Individual differences in neural mechanisms of selective auditory attention in preschoolers from lower socioeconomic status backgrounds: an event-related potentials study
Abstract
Selective attention, the ability to enhance the processing of particular input while suppressing the information from other concurrent sources, has been postulated to be a foundational skill for learning and academic achievement. The neural mechanisms of this foundational ability are both vulnerable and enhanceable in children from lower socioeconomic status (SES) families. In the current study, we assessed individual differences in neural mechanisms of this malleable brain function in children from lower SES families. Specifically, we investigated the extent to which individual differences in neural mechanisms of selective auditory attention accounted for variability in nonverbal cognitive abilities in lower SES preschoolers. We recorded event-related potentials (ERPs) during a dichotic listening task and administered nonverbal IQ tasks to 124 lower SES children (77 females) between the ages of 40 and 67 months. The attention effect, i.e., the difference in ERP mean amplitudes elicited by identical probes embedded in stories when attended versus unattended, was significantly correlated with nonverbal IQ scores. Larger, more positive attention effects over the anterior and central electrode locations were associated with superior nonverbal IQ performance. Our findings provide initial evidence for prominent individual differences in neural indices of selective attention in lower SES children. Furthermore, our results indicate a noteworthy relationship between neural mechanisms of selective attention and nonverbal IQ performance in lower SES preschoolers. These findings provide the basis for future research to identify the factors that contribute to such individual differences in neural mechanisms of selective attention.
Research highlights
- We assessed the neural indices of selective auditory attention by measuring the differences in ERP mean amplitudes elicited by identical probes embedded in stories when attended versus unattended.
- There are prominent individual differences in neural indices of selective auditory attention in preschoolers from lower SES backgrounds.
- Neural indices of selective auditory attention are significantly correlated with nonverbal IQ scores in preschoolers from lower SES families.
- Larger, more positive attention effects over the anterior and central electrode locations are associated with better nonverbal IQ performance.
- Our results indicate a noteworthy relationship between neural mechanisms of selective attention and the development of nonverbal IQ in preschoolers from lower SES families.
Introduction
There is overwhelming evidence for an educational achievement gap between children from lower and higher socioeconomic status (SES) families (Hout & Janus, 2011; Reardon, 2011). Regrettably, the SES achievement gap has widened markedly over the last three decades (Reardon, 2013). Precursors of this gap can be detected as early as kindergarten, with children from lower SES families performing worse on standardized tests of school readiness (Duncan & Magnuson, 2011). The existence of SES disparities at such an early stage in education underscores the importance of characterizing brain functions that are associated with fundamental cognitive abilities during preschool years.
SES disparities have been documented for several aspects of brain development (Hackman & Farah, 2009; Noble, Houston, Kan & Sowell, 2012; Raizada & Kishiyama, 2010). In this study, we focused on neural mechanisms of one specific cognitive function, selective attention. This fundamental ability is posited as a critical component of foundations for education (Stevens & Bavelier, 2012). Neural mechanisms of selective attention are particularly vulnerable in children from lower SES families (Stevens, Lauinger & Neville, 2009; Stevens, Paulsen, Yasen & Neville, 2015). However, despite such heightened vulnerability, these neural mechanisms can also be enhanced via targeted interventions (Neville, Stevens, Pakulak, Bell, Fanning et al., 2013). Here we assessed individual differences in this malleable brain function in children from lower SES families. Specifically, we investigated the extent to which such individual differences in neural mechanisms of selective attention would account for nonverbal cognitive abilities in lower SES preschoolers, drawing parallels from previous research that linked various aspects of attention to fundamental nonverbal cognitive skills, as discussed below.
Attention as a critical element of nonverbal cognitive abilities
Engle and Kane (2004) argue that the domain-general ability to control attention, especially in the presence of internal and external distractors, is a critical element of higher-order cognitive abilities, such as language comprehension and fluid reasoning. In accordance with this proposition, studies of adults have provided ample evidence linking various aspects of attention to nonverbal cognitive abilities. For instance, behavioral and event-related brain potential (ERP) measures of attentional capture were linked to visual working memory performance, as high capacity individuals showed stronger resistance to and faster recovery from capture of attention (Fukuda & Vogel, 2009, 2011). Furthermore, in a study that used steady-state visual evoked potentials (SSVEPs), superior attentional control, specifically early suppression of irrelevant information, was associated with higher visual working memory capacity (Gulbinaite, Johnson, de Jong, Morey & van Rijn, 2014). Similarly, in a recent ERP study, neural indices of auditory selective attention were linked to individual differences in visual working memory capacity in adults (Giuliano, Karns, Neville & Hillyard, 2014). In line with these findings, adults who performed better on attentional control tasks were found to have higher fluid intelligence scores (Unsworth, Fukuda, Awh & Vogel, 2014; Unsworth & Spillers, 2010). Moreover, individual differences in self-reports of pre-trial attentional state were linked to individual differences in general fluid intelligence (Unsworth & McMillan, 2014). Participants who reported higher focus before trials, and fewer fluctuations in attentional state during the tasks, demonstrated better fluid intelligence performance. These studies provide converging evidence for links between various aspects of attention and cognitive abilities in adults.
Similar relations between attention skills and nonverbal cognitive abilities have also been reported in studies of children. For example, the ability to sustain attention for targets in the presence of distractor items was linked to both parent ratings and laboratory measures of inhibitory control in 3- to 6-year-old children (Reck & Hund, 2011). Moreover, individual differences in the ability to orient attention were linked to variability in visual short-term memory and visuospatial working memory in children (Astle, Nobre & Scerif, 2010; Shimi, Nobre, Astle & Scerif, 2014b). Children who were better at using spatial attention cues presented during maintenance of items in visual short-term memory had higher visual short-term memory scores and larger visuospatial working memory spans. Furthermore, in a recent ERP study, individual differences in the neural markers of attentional orienting were linked to visual short-term memory capacity in 10-year-old children (Shimi, Kuo, Astle, Nobre & Scerif, 2014a). Children who were more similar to adults in their neural responses elicited by spatial cues, which were presented prior to arrays of items to be remembered, had higher visual short-term memory capacity. Together these findings provide evidence for the pivotal role of attentional skills in nonverbal cognitive abilities in children.
In accordance with the considerable evidence linking various aspects of attention to numerous nonverbal cognitive abilities in adults and children, we posited that selective attention would account for variability in nonverbal cognitive performance in preschoolers from lower SES families. To test this hypothesis, we assessed neural markers of selective attention, a brain function that is both vulnerable and enhanceable in lower SES children (Neville et al., 2013; Stevens et al., 2009; Stevens et al., 2015).
Neural indices of auditory selective attention in lower SES children
Selective attention refers to the ability to enhance the processing of particular input while suppressing the information from other concurrent sources (Desimone & Duncan, 1995; Hillyard, Hink, Schwent & Picton, 1973; Serences & Kastner, 2014; Yantis, 2008). Via an ecologically valid and child-friendly dichotic listening paradigm, the typical neural indices of selective auditory attention have been characterized in children from various SES backgrounds (Coch, Sanders & Neville, 2005; Sanders, Stevens, Coch & Neville, 2006; Stevens et al., 2009). In this paradigm, children are instructed to attend selectively to one of two simultaneously presented stories. Event-related potentials (ERPs) are recorded to the same probe stimuli superimposed on both the attended and unattended, or ignored, stories. Neural indices of selective attention are measured by comparing the mean amplitudes of ERPs elicited by the identical probes in the attended and unattended stories. Earlier studies that used this paradigm reported a significant effect of selective attention on ERPs as early as 100 to 200 ms in typically developing children from higher SES families (Coch et al., 2005; Sanders et al., 2006; Stevens et al., 2009). This attention effect was characterized as larger, more positive ERP mean amplitudes for probes embedded in the attended stories versus probes in the unattended stories.
Using this paradigm, Stevens and colleagues (2009) investigated SES-related disparities in neural indices of auditory selective attention in children. In line with previous studies that used this paradigm (Coch et al., 2005; Sanders et al., 2006), a significant effect of attention on ERPs was documented in both higher and lower SES children. However, when the groups were compared, the magnitude of the attention effect was significantly reduced in the lower SES group compared to the higher SES group. These results suggested that neural mechanisms of selective attention were vulnerable in lower SES children.
Present study
In the current study, we proposed that, despite their aggravated vulnerability in children from lower SES families, neural mechanisms of selective attention would still account for variability in nonverbal cognitive abilities among this at-risk population. To test this proposal, we employed an individual differences approach to evaluate neural indices of selective auditory attention. To measure selective attention, we recorded ERPs in the child-friendly dichotic listening task described above. This task allowed us to focus on neural mechanisms of selective attention without overt response demands.
The associations between neural mechanisms of attention and nonverbal cognitive abilities have been mainly reported in studies of visual attention (Fukuda & Vogel, 2011; Gulbinaite et al., 2014; Shimi et al., 2014a). However, we employed an auditory selective attention paradigm because the ERP indices of this paradigm have been well characterized in young children (Coch et al., 2005; Sanders et al., 2006; Stevens, Sanders & Neville, 2006). Furthermore, in a recent study that used this ERP paradigm with university students, the magnitude of the early auditory selective attention effect (the P1 component) was associated with individual differences in visual working memory; adults who had a larger, more positive P1 also had higher visual working memory capacity (Giuliano et al., 2014).
Using this well-established paradigm, we assessed individual differences in selective attention in young children from lower SES families. Based on the previous research that documented links between various measures of attention and tests of nonverbal cognition, we expected a notable association between neural markers of selective attention and performance on nonverbal tests of cognition in lower SES children. Specifically, we anticipated that a larger (i.e., more positive in mean amplitude) ERP attention effect would predict better nonverbal cognitive abilities, as measured by tasks of nonverbal intelligence.
Method
Participants
Participants were 124 children (77 females) between the ages of 40 and 67 months (Mean = 54 months, SD = 6.5 months). They were recruited from 12 Head Start (HS) preschool sites in Oregon, a program for families living at or below the poverty line. Based on parent report, children with diagnosed behavioral or neurological problems (e.g., ADHD, specific language impairment, epilepsy) and children taking psychoactive medications were excluded from the present study. All children included in the ERP analyses were right-handed, monolingual, native English speakers who passed a hearing screening at 20 dB HL at 500, 1000, 2000 and 4000 Hz in both the right and left ears. From a total of 158 children who met these criteria, 23 were excluded due to low ERP data quality (excessive EEG artifacts and/or less than 75 trials per condition), and 11 were excluded for having less than 50% accuracy on the comprehension questions presented during the ERP task. In the final sample, 58% of the children were White/Caucasian, 1% Black/African American, 4% American Indian, 16% more than one ethnicity, and 21% unreported.
Informed consent was obtained from parents or other caregivers. In addition, verbal assent was obtained from child participants. Behavioral measures and ERPs were collected in two different sessions, separated by no more than 30 days. All families were paid for participation. Study procedures were approved by the University of Oregon Institutional Review Board.
Socioeconomic status (SES)
Parents/caregivers filled out a questionnaire which gathered information on the education level and profession of the primary caregivers. Socioeconomic status (SES) of the child was coded by trained research assistants according to the Hollingshead Four Factor Index of Social Status (Hollingshead, 1975). SES questionnaire information was incomplete or missing for 13 children. However, since all participants were recruited from a Head Start program, which requires at least 90% of the enrolled children to be from low-income families, and we used HS enrollment as a proxy for lower SES, we did not exclude these children from our final sample. For the children for whom we had complete SES questionnaire information (n = 111), the mean SES was 29.80 (SD = 11.43) according to the Hollingshead index (range = 8–66). There was one SES outlier (3 SD above the mean), therefore all the analysis that included the SES variable were conducted with and without this child. Exclusion of this child did not change the direction or strength of the results. Since this child was originally selected for our study based on HS enrollment, which we used as the proxy for lower SES, we included this child in all analyses.
Nonverbal intelligence
As a part of a battery of behavioral measures, children completed three subtests of the Stanford-Binet 5th Edition (SB-5) nonverbal IQ scale Roid (2003). The subtests included were Fluid Reasoning, Quantitative Reasoning, and Working Memory. The nonverbal Fluid Reasoning subtest measured the child's ability to identify sequences of pictured objects and complete matrices of figures and geometric patterns. The nonverbal Quantitative Reasoning subtest focused on mathematical reasoning. The lower levels of this subtest, which were designed for younger children, measured basic concepts (e.g., bigger/smaller), counting, addition using objects and pictures, and recognition of numbers. The nonverbal Working Memory subtest consisted of a Delay Response activity, which used a memory paradigm of hiding objects under cups, and the Block Span activity, which required children to recall a sequence of block taps. The range of possible raw scores on each subtest was between 0 and 19. A composite nonverbal IQ score was obtained by averaging the raw scores from the three subtests.
Electrophysiological assessment of selective auditory attention
Stimuli
Four narrative stories from the Blue Kangaroo series (Clark, 1999, 2001a, 2001b, 2002), four from the Harry the Dog series (Zion & Graham, 1956, 1958, 1960, 1965), four from Max and Ruby series (Wells, 1991, 1997, 2000, 2002) and four from the Classic Munch series (Munsch & Martchenko, 1988, 1989, 1992; Munsch & Petricic, 2004) were digitally recorded (16 bit, 22 kHz) using an Electro Voice 1750 microphone connected to a Macintosh computer running a sound-editing program (SOUNDEDIT 16, Version2). Either a male or a female narrator read the stories at a normal speaking rate in a child-directed manner. Pauses were edited such that they did not exceed 1 s in order to lessen the opportunity to switch attention to the other channel and to equate the length of pairs of stories. The average amplitude of each story was equated and high amplitude noises created by bursts of airflow were deleted. Following editing, 16 stereo files (four stories × four series) were created. The stereo files differed in location (left/right speaker) and narration voice (male/female). Each file was 2.5–3.5 min in length. The stories were presented at an average of 60 dB SPL (A-weighted).
Two probe stimuli were created by digitizing a token of the syllable ba spoken by a female voice (different from the female narrators) and scrambling the order of 4–6 ms segments of that token to create a nonlinguistic sound with similar acoustic characteristics. The two probes were 100 ms in length and were presented at 70 dB SPL. Across the stories, an equal number (N~180–206) of linguistic and non-linguistic probes were presented in each channel. The probes were presented in a pseudo-random order at an interstimulus interval (ISI) of either 200, 500, or 1000 ms in one of the two channels. Probes were never presented simultaneously in the attended and ignored channels.
Color pictures from the 16 stories were scanned and edited such that the image presented on the computer monitor directly in front of the participant subtended a visual angle of no more than 5o in the vertical or horizontal directions. Fifteen to 20 images were selected from the attended story and presented for 5–15 s at points relevant to the content of the story. A small green arrow pointing to the left or right was superimposed at the bottom of each image to indicate the attended side.
Procedure
Children arrived at the laboratory with their parents and were provided time to acclimate to their environment before placement of the electrode cap began. Once the EEG cap was in place, children were seated in a comfortable chair in an electrically shielded, sound-attenuating booth. They were instructed not to move or lean from side to side. Two speakers were placed on either side of the participant (90o to the left and right of the chair). A computer monitor was positioned approximately 145 cm in front of the participant. Before the data were recorded, participants received instructions to attend to the story played from one speaker while ignoring the story presented from the other speaker. They were told that either a male or a female speaker would narrate the story. An arrow at the bottom of the screen would point to the speaker they should attend to and the attended story would correspond to the pictures on the screen. They were also instructed that unrelated sounds (‘bas’ and ‘buzzes’) would be presented but should be ignored.
At the beginning of each story, participants were presented with a sound sample of the narrator to which they should attend. They were instructed to listen carefully to the story from this narrator and ignore the other voice. Participants attended to a total of four narratives selected from the four story sets, attending twice to the right side and twice to the left side (order either RLLR or LRRL). All participants were presented with two stories narrated by a female and two stories narrated by a male. For the duration of the experiment, participants were monitored by an intercom system and a video camera. Throughout the experiment, an adult trained for behavioral management accompanied the child in the booth. The stories were paused briefly if excessive EEG artifacts were present, such as too many saccades or drifts, or if an electrode needed adjustment. After each story, the experimenter asked the participants three basic comprehension questions about the attended story. These questions were designed to reinforce to the child to attend to a single story at a time, and also to assess whether the child's comprehension of the stories was above chance (6 or more correct answers out of 12). The comprehension questions were always about the attended story and had two alternatives. A response of ‘I don't know’ was considered an incorrect response. Only children who performed with at least 50% accuracy on the comprehension questions were included in the EEG analyses.
EEG recording and analysis
EEG was recorded at a sampling rate of 1024 Hz from 32 Ag-AgCl electrodes attached to an electrode cap and arranged according to the 10/20 system. Recordings were made using the Active-Two system (Biosemi, Amsterdam, Netherlands), which does not require impedance measurements, an online reference, or gain adjustments. The electrode configuration for event-related brain potential recordings is illustrated in Figure 1. Additional electrodes were placed on the left and right mastoid, at the outer canthi of both eyes and below the right eye. Scalp signals were recorded relative to the Common Mode Sense (CMS) active electrode and then re-referenced off-line to the algebraic average of the left and right mastoid. Left and right horizontal eye channels were re-referenced to one another.
ERP analyses were carried out using EEGLAB (Delorme & Makeig, 2004) and ERPLAB (Lopez-Calderon & Luck, 2014). Data were down-sampled to 256 Hz to speed computation and band-pass filtered from 0.1 to 40 Hz. The EEG data were epoched offline between 100 ms prior to and 500 ms after stimulus onset, using the 100 ms pre-stimulus onset for baseline correction. Artifact rejection was initially executed using a 200 ms window moving at 50 ms increments with peak-to-peak rejection criteria of 100 μV for the eye channels and 200 μV for all other channels for almost all included participants. However, on the basis of the visual inspection of the epoched EEG data, individual artifact rejection parameters were selected if the automatic rejection protocol was not sufficient. Moreover, trained research assistants performed a subsequent artifact rejection step to exclude additional epochs containing eye movements and muscle artifacts from further analysis. Out of ~ 400 trials per condition, an average of 253 trials (SD = 58) per participant were accepted for the attended condition, and 253 trials (SD = 55) were accepted for the unattended condition.
For a total of four participants with otherwise clean EEG data, faulty electrodes were replaced with the average mean amplitude of the three neighboring electrodes. The neighboring electrodes were determined based on the rows described below, within the hemisphere of interest.
The mean amplitudes of ERPs were measured between 100 and 200 ms post-stimulus onset, collapsed across the linguistic and nonlinguistic conditions, consistent with previous studies using this paradigm with young children from lower SES families (Neville et al., 2013; Stevens et al., 2009; Stevens et al., 2015). In line with a recent study that measured ERPs in lower SES preschoolers with the same paradigm (Neville et al., 2013), three electrode aggregates were created as follows: anterior: F7/8, F3/4, FT7/8, FC5/6; central: T7/8, C5/6, CP5/6, C3/4; posterior: P7/8, P3/4, PO3/4, O1/2.
The ERP effect of selective attention was operationalized as the mean amplitude difference between ERPs to probes embedded in attended versus unattended stories. Difference waves are a strong method for isolating an individual component of interest (Luck, 2014), in this study, the attention effect. Furthermore, the difference wave approach was consistent with a recent study that measured ERPs with the same dichotic listening paradigm in adults and used difference waves to assess correlations between ERPs of selective auditory attention and visual working memory (Giuliano et al., 2014).
To assess the link between neural indices of selective attention and nonverbal IQ, first, separate correlational analyses were conducted for each electrode aggregate. These correlation analyses avoid a multicollinearity problem that would occur in the presence of highly correlated predictors (electrode aggregates). Such high correlations between predictors would increase the instability of the regression model, inflate the standard errors, and create erroneous beta values for the predictors. Therefore, to avoid the multicollinearity problem, we reported the correlation coefficients between the electrode aggregates and nonverbal IQ. These initial analyses allowed us to demonstrate the contributions of frontal, central, and posterior rows of electrodes in a straightforward fashion. In addition, to control for multiple comparisons resulting from three levels of scalp distribution, the significance level was set at α = .017.
However, in a supplementary analysis that utilized a multiple regression approach, to avoid the collinearity problem due to the high correlation between ERP mean amplitudes of anterior and central electrode locations (r = .80), these electrode aggregates were averaged together to form a single fronto/central predictor.
To further illustrate the association between neural mechanisms of selective attention and nonverbal IQ, an omnibus ANOVA was also included with the between-subject factor of IQ group (described below; top, middle, bottom) and within-subject factors of attention (attended, unattended) and electrode location (anterior, central, posterior). Greenhouse-Geisser corrections were applied when the degrees of freedom were greater than 1. Uncorrected degrees of freedom but corrected p-values are reported.
Results
There was a significant zero-order correlation between the raw nonverbal IQ composite scores and age, r (124) = .40, p < .001. Therefore, we regressed nonverbal IQ on age and used the residuals for further analyses of the relationship between neural mechanisms of selective attention and nonverbal IQ. Then, we created three aggregate measures of ERPs by averaging across eight electrodes within the anterior, central, and posterior rows (electrodes included in each row are detailed in the Method section). The zero-order correlations between age and these aggregates were not significant, all ps > .10. There were no outliers ± 3 SD for the nonverbal IQ measure. There were also no outliers ± 3 SD for the anterior electrode aggregate of the ERPs. However, there was one outlier for the central electrode aggregate (< −3 SD) and one outlier for the posterior electrode aggregate (< −3 SD). Initially, all analyses were conducted with and without these outliers. The direction and strength of the results were consistent with and without these outliers. To be inclusive and reflect the whole spectrum of results, we reported the results from the complete data set.
The associations between neural mechanisms of selective attention and nonverbal IQ were first assessed with correlation analyses. For these correlation analyses, to control for multiple comparisons resulting from three levels of scalp distribution, the significance level was set at α = .017. The mean amplitude difference between ERPs to probes embedded in attended versus unattended stories was significantly correlated with nonverbal IQ scores (see Table 1 for the correlation statistics). These significant links were observed only for the ERPs measured over the anterior and central rows of electrodes (as illustrated in Figures 2a and 2b), but not the posterior rows of electrodes (Figure 2c). Overall, the more positive in amplitude was the difference between the attended and unattended conditions, the higher the nonverbal IQ scores.
| Measure | 1 | 2 | 3 | 4 |
|---|---|---|---|---|
| 1. Nonverbal IQ | – | |||
| 2. ERPs over anterior electrodes | .24** | – | ||
| 3. ERPs over central electrodes | .24** | .80*** | – | |
| 4. ERPs over posterior electrodes | −.02 | .14 | .49*** | – |
- Note: For the nonverbal IQ measure, higher scores are indicative of better performance. For the ERP measures, which reflect the mean amplitude difference between attended and unattended conditions, higher values are indicative of larger selective attention effects. **p < .01; ***p < .001.
In addition, a multiple regression analysis was conducted to assess the extent to which the neural mechanisms of selective attention accounted for individual differences in nonverbal IQ, above and beyond the effects of age, gender, and SES. In this analysis, only children for whom we had complete SES questionnaire information were included (n = 111). Age, gender, and SES were entered as covariates, and neural mechanisms of selective attention were entered as the predictors. As described in the Method section, to avoid the collinearity problem due to the high correlation between ERP mean amplitudes of anterior and central electrode locations (r = .80), the frontal and central electrode aggregates were averaged together to form a single fronto/central predictor. Along with this fronto/central electrode aggregate, the posterior electrode aggregate was also entered as a predictor in this analysis. Since age was included as a covariate, the dependent variable was raw nonverbal IQ score.
The results of this regression analysis are summarized in Table 2. This analysis indicated that age, gender, and SES together explained a significant portion of variance in nonverbal IQ scores, R2 = .14, F(3, 107) = 5.87, p < .01. Among the covariates, while age was a significant predictor of nonverbal IQ scores (p < .01), gender and SES were not significant predictors (p = .91 and p = .16, respectively). The addition of fronto-central and posterior ERP variables significantly contributed to the model, ΔR2 = .07, F(2, 105) = 4.87, p = .01. In this model, only the fronto-central ERP index of selective attention was a significant predictor of nonverbal IQ, p < .01. Larger, more positive ERP mean amplitudes for the selective attention effect were associated with higher nonverbal IQ scores. The posterior ERP index of selective attention was not a significant predictor of nonverbal IQ, p = .12.
| Variable | Model 1 | Model 2 | ||||||
|---|---|---|---|---|---|---|---|---|
| B | SE B | β | p | B | SE B | β | p | |
| Age | 1.22 | .32 | .34 | < .001 | 1.15 | .31 | .32 | < .001 |
| Gender | −.04 | .36 | −.01 | .91 | −.13 | .35 | −.03 | .72 |
| SES | .02 | .02 | .13 | .16 | .02 | .02 | .10 | .29 |
| Fronto-central ERPs | .42 | .14 | .29 | < .01 | ||||
| Posterior ERPs | −.17 | .11 | −.15 | .12 | ||||
To further illustrate the relationship between neural indices of selective attention and nonverbal IQ, participants were ranked from high to low performance based on the nonverbal IQ residuals, then divided into three groups based on their ranking as follows: higher (top third), middle, and lower (bottom third) IQ performance. Univariate ANOVAs revealed that there were no significant between-group differences in age, accuracy on the comprehension questions asked during the selective attention task, or the average number of ERP trials included in the analyses. For the subset of children for whom we had complete SES questionnaire information (n = 111), SES also did not differ between nonverbal IQ groups. Descriptive information on age, SES, comprehension question accuracy, nonverbal IQ scores, and average number of accepted ERP trials are presented in Table 3.
| Bottom IQ group | Middle IQ group | Top IQ group | ||||
|---|---|---|---|---|---|---|
| Variable | n | Mean SD | n | Mean SD | n | Mean SD |
| Age | 41 |
4.60 .58 |
42 |
4.48 .51 |
41 |
4.53 .57 |
| SES | 37 |
27.93 10.43 |
39 |
30.82 11.67 |
35 |
30.64 12.22 |
| Comprehension accuracy | 41 |
8.27 1.47 |
42 |
8.52 1.44 |
41 |
8.59 1.58 |
| Nonverbal IQ | 41 |
10.36 1.20 |
42 |
12.02 .91 |
41 |
14.19 1.30 |
| ERP trials | 41 |
481.22 118.66 |
42 |
514.69 136.90 |
41 |
505.10 108.50 |
Using these groups as a between-subjects factor in a mixed model ANOVA, we evaluated the effect of selective auditory attention on ERPs as a function of IQ performance group. This ANOVA included the between-group factor of IQ performance group (top, middle, bottom), and two within-group factors: attention (attended, unattended) and electrode location (anterior, central, and posterior). The ANOVA statistics are reported in Table 4. There was a significant main effect of electrode location and a significant interaction between attention and electrode location. There was also a significant interaction between attention and IQ group. There were no other significant main effects or interaction effects.
| F | df | p | partial η 2 | |
|---|---|---|---|---|
| Age | .49 | 2, 121 | .61 | .01 |
| SES | .74 | 2, 108 | .48 | .01 |
| Comprehension accuracy | .52 | 2, 121 | .60 | .01 |
| ERP trials accepted | .83 | 2, 121 | .44 | .01 |
| ERP indices of attention | ||||
| Attention | 3.29 | 1, 121 | .07 | .03 |
| Electrode | 199.69 | 2, 242 | < .001*** | .62 |
| IQ group | .51 | 2, 121 | .60 | .01 |
| Attention × electrode | 5.32 | 2, 242 | .02* | .04 |
| Attention × IQ group | 3.44 | 2, 121 | .04* | .05 |
| IQ group × electrode | 1.89 | 4, 242 | .14 | .03 |
| IQ group × attention × electrode | 2.64 | 4, 242 | .06 | .04 |
- *p < .05; ***p < .001.
Subsequent step-down analyses were conducted to unpack the interaction between attention and electrode location. Paired samples t-tests revealed that the mean amplitude of ERPs was significantly larger, more positive, for the probes in the attended versus unattended stories over the anterior and central electrodes (t(123) = 2.40, p = .02; t(123) = 2.57, p = .01, respectively). The ERPs for the attended versus unattended stories did not differ over the posterior electrodes (t(123) = −.48, p = .63).
In addition, subsequent step-down analyses were conducted to unpack the interaction between attention and IQ group. Since there was not a significant three-way interaction between Attention × Electrode × IQ Group, these step-down analyses were conducted with an aggregate measure of all channels included in the analyses. In addition, to provide a more detailed account of what we observed in the grand average plots, mean amplitude differences and 95% confidence intervals are reported for the three rows of electrodes (anterior, central, and posterior) in Table 5. The ERP grand average plots are shown in Figure 3 for the bottom IQ group, Figure 4 for the middle IQ group, and Figure 5 for the top IQ group.
| Mean (μV) | SD | 95% CI | ||
|---|---|---|---|---|
| LL | UL | |||
| Bottom IQ group (n = 41) | ||||
| Anterior electrodes | −.24 | 1.23 | −.62 | .15 |
| Central electrodes | −.16 | 1.21 | −.54 | .23 |
| Posterior electrodes | −.09 | 1.68 | −.62 | .44 |
| All electrodes | −.16 | 1.10 | −.51 | .18 |
| Middle IQ group (n = 42) | ||||
| Anterior electrodes | .42 | 1.60 | −.08 | .92 |
| Central electrodes | .32 | 1.60 | −.18 | .85 |
| Posterior electrodes | −.08 | 1.72 | −.61 | .46 |
| All electrodes | .22 | 1.35 | −.20 | .64 |
| Top IQ group (n = 41) | ||||
| Anterior electrodes | .78 | 1.42 | .33 | 1.22 |
| Central electrodes | .80 | 1.17 | .43 | 1.17 |
| Posterior electrodes | −.04 | 1.40 | −.48 | .41 |
| All electrodes | .51 | 1.03 | .19 | .84 |
The paired samples t-tests revealed that there was no significant effect of attention on ERPs (i.e., no significant differences between the ERPs to probes in the attended versus unattended stories) in the bottom IQ group (Figure 3), t(40) = −.95, p = .35. Although the grand average plot hinted at an emerging attention effect for the middle third IQ group (Figure 4), there was no significant effect of selective attention on ERPs in this group either, t(41) = 1.06, p = .29. In contrast, there was a significant selective attention effect on ERPs in the top IQ group, t(40) = 3.19, p < .01. The mean amplitudes of the ERPs were more positive for the attended condition versus the unattended condition. As illustrated in Figure 5, this attention effect was most evident over the anterior and central rows of electrodes.1
To complement these analyses and directly compare the nonverbal IQ groups, Helmert contrasts were conducted with the ERP difference waves. Helmert contrasts revealed that the top third IQ group exhibited a larger attention effect than the middle and bottom third IQ groups, t(121) = 2.16, p = .03. The attention effect did not significantly differ between the middle and bottom third IQ groups, t(121) = 1.50, p = .14. Figure 6 illustrates the difference waves (attended – unattended) for the three IQ groups overlaid at representative frontocentral and central electrode sites.
Discussion
The present study provided evidence for a noteworthy relationship between neural mechanisms of selective attention and nonverbal IQ performance in young children from lower SES families. We documented prominent individual differences in neural indices of sustained selective auditory attention in lower SES children. These individual differences, as measured by ERPs in a dichotic listening paradigm, were associated with nonverbal IQ scores. In particular, higher nonverbal IQ scores were observed in children who displayed larger mean amplitude differences between ERPs elicited by identical probes embedded in attended versus unattended stories.
Previous studies with higher SES children documented the effects of selective auditory attention as being largest over the anterior and central electrode locations in the 100–200 ms time window (Coch et al., 2005; Sanders et al., 2006). Consistent with these findings, we also documented the effects of selective auditory attention over the anterior and central electrodes in this sample of preschoolers from lower SES households. Furthermore, the associations between neural indices of selective auditory attention and nonverbal IQ performance were also observed over these anterior and central electrode locations, where the effects of selective auditory attention were evident.
Selective auditory attention and nonverbal cognitive performance
Our results align with the argument that attentional abilities are critical for performance in various aspects of nonverbal cognition (Engle & Kane, 2004). Previously, studies of adults and children linked various aspects of attention to nonverbal cognitive abilities using behavioral measures (Astle et al., 2010; Shimi et al., 2014b; Unsworth et al., 2014; Unsworth & Spillers, 2010). In addition, several studies associated neural mechanisms of attention with performance on tasks of nonverbal cognition in adults (Fukuda & Vogel, 2011; Giuliano et al., 2014; Gulbinaite et al., 2014; Kuo, Stokes & Nobre, 2012), and a recent developmental cognitive neuroscience study demonstrated a similar link in children (Shimi et al., 2014a). Here we extend these findings to young children from lower SES families, a population at elevated risk for poorer attentional skills than their higher SES peers (Mezzacappa, 2004; Stevens et al., 2009).
Due to the high temporal resolution of ERPs, we were able to pinpoint relatively early mechanisms of selective attention (100–200 ms) in lower SES preschoolers and demonstrate a relationship between this early selective attention effect and nonverbal cognitive performance. The associations between neural mechanisms of attention and nonverbal cognitive abilities have been mainly reported in studies of visual attention (Fukuda & Vogel, 2011; Gulbinaite et al., 2014; Shimi et al., 2014a). Our results extend these findings to the auditory modality in young children, in line with a recent study that linked a relatively early index of auditory selective attention to visual working memory capacity in adults (Giuliano et al., 2014).
The current study is the examination of ERPs from a large group of young children from lower income families. Our sample size was relatively large compared to many developmental neuroscience studies conducted with similar age groups. This large sample provided a representative portrayal of variability in neural mechanisms of selective attention in lower SES children. Yet, it is important to note that the effect sizes (as indicated by the correlation coefficients) we report here are in general smaller than what have been reported in previous studies that linked neural mechanisms of attention to nonverbal cognitive abilities (Giuliano et al., 2014; Shimi et al., 2014a). One speculation is that the young children in our study might have provided noisier ERP data in general. Noisier ERP data, which would lead to lower signal-to-noise ratio, might have reduced our statistical power to detect stronger associations between selective attention and nonverbal IQ. Another possibility is that the association between neural mechanisms of selective attention and nonverbal cognitive abilities may be weaker in lower SES children compared to higher SES populations. Nevertheless, our results provide an important, initial line of evidence associating individual differences in neural mechanisms of selective attention to nonverbal IQ performance in children from disadvantaged backgrounds.
Interestingly, among this sample of children, SES was not a significant predictor of nonverbal IQ. There may be at least two explanations for this finding. First, our study included only lower SES children and consequently, a restricted range of SES. Within this restricted range, SES may not be a significant predictor of nonverbal IQ. Second, the questionnaire we used for the assessment of SES may not be sensitive enough to distinguish SES disparities in neural mechanisms of selective attention or nonverbal cognition within such a restricted range of SES.
It should also be noted that the sample of children we tested had similar developmental and sociodemographic characteristics (typically developing, right-handed monolingual native speakers of English, from primarily Caucasian households). Although this stringent selection allowed us to avoid numerous potential confounds of ERP studies, we acknowledge that we cannot be certain of the degree to which these results would generalize to other groups of lower SES children. However, given that our findings linked attention to nonverbal cognition, in line with previous research conducted with adults and children on this topic, we would expect our results to extend at least to other populations of typically developing children from lower SES households, despite their differing sociodemographic characteristics.
The current results may imply that neural mechanisms of selective attention support nonverbal cognitive development in lower SES children. However, due to the design of our study, we cannot establish the direction of causality between selective attention and nonverbal IQ. It is possible that there are more basic underlying mechanisms responsible for the subpar performance in both domains of cognition. For instance, in a recent study with adolescents, lower maternal education levels were linked to less efficient auditory processing, as indexed by weaker, more variable, and noisier neural responses to auditory stimuli (Skoe, Krizman & Kraus, 2013). SES disparities in perceptual mechanisms may drive differences in performance on laboratory measures of higher cognitive abilities, such as attention and nonverbal IQ. Future research would benefit from the inclusion of more specific and diverse cognitive tasks to determine which other cognitive factors may mediate or account for the association between neural mechanisms of selective attention and nonverbal intelligence. Furthermore, an intervention design would be necessary to establish a causal relationship between selective attention and nonverbal IQ.
Individual differences in selective attention in lower SES children
Recently, many studies have reported SES disparities in the neural indices of various cognitive functions (Gianaros, Manuck, Sheu, Kuan, Votruba-Drzal et al., 2011; Kishiyama, Boyce, Jimenez, Perry & Knight, 2009; Sheridan, Sarsour, Jutte, D'Esposito & Boyce, 2012; Stevens et al., 2009). However, there has been a paucity of research addressing individual differences in brain functioning among children growing up in lower SES families. Our findings contribute to and extend this understudied area of research. Here we document a considerable amount of variability among lower SES children for neural indices of selective attention. When children were divided into three groups based on their nonverbal IQ scores, only children in the top third group showed a significant effect of attention on ERPs. In contrast, we did not find a significant modulation of ERPs by selective attention in children whose nonverbal IQ scores fell into the middle or bottom thirds. Likewise, when we directly compared the ERP attention effects between these IQ groups, children in the top IQ group displayed more enhanced neural indices of selective attention compared to the children in the middle and bottom IQ groups.
It is important to note that we created these categorical groups to better illustrate our ERP results. Due to testing time constraints, we were able to administer only a subset of tasks from the full battery of the nonverbal IQ assessment. While keeping the testing session relatively short with such young children allowed us to reduce fatigue and improve performance, the lack of a full IQ battery precludes us from considering clinical classifications or cut-off points. The terciles in our study are based on the scores of the sample rather than previously established norms or standards; therefore, the exact categorization criteria of the groups should be interpreted with caution. Nevertheless, our results emphasize that lower SES children do not constitute a homogenous group of at-risk children who show similar levels of alterations in neural mechanisms of selective attention.
The underlying mechanisms of these individual differences in selective attention remain to be investigated in order to understand the interactions between genetic and familial factors. Polymorphisms of several candidate genes have been linked to individual differences in attention abilities in typically developing individuals (Blasi, Mattay, Bertolino, Elvevåg, Callicott et al., 2005; Fan, Fossella, Sommer, Wu & Posner, 2003; Green, Munafò, DeYoung, Fossella, Fan et al., 2008; Parasuraman, Greenwood, Kumar & Fossella, 2005; Rueda, Rothbart, McCandliss, Saccomanno & Posner, 2005). However, it is unclear how genetic effects might manifest in lower SES children. While some studies found stronger genetic influences in lower SES populations (Nobile, Giorda, Marino, Carlet, Pastore et al., 2007; Nobile, Rusconi, Bellina, Marino, Giorda et al., 2010; Sadeh, Javdani, Jackson, Reynolds, Potenza et al., 2010; Williams, Marchuk, Siegler, Barefoot, Helms et al., 2008), others reported suppressed genetic influences in lower SES households (Rhemtulla & Tucker-Drob, 2012; Tucker-Drob, Rhemtulla, Harden, Turkheimer & Fask, 2011; Turkheimer, Haley, Waldron, D'Onofrio & Gottesman, 2003). Further research is warranted to understand the potential genetic underpinnings of individual differences in selective attention in lower SES children.
Furthermore, the role of familial factors in the emergence of individual differences in selective attention has yet to be explored. Previous research has demonstrated that lower SES children are at greater risk for unfavorable household characteristics, such as low cognitive stimulation, high stress, and poor parenting skills (Bradley & Corwyn, 2002; Brooks-Gunn & Duncan, 1997; Evans, 2004) and a few training and intervention studies have demonstrated that some of the negative outcomes associated with such adversity could be ameliorated (Bierman, Nix, Greenberg, Blair & Domitrovich, 2008; Campbell, Pungello, Miller-Johnson, Burchinal & Ramey, 2001; Mackey, Hill, Stone & Bunge, 2011; Neville et al., 2013). It remains crucial to delineate which familial characteristics compromise versus promote the neural mechanisms of selective attention in lower SES children.
Conclusions
The present study provides initial evidence for noteworthy individual differences in neural indices of selective attention in young children from lower SES families. These individual differences in neural indices of selective attention were associated with nonverbal IQ performance. Children who revealed more pronounced attention effects, as measured by ERPs, also demonstrated superior nonverbal cognitive abilities. Further research is warranted to pinpoint the factors that account for the variability observed in neural mechanisms of selective attention in children from disadvantaged backgrounds.
Acknowledgements
This work was supported through Department of Education/Institute of Education Science Grant R305B070018 to HJN. We thank the members of the Brain Development Lab for their support in data acquisition and processing, Courtney Stevens, Theodore Bell and Jason Isbell for helpful comments on an earlier version of the manuscript, and Head Start of Lane County for its continued cooperation in this research.
