“Increased exposure to environmental risk factors alters neural oscillatory activity during language learning in children.“
Analia Marzoratti, The University of Texas at Dallas, School of Behavioral and Brain Sciences
Children from low socioeconomic (SES) families tend to enter school with lower language abilities, often leading to academic problems, and likely linked with exposure to more environmental factors known to cause stress (Sheridan & McLaughlin, 2016). These factors may affect language learning skills, and explain outcomes in ways SES alone cannot (Evans et al., 2013). Vocabulary, a factor essential for all learning, appears to be especially affected. In this study, we use EEG to examine processes involved in successful word learning, and differences based on risk exposure.
Children aged 8-15, 28 exposed to more risk factors and 23 exposed to fewer, performed a word learning task where they were presented sets of three sentences ending in the same pseudoword and asked to define it using context. We studied how neural oscillations underlying language processing change as a word moves from ‘unknown’ to ‘known’, focusing on theta, associated with semantic retrieval and integration. Both groups exhibited increased theta as words became ‘known’. However, our lower risk group showed the largest changes earlier, at the second presentation. This indicates that word learning may require fewer exposures to the word for children experiencing lower risk, which has important implications for how vocabulary is taught.
The academic achievement gap associated with low socioeconomic status (or SES) is very well studied, but there is an amount of variability in learning outcomes that SES alone cannot account for (Sheridan & McLaughlin, 2016). There are multiple environmental qualities that are highly correlated with SES despite the wide varieties of individuals SES categories encompass, such as poorer housing conditions, reduced exposure to language, and insufficient sleep (Cadima et al, 2009; Dilnot et al, 2016). Under the cumulative risk model, children with a greater cumulative number of such environmental risk factors may show differences in cognitive development that could lead to problems with language skills and thus school readiness and long-term academic achievement (Evans et al., 2013).
Differences in cognitive skill development would be reflected in changes to neural structure and in turn functional connectivity, potentially resulting in reliance on different cognitive structures and abilities when processing and storing new information (Ralph, et. al, 2020). EEG recordings are among the most-used and least invasive methods for studying neural function in children, and could capture this effect by allowing us to observe (Kuhl, 2010; Maguire & Abel, 2013). Namely, ERSP analysis is a form of time-frequency analysis that decomposes the EEG signal into individual frequencies including theta, a band at 4-8Hz thought to index word learning (Maguire et al., 2018). Differences in spectral activity reflect differences in cognitive processes engaged (Maguire & Abel, 2013), thus analysis of this frequency implicated in word-learning would clarify the neurocognitive basis behind SES-related differences in language development.
In my study I investigated what theta activity patterns occur across exposures to an unknown word during a word-learning task, and whether this differs based on cumulative risk factor exposure. Behavioral studies have shown a negative correlation between higher levels of cumulative risk and literacy skills and a negative linear relationship between risk factor exposure and outcome (Cadima et al, 2010; Evans et al, 2013). Because of this we expected that our low risk group would show better performance on our behavioral tests of cognitive ability, demonstrating more effective strategies for verbal processing. As increases in theta power are thought to reflect neural processes critical to language learning, given the advantage we predicted for our low risk group, we expected them to show a faster increase in theta power across sentence presentations (Maguire et al, 2018; Ralph et al 2020). These results would provide neurological evidence for a difference in the cognitive strategies adopted by children exposed to fewer risk factors that underlies their improved academic performance compared to their higher risk peers.
Participants included 51 right-handed, multilingual, English-speaking children children between the ages of 8-15. Risk group was determined based on the number of risk variables a participant possessed out of a possible nine, which had varying point values. These variables included high levels of home chaos, high residential mobility, few books in the home, time per day spent reading or being read to, high home density, sharing a bedroom, living in a neighborhood with high poverty, low maternal education, and qualification for free or reduced lunch (Ferguson et al, 2013; Logan et al, 2019). Because income and maternal education are particularly influential on cognitive outcomes, their corresponding variables were weighted more heavily and included levels of severity. The other seven variables were valued at one point. The low risk group contained 23 children with 0-2.5 points (14 males; Mage: 11.26, SD: 2.28) and the high risk group contained 28 children with 6.5-11 points (8 males; Mage: 11.36, SD: 2.32).
During the EEG recording, children completed a word learning task in which they were presented with 50 sets of three sentences each ending in the same pseudo word corresponding to a single target word. For example, for the target word bicycle we could have presented: “He loaned his older brother his huth. He took a ride on his huth. He taught his son to ride a huth.”Participants were asked to discern the meaning of the pseudoword given the context of these sentences.
Behavioral measures of cognitive ability include the percent of correct responses the child gave on this word learning task as well as a battery of cognitive and language assessments, counterbalanced to occur before or after the EEG task. This included the Peabody Picture Vocabulary Test–Fourth Edition (PPVT-4; Dunn & Dunn, 2007) as a measure of receptive vocabulary, the Gray Oral Reading Tests–Fifth Edition (GORT-5; Wiederholt & Bryant, 2012) as a measure of reading comprehension, and the Test of Word Reading Efﬁciency–Second Edition (TOWRE-2; Torgensen, Wagner, & Rashotte, 2012) as a measure of reading efﬁciency. Demographic information was collected from parents on the day of the EEG recording.
Data were collected using a 62 electrode EEG Cap referenced at the central electrode, FCZ; and average reference was computed for all electrodes to isolate power fluctuations. EEG data was epoched from 500 msec before to 1500 msec after target word onset for each of the three sentences, and the mean ERSP for the theta frequency (4-8Hz) was computed for all data channels. Only trials where participants correctly responded were analyzed, allowing us to study processes underlying successful word learning. We used a 2 (Risk Groups) × 3 (Sentence 1, 2, or 3) ANOVA to evaluate the EEG data and a Student’s two-sample t-test for behavioral scores. Results were significant at p>0.05.
As predicted, the low risk group performed better on all of the cognitive and language assessments. Importantly, they showed a significantly greater percentage of correct responses on the word learning task compared to the high risk group.
To interpret the EEG results, we produced scalp maps comparing theta oscillatory activity across sentences within each group, and for each sentence between the two groups, allowing us to localize effects to specific areas of the scalp. We also produced spectrograms for significant electrodes for the whole epoched time window (-500 to 1000 ms) to examine the timing of periods of significance.
The high risk group showed significant differences in theta power between sentences from 100 to 550ms after the presentation of the final word, with significant electrodes lateralized to the right and right-frontal areas of the scalp. Spectrograms of a representative right-frontal electrode showed that this difference was driven by the increased theta power the high risk participants displayed in the third sentence compared to the first and second, demonstrating a slow escalation across sentences.
The low risk group showed a later period of significant difference in theta power across sentences, from 600-850ms after the final word was presented and in more localized fronto-central electrodes. The spectrograms of a representative frontal electrode show that the main source of these differences is a large spike in theta power in the second sentence, with greater theta power than the high risk group displayed at any point in the task. The later timing of the significance for the low risk group reflects a longer period of theta activity compared to the high risk group, confirmed by the length of the significant activation shown in the spectrogram.
Between the high and low risk groups, there were significant differences in fronto-central theta activity in the second sentence both at 50-300ms and 700-800ms after the final word. Similarly to the differences across sentences in the low-risk group, the spectrograms show that this effect was driven by the spike in fronto-central theta activity the low-risk group displayed in the second sentence. The high-risk group in the spectrograms also shows a pattern of theta activation that slowly increases across sentences, and peaks at the third sentence, where it begins to approximate the pattern the low-risk group shows at the second sentence.
This study sought to decompose SES into the individual environmental qualities uniquely associated with its various levels in order to examine whether these environmental factors could mediate the link between SES and children’s cognitive and academic outcomes. Specifically, we examined the relationship between the cumulative number of risk factors a child was exposed to and their cognitive performance during a word-learning task, using time-frequency analysis of EEG recordings to isolate differences in theta power, a frequency band implicated in word-learning.
The fact that the low-risk group performed significantly better than the high-risk group on all of the behavioral measures of language ability, especially the word-learning task, suggests that they engage more effective cognitive processes for language processing. This effect is supported by much of the existing literature on the effect of environmental risk on learning outcomes (Cadima et al, 2009; Dilnot et al, 2016; Sheridan & McLaughlin, 2016).
The differences in theta activation between the high and low risk groups appear to be driven by a large peak in theta activation for the low risk group in the second sentence, suggesting that they are successfully integrating the word in context by this point. The high-risk group on the other hand never reaches the same level of activation, and the highest theta power it reaches occurs in sentence three, requiring two sentences prior to slowly increase to this level. This suggests that the low-risk group may be able to learn the word after fewer presentations than the high-risk group, supporting the behavioral results in implying that they engage more efficient word learning processes.
Differences in theta activity across sentences within the two groups provide us a window into what these cognitive processes being engaged might look like. The disparities in the timing and localization of these significant effects for the low and high risk groups support my inference that they may have used different cognitive mechanisms to complete the word learning task. For the low risk group, differences in theta activity between sentences appear in focal, frontal electrodes. They also occur later in time after the presentation of the word being “learned”, which suggests theta activation lasts for a longer period of time for this group. In contrast, the high risk group shows its significant differences in theta between sentences in more diffuse, right-frontal electrodes, occurring earlier in the time period after the final word, and this theta activation appears to last for a shorter period of time.
In combination, the behavioral and physiological results imply that the high and low risk groups engage different cognitive strategies during word-learning, and that of the two, the one adopted by the low-risk group is the more mature and effective one. This means it may take fewer exposures to a word for low-risk children to learn it compared to high-risk children, leading to differences in vocabulary that could impact long-term learning outcomes.
Overall, this study took important steps in showing that cumulative environmental risk may lead to neural effects that mediate learning differences that could help to explain some of the variability in cognitive outcomes for low SES children. In isolating the specific environmental factors that affect cognitive development, we can help to more effectively target interventions and policy aimed at ensuring that children are developing in optimal environments for learning. By understanding the mechanisms underlying these word learning differences, we can also develop better strategies for teaching vocabulary to children from all backgrounds, beginning the long process of narrowing the academic achievement gap.