Bold the Key Words in Your Notes? 103 Studies Say It Helps You Remember — and the Wrong Bold Makes You Worse

A meta-analysis of 103 studies and 12,201 participants found that author-placed signaling cues — bold type, headings, and typographic emphasis — raise retention by half a standard deviation over unsignaled text.¹ The catch: bold placed on the wrong words measurably lowers comprehension, and warning readers in advance does not undo the harm.²

This is not the same finding as "highlighting is low utility." That verdict applies to self-highlighting while reading passively — a different intervention entirely, covered separately in How to Remember What You Read. What the meta-analysis above measures is emphasis placed by the note's author for a later reader — including yourself.

The distinction matters. When you write a note and bold a term, you are acting as the author of a signal, not as a passive reader reaching for a yellow pen. The evidence says those two acts have opposite effects on memory.

What 103 studies actually measured

Author-placed signaling — bold type, headings, typographic variation, text color, and similar cues — raises immediate recall (g = 0.53, 139 comparisons) and transfer of knowledge to new problems (g = 0.33, 70 comparisons) compared with unsignaled text. These are consistent, positive effects across five decades of classroom and laboratory research.¹

The term of art is signaling. In the educational-psychology literature, it means any device an author places in a text to tell a reader: this part matters. "Signals presumably evolved as a means for an author to guide readers' processing," write Lorch, Lemarié, and Grant in a 2011 review of the theoretical framework.³

The corpus behind these numbers spans studies published from 1970 to 2016, covering textbooks, lecture slides, and instructional materials of many kinds. These are mostly lab studies with student populations — the transfer to a note you write for your own future self is an inference, not a direct experimental finding. Keep that caveat in mind.

The largest effect: headings and bold specifically

Not all signaling cues are equal. The meta-analysis breaks the data by cue type, and the numbers diverge sharply. Structure cues — headings and bold-printed words — produce the largest retention effect in the corpus: g = 0.71 across 57 comparisons, nearly a full standard deviation above baseline.¹

That is a large effect by the standards of educational research. It is also the cue type most directly relevant to note-takers: Markdown's ## heading syntax and **bold** emphasis syntax map exactly onto what the signaling literature calls structure cues.

One mechanism is attentional. Eye-fixation studies show that headings increase the number of topics readers recall in summaries — not by adding recall across the board, but by concentrating it.⁴ Lorch, Lorch, and Klusewitz found in 1995 that signaling cues "increase attention only to the signaled content" rather than producing a general alerting effect.⁵ You are not making everything more memorable. You are making the signaled thing more memorable, at the cost of some de-emphasis on what you do not mark.

That asymmetry is the most useful thing to understand about bold. It redistributes attention.

The twist: bold on the wrong words actively hurts

Pre-existing inappropriate highlighting interferes with reading test performance — and warning readers in advance that the emphasis is unreliable does not eliminate the harm. That finding, from a 1997 study by Vicki Silvers Gier and David S. Kreiner, was notable enough to earn the Ig Nobel Prize for Literature in 2002, awarded for "their colorful report" on reading comprehension.⁶

The study tested what happens when a reader encounters a text already highlighted by someone else, badly. Experiment 1 found that pre-existing inappropriate highlighting interfered with performance. Experiment 2 found that advance warning did not fix it.²

The harm is not trivial. Comprehension drops when the emphasis in the text points to the wrong things. The attention system follows the marks on the page whether or not the reader knows to distrust them. Knowing is not enough.

The fix: re-highlight

A later series of studies by Gier, Kreiner, and colleagues tested whether readers could recover from pre-existing bad highlighting. The finding, from both a 2010 study published in Reading Psychology and a 2011 study in the Journal of College Reading and Learning, is that re-highlighting works. When readers used their finger or a digital highlighter to mark what they considered important, the harm from the pre-existing poor emphasis was eliminated.⁷

The 2011 study found specifically that "using the electronic highlighter eliminated negative effects of pre-existing, poor highlighting on both comprehension and metacomprehension."⁸ Metacomprehension is the reader's own estimate of how well they understood the text — both the actual and the perceived understanding improved when readers could impose their own signal on top of the bad one.

The practical translation: if you encounter a note — your own or someone else's — with emphasis on the wrong things, the fix is not to read more carefully. The fix is to re-mark it.

Boundary conditions: what signaling does not fix

Three boundary conditions from the meta-analysis are worth stating. Signaling cues do not produce equal effects in every domain or with every format, and treating the overall g = 0.53 as a universal guarantee would misread the evidence.¹

Mathematics. Signaling cues in mathematical content produce a near-zero effect: g = 0.08, not statistically significant.¹ The authors attribute this to the nature of mathematical processing. Mathematical structure may already be organized by its own inherent signals — equations, numbering, logical dependencies — leaving little work for typographic emphasis to do. Bolding a term in a proof likely adds nothing.

Animated or blinking cues. The meta-analysis includes a small group of studies testing dynamic emphasis — cues that flash or blink. The retention effect for blinking cues is g = −0.56 and the transfer effect is g = −0.47.¹ These numbers suggest potential harm, but the evidence base is thin: k = 3 comparisons in each direction, well below the threshold for a reliable estimate, and the effects are not statistically significant. The result is suggestive, not conclusive.

Permanent versus temporary cues. Permanent formatting — bold, headings, underlining — produces a retention effect of g = 0.58. Temporary cues, such as arrows or pointers presented alongside the text, produce g = 0.46.¹ Persistent formatting outperforms temporary callouts, though both are positive.

What this means for your notes

The signaling literature studied instructional materials — textbooks prepared for students, lecture slides prepared for audiences. The transfer to a Markdown note you write for your own future self is an inference from the evidence, not a finding of a controlled study on personal notes. With that caveat stated, the evidence pattern is consistent enough to suggest two working rules.

Bold the term, not the sentence. Structure cues redistribute attention. A bolded sentence is not a signal; it is a wall of emphasis that tells the reader nothing is more important than anything else. A bolded noun or brief phrase marks the specific thing worth returning to.

Headings are the highest-effect cue. The meta-analysis finds that text-structure cues — headings first, bold second — produce the largest effects in the corpus. A note with clear heading structure is performing the same function as a well-signaled textbook: it tells the future reader where the main topics are, and eye-fixation evidence shows this increases what they recall from summaries.⁴

Markdown's **bold** and ## heading syntax are the two tools the signaling literature specifically measures as effective. The habit of sparing, deliberate emphasis aligns with what the evidence recommends. Over-bolding is a form of no-bolding — the signal disappears into noise.

Frequently Asked Questions

Does bolding key words in notes help memory?

Yes, with important caveats. A meta-analysis of 103 studies found that author-placed bold and headings raise recall by g = 0.53 (139 comparisons) and transfer of knowledge by g = 0.33 (70 comparisons) compared with unsignaled text.¹ The studies used instructional materials with student populations, so applying these results to personal notes is an inference. The effect for structure cues specifically — headings and bold — is g = 0.71 for retention.

Is highlighting bad for learning?

The "low utility" verdict applies to self-highlighting while passively reading — the act of running a pen across someone else's text as you encounter it for the first time. That research is covered in How to Remember What You Read. Author-placed signaling — emphasis a writer places deliberately to guide a later reader — is a different act with a different and more positive evidence base.

Should I use headings in my notes?

The evidence from the meta-analysis suggests yes. Headings are classified as structure cues and produce the largest retention effect in the corpus: g = 0.71 across 57 comparisons.¹ Eye-fixation studies show headings increase the number of topics readers recall in summaries.⁴ A note with clear heading structure gives your future self a navigational map and a recall scaffold.

What happens if I highlight the wrong words?

Comprehension drops. Silvers and Kreiner's 1997 study found that pre-existing inappropriate highlighting interfered with reading test performance, and that warning readers in advance about the poor highlighting did not eliminate the effect.² The attention system follows the marks on the page regardless of the reader's intentions.

How do I fix a note with bad highlighting?

Re-mark it. Studies by Gier, Kreiner, and colleagues found that having readers apply their own emphasis — even using only a finger as a highlighter — eliminated the comprehension harm from pre-existing poor emphasis.⁸ The fix is not more careful reading; it is imposing a new signal on top of the bad one.

Does bolding help in math notes?

Probably not. Signaling cues in mathematical content produce a near-zero effect in the meta-analysis: g = 0.08, not statistically significant.¹ Mathematical structure may already be organized by its own inherent signals — equations, numbering, logical dependencies — leaving little work for typographic emphasis to do.

Why does bold help retention at all?

The best-supported explanation is attentional redistribution. Lorch, Lorch, and Klusewitz found in 1995 that signaling cues "increase attention only to the signaled content" rather than raising overall alertness.⁵ Bold does not make you read more carefully in general; it makes the marked content more likely to be processed and encoded. Unmarked content is correspondingly less attended to — which is why over-bolding is a form of no-bolding.

Emphasis is a promise you make to your future self: this word is worth the fixation. The research says that promise is worth making — carefully.

MNMNOTE stores your notes as plain Markdown on your own device, with no account required. The **bold** and ## heading syntax you use in every note are the same structure cues the signaling literature specifically measures. mnmnote.com