradual fading outperformed an abrupt switch.

Can you scaffold too much?

Yes, in two ways. A scaffold can get so complete that filling it in becomes the assignment, which replaces the standard rather than supporting it. And intervening too often appears to undercut even well-tuned support: when teachers helped frequently, less contingent support produced higher achievement than highly contingent support in one secondary study.

Won’t students feel singled out if some of them get extra support?

They will if you hand it to them. Offering the scaffold to the whole class and letting each student decide whether to pick it up removes the public sorting while keeping the support available, and choosing it is itself the transfer of responsibility the research describes. Accommodations written into an IEP or 504 plan are a separate matter and are not a teacher’s to withdraw.

Is it a problem if students keep using a scaffold after you fade it?

Not necessarily. Rosenshine notes students may still lean on a scaffold for especially hard problems after outgrowing it on routine ones, and that is appropriate use rather than regression. The useful move is to fade it for the ordinary version of the task and leave it available on request for the harder version.

About Clay Shumate

Clay Shumate is a certified secondary Social Studies teacher in the public schools of West Alabama, with seven years of classroom experience, a B.A. in History, and an M.Ed. in Secondary Education. He writes about project-based learning, student responsibility, respect, and practical ways to hold young people to a higher standard while giving them room to learn from mistakes. He is a member of the Society of Professional Journalists and writes to its Code of Ethics; this site’s editorial standards and corrections policy are published in full. More about Clay.

November 1, 2026

Co-Teaching Models: Which of the Six Actually Works in Grades 6–12

By Clay Shumate

Co-teaching models are the six ways two teachers can share one classroom: one teach one observe, one teach one assist, station teaching, parallel teaching, alternative teaching, and teaming. The first two are the most common and the cheapest to plan. The three that change what students actually experience — stations, parallel and teaming — cost real planning time.

Key Takeaways

There are six co-teaching models, and they are not interchangeable. Georgia’s state guidance rates one teach one observe, one teach one assist and station teaching as low planning cost, parallel teaching as medium, and teaming and alternative teaching as high.
The measured achievement effect is small. Across eleven years of Massachusetts data in grades 3–8, co-teaching raised reading scores for students with disabilities by 0.016 standard deviations and math by 0.026. Middle-grades math was the biggest single effect at 0.036. For students without disabilities, math gained 0.012 and reading showed nothing.
The widely quoted number is roughly fifteen times larger, and both deserve to be on the table. A 2001 meta-analysis averaged about 0.40 standard deviations — but only six studies out of 89 met its inclusion criteria, and few compared co-taught students against a pull-out control group.
One teach, one assist dominates, and that is the problem. In a synthesis of 32 qualitative studies covering 454 co-teachers, 25 of them described the special educator in an assistant-like or subordinate role, and differentiated instruction was rarely observed at all.
Planning time is the binding constraint, not goodwill. One study inside that synthesis found co-teaching teams averaged 45.5 minutes of shared planning a week and said they needed close to three times that.

Two adults in one room is one of the most expensive instructional decisions a school makes. It is also one of the least examined. Schedules get built, names get paired, and then the two teachers are left to work out what the second adult is for. The default answer — one teaches, the other walks around — is the model that research finds most often and recommends least.

This guide sets out the six co-teaching models, what the evidence says about whether co-teaching works, and which model fits which lesson in grades 6–12. The structures and timings below are compiled from the sources listed at the end, so read them as recommendations rather than a record of any one classroom. Where the research population was elementary, it says so.

What Are the Six Co-Teaching Models?

The six co-teaching models are one teach one observe, one teach one assist, station teaching, parallel teaching, alternative teaching, and teaming. They differ in how students are grouped, which adult is delivering instruction, and how much shared planning each one requires before the lesson starts.

Co-teaching is two certified teachers sharing responsibility for instruction, assessment and management in the same classroom at the same time. Georgia’s state guidance names all six and rates each one for planning cost, which is the detail most lists leave out and the one that decides whether a model survives past October.

Graphic describing the six co-teaching models in one line each: one teach one observe, one teach one assist, station teaching, parallel teaching, alternative teaching and teaming, each labelled with its planning cost.

Model	What happens	Planning cost	Best for	Main drawback
One teach, one observe	One teacher leads the whole group; the other watches named students and records what they do	Low	Finding out why a specific student stalls, before you change anything	Produces data, not instruction. Useless if nobody reads the notes
One teach, one assist	One teacher instructs; the other circulates and supports	Low	Covering a one-off lesson, or a first week while a pair is still learning each other	The second adult becomes an aide. This is the default and the weakest use of a licensed teacher
Station teaching	Small groups rotate through two or three stations, each teacher owning one	Low	Practice, review, labs, document work — anything with discrete tasks	Noise and timing. One slow station derails the rotation
Parallel teaching	The class splits in half; each teacher teaches the same content to one group	Medium	Discussion, questioning, anything where halving the group doubles airtime	Both halves must arrive at the same place. Requires genuinely shared planning
Alternative teaching	One teacher re-teaches or extends with a small group while the other works with the rest	High	Students who missed a prerequisite, and students who already have it	Visible grouping. If the same students always get pulled, you have built a tracked class inside a period
Teaming	Both teachers present, question and model to the whole group, trading roles	High	Modeling disagreement, two-method explanations, debate and seminar framing	The most planning of any model, and it fails flat if the pair has not rehearsed

Model names and planning-cost ratings follow Georgia Department of Education guidance. The “best for” and “drawback” columns are compiled for this guide and are not drawn from a study.

Notice what the planning column does to the list. The two models that need almost no preparation are the two that leave one teacher doing nothing a trained paraprofessional could not do. The models that justify the second salary are the expensive ones. That tension is the whole subject.

Does Co-Teaching Actually Improve Achievement?

Modestly, and by less than the usual claim. The largest study to date, covering eleven years of Massachusetts testing data in grades 3–8, found co-teaching raised reading scores for students with disabilities by 0.016 standard deviations and math by 0.026. Those are real, statistically significant and small.

Nathan Jones and Marcus Winters compared the same student’s scores in co-taught and single-teacher classrooms over time, using student and school fixed effects on state administrative data from 2007–08 through 2017–18. The math sample alone runs to roughly 129,700 observations for students with disabilities and 285,700 for students without. Their findings, in standard deviations:

Students with disabilities, reading: +0.016, significant, with no difference between elementary and middle grades.
Students with disabilities, math: +0.026 overall — 0.036 in middle grades against 0.014 in elementary, and that difference is itself significant.
Students without disabilities, math: +0.012, significant.
Students without disabilities, reading: no significant effect.

The authors are blunt about the comparison that matters. Their estimate for students with disabilities is far below the roughly 0.40 standard deviations averaged by Murawski and Swanson’s 2001 meta-analysis, and they place their own result around the 30th to 40th percentile of randomized-trial effect sizes — “small” by the standard taxonomy. If you have seen co-teaching sold on the 0.40 figure, this is the number that followed it.

Why the older number is so much bigger

The 2001 meta-analysis is not fraudulent; it is thin. A Virginia Commonwealth University summary of it, which is the version I was able to read in full, records that 89 studies were found, 52 were excluded for lacking substantial data, and only six met every inclusion condition, including a co-teaching treatment lasting more than two weeks. Those six produced a large reported gain in reading and language arts, a moderate gain in math, and moderate reductions in absences and referrals. The same summary records the authors’ own caveat: few of the studies included a control group of students with disabilities in self-contained or pull-out settings, which is the comparison the whole question rests on.

Solis, Vaughn, Swanson and McCulley’s 2012 review of six syntheses is often cited alongside it as independent confirmation. It is not. For student outcomes it draws on the same studies. Two citations, one evidence base.

So the honest position is this. Co-teaching, as actually implemented across an entire state, produces small positive effects for students with disabilities and a small math effect for everyone else. The larger figures come from a handful of studies of deliberately good practice. Jones and Winters say so themselves: their data cannot see what teachers did in those rooms, so their estimate is the average of statewide implementation, not of co-teaching done well.

What a small effect does and does not justify

A district reading those numbers will notice immediately that co-teaching puts two salaries in one room for a gain the authors themselves call small. That is a fair observation and it leads to the wrong conclusion, for a reason the achievement data cannot show. For most schools co-teaching is not an instructional experiment they chose; it is the structure through which students with disabilities receive services in a general education setting rather than a separate one. Dropping it does not return the money. It changes where a student sits all day.

Which makes the implementation finding the actionable one rather than the effect size. Jones and Winters are measuring average practice, and average practice — per the qualitative research in the next section — is one adult teaching and one adult circulating. The arithmetic a district should run is not “is 0.026 worth two salaries,” it is “we are already paying for this; what would it take to move it off the cheapest model.” The research answer to the second question is specific and unglamorous: shared planning time with an agenda, training the pair attends together, and pairing by content rather than by schedule.

Why Is “One Teach, One Assist” So Common, and Why Is That a Problem?

Because it needs no shared planning and no agreement about who is in charge. The cost is that the second teacher stops teaching. In a synthesis of 32 qualitative studies, 25 described the special educator in an assistant-like or subordinate role, and whole-class instruction largely persisted.

Scruggs, Mastropieri and McDuffie’s metasynthesis of 32 qualitative reports pulled together the views of 454 co-teachers, 42 administrators, 142 students, 26 parents and five support staff. Fifteen of the studies were elementary or earlier, fourteen were junior high, middle or high school, and three covered both — so unlike much of this literature, roughly half of it is about rooms like the ones this site is written for.

Their findings are uncomfortable reading, and they have not been superseded:

One teach, one assist was by far the most common model. Parallel, station and team teaching appeared much less often.
In 25 investigations the special educator’s role was described as subordinate, sometimes because of weaker content knowledge in the subject. One study found the special educator “seldom took (or was permitted to take) the lead in instruction.”
Differentiation was rarely observed. In one study teachers used whole-class activities 100 percent of the time. Another reported “little differentiation of instruction to address individual needs.”
Strategy instruction, mnemonics and peer mediation — the practices the literature recommends most — were seldom observed.

The authors’ conclusion is the sentence worth taking to a faculty meeting: the ideal of two equal partners has “largely not been met.” Students with disabilities were receiving good general education instruction with assistance, but possibly not special education. They borrow Hargreaves’ phrase and call much of the practice “contrived collaboration” — partnership produced by a schedule rather than by two people choosing it.

One caveat in their favour, which they raise themselves: ten of the 32 studies deliberately examined outstanding examples of co-teaching. If anything, the picture is more flattering than typical practice.

A smaller 2017 case study makes the same point at close range. Brendle, Lock and Piazza observed two co-teaching pairs and found that neither pair could name the model they were using. Co-planning covered lesson modification and rarely touched how instruction would be delivered or assessed. Only one of the four teachers had ever attended co-teaching training, and none of them had trained together. That study was in a single elementary school, with fourth- and fifth-grade pairs, so treat it as illustrative rather than as evidence about secondary rooms.

Which Co-Teaching Model Should You Use, and When?

Match the model to the lesson’s job. Use stations for practice and discrete tasks, parallel teaching when you need discussion airtime, alternative teaching when a group is missing a prerequisite, and teaming when the point of the lesson is to show two ways of thinking. Keep one teach, one assist for the weeks you have not planned together.

Graphic showing measured co-teaching effects in standard deviations: plus 0.016 reading and plus 0.026 math for students with disabilities, plus 0.012 math for students without, and no significant reading effect for students without disabilities.

The useful question is not “which model is best.” It is “what does this lesson need that one adult cannot do?” Three answers come up repeatedly in grades 6–12.

The lesson needs more student talk

Thirty students and one teacher means most of them say nothing. Parallel teaching halves the group, which roughly doubles every student’s airtime. It is the model that pairs best with a seminar or any structured discussion format, and the one that most obviously justifies the second teacher to a student watching the room.

Some students are missing a prerequisite

Alternative teaching exists for this and is the most easily abused model on the list. The rule that keeps it honest is that the pulled group changes. If the same six names go to the back table every week, the period now contains a tracked class, and the students know it before you do. Pairing it with clear learning targets makes the grouping legible: students can see which target they are working on and when they are done with it.

It also matters how the grouping is named out loud. “Come with me” aimed at the same table every Tuesday is a label. “Anyone who wants another pass at the second target, come over” is an invitation, and it lets a student who has it wrong but would never say so take the help without being assigned to it. Grades 6–12 students will refuse useful support rather than be seen receiving it, which is a predictable behaviour and not a character flaw. Build the model so accepting help is a choice they make.

The content has two defensible approaches

Teaming is where two teachers earn their keep. Two adults solving the same problem two ways, or openly disagreeing about an interpretation and then reasoning it out, is a lesson a single teacher cannot deliver. It is also the highest planning cost on the Georgia list, and it does not survive being improvised.

What all three have in common is that they need agreement before the bell, not during the lesson. That is also why differentiation that does not need five lesson plans matters more in a co-taught room than a solo one: the cheapest differentiation moves are the ones two teachers can run without a second plan.

How Much Planning Time Does Co-Teaching Really Need?

More than schools give it. Inside the 32-study synthesis, one investigation found co-teaching teams averaged 45.5 minutes of shared planning a week and reported needing close to three times that. Another found 40 minutes a week still felt insufficient.

Planning time was the most frequently raised need in the synthesis, appearing in nearly every study. That is not a complaint about workload. It is a structural fact about the models: station, parallel, alternative and teaming all require two people to agree in advance on who does what, with which students, to what standard. Without that agreement the pair defaults to one teach, one assist, because it is the only model that works with no conversation.

Graphic on co-teaching planning time: teams averaged 45.5 minutes a week, said they needed close to three times that, planning time was the most frequently raised need, and protected time with no agenda becomes modification time.

The 2017 case study shows how time alone does not fix it. Administrators there scheduled one day every nine weeks plus a half day every two weeks for collaboration, and the teachers valued it — but they received no guidance on what co-planning should cover, so the time went on modifying worksheets rather than on deciding how instruction would be delivered. Protected time with no agenda becomes modification time.

What if your co-teacher will not plan with you?

This is the question the model diagrams never answer, and teacher compatibility is named as a concern throughout the qualitative research. If the shared planning is not happening, the realistic move is not to fix the relationship first. It is to pick the one model that works with almost no negotiation and run it well.

Station teaching is that model. You can hand someone a single station with a clear task, a time limit and a defined output, and it functions even if the two of you never agree on anything else. It is low planning cost by design, and unlike one teach one assist it puts the second teacher in front of students doing instruction. Start there, keep a written record of what you proposed, and take the pattern to an administrator if it does not move — a pair that cannot plan is a scheduling problem, not a personality one, and it is fixable at the level that built the schedule.

If you are the one asking an administrator for something, ask for the two things the research actually points at: shared planning time with a named agenda, and training the pair attends together. Professional development that only one of the two attends cannot build a shared model, which is exactly what the case study found.

What Does Co-Teaching Have to Do With Treating Students Like Young Adults?

Students read the hierarchy in the room long before they read the lesson plan. When one of two licensed teachers spends every period handing out papers, students learn which adult counts. That lesson undercuts everything a co-taught class is supposed to signal about shared responsibility.

This is where I will put my own view plainly, because the research does not reach it. My position on respect in a classroom is that it is mostly mechanical: you do what you said you would do, every time, and students decide you are credible. A co-taught room makes that harder for two people at once, because now two adults have to say the same thing about deadlines, consequences and what counts as finished. If they do not, students will find the gap within a week and work it.

The other half of it is about teachers, and I have said this before on this site. Keeping your ears open and your mouth closed sometimes pays off; keeping your ears open and asking questions really pays off. In my first year I picked up a review format from another teacher on my hall because he noticed I was short on them and offered. That is what co-teaching is supposed to be an institutional version of — and a pair where one person is permanently the assistant has lost the part that made the informal version work.

The practical test I would apply, as a teacher rather than as a reader of studies: could a student in that room name what the second teacher is for? If the answer is “she helps,” the model is one teach, one assist whatever the schedule calls it.

How Do You Start Co-Teaching Well in Grades 6–12?

Start by agreeing on the non-instructional decisions, then add one new model at a time. The research points at three failure points — no shared planning agenda, no joint training, and no named model — so the first steps are the ones that close those.

Name the model before the lesson, out loud. “Tuesday is stations; you take the document group.” The 2017 case study found pairs who could not name what they were doing, and a model nobody names cannot be improved.
Agree on the boring things first. Who answers a question about a grade. Who calls a parent. Whose consequence applies when both of you saw it. Disagreement here is what students exploit, and it is cheaper to settle in August. A single written routine for late work, missing work and consequences gives both of you the same script to read from.
Take one low-planning model and one high-planning model. Stations cost little and immediately stop one adult from standing still. Pick one of parallel or teaming as the model you will actually build, and leave the others alone until that one is reliable.
Put a standing agenda on the shared planning time. Not “what are we doing” but: which model, who leads, how we will know students got it, and who is being re-taught, and who grades it. Deciding the check in advance is what turns two adults into two sources of information instead of one teacher and one observer. The 2017 case study found co-planning almost never reached assessment, which is how one teacher ends up owning the gradebook and the other ends up owning nothing.
Tell the students what the second teacher is for. Say it in the first week and say it again when the model changes: both of us teach, both of us grade, either of us can answer you. Students work out the hierarchy on their own if nobody states it, and the version they work out is usually the one you did not want.
Rotate who leads. If one of you has never opened a lesson by November, that is the subordinate-role finding happening in your room. Swap on content where the other person is strong.
Audit the pulled group once a month. List the names who have gone to alternative teaching. If the list barely changes, change the grouping or stop using the model.

Where Co-Teaching Goes Wrong

The second teacher is treated as capacity, not expertise. This is the subordinate-role finding, and it is the single most replicated result in the qualitative literature.
Pairing is done by schedule rather than by content. A special educator asked to co-teach four different subjects cannot hold four content loads, which is one reason the literature keeps finding weaker content knowledge as the cause of the subordinate role. The three loads that drive special education burnout are the same loads that make this worse.
Nobody reads the observation notes. One teach, one observe is the only model whose entire output is data. If it is not read and acted on, it is a teacher watching a class.
Alternative teaching hardens into tracking. Same names, same table, every week.
The pair is split up after one year. Teachers in the 2017 case study asked specifically to stay paired. Rebuilding a working pair costs a year of the thing that was working.

What to Do Next

Pick one lesson next week that needs something one adult cannot do, and run stations or parallel teaching for it on purpose. Name the model to your co-teacher and to the students. Then look at the pulled-group list and at who opened the last five lessons. Those two lists will tell you more about your co-teaching than any model diagram.

If you are building the structures a second adult plugs into, the next useful piece is the procedures that carry the most weight in grades 6–12. They are easier to run with two people and they collapse faster when the two people disagree about them.

Sources

Jones, N., & Winters, M. A. (2022). Are two teachers better than one? The effect of co-teaching on students with and without disabilities (Working Paper 2020–1, updated January 2022). Boston University Wheelock Policy Center. Massachusetts administrative data, 2007–08 to 2017–18, grades 3–8. Read in full.
Scruggs, T. E., Mastropieri, M. A., & McDuffie, K. A. (2007). Co-teaching in inclusive classrooms: A metasynthesis of qualitative research. Exceptional Children, 73(4), 392–416. 32 qualitative reports; 14 of them junior high, middle or high school. Read in full.
Georgia Department of Education. Co-teaching model one pager. Source for the six model names and the planning-cost ratings. The document states no advantages or cautions per model; those columns in the table above are compiled for this guide.
Brendle, J., Lock, R., & Piazza, K. (2017). A study of co-teaching: Identifying effective implementation strategies. International Journal of Special Education, 32(3), 538–550. Four teachers, two pairs, one elementary school. Read in full.
Virginia Commonwealth University RRTC. Co-teaching efficacy: What does the research say? Used as the readable summary of Murawski & Swanson (2001), Remedial and Special Education, 22(5), and of Solis, Vaughn, Swanson & McCulley (2012). The primary meta-analysis was not reachable; this summary of it was, and the six-of-89 inclusion figures and the verbal effect labels come from the summary, not from the paper itself.

What are the six co-teaching models?

One teach one observe, one teach one assist, station teaching, parallel teaching, alternative teaching, and teaming. Georgia’s state guidance rates the first three as low planning cost, parallel teaching as medium, and alternative teaching and teaming as high. They differ in how students are grouped and which adult is delivering instruction.

Does co-teaching improve test scores?

By a small amount. The largest study, covering eleven years of Massachusetts data in grades 3 through 8, found gains of 0.016 standard deviations in reading and 0.026 in math for students with disabilities, and 0.012 in math for students without. Reading showed no significant effect for students without disabilities.

Why is the 0.40 effect size for co-teaching so much larger?

It comes from a 2001 meta-analysis in which only six studies out of 89 found met every inclusion condition, and few compared co-taught students against a pull-out control group. The 0.40 figure describes a handful of studies of deliberately good practice; the 0.016 to 0.036 range describes an entire state’s typical implementation.

What is the most common co-teaching model?

One teach, one assist, by a wide margin. A synthesis of 32 qualitative studies found it dominant, with parallel, station and team teaching appearing much less often. It requires no shared planning, which is why it wins by default, and it leaves one licensed teacher doing work a paraprofessional could do.

How much shared planning time do co-teachers need?

More than they get. One study inside the 32-study synthesis found teams averaging 45.5 minutes a week and saying they needed nearly three times that. A separate case study found that protected time without a named agenda gets spent modifying worksheets instead of deciding how instruction will be delivered.

Is co-teaching only for students with disabilities?

No, though that is its usual purpose. The Massachusetts study found a small significant math gain for students without disabilities as well, and the qualitative research reports social benefits for them more often than academic ones. Any two teachers sharing a room are using one of the six models whether they name it or not.

My child is in a co-taught class. What should I ask the school?

Ask who the two teachers are and what each one is responsible for, whether your child is being pulled into a small group and how often, and who to contact about grades. A co-taught class is not a lower-level class. If the answer is that one teacher helps the other, the school is using the weakest of the six models and it is fair to ask why.

Which co-teaching model should a new pair start with?

Station teaching. It costs little to plan, it immediately stops one adult from standing still, and it works with content you already have. Then choose one high-planning model, parallel teaching or teaming, and build that one properly rather than sampling all six.

About Clay Shumate

October 30, 2026

Vocabulary Instruction in Grades 6–12: What Actually Moves Comprehension

By Clay Shumate

Vocabulary instruction is the explicit teaching of word meanings and of strategies for figuring out unfamiliar words. In grades 6–12 it reliably improves comprehension of the text that carries the taught words, and barely moves standardized reading scores. The version that works is one or two essential words per text, taught through use, not a weekly list of ten.

Key Takeaways

The same vocabulary instruction produces two very different numbers depending on which test you give. Across 37 studies, Elleman and colleagues found an effect of d = 0.50 on comprehension measures built for the study and d = 0.10 on standardized reading tests.
Students who struggle with reading gain the most. On those study-built measures, controlling for method variables, students with reading difficulties gained d = 1.23 against d = 0.39 for students without — more than three times as much — while gains on the vocabulary tests themselves were similar for both groups.
The federal practice guide for grades 4–9 tells you to teach one or two essential words before a text, not a list. Its comprehension recommendation carries a strong evidence rating on 34 studies.
A school-randomized trial of academic vocabulary across 28 schools and 1,554 students in grades 6–8 moved the taught words (g = 0.25) and did nothing measurable to general vocabulary. What it moved most was the quality of classroom discussion (d = 0.62).
Teaching one or two word-attack strategies is the part with the least support. A review of 36 studies found no empirical evidence that it improves general comprehension. Context clues as a standalone unit is the weakest thing on most vocabulary plans.

Walk into most secondary classrooms the week before a unit test and you will find a vocabulary list. Ten words, definitions copied, sentences written, quiz Friday. It is the most common literacy practice in grades 6–12 and it is mostly a waste of a class period.

That is not an argument against teaching words. The research on vocabulary is better than the research behind most things teachers are told to do. It is an argument against the format and the promise attached to it.

What is vocabulary instruction?

Vocabulary instruction is the deliberate teaching of word meanings and of strategies for working out unfamiliar words, as opposed to leaving students to pick up words from reading and talk alone. It splits into two things researchers treat separately: direct teaching of specific word meanings, and teaching generalizable word-learning strategies like using context or breaking a word into parts.

That distinction matters more than any brand-name routine, because the two halves have very different track records. Direct teaching of specific words has a good one; strategy teaching on its own does not. Most programs sell both in the same box.

You will also hear words sorted into tiers — Tier 1 everyday, Tier 2 high-utility academic words that cross subjects, Tier 3 technical words belonging to one discipline. That comes from Isabel Beck and Margaret McKeown’s work and it is a useful way to triage a reading. It is a sorting scheme, not a finding: no effect size attaches to it.

Does vocabulary instruction actually improve reading comprehension?

It improves comprehension of the passage containing the taught words, consistently and by a lot. It barely improves performance on a standardized reading test. Elleman and colleagues meta-analyzed 37 vocabulary intervention studies from pre-K through grade 12 and found d = 0.50 on comprehension measures designed for the study and d = 0.10 on standardized measures.

Read those two numbers together and the field makes sense. Teaching a student mercantilism helps that student read a passage about mercantilism. It does not make them a better reader of a passage about tide pools.

What you measure	Effect of vocabulary instruction	What that means for your Monday
Comprehension of the text containing the taught words	d = 0.50 across 37 studies	Real and worth the time. This is the claim to make.
Score on a standardized reading test	d = 0.10 across the same studies	Close to nothing. Do not promise this to an administrator.
Knowledge of the specific words taught	Gains similar for strong and weak readers	Reliable. Students learn the words you teach.
General vocabulary beyond the taught words	No measurable effect in a 28-school randomized trial	Word lists do not grow a student’s overall word stock.

Effect sizes from Elleman et al. (2009); the general-vocabulary row is the null result from Lawrence et al. (2015).

Wright and Cervetti’s 36-study review lands in the same place from a different direction: teaching word meanings supported comprehension of the texts containing those words in almost every study, while evidence for generalized comprehension was — in their words — very limited, even in long interventions covering many words. They also could not determine how much instruction per word is enough, which is worth noticing in a field full of six-step routines.

So the defensible claim is narrow and still useful: vocabulary instruction is how you make a hard text readable, not a lever on reading ability. Which leaves a procurement question worth asking out loud. If a program’s own evidence rests on measures it built, the claim you can hold it to is students understand assigned texts better, measured on your own unit assessments. That is a real, checkable commitment. A state-test target written into a vocabulary plan is a commitment the published evidence says will not be met, and writing it down anyway sets up the teachers delivering it to take the blame.

How many vocabulary words should you teach per week?

Far fewer than most schedules assume. The federal practice guide for grades 4–9 recommends selecting one or two essential words to teach before a reading — words that are critical to the passage and likely to be hard — then giving quick meanings for other important words as they come up, so reading is not interrupted.

One or two. Not ten. And run the arithmetic from the student’s side before defending the list: a seventh grader with six teachers who each assign ten words is handling sixty words a week, none of which were chosen with the others in mind. No adult in the building carries that load. The reasoning is not that students cannot learn ten words; it is that deep instruction on two words costs about as much class time as shallow instruction on ten and produces something students can actually use. Wright and Cervetti found that instruction involving active processing generally beat definitions and dictionary work for understanding the text, which is precisely what a ten-word list has no room for.

You will see other credible guidance say three or four, and the difference is worth naming rather than hiding. Elsewhere on this site, the digital reading routine budgets two minutes for three or four pre-taught words, because a screen reading loses the edges a printed page gives for free and the words are part of what a Lexile measure is detecting. The practice guide is stricter because it is describing deep instruction on an essential word, not a quick gloss. Both can hold at once: one or two words get the full treatment, and another two or three get five seconds each. What nobody’s evidence supports is fifteen.

That guide’s broader comprehension recommendation is rated strong on 34 studies, 23 of which met What Works Clearinghouse standards without reservations. One caveat worth stating: the guide covers grades 4–9, so the top half of a secondary teacher’s roster sits outside its range. Nothing in it is implausible for grades 10–12, but it has not been tested there.

There is also a reason not to despair about the arithmetic. Hiebert, Goodwin and Cervetti mapped a core vocabulary of 2,451 morphological families containing 11,298 words, and found those families account for about 89 percent of the words in texts pitched at grade 11 and college readiness. By middle school roughly 20 percent of the distinct words in a text sit outside that core; in high school it is about 25 percent. That is the real size of the problem. It is not an infinite list, and it is not two words either.

Those two numbers are in open tension and the honest thing is to say so. If a quarter of the distinct words in a high school text sit outside the core vocabulary, teaching one or two words per reading cannot possibly close that gap. Both claims are still true, because they answer different questions. The one-or-two rule is about the words that block comprehension of this passage, which is a problem instruction can solve this week. The long-run size of a student’s word stock is a different problem, and the evidence above does not show that any amount of word-list instruction solves it — the 28-school trial found exactly nothing on general vocabulary after a full year. Anyone who tells you a vocabulary curriculum closes the word-gap is going beyond what has been measured.

Which words should you choose for instruction?

Choose the words that carry the meaning of this text and that students will meet again. A word qualifies if removing it would break a student’s understanding of the passage. That test does more work than any tier chart, because it is answered by the reading in front of you rather than by a list someone else made.

Four questions to run over a reading before you teach it:

Would a student who skipped this word misunderstand the paragraph? If not, it is not essential. Give it a synonym in passing and move on.
Will they meet it again this year? Infer, precedent and equilibrium earn real instruction. A term appearing once in one chapter does not.
Is the hard part the word or the idea? Photosynthesis is a concept with a label on it. Teaching the label without the concept is how students write definitions they cannot use.
Does it have usable parts? A word built from a prefix, suffix or root the class already knows is worth teaching twice over, because it pays into the next word too.

Four questions to run over a reading before teaching it: would a student who skipped this word misunderstand the paragraph; will they meet it again this year; is the hard part the word or the idea; does it have usable parts. — Four questions that decide which words earn real instruction.

The guide’s sequence on word parts is worth keeping: common prefixes and suffixes first, then Latin and Greek roots, using word maps. Prefixes and suffixes are regular and few. Roots are neither.

What does vocabulary instruction look like in a grades 6–12 content classroom?

It looks like about four minutes before the reading, a few seconds during it, and one short task a week after it. Here is the routine the practice guide describes, written as steps you can run tomorrow in history, biology, algebra or English.

Before reading, pick one or two essential words. Give a student-friendly meaning in plain language — not the dictionary entry. Then give one example and one non-example, because the boundary is where the meaning lives. A precedent is an earlier decision that sets a pattern for later ones. The first state to lower a voting age sets a precedent. A state that lowers it afterward is following one. The non-example is the step teachers skip and it is the one doing the work — a student who can tell you what does not count has the boundary, and a student with only the definition usually does not.
Make students use it once, immediately, out loud or in writing. One sentence that requires the meaning, not one that merely contains the word. “Name something in this chapter that set a precedent and say what followed it” works. “Use precedent in a sentence” produces “The precedent was very important.”
During reading, hand over other hard words in under five seconds. A synonym or a six-word definition. The point is to keep the student in the text, not to teach a second word.
Once or twice a week, make them do something with the words. Sort them, argue about non-examples, use two in an answer to a content question, add them to a running word log. This is the active processing that the reviews found outperforms definitions.
Teach the parts when the parts repeat. When three words in a unit share a prefix or a root, stop and map it. That is the only point at which strategy instruction has an argument behind it.

Numbered five-step routine: pick one or two essential words before reading with a plain-language meaning, an example and a non-example; make students use the word once immediately; hand over other hard words in under five seconds; once or twice a week do something with the words; teach word parts when the parts repeat. — The whole routine, which fits in the margins of a lesson you were already teaching.

On prep cost: the only real labor here is reading the text as a student would and marking the words that would stop one. That takes about ten minutes the first time you teach a reading and nothing at all afterwards, provided you keep the marked copy. If you teach six sections of two courses, you are marking two readings, not twelve.

Notice what is missing: the quiz. Keep one if your gradebook needs it, but do not confuse it with the instruction — matching ten words to ten definitions on Friday demonstrates the one thing the research says transfers least. So collect different evidence. The question worth answering is whether a student can answer a content question that cannot be answered without the word. “Which of these decisions set a precedent, and what followed from it?” tells you something a matching quiz cannot. That is also the easiest thing for a coach or an administrator to look for in a ten-minute walkthrough: not whether words are posted, but whether a student had to use one on the content inside the first fifteen minutes.

If you already run regular checks on understanding, this folds into them rather than adding a system. The same logic drives formative assessment strategies for reading: find out what the text actually cost students before you move past it. And when a class spans a wide range of reading levels, the word-choice question above is one of the cleanest places to differentiate instruction without writing two lesson plans.

Do word-learning strategies like context clues and roots actually work?

Teaching one or two word-attack strategies has the weakest evidence in the whole field. Wright and Cervetti’s review of 36 studies found no empirical evidence that teaching one or two strategies for working out word meanings improves generalized comprehension. That is a null result on the practice most schools spend the most instructional time on.

This deserves care, not a victory lap: null evidence for a narrow version of a practice is not proof the practice is useless. What the reviewers found promising was the fuller version — teaching students to monitor whether they understand a word and move flexibly among several strategies rather than drilling one. They call that a promising direction for future research, which means nobody has shown it works yet either.

The practical read: a September context-clues unit that never comes back is the unsupported version. Prefixes and roots taught at the moment three words in a unit share them, repeatedly across a year, is the version with an argument. One is a week of lessons. The other is thirty seconds, forty times.

Why doesn’t vocabulary instruction raise standardized reading scores?

Because a standardized reading test samples the whole language while your instruction samples your unit. Teach forty words well in a semester and a student knows forty more words, against a test drawing from tens of thousands. The signal is real and it is swamped.

Four-row comparison graphic: comprehension of the text with the taught words d = 0.50 across 37 studies; score on a standardized reading test d = 0.10 across the same studies; knowledge of the specific words taught shows gains similar for strong and weak readers; general vocabulary beyond the taught words shows no measurable effect in a 28-school randomized trial. — Same instruction, four different outcomes, depending on what you measure.

The Word Generation trial shows this cleanly. Lawrence, Crosson, Paré-Blagoev and Snow randomly assigned schools within matched pairs across two districts — 28 schools, 1,554 students in grades 6–8 with pre- and posttest, one full school year, delivered by whole grade-level teams. A significant effect on the taught academic words (Hedges’ g = 0.25), and no significant effect on general vocabulary on the Gates–MacGinitie, estimated at roughly −0.01.

The most interesting finding there was not about words. Discussion quality improved substantially in treatment schools — d = 0.62 overall and d = 1.13 in math — and mediated about 14 percent of the effect on word learning. A vocabulary program’s most visible consequence was that students argued about ideas more often.

The authors are candid about the study’s weaknesses, and they matter: posttest completion was 49 percent in treatment schools and 35 percent in control schools, treatment schools started with higher vocabulary scores, observers could not be blinded, and the word test covered only 36 of the 120 words taught. Attrition at that level can bend a result in either direction. Read g = 0.25 as a reasonable estimate rather than a settled number.

Which students gain the most from vocabulary instruction?

Students who struggle with reading, by a wide margin. In Elleman’s meta-analysis, looking only at the study-built comprehension measures and controlling for method variables, students with reading difficulties gained d = 1.23 against d = 0.39 for students without reading problems. More than three times the benefit from the same instruction.

Two qualifications. That contrast comes from the custom measures, the generous end of the evidence. And gains on the vocabulary tests were similar regardless of reading ability — both groups learned the words. The difference was in what the words did for comprehension, which points at word knowledge being the binding constraint for the weaker readers and not the stronger ones. Across studies reporting both outcomes the correlation between vocabulary effects and comprehension effects was r = 0.43 — modest. Learning the words helps; it is not the only thing in the way.

What about English learners, and classes where reading levels vary widely?

The honest answer on English learners is that this article cannot give you one. It is one of the most common things teachers ask about vocabulary. Neither review above reports results for English learners as a subgroup, and no English-learner-specific source was read for this article, so nothing here is an evidence claim about them. If somebody hands you a vocabulary program marketed on its results for English learners, ask which studies and which measures — and expect the custom-versus-standardized gap to show up there too.

For a class with a wide spread of reading levels, word selection is the cheapest lever you have: the same two words serve everybody while the amount of support varies. That is the logic behind differentiated instruction strategies generally: change the scaffold, not the destination.

And there is a real risk attached to the finding that weaker readers gain most, worth naming before a district acts on it. “Front-load vocabulary where reading is hardest” is a sound instructional read. “Those classes do words while the honors classes do ideas” is how that read turns into a lower-rigor track, and nothing in this evidence supports it. Elleman’s finding is that word knowledge was the binding constraint for those readers on the taught text — remove it and they are reading the same text about the same ideas. The whole point of clearing the word is to get to the argument.

If a parent asks why the weekly list went away, the one-sentence version is: the list taught words for a Friday quiz, and we would rather spend that time making sure the reading itself is understandable. For what a parent can do at home, the truthful answer is that no source in this article tests home vocabulary work, so treat any specific routine as a guess. Talking about what a kid read, in ordinary language, is a reasonable thing to do and is not a studied intervention.

Clay’s take: the list is a ritual, and rituals are hard to kill

The ten-word Friday list survives for reasons that have nothing to do with learning: easy to assign, easy to grade, easy to show a parent, and it produces a number in the gradebook every week. Those are real problems. None of them is comprehension. The honest version is smaller and harder to show off. Two words taught properly before a hard reading. A few seconds of definition handed over during it so nobody stalls. One short task a week using the words on content. That is the whole thing, and it fits in the margins of a lesson you were already teaching.

Where I would push back: the standardized-test number gets used to argue vocabulary work does not matter, and that is a misreading. A d of 0.10 on a test of everything is perfectly compatible with a d of 0.50 on the text in front of the student — and the text in front of the student is the job. A student who can read today’s document can participate in today’s class, and participation is not a consolation prize. It is the condition for everything else.

And the thing I would not do is use a vocabulary gap as an explanation for a student’s behavior or ceiling. The evidence says word knowledge was the binding constraint for weaker readers on the taught text. It says nothing about what those students can do. Treat the gap as a thing you remove from in front of them, not as a description of them. Students hear the difference immediately. A room where hard words get handed over without comment is a room where asking is normal; a room where not knowing a word is treated as a finding about you is a room where students stop asking and start guessing.

What to do next

Pick the hardest reading on next week’s plan. Read it as a student would and mark every word that would stop one. Cut the list to the one or two that carry the paragraph, write a plain-language meaning plus one non-example for each, and plan the thirty seconds where students have to use them. Keep a word log for the unit and spend five minutes on Friday making students do something with it other than recite.

If you want the rest of the system this sits inside, the list of formative assessment strategies covers how to find out what a text cost before you move on, digital reading strategies for students deals with the screen-reading problem that makes hard words easier to skip, and formative assessment strategies for listening covers the half of comprehension that never gets checked. And if the hardest readings in your course are documents rather than textbook pages, primary source analysis covers how to excerpt and frame them so the words are not the obstacle. Everything on this site is free and nothing asks for your email address.

Where a single concept carries a whole unit, a dedicated organizer earns its minutes in a way a routine cannot — when four boxes beat a definition list covers that case, how to score it, and what the evidence will and will not carry.

Frequently Asked Questions

What is vocabulary instruction?

Vocabulary instruction is the deliberate teaching of word meanings and of strategies for working out unfamiliar words, rather than leaving students to absorb words from reading alone. Researchers split it into direct teaching of specific word meanings and teaching generalizable word-learning strategies. The two halves have very different evidence behind them, with direct teaching of specific words faring much better than strategy instruction.

How many vocabulary words should I teach per week?

The federal practice guide for grades 4-9 recommends selecting one or two essential words to teach before a reading, then giving quick meanings for other important words as they appear so reading is not interrupted. One or two taught deeply costs about the same class time as ten taught shallowly and leaves students with something they can use. Add one short word task once or twice a week.

Does vocabulary instruction improve reading comprehension?

It improves comprehension of the text containing the taught words. Across 37 studies, Elleman and colleagues found d = 0.50 on comprehension measures built for the study and only d = 0.10 on standardized reading tests. So vocabulary instruction is a reliable way to make a specific hard text readable, and a weak lever on general reading ability.

Which words should I choose for vocabulary instruction?

Choose words that carry the meaning of the passage and that students will meet again. The working test is whether a student who skipped the word would misunderstand the paragraph. High-utility academic words that cross subjects, such as infer, precedent or equilibrium, earn real instruction. A technical term that appears once in one chapter usually does not.

Do context clues and root words actually work?

A review of 36 studies found no empirical evidence that teaching one or two strategies for working out word meanings improves generalized comprehension. The version with more promise, which the reviewers call a direction for future research rather than a finding, teaches students to monitor their own understanding and move flexibly among several strategies. A one-off context-clues unit is the weakest part of most vocabulary plans.

Which students benefit most from vocabulary instruction?

Students who struggle with reading. On study-built comprehension measures, controlling for method variables, students with reading difficulties gained d = 1.23 against d = 0.39 for students without reading problems. Both groups learned the taught words at similar rates, so the difference was in what knowing the words did for their comprehension of the passage.

Is Marzano’s six-step vocabulary process evidence-based?

Treat named routines carefully. The reviews find that instruction involving active processing beats definitions and dictionary work for understanding the text, which supports the general shape of multi-step routines. What the research does not establish is how much instruction per word is enough, or that any specific branded sequence outperforms another. Wright and Cervetti say so directly.

Will a vocabulary program raise our state reading scores?

Probably not measurably. A school-randomized trial across 28 schools and 1,554 students in grades 6-8 found a significant effect on the academic words taught and no significant effect on general vocabulary. A standardized test samples the whole language while your instruction samples your unit. Promise better comprehension of assigned texts instead.

Sources

Elleman, A. M., Lindo, E. J., Morphy, P., & Compton, D. L. (2009). The impact of vocabulary instruction on passage-level comprehension of school-age children: A meta-analysis. Journal of Research on Educational Effectiveness, 2(1), 1–44. 37 studies, pre-K through grade 12. d = 0.50 custom comprehension measures, d = 0.10 standardized; reading difficulties d = 1.23 vs. d = 0.39 on custom measures controlling for method variables; r = 0.43 between vocabulary and comprehension effects. Read at ERIC abstract level — the full article is paywalled, so no figure beyond those reported in the abstract is quoted here, and the grade distribution across the 37 studies is not reported.
Wright, T. S., & Cervetti, G. N. (2017). A systematic review of the research on vocabulary instruction that impacts text comprehension. Reading Research Quarterly, 52(2), 203–226. 36 studies. Teaching word meanings supported comprehension of texts containing the taught words in almost all cases; active processing beat definitions and dictionary use; evidence for generalized comprehension very limited; no evidence that teaching one or two word-solving strategies improves generalized comprehension; amount of instruction needed indeterminate. Read via the ERIC record and the University of Michigan Deep Blue repository record; the Wiley full text is paywalled.
Institute of Education Sciences / What Works Clearinghouse (2022). Providing Reading Interventions for Students in Grades 4–9 (WWC 2022007). Recommendation 3 (“Routinely use a set of comprehension-building practices”) rated strong on 34 studies, 23 meeting WWC standards without reservations and 11 with reservations. Part A specifies selecting one or two essential words before reading, brief meanings during reading, word work once or twice a week with a word log, and prefixes and suffixes before Latin and Greek roots. Covers grades 4–9; grades 10–12 are outside its tested range. Full guide read.
Lawrence, J. F., Crosson, A. C., Paré-Blagoev, E. J., & Snow, C. E. (2015). Word Generation randomized trial: Discussion mediates the impact of program treatment on academic word learning. American Educational Research Journal, 52(4), 750–786. Schools randomized within matched pairs in two districts; 28 schools, 1,554 students grades 6–8 with pre- and posttest, one school year. Taught academic vocabulary g = 0.25 (p < .01); general vocabulary (Gates–MacGinitie) not significant; discussion quality d = 0.62 overall and 1.13 in math; discussion mediated about 14 percent of the effect. Authors’ stated limitations include posttest completion of 49 percent (treatment) and 35 percent (control), higher treatment-school pretest scores, unblinded observers, and a word test covering 36 of 120 taught words. Full text read.
Hiebert, E. H., Goodwin, A. P., & Cervetti, G. N. (2018). Core vocabulary: Its morphological content and presence in exemplar texts. Reading Research Quarterly, 53(1). 2,451 morphological families containing 11,298 words; coverage of 97.1 percent of tokens in grades K–1 exemplar texts falling to 89.1 percent at grade 11/college-and-career-ready, and 95.6 percent to 74.9 percent for unique words; roughly 20 percent of unique words outside the core by middle school and 25 percent in high school. Read as the authors’ pre-copyedit manuscript posted by TextProject.
Tier 1/2/3 word classification is attributed to Beck and McKeown’s work on robust vocabulary instruction and is described above as a sorting framework rather than an evidenced intervention. No effect size is claimed for it here, and no source above tests it against an alternative.
The four word-selection questions and the example wording for precedent are written for this article to illustrate the practice guide’s criteria. They are worked examples, not findings, and not a report of any teacher’s classroom.

About Clay Shumate

October 25, 2026

Executive Functioning Strategies for Students: What Works in Grades 6–12

By Clay Shumate

The executive functioning strategies for students with the best evidence change the environment rather than training the brain. A meta-analysis found training one executive function transfers to that skill (g = 0.44) but not to untrained ones (g = 0.11, not significant). What works in grades 6–12 is externalizing the plan and building in checkpoints.

Key Takeaways

Executive function training mostly trains itself. Across 43 studies of near transfer and 17 of far transfer, training working memory, inhibition or cognitive flexibility produced g+ = 0.44 on the trained component and g+ = 0.11 (p = .11) on untrained ones.
Even the gains that appear may not stay. A separate meta-analysis of 90 studies and 8,925 children found executive function can be fostered, but reported no convincing evidence that benefits remained at follow-up.
The strongest classroom result here is organizational, not cognitive. In 280 middle schoolers with ADHD, two brief interventions delivered by school staff in under 20 minutes a session beat a waitlist on parent-rated homework problems and organization with large effect sizes.
Teaching students to plan and monitor their own work does have real secondary-school numbers. Self-regulated learning training at secondary level averaged 0.54 on academic performance and 0.88 on strategy use, with metacognitive reflection the strongest single component.
The same evidence carries a warning for teachers: researcher-delivered versions of these programs outperformed teacher-delivered ones. The gap is implementation and training, which means a half-taught system is a predictable way to get nothing.

A student who has not started an assignment by minute twelve is not refusing. Usually they cannot find the first move. The gap between knowing what to do and beginning to do it is the single most common thing I see go wrong in grades 6 through 12, and almost nothing in a teacher’s training addresses it directly.

The industry built around this problem sells brain training — games, apps, working-memory programs, cognitive exercises. The research on that is unusually clear and unusually disappointing. What it supports instead is less exciting and considerably cheaper: stop asking the student to hold the plan in their head. Below: what these skills are, what the transfer research found, the strategies with classroom evidence behind them, how to run this for 150 students, and where the evidence runs out.

What are executive functioning strategies?

Executive functioning strategies are the supports that help a student plan, start, organize, hold information, resist distraction and switch tasks — either by teaching a routine or by moving the demand out of the student’s head and into the room.

The research usually splits executive function into three components: working memory, inhibitory control and cognitive flexibility. In a classroom those show up as holding multi-step directions, not responding to the first distraction, and changing approach when the first one fails.

The practical taxonomy matters less than the split between two kinds of strategy, because the evidence treats them completely differently. Training strategies try to make the underlying skill stronger — working-memory games, cognitive exercises, apps. Scaffolding strategies reduce how much of the skill the task requires — a written plan, a visible checklist, a checkpoint, a materials system. Most of what gets marketed to schools is the first kind. Nearly all of the evidence favors the second.

Can executive functioning skills be taught?

They can be improved on the thing you train, and that improvement mostly does not spread. Training one executive function produced g+ = 0.44 on the trained component and g+ = 0.11 on untrained components, which was not statistically significant.

Kassai, Futo, Demetrovics and Takacs (2019), in Psychological Bulletin, pooled 43 studies for near transfer and 17 for far transfer across working memory, inhibitory control and cognitive flexibility. Near transfer was clear: interventions did train the component they targeted, g+ = 0.44, p < .001. Far transfer was not: g+ = 0.11, p = .11. Their conclusion is worth quoting in substance — the absence of benefit beyond the trained component calls into question the practical value of training executive functions in isolation.

The companion paper is gentler and still sobering. Takacs and Kassai (2019) synthesized 100 effect sizes from 90 studies covering 8,925 children and concluded executive function can be fostered in childhood. But they also reported that they did not find convincing evidence for the benefits remaining at follow-up assessment. For typically developing children, mindfulness practices showed moderate effects; for children with neurodevelopmental disorders or behavior problems, learning new self-regulation strategies worked best. Explicit practice on executive-function tasks — computerized or not — produced small to moderate effects, and those effects were consistently weaker for the non-typically-developing children who are usually the reason the program was purchased.

Both of those studies capped at age 12. That is a real limitation on a grades 6–12 site and it cuts both ways: the discouraging transfer findings are drawn from children younger than most of this audience, and so is the encouraging part. Nobody should be quoting these numbers as high school results, including me.

Why don’t executive functioning games and apps work?

Because getting better at a working-memory game makes you better at that game. The skill does not generalize to the untrained skills, and the transfer meta-analysis is explicit that this is why working-memory training has not produced academic gains.

Kassai and colleagues point directly at that connection, citing the working-memory training literature as the obvious case. If training working memory does not reliably improve inhibitory control — a close cousin measured in the same lab in the same session — there is no reason to expect it to improve a student’s ability to write a document-based essay three weeks later.

This is not an argument that executive function does not matter. It matters enormously. It is an argument about where the leverage is. A student with weak task initiation does not need a stronger brain; they need the first step written down where they can see it. The second is achievable in a Tuesday lesson and the first is not achievable by any method anyone has demonstrated.

Which executive functioning strategies actually work in grades 6-12?

The ones that move the load into the environment: a written first step, a visible plan with times, one materials system taught to mastery, dated checkpoints instead of a single deadline, and a short reflection before submission. Each substitutes a structure for a skill the student does not have yet.

The strongest school-based evidence here is organizational. Langberg and colleagues (2017), in the Journal of Consulting and Clinical Psychology, randomized 280 middle school students with ADHD in Virginia to one of two brief interventions or a waitlist. Both were delivered by school mental health providers during the school day, pulling students from electives, in sessions averaging under 20 minutes. Both beat the waitlist on parent-rated homework problems and organizational skills with large effect sizes. The Homework, Organization and Planning Skills program also produced moderate-effect improvements on teacher-rated materials management and organized action behaviors, and outperformed the comparison program on organized actions specifically.

Note what that intervention consists of. Not cognitive exercises. A binder system, a planner routine, and someone checking it. Students with more severe behavioral dysregulation responded better to the more structured of the two programs, which is the opposite of the usual assumption that struggling students need looser expectations.

Six rows pairing an executive function that commonly breaks in grades 6 to 12 with an environmental fix: task initiation, planning, materials organization, working memory, self-monitoring and cognitive flexibility. — Six breakdowns, six structures instead of six skills.

What is breaking	What it looks like in grades 6–12	The environmental fix	What the evidence says
Task initiation	Twelve minutes in, nothing on the page, not defiant	The first physical move written on the board: “title and date, then read the first question aloud to your partner”	No direct trial. Strongest inference from organizational interventions and from strategy instruction
Planning	Starts the project the night before it is due	Three dated checkpoints with a product due at each, not one deadline	Checkpoint structure is standard in the organizational programs that worked
Materials organization	The assignment was done and cannot be found	One system, taught to mastery, checked weekly — not three competing systems	Directly supported: teacher-rated materials management improved with moderate effect sizes
Working memory	Loses step three of a four-step direction	Steps posted and left posted; never deliver four steps verbally	Training working memory does not transfer (g+ = 0.11, ns). Reduce the demand instead
Self-monitoring	Hands in work they believe is finished and is not	A two-question reflection before submission: what did you do, what is still missing	Metacognitive reflection was the strongest component at secondary level (B = 0.82 on performance)
Cognitive flexibility	First approach fails and they stop rather than switch	Two worked approaches modeled, and permission to abandon one named out loud	Weakest evidence of the six. Treat as reasonable practice, not a result

How do you help a student who cannot start?

Make the first move smaller than the task and physical rather than cognitive. “Start your essay” is a planning problem. “Write the date and copy the question at the top of the page” is not a problem at all, and it ends the frozen state.

Task initiation is where teachers most often misread the behavior, because a student sitting still with a blank page looks like refusal and is usually paralysis. The two call for opposite responses. A consequence applied to paralysis adds a second problem to the first.

The sequence that works in my room, and it is three minutes:

Name the state, not the character. “You haven’t started.” Not “you’re wasting time.” One is a fact you can both work from.
Ask which step, not whether they understand. “Which part is the one you can’t get past?” Students can almost always answer this and almost never volunteer it.
Shrink the first move until it is not a decision. Write the title. Copy the question. Read the first source aloud. A move with no judgment in it cannot be gotten wrong, which is what makes it startable.
Stay for the first one. Thirty seconds standing there while they do it, then leave. Leaving before the first move happens wastes the whole conversation.
Come back once. Four or five minutes later, one look, no comment if it is moving. The check is what makes the next one faster.

The reason this is written as a teacher routine rather than a student skill is the transfer evidence. A student who gets unstuck fifteen times with help has had fifteen models of how to get unstuck. A student who is told to work on their executive function has had none.

Numbered graphic of five teacher moves for a student who cannot start: name the state not the character, ask which step, shrink the move until it is not a decision, stay for the first one, and come back once. — Paralysis is not refusal. Three minutes, five moves.

Does teaching students to plan and self-monitor actually work?

Yes, and this is where the secondary-school numbers finally appear. Self-regulated learning training at secondary level averaged 0.71 across all outcomes, 0.54 on academic performance and 0.88 on strategy use, with metacognitive reflection the strongest single component.

Dignath and Büttner (2008), in Metacognition and Learning, pooled 357 effect sizes, 94 of them from secondary school. The secondary results were not weaker than primary — strategy use was actually higher (0.88 against 0.72). In their meta-regression at secondary level, metacognitive reflection carried the strongest effects on academic performance (B = 0.82) and on strategy use (B = 1.45), with motivation strategy instruction next.

Two findings in that paper deserve to be uncomfortable. First, researcher-directed training outperformed teacher-directed training on both academic performance (B = −0.80) and strategy use (B = −0.64). Second, group work was positive overall at secondary level but negative for secondary mathematics performance (B = −0.65). The motivation figure for secondary rests on a single study and should not be quoted at all.

The Education Endowment Foundation’s evidence review of metacognition and self-regulation reaches the same general conclusion and adds the same caveat from another direction: most of the underlying research is cross-sectional, so causal claims are hard, and one study it cites suggests higher attainment may drive better strategy use rather than the reverse. It also notes a secondary science intervention that showed no impact alongside poor implementation in many schools, and that programs were effective mainly where they were implemented well with extensive teacher development.

Which is the honest headline for a teacher reading this. The effect is real and it is contingent on doing the thing properly for long enough. A reflection sheet handed out twice and abandoned is not a small dose of a 0.54 effect. It is zero.

That researcher-versus-teacher gap is worth reading as an instruction rather than an insult. The difference between the two conditions is training, feedback and somebody watching the implementation — which is the job description of an instructional coach. If your building has one, this is exactly the kind of work to bring them, and the honest ask is not “observe me” but “watch whether my version of this is actually the thing the studies ran.” Nearly every system in this article is easy to half-implement in a way that looks right.

And on the reflection itself: it is a gate, not a graded item. Two questions on an index card or a form before submission, read in the thirty seconds before you accept the work, not scored. Turn it into a graded assignment and you have added a hundred and fifty things to grade in order to support students who are already behind on grading, which is how this dies.

How do you do this for 150 students?

By building the supports into the lesson structure rather than into individual plans. A posted plan, one materials system, dated checkpoints and a closing reflection are classwide and cost no additional minutes once they are routine.

The individualized version of this work is real and it belongs to special education and school mental health teams. The Langberg interventions were delivered one-to-one by trained providers to students with a diagnosis. A general-education teacher with five sections cannot replicate that and should not pretend to.

What a classroom teacher can do is remove the executive-function tax from the whole room. If the plan is on the board with times on it, nobody has to hold it. If there is one place assignments live, nobody has to remember which of three places. If a project has three dated checkpoints, no student has to plan backward from a single date, which is the hardest planning task there is and the one we assign most casually.

One hard line, and it is the most important sentence in this article for anyone with a caseload. A classwide scaffold is not a substitute for a service a student is legally entitled to. Posting the plan does not discharge an accommodation on an IEP or a 504 plan, and “we do checkpoints for everybody” is not an answer to extended time, a separate setting, chunked assignments or organizational support written into a document. Build the classwide version because it helps everyone; deliver the individual services because they are owed. If the two ever get presented as alternatives in a meeting, they are being confused.

This also reframes who benefits. A posted plan is usually described as an accommodation. It is not — it helps the organized students too, and it means the student who needed it never had to identify themselves to get it. The same logic runs through differentiation that does not require five lesson plans and through the routines worth teaching explicitly.

One caution on the late-work side. Executive function explains why work arrives late; it does not settle what should happen next. That is a separate argument with its own retracted-study problem, and the evidence on late-work policy is worth reading before you decide a deadline is a scaffold.

Where does the evidence run out?

Early, and in four specific places: the transfer research stops at age 12, the classroom trials are clinical samples, nobody has tested the classwide version, and the one measured teacher-versus-researcher comparison went against teachers.

Five rows summarising the executive function research and naming each sample: the Kassai transfer meta-analysis at 0.44 near and 0.11 far, the Takacs review of 90 studies with no evidence of benefits at follow-up, the Langberg trial of 280 middle schoolers, the Dignath and Buttner secondary-level figures of 0.54 and 0.88, and the Education Endowment Foundation's implementation warning. — Two meta-analyses say don’t train it. Two say scaffold it.

Age ceiling. Both Psychological Bulletin meta-analyses covered children up to 12. Everything said here about transfer is extrapolated upward into grades 6–12, and that extrapolation is an assumption.
Clinical samples. The 280-student trial was students with ADHD, delivered by trained mental health providers. Whether the same binder-and-planner routine helps a general-education student with ordinary disorganization is untested, though it is hard to see the harm.
No test of the classwide version. Nobody has randomized classrooms to “post the plan with times and run three checkpoints” versus business as usual. The table above is assembled from adjacent evidence, not drawn from a trial of itself.
Delivery is a measured weakness. Self-regulated learning training worked better in researchers’ hands than in teachers’. That is the most relevant single finding in this article for anyone planning to run it alone with no training and no coach.
Durability is unestablished. One meta-analysis explicitly found no convincing evidence of benefits persisting to follow-up. Nothing here should be sold as building a skill a student keeps for life.
Nobody has studied how to take the scaffold off. This is the gap that matters most on a site about student responsibility, and there is no honest research answer to give.

That last one deserves more than a bullet, because a reasonable student will ask it. If the plan is always posted and the deadline is always broken into three, when does anyone learn to do it themselves? I do not have a study to point at. What I can say is that the alternative — withholding the support so students build the skill under pressure — has no evidence behind it either, and the transfer research gives no reason to think struggling through an unsupported task builds a transferable capacity.

The defensible position is that the scaffold should narrow rather than vanish. Three checkpoints in the first project, two in the second, one in the third, and name out loud what you are doing and why, so the student is tracking their own progress rather than noticing that the help disappeared. That is a judgment, not a finding, and it is labeled as one. It is also the version that is consistent with how this site thinks about what responsibility actually means — a standard that stays while the support thins out, not a cliff.

Clay’s take: the room does the remembering

My classroom runs on lamps. I do not turn the overhead fluorescents on — the light comes from torchiere lamps around the perimeter and a few desk lamps, and the effect is warm and low instead of institutional. I did it because the room is meant to be enjoyed. What I did not expect was how much it changed the start of class.

The point generalizes, and it is the whole argument of this article. I did not make anyone better at ignoring distraction. I lowered the amount of distraction there was to ignore. That is a worse-sounding intervention and a better-working one, and almost everything in the table above has the same shape.

The other thing I would say is about the posted plan, because I came to it from a different direction. My method is organization to the minute, and then you pause — the plan goes on the framed whiteboard at the front with times attached, and then you stop talking and let the room settle instead of filling the silence. I started posting times to control pacing. The unintended result is that no student has to carry the sequence in their head, and the ones who cannot carry it stopped being visible as the ones who cannot.

I will also say what the posted times are not, since this is where it goes wrong. They are a budget and a promise, not a script. When I miss them I say so out loud and give the reason. A schedule nobody acknowledges missing teaches students that the posted plan is decoration, and then it stops working as a support for the kids who were actually using it.

And one honest admission. Everything above is cheap to start and easy to quit. The posted plan survives in my room because it is tied to how I run pacing, not because I decided to support executive function. The supports that stick are the ones that make your own teaching easier. That is not a research finding. It is the reason most of these systems die in November.

What to do next week

Pick one: post the plan with times every day for two weeks, or convert the next project from one deadline into three dated checkpoints with something due at each. Do not do both. Then add the two-question reflection before submission — what did you do, what is still missing — because metacognitive reflection is the component with the strongest secondary-level numbers behind it.

Skip the brain-training app. If someone is proposing to buy one, the transfer meta-analysis above is the document to put in front of them, and it is one page of abstract. And if you are a parent who already bought one, that is not a parenting failure — the marketing for these products is genuinely persuasive and cites real neuroscience, just not neuroscience that says the product works. The planner routine and the one place assignments live are free, and they are the parts with the evidence.

Related reading: note taking strategies that survive the research is the working-memory half of this problem, what student responsibility looks like in practice covers the part where the scaffold comes back off, and holding teenagers to a standard without treating them like children is the line this whole topic has to walk. Everything on this site is free and nothing asks for your email address.

When the skill gap is the obstacle: see eight scaffolds that hold up in a secondary classroom — and the contingency finding that explains why most classroom scaffolding never actually fades.

Frequently Asked Questions

What are executive functioning strategies for students?

They are supports that help a student plan, start, organize, hold information, resist distraction and switch tasks. They split into two kinds: training strategies that try to strengthen the underlying skill, and scaffolding strategies that reduce how much of the skill a task demands. The evidence strongly favours the second kind, which includes posted plans, one materials system, and dated checkpoints.

Can executive functioning skills be improved?

On the specific thing you train, yes. Beyond it, the evidence is weak. A meta-analysis of 43 near-transfer and 17 far-transfer studies found g+ = 0.44 on the trained component but only g+ = 0.11 on untrained components, which was not statistically significant. A companion review also found no convincing evidence that gains remained at follow-up.

Do executive function games and apps work?

They improve performance on tasks resembling the game and do not reliably improve anything else. The far-transfer meta-analysis points to this directly as the reason working-memory training has not produced academic gains. Before a school buys a program, ask what untrained outcome it has improved in a randomized trial, and read the answer carefully.

How do you help a student who cannot start an assignment?

Shrink the first move until it contains no decision. “Start your essay” is a planning task; “write the date and copy the question at the top” is not. Name the state rather than the character, ask which specific step they cannot get past, stay the thirty seconds it takes for the first move to happen, and come back once a few minutes later.

Which executive functioning strategies have the best classroom evidence?

Organizational ones. A randomized trial of 280 middle schoolers with ADHD found two brief school-based programs — binder systems, planner routines and regular checks, under 20 minutes a session — beat a waitlist on parent-rated homework problems and organization with large effect sizes, and improved teacher-rated materials management with moderate effects.

Is teaching metacognition worth the class time?

At secondary level it has the better numbers in this whole topic. A meta-analysis of self-regulated learning training reported an average of 0.54 on academic performance and 0.88 on strategy use at secondary level, with metacognitive reflection the strongest component. The caveat is that researcher-delivered versions outperformed teacher-delivered ones, so a half-run version should not be expected to pay.

Are these strategies only for students with ADHD or an IEP?

The individualized versions belong to special education and school mental health teams, and the strongest trials used diagnosed samples. But the classwide supports — a posted plan with times, one place assignments live, three checkpoints instead of one deadline — cost nothing extra and help the organized students too. They also mean a student who needs the support never has to identify themselves to get it.

Do classwide supports replace an IEP or 504 accommodation?

No, and treating them as alternatives is a mistake. A posted plan helps everyone and discharges nothing. Extended time, chunked assignments, a separate setting or organizational support written into a document are services a student is entitled to, and “we do checkpoints for the whole class” is not an answer to any of them. Build the classwide version anyway, because it helps.

Sources

Kassai, R., Futo, J., Demetrovics, Z., & Takacs, Z. K. (2019). A meta-analysis of the experimental evidence on the near- and far-transfer effects among children’s executive function skills. Psychological Bulletin, 145(2), 165–188. 43 studies for near transfer, 17 for far transfer. Near transfer g+ = 0.44 (p < .001); far transfer g+ = 0.11 (p = .11), no convincing evidence. Authors conclude the lack of benefit beyond trained components questions the practical value of training executive functions in isolation. Children only; not a secondary-school sample.
Takacs, Z. K., & Kassai, R. (2019). The efficacy of different interventions to foster children’s executive function skills: A series of meta-analyses. Psychological Bulletin, 145(7), 653–697. 100 independent effect sizes from 90 studies, 8,925 children up to age 12. Executive function can be fostered, but “we did not find convincing evidence … for the benefits to remain on follow-up assessment.” Mindfulness moderate for typically developing children; strategy-teaching best for non-typically-developing children; explicit task practice small to moderate and weaker for non-typically-developing children.
Langberg, J. M., Dvorsky, M. R., Molitor, S. J., Bourchtein, E., Eddy, L. D., Smith, Z. R., Oddo, L. E., & Eadeh, H.-M. (2017). Overcoming the research-to-practice gap: A randomized trial with two brief homework and organization interventions for students with ADHD as implemented by school mental health providers. Journal of Consulting and Clinical Psychology. 280 middle school students with ADHD in Virginia; HOPS and CHIEF versus waitlist; sessions averaging under 20 minutes. Both beat waitlist on parent-rated homework problems and organizational skills, large effect sizes; HOPS improved teacher-rated materials management and organized action behaviors at moderate effect sizes. Numeric effect sizes are reported as magnitude labels in the record read; the published article carries the values.
Dignath, C., & Büttner, G. (2008). Components of fostering self-regulated learning among students: A meta-analysis on intervention studies at primary and secondary school level. Metacognition and Learning, 3, 231–264. 357 effect sizes (263 primary, 94 secondary). Secondary: 0.71 across all outcomes, 0.54 academic performance (n = 34), 0.88 strategy use (n = 54). Metacognitive reflection strongest at secondary level (B = 0.82 performance, B = 1.45 strategy use). Researcher-directed training outperformed teacher-directed (B = −0.80 performance). Group work negative for secondary mathematics (B = −0.65). The secondary motivation figure (0.17) rests on one study and is not quoted here as a finding.
Muijs, D., & Bokhove, C. (2020). Metacognition and Self-Regulation: Evidence Review. London: Education Endowment Foundation. Concludes improving self-regulated learning and metacognition can raise attainment, particularly where multiple elements are included. Caveats recorded here: most studies are cross-sectional so causal claims are hard; one study suggests higher attainment may drive strategy use rather than the reverse; a secondary science intervention showed no impact with poor implementation; programs were effective mainly where implemented well with extensive teacher development.

About Clay Shumate

October 23, 2026

Note Taking Strategies for High School Students: What the Research Supports

By Clay Shumate

The note taking strategies with the best evidence for high school students are the ones that change what a student does with the page: partially completed guided notes during the lesson, and self-quizzing from those notes over several short sessions afterward. Handwriting versus typing matters far less than most advice claims.

Key Takeaways

Guided notes — a partially completed handout students fill in — have the strongest numbers here. A meta-analysis of 12 studies and 1,529 students found an overall effect of d = 0.546 against student-generated notes, though every study was post-secondary.
The famous pen-beats-laptop study was run on 65 Princeton and 151 UCLA subject-pool students, and a 2019 direct replication did not reproduce it. Two experiments found no consistent difference, and the effect shrank further once students studied their notes.
Note-taking’s own benefit is real but modest. Across 57 comparisons of note-taking against no note-taking, the effect was described as positive but modest — and higher grade levels did not increase it.
What a student does afterward is where the big, well-replicated effects live. Practice testing and spacing study across sessions were both rated high utility in a major review; summarization, highlighting and rereading were all rated low.
Teach one format, teach it explicitly, and then spend your time on the review step. There is no credible secondary-school trial showing one template beats another.

Almost every article on this subject answers a different question than the one students are asking. Teenagers want to know what to put on the page; the research is mostly about what happens to the page afterward. That gap explains why so much note-taking advice feels true and changes nothing.

Below is what holds up: five strategies ordered by what they cost a teacher, the honest version of the handwriting-versus-laptop fight, the one intervention with real numbers behind it, and the review step that does most of the actual work.

What are the best note taking strategies for high school students?

Five, in order of what they cost you to run: guided notes, a two-column page with a question margin, a hard stop for processing, pair-and-compare, and self-quizzing from the notes later in the week. The last one carries the most evidence.

Guided notes — costs one planning block, once. Hand out the lesson’s structure with the key terms, figures and transitions left blank. Put the blanks on the items that carry the reasoning, not on incidental detail — a page of fill-in-the-blank trivia is a worksheet wearing a strategy’s name. Students write less and think more, and nobody loses the thread because they were still copying the previous slide.
A question margin — costs nothing. Students rule off the left third of the page and write questions there, not content. The content goes on the right. The margin turns the page into a quiz by Friday instead of a transcript, which is the whole point of the two-column formats.
A hard stop to process — costs three minutes. Stop talking twice a lesson and have students finish a sentence, fill a gap, or write the one thing they are not sure about. Taking notes while listening is a split-attention task; a pause is how you stop charging students for your pacing.
Pair and compare — costs two minutes. Students swap pages and add what they missed in a different colour. It surfaces gaps immediately, it is the cheapest formative check in the room, and it runs on the same mechanics as any other structured student talk.
Self-quizzing from the notes — costs the student time, not you. Cover the right column, answer the questions in the margin, check. Three short sessions across the week, not one long one the night before. This is the step with the strongest research in the entire article and the one students skip.

Notice that four of the five change the page and the fifth changes the week. If you only have the appetite for one, take the fifth. And be realistic about the first: one guided-notes page for the hardest lesson in a unit is achievable and reusable next year; one for every lesson is not, and a plan that assumes it will collapse by October.

Numbered graphic of five note taking strategies ordered by teacher cost: guided notes, a question margin, a hard stop to process, pair and compare, and self-quizzing from the notes. — Five strategies, ordered by what they cost you to run.

Is handwriting notes better than typing on a laptop?

Probably a little, in some conditions, on a question nobody has properly tested on high schoolers. The original finding was real; the direct replication did not reproduce it, and both studies ran on university students.

Pam Mueller and Daniel Oppenheimer’s 2014 study in Psychological Science is the source of every “the pen is mightier” headline. In Study 1, 65 students from the Princeton subject pool took notes on five TED Talks. On factual questions there was no difference at all. On conceptual questions the longhand group did better. Laptop notes averaged 309.6 words against 173.4 longhand, with 14.6 percent verbatim overlap with the talk against 8.8 percent — which is the mechanism the authors proposed: typing fast enough to transcribe means you can skip the thinking.

Study 2, with 151 UCLA subject-pool students, added a group told explicitly not to transcribe. The instruction did nothing — verbatim overlap was 12.07 percent with the warning and 12.11 percent without it. Study 3 tested a week’s delay and found no main effect of method at all; the only significant result was an interaction, where longhand students who got ten minutes to study their notes outperformed the other three groups.

Then in 2019, Kayla Morehead, John Dunlosky and Katherine Rawson published a direct replication and extension in Educational Psychology Review. In their first experiment longhand students did better on factual questions and showed no advantage on conceptual ones — the opposite pattern to the original. Their second experiment found no significant difference at all on an immediate test, and a group that took no notes did not perform worse than the note-takers. Across two delayed tests, method had no effect. Their meta-analysis of the direct replications found small, non-significant effects favouring longhand, and they concluded that naming a superior method “seems premature.”

Study	Who it was run on	What it found	What it means for a 10th grader
Mueller & Oppenheimer (2014), Study 1	65 Princeton subject-pool students	No difference on factual questions; longhand better on conceptual. Laptop notes 309.6 words vs 173.4, 14.6% vs 8.8% verbatim	The mechanism is transcription, not the pen. A typist who paraphrases is not the person in this study
Mueller & Oppenheimer (2014), Study 2	151 UCLA subject-pool students	Telling laptop users not to transcribe changed nothing — 12.07% vs 12.11% verbatim	Do not expect a reminder to fix it. Change the task instead
Mueller & Oppenheimer (2014), Study 3	109 UCLA subject-pool students, one-week delay	No main effect of method. Longhand plus ten minutes of study beat everything else	The study step is doing the work in the one result that survived a delay
Morehead, Dunlosky & Rawson (2019)	Undergraduates, two experiments plus a replication meta-analysis	Longhand advantage on factual, not conceptual; no consistent difference overall; a no-notes group matched the note-takers	Nobody should be telling your students the question is settled
Any of them, on teenagers	—	Not tested	Every number above comes from a university lab. Treat the transfer as an assumption

The defensible position for a high school teacher: if a student is transcribing instead of thinking, that is a problem whatever they are typing on, and the fix is a page that cannot be transcribed — which is what guided notes and a question margin are for. The same requirement, applied to the text itself rather than a separate page, is what separates annotation that works from highlighting. Banning laptops on the strength of a Princeton lab study is more certainty than the evidence supports. None of which overrides a device policy your school has already set; if yours restricts laptops, that is a conversation to have with an administrator rather than a rule to quietly ignore.

Does taking notes actually help, or just feel productive?

It helps, modestly. Across 57 studies comparing note-taking with no note-taking, the benefit of the act of writing was positive but small, and it did not get larger in higher grades.

Keiichi Kobayashi’s meta-analytic examination of the encoding effect in Contemporary Educational Psychology looked at exactly this. Three of its findings are worth a teacher’s time. Interventions designed to improve note-taking did not reliably increase the benefit, and neither did a higher schooling level. Visual presentation of the material interfered with note-taking. And recall tests picked up the effect more readily than recognition or higher-order tests did — so the gain is clearest on the kind of question that asks students to produce something, and least clear on the kind we say we care about most.

That middle finding deserves a sentence on its own. If you are projecting a dense slide and talking over it while students write, you are running the condition that interfered. The three minutes of silence in strategy 3 above is not a courtesy; it is the correction, and it is the same reason the first five minutes of a lesson are worth protecting.

Five rows summarising the laptop versus longhand research, naming the sample for each: 65 Princeton students, 151 UCLA students, 109 UCLA students at a one-week delay, the 2019 undergraduate replication, and a final row noting teenagers were never tested. — Every number here comes from a university lab.

Do guided notes work better than letting students take their own?

Yes, by a clear margin against blank-page notes, and by much less against a full set of teacher-provided notes. That second comparison is the one that tells you what to hand out.

Karen and David Larwin’s meta-analysis of guided notes pooled 12 studies, 27 effect sizes and 1,529 students. The overall effect was d = 0.546 (95% CI 0.342 to 0.749), which the authors describe as the average guided-notes student outperforming about 72 percent of students without them. Twenty-one of the 27 effects were positive.

The moderator analysis is the practical part. Against students generating their own notes from scratch, guided notes returned d = 0.781. Against a full set of professor-provided lecture notes, d = 0.265 — small, though still significant. In other words most of the value is in giving students structure, and you lose a lot of it by filling the structure in for them. The blanks are the intervention.

Two honest wrinkles. Every included study was post-secondary — the authors found no graduate studies and, relevant here, no secondary ones. And the discipline breakdown was not uniform: education d = 0.749, physics d = 0.614, psychology d = 0.603, but mathematics came out at d = −0.355 and dentistry at −0.142, neither significant. A procedural subject where the working is the content may not behave like a content-heavy one, and a maths teacher should treat this as untested rather than as support. Scaffolding the task without lowering the standard is the same principle that governs any other differentiation decision.

What should students do with their notes after class?

Quiz themselves from the notes, in short sessions spread across several days. Those two moves have the best-replicated evidence in the whole study-skills literature, and they are the two students almost never make.

Dunlosky and colleagues’ review of ten learning techniques in Psychological Science in the Public Interest rated practice testing and distributed practice high utility — the authors note these benefited students across many ages and ability levels. Rated low utility: summarization, highlighting and underlining, rereading, keyword mnemonics and imagery for text. On highlighting, the review found most studies show no benefit beyond simply reading the text.

Read that list next to what students actually do with notes, and the problem is obvious. Rereading the page and highlighting it are the two default behaviours, and they are both in the low column. So the review routine worth teaching is four steps:

Cover the content. The question margin stays visible; everything else is hidden.
Answer out loud or in writing. Produce the answer before checking. Looking first converts a test into a reread, which is exactly the swap that kills the benefit.
Check and mark the misses. Only the misses. A short list of what you got wrong is the study list.
Come back in two days, then before the test. Three ten-minute passes beat one thirty-minute pass, and the spacing is doing the work, not the total. It is also less total time than cramming, which is worth saying to a student carrying six other subjects.

For families asking what to buy: nothing. No pen, app or subscription in this literature outperforms a covered page and a question. If a parent wants to help, the useful job is asking their teenager to answer three questions from the margin out loud on Wednesday — not checking whether the notes look neat, which is the thing that is easy to see and does not matter.

If you want this to happen rather than to be recommended, put the first pass in class. Two minutes of covered-column quizzing as a starter costs almost nothing and is the same trade as retrieval on old material for early finishers. Students who have done it once with you are considerably more likely to do it alone.

Should I teach Cornell notes or another format?

Teach one format explicitly, then stop optimising the format. There is no credible secondary-school trial showing that a particular template beats another, and the research weight sits on use rather than layout.

Cornell’s structure — a cue column, a notes column, a summary strip — is sensible and the cue column is doing the job the question margin above does. What it does not have is a body of randomized evidence in grades 6–12 saying it outperforms a well-run two-column page. Treat it as a reasonable default rather than an evidence-based one, and be honest with students that you are choosing it for consistency.

What does matter, and what teachers routinely skip, is teaching the format like content: model a page under a document camera, show a bad example next to a good one, and look at students’ actual pages more than once in the first fortnight. A format assigned and never inspected is a format that quietly becomes a transcript again.

Three things to look for when you do inspect one, because “are these good notes?” is not a question you can answer by glancing: are any of the words the student’s own rather than yours, is there at least one question on the page, and is anything marked as not understood. A page with none of those is a transcript, whatever template it is on. If the format in question is a structure students keep for a whole term, the same inspection question runs through our piece on what to require in an interactive notebook.

And leave a door open. A student who already has a system that works — and some do, including ones that look like nothing you would recommend — should be allowed to keep it once they can show you it passes those three tests. Forcing a template onto a method that is already working is the kind of compliance exercise this site exists to argue against.

Numbered graphic of a four-step note review routine: cover the content, answer before you check, mark only the misses, and come back in two days and again before the test. — The step students skip, in four moves.

Where does the evidence run out?

Almost immediately, once you ask about teenagers specifically. Every quantitative claim in this article comes from a population that is not a high school class.

The laptop-versus-longhand literature is university students — Princeton and UCLA subject pools in the original, undergraduates in the replication. There is no equivalent trial on 15-year-olds, and the replication failed on the adults it was run on.
The guided-notes meta-analysis contains no secondary studies at all. Its authors say they found none at graduate level and the included work is undergraduate. The d = 0.546 is a real number about a different population.
Kobayashi’s review reports the gain most clearly on recall tests and least clearly on higher-order ones. If what you care about is analysis rather than retrieval, the encoding effect is on weaker ground.
Nothing above is broken out for students with disabilities, or for English learners, or for students who cannot write at speed. Guided notes are widely recommended as an access measure and that recommendation is older than the evidence base quoted here; check the accommodations on file rather than inferring from a meta-analysis.
No study here tested a format. Cornell, outline, mapping, charting — the comparative question is effectively unanswered in secondary classrooms.

Clay’s take: I have taught this at three levels of device access

The laptop question looks different to me than to most people arguing about it, because I have taught at all three levels of device access.

My second year, devices meant one cart shared between two classrooms. By my final year before my current school it was one device per student, and it is 1:1 again now.

What that taught me is that the device is rarely the variable. On the shared-cart years students wrote everything by hand and most of it was still a transcript. The pages that were worth anything were the ones where the task made transcription impossible — blanks to fill, a question to write, a partner’s page to compare against. That is a design choice, not a stationery choice.

I also take the research’s own hint seriously: I am circulating for most of the period, which means I can see what is actually on the page while there is still time to do something about it. A page that is three slides behind tells me something a quiz on Friday cannot, and it is the same reason pulling a few students to a table beats reteaching the whole room.

And I would rather be honest with students than tidy. When a teenager asks whether handwriting makes them remember more, the true answer is that the study everyone quotes was run on college students and did not replicate, and that what they do on Wednesday with Monday’s notes matters more than which hand they used. Treating them as people who can hear that is most of what this site is about.

What to do next week

Five moves, and none of them needs a purchase or a new system. Build one guided-notes page, add two silent stops, teach the question margin once, run the first review pass in class, and then look at four pages to see whether any of it landed.

Pick your busiest content lesson and build one guided-notes page for it. Leave the key terms, the figures and the transitions blank, and nothing else.
Add two three-minute stops to that lesson where you are not talking.
Teach the question margin once, explicitly, with a bad example beside a good one.
Put a two-minute covered-column quiz at the start of the next lesson so students have done the review step with you before you ask them to do it alone.
Walk the room and look at four pages. If they are transcripts, the task needs changing before the format does.

If you want the next step after notes, the same retrieval logic is what makes test corrections teach instead of hand back points, and the quickest checks for whether any of it landed are the five-minute formative ones. Everything on this site is free and nothing asks for your email address.

Good notes are the raw material; what a student does with them afterwards is the other half. The study skills evidence ranks ten common review methods and puts rereading those notes near the bottom — well below testing yourself on them.

Frequently Asked Questions

What is the best note taking strategy for high school students?

Guided notes during the lesson plus self-quizzing from them afterward. The guided handout gives students structure without letting them transcribe, and the quizzing is where the best-replicated evidence sits: practice testing and spacing study across several short sessions were both rated high utility in a major review of learning techniques, while rereading and highlighting were rated low.

Is it better to write notes by hand or type them?

The honest answer is that nobody has shown this clearly for high schoolers. The well-known study favouring handwriting used Princeton and UCLA subject-pool students, and a 2019 direct replication did not reproduce it — in one experiment a no-notes group did as well as the note-takers. If a student is transcribing rather than thinking, change the task rather than the tool.

What are guided notes?

A partially completed handout that carries the lesson’s structure with key terms, figures and transitions left blank for students to fill in. A meta-analysis of 12 studies and 1,529 students found d = 0.546 against student-generated notes but only d = 0.265 against a full set of teacher-provided notes, which suggests the blanks are what matters.

Are Cornell notes evidence-based?

Cornell’s cue column is a sound idea and the format is a reasonable default, but there is no credible randomized trial in grades 6–12 showing it beats another well-run two-column page. Pick one format for consistency, teach it explicitly, and spend the rest of your effort on whether students actually review from it.

How should students review their notes?

Cover the content, answer the questions in the margin from memory, check, and mark only what was missed. Then repeat in two days and again before the test. Three ten-minute passes beat one thirty-minute session, because the spacing is doing the work rather than the total time.

Does taking notes even make a difference?

A modest one. Across 57 studies comparing note-taking with no note-taking, the benefit of the writing itself was described as positive but modest, and higher grade levels did not increase it. The same review found visual presentation of the material interfered with note-taking, which is a reason to stop talking over a dense slide.

How do I help a student who cannot write fast enough to keep up?

Start with the accommodations or plan on file rather than improvising, then give that student the guided-notes page with more filled in and the same blanks everyone else has for the key ideas. Reducing the transcription load is not reducing the standard, and a page the student can finish is a page they can revise from.

Sources

Mueller, P. A., & Oppenheimer, D. M. (2014). The pen is mightier than the keyboard: Advantages of longhand over laptop note taking. Psychological Science, 25(6), 1159–1168. Study 1: 65 Princeton subject-pool students. Study 2: 151 UCLA subject-pool students. Study 3: 109 UCLA subject-pool students with a one-week delay. No high school sample in any study; the Study 3 main effect was null and the surviving result was an interaction with a ten-minute study period.
Morehead, K., Dunlosky, J., & Rawson, K. A. (2019). How much mightier is the pen than the keyboard for note-taking? A replication and extension of Mueller and Oppenheimer (2014). Educational Psychology Review, 31(3), 753–780. Abstract read via ERIC; the experiment-level detail quoted here (the factual versus conceptual reversal, the no-notes group, the effect of studying notes) was read in the Learning Scientists’ summary of the paper and is attributed to that summary rather than to a reading of the full text, which sat behind a publisher paywall.
Kobayashi, K. (2005). What limits the encoding effect of note-taking? A meta-analytic examination. Contemporary Educational Psychology, 30(2), 242–262. 57 note-taking versus no note-taking comparisons. Abstract read; it describes the overall effect as positive but modest without giving a figure in the abstract, so no number is quoted here. Schooling level did not increase the benefit; visual presentation interfered; recall tests detected the effect more than recognition or higher-order tests.
Larwin, K. H., & Larwin, D. A. (2013). The impact of guided notes on post-secondary student achievement: A meta-analysis. International Journal of Teaching and Learning in Higher Education, 25(1), 47–58. 12 studies, 27 effect sizes, 1,529 participants, all post-secondary. Overall d = 0.546; d = 0.781 against student-generated notes and d = 0.265 against full provided notes; mathematics d = −0.355 and dentistry d = −0.142, neither significant.
Dunlosky, J., Rawson, K. A., Marsh, E. J., Nathan, M. J., & Willingham, D. T. (2013). Improving students’ learning with effective learning techniques. Psychological Science in the Public Interest, 14(1), 4–58. Practice testing and distributed practice rated high utility; summarization, highlighting/underlining, rereading, keyword mnemonics and imagery rated low; elaborative interrogation, self-explanation and interleaved practice rated moderate.

About Clay Shumate

October 22, 2026