New evidence for the power of reading

The Sullivan and Brown Reading Study: New evidence for the power of reading, the effect of reading on poverty, and evidence for late intervention.

S. Krashen

As part of a large ongoing longitudinal study, Sullivan and Brown (2013) studied the performance of several thousand children in the UK on a variety of tests given when they were 16, and analyzed the effect of a number of predictors on their test scores using multivariate techniques.  I focus here on a few of the results of this very important study.

Overall results: Pleasure reading counts

Sullivan and Brown's main finding was that more reported pleasure reading of books at ages 10 and at age 16 was significantly related to scores on vocabulary, spelling and math tests given at age 16. The vocabulary and spelling results are consistent with those of  many studies, as Sullivan and Brown note.

Table 1 presents their results for vocabulary tests given at age 16, limited to only a handful of the many predictors they included in the analysis.

Table 1: Predictors of vocabulary test scores at age 16

Variable	beta	p-value
SES: higher job status	-0.012	0.613
Parent has degree	0.255	0
Higher income family	0.02	0.542
Read to everyday at age 5	0.115	0.01
Reads books often at age 10	0.313	0
Visits library often at age 10	0.009	0.791
Reads newspapers more than once/week at age 16	0.183	0
Reads comics/magaines more than once/week at age 16	-0.074	0.021
Reads books more than once/week at age 16	0.353	0
Reading proficiency at age five	0.039	0
Pictoral vocab at age five	0.091	0
Reading proficiency at age 10	0.117	0

From Sullivan and Brown (2013), table 7, model 4, based on 3,424 subjects.

SES job status: Levels 1 to 3 in the Goldthorpe Schema, which consists of seven levels. 1 includes "higher grade professionals", 2 includes higher-grade technicians, managers in enterprises, 3 includes routine non-manual jobs, 7 includes semi-skilled and unskilled workers.

The "betas" in table 1 tell us the impact of each variable on vocabulary scores. Betas can be compared with each other in the same study.  The betas for the subjects' book reading habits show that they are the strongest predictors of vocabulary knowledge at age 16. Early read-alouds also contribute. Lighter reading does not, but this makes sense: Very light reading makes its greatest contribution for less advanced readers. 

[The p-value indicates the odds that the effect is real, or "statisitically significant." P-values of .05 or less are generally considered to be significant. When a p-value is given as zero, it means that the value was very small.]

Surprisingly, "visits to the library" was not a significant predictor, which appears to be in conflict with previous research showing consistent relationships between library quality and reading proficiency (Lance, no date; Krashen, Lee and McQuillan, 2012).  There are several possible reasons for this: Libraries may have their strongest effect for higher poverty groups, "visits" may be less valid a predictor than measures reflecting actual use of the library (e.g. books taken out), and of course measures of library quality. It is also possible that a path analysis would show that library visits are a predictor of amount of book reading reported. Finally, library visits by a ten-year-old are not always up to the ten-year-old. If the library is far from home, the ten-year-old will require help getting there.

Reading as part of the cure for poverty

When Sullivan and Brown did not include reading exposure variables in their analysis, socio-economic variables were significant predictors of vocabulary scores (model 1). But, as Sullivan and Brown note, when reading exposure variables were added, socio-eonomic variables were no longer significant predictors and their impact declined (as reflected in the betas in table 2). In fact, as more reading exposure variables were included from model 1 to model 3, the job status betas declined linearly, and the same tendency is present for the higher income variable. This result is consistent with the results of previous studies showing that access to books and libraries can counter the negative effects of poverty on literacy development (Krashen, Lee and McQuillan, 2012; Krashen, 2011a).

Table 2: Disappearing effect of socio-economic factors

		job status		higher income
model	description	beta	p-value	beta	p-value
1	no reading exposure variables	0.061	0.029	0.081	0.038
2	family reading behavior	0.03	0.271	0.044	0.248
3	student reading behavior	0.015	0.551	0.048	0.179
4	adds earlier test scores	-0.012	0.613	0.02	0.542

The grade 3 fallacy

Sullivan and Brown's report presents counterevidence to the claim that grade 3 is magic, that if a child is not reading well by grade 3, the child will be "behind" forever.

Table 3 repeats some of the results from table 1.

Table 3: The impact of early reading proficiency

Variable	beta	p-value
Reads often at age 10	0.313	0
Reads books more than once/week at age 16	0.353	0
Reading proficiency at age five	0.039	0
Pictoral vocab at age five	0.091	0
Reading proficiency at age 10	0.117	0

Reading proficiency at age 10 is indeed a significant predictor of vocabulary knowledge at age 16.  But note that reading frequency is a stronger predictor than reading proficiency at age 10, both at age 10 and age 16 (compare the betas).  This supports the claim that we can improve in literacy development and "catch up" anytime, and that the way to do it is free voluntary reading (Krashen and McQuillan, 2007; Krashen, 2011b).

Causality

Sullivan and Brown point out that their analysis found that the students' own reading was a significant predictor even when reading proficiency measured at ages 5 and 10 was controlled (model 4; see table 1). This suggests "that the positive link between leisure reading and cognitive outcomes is not purely due to more able children being more likely to read a lot, but that reading is actually linked to increased cognitive progress over time." We first don't learn our skils and then use them for reading. Rather, our literacy development is the result of reading.

Krashen, S. 2011a. Protecting students against the effects of poverty: Libraries. New England Reading Association Journal 46 (2): 17-21.

Krashen, S. and McQuillan, J. 2007. Late intervention. Educational Leadership 65 (2): 68-73.

Krashen, S. 2011b. Need Children Read "Proficiently" by Grade 3? Some Possible Misinterpretations of the "Double Jeopardy" Study.  Language Magazine 11,2: 24-27.

Krashen, S., Lee, S.Y. and McQuillan, J. 2012. Is the library important? Multivariate studies at the national and international level. Journal of Language and Literacy Education, 8(1): 26-36.

Lance, Keith. The Impact of School Libraries on Student Achievement. http://www.lrs.org/impact.php

Sullivan, Alice and Brown, Matt. 2013. Social inequalities in cognitive scores at age 16: The role of reading. London: Centre for Longitudinal Studies,
Institute of Education, University of London   www.cls.ioe.ac.uk

Postscript: At about the same time this study with its strong evidence for the power of self-selected book reading was released, American Libraries  (Sept/Oct 2013) carried an article by Steve Coffman, "How low can our book budgets get?"

Coffman presented data showing that the public associates libraries primarily with books, but public libraries spend only 11.4% of their budget on books (in contrast, Netflix spends 56% of its budget on content), the number of books on public library shelves has dropped (in 1989, libraries bought 41.3 million volumes, in 2009, only 27.9 million), and public libraries' share of the book market has dropped from 4% in 1989 to 1% in 2009. 

All this while the number of books published in the US has increased from about 47,000 in 1990 to about 325,000 in 2010 and book sales have increased from 4.4 per capita in 1975 to 10 per capita in 2009, and of course while evidence for the positive effects of reading has been growing.