In sum, four systematic reviews with meta-analyses have estimated the strength of the association between home-based shared book reading and cognitive developmental outcomes in children up to the age of 9. The results suggest small-to-moderate significant associations between home-based shared book reading and spoken language abilities and emerging literacy skills. While these estimates are encouraging, they provide limited insight into the question of whether home-based shared book reading has an impact on a child's development before school for three reasons. First, each meta-analysis included data from children who had started school and hence had received compulsory, formal classroom intensive instruction in spoken language and literacy. These school-based developmental gains would obscure those related specifically to home-based shared book reading. Second, no review measured development outcomes beyond spoken language and emerging literacy skills, such as socio-emotional development. Thus, we do not know if the impact of home-based shared book reading goes beyond language and literacy. Third, we do not know if specific components of shared book reading are associated with specific components of child development as this has not been considered by previous reviews. Different specific measures of shared book reading such as duration and frequency may have varying impacts on a child's development (McElvany et al., 2009) therefore investigation into these would provide greater insight thus filling a gap in knowledge. For example, research has shown that the number of books in the home is associated with children's early literacy experiences (de Bondt et al., 2020) and a child's receptive vocabulary (Bracken and Fischel, 2008). However, because the number of books was incorporated into an overall composite outcome and not reported specifically, it is not possible to determine its individual relationship with a child's development but rather only the composite outcomes relationship with the child's development.

Therefore, the aim of this systematic review was to address these limitations whilst answering the question of whether home-based shared book reading has an impact on a child's development. To this end, we conducted a primary meta-analysis to estimate the strength of the relationship across all measures (composite and specific) of home-based shared book reading and all measures (composite and specific) of child development in children with a mean age < 4 years. In a series of secondary meta-analyses, we estimated the relationships between specific measures of home-based shared book reading and child development. Based on the available evidence, we predicted small-to-moderate significant relationships between shared book reading and spoken language and emerging literacy skills overall and more specific measures of shared book reading (e.g., frequency, duration, number of books in the home) and spoken language (e.g., language, vocabulary, receptive vocabulary, expressive vocabulary) and emerging literacy skills (e.g., concepts about print, letter names, rhyming). We made no predictions about the relationship between shared book reading and any other cognitive or socioemotional development outcomes due to the lack of relevant data.

Due to the high variability of the measurement and reporting of shared book reading, as noted above, this research cast a wide arc of inclusion for studies. It considered all types of studies including randomized controlled trials (RCTs), longitudinal studies, and cross-sectional studies. A study was included if it was written in English and included at least one composite or specific measure for home-based shared book reading and at least one cognitive or socioemotional developmental outcome. Each study included at least 10 child participants who were not yet attending any compulsory, formal schooling and who spoke English as a primary or additional language. Each group of children had to have a mean age of < 4 years. For example, if a group of children from a study had an age range from 3 to 6 years, no child in that group was attending any compulsory, formal schooling. These children were read to by a parent or caregiver at home.

A study was excluded if it assessed (1) outcomes after a child had started school, (2) reading at other locations (e.g., preschool, daycare center) and not the home, (3) storytelling rather than book reading, (4) shared reading with books without written words (5) multiple interventions (e.g., book reading plus television) and (6) interventions delivered by specialists on specific reading techniques (e.g., dialogic reading). In the one instance in which a study included both a disability cohort and a typically developing group, only the latter was included as the singular group would not have been a proportional representation with respect to the overall sample size (N = 47 studies) (Bills and Mills, 2022).

The search was conducted using comprehensive databases of relevant research disciplines: ERIC, APA Psycinfo, MEDLINE, EBSCOHost and Embase Ovid. All searches included Boolean operators, truncation, Medical Subject Headings (MeSH terms) and proximity operators and allowed the inclusion of studies from 1980 onwards. Keywords were used to represent family (parent, sibling, family, mother, father, brother or sister), location (home), activity (reading, literature, and picture book) and age (early child, infant, preschool). The search was conducted in November 2024. Reference lists of included articles were screened to determine if any critical studies had been missed by the search terms, and a further four studies were thus identified.

Screening of the studies involved a series of steps (see Figure 1 for the flowchart): search terms were agreed upon by the team of authors which were used by CG to search the databases. Titles and abstracts (not authors) of identified studies were uploaded by CG into SYRAS (https://syras.org/), where duplicates were removed. CG used SYRAS to include or exclude abstracts according to criteria, and then the team reconvened to discuss the validity of criteria. These steps were then repeated (twice) until the criteria were clearly valid and the final pool of abstracts identified. Full-text articles for included abstracts were downloaded by CG (N = 158), which were independently screened by pairs of authors (CG-GM, CG-AJ, CG-KK, CG-AS). CG identified any inconsistent decisions between authors, which were resolved via discussion. CG contacted the authors of studies that were missing relevant information (N = 55), and integrated relevant data for studies that fulfilled our criteria. Studies that did not meet criteria or did not provide requested data were excluded from further analysis (N = 52). Reference lists of included studies were screened by CG to identify if any potentially relevant studies had been missed (N = 4).

Data from included studies were extracted by CG and double-checked by an independent research assistant. No discrepancies were found.

Effect sizes and sample sizes of included studies were imported into SPSS, and a Fisher's Z transformation was used to calculate the standard error. Meta-analyses were performed in SPSS using random-effects weights based on the inverse-variance, including both within- and between-study variances. Restricted maximum likelihood was used, applying the iterative method and calculating the restricted maximum likelihood estimator. No standard error adjustment was made. A value of p < 0.05 was considered statistically significant in all relevant analyses. In line with the recommendations of Gignac and Szodorai (2016) and Funder and Ozer (2019), we interpreted r-values of 0.10, 0.20, and 0.30 as small, moderate and strong relationships, respectively. In line with recommendations from Cohen (1998), we interpreted Cohen's d values of 0.2, 0.5, and 0.8 as small, moderate and strong relationships, respectively.

Forest plots were generated in SPSS, centered on the mean effect size. Funnel plots were generated to assess the studies for publication bias and examination of Egger's test (Egger et al., 1997). The “trim and fill” method was used to estimate the number of missing studies and compute any artificial studies needed to determine an unbiased estimate of the mean effect size. I² was calculated as the measure of heterogeneity for each meta-analysis (Francis et al., 2019) to determine the proportion of variability explained by the differences between the studies included in each of the meta-analyses. Given the expected high heterogeneity based on the variety of study designs and outcome reporting, the random-effects model was chosen as this assumes that each study estimates a different underlying true effect (Tufanaru et al., 2015).

We did two types of sensitivity analyses for each meta-analysis. First, we repeated each meta-analysis with low-quality studies removed. Second, we repeated each meta-analysis with highly variable outcomes removed (i.e., SE > 0.15). We concluded that an original meta-analysis had adequate sensitivity if there was little change in the overall estimated mean (?).

Included studies were assessed for risk of bias using the Newcastle-Ottawa Scale (Wells et al., 2000), which is recommended by the Cochrane Collaboration for non RCT studies (Francis et al., 2019). Three categories were assessed: selection (maximum score of 4), comparability (maximum 2), and outcome (maximum 3). A total raw score of 0–3 was classified as poor quality, 4–6 as fair and 7–9 as good (see Table 2).

An adaptation of the Grading of Recommendations Assessment, Development and Evaluation (GRADE) criteria was applied to the studies included in the meta-analyses to determine a quality-of-evidence rating. The quality of evidence for each outcome was based on: (1) median GRADE values: I² (0–50 high quality, 51–74 moderate quality, ≥75 low quality; Francis et al., 2019); (2) the number of studies included in the meta-analysis (≥10 high quality, 5–9 moderate quality, < 5 low quality); (3) the total sample size of the meta-analysis (N ≥ 1,000 high quality, N = 51–999 moderate quality, N ≤ 50 low quality); (4) the risk-of-bias rating (majority good—high quality, majority fair—moderate quality, majority low—low quality); and (5) the publication bias based on the Eggers test (>0.05 high quality, ≤ 0.05 low quality; see Table 3).

Guided by the criteria provided by Glenton et al. (2010), we used predefined wording to express the overall quality of evidence for each outcome. Specifically, if an outcome was supported by high quality of evidence, we concluded that there was “a relationship” between home-based shared book reading and a developmental outcome. If an outcome was supported by moderate quality evidence, then we concluded that there was a “a probable relationship.” And if an outcome was supported by low quality evidence, then there was an “uncertain relationship.”

A total of 10,605 studies were identified. Having removed 3,240 duplicates, we screened the titles and abstracts of 9,365 articles against the inclusion criteria. A total of 9,202 were identified as irrelevant and were excluded, leaving 163. We excluded a further 121 articles after accessing the full manuscript (see Figure 1), leaving 42 articles for inclusion. The reference lists of these 42 articles were screened, where a further four articles met the inclusion criteria and were therefore included in the research. A total of 46 articles met the inclusion criteria for the systematic review (see Figure 1).

Of those 46 studies, 28 provided correlation coefficients between measures of home-based shared book reading and developmental outcomes and were included in the meta-analyses (see Table 1). Of the 18 remaining studies, 17 reported effect sizes that could not be converted into correlation coefficients (Note: provided means or beta statistics without standard deviations; see Supplementary Table 2). An additional study by Mendelsohn et al. (2020) did report Cohen's d, but we could not convert this into a correlation coefficient with confidence due to the dichotomisation of a continuous variable to create the treatment and control groups (see Supplementary Table 2 for the findings of the 18 studies that could not be included in the meta-analyses).

It is noteworthy that only one included study recruited participants with a disability (hearing loss). We used data from the typically developing control group to represent this study. Thus, the selection procedure identified 28 studies that provided relevant data and could be included in the meta-analyses.

For the N = 28 studies the total sample size was 24,859 (mean sample size per study of 887). The mean age of the children was 30.2 months (Range 10–79 months) and the proportion of people who identified as female and male were 52% and 48% respectively. The mean percentage of participants reported as Caucasian or white was 45%, African-American or Black 22%, Asian 16%, and Other 18%.

Five home-based shared book reading measures were reported: home-based shared book reading (composite measure; 47%), frequency (specific measure; 38%), duration (specific; 9%), number of books in the home (specific; 3%) and age of onset (specific; 3%).

Six measures of child development were reported: composite measure (48%), language (composite; 11%), vocabulary (composite; 10%), receptive vocabulary (specific; 12%), expressive vocabulary (specific; 12%), cognitive development (composite; 6%) and recognizing letters (1%).

If a study measured home-based shared book reading using both composite and specific scores, we selected the former. In addition, if a study measured multiple developmental outcomes, we calculated the relationship between home-based shared book reading and each of these outcomes. This produced N = 45 estimates across the 28 included studies (see Table 1).

In total, there were 18 combinations of home-based shared book reading measures (N = 5) and developmental measures (N = 6). However, only six combinations met our criteria for statistical power (i.e., at least five studies and 783 participants; see Table 4 for excluded combinations):

When assessing bias on the N = 28 studies included in the meta-analyses all studies were either longitudinal or cross-sectional with the majority assessed as fair quality (21/28, 75%), four as low quality (18%) and the remainder as good (7%; see Table 3).

The following results are presented in two sections. Firstly, Section 1 presents the results of the meta-analyses relating to home-based shared book reading when measured as either a composite and specific outcome and the associations with developmental outcomes. Secondly, Section 2 presents the results of the meta-analyses relating to the specific measure of the frequency of home-based shared book reading and the associations with developmental outcomes.

We found 28 studies (N = 24,859 participants) that reported the strength of the relationship between home-based shared book reading and at least one developmental outcome. These studies used a variety of composite and specific measures of shared book reading (frequency, duration, age of the child when the caregiver commenced reading, number of books in the home) and developmental outcomes (language, vocabulary, receptive vocabulary, expressive vocabulary, cognitive development). In our first meta-analysis, we found a strong relationship between shared book reading and developmental outcomes overall. This average was very similar to estimates produced by meta-analyses of the mean correlation coefficients between any developmental outcome measure (composite or specific) and a composite measure of home-based shared book reading or the specific measure of the frequency of home-based shared book reading. These results show that home-based shared book reading and early child development are related regardless of frequency, duration, or other measures such as number of books in the home. At first glance, this suggests that future research could opt to use any measure of home-based shared book reading. However, this conclusion would be premature for reasons outlined in the limitations and future directions section.

It was interesting to learn from this systematic review that most studies (24 of the 28) that investigated the relationship between home-based shared book reading and child development have focused on spoken language. Spoken language was measured in numerous ways across the included studies. Our analyses revealed a strong relationship between home-based shared book reading and spoken language, which exceeded our prediction of a small-to-moderate relationship. It also exceeded the estimate calculated by Barone et al. (2019) who found a small, yet significant, relationship between home-based shared book reading and a composite measure of language [d = 0.22, 95% CI = (0.13, 0.31); N = 3,556 participants] in children aged 0–6 years. We also found, for the first time, a strong relationship between home-based shared book reading and vocabulary development. This finding is encouraging given research has shown that vocabulary at age 3 years is a major determinant of school readiness (Camp et al., 2010). This coupled with research showing that children who enter school with a larger vocabulary are more likely to become successful readers in school (Noble et al., 2019) and that language abilities have been shown to be at the core for school success (DeBaryshe, 1995) suggests that shared book reading with children in the early years is important.

Finally, we found a moderately strong relationship between the frequency of home-based shared book reading and expressive vocabulary. This finding appears stronger than the review by Barone et al. (2019) who found a small, yet significant relationship between home-based shared book reading and expressive language [d = 0.21, 95% CI = (0.10, 0.29)]. Although this was similar to the review by Dowdall et al. (2020) who found a moderate, significant relationship between home-based shared book reading and expressive language [N = 2,594 parent-child dyads; d = 0.41, CI = (0.20, 0.61)] in children aged from 1 to 6 years. It is noted that Barone et al. (2019) and Dowdall et al. (2020) included an array of measures for shared book reading and did not specifically assess frequency which may account for the variation in findings.

The positive association between the frequency of shared book reading and expressive vocabulary is important as previous research has shown that expressive vocabulary, i.e., the words a child can produce (Turnbull et al., 2022), is associated with pre-reading skills (Wise et al., 2007), which are predictive of both word reading (Liu et al., 2017) and a child's socioemotional skills (Wirth et al., 2020).

In sum, the results of this systematic review suggest that home-based book reading is probably related to childhood developmental outcomes, related to spoken language in general, and vocabulary most specifically, in children with a mean age < 4 years who had not yet started compulsory, formal schooling. When considering the strength of this suggestion, it is important to consider five methodological limitations of this review and the studies that it was able, and unable, to include.

The first limitation was the 12 meta-analyses that we could not conduct due to the lack of studies and/or participants (see meta-analyses section). For example, this systematic review was not in a position to quantitatively evaluate the relationship between home-based shared book reading and cognitive development. Two studies reported home-based shared book reading using a composite outcome, which demonstrated small (r = 0.17; Rodriguez et al., 2009) to moderate (r = 0.29; Tomopoulos et al., 2006) relationships. Another three studies used a specific measure (i.e., frequency of home-based shared book reading) which showed moderate (r = 0.259, r = 0.24: Dexter and Stacks, 2014; Tomopoulos et al., 2006,) or large (r = 0.64; Li et al., 2021) relationships. These findings suggest that home-based shared book reading may be positively associated with cognitive development, although it is not possible to draw definitive conclusions until there is sufficient data to undertake a meta-analysis.

Another meta-analysis that could not be conducted aimed to look at the relationship between the specific measure of duration of home-based shared book reading and developmental outcomes. Three studies reported small (r = 0.16; Farrant and Zubrick, 2011) to moderate positive relationships (r = 0.23; Schlesinger et al., 2019, r = 0.204; Shen and Del Tufo, 2022) which provides further evidence of the importance of investigating the specific measures of home-based shared book reading and not only composite outcomes. It would be particularly beneficial if future research could conduct well-powered and high-quality studies investigating the relationship between composite and specific measures of home-based shared book reading and composite and specific measures of developmental outcomes.

However, when evaluating the findings from the meta-analyses that could be undertaken (N = 6), the composite measures of home-based shared book reading tended to produce similar relationships with child development outcomes to the specific frequency measure. On the one hand, this may recommend the use of a composite measure over frequency as a more sensitive measure which aligns with previous shared book reading research (Payne et al., 1994). Given that composite measures index multiple aspects of shared book reading at the same time, this may well be true. However, this sensitivity comes at the cost of specificity because it is not possible to determine which aspects of shared book reading may be contributing most to the outcomes. This conundrum suggests that future research choose measures of shared book reading that best address the aims of each study. If the choice is not clear, we suggest a parsimonious approach whereby a composite measure is used as a more sensitive “screener” of shared book reading, and more specific measures of frequency, number of books and duration are used to understand if some aspects of shared book reading are more important than others. It is noted that although Payne et al. (1994) recommends the use of composite outcomes it is important that they are aggregated appropriately. The conceptualization of applying a composite outcome to measures of shared book reading generally assumes that each specific component contributes equally to the overall value (Burgess et al., 2002). We recommend further statistical investigation on large sample sizes to determine which specific components contribute to child development and how they could be combined to produce a reliable composite outcome.

Another limitation for this research was our inability to test our prediction of a moderate and significant relationship between shared book reading and emerging literacy skills. Only one study investigated the relationship between home-based shared book reading and emerging literacy skills in children with a mean age < 4 years, which points to a significant gap in our knowledge that needs filling. Given previous research has shown that shared book reading is an important contributor to a child's emerging literacy skills (Celano et al., 1998; Mascarenhas et al., 2017; Sinclair et al., 2018) we recommend research in this area. Further to this, no studies were identified that investigated socioemotional skills. Therefore, we were unable to undertake any meta-analyses on this area suggesting another significant gap in knowledge that also needs filling, especially as research has shown that a child's early language is an important predictor of their socioemotional skills (Wirth et al., 2020).

The third limitation was that although the inclusion criteria for the 6 meta-analyses conducted in this research was met (i.e., at least 5 studies and 783 participants), and the sensitivity analyses did not change the overall estimated mean, the variability in sample sizes between studies was notably high (composite: N = 44–6,050; frequency: N = 28–4,109). Given the importance of statistical power, we suggest that future studies that investigate shared book reading use an adequate sample size to avoid adding further noise to the data.

A fourth limitation of this review was our inability to determine if children were receiving any formal pre-school education during a study. The majority of studies (15/28; 54%) did not report on preschool attendance. The remaining 13 studies reported that (1) no children were attending any formal education (four studies; 14%); (2) some children may have been attending preschool (two studies; 7%) or early intervention programs, community programs or preschool (seven studies; 25%). We recommend that future research on shared book reading include information about whether participating children are attending preschools that provide instruction in reading-related skills.

The final limitation of this research was the high heterogeneity of the results in the meta-analyses which raises concerns around the conclusion of the relationships. One could assume that the largest studies would be producing the largest effect, however this was not always the case as can be seen when a composite measure was used for shared book reading and the relationship with vocabulary. Providing a precise estimate is dependent on the interaction between the nature of the measure being used for the outcome and the number of participants—the more precise the outcome measure, the more precise the relationship.

Notably, although this review did not consider demographic factors such as socioeconomic status (SES) in the meta-analyses, we investigated this variable across the 28 manuscripts included. The majority of studies (15/28; 54%) either did not mention SES or recognized it as a variable of interest but did not consider it within the analysis. The remaining 13 studies (46%) did include SES as a variable of interest with the majority (10/13; 77%) demonstrating an association with the home literacy environment or child developmental outcomes. Given this, it would suggest that SES is an important factor in this field and recommend that future research on shared book reading include information about the socio-economic status of participating caregivers both in reporting and analysis.

We recommend the following steps for future research into shared book reading: (a) large sample sizes, (b) apply the clear outline of how to measure and report shared book reading as mentioned previously, (c) small age ranges with schooling specified if applicable as well as socioeconomic status and (d) precise developmental outcome measures.