The Complexity That Counts: How Word Support Makes Demanding Text Comprehensible

by Elfrieda (Freddy) H. Hiebert, TextProject CEO

A year ago, I wrote about the decline in NAEP reading scores among our lowest-performing students (Hiebert, 2025a). I suggested that the decline might be traced, in part, to more than a decade of mandates and standards requiring students to read primarily complex texts, whether or not they had the proficiency to read them independently. Now my name is on a study with the words complex text in its title, and that study reports that the children who read the complex texts came out ahead (Downs et al., 2026). Read side by side, the two pieces can look as though I am arguing both sides of the question.

I am not. I took part in this study for two reasons: (a) complexity is not one thing, and (b) many of the challenging words were taught. Further, the finding that accuracy improved but rate did not lend support to an argument I have made elsewhere: complex text on its own is not enough for struggling readers to develop the automaticity underlying proficient comprehension.

Complexity is not one thing

Text complexity often comes down to words: the more unfamiliar or rare words a text contains, the greater the challenge—especially for students who have read less. The intervention texts in the Downs et al. study may look harder on the Flesch-Kincaid, but readability formulas—Lexiles included—can’t tell us what matters most here: how many words in a text are likely to be new or rare for readers at a particular point in their development.

That question matters especially for third graders, who may have read only 200,000 words in their lives so far, or even fewer. For these readers, the number of rare words—words that appear less than once per million words of text—is a crucial indicator of how complex a text really is. By that measure, the intervention texts did contain more rare words than the comparison texts—2.5 versus 1.0 per 100 words—but far fewer than the texts students encountered in their core reading program, which averaged 4.4 rare words per 100 words.

Length also changes what those proportions mean in practice. A typical core reading program text ran 1,084 words, so students encountered roughly 48 rare words in a single passage. The intervention passages averaged 287 words—about 7 rare words per passage. Seven rare words and 48 rare words represent very different reading demands. But there is another difference that may matter even more: whether students are taught those words or left to figure them out for themselves.

The challenging words were taught

The second reason for my support of the study’s findings is that the intervention did not simply expose students to challenging words. It taught them. Of the roughly seven rare words in a typical intervention text, three were taught through explicit strategy work and repeated reading, both before and during students’ reading of the passage. That left about four rare words for students to negotiate without direct instructional support, as illustrated in Figure 1.

The contrast with a typical core reading program is substantial. Typically, six to eight words are selected for instruction. Even if all 6 to 8 are rare words—which is unlikely—a text containing 48 rare words would still leave around 40 rare words without instruction. In other words, students reading the core-program text could be left to negotiate about ten times as many unsupported rare words in a single text as students reading an intervention passage.

The issue, then, is not simply whether a text contains challenging words. Students need challenging words if their vocabularies are to grow. What matters is the density of those words, the length of the text in which they occur, and the instructional support students receive when they encounter them.

The instruction also included a strategy for parsing multisyllabic words. We do not know how far students carried that strategy to other multisyllabic words in the intervention and assessment texts because the study did not include a measure of multisyllabic decoding. Still, such generalization is worth considering as one reason the intervention students did better on a distal measure of accuracy that was not limited to the words they had been taught.

Accuracy improved while rate did not

Intervention students gained in accuracy but not in rate. That result may look like a disappointment. It is not surprising, though, given the view of automaticity for which I have argued (Hiebert, 2025b). Reading speed comes from recognizing known words rapidly, without frequent stops at unfamiliar ones. Explicit strategy instruction can raise accuracy even in demanding text, and in this study it did. Reading rate, however, depends on text that provides substantial exposure to words shared across many texts, especially medium-frequency words, many of which are general academic vocabulary.

A developmental view of text complexity

When we sort texts by their vocabulary demands rather than by a single label, the kinds of text a striving reader needs become clearer. They differ from one another in ways that matter for instruction.

The first kind is the conventional text of instruction, which has risen in complexity over recent decades (Hiebert, 2025c). When such text is left unsupported, a common and counterproductive response follows: teachers or devices read the text aloud for students, and the students themselves read less.

The second kind is text designed to support automaticity (Hiebert, 2025b). Most of its words are among the most frequent in written English, giving students the repeated exposure needed to build reading speed.

The third kind is text like the ones in the Downs et al. intervention. It is hard on the surface but heavily scaffolded. Text of this kind can raise accuracy and build knowledge even when, on its own, it may not increase reading rate.

These three kinds differ sharply from a fourth: the tightly controlled decodable texts used in Corrective Reading. Those texts are built around narrow spelling patterns and often improbable storylines. They develop little of the lasting vocabulary or knowledge that students draw on in later reading.

For a developing reader, complexity is not simply a property of the text. It is a relationship between the words a text contains and the words a reader already knows. That is why the Downs et al. findings refine my earlier argument rather than contradict it. Accuracy and knowledge can advance in demanding text when the hard words are relatively manageable in number and taught. Automaticity in rate, on the other hand, is built with text in which nearly all the words are already known. The two are distinct accomplishments, and they require distinct texts.

The question, then, is not whether children should read complex text, but what kind of complexity they are ready to take on—and what support makes that complexity productive.

References
Downs, J., Hiebert, E. H., Conradi Smith, K., & Martz, K. (2026). Promoting reading accuracy and fluency outcomes with complex texts: A grade 3 intervention comparison. Reading Research Quarterly, 61(2), e70117.

Hiebert, E. H. (2025a). How Misunderstandings of Text Complexity May Have Widened the Achievement Gap. https://textproject.org/frankly-freddy/misunderstanding-text-complexity

Hiebert, E. H. (2025b). Unpacking automaticity: Scaffolded texts and comprehension. Journal of Adolescent & Adult Literacy, 68(4), 369–379.

Hiebert, E. H. (2025c). Flattening the developmental staircase: Lexical complexity progression in elementary reading texts across six decades. Education Sciences, 15(11), 1546.

The Complexity That Counts: How Word Support Makes Demanding Text Comprehensible