Contributions of print exposure to first and second grade oral language and reading in Chile
Abstract
This study examines the contribution of print exposure to oral language (expressive vocabulary and listening comprehension) and reading (word reading and reading comprehension) in first and second grade in Chile, and tests whether the contribution of print exposure to reading comprehension is mediated by language and word reading skills. Two‐hundred and eighty one children (mean age 6.55 years) participated. Print exposure was measured with a book‐cover recognition task in first grade, and outcomes were measured both in first and second grade. Print exposure had direct effects on all outcomes in first grade and indirect effects in second grade. Effects on first grade reading comprehension were partially mediated by listening comprehension and word reading, but not vocabulary. We discuss the importance of the findings for improving reading comprehension in countries with low access to books.
Contributions of print exposure to first and second grade reading in Chile
Exposure to print at an early age has long been known to contribute to oral language, word reading, and reading comprehension (Bus, Van Ijzendoorn & Pellegrini, 1995; Cunningham & Stanovich, 1991, 1997, 1998; Mol & Bus, 2011; Stanovich & Cunningham, 1992). However, the mechanisms through which print exposure contributes to reading comprehension have not been thoroughly studied. Furthermore, most studies of print exposure have been conducted in countries with wide access to books and where sharing books with young children is commonplace. This study examines the effects of print exposure in the context of a country with limited access to books. We measure the contribution that print exposure makes to vocabulary, word reading, listening comprehension, and reading comprehension in first grade, and we test the hypothesis that the effects of print exposure on reading comprehension are mediated by vocabulary, listening comprehension, and word reading. Finally, we follow children into second grade and evaluate whether print exposure measured in first grade continues to contribute to oral language and reading skills assessed after one year.
As evidenced by international tests such as the Programme for International Student Assessment (PISA), the reading comprehension level of children is in a critical state in Chile as well as in other countries in Latin America. In the last version of the PISA test, 33% of Chilean students did not reach level 2, which represents the minimum competence for extracting literal information from a written text (Agencia de Calidad de la Educación, 2014). To address this problem, Chilean policymakers have created incentives for schools to focus intensively on skills found to support reading comprehension, especially decoding skills such as word reading and reading speed (see Aikens & Barbarin, 2008; Bravo, 2002; Bravo, Villalón & Orellana, 2006; Kim & Pallante, 2012; Schatschneider, Fletcher, Francis, Carlson, & Foorman, 2004). The most salient of these incentives is a conditional subsidy given to schools if they measure reading speed and report it back to the Ministry of Education (Organization for Economic Cooperation and Development, 2010). However, predictors of reading comprehension other than word reading have not received equal attention in Chilean educational policy, in spite of the fact that listening comprehension and vocabulary have long been known to explain substantial variance in children's reading comprehension, over and above word reading (this is true in English‐language studies, e.g. Braze, Tabor, Shankweiler & Mencl, 2007; Cromley, Snyder Hogan & Luciw‐Dubas, 2010; Ouellete & Beers, 2010; Perfetti, 2007; Protopapas, Sideridis, Mouzaki & Simos, 2007; Tilstra, McMaster, Van den Broek, Kendeou & Rapp, 2009; Verhoeven & Van Leeuwe, 2008; as well as Spanish‐language ones, e.g. Canet‐Juric, Burin, Andrés & Urquijo, 2013; Caravolas, et al., 2012; Caravolas, Lervåg, Defior, Málková, & Hulme, 2013; Miranda, Fernández, García, Roselló, & Colomer, 2011). Therefore, in addition to focusing on word reading skills, educational policies that target listening comprehension and vocabulary may be relevant in promoting reading comprehension in countries with poor performance, such as Chile.
Increasing access to books may be such a policy. Exposure to print is a well‐known predictor of verbal skills ranging from vocabulary to written passage comprehension. Classic findings by Stanovich and Cunningham showed that measures of print exposure are associated with general knowledge (Cunningham & Stanovich, 1991, 1993, 1997; Stanovich & Cunningham, 1992), receptive vocabulary (Cunningham & Stanovich, 1991, 1997; Stanovich & Cunningham, 1992), word reading (Cunningham & Stanovich, 1991; Stanovich & Cunningham, 1992), spelling (Cunningham & Stanovich, 1991; Stanovich & Cunningham, 1992), and reading comprehension (Cunningham & Stanovich, 1997). These findings have been extended in subsequent studies with participants of different ages (Chateau & Jared, 2000; Davidse, de Jong, Bus, Huijbregts & Swaab, 2011; Ecalle, & Magnan, 2008; Gustafson, 2001; Martin‐Chang & Gould, 2008; McBride‐Chang, Manis, Seidenberg, Custodio, & Doi, 1993), as well as in two meta‐analyses demonstrating that exposure to written texts via shared reading or independent reading is a significant concurrent predictor of expressive and receptive vocabulary as well as word‐reading skills (Bus, et al., 1995; Mol & Bus, 2011).
Studies on print exposure, however, have been conducted only in a small number of countries (specifically Canada, France, USA, and the Netherlands). This presents a problem because exposure to print in the early years is likely to vary across societies. Contextual variables such as attitudes about reading or access to books may influence the frequency with which parents read to their children (Britto, Fuligni & Brooks‐Gunn, 2002; Coddington, Mistry & Bailey, 2014; Neuman, 1999). For example, placing a low value on reading for pleasure is likely to diminish a parent's interest in reading with their child. In Chile, this factor may be relevant, because 52.4% of adults surveyed in a representative sample declared that they had not read any books in the previous year (Consejo Nacional de la Cultura y las Artes, 2013). Consistently, the frequency of shared book reading among parents and young children in Chile and other Latin–American countries has been shown to be low (Strasser & Lissi, 2009; Romero‐Contreras, 2007; Romero‐Contreras, Arias, & Chavarría, 2007). In addition to low interest in reading, factors that reduce access to books may contribute to low frequency of book sharing with children. Cost is one such factor. In the United States, a paperback storybook may cost between 7 and 12 USD (US dollars), while the same kind of books usually start at 12 USD in Chile. This difference in price is aggravated by the fact that while the USA's GDP per capita (PPP) in 2014 was of 54,629 USD, Chile reached only 22,333 USD, according to the International Monetary Fund. Public libraries do not necessarily compensate the high cost of books. Chile does not have a law regulating the existence of public libraries, which has resulted in heterogeneous access to this resource, and large urban areas exist where it is impossible to find a library within a radius of several kilometres. For example, the Metropolitan Region of Santiago de Chile has a population of 7.2 million people and only about 62 libraries (Consejo Nacional de la Cultura y las Artes, 2011). This makes for about one library for every 117 thousand inhabitants, in contrast to an estimated 1,081 public libraries in the city of New York for a population of 8.2 million inhabitants – one public library per 7 thousand inhabitants (Source: our elaboration on public library listings in www.publiclibraries.com).
Given this background, it is reasonable to question whether print exposure is as strong a factor in determining Chilean children's language and reading as it is in countries with wide access to books. Thus, our first goal is to determine whether a measure of print exposure explains significant variance in the vocabulary, listening comprehension, word reading, and reading comprehension of Chilean first graders.
We also aim to examine whether the possible contribution of print exposure on reading comprehension is mediated by vocabulary, listening comprehension, and word reading. Researchers have proposed that the effects of print exposure on reading comprehension may occur through at least three intervening variables: word reading, vocabulary (Cunningham and Stanovich, 1991; Ecalle & Magnan, 2008; Stanovich, 1986) and listening comprehension (Hedrick & Cunningham, 2002). Given these proposals, it is to be expected that the effect of print exposure be fully mediated by those variables. There is some evidence that those variables could mediate the relation between print exposure and reading comprehension. For example, Kail, Hall and Caskey (1999) found that the contribution of print exposure to reading comprehension in elementary school children disappeared when word reading was controlled, suggesting mediation. Sénéchal and LeFevre (2002) found evidence that a factor composed of receptive vocabulary and listening comprehension mediated the relation between parents' reports of reading at home with their children, and a factor including three literacy skills in third grade. However, vocabulary, listening comprehension and word reading have only been studied separately as mediators of print exposure effects on reading comprehension, and never in the same model. Therefore, our second goal is to examine a mediational model in which the contributions of print exposure on reading comprehension are effected through listening comprehension, vocabulary and word reading skills.
Finally, our third goal is to evaluate whether print exposure measured in first grade continues to affect reading comprehension one year later (second grade), and the mechanisms through which it does. Stanovich (1986) proposed that the effects of print exposure should intensify during development because children who are exposed to more books develop both better oral language and better word reading skills, which in turn increase their independent reading, creating a virtuous cycle. Consistently, several studies have found that measures of print exposure predict reading comprehension growth. Echols, West, Stanovich and Zehr (1996) showed that children's responses to a book title checklist accounted for significant variance in receptive vocabulary, spelling, sight vocabulary, and reading comprehension measured in fifth, sixth, and seventh grade, even after controlling for children's performance in the previous year. Cipielewski and Stanovich (1992) found similar results for fifth grade reading comprehension, controlling for third grade reading comprehension. These results suggest that print exposure does explain not only initial reading ability, but also growth. Yet these studies all start in third grade at least, at an age where most print exposure comes from independent reading. In beginning readers, however, things may be different, because children would presumably get the majority of their print exposure through being read to (Bus, 2003; Mol & Bus, 2011). Sénéchal and LeFevre (2002) showed that the number of books children recognized from key illustrations in first grade predicted their third‐grade reading comprehension, after controlling for first‐grade receptive language, emergent literacy, and word reading. But, because that study did not control for first‐grade reading comprehension, it is not clear that print exposure could explain growth in that skill at this early age. Additionally, those studies do not test the independent effects of print exposure and reading ability on reading comprehension improvement. According to Stanovich's hypothesis (1986), as children improve in their ability to read, this improves both their vocabulary and their word reading. This should be reflected in an effect of reading comprehension skill on the second grade predictors of reading comprehension. Thus, we aim to compare two models of print exposure effects on second grade reading comprehension, one that includes only effects of print exposure on grade two vocabulary, listening comprehension, and word reading, and one that also includes effects of grade one reading comprehension on those variables.
In sum, the study had three main goals: to evaluate the contribution of print exposure to vocabulary, listening comprehension, word reading, and reading comprehension in first grade and second grade in Chile, to test whether the contribution to reading comprehension in first grade occurs through vocabulary, listening comprehension, and word reading, and finally, to determine whether print exposure measured in first grade continues to influence vocabulary, listening comprehension, word reading and reading comprehension in second grade, and how. Given previous research and theory, we hypothesized that print exposure would contribute significantly to oral language and reading in first grade, and that its contribution to reading comprehension would be mediated via oral language and word reading. As for second grade, we expected print exposure measured in grade 1 to make both direct and indirect contributions to grade 2 oral language and reading.
Methodology
Participants
Participants were 281 first grade children (42% female), between the ages of 66 and 89 months, from seven voucher (82.9%) and one private school (17.1%) in the Santiago de Chile Metropolitan Region, with an average age of 6.55 years at the time of the reading assessments in first grade (SD = 3.92 months).
The private school was slightly more affluent than the seven voucher schools included in the sample. The average family monthly income of the seven voucher schools was US$1,000, while that of the private school families was US$1,360. Average parental education was 12 years in the voucher schools and 14 years in the private school. Although private schools in Chile can charge as much as US$1,200 in monthly tuition, this specific private school's tuition was only US$175, close to the monthly tuition of the most expensive voucher school in our sample (US$119). Additionally, the voucher and private schools were similar in curriculum and teaching. Chile has a national curriculum that is quite prescriptive and followed closely by the vast majority of private schools. Furthermore, empirical studies of actual classroom activities and teaching practices show that in Chile, teaching practices are very similar in private, public, and voucher schools (Preiss, Grau, Nuñez, Alegría, Espinoza, & Ponce, 2011 2011; Strasser & Lissi, 2009). Although many private schools are bilingual, the private school in our sample was not (they only provided 6 h per week of English‐as‐a‐second‐language instruction).
Data on individual family income were unavailable. However, we used school tuition as a proxy for children's Socio‐Economic Status (SES), including it as a control in all analyses. This approach is justified by the high socioeconomic segregation of the Chilean educational system (Valenzuela, Bellei & de los Ríos, 2013), which ensures that children attending a school will be very similar to each other in terms of SES. Studies have shown that in Chilean schools, the school's socioeconomic composition explains as much variance in educational achievement as individual socioeconomic characteristics (Mizala & Torche, 2012).
In first grade, 71 children could not take the reading comprehension measure, because they were not reading words yet. As expected, children who could not take the reading comprehension test in first grade had significantly lower scores than their peers in several measures at time one: letter–word identification (t = 6.69, df = 279, p = .000), and vocabulary (t = 2.41, df = 279, p = .014). They were similar to their peers in their exposure to books and listening comprehension at time one, and they were not more likely to be female than children who completed the first grade reading comprehension measure, but they were more likely to come from a voucher school (X2 = 9.84, df = 1, p = .001). Given that missing data in the first grade reading comprehension measure is dependent on word reading and vocabulary, and those were included in the models, the missing data is conditionally random (MAR). Missing data was treated using full information maximum likelihood, which provides unbiased estimates in these cases (Graham, 2008).
Of the 281 children assessed in first grade, 233 were assessed again in all variables in second grade (41.8% female, 81.8% voucher, Mean age = 6.56 months, SD = 3.96 months). These children constitute the longitudinal sample. The majority of the 48 children who dropped out of the study moved to a different school (N = 43) and the other five could not be assessed because of repeated absences. Children who dropped out of the study between first and second grade had significantly lower vocabulary scores (t = 2.87, df = 279, p = .004) than the ones that remained. Children who dropped out were not more likely to be female or to come from a voucher school than children who remained, and their average tuition value was the same as that of children assessed at both times.
Procedures
Principals were contacted by phone and then in person to explain the procedures of the study. Parents were contacted during class meetings to explain study procedures, and they were asked written consent for their children's participation. About two thirds of the contacted parents gave consent. Children were asked at the beginning of each assessment session if they wanted to participate. No child refused assessment. The measures were administered in three sessions of 20 to 45 min of duration, depending on the child. Because of school time and space constraints, we could only assess a few children per day per school. Some of the schools administer standardized assessments in first grade, and all second grade children receive a national standardized test that has relevant consequences for the schools. This meant that teachers and administrators were reluctant to let children out of the classroom during math, language, and science periods, and only allowed testing during physical education, music and orientation. This, added to space constraints that only allowed us to test one child at a time in each school, meant that the evaluation schedule had to be extended for several months. In first grade, assessments were administered during the first three quarters of the school year, as follows: Session 1 (months 2 and 3 of the school year): print exposure, word reading, expressive vocabulary; Session 2 (months 4 and 5): listening comprehension; Session 3 (months 7 and 8): reading comprehension. Assessment in the second grade was conducted in the same order to keep the time between measures constant (12 months).
Measures
Word reading
To assess word‐reading skills, we used the Letter–Word Identification Scale of the Woodcock–Muñoz Batería III. This is the Spanish‐language version of the Woodcock–Johnson Achievement battery, developed and tested with native Spanish speakers from several Latin American countries. The scale has 76 items. Reported reliability on this subscale is approximately .9 for this age range (Muñoz‐Sandoval, Woodcock, McGrew, & Mather, 2005). No Chilean norms exist, so raw scores were used, and age of children at the time of the assessment was used as a control.
Expressive vocabulary
The Vocabulary scale of the Chilean version of the Wechsler Intelligence Scale for children (WISC IIIv.ch.) was administered in first and second grade. This is an expressive vocabulary scale that requires children to define 26 words (administration is suspended after four consecutive errors). Each definition receives a score of 0, 1, or 2 depending on completeness and form. Raw scores were used because they presented a better distribution than Chilean standard scores. Reliability reported by the authors (Cronbach's alpha) is .79 (Ramírez & Rosas, 2007).
Listening comprehension
Listening comprehension was measured with different instruments in first and second grade, because the instrument used in first grade approached a ceiling effect and therefore had to be replaced for second grade. In first grade, we used a translated and adapted version of the Test of Story Comprehension (Petersen & Spencer, 2012). This measure consists of a pool of 40 short stories (about 70 words long) with the same underlying structure (main character has a goal, faces a problem, attempts a solution, and finally solves the problem, experiencing different emotions along the way). Children are asked six questions about each story, referring to the characters, problems, resolution, and emotions. For the present study, we translated and piloted several of the stories and chose three of them to administer in the study. Each answer could receive a score of 0, 1, or 2 points. The complete test has a total of 18 questions and a maximum score of 36 points. Answers were double coded and interrater reliability calculated with Cohen's Kappa coefficient. All Kappa coefficients were significant and reached .7 or superior. Internal consistency of the 18 questions was .73 (Cronbach's alpha). In second grade, we used an auditory version of the reading comprehension measure (the passage comprehension subscale of the LEE reading achievement test). To do this, we analysed the passages in the written test and constructed similar passages in topic, structure, number of words, and legibility using the Spanish perspicuity formula (Szigriszt‐Pazos, 1993). The resulting listening comprehension test, like the original reading comprehension test, had three passages, one informative and two narrative. Children were presented the passages in a recorded format, twice, to ease the burden on working memory. The presentation of the questions follows the same procedure as the written passage test. For each passage, children have to answer eight questions (six open and two multiple choice) requiring retrieval of literal information as well as drawing inferences and summarizing. Questions are read aloud, and children are told that they can listen to the passages again if they wish, although no children chose this option. Cronbach's alpha of the 24 questions, calculated from our own data, is .68, lower than the written passage version of the test.
Print exposure
Print exposure was measured with a book‐cover recognition scale. The use of illustrations from storybooks has previously been found to be an appropriate and unbiased measure of print exposure in non‐readers (Sénéchal, LeFevre, Hudson & Lawson, 1996), and book covers specifically have been shown to behave as an adequate material to assess this construct (Davidse et al., 2011). We chose to use book covers because we wanted our measure to be less conservative in that it would capture books that children may have seen or manipulated at home or in the library but not necessarily read or been read. Given our expectation that book knowledge would be very low, it was important that we captured as much of the variance of children's exposure to books as possible. Accordingly, scoring was done so that contact with the book, rather than actual reading of it, was privileged. Thus, children were given one point if they gave the title or any other relevant information about the book that could not be inferred directly from the cover illustration. The test consisted of a booklet with covers of 41 books from which the title was deleted. The 41 titles were chosen from bookstore recommendations and bestseller lists, and the Ministry of Education recommended reading list. Most of the books chosen had only one possible cover illustration, but five were fairy tales that could have many different covers; we chose a cover that included all the main elements of the story (for example, for Little Red Riding Hood you can see the main character with her basket, and the wolf in the background, peeking out from behind a bush). We excluded titles that had TV shows or movies. The final instrument included 22 titles from the Ministry of Education recommendations (13 recommended for 1st grade; six for 2nd to 6th grade; and two for preschool); six comic books popular in bookstores at the time; six classical fairy tales; and seven storybooks recommended by bookstores for this age range. Internal consistency of this instrument, calculated from our own data, was .64 (Cronbach's alpha).
Reading comprehension
The text comprehension subtest of the LEE reading achievement test in Spanish was administered in first and second grade (Defior, Fonseca, Gottheil, Aldrey, Rosa, Pujals, et al., 2006). The test includes three passages (two narrative, one expositive). For each passage, children have to answer eight questions (six open and two multiple choice) requiring retrieval of literal information as well as drawing inferences and summarizing. Questions are read aloud, and children are allowed to go back to the passages as they wish. Cronbach's alpha of the 24 questions, calculated from our own data, is .94.
Results
Descriptive statistics
Table 1 shows descriptive statistics. Several dependent variables showed different degrees of deviation from normality‐specifically skewness. Upon examination of distributions, raw vocabulary scores proved to be more normally distributed than standardized vocabulary scores in both years, so raw scores were used, given that the correlation between them was higher than .9 in both cases. For the rest of the variables, the most serious departures from normality were those of the letter–word identification scores in first grade, which was positively skewed, and those of the two listening comprehension variables, which were negatively skewed. All three variables were transformed to achieve more symmetry, and analyses were performed with both the original and transformed variables, and because significance results did not change, all analyses are reported with the original, untransformed variables.
| N | Mean(SD) | Age(SD)a | Skewness | Kurtosis | Min | Max | |
|---|---|---|---|---|---|---|---|
| Monthly school tuition in USD | 281 | 39.89(36.42) | — | 1.16 | −0.46 | 17 | 110 |
| Print exposure in Grade 1 | 281 | 5.83(2.80) | 78.60(3.89) | 0.60 | 0.76 | 0 | 17 |
| Word reading Grade 1 | 281 | 23.49(16.05) | 78.60(3.89) | 1.02 | 0.09 | 4 | 71 |
| Word reading Grade 2 | 238 | 49.08(14.18) | 90.28(3.91) | −0.13 | −0.77 | 14 | 75 |
| Listening comprehension Grade 1 | 281 | 27.17(4.98) | 80.47(3.86) | −0.74 | 0.82 | 7 | 36 |
| Listening comprehension Grade 2 | 233 | 32.98(5.99) | 92.41(3.91) | −0.72 | 0.80 | 12 | 45 |
| Expressive vocabulary Grade 1 | 281 | 13.57(5.10) | 78.60(3.89) | 0.14 | 0.51 | 0 | 32 |
| Expressive vocabulary Grade 2 | 238 | 18.26(5.29) | 90.28(3.91) | −0.06 | −0.48 | 5 | 31 |
| Reading comprehension Grade 1 | 210 | 29.90(6.15) | 82.21(3.88) | −0.24 | −0.55 | 13 | 43 |
| Reading comprehension Grade 2 | 233 | 34.65(6.10) | 95.41(3.92) | −0.56 | 0.07 | 16 | 48 |
- Note.
- a Age = age in months at which the measure was administered.
As shown in Table 1, on average, children recognized about six of the 41 book covers (M = 5.83, SD = 2.80). The frequency distribution of this variable shows that 51.3% (n = 144) of children recognized 5 or fewer titles, while 89.7% (n = 252) recognized nine or fewer book covers. A title‐by‐title analysis shows that the category most recognized was classical fairy tales/fables, where children recognized an average of 2.46 titles out of 6 (SD = 1.36, min = 0, max = 6) (the six fairy tales were Ugly Duckling, Little Red Riding Hood, Lead Soldier, Golden Egg Hen, Turtle and the Hare, and Goldilocks), followed by comic books, of which children recognized an average of 1.48 titles out of 6 (SD = 0.96, min = 0, max = 5). Among the titles taken from the reading list recommended by the Ministry of Education for first grade, children recognized an average of 1.55 titles out of a possible 13 (SD = 1.19, min = 0, max = 7).
Contributions of print exposure to reading comprehension in first grade
Our first goal was to determine whether a measure of print exposure explains significant variance in the vocabulary, listening comprehension, word reading, and reading comprehension of Chilean first graders. We first analysed correlations between print exposure and outcome measures. As shown in Table 2, print exposure has significant correlations with all language and reading measures, except word reading in second grade. The correlations with print exposure are similar in first and second grade for listening comprehension and expressive vocabulary, but significantly greater in first than second grade in the case of word reading (Z = 2.87, p = .004) and reading comprehension (Z = 2.28, p = .011).
| 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | |
|---|---|---|---|---|---|---|---|---|---|---|
| 1.Monthly school tuition in USD | ||||||||||
| 2.Print exposure in Grade 1 | −.09 | |||||||||
| 3.Word reading Grade 1 | .14 | .29 | ||||||||
| 4.Word reading Grade 2 | .09 | .04 | .37 | |||||||
| 5.Listening comprehension Grade 1 | .02 | .21 | .09 | .03 | ||||||
| 6.Listening comprehension Grade 2 | .12 | .19 | .25 | .10 | .15 | |||||
| 7.Expressive vocabulary Grade 1 | −.03 | .26 | .22 | .10 | .19 | .30 | ||||
| 8.Expressive vocabulary Grade 2 | .06 | .25 | .31 | .21 | .17 | .25 | .28 | |||
| 9.Reading comprehension Grade 1 | .10 | .41 | .41 | .17 | .29 | .45 | .25 | .37 | ||
| 10.Reading comprehension Grade 2 | −.00 | .20 | .37 | .21 | .18 | .59 | .29 | .27 | .48 |
- Note.
- N = 281 for correlations among first grade variables, except for Reading Comprehension where N = 210.
- N = 233 for correlations involving first and second grade variables, except for those including G1 Reading Comprehension where N = 179.
- rs > .13 are significant at p < .05.
Next, we conducted a path analysis to identify direct and indirect effects of print exposure on reading comprehension in first grade. The model was tested with MPlus (Muthen & Muthen, 1998–2012), using a bootstrapping technique with 10,000 iterations to generate significance tests and confidence intervals for the estimated parameters (Hayes, 2013). We included gender and tuition fees as predictors of reading comprehension in all path analyses. For all measures we included as a control the age at which the measure was administered, to account for the fact that we used raw scores and that measures were administered with differences of between two and six months. In the first grade model, print exposure was used to predict expressive vocabulary, listening comprehension and word reading. Additionally, expressive vocabulary was used to predict listening comprehension in the model because of its known role as a predictor of that variable (Strasser, del Río, & Larrain, 2013; Florit, Roch, Altoe, & Levorato, 2009; Florit, Roch, & Levorato, 2011; Kim, 2015; Lepola, Lynch, Laakkonen, Silvén, & Niemi, 2012). Print exposure, expressive vocabulary, word reading, and listening comprehension (plus controls) were used to predict Reading Comprehension. The model tested is shown in Figure 1. Results from the analysis are shown in Table 3.
| b | SE | b/SE | Bootstrapped 95% confidence intervals for b | |
|---|---|---|---|---|
| Direct effects on G1 word reading | ||||
| G1 print exposure | 0.27** | 0.05 | 5.05 | [0.16 – 0.38] |
| Age | 0.11 | 0.06 | 1.79 | [−0.01 – 0.22] |
| Direct effects on G1 vocabulary | ||||
| G1 print exposure | 0.26** | 0.06 | 4.53 | [0.14 – 0.37] |
| Age | 0.03 | 0.07 | 0.47 | [−0.10 – 0.15] |
| Direct effects on G1 listening comp. | ||||
| G1 print exposure | 0.17** | 0.06 | 2.78 | [0.04 – 0.28] |
| G1 vocabulary | 0.14* | 0.07 | 2.13 | [0.01 – 0.27] |
| Age | 0.08 | 0.06 | 1.24 | [−0.05 – 0.19] |
| Direct effects on G1 RC | ||||
| G1 print exposure | 0.29** | 0.06 | 4.94 | [0.17 – 0.41] |
| G1 word reading | 0.27** | 0.06 | 4.19 | [0.14 – 0.39] |
| G1 vocabulary | 0.07 | 0.07 | 1.01 | [−0.06 – 0.19] |
| G1 listening comprehension | 0.21** | 0.06 | 3.78 | [0.10 – 0.32] |
| Age | 0.09 | 0.06 | 1.39 | [−0.04 – 0.21] |
| Gender (1 = female, 0 = male) | −0.04 | 0.06 | −0.74 | [−0.16 – 0.07] |
| Tuition fees | 0.03 | 0.07 | 0.49 | [−0.09 – 0.16] |
| Indirect effects of PE on G1 RC | ||||
| Through word reading | 0.07** | 0.02 | 3.36 | [0.04 – 0.12] |
| Through vocabulary | 0.02 | 0.02 | 0.96 | [−0.01 – 0.06] |
| Through listening comprehension | 0.04* | 0.02 | 2.23 | [0.01 – 0.08] |
- Note.
- N = 281; G = grade; PE = print exposure; RC = reading comprehension.
- * p < .05.
- ** p < .01.
The first grade model had a good fit (χ2 = 29.247, df = 20, p = .083; RMSEA = 0.041, CFI = 0.938, TLI = 0.894). Table 3 presents the path coefficients with confidence intervals. Print exposure has significant unique effects on all four first grade dependent variables. Our second goal was to examine whether the effect of print exposure on reading comprehension in first grade is mediated by vocabulary, listening comprehension, and word reading, as has been proposed in the literature. According to most accounts, children who are exposed to more books should develop better vocabulary and word reading skills, which in turn affect their reading comprehension. The effect of print exposure on reading comprehension in first grade was indeed partially mediated, but only by word reading and listening comprehension, not by vocabulary. After accounting for these indirect effects, there remained a significant direct effect of print exposure on grade one reading comprehension. Effect sizes (Cohen's d) for the effects of print exposure are 0.357 for word reading, 0.521 for expressive vocabulary, 0.545 for the direct effect on reading comprehension, and 0.601 for listening comprehension. These effects range from small to medium sized.
Contributions of first grade print exposure to expressive vocabulary, listening comprehension, word reading, and reading comprehension in second grade
Finally, our third goal was to evaluate whether the effects of print exposure measured in first grade continue to affect reading comprehension one year later, and to compare two alternative models of how this influence may occur. The two models tested are shown in Figure 2. In model 2A, the effect of print exposure measured in first grade on second grade reading comprehension occurs through first grade reading comprehension and through second grade language mediators (vocabulary, word reading, and listening comprehension). However, Stanovich (1986) proposed that reading competency also influences the subcomponents of reading, that is, that children who are better at reading would in turn increase their vocabulary and word reading skills at a faster rate. Therefore, we tested a second model with paths from first‐grade reading comprehension to expressive vocabulary, listening comprehension, and word reading in second grade (model 2B).
The models were tested using the same technique as the first grade model (bootstrapping with 10,000 iterations). The chi‐square difference test (χ2diff) comparing the fit of both models indicated a statistically significant better fit of model 2B, Δχ2 = 58.878, Δdf = 3, p = .000. Thus, the model that includes a path from first grade reading comprehension to mediators of reading comprehension in first grade describes the data better. Fit for the final model was adequate (χ2 = 44.567, df = 26, p = .013; RMSEA = 0.055, CFI = 0.928, TLI = 0.875). Results for the final model are presented in Table 4.
| b | SE | b/SE | Bootstrapped 95% confidence intervals for b | ||
|---|---|---|---|---|---|
| Direct effects on G2 word reading | |||||
| G1 print exposure | −0.06 | 0.07 | −0.90 | [−0.20 | 0.08] |
| G1 reading comprehension | 0.23** | 0.07 | 3.12 | [0.08 | 0.38] |
| Age | −0.03 | 0.06 | −0.53 | [−0.16 | 0.09] |
| Direct effects on G2 vocabulary | |||||
| G1 print exposure | 0.08 | 0.06 | 1.35 | [−0.03 | 0.20] |
| G1 reading comprehension | 0.36** | 0.07 | 5.22 | [0.22 | 0.49] |
| Age | 0.07 | 0.06 | 1.12 | [−0.05 | 0.19] |
| Direct effects on G2 listening comp. | |||||
| G1 print exposure | 0.00 | 0.06 | 0.11 | [−0.11 | 0.12] |
| G1 reading comprehension | 0.44** | 0.07 | 6.35 | [0.30 | 0.57] |
| G1 vocabulary | 0.09 | 0.06 | 1.52 | [−0.02 | 0.21] |
| Age | −0.09 | 0.06 | −1.42 | [−0.21 | 0.03] |
| Direct effects on G1 reading comp. | |||||
| G1 print exposure | 0.37** | 0.06 | 6.03 | [0.24 | 0.49] |
| Direct effects on G2 RC | |||||
| G1 print exposure | −0.03 | 0.05 | −0.60 | [−0.12 | 0.07] |
| G1 reading comprehension | 0.26** | 0.06 | 4.06 | [0.13 | 0.39] |
| G2 word reading | 0.11* | 0.05 | 1.98 | [0.00 | 0.22] |
| G2 vocabulary | 0.02 | 0.05 | 0.55 | [−0.07 | 0.13] |
| G2 listening comprehension | 0.46** | 0.05 | 7.98 | [0.34 | 0.57] |
| Age | 0.07 | 0.05 | 1.44 | [−0.02 | 0.18] |
| Gender (1 = female, 0 = male) | −0.01 | 0.05 | −0.32 | [−0.11 | 0.08] |
| Tuition fees | −0.11* | 0.05 | −2.05 | [−0.21 | −0.00] |
| Indirect effects of PE on G2 RC | |||||
| Through G1RC | 0.09** | 0.02 | 3.43 | [0.05 | 0.16] |
| Through G1RC and word reading | 0.01 | 0.00 | 1.63 | [0.00 | 0.02] |
| Through G1RC and vocabulary | 0.00 | 0.00 | 0.53 | [−0.01 | 0.02] |
| Through G1RC and listening comp. | 0.07** | 0.02 | 3.77 | [0.04 | 0.12] |
- Note.
- N = 233.
- * p < .05.
- ** p < .01.
- G = grade.
- PE = print exposure.
- RC = reading comprehension.
As shown in Table 4, after controlling for the first grade reading comprehension score, print exposure does not have direct effects on expressive vocabulary, listening comprehension, word reading, or reading comprehension in second grade. This means that, in this sample, the effect of print exposure that children had in first grade does not continue to act on their skills' growth, but rather, it affects the starting point of their reading skills, and that is what predicts later reading and language scores.
Discussion
Our results show that exposure to print is an important factor in explaining young children's oral language and reading comprehension even in a context where access to books for beginning readers is limited, as is the case in many developing countries. This is relevant given that it suggests that facilitating access to books may be a useful strategy for these countries to increase their reading comprehension results. We also found that the contribution of print exposure to reading comprehension in first grade was partially mediated by listening comprehension and word reading. Finally, the longitudinal results show that print exposure measured in first grade has an indirect effect on listening comprehension, expressive vocabulary, word reading, and reading comprehension in second grade through reading comprehension in first grade. However, first grade print exposure did not have a direct effect on any of the second grade variables. A model with indirect effects of print exposure on second grade variables (through first grade reading comprehension) fitted the data better than a model that only included direct effects of print exposure to second grade variables. This suggests that the reading comprehension achieved in first grade is determinant in the development of expressive vocabulary, listening comprehension, and word reading. The only significant effects on second grade reading comprehension were first grade reading comprehension and second grade listening comprehension, although word reading approached significance at p = .082. Expressive vocabulary did not have a significant unique contribution in this model.
The first grade results are consistent with previous findings in Europe, Canada, and the USA (Chateau & Jared, 2000; Davidse et al., 2011; Ecalle & Magnan, 2008; Kail et al., 1999; Sénéchal & LeFevre, 2002), and they expand on those findings by explicitly testing the mediation mechanism proposed by Stanovich (1986) in first grade. Furthermore, our findings show that subsequent effects of print exposure on second grade reading comprehension do not only occur through autoregressive effects of first grade reading comprehension on second grade reading comprehension, but also through effects of first grade reading comprehension on second grade expressive vocabulary, word reading, and listening comprehension.
The absence of an indirect effect through expressive vocabulary in first grade was surprising, given that vocabulary has been found to be related to both reading comprehension and print exposure in the literature. In our model, this result may be because of shared variance of expressive vocabulary with either listening comprehension or word reading, both of which had significant bivariate correlations with expressive vocabulary. The stronger role of listening comprehension and word reading in the mediation process suggests that when children are exposed to books, some of the most important advantages they gain are word‐reading skills and familiarity with textual structures and syntactic choices that are typical of the written register.
In this Spanish‐language sample, word reading proved to be a mediator of the effect of print exposure on reading comprehension in first grade, similar to previous results obtained in an English‐language sample (Kail et al., 1999). This is consistent with research showing that although in Spanish and other transparent orthographies children acquire decoding relatively faster than in opaque ones, the relative contributions of word reading and oral language to reading comprehension are similar in both types of writing systems (Caravolas et al., 2012; Caravolas et al., 2013). The fact that print exposure makes positive contributions to both word reading and oral language (expressive vocabulary and listening comprehension) in first grade emphasizes the potential of policies that strengthen book access as a means for achieving gains in both areas.
The direct effect of print exposure on first‐grade reading comprehension after controlling for its indirect effects was surprising. Various researchers have proposed that print exposure would affect reading comprehension through improvements in word reading, vocabulary, and listening comprehension (Cunningham and Stanovich, 1991; Ecalle & Magnan, 2008; Hedrick & Cunningham, 2002; Stanovich, 1986), and therefore we expected that once those measures were included in the model, the effect of print exposure would cease to be significant. One possibility is that our measures did not capture enough variance in children's word reading, vocabulary, and listening comprehension. For example, we used an expressive vocabulary measure and did not include a receptive vocabulary measure. Similarly, our listening comprehension measure had ceiling effects in first grade, and therefore it may have failed to capture variance in children at the top of the distribution. Additionally, we did not include a measure of pseudo word reading, which may account for additional variance in children's reading ability. Future research that aims at fully explaining mediation should use a wider pool of measures of reading comprehension predictors. Another explanation for the remaining direct effect of print exposure on reading comprehension is that our print exposure measure is indexing other student characteristics that explain additional variance in reading comprehension, such as cultural capital or SES, because we only controlled for school tuition as a proxy for SES. A design that includes an individual measure of contextual or demographic characteristics is needed to examine this issue.
Effect sizes for the contributions of print exposure to first‐grade language and reading skills in the present study are small to moderate. Comparison to other studies is difficult because of variations in age, measure of print exposure, and control variables. Perhaps, the most comparable study is that of Davidse et al. (2011), who used recognition of book covers to measure print exposure in kindergarten children, and found effect sizes of 0.49 on letter knowledge and 0.42 on vocabulary (respectively, slightly larger and slightly smaller than the ones in the present study). Another study that uses a similar measure to ours is that of Sénéchal et al. (1996), who measured storybook exposure with book illustrations in children from 2 to 5 years of age, finding an effect size of 1.04 on expressive vocabulary, much larger than the one in our study. Finally, another somewhat comparable study is that of Sénéchal et al. (2002), who measured storybook exposure with book illustrations at the end of first grade and evaluated its effect on reading in third grade, finding an effect size of 0.57, whereas we found no direct longitudinal effect from first to second grade. Studies that use measures administered to parents of preschool children (e.g. Sénéchal et al., 1998), as well as studies that use title recognition checklists administered to older children (e.g. Davidse et al, 2011, Ecalle and Magnan, 2008) find effect sizes ranging from 0.2 to 0.74, similar to the range found in the present study.
The second grade model shows no direct effects of print exposure on expressive vocabulary, listening comprehension, word reading, or reading comprehension. The only effects of print exposure occur indirectly through reading comprehension in first grade, which does explain significant variance in all second grade dependent variables. The lack of direct effects of print exposure on second grade variables is inconsistent with longitudinal findings in older children that show that print exposure measured through title recognition checklists explained the reading comprehension of children in fifth, sixth, and seventh grade, after controlling for the autoregressor (Echols et al., 1996, Cipielewski & Stanovich, 1992). If the effects of print exposure do indeed intensify as differences between poor readers and good readers become larger (Stanovich, 1986), it is possible that at this early stage in development it is more difficult to detect effects on growth than it is at later ages, when differences have grown more acute. Additionally, sources of print exposure change with development. Our print exposure measure was administered at the beginning of first grade, and it focused strongly on storybooks, probably capturing children's past experiences with parent‐book sharing in the home, which in turn reveal parents' motivation to expose children to books (Bus, 2003). That young children's exposure to books is associated with their parents' behaviour is supported by the fact that Sénéchal et al. (1996) found correlations of about .45 between preschool children's recognition of book illustrations and parents' recognition of children's titles and/or authors. In contrast, print exposure measures administered to older children partly assess leisure time reading motivated by the children themselves (Mol and Bus, 2011). Therefore, both the magnitude and the sources of print exposure variance may change between first grade and higher grades. Longitudinal designs with different measures of print exposure at different times may be needed in order to determine whether print exposure effects change as children progress from being read to, towards reading independently, and whether print exposure effects intensify with age.
Children in our sample recognized few books. This is problematic given our results that book exposure may be an important factor in determining reading success. The Chilean Ministry of Education has implemented some policies targeting book access. Most notably, since 2004, a program is in place to provide all public schools with libraries. Current ministry's statistics indicate that 85% of public primary schools have a library, but data about how those libraries are being used is scarce. One study showed that on average, primary students check out six books per year and visit the library around once a month (MINEDUC, 2014), suggesting that the school libraries are not compensating for children's lack of book exposure at home. This would be consistent with the fact that children in our sample recognized only about one book out of 13 ministry‐recommended titles, in spite of those titles being stocked in all school libraries. More studies about the use of school libraries in Chile are needed to make a more efficient use of these resources.
It is possible that in contexts with reduced access to print oral narratives can provide similar effects as book exposure. Reese and Leyva (2010) found that training low income mothers in the United States to conduct elaborative reminiscing was more effective for boosting preschool children's oral language than training mothers in dialogic reading. However, this study was conducted in the United States, where even low income families share books more frequently than Chilean middle‐income families (Strasser & Lissi, 2009). The few existing studies on Chilean parent‐child oral narratives (Nolivos & Leyva, 2013; Leyva, Berrocal & Nolivos, 2014; Leyva & Nolivos, 2015) have not sought to determine whether frequency or style of oral storytelling is related to children's gains in language. Therefore, there is no evidence that oral storytelling can replace book exposure as a language development strategy in contexts such as Chile. Future research could examine whether oral storytelling in Chilean homes or classrooms contributes to children's oral language. Nevertheless, the positive effects of book exposure, which have been replicated across several countries (now including Chile), should not be dismissed as a potential policy cornerstone for improving reading outcomes.
This study presents several limitations that constrain its conclusions. Seventy one children were not able to take the reading comprehension test in first grade, potentially biasing results. A second limitation is the lack of an individual socioeconomic status measure, which in this study was replaced by the school's monthly tuition. Although the segregated nature of the Chilean school system warrants such an approach, an individual measure of income or parental education is the best option when dealing with a variable that is so correlated with socioeconomic factors.
These results expand our understanding of how print exposure influences the reading comprehension of beginning readers, and highlight the importance of print exposure, especially in countries with limited access to books. For educational systems that need to boost their students' reading comprehension, creating policies that bring all children closer to books may be key. Where public libraries are insufficient, school libraries need to play a more active role in the lives of children, motivating children to take the books home and parents to interact around them. Additionally, policies such as lowering taxes on children's literature may be helpful. Improving book access may be a high‐impact measure for poor countries, one which could provide an alternative to more complex and expensive policies such as curriculum reform or extensive in‐service teacher training.
Acknowledgements
Support for this research was provided by the Chilean National Fund of Scientific and Technology Development (Fondecyt) through grant N 1110857.
Biographies
Katherine Strasser: PhD in Education and Psychology (University of Michigan), works as an Associate Professor of Psychology at Pontificia Universidad Católica de Chile. Main research lines include language development in early childhood, preschool education, and development of reading comprehension. Works occasionally as a consultant for the Ministry of Education in matters related to preschool education.
Daniela Vergara: Ms in Educational Psychology (Pontificia Universidad Católica de Chile), research coordinator Pontificia Universidad Católica de Chile.
M. Francisca del Río: PhD in Psychology (Pontificia Universidad Católica de Chile), Associate Professor at the Faculty of Education in Universidad Diego Portales in Santiago de Chile. Among her research interests are language and cognitive development in early childhood, social representations in preschool children, and the development of gender and social status stereotypes associated with academic achievement and behavior.
