About half of the participants returned the questionnaire that was sent by mail. However, the percentage of data present was approximately equally distributed among the groups (FR dyslexia: 28/50, 56 %; FR no-dyslexia: 42/82, 51 %, and controls: 32/64, 50 %)
1. Missing-value analyses showed that the subsamples with and without missing questionnaire data did not differ significantly on parental education,
t(189.9) = −0.57,
p = 0.572, word-reading fluency of the weakest-reading parent,
t(192.2) = −0.17,
p = 0.865, or word-reading fluency of the child,
t(193.0) = −1.52,
p = 0.131. Hence, further analyses of the questionnaire data were deemed appropriate. The preliteracy data were present for 96 % of the children. Grade 3 data were complete, as this was a requirement for inclusion into the study. One outlier on rapid naming in the control group was removed because the score was 3.4 standard deviations above the control group’s mean. Distributions were close to normal, unless stated otherwise.
Differences among the three outcome groups on continuous measures were evaluated using one-way ANOVAs (unless stated otherwise), followed by pairwise comparisons with Tukey’s correction for multiple testing. Group means, one-way ANOVA results, and effects sizes are presented in Tables
2,
3, and
4. Results are described below for group comparisons on family and child characteristics, followed by predictions of children’s reading and arithmetic skills.
Table 2
Parental literacy according to children’s literacy outcome
Print exposure |
Father | 7.81a
| 3.36 | 7.53a
| 3.08 | 9.81b
| 3.06 | 95 | 4.99 | (2, 92) | 0.009 | −0.09 | 0.65 | 0.75 |
Mother | 7.96a
| 2.79 | 8.10a
| 3.00 | 9.03a
| 2.87 | 98 | 1.24 | (2, 95) | 0.293 | 0.05 | 0.37 | 0.32 |
Dyslexic parent | 7.56a
| 2.98 | 7.08a
| 3.08 | | | 66 | < 1 | (1, 64) | 0.532 | −0.16 | | |
Non-dyslexic parent | 8.23a
| 3.15 | 8.56a
| 2.84 | | | 65 | < 1 | (1, 63) | 0.659 | 0.11 | | |
Literacy difficulties |
Father | 6.68a
| 1.52 | 6.20a
| 1.83 | 3.91b
| 0.93 | 97 | 29.91 | (2, 94) | <0.001 | −0.52 | −2.98 | −2.46 |
Mother | 6.00a
| 1.94 | 5.10a
| 2.07 | 3.84b
| 1.17 | 101 | 10.83 | (2, 98) | <0.001 | −0.77 | −1.85 | −1.08 |
Dyslexic parent | 7.56a
| 0.96 | 7.24a
| 1.11 | | | 66 | 1.39 | (1, 64) | 0.244 | −0.30 | | |
Non-dyslexic parent | 5.19a
| 1.57 | 4.02b
| 1.31 | | | 68 | 10.88 | (1, 66) | 0.002 | −1.11 | | |
Table 3
Preliteracy skills at 6 years and school achievement at 9 years according to literacy outcome
Preliteracy skills (6 years) |
Rapid naming colours | 0.58a
| 0.16 | 0.72b
| 0.18 | 0.75b
| 0.18 | 185 | 14.19 | (2, 182) | <0.001 | 0.79 | 0.96 | 0.17 |
Phonological awareness |
Blending | 7.57a
| 5.70 | 11.56b
| 6.25 | 14.45c
| 5.14 | 189 | 19.51 | (2, 186) | <0.001 | 0.78 | 1.34 | 0.56 |
Segmentation | 5.08a
| 4.97 | 9.46b
| 6.67 | 12.32c
| 6.09 | 189 | 19.49 | (2, 186) | <0.001 | 0.72 | 1.19 | 0.47 |
Letter knowledge |
Receptive | 13.61a
| 6.58 | 21.27b
| 6.59 | 22.73b
| 7.20 | 189 | 27.91 | (2, 186) | <0.001 | 1.06 | 1.27 | 0.20 |
Productive | 9.02a
| 6.28 | 17.66b
| 8.13 | 18.50b
| 8.22 | 188 | 24.89 | (2, 185) | <0.001 | 1.05 | 1.15 | 0.10 |
School achievement (9 years) |
Reading | 28.64a
| 8.21 | 54.72b
| 12.10 | 61.53c
| 12.24 | 196 | 129.98 | (2, 193) | <0.001 | 2.13 | 2.69 | 0.56 |
Arithmetic | −1.35a
| 1.61 | 0.15b
| 1.79 | 0.90c
| 1.75 | 196 | 24.22 | (2, 193) | <0.001 | 0.86 | 1.29 | 0.43 |
Table 4
β-weights and total R
2 of the multiple regressions predicting school achievement at 9 years from preliteracy skills at 6 years (left hand side) and pooled within-group correlations (right hand side)
Controlling variable |
Risk | −0.29*** | −0.24*** | −0.28*** | −0.08 | | | | |
Arithmetic | – | 0.25*** | – | – | – | | | |
Reading | – | – | – | 0.38*** | 0.45*** | – | | |
Preliteracy skills (6 years) |
Rapid naming colours | 0.24*** | 0.17** | 0.28*** | 0.19** | 0.40*** | 0.48*** | – | |
Phonological awareness | 0.09 | 0.10 | −0.03 | −0.06 | 0.28*** | 0.46*** | 0.38*** | – |
Letter knowledge | 0.37*** | 0.31*** | 0.26** | 0.12 | 0.39*** | 0.56*** | 0.45*** | 0.71*** |
Total R
2
| 0.51*** | 0.56*** | 0.27*** | 0.34*** | | | | |
Family Characteristics According to Risk and Literacy Status
Parental print exposure and literacy difficulties of the groups can be found in Table
2. The fathers of the control children spent significantly more time on reading and writing than those of the FR children, but maternal print exposure was not to be related to children’s group. These differential patterns could indicate a parent by group interaction. However, in a multivariate analyses (with data of both parents analysed simultaneously) this interaction was not significant,
F(2, 90) = 1.63,
p = 0.202, and there was only an effect of group,
F(2, 90) = 3.73,
p = 0.028.
To investigate our hypothesis regarding the effects of the dyslexic and the non-dyslexic parent in the FR sample, we subdivided parent couples according to their reading status (dyslexic vs. non-dyslexic
2). Parental print exposure was unrelated to reading outcome of FR children (see third and fourth line in Table
2). Non-dyslexic parents (
M = 8.43,
SD = 2.95) appeared to read and write approximately equally frequent as the parents of control children (
M = 9.42,
SD = 2.62),
t(94) = 1.59,
p = 0.115.
Concerning literacy difficulties (also in Table
2), parents in the control group reported fewer problems with reading and spelling, as was expected given the selection criteria. The only FR parents that reported similar levels of literacy as the control parents (
M = 3.88,
SD = 0.67) were the non-dyslexic parents of the non-dyslexic children (
M = 4.02,
SD = 1.31),
t(71) = 0.59,
p = 0.560. Interestingly, within the FR sample parental self-reported literacy difficulties (literacy difficulties, for short) seemed to differentiate children with and without dyslexia. This difference was large and significant (Cohen’s
d = −1.11,
p = 0.002) for the non-dyslexic parent. We conducted a follow-up 2 × 2 ANOVA on the four means and standard deviations in the bottom left corner of Table
2. The interaction between parental status (dyslexic vs. non-dyslexic) and child outcome (dyslexic vs. non-dyslexic) approached significance,
F(1, 63) = 3.85,
p = 0.054.
Regarding home literacy environment, the percentages of fathers/mothers subscribed to a magazine/newspaper were 68 %/68 % in the FR dyslexia group, 73 %/76 % in the FR no-dyslexia group, and 100 %/91 % in the control group. The differences were significant for fathers, suggesting more subscriptions in control families, χ
2(2, N = 101) = 11.86, p = 0.003, but not for mothers, χ
2(2, N = 101) = 4.82, p = 0.090. A table with detailed descriptive statistics on number of books in the home, shared reading, and cognitive stimulation can be obtained from the first author. Here we only present analytic results in the interest of space. The variable about number of books in the home was strongly skewed and therefore analysed with the nonparametric Kruskal-Wallis test. Overall, the difference between groups was significant, K(2, N = 102) = 7.01, p = 0.030, suggesting more books in the homes of control families (M = 4.62, SD = 0.75) compared to FR dyslexia (M = 4.04, SD = 1.14) and FR no-dyslexia (M = 4.07, SD = 1.11), but pairwise comparisons with adjusted p-values were not significant. The frequency of storybook reading was virtually the same over groups for fathers, F < 1, and mothers, F(2, 97) = 1.09, p = 0.340. The only difference was that mothers read more than fathers. Furthermore, parents provided similar levels of cognitive stimulation, F < 1.
Children’s Characteristics According to Risk and Literacy Status
Group means on precursors of reading at the end of kindergarten are shown in Table
3. All group effects on the preliteracy tasks were highly significant (
ps < 0.001). The FR dyslexia group was slower on rapid naming and knew fewer letters compared to the two non-dyslexic groups, which were statistically indistinguishable. However, the group means on phonological awareness showed a stepwise pattern, with the FR dyslexic children performing lowest, followed by the FR non-dyslexic, and thereafter by the control children. Group differences on preliteracy were not attributable to the group differences on parental education (Table
1), as ANCOVAs with parental education as covariate also yielded highly significant group effects (
ps < 0.001) and nonsignificant covariate effects (
ps < 0.288). Controlling for IQ differences (Table
1) in ANCOVAs showed significant effects of IQ (
ps < 0.001), but all group effects remained highly significant (
ps < 0.001).
As can be seen in Table
3, groups did not only differ on reading but also on arithmetic ability. When applying a similar criterion for dyscalculia (≤10, or Wechsler scale score ≤6.2, norm scores taken from Melis
2002) as for dyslexia, the percentages of children identified with dyscalculia were found to differ significantly: 42 % (21/50) in the FR dyslexic group, 20 % (16/82) in the FR non-dyslexic group, and 8 % (5/64) in the control group. These differences were confirmed by a chi-square test,
χ
2(2,
N = 196) = 19.79,
p < 0.001.
Prediction of Children’s Reading Skills
After confirming comorbidity, we examined in the full sample how much of the variance in school achievement can be explained by the preliteracy skills, and whether these skills are specifically related to reading or arithmetic. To ensure that we could collapse the FR and noFR samples, we checked whether the relations between preliteracy skills and school achievement were similar in the two samples. Therefore, we ran regressions with reading or arithmetic as dependent variable, entered one of the preliteracy skills and risk status (coded as 0 = noFR; 1 = FR) in the first step, and checked in the second step whether the interaction between the latter two explained additional variance. None of the interactions was significant (0.172 ≤ p’s. ≤ 923), indicating similar relations in the two samples. Therefore, it was sufficient to include only the main effect of risk in the regression analyses presented below.
The pooled within-group correlations between preliteracy and school skills and results of multiple regression analyses are presented in Table
4. Correlations among all variables were significant. The preliteracy skills were related to both arithmetic and reading skills, albeit seemingly more so with reading.
The first multiple regression showed that risk status and preliteracy skills together explained 51 % of the variance in reading; letter knowledge (akin to an autoregressor) made the strongest contribution. Phonological awareness did not explain variance in reading above that accounted for by rapid naming and letter knowledge. However, when letter knowledge (which correlated 0.71 with phonological awareness) was excluded, rapid naming (β = 0.32, t[185] = 5.36, p < 0.001) as well as phonological awareness (β = 0.33, t[185] = 5.41, p < 0.001) were significant.
When risk and arithmetic ability were added to the model (Table
4, second column), rapid naming and letter knowledge remained significant predictors, so they explained unique variance. This confirms the specific relation between preliteracy and literacy skills. With respect to arithmetic, risk and preliteracy skills explained 27 % of the variance. Rapid naming made a significant unique contribution to arithmetic after risk and reading were controlled.
Lastly, we examined the utility of parental literacy difficulties in predicting children’s reading skills. Again, we first checked the interactions between predictors and risk status in separate regressions, but they were not significant (0.172 ≤ p’s. ≤ 911). Hence, only risk status was accounted for in the regressions below.
Three hierarchical regression models were specified (see Table
5). In line with our research question regarding the non-dyslexic parents of the FR children, we again subdivided parents according to reading ability. The weakest-reading parent refers to the dyslexic parents of the FR children and the weakest of the parent couple of the noFR children (although still reading at least average).
3
Table 5
Hierarchical regression models predicting children’s reading at 9 years from parental literacy difficulties
1 (N = 98) | 1 | Risk | 0.11*** | −0.10 |
2 | Literacy of weakest-reading parent | 0.02 | −0.17 |
3 | Literacy of best-reading parent | 0.11*** | −0.34*** |
2 (N = 98) | 1 | Risk | 0.11*** | −0.10 |
2 | Parental education | 0.03 | 0.06 |
3 | Literacy of weakest-reading parent | 0.01 | −0.16 |
4 | Literacy of best-reading parent | 0.09** | −0.32** |
3 (N = 86) | 1 | Risk | 0.14*** | −0.04 |
2 | Parental education | 0.43*** | 0.15 |
Children’s preliteracy skills |
Rapid naming | | 0.28** |
Phonological awareness | | 0.14 |
Letter knowledge | | 0.34** |
3 | Literacy of weakest-reading parent | 0.01 | −0.19 |
4 | Literacy of best-reading parent | <0.01 | −0.06 |
Risk status was mainly based on the weakest-reading parent. For that reason we anticipated that knowledge of the self-reported literacy difficulties of the weakest parent would not add much over and above risk status. Therefore, in subsequent analyses literacy difficulties of the weakest-reading parent were entered before those of the best-reading parent. This allowed us to test whether literacy difficulties of the best-reading parent (which were not crucial in sample selection) are related to their offspring’s reading skills, controlling for literacy difficulties of the weakest-reading parent. It appeared that literacy difficulties of the weakest-reading parent indeed did not explain a significant amount of variance in children’s reading fluency, but differences in literacy difficulties of the best-reading parent explained an additional 11 % of the variance. Over and above risk, parental education, and literacy difficulties of the weakest-reading parent, the literacy difficulties of the non-dyslexic parent accounted for an additional 9 % in children’s reading fluency. In a final regression analysis, differences in risk, parental education, and children’s preliteracy skills together explained an impressive 57 % of the variance in children’s reading fluency, yet parental literacy difficulties did not significantly add to this prediction.
The regression analyses show the impact of the best-reading parent on children’s reading outcome. Within the FR sample this pertains to the effect of the non-dyslexic parent on their offspring’s risk for dyslexia. To illustrate this effect, we dichotomised the literacy-difficulty measure: ‘non-dyslexic’ parents scoring ≥6
4 were categorized as having literacy difficulty. It appeared that within the group of FR children with a non-dyslexic parent
without literacy difficulties 30 % (16/54) developed dyslexia, compared to 79 % (11/14) in the group of FR children with a ‘non-dyslexic’ parent
with literacy difficulties.