The N400 response indexes word learning from linguistic context in children

Abstract

Word learning from linguistic context is essential for vocabulary growth from grade school onward; however, little is known about the mechanisms underlying successful word learning in children. Current methods for studying word learning development require children to identify the meaning of the word after each exposure, a method that interacts with the act of learning. In this study, school-aged children (11–14 years) performed a word learning task as their EEG was recorded. The word learning task required children to identify the meaning of new words presented in sentence triplets that either provided enough context to support word learning or did not provide a supportive context. Children displayed a significant attenuation of the N400 for words for which they identified meanings compared to those for which they were unable to identify meanings. Additionally, the N400 to the final presentation of learned words paralleled that of a known real word. These results indicate that the same mechanisms related to the N400 for extracting word meaning may be associated with word learning in children.

Full text

Full Terms & Conditions of access and use can be found at
http://www.tandfonline.com/action/journalInformation?journalCode=hlld20
Download by: [SDSU San Diego State University] Date: 29 November 2017, At: 10:43
Language Learning and Development
ISSN: 1547-5441 (Print) 1547-3341 (Online) Journal homepage: http://www.tandfonline.com/loi/hlld20
N400 Response Indexes Word Learning from
Linguistic Context in Children
Alyson D. Abel, Julie Schneider & Mandy J Maguire
To cite this article: Alyson D. Abel, Julie Schneider & Mandy J Maguire (2017): N400 Response
Indexes Word Learning from Linguistic Context in Children, Language Learning and Development,
DOI: 10.1080/15475441.2017.1362347
To link to this article: https://doi.org/10.1080/15475441.2017.1362347
Published online: 27 Nov 2017.
Submit your article to this journal
Article views: 1
View related articles
View Crossmark data

N400 Response Indexes Word Learning from Linguistic Context in
Children
Alyson D. Abel
a
, Julie Schneider
b
, and Mandy J Maguire
b
a
School of Speech, Language & Hearing Sciences, San Diego State University;
b
Callier Center for Communication
Disorders, University of Texas at Dallas
ABSTRACT
Word learning from linguistic context is essential for vocabulary growth
from grade school onward; however, little is known about the mechanisms
underlying successful word learning in children. Current methods for study-
ing word learning development require children to identify the meaning of
the word after each exposure, a method that interacts with the act of
learning. In this study, school-aged children (11–14 years) performed a
word learning task as their EEG was recorded. The word learning task
required children to identify the meaning of new words presented in
sentence triplets that either provided enough context to support word
learning or did not provide a supportive context. Children displayed a
significant attenuation of the N400 for words for which they identified
meanings compared to those for which they were unable to identify mean-
ings. Additionally, the N400 to the final presentation of learned words
paralleled that of a known real word. These results indicate that the same
mechanisms related to the N400 for extracting word meaning may be
associated with word learning in children.
Word learning is essential to cognitive and social development across the lifespan (Frishkoff, Collins-
Thompson, Perfetti, & Callan, 2008). Young children can quickly map new words onto objects or
actions in their environment; however, learning using only surrounding linguistic context, the
primary method for word learning available to older children and adults, is substantially more
difficult and less well-researched (Fukkink, 2005). For example, how does one identify the meaning
of the word gossamer when encountering this sentence from David Herbert Lawrence’sSons and
Lovers “Fascinated, he watched the heavy dark drop hang in the glistening cloud, and pull down the
gossamer”? It is estimated that 4
th
graders only learn 3–5% of the unfamiliar words that they first
encounter when reading, a number that increases slowly with age (Swanborn & De Glopper, 1999).
Further, many children, including those who are raised in poverty and those who have language
disorders like Specific Language Impairment, exhibit slow vocabulary growth over the course of
grade school (Hart & Risley, 1995; Hoff, 2003; Steele & Mills, 2011). Identifying how average,
typically developing children can learn a new word using only the linguistic context, and the
underlying neural processes engaged when doing so, will greatly increase our understanding of
vocabulary growth during the school years, when learning is imperative to academic success.
Behavioral research indicates that word learning from context is a slow, deliberate process that
unfolds over multiple exposures to the word (Frishkoff et al., 2008; Fukkink, 2005). Studies of word
learning from context often ask participants to provide a possible word meaning after each exposure
to the novel word. While this method can identify changes in one’s assumptions about a word’s
CONTACT Alyson D. Abel alyson.abel@mail.sdsu.edu School of Speech, Language & Hearing Sciences, San Diego State
University, San Diego, CA 92182.
Color versions of one or more of the figures in the article can be found online at www.tandfonline.com/hlld.
© 2017 Taylor & Francis Group, LLC
LANGUAGE LEARNING AND DEVELOPMENT
https://doi.org/10.1080/15475441.2017.1362347
Downloaded by [SDSU San Diego State University] at 10:43 29 November 2017

meaning over exposures, how many exposures, and what information provides enough support for
successful word learning, verbalizing a word’s potential meaning during the learning process likely
influences the act of learning. To address this possible limitation, studies with adults have utilized the
N400 Event Related Potential (ERP) component as a non-invasive, objective way to study the process
of word learning (e.g., Batterink & Neville, 2011; Borovsky, Kutas, & Elman, 2010; Mestres-Missé,
Rodriguez-Fornells, & Münte, 2007; Perfetti, Wlotko, & Hart, 2005; Torkildsen et al., 2008). The
N400 component is represented by a negative amplitude at around 300–500 milliseconds (msec) that
is greater for a novel or unexpected stimulus and has been identified as playing a role in semantic
processing (King & Kutas, 1995; Kutas & Federmeier, 2011; Neville, Mills, & Lawson, 1992;
Pulvermüller, Lutzenberger, & Birbaumer, 1995). The fact that the N400 response to an unknown
word is significantly larger than to a known word has made it ideal for studying the processes
underlying word learning, specifically whether and when the N400 response to a word being learned
attenuates to resemble the N400 to known words (Batterink & Neville, 2011; Borovsky et al., 2010;
Frishkoff, Perfetti, & Collins-Thompson, 2010; McLaughlin, Osterhout, & Kim, 2004; Mestres-Missé
et al., 2007). This attenuation is interpreted as the semantic representation underlying the unknown
word becoming more robust and semantically rich through learning.
In the current study we investigated whether school-aged children (11–14 years) demonstrate
neural correlates of word learning from context similar to what has been shown in adults.
Specifically, do they exhibit an attenuation of the N400 over three or fewer exposures to the novel
word and does the N400 then mirror that of a real word? This pattern of findings would provide new
information about the speed, process, and mechanisms underlying word learning in typically
developing school age children. Most notably, it would indicate that when children successfully
learn a new word from context, the neural signal underlying that word quickly mirrors that of words
already in their lexicon. These middle-school years, between ages 11 and 14, incorporate an
important and relatively understudied developmental period for studying vocabulary growth.
Middle school highlights a period when English Language Arts programs have shifted to emphasize
literacy analysis over direct instruction for vocabulary, thus learning from context becomes a
primary strategy for word learning during these years (Kelley, Lesaux, Kieffer, & Faller, 2010).
In this study, adapted from methods used with adults by Mestres-Missé et al. (2007) and Batterink
and Neville (2011), children read sentence triplets in which each sentence ended with the target
word. Triplets were organized in three conditions. The experimental condition (Meaning) supported
word learning such that, across the sentence triplet, the linguistic context increasingly constrained
the target novel word’s meaning. There were two control conditions. The No Meaning condition
controlled for repeated exposures to a novel word without contextual support for the word’s
meaning. The Real Word condition provided a means to compare the N400 amplitude of the
novel word to a real word provided in the same context.
In addition to ensuring that the stimuli included only early-acquired words, we made two changes
to the previous adult studies. First, to determine if the word had been learned, we asked the
participants to verbally identify the meaning of the word, if possible, after each sentence triplet.
An important difference between our task and previous behavioral research is that we did not ask for
a response after each exposure to the word; instead, participants provided their response at the end
of each triplet. This decision was based on our belief that explicitly providing a response after one
exposure forces the participant to pick a meaning early in the learning process, which may influence
how they process the following sentence. Second, we analyzed N400 responses based on whether
children provided a meaning for the novel word. Thus, in the Meaning and No Meaning conditions,
if the children provided a meaning, the item was classified as a learned word (Learned Word). Items
for which children did not provide a meaning were classified as not learned words (Not Learned
Word). We took this somewhat innovative approach to the analysis because it more directly
addresses our goal of clearly identifying changes in the brain that correspond to building a robust
semantic representation of a new word over multiple exposures, regardless of the word. Thus, the
Learned Word classification includes items for which the child has built such a representation, while
2A. D. ABEL ET AL.
Downloaded by [SDSU San Diego State University] at 10:43 29 November 2017

the Not Learned Word classification includes items for which no representation was created. Our
conditions (Meaning, No Meaning) were defined by whether a representation should have occurred,
which would be less accurate for distinguishing when a semantic representation did or did not occur,
thus adding noise to the data. Using the Learned/Not Learned classification allowed us to address
two questions related to the neural underpinnings of word learning in children. First, do school-aged
children, similar to adults in previous studies, show an N400 amplitude attenuation for words that
gain meaning across exposures compared to words that do not? Second, does the N400 correspond-
ing to Learned Words parallel the N400 to known, real words?
Methods
Participants
Participants included 28 right-handed, monolingual, English-speaking children between 11 and
14 years old (12 male, 16 female; M= 13, Range = 11;2–14;9). Five participants were lost due to
inability to attend to stimuli (1), inability to read at the presentation rate (1), and having data
containing too many artifacts (3); therefore, a total of 23 participants were included in our analysis
(8 male, 15 female; Mage = 12;7, Range = 11;2–14;9). The 11–14 age range corresponds with the
middle school years, an age at which, as previously discussed, there is a shift in vocabulary instruction
that makes learning from context important for academic success. Exclusion criteria, obtained via
parent report, included: history of significant neurological issues (traumatic brain injury, CVA, seizure
disorders, history of high fevers, tumors, or learning disabilities), use of controlled substances within
24 hr of testing, and medications other than over-the-counter analgesics. No parent reported a history
of reading disability or developmental language delay. Based on parental reports, 7 of the 23 children
came from low-income homes, qualifying for free or reduced lunch.
Word Learning from Context Task
Stimuli.
In the experimental word learning task, participants read sets of sentence triplets in which the target
word appeared in the sentence-final position. For two of the three conditions (Meaning, No Meaning)
the target word was a novel word and in the third condition (Real Word) the target word was a real word.
In the Meaning condition, the sentence triplets supported the meaning of the novel word by increasing in
cloze probability across the three sentences (low, medium, high). The calculation of cloze probability is
described below. In the No Meaning condition, all sentences were classified as low cloze probability, and
each of the three sentences was composed to end in a different target word. In this way the No Meaning
condition served as a control for repeated exposure to the target word without providing support for the
novel word’s meaning. The Real Word condition mirrored the Meaning condition in terms of increasing
cloze probability across each triplet; however, the target words were real words instead of novel words.
This condition served as a comparison for the third presentation of the novel word in the Meaning
condition to determine whether processing of the novel word paralleled that of a real word.
Sentences were chosen from a pool of 411 sentences, 6–9 words in length, which varied in how
well the context predicted the target word. Target words were all concrete nouns considered to be in
an average child’s productive vocabulary by 30 months of age (MacArthur-Bates Communicative
Developmental Inventories, MBCDI; Fenson et al., 2006). Target words appeared in the sentence-
final position and were preceded either by a possessive (i.e., my, your, etc.) or a determiner (aor the).
To control for how well each sentence constrained the meaning of the target word, the cloze
probability of each sentence was calculated by removing the target word and asking 238 under-
graduate students to provide the word that they thought best completed each sentence. From this,
sentences were classified as being low (M= 4.14%, SD = 6.19), medium (M= 44.53%, SD = 11.04), or
high (M= 88.51%, SD = 10.89) cloze probability based on the percentage of correct responses
provided. Novel words used in the Meaning and No Meaning conditions came from Storkel’s(2013)
corpus of consonant-vowel-consonant sequences, which includes phonetic transcription of all
LANGUAGE LEARNING AND DEVELOPMENT 3
Downloaded by [SDSU San Diego State University] at 10:43 29 November 2017

sequences. The sequences were orthographically transcribed by the first author and the transcription
was confirmed by two independent readers. Examples of a triplet for each condition are provided in
Table 1. Participants were randomly assigned to one of eight randomized orders of 63 sentence
triplets, 21 triplets per condition.
Although replacing a real word with a novel word does not directly parallel real-world word learning,
school-aged children often learn new words with nuanced meanings for concepts they already have
names for (Borovsky, Elman, & Fernald, 2012). This is done by anchoring the novel word’s meaning to a
known word or synonym then identifying differences between the known and novel words’meanings
with exposure. As such, the current study addresses that first, critical point in word learning.
Procedure
Participants sat in a chair 1 m from a computer monitor. They were told that they would read sets of three
sentences presented word-by-word with a target word, either real or novel, as the last word in each sentence.
Sentences were presented word-by-word with each word appearing for 500 ms and a blank screen between
words appearing for 300 ms. The blank screen directly preceding the target word was presented for 600 ms
to establish a baseline for analysis of the novel word. The target word was presented for 600 ms.
At the end of each sentence triplet, participants answered a test question, which differed accord-
ing to condition. For the Meaning and No Meaning conditions, children were asked whether the
novel word represented a real word and, if so, what the real word was. For the Real word condition,
children were asked to provide a synonym for the real word. A trained examiner transcribed test
question responses on-line. The first author scored the responses off-line, considering the following
criteria. Given the interest of this study in whether the N400 is specific to word learning, for the
Meaning and No Meaning conditions, test question answers were classified as a Learned Word,
indicating that the participant provided a meaning for the novel word, regardless of correctness, or
as a Not Learned Word, indicating that no meaning was provided. Consistent with the study design,
more items from the Meaning condition (79% vs. 21% from No Meaning) were classified as Learned
Words and more items from the No Meaning condition (82% vs. 18% from Meaning) were classified
as Not Learned Words. As mentioned, the goal was to differentiate the processes underlying the
creation of a robust semantic representation over exposures, compared to the same type of context
but with no robust representation occurring. Thus, we feel that classifying the items based on the
child’s report of a representation (Learned Word) vs. no representation (Not Learned Word) is more
consistent with the study goals than using our experimental conditions of Meaning vs. No Meaning.
Prior to the test session, participants completed a training session consisting of an example of one
triplet for each condition. Feedback was provided during the training session and repetition was
provided if requested. Feedback was not provided during the test session.
EEG acquisition
EEG was collected from 64 silver/silver-chloride electrodes mounted within an elastic cap
(Neuroscan Quickcap) which are placed according to the International 10–20 electrode placement
standard (Compumedics, Inc.). EEG data were recorded continuously using a Neuroscan SynAmps2
amplifier and Scan 4.3.2 software sampled at 1 kHz with impedances typically below 5 kΩ.
Table 1. Example stimuli for meaning and no meaning conditions.
Meaning Condition No Meaning Condition
Sentence 1 Her parents bought her a pav. Her favorite toy of all time is the zat.
Sentence 2 The sick child spent the day in his pav. He had a lot of food on his zat.
Sentence 3 Mom piled the pillows on the pav. Before bed I have to take a zat.
Real word condition: Identical to Meaning condition but with the real word instead of a novel word.
(For example, the word bed would replace the word pav in the example above)
4A. D. ABEL ET AL.
Downloaded by [SDSU San Diego State University] at 10:43 29 November 2017

EEG pre-processing
Data were recorded with the ground at Fz and the reference electrode located near the vertex,
resulting in small amplitudes over the top of the head. To eliminate this effect, data were re-
referenced offline to the average potential over the entire head, approximating the voltages relative
to infinity (Nunez & Srinivasan, 2005). In order to minimize a small bias in the electrode-based
average reference (Junghöfer, Elbert, Tucker, & Braun, 1999), a spline-based estimate of the average
scalp potential (Ferree, 2006) was computed using spherical splines (Perrin, Pernier, Bertrand, &
Echallier, 1989). In participants with a small number of bad electrodes, the splines were used to
interpolate those electrodes, yielding a total of 62 data channels in every subject. The validity of this
method of interpolation is supported theoretically for 64 or more electrodes (Srinivasan, Tucker, &
Murias, 1998).
Blinks and eye movements were monitored via electrodes mounted above and below the left eye.
The data were processed to remove ocular and muscle artifacts following three steps. First, poorly
functioning electrodes were identified visually and removed. Second, eye blink artifacts were
removed by a spatial filtering algorithm in the Neuroscan Edit software using the option to preserve
the background EEG. Third, time segments containing significant muscle artifacts or eye movements
were rejected.
After ocular and muscle artifact removal, data were separated according to condition or classifica-
tion (Real Word, Learned Word, Not Learned Word) and sentence number within the triplet (1, 2, 3).
On average, each child had 39 epochs for the Learned Words classification and 25 epochs for the Not
Learned Words classification. EEG data were then segmented into epochs spanning 500 msec before to
1500 msec after the presentation of the target word. A semi-automatic artifact rejection procedure
following two rejection criteria: (1) amplitudes ±75µV and (2) voltage differences between two adjacent
time points >50 µV.
ERP calculation
ERPs were time-locked to the onset of the target word in the sentence-final position. For each item,
the mean amplitude of the prestimulus interval (−100 msec-0 msec) was subtracted from each time
point and those data were averaged across trials to create the ERP. We took a focused approach to
our analysis of the N400 and determined the location of the N400 based on previous developmental
research all of which identifies the N400 effect in frontal sites in children (e.g., Atchley et al., 2006;
Henderson, Baseler, Clarke, Watson, & Snowling, 2011). Thus, we computed the average amplitude
for each response type at widespread frontal and central electrodes (FC3, FCz, FC1, FP2, FPz, F3,
FC5, FC2, C3, AF3, F5, Cz, CPz, C1, C2, Fz, F1, AF4) between 300–500 msec after presentation of
the target word (Batterink & Neville, 2011; Mestres-Missé et al., 2007).
Results
Behavioral findings
The study was designed with 21 triplets in each of the three conditions (Meaning, No Meaning, Real
Word). Children correctly identified the meaning of the novel word 72.4% of the time in the
Meaning condition and responded that there was no meaning for the novel word 70.4% of the
time in the No Meaning condition. As noted above, the EEG analysis focused on Learned Words,
items for which the child provided a meaning for the novel word, vs. Not Learned Words, items for
which the child did not provide a meaning. 79% of items in the Meaning condition were classified as
Learned Words and 82% of items from the No Meaning condition were classified as Not Learned
Words. This pattern supports the study design, specifically that the Meaning condition was designed
such that participants could identify a meaning for the novel word and that the No Meaning
condition did not support meaning identification.
LANGUAGE LEARNING AND DEVELOPMENT 5
Downloaded by [SDSU San Diego State University] at 10:43 29 November 2017

Event-related potentials
Our analyses addressed the following two questions: (1) Do children show an N400 amplitude
attenuation for Learned Words compared to Not Learned Words? And (2) Does the N400 corre-
sponding to Learned Words parallel the N400 to known words? To address the first of these
questions, a 2 (Classification: Learned, Not Learned) x 3 (Sentence: 1,2,3) Repeated Measures
ANOVA revealed a significant interaction, F(2,44) = 3.25, p< 0.05. No significant main effects
were found for Classification or Sentence. Based on our predictions and the significant omnibus
effect, we conducted posthoc analyses to identify whether the interaction was the result of differences
in the N400 attenuation between presentations of the Learned or Not Learned Words. A one-way
ANOVA for Learned Words across presentations was significant [F(2,44) = 13.82, p<0.01], while
there was no significant attenuation of the N400 effect across presentations for the Not Learned
Words [F(2,44) = 0.93, p= NS]. To further delineate where the N400 attenuated between presenta-
tions for Learned Words, we conducted two t-tests. While controlling for multiple comparisons
using a Bonferroni correction, there were no significant differences for Learned Words between
presentation 1 and 2 [t(22) = -.78, p= NS]; however, there was a significant difference between
presentation 2 and 3 [t(22) = 2.81, p=0.01]. As shown in Figures 1 and 2, the N400 amplitude
attenuates across the sentence triplet for Learned Words, with the greatest change occurring between
the 2nd and 3rd sentences in the triplet, with no such pattern of N400 change for the Not Learned
Words. Providing further support for this finding, we conducted three t-tests comparing the Learned
Words and Not Learned Words at each of the three presentations. There was no significant
difference between the conditions at presentation 1 [t(22) = -.64, p= NS] or presentation
2[t(22) = -.80, p= NS] but the N400 did diverge at presentation 3 [t(22) = 2.1, p<0.047].
To address the second question, we conducted a t-test comparing the N400 amplitude of the novel
word in sentence 3 of the Learned condition to the N400 amplitude of the real word in sentence 3 of
Figure 1. ERPs across sentence presentations 1, 2 and 3 for Learned and Not Learned Words. Learned words (blue) attenuate
between 300-500 msec while Not Learned words (black) do not show the same attenuation.
The red boxes on the headmap denote the electrodes plotted in this figure.
6A. D. ABEL ET AL.
Downloaded by [SDSU San Diego State University] at 10:43 29 November 2017

the Real Word condition. The two conditions did not differ [t(22) = 0.19, p=NS] indicating that by
the third presentation, children process a newly learned word like a real word (see Figure 3). A one-
way ANOVA of the Real Word condition also was not significant [F(2,44) = 0.94, p= NS], indicating
that there was no attenuation of the N400 across presentations for Real Words, as shown in Figure 4.
Figure 2. Line graph demonstrating N400 amplitude averages across classification (Learned/Not Learned) and presentation (1,2,3).
Figure 3. ERPs for the third sentence presentation of Learned and Real Words. Learned words (blue) showed no significant
amplitude difference between 300-500 msec compared to Real words (red). The red boxes on the headmap denote the electrodes
plotted in this figure.
LANGUAGE LEARNING AND DEVELOPMENT 7
Downloaded by [SDSU San Diego State University] at 10:43 29 November 2017

Discussion
School-aged children learn most of their new vocabulary by utilizing information in the surrounding
linguistic context to infer a new word’s meaning. Using an experimental word learning from context
task, the current study demonstrates that it only takes three exposures for 11–14-year old children to
show brain responses to newly learned words that parallel those to known, real words. Conversely,
they do not show such brain responses for Not Learned Words. These findings support the use of
EEG as a non-invasive means of bridging the gap between brain and behavioral research to examine
the process of successful word learning from context in children.
Unlike other EEG word learning from context tasks used with adults, our study asked children to
verbally provide a meaning for the target word, if possible. Including this measure allowed us to (1)
examine whether children were able to learn the words in the experimental task and (2) focus EEG
analyses on those words that children did or did not learn. By classifying items as Learned Words or
Not Learned Words, we were able to investigate word learning from context, while also identifying
the process of building a semantic representation for a new word. In this way, this study is able to
explore learning at a deeper and subconscious level than what traditional behavioral methods offer.
This study supports the findings of Mestres-Missé et al. (2007) and Batterink and Neville (2011)
that the N400 response is sensitive to word learning from context and that words learned well
enough to support an explicit behavioral response, indicating learning, show different brain
responses than words not learned enough to support an explicit response. Our findings support
claims that the development of a semantic representation is dependent on high quality input, not just
exposure to the novel word. Specifically, the N400 did not attenuate for the words that were not
learned, despite having been presented to the participant three times. Most (82%) of the words
classified as Not Learned classification were from the No Meaning condition, which did not provide
Figure 4. ERPs across sentence presentations 1, 2 and 3 for Real Words. Real words (red) showed no significant amplitude
differences between 300-500 msec across sentence presentations 1, 2 and 3. The red boxes on the headmap denote the electrodes
plotted in this figure.
8A. D. ABEL ET AL.
Downloaded by [SDSU San Diego State University] at 10:43 29 November 2017

enough semantic information to support word meaning. Thus, we conclude that quality of input and
the ability to use that input to determine word meaning is likely a key component of successful word
learning. Additionally, the N400 response has been shown to attenuate in situations where the word
is easier to predict (due to high cloze probability; Federmeier & Kutas, 1999b; Federmeier,
McLennan, De Ochoa, & Kutas, 2002). In this study, changes for the Learned Words replicate
those findings, further implicating the importance of high constraining input in word learning. The
similarity between our results and those of studies with adults offers a preliminary suggestion that
school-aged children draw on similar neural networks to learn new words from context.
An alternative explanation for the finding that the N400 amplitude to the Not Learned Words
does not attenuate similar to that to the Learned Words is that the design did not fully capture
possible effects of repeated exposure. Recall that the majority (82%) of the words classified as Not
Learned were from the No Meaning condition. The No Meaning condition was designed such that
the target word differed across the three presentations and this variability in target word could
prevent the attachment of a meaning to the novel word. We consider this possibility unlikely because
the sentences in the No Meaning condition were all low cloze probability (M = 4.14%), making it
difficult to determine a meaning for the novel word in any of the sentences let alone determining
that the sentences each had a different target word. However, to address this possibility, a future
direction of this research will be to constrain further the No Meaning condition by using the same
target word across sentences. Additionally, it could be the case that many low cloze probability
sentences would lead to learning that looks similar to what we observe here (attenuation of the
N400). Given the difficulty of identifying the point of learning in so many trials, it was beyond the
scope of the current study but may be addressed in future projects.
Additional planned changes to the methodology include manipulating the number of sentences in
each condition, the amount of constraint in the sentences in the Meaning condition and using this
protocol to examine verb learning. We also plan to extend this line of research to include children
with language impairments and those raised in poverty.
The current study is the first to examine the neural mechanisms associated with successful word
learning from context in school-aged children. As mentioned, during the middle school years, the
focus of formal education begins to shift from vocabulary growth to literary analysis, leaving
children to learn new words primarily from the surrounding linguistic context. In this study, we
find that 11–14-year old children show brain responses to learned words that look like those of
known words after only three exposures to the new word when meaning is successfully identified.
Through this methodology we can identify word learning behaviorally and neurally in typically
developing school-aged children, which can build a foundation for studying vocabulary development
in younger children and atypical populations.
Declaration of interest
The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the article.
Funding
This work was supported by the National Science Foundation [BCS-1551770];
References
Atchley, R. A., Rice, M. L., Betz, S. K., Kwasny, K. M., Sereno, J. A., & Jongman, A. (2006). A comparison of semantic
and syntactic event related potentials generated by children and adults. Brain and Language,99, 236–246.
doi:10.1016/j.bandl.2005.08.005
Batterink, L., & Neville, H. (2011). Implicit and explicit mechanisms of word learning in a narrative context: An event-
related potential study. Journal of Cognitive Neuroscience,23, 3181–3196. doi:10.1162/jocn_a_00013
LANGUAGE LEARNING AND DEVELOPMENT 9
Downloaded by [SDSU San Diego State University] at 10:43 29 November 2017

Borovsky, A., Elman, J. L., & Fernald, A. (2012). Knowing a lot for one’s age: Vocabulary skill and not age is associated
with anticipatory incremental sentence interpretation in children and adults. Journal of Experimental Child
Psychology,112, 417–436. doi:10.1016/j.jecp.2012.01.005
Borovsky, A., Kutas, M., & Elman, J. (2010). Learning to use words: Event-related potentials index single-shot
contextual word learning. Cognition,116, 289–296. doi:10.1016/j.cognition.2010.05.004
Federmeier, K. D., & Kutas, M. (1999b). A rose by any other name: Long-term memory structure and sentence
processing. Journal of Memory and Language,41, 469–495. doi:10.1006/jmla.1999.2660
Federmeier, K. D., McLennan, D. B., De Ochoa, E., & Kutas, M. (2002). The impact of semantic memory organization
and sentence context information on spoken language processing by younger and older adults: An ERP study.
Psychophysiology,39(2), 133–146. doi:10.1111/1469-8986.3920133
Fenson, L., Marchman, V., Thal, D., Dale, P., Reznick, S., & Bates, E. (2006). The MacArthur communicative
development inventories: User’s guide and technical manual (2nd ed.). Baltimore, MD: Brookes.
Ferree, T. C. (2006). Spherical splines and average referencing in scalp electroencephalography. Brain Topography,19,
43–52. doi:10.1007/s10548-006-0011-0
Frishkoff, G. A., Collins-Thompson, K., Perfetti, C. A., & Callan, J. (2008). Measuring incremental changes in word
knowledge: Experimental validation and implications for learning and assessment. Behavior Research Methods,40,
907–925. doi:10.3758/BRM.40.4.907
Frishkoff, G. A., Perfetti, C. A., & Collins-Thompson, K. (2010). Lexical quality in the brain: ERP evidence for robust
word learning from context. Developmental Neuropsychology,35, 376–403. doi:10.1080/87565641.2010.480915
Fukkink, R. (2005). Deriving word meaning from written context: A process analysis. Learning and Instruction,15,23–
43. doi:10.1016/j.learninstruc.2004.12.002
Hart, B., & Risley, T. R. (1995). Meaningful differences in the everyday experience of young American children.
Baltimore, MD: Paul H Brookes.
Henderson, L. M., Baseler, H. A., Clarke, P. J., Watson, S., & Snowling, M. J. (2011). The N400 effect in children:
Relationships with comprehension, vocabulary and decoding. Brain and Language,117,88–99. doi:10.1016/j.
bandl.2010.12.003
Hoff, E. (2003). The specificity of environmental influence: Socioeconomic status affects early vocabulary development
via maternal speech. Child Development,74, 1368–1378. doi:10.1111/cdev.2003.74.issue-5
Junghöfer, M., Elbert, T., Tucker, D. M., & Braun, C. (1999). The polar average reference effect: A bias in estimating
the head surface integral in EEG recording. Clinical Neurophysiology,110, 1149–1155. doi:10.1016/S1388-2457(99)
00044-9
Kelley, J. G., Lesaux, N. K., Kieffer, M. J., & Faller, S. E. (2010). Effective academic vocabulary insturction in the urban
middle school. The Reading Teacher,64(1), 5–14. doi:10.1598/RT.64.1.1
King, J. W., & Kutas, M. (1995). Who did what and when? Using word-and clause-level ERPs to monitor working
memory usage in reading. Journal of Cognitive Neuroscience,7, 376–395. doi:10.1162/jocn.1995.7.3.376
Kutas, M., & Federmeier, K. D. (2011). Thirty years and counting: Finding meaning in the N400 component of the
event-related brain potential (ERP). Annual Review of Psychology,62, 621–647. doi:10.1146/annurev.
psych.093008.131123
McLaughlin, J., Osterhout, L., & Kim, A. (2004). Neural correlates of second-language word learning: Minimal
instruction produces rapid change. Nature Neuroscience,7, 703–704. doi:10.1038/nn1264
Mestres-Missé, A., Rodriguez-Fornells, A., & Münte, T. (2007). Watching the brain during meaning acquisition.
Cerebral Cortex,17, 1858–1866. doi:10.1093/cercor/bhl094
Neville, H. J., Mills, D. L., & Lawson, D. S. (1992). Fractionating language: Different neural subsystems with different
sensitive periods. Cerebral Cortex,2, 244–258. doi:10.1093/cercor/2.3.244
Nunez, P. L., & Srinivasan, R. (2005). Electric fields of the brain: The neurophysics of EEG. New York, NY: Oxford
University Press, USA.
Perfetti, C. A., Wlotko, E. W., & Hart, L. A. (2005). Word learning and individual differences in word learning
reflected in event-related potentials. Journal of Experimental Psychology: Learning, Memory, and Cognition,31,
1281–1292.
Perrin, F., Pernier, J., Bertrand, O., & Echallier, J. (1989). Spherical splines for scalp potential and current density
mapping. Electroencephalography and Clinical Neurophysiology,72, 184–187. doi:10.1016/0013-4694(89)90180-6
Pulvermüller, F., Lutzenberger, W., & Birbaumer, N. (1995). Electrocortical distinction of vocabulary types.
Electroencephalography and Clinical Neurophysiology,94(5), 357–370. doi:10.1016/0013-4694(94)00291-R
Srinivasan, R., Tucker, D. M., & Murias, M. (1998). Estimating the spatial Nyquist of the human EEG. Behavior
Research Methods, Instruments, & Computers,30,8–19. doi:10.3758/BF03209412
Steele, S. C., & Mills, M. T. (2011). Vocabulary intervention for school-age children with language impairment: A
review of evidence and good practice. Child Language Teaching and Therapy,27, 354–370. doi:10.1177/
0265659011412247
10 A. D. ABEL ET AL.
Downloaded by [SDSU San Diego State University] at 10:43 29 November 2017

Storkel, H. L. (2013). A corpus of conssonant–vowel–consonant real words and nonwords: Comparison of phonotactic
probability, neighborhood density, and consonant age of acquisition. Behavior Research Methods,45, 1159–1167.
doi:10.3758/s13428-012-0309-7
Swanborn, M. S., & De Glopper, K. (1999). Incidental word learning while reading: A meta-analysis. Review of
Educational Research,69, 261–285. doi:10.3102/00346543069003261
Torkildsen, J. V. K., Svangstu, J. M., Hansen, H. F., Smith, L., Simonsen, H. G., Moen, I., & Lindgren, M. (2008).
Productive vocabulary size predicts event-related potential correlates of fast mapping in 20-month-olds. Journal of
Cognitive Neuroscience,20, 1266–1282. doi:10.1162/jocn.2008.20087
LANGUAGE LEARNING AND DEVELOPMENT 11
Downloaded by [SDSU San Diego State University] at 10:43 29 November 2017