← All articles

Learning to read: what science says about the method

· By the Popinz team

Reading methods have clashed in public debate for decades. The debate has had hard numbers behind it since April 2000: the National Reading Panel synthesis, convened in 1997 on a mandate from the US Congress, measures the effect of systematic decoding instruction at 0,41 standard deviations, rising to 0,55 when it starts in preschool or in CP (National Reading Panel, 2000). A synthesis of 333 studies published in 2023 pinpoints who benefits most from it: the students who struggle to read (Burns, Duke and Cartwright, 2023).

A verdict built on 38 experiments

The report Teaching Children to Read appeared in April 2000. On decoding, the literature search screened 1 373 studies published since 1970; 38 met the panel's methodological criteria, enough to yield 66 comparisons between groups of students. Each comparison sets systematic teaching of letter-sound correspondences against another approach (National Reading Panel, full report, 2000).

The results of this strand were published in 2001 in Review of Educational Research, by Linnea Ehri, Simone Nunes, Steven Stahl and Dale Willows: an average effect of 0,41 standard deviations, measured at the end of the program (Ehri et al., 2001). The standard deviation serves as the unit of comparison between interventions; the panel calls this effect moderate.

The 38 studies cover students aged about 5 to 12, mostly in ordinary classrooms, with tests before and after (National Reading Panel, 2000). Several programs studied are widespread in English-speaking schools, such as Jolly Phonics or Reading Mastery.

What decoding does, and does not do

Behind the average of 0,41, the measurements vary. Gains peak at 0,67 standard deviations on regular words and 0,60 on pseudowords, those strings of letters with no meaning. Comprehension improves by 0,27 and oral reading by 0,25, both levels above zero (National Reading Panel, 2000).

The panel also flags one condition: programs that stick to letter-sound correspondences without putting them into practice achieve little effect; reading texts remains the goal (National Reading Panel, 2000). Among older students, progress in decoding does not come with a significant improvement in comprehension.

In the classroom, this gives two linked phases: explicit teaching of the code, and its use in texts that students read and write.

Starting early doubles the effect

In preschool and in CP, the average effect reaches 0,55 standard deviations; it falls to 0,27 when instruction only starts after CP (National Reading Panel, 2000). The report concludes that it is worth starting early, and continuing for two to three years.

The panel also compared delivery formats. One-to-one tutoring shows 0,57 standard deviations; whole-class instruction stays at 0,39. The differences between these formats are not statistically significant (National Reading Panel, 2000).

Classroom teaching, systematic and sustained for two to three years, therefore produces an effect comparable to one-to-one tutoring (National Reading Panel, 2000). No special setup is needed to start.

Struggling students first

In comparisons sorted by social background, the effect peaks among children from disadvantaged backgrounds: 0,66 standard deviations, against 0,44 for children from middle-class backgrounds (National Reading Panel, 2000). Among children identified as struggling with reading, it clearly improves word reading; gains on text processing are smaller.

The synthesis by Burns, Duke and Cartwright, published in School Psychology in 2023, reanalyzed 333 studies drawn from 26 meta-analyses: among students who struggle to read, decoding interventions rank at the top, with a median effect of 0,48 on comprehension or reading outcomes. Among average readers, the same effect falls to 0,19.

A support plan can lean on these figures: systematic decoding first, from preschool onwards, for the most fragile students.

What these figures change in a school

The report recommends putting decoding instruction in place in preschool and in CP, for all students, and continuing it for two to three years (National Reading Panel, 2000).

The panel also asks each program to provide teachers with tests, to identify the children who need extra systematic teaching (National Reading Panel, 2000).

Four months to a year after the end of a program, the follow-up effect has dropped to 0,27, against 0,51 at the end of the sessions (National Reading Panel, 2000). Real reading, practiced afterwards, sustains these gains.

The report also leaves questions open: the ideal length of a session, the number of correspondences to cover in the year. Those trade-offs are for teaching teams to make.

Sources