Attention and Executive Functioning in Infancy: Links to Childhood Executive Function and Reading Achievement

Abstract

This study provides the first analyses connecting individual differences in infant attention to reading achievement through the development of executive functioning (EF) in infancy and early childhood. Five-month-old infants observed a video, and peak look duration and shift rate were video coded and assessed. At 10 months, as well as 3, 4, and 6 years, children completed age-appropriate EF tasks (A-not-B task, hand game, forward digit span, backwards digit span, and number Stroop). Children also completed a standardized reading assessment and a measure of verbal intelligence (IQ) at age 6. Path analyses on 157 participants showed that infant attention had a direct statistical predictive effect on EF at 10 months, with EF showing a continuous pattern of development from 10 months to 6 years. EF and verbal IQ at 6 years had a direct effect on reading achievement. Further, executive functioning at all time points mediated the relation between 5 month attention and reading achievement. These findings may inform reading interventions by suggesting earlier intervention time points and specific cognitive processes (i.e., 5 month attention).

Infancy is a period of development characterized by rapid physical, socio-emotional, and cognitive growth. Individual differences during infancy are often studied for their stability and their foundation for later development (Bornstein, 2014). One cognitive construct that is the focus of much research during infancy is attention because individual differences in early attention predict performance on later cognitive tasks, such as those focused on executive functioning (Colombo & Cheatham, 2006; Cuevas & Bell, 2014; Rose & Feldman, 1997; Rose, Feldman, & Jankowski, 2012). Executive functioning (EF) refers to a series of interrelated processes (i.e., inhibitory control, working memory, cognitive flexibility; Diamond, 2013) that are essential for cognitive control. As such, EF is directly associated with school readiness (e.g., Blair & Raver, 2015; Welsh, Nix, Blair, Bierman, & Nelson, 2010) and academic success (e.g., St. Clair-Thompson & Gathercole, 2006). Less is known, however, of the connection between infant attention and later reading achievement. Given that infant attention is linked with child EF, which in turn is linked with educational attainment, we examined a developmental cascade connecting infant attention to early reading achievement through childhood EF at multiple developmental time points.

Individual differences in infant looking behavior are often observed and discussed in terms of attentional processing. Development of the orienting network, a neural attentional network, may help explain the wide-ranging individual differences observed in looking behaviors (Posner & Rothbart, 2007). The orienting network allows for disengagement and shifting of attention (Frick et al., 1999). During infancy and early childhood, this attentional network provides support for the developing executive network, which is involved in attentional conflict resolution, a skill also linked with EF (Nakagawa & Sukigara, 2013; Rothbart, Sheese, Rueda, & Posner, 2011). Because of developments in executive attention, infants initially rely on orienting attention (i.e., exogenously driven attention) as a means of regulation (Harman, Rothbart, & Posner, 1997). After executive networks emerge, typically by the end of the first year, older infants and toddlers begin to exhibit a shift to endogenously driven regulation (Rueda et al., 2005), suggesting that brain systems underlying EF and attention work simultaneously and recursively to allow for successful self-regulation (Blair 2016; Blair & Raver, 2015; Johansson et al., 2015; Papageorgiou et al., 2014). This line of research suggests that the development of early attentional processes provides a foundation for later EF through the larger mechanism of self-regulation (Blair, 2016; Raver et al., 2012; Sheese et al., 2008). We examined attention, measured through look duration (i.e., sustained attention) and shifting (i.e., orientation attention) at 5 months to capture a developmental time when infants utilize the orienting attentional network prior to the maturation of the executive attention network (Posner & Rothbart, 2007).

EF undergoes significant development throughout early childhood (for review see Garon, Bryson, & Smith, 2008) and displays a linear developmental trajectory across early and middle childhood (e.g., 2–3 years: Carlson, Mandell, & Williams, 2004; 5–12 years: Polderman, Posthuma, De Sonneville, Verhulst, & Boomsma, 2007). EF development during infancy is relatively less known given the lack of a wide variety of age appropriate measures. What is known, however, is that the executive network and the prefrontal cortex are operative within the later half of the first year (Diamond, 1990, 2013), and infants are able to complete tasks designed to measure EF (Gottwald et al., 2016; Holmboe et al., 2018; Johansson, Marciszko, Brocki, & Bohlin, 2015; Johansson, Marciszko, Gredeback, Nystrom, & Bohlin, 2015). One task that has been argued to elicit EF-like abilities during infancy is the A-not-B task (Bell, & Adams, 1999; Diamond, 1985), a task originally designed by Piaget (1954) to assess lack of object permanence. The task requires infants to both maintain representations of information while objects are hidden (i.e., working memory), as well as resolve conflict when objects are moved to a new location (i.e., inhibitory control). Some infants as young as 8 months are successful at the A not B task (Bell, 2012; Bell & Adams, 1999; Bell & Fox, 1992; Diamond, 1985; Johansson, Forssman, & Bohlin, 2014), suggesting they may possess some rudimentary form of EF. We examined EF using the A-not-B task at 10 months, as infants at this age typically have executive attention abilities (Rothbart, Sheese, & Posner, 2007). We also collected EF data at 3 years, 4 years, and 6 years based on literature suggesting large improvement in EF from 3–4 years (e.g., Epsy, 1997) and 4–6 years (e.g., Rueda, Rothbart, McCandliss, Saccomanno, & Posner, 2005).

EF is one of many skills needed to excel in school (Altemeier, Abbott, & Berninger, 2008; Toll, Van der Ven, Kroesbergen, & Van Luit, 2011). Cascade effects of EF connect these higher order cognitive processes to multiple aspects of academic success (Rose, Feldman, & Jankowski, 2011). Clearly, inhibiting distracting stimuli and planning for future assignments are crucial for school success, but the developmental mechanisms associated with these skills are less clear.

Reading achievement, is critical for overall school success (Cantin, Gnaedinger, Gallaway, Hesson-McInnis, & Hund, 2016; Nolen, 2003). From the earliest school years, reading achievement displays wide ranging individual differences, and this variability has been proposed to reflect developments in attentional control and EF (e.g., McVay & Kane, 2012). Indeed, children with reading difficulties struggle with EF tasks compared to children with age-appropriate reading skills (Engel de Abreu et al., 2014). However, the mechanism linking attention, EF, and reading achievement is unclear. Given reports that infant attention lays the foundation for early self-regulation and thus early EF, we examined a potential mediating effect of EF, across early development, on the relation between infant attention and later reading achievement.

As noted, there are both conceptual and empirical links between infant attention and early EF (e.g., Colombo & Cheatham, 2006; Cuevas & Bell, 2014), as well as between EF and reading achievement (e.g., Arrington, Kulesz, Francis Fletcher, & Barnes, 2014; Baddeley, 1979; Siegel, 1994). What is missing from the literature is evidence within the same sample of these developmental linkages. We examined 5 month attention during infancy in relation to EF during infancy 10 month and early childhood at ages 3 years, 4 years, and 6 years, with an outcome of reading achievement at 6 years. Our primary hypotheses were: 1) EF throughout development will mediate the relation between infant attention at 5 months and reading achievement at 6 years (i.e., infant attention predicts early childhood EF, which predicts early reading). 2) EF performance will display continuity from 10 months to 6 years.

Method

Participants

A final sample of 157 children participated in the study. The children were part of an ongoing longitudinal study on cognition-emotion relations across early development. Initially, a total of 299 infants were recruited at 5 months of age (M = 5.42, SD = .23). The children represent two cohorts, or approximately 75%, of the participants of a larger longitudinal study examining cognition-emotion links across early development; the remaining 25% of the participants in the larger study represented a third cohort who did not have a research visit at age 6. The cohorts were recruited by two research locations. The Blacksburg, VA research location and the Greensboro, NC location each recruited half of the participants in the longitudinal study.

Of the original sample of 299 infants at 5 months, 268 returned and five additional infants were recruited at 10 months (M=10.46, SD= .38), 224 returned at 3 years (M = 3.08, SD = .08), 205 returned at 4 years (M = 4.39, SD = .10), and 194 returned at 6 years. Of the 194 participants seen at the age 6 visit (M = 6.56, SD = .38), 157 completed the Woodcock Johnson letter-word identification measure (i.e., reading achievement). The letter-word identification task was added into the protocol after a subset of children were seen (n = 37) due to the inability of some children to complete other more difficult reading measures (i.e., Woodcock Johnson reading fluency subscale).

Of the 157 participants (86 males) who provided reading achievement data for the analyses, 74% were Caucasian, 19% African American, and 7% were of other or mixed race. The mothers of the children came from diverse educational backgrounds (3% some high school, 10% high school graduate, 23% some college, 64% college graduate), as did the fathers (2% grade school, 9% some high school, 15% high school graduate, 29% some college, 45% college graduate). Performance on all tasks measured before age 6 did not differ between those participants with reading achievement data (n = 157) and those without (n = 37; all ts between .02-1.8, and all ps between .07-.98)

Procedure

Data were collected at both research locations using identical protocols. Research assistants from each location were trained together by the project’s Principal Investigator on protocol administration, as well as on behavioral data collection and coding. To ensure that identical protocol administration was maintained between the labs, the Blacksburg, VA site periodically viewed DVD recordings collected by the Greensboro, NC lab. To ensure that identical coding criteria were maintained between labs, the Blacksburg, VA lab provided reliability coding for behavioral data coded by the Greensboro, NC lab.

Upon arrival at the research lab for each visit, children and mothers were greeted by research assistants, study procedures were described, and signed consent was obtained from the mothers and verbal assent from the children beginning when they were 3 years of age. For each assessment, mothers were given an honorarium. Children began receiving a small gift during the early childhood lab visits.

Infants participated in a looking task at 5 months and the A-not-B task at 10 months, as well as other cognitive and self-regulatory tasks not reported here. The children returned at 3, 4, and 6 years of age to complete a battery of EF tasks, as well as other cognitive and self-regulatory tasks not reported here. At 6 years the participants also completed a measure of reading achievement and verbal IQ. For all tasks at all ages, interrater reliability was established, with at least 20% of our sample (for all tasks, Cronbach’s α ≥ .90).

5 month Attention

A brief video clip (45 s) from Sesame Street (Cecile – Up, Down, In, Out, Over, and Under) was presented while each infant was seated on their mother’s lap. Video cameras were placed to the side of the infant and above the monitor displaying the video clip. This allowed a close up view of both the video and the infant’s face during the display. A research assistant coded each infant’s looking behavior from the DVD of the laboratory session using Video Coding System software developed by the James Long Company (Caroga Lake, NY). Peak look duration (i.e., longest look at video) and shift rate (i.e., number of looks at video) were the variables of interest. The peak look duration variable was reverse scored based on literature suggesting that longer looking times are inversely related to attentional control and encoding speed (Colombo, 2001; Colombo Mitchell, Coldren, & Freeseman, 1991; Rose, Feldman, & Jankowski, 2005; Rose et al., 2012). Shift rate has similarly been linked to encoding speed (Rose et al., 2005, 2012).

EF Measures

A-not-B task (10 months).

Our A-not-B task was a looking version of the task used in previous studies with infants (e.g., Bell, 2001, 2012; Bell & Adams, 1999; Cuevas & Bell, 2010). Infants watched as an experimenter hid a toy in one of two locations (nonreversal trials). The gaze of the infants was then broken for a brief period and the infants were given the opportunity to look for the hidden object. If the infant responded correctly on two consecutive nonreversal trials, the location of the object was moved (reversal trials). Infants received a mean of 12 trials. The variable of interest was the proportion of correct responses across nonreversal and reversal trials.

Forward Digit Span (3 years, 4 years).

A forward digit span (FDS) task was administered to assess working memory at both the 3- and 4-year lab visits. Children were initially presented with two digits and instructed to repeat the sequence. Two practice trials were given to ensure understanding and then the task began. Attempt at recall of the same digit span with at least one correct trial for two trials was required before lengthening the span by one digit. The digit span was lengthened until errors were produced on two consecutive trials of the same span. The variable of interest was digit span, which accounts for nonconsecutive errors.

Hand Game (3 years, 4 years).

The hand game was used as the measure of inhibitory control (Luria, Pribham, & Homskaya, 1964). Children were shown two hand gestures (i.e., flat open hand or fist), and instructed to make the opposite gesture the experimenter made (e.g., flat open hand = fist). Practice trials were given to ensure understanding. Children were given a total of 16 trials, and the variable of interest was proportion correct.

Backwards Digit Span (6 years).

A backwards digit span (BDS) task was administered to assess working memory at 6-years. The task administration was identical to the FDS with one exception, the children were instructed to repeat the sequences backwards rather than forwards. The variable of interest was digit span.

Number Stroop (6 years).

A number-based computerized Stroop task was used to assess inhibitory control (Ruffman et al., 2001). The task had three conditions: letters, numbers, and mixed (both letter and numbers). Emphasis was placed on the mixed condition, which is considered to involve the most conflict. Children were instructed to count either letters (“AAA”=3) or number digits (“555” = 3) and to indicate their responses on the keyboard. Practice trials were provided. There were a total of 75 task trials (25 per condition). The variable of interest was proportion correct for mixed condition.

6-year Reading Achievement

A reading subscale (letter-word identification; test 1) from the Woodcock Johnson (WJ) III Test of Achievement was used to measure reading achievement (Woodcock et al., 2001). As noted above, we began our study with the reading fluency scale but some of the children were unable to complete this more difficult reading measure. Thus, after 35 children were seen in the lab, we changed to the letter-word identification in order to have data from each of the remaining children who had a 6-year lab visit. The variable of interest on the letter-word identification subscale was total correct. The WJ III subtests demonstrate reliabilities of .80 or higher (Nelson, Brenner, Lane, & Smith, 2004).

Covariate measures

The Peabody Picture Vocabulary Test IV (PPVT; Dunn & Dunn, 2012) was administered at age 6 as a proxy for verbal IQ. PPVT is a nationally standardized instrument, and the measure of interest was participants’ standardized scores. Demographic information regarding child age, sex, and race was measured through parent report.

Analytic Plan

SPSS software was used to examine the descriptive statistics of the variables of interest. A composite of infant attention was generated by standardizing and averaging two attention measures at 5 months. This composite included longest look duration (reversed scored) and shift rate, which were positively correlated (r = .67, p < .001). Composite EF scores at ages 3, 4, and 6 years were generated by standardizing and averaging the EF variables at each age. Composite scores were generated because composite variables are more reliable measures of EF than single assessments (Carslon et al., 2004). The zero-order correlations for the EF variables used in the composites were not statistically significant, yet theoretically represent two constructs that are included in EF (i.e., inhibitory control and working memory). At 3 and 4 years the tasks included FDS and hand game (r = −.08, p = .66; r = .16, p = .10, respectively) and at 6 years the tasks included BDS and number Stroop (r = .099, p = .23). Others have reported nonsignificant to low correlations among various EF inhibitory control and working memory tasks in early childhood (e.g., Wiebe et al., 2011). The relation between inhibitory control and working memory increases with age (Kim-Spoon, Deater-Deckard, Calkins, King-Cases, & Bell, 2019), so the lack of significant correlations may be a result of our focus on early childhood. For descriptive statistics see Table 1.

Table 1.

Descriptive Statistics

Measure	M	SD	Min	Max
5 months
Longest Look Duration	17.76	10.87	1.70	49.25
Shift Rate	8.66	4.49	1.00	26.00
Attention composite	−0.04	0.92	−2.35	2.58
10 months
A not B	0.58	0.17	0.18	0.94
3 years
Forward Digit Span	2.81	0.92	1.00	4.00
Hand Game	0.66	0.29	0	1.00
EF composite	0	0.96	−2.12	1.32
4 years
Forward Digit Span	3.74	0.76	2.00	6.50
Hand Game	0.82	0.17	0.25	1.00
EF composite	0.02	0.74	−2.07	1.83
6 years
Backwards Digit Span	2.98	0.84	1.00	4.50
Number Stroop	0.90	0.14	0.39	1.00
EF composite	−0.01	0.79	−2.44	1.30
Reading Achievement	32.80	9.39	14.00	63.00
Verbal IQ	110.89	12.48	82.00	139.00

Note: Final n = 157; M = Mean and SD = Standard Deviation. All values, other than composites, represent values before standardizing. Infant attention (i.e., longest look duration) was measured in seconds; shift rate was number of looks to video. A not B, Hand Game, and Number Stroop values represented proportion correct. Forward Digit Space and Backwards Digit Span represent span values (dependent on number correct). Reading achievement represents number correct on Woodcock Johnson assessment. Finally, Verbal IQ represents a standardized score.

Results

Q-Q plots were suggestive of multivariate normality, and there were no suggestions of multicollinearity in our model (VIF < 2). All EF composite measures were positively correlated with each other (see Table 2). Infant attention was positively correlated with EF at 10 months. EF was positively correlated with the EF measure at the next testing age (i.e., 10 months and 3 years; 3 years and 4 years; 4 years and 6 years). EF at 6 years was positively correlated with reading achievement at age 6. Finally, child sex and race were not correlated with reading achievement, therefore, these covariates were not included in the model.

Table 2.

Correlations

Task	1	2	3	4	5	6	7	8	9	10
1. 5 month Attention	--
2. 10 month EF	.193*	--
3. 3 year EF	.114	.219**	--
4. 4 year EF	−.025	.054	.272**	--
5. 6 year EF	−.096	−.027	.067	.385***	--
6. Reading Achievement	−.135	.057	.128	.273***	.328***	--
7. Verbal IQ	−.013	.104	.284**	.273***	.310***	.279***	--
8. Age	−.149	.103	−.064	−.114	.168*	.403***	−.020	--
9. Sex	.102	.143	.003	.037	.004	−.113	−.005	−.014	--
10. Race	−.071	−.133	−.104	.072	−.064	.111	.145	.057	−.062	--

Note:

^***p ≤. 001

^**p≤.01

^*p≤.05;

EF = executive functioning. Values represent correlation prior to imputation.

Path Model

We conducted longitudinal mediation analyses (see Figure 1) using structural equation modeling (SEM) in Mplus version 8 (Muthén & Muthén, 1998–2017), with Full Information Maximum Likelihood (FIML) estimation with robust standard errors (MLR) to account for missing data and non-normal distributions. For testing indirect effects, we calculated bias-corrected bootstrap confidence intervals (CIs) with maximum likelihood estimation (bootstrapping is not available for MLR) using 10,000 bootstrapping samples (Preacher & Hayes, 2008). Model fit was medicore, χ² (16, N=157) = 29.68, p = .020, RMSEA = .07, CFI= .85. The direct paths from EF at 10 months and EF 48 months to Verbal IQ (b = −0.47, SE = .72; b = 0.52, SE = .29; all ps > .05), and the covariation between age to infant attention (b = −5.52, SE = 20.85, p = .791) were not significant. We removed these nonsignificant paths to allow for a better-fitting, more parsimonious model. The fit of the trimmed model was acceptable, χ² (19, N = 157) = 31.80, p = .033, RMSEA = .07, CFI= .89. All regression paths in the revised model were significant (p < .05; see Table 3). The nested model comparison between the full model and the trimmed model (removing nonsignficant paths) using the Satorra-Bentler scaled chi-square statistic (Satorra & Bentler, 2001) indicated that the trimmed model was the preferred, more parsimonious model over the full model (Satorra-Bentler Δχ² = 5.93, Δdf = 3, p = .115).

Figure 1. Conceptual Path Model.

Note: EF represents executive functioning at varying ages; reading represents reading achievement; Verbal IQ represents Peabody Picture Vocabulary Test as a proxy for verbal IQ.

Table 3.

Direct and Indirect Standardized Values for Revised Model.

	Direct effect	Indirect effect
Dependent variable: 10 month EF
1. 5 month Attention	.186**	--
Dependent variable: 3 year EF
1. 5 month Attention	--	[0.001; 0.103]
2. 10 month EF	.188*	--
Dependent variable: 4year EF
1. 5 month Attention	--	[0.000; 0.030]
2. 10 month EF	--	[0.015; 0.605]
3. 3 year EF	.264**	--
Dependent variable: 6year EF
1. 5 month Attention	--	[0.000; 0.014]
2. 10 month EF	--	[0.008; 0.288]
3. 3 year EF	--	[0.021; 0.171]
4. 4 year EF	389***	--
Dependent variable: Reading Achievement
1. 5 month Attention	--	[.000; .039]
2. 10 month EF	--	[.018; .859]
3. 3 year EF	--	[0.036; 0.545]
4. 4 year EF	--	[0.264; 1.969]
5. 6 year EF	.196**	--
6. Verbal IQ	.229**	--
7. Age	.378***	--

Note: Final n = 157. Completely standardized solutions presented for direct effect. Bias-corrected bootstrap 95% confidence intervals presented for indirect effect.

^*p ≤ .05

^**p ≤ .01

^***p ≤ .001.

In terms of indirect paths for the revised model, infant attention at 5 months was positively related to EF at 10 months (b = 0.034, SE = .013, p = .010). EF at 10 months was positively related to EF at 3 years (b = 1.083, SE = .545, p = .047), which in turn was positively related to EF at 4 years (b = 0.206, SE = .081, p = .011). EF at 4 years was positively related to EF at 6 years (b = 0.414, SE = .090, p < .001). EF 6 years significantly predicted reading achievement at 6 years (b = 2.311, SE = .819, p = .005). Additionally, Verbal IQ at 6 years and child age were positively related to reading achievement at 6 years (b = 0.170, SE = .055, p = .003; b = 0.028, SE = .004, p < .001 respectively). The indirect effect of attention at 5 months on reading achievement at 6 years via EF at 10 months, EF at 3 years, EF at 4 years, and EF at 6 years was significant (bias-corrected bootstrap 95% CI [0.000; 0.039]). Finally, the indirect effect of EF at 10 months on EF at 6 years via EF at 3 and 4 years was significant (bias-corrected bootstrap 95% CI [0.018; 0.859]). See Table 3 for all direct, indirect, and total coefficient values.

Discussion

This is the first study to show a possible mechanism linking individual differences in infant attention to early childhood reading achievement. We provide evidence for a sustained developmental effect of infant attention, through visual attention at 5 months, on 6-year reading achievement via EF abilities throughout early childhood. Specifically, these results support our primary hypotheses that infant attention at 5 months has a direct effect on infant EF at 10 months; EF remains related from infancy through early childhood, and that EF across early development mediates the relation between infant attention and early childhood reading achievement.

The relation of 5 month attention to 10 month EF was expected, given the reliance on the orienting network (i.e., attention) for self-regulation during early development (Nakagawa & Suikigara, 2013; Rothbart et al., 2011). Further, the relation between infant attention and early EF abilities was also expected given previous findings (Cuevas & Bell, 2014). Finally, existing research suggests a connection between early childhood EF and early academic success (e.g., Blair & Razza, 2007; Liew, 2012). Our data bridges this previous work by demonstrating that the relation between 5 month attention and reading achievement is significantly mediated by EFs at 10 months, 3, 4, and 6 years.

Our results suggest that 5 month attention impacts early EF; EF displays continuity with age; and EF performance across development mediates the relation between infant attention and childhood reading achievement. Indeed, the critical role of infant cognition (5 month attention) on subsequent cognition was supported, providing evidence that infant cognition is the foundation for the development of later and more complex cognitive abilities (Colombo & Cheatham, 2006; Reynolds & Romano, 2016).

The continuity of EF was expected, given that previous research suggests EF performance is continuous throughout childhood (Carlson et al., 2004; Cuevas & Bell, 2014; Garon et al., 2008). Because of the established relation between EF and reading performance, the direct and indirect paths from EF to reading achievement were also expected (e.g., Altemeier, Abbott, & Berninger, 2008). Our study is unique, however, in that we established a pathway from infant attention, measured through behavioral measures, to infant and early childhood EF. Furthermore, we established a connection between EF at multiple developmental periods, including a 10 month infant measure, and reading achievement. Our results suggest that individual differences in attentional control in infancy are crucial for infant and early childhood EF performance, which is essential for later reading skills. Critically, these findings persisted after controlling for both verbal IQ and age, suggesting that these findings are not confounded by intelligence or general cognitive development. Indeed, this study is the first to combine these multiple facets of cognitive development to form a cohesive model linking infant attention and reading achievement at age 6.

There are some limitations to our work. First, although standardized assessments are highly informative, they do not provide a complete picture of reading ability. Future studies should include school performance assessments in order to gain more insight into reading achievement. Second, the reading achievement measure was specific to letter-word identification in order to collect data from each child in our study. Letter-word identification is foundational for reading achievement and targets a very specific pre-reading skill. Future studies, however, should utilize additional reading tasks, taking care when working with 6-year-old children to have tasks that can be accomplished by children in the study. Similarly, while the looking version of the A-not-B task has been argued to measure EF during infancy (Bell, 2001, 2012; Bell & Adams, 1999; Cuevas & Bell, 2010), arguments exist for other cognitive mechanisms (e.g., object permanence, memory representations, spatial understanding) driving A-not-B performance (e.g., Smith, Thelen, Titzer, & McLin, 1999). Future studies should incorporate multiple measures (e.g., hide and seek or simplified reverse categorization; Carlson, Mandell, & Williams, 2003; Garon, Smith, Bryson, 2014; Johansson, Marciszko, Brocki, & Bohlin, 2015) when examining infant EF. Third, EF was analyzed using composite measures, however, EF may be conceptualized as including multiple components (i.e., inhibitory control, set-shifting, working memory; Miyake et al., 2000). We used composites to ensure we were measuring similar constructs across development, as the components of EF may emerge throughout early development (Garon et al., 2008). Future studies, however, should examine how components of EF differentially mediate the relation between infant attention and reading achievement. Related, our EF measures were selected to analyze various and similar components of EF (inhibitory control and working memory) at various ages (hand game, number Stroop, and digit span tasks). The use of similar but varied measured was beneficial because it allowed us to analyze EF at various ages, but it may have added noise to our model. For example, our inhibitory control measure at ages 3–4 involved motor control (hand game), while our 6-year measure required less motor control (number Stroop). Inhibitory control that requires suppression of a motor response is known to be more difficult (e.g., Wessel, 2018). Further, the number Stroop task used digit stimuli, which may have inflated the relation between 6-year EF and reading achievement. Our reading achievement measure relied on letter and word identification, which is more similar to identifying digit amount than hand motions. Future studies should be careful to choose tasks that use similar stimuli across age groups. Fourth, the data were correlational; although the longitudinal design allowed us to estimate temporally lagged statistical predictive effects, caution is warranted with regard to making causal inferences. Finally, the video used to measure looking time included stimuli that is social in nature (i.e., a claymation ball with a moving mouth, but no eyes). This may have engaged attention in infants who might not have been engaged with less social stimuli (i.e., static nonsocial images; Colombo et al., 1991). Future studies should examine infant looking behaviors using various stimuli and attempt to replicate our results.

In summary, our study demonstrated that cognition in infancy generates a sustained developmental effect through early childhood EF, relating to later 6-year reading achievement. Further, our results support the continuous development of individual differences in EF throughout early development, and further support the well-known relation between early EF and reading achievement. These findings may inform interventions targeting attention and EF abilities by suggesting earlier intervention time points and cognitive processes (i.e., 5 month orienting attention). Further research is necessary to establish possible causal links.

Highlights.

• We examined a path model linking infant attention to reading achievement through executive functioning at multiple developmental time points.

• Executive functioning at 10 months, 3, 4, and 6 years mediated the relation between 5 month attention and 6 year reading achievement.

• Executive functioning was continuously related both directly and indirectly from age 10 months to 6 years.