The Relationship Between Word Learning and Executive Function in Preschoolers With and Without Developmental Language Disorder

Abstract

Purpose

The purpose of this study was to examine the relationship between executive functioning and word learning among preschoolers with and without developmental language disorder (DLD).

Method

Forty-one preschool-age children with DLD were matched to typically developing children on age and sex. Participants were exposed to 10 novel pseudowords, half of which referred to familiar objects and half of which referred to unfamiliar objects. Their ability to produce, recognize, and comprehend the novel words was tested, and they completed executive function tasks measuring sustained selective attention, short-term memory, working memory, inhibition, and shifting.

Results

Preschoolers with DLD performed worse compared to typically developing peers on all measures of executive function and novel word learning. Both groups showed a relative weakness in producing pseudowords that corresponded with familiar objects versus pseudowords for unknown objects. Executive function accounted for statistically significant variance in word learning beyond group membership, with inhibition as a significant predictor of all word learning outcomes and short-term memory as a significant predictor of novel word comprehension. Executive function explained significant variance in novel word production and recognition even after accounting for variance explained by group differences in IQ and receptive vocabulary.

Conclusion

Findings replicate previous research reporting deficits in word learning and executive function in children with DLD, indicate that preschoolers are disadvantaged in learning new words for familiar objects, and support a relationship between executive function and word learning for children with and without DLD. Future research should examine the directionality of the relationship between these variables.

Compared to typically developing (TD) peers, children with developmental language disorder (DLD), which is also referred to as specific language impairment, are worse at word learning (Kan & Windsor, 2010). Although a word learning deficit has been established in this population, less is known about what factors may contribute to word learning deficits among children with DLD. Previous research has identified significant relationships between participants' word learning abilities and their receptive vocabulary, articulation abilities, nonverbal IQ, and phonological short-term memory (Alt & Plante, 2006; Chen & Liu, 2014; Gray, 2004), but additional cognitive and linguistic variables are also likely related. Among TD children, significant relationships have been identified between word learning abilities and executive function (Kapa & Colombo, 2014; Weiland et al., 2014; White et al., 2017). Although, as a group, children with DLD demonstrate executive function deficits (Kapa & Erikson, 2019) relative to typical peers, the potential contribution of executive function to word learning in this population remains relatively unexplored.

Word Learning in DLD

Compared to children with typical language abilities, children with DLD show deficits in learning new words. When presented with the same word learning context, children with DLD show reduced accuracy in producing novel words (Alt & Spaulding, 2011; Dollaghan, 1987; Gray, 2004, 2005, 2006; Nash & Donaldson, 2005; cf. Gray & Brinkley, 2011), comprehending novel words (Gray, 2003, 2004, 2005, 2006; Nash & Donaldson, 2005; Rice et al., 1990; cf. Chen & Liu, 2014), and recognizing the phonological form of novel words (Alt & Plante, 2006; Alt et al., 2004; Alt & Spaulding, 2011). Kan and Windsor (2010) confirmed these findings in a meta-analysis of word learning abilities in children with DLD, which supported that, across studies, outcomes were significantly poorer for participants with DLD relative to TD age-matched peers on all types of word learning outcome measures.

Beyond comparing word learning between children with and without DLD, previous studies have assessed whether participants' linguistic and/or cognitive abilities, including vocabulary, nonverbal cognition, and phonology/articulation, are significantly associated with their performance on word learning tasks. Reports are mixed regarding the relationship between children's vocabularies and their word learning abilities. Some research has found no significant relationship between standardized vocabulary performance and word learning (Gray, 2003, 2006; Nash & Donaldson, 2005; Rice et al., 1990). In contrast, other research reports that, among children with and without DLD, participants' vocabulary, as measured by standardized vocabulary tests, is significantly associated with novel word production (Chen & Liu, 2014; Gray, 2004), comprehension (Chen & Liu, 2014; Gray, 2004), and recognition (Alt et al., 2004).

Although less systematically studied than vocabulary, nonverbal IQ and phonological short-term memory (i.e., ability to store speech sound sequences in memory) are also associated with word learning. Gray (2004) reported that participants' nonverbal IQ positively predicted novel word production performance. Alt and Plante (2006) found that children's ability to recognize the correct phonological form of a target word during a fast mapping task was positively correlated with phonological short-term memory as assessed by performance on a nonword repetition task. A positive relationship between phonological short-term memory and fast mapping production has also been identified among TD children (Gray, 2006; see Gathercole, 2006, for a review).

Another potential factor contributing to differences in word learning between children with and without DLD may stem from differences in lexical constraints during word learning. Children with DLD may be less efficient word learners relative to TD peers, in part, because they do not use lexical constraints employed by TD children. For example, TD children rely on mutual exclusivity to assume that a novel word refers to an unknown referent (Merriman et al., 1989), which reduces the number of possible referents for a new word. However, preschool-age children with DLD rely less on mutual exclusivity to map novel labels to unknown objects relative to same-age peers with typical language and older children with DLD (Estis & Beverly, 2015). In contrast, Gray et al. (2013) reported that 3- to 5-year-olds with and without DLD showed an interaction effect between word properties and referent familiarity where children from both groups were best at learning novel words with low phonotactic probability paired with unfamiliar referents, which suggests that both groups benefited from assuming that a novel word labeled a novel object. If preschoolers with DLD do not demonstrate mutual exclusivity, then referent familiarity should not affect word learning, whereas if they do use mutual exclusivity, we would anticipate better learning outcomes for words with unfamiliar referents. Furthermore, the process of relabeling a known object may tax inhibition skills as learners inhibit retrieval of the conventional label in favor of the novel word (Gray, 2006).

Executive Function in DLD

In addition to difficulties with word learning and other aspects of language acquisition (e.g., syntax, morphology, phonology), children with DLD also have deficits in executive function (Kapa & Plante, 2015). Executive function refers to a set of cognitive processes that control and coordinate attention and memory, enabling higher-order cognition including reasoning, problem-solving, and planning (Diamond, 2013). Although executive function is related to IQ, they are not synonymous, as evidenced by relatively weak correlations between measures of the two constructs (Ardila et al., 2000).

Executive function comprises multiple skills, which have been described for both adults (Miyake et al., 2000) and preschool-age children (Garon et al., 2008) within an integrative framework in which executive function skills are related but distinct, resulting in three commonly identified executive function components: updating, inhibition, and shifting. Updating, which is also called working memory, refers to one's ability to manipulate information held in memory for a short time. Short-term memory is a related skill that refers to the storage of information over a brief time. Working memory tasks that require the storage and manipulation of information, such as a repetition task in which items must be reordered, require both short-term and working memory (Diamond, 2013). In contrast, tasks that require storage without manipulation, such as a rote repetition task, tax only short-term memory. Inhibition is controlling one's attention by ignoring distracting information (i.e., interference control) and/or withholding inappropriate responses (i.e., response inhibition; Diamond, 2013). Shifting is required to move attention between tasks, stimuli, or stimulus properties (Miyake et al., 2000). Additionally, although not included as a formal executive function component in most models, Garon et al. (2008) point out the necessity of sustained selective attention, which is the ability to focus attention on a specific task and/or stimulus across time, in order to employ any of the other, more complex executive function components.

Following integrative models of executive function, researchers have compared measures of different executive function components between children with and without DLD. Children with DLD perform worse compared to peers with typical language on measures of sustained selective attention (Ebert & Kohnert, 2011), short-term and working memory (Vugs et al., 2013), attention shifting, and inhibition (Pauls & Archibald, 2016). Although children with DLD have documented deficits in both executive function and language abilities, the relationship between these variables remains unclear. However, research with TD children provides evidence of a relationship between language and executive function that can inform predictions about how they are related in individuals with disordered language.

Executive Function and Language Development

Previous research with TD populations has linked executive function abilities with receptive language knowledge and syntax (Kaushanskaya et al., 2017), morphosyntax (Gandolfi & Viterbori, 2020), and language learning in a small artificial language (Kapa & Colombo, 2014) and in novel word learning tasks (White et al., 2017). Likewise, among children with DLD, significant correlations have been identified between executive function and vocabulary, language comprehension, and syntax (Kapa et al., 2017; Vugs et al., 2015).

Establishing a directional relationship between executive function and language abilities is a difficult task because both language and executive function emerge in infancy and develop simultaneously across childhood. Therefore, one skill does not clearly precede the other developmentally, which makes it difficult to ascertain whether the reported relationship between the two variables is due to executive function influencing language acquisition, language abilities influencing the development of executive function, or the two skills simply co-occurring. To address this issue of directionality in the relationship between executive function and receptive vocabulary abilities in TD preschoolers, Weiland et al. (2014) tested the skills at two time points and used structural equation modeling to measure the relationships between earlier and later executive function and receptive vocabulary. The predictive relationship between earlier executive function and later vocabulary was positive and significant, but earlier vocabulary did not predict later executive function abilities. Based on this structural equation modeling model, executive function abilities contribute to vocabulary development, but vocabulary skills do not affect the development of executive function. One possible mechanism to account for the directional relationship between executive function and vocabulary is a cognitive model purporting that executive function processes are needed for controlling actions and thoughts, which is necessary for learning (Diamond, 2013).

An additional factor that contributes to the difficulty in establishing a directional relationship between executive function and language development is the presence of co-occurring variables, such as socioeconomic status (SES), which reportedly affect both executive function and language development. Fernald et al. (2013) reported differences in language processing between toddlers from families with high versus low SES. Likewise, researchers have reported a positive relationship between SES and executive function development (Noble et al., 2007). Thus, in order to ensure that any relationships found between executive function and language development are not actually SES effects, this variable must be matched between participant groups and/or addressed in analyses.

Current Study

The purpose of the current study was to examine the relationship between executive function components and word learning among preschoolers with DLD and those with typical language abilities. Previous research has established deficits in both word learning (Kan & Windsor, 2010) and executive function (Kapa et al., 2017; Vissers et al., 2015) in preschool-age children with DLD. Based on a model from TD preschoolers that supports the role of executive function abilities in vocabulary development (Weiland et el., 2014), we asked whether group differences in executive function between preschoolers with and without DLD are related to differences in word learning. We expected executive function measures to be positive predictors of word learning performance in both groups of children based on regression analyses. Because researchers have reported that nonverbal IQ and vocabulary abilities are related to word learning outcomes among children with DLD (Chen & Liu, 2014; Gray, 2004; Kan & Windsor, 2010), and IQ and receptive vocabulary differed between our DLD and TD groups, we also asked whether executive function performance explains unique variance in word learning outcomes after controlling for these potentially confounding factors.

Finally, we examined the effects of referent familiarity on word learning, as this may differ between groups due to differences in inhibition. To test the effects of referent familiarity, we paired half of the pseudowords with unfamiliar referents and half with familiar referents (e.g., ball, duck). We hypothesized that children with DLD may show particular difficulty learning novel words with familiar referents as this may especially tax inhibition and/or cognitive flexibility for children to avoid retrieval of the previously learned label. Gray (2006) noted that the real-world process of word learning includes learning both novel labels for unfamiliar referents and novel labels for familiar referents as children learn hyponyms and synonyms. However, it was also possible that children with DLD would show no effect of referent familiarity if they did not rely on mutual exclusivity in word learning and therefore did not differentiate between learning labels for familiar versus unfamiliar referents.

In sum, we addressed the following research questions:

1. Does executive function predict significant variance in novel word learning for preschoolers with and without DLD?

2. Does executive function remain a significant predictor of word learning after accounting for participants' nonverbal IQ and receptive vocabulary performance?

3. Do group differences in novel word learning between children with and without DLD vary as a function of referent familiarity?

Although the current study design does not allow us to draw conclusions about the directional relationship between executive function and word learning, it addresses a more basic question about whether or not executive function and word learning are related skills in young children with DLD, which has not been clearly established. Documenting the relationship between executive function and word learning improves our understanding of the nature of DLD and provides an important foundation for future treatment research targeting word learning in children with language and executive function deficits.

Method

Participants

The current study included eighty-two 4- and 5-year-old participants who all passed a binaural pure-tone hearing screening (1000, 2000, and 4000 Hz at 20 dB), received a standard score of 75 or above on a nonverbal IQ measure, and were not diagnosed with other developmental disorders (e.g., attention-deficit/hyperactivity disorder, autism spectrum disorder, Down syndrome). Participants completed a battery of standardized language and cognitive measures in order to determine eligibility and to classify each participant in either the DLD or TD group. Select demographic information (e.g., age, maternal education, home languages) was collected via a brief parent questionnaire that was distributed as part of the study information/consent packet.

Half of the participants (15 female, 26 male) were classified as having DLD based on receiving a standard score below 87 ¹ on the Structured Photographic Expressive Language Test–Preschool: Second Edition (SPELT-P:2; Dawson et al., 2005). The other 41 participants (15 female, 26 male) had typical language abilities. Each child in the typical language group was matched to a DLD group member on age (± 3 months) and sex, received a standard score above 87 on the SPELT-P:2, had no reported parent or teacher concerns about his or her language abilities, and had no history of speech or language therapy. An additional 24 participants with typical language completed the study procedures but were not included in our final sample because they were not an age/sex match for any of the participants with DLD (n = 23) or because they did not complete both days of the word learning paradigm (n = 1). All participants interacted via standardized testing and conversation with a certified speech-language pathologist who was part of our research team and confirmed that her professional opinion of the participants' language abilities matched their group classification based on their standardized SPELT-P:2 scores. In cases of a discrepancy between a potential participant's SPELT-P:2 score and the clinician's impression, the child was excluded from the study sample (n = 3).

Nonverbal IQ was tested using the nonverbal index of the Kaufman Assessment Battery for Children–Second Edition (KABC-II; Kaufman & Kaufman, 2004), articulation was assessed using the Goldman-Fristoe Test of Articulation–Second Edition (Goldman & Fristoe, 2000), and expressive and receptive language abilities were assessed via the SPELT–P:2 and the Peabody Picture Vocabulary Test–Fourth Edition (PPVT-4; Dunn & Dunn, 2007).

Sixteen children were reported by a parent to live in a home where another language was spoken in addition to English (13 Spanish, one Arabic, one Lao, and one Turkish), but the amount of exposure and the child's proficiency in the other language were not documented. Of these children whose families spoke a language in addition to English, 12 were in the DLD group and four were in the TD group. All children spoke only English as a native language according to parent report and attended English-speaking preschools or day cares, where they were exposed to Mainstream American English during the day. All testing procedures for this study were conducted in English.

Table 1 summarizes the demographic, linguistic, cognitive, and executive function performance of participants. Mean comparisons with Bonferroni corrections for multiple comparisons (adjusted α = .008) revealed that the DLD and TD groups were equivalent on age, t(80) = −.146, p = .1, d = 0.03, and maternal education (proxy for SES), t(78) = −2.25, p = .01, d = 0.56. The TD group's mean score was higher on SPELT-P:2, t(80) = −18.36, p < .001, d = 4.01; KABC-II, t(80) = −4.68, p < .001, d = 1.03; PPVT-4, t(80) = −6.0, p < .001, d = 1.33; and Goldman-Fristoe Test of Articulation–Second Edition, t(80) = −9.93, p < .001, d = 2.22.

Table 1.

Demographic, standardized test, and executive function task means (standard deviations) and ranges.

Variable	DLD	TD
Age (months)	59.44 (5.22) 50–68	59.61 (5.38) 48–70
Maternal education a	14.15 (1.58) 12–17	15.10 (1.76) 10–17
KABC-II**	98.85 (12.22) 79–130	110.56 (10.34) 81–130
SPELT-P:2**	67.59 (10.34) 45–85	109.17 (10.17) 89–140
PPVT-4**	91.71 (12.38) 62–118	109.59 (14.52) 85–130
GFTA-2**	77.20 (15.41) 42–109	104.68 (8.74) 83–117
Forward word span b , *	2.98 (1.31) 0–8	4.02 (0.22) 2–8
Backward word span b , *	0.66 (0.18) 0–4	1.9 (0.18) 0–4
Day–night c *	0.54 (0.03) .03–.97	0.73 (0.05) 0–1.0
Selective sustained attention c , *	0.88 (0.03) 0.37–1.0	0.96 (0.02) 0.37–1.0
Dimensional Change Card Sort task b , *	8.58 (0.45) 6–12	11.63 (0.18) 6–12

Note. DLD = developmental language disorder; TD = typically developing; KABC-II = Kaufman Assessment Battery for Children–Second Edition; SPELT-P:2 = Structured Photographic Expressive Language Test–Preschool: Second Edition; PPVT-4 = Peabody Picture Vocabulary Test–Fourth Edition; GFTA-2 = Goldman-Fristoe Test of Articulation–Second Edition.

^a Reported in years of education between 8 and 16.

^b Number correct with a minimum possible score of 0 and a maximum score of 12.

^c Percentage of accuracy.

^* p ≤ .01.

^** p < .008 (Bonferroni-adjusted alpha for multiple comparisons).

Procedure and Measures

All study procedures and materials were reviewed by The University of Arizona Human Subjects Protection Program and approved by the institutional review board. Participants' parents/guardians provided written informed consent for participation before any research activities began. All participants in the TD group and nine participants in the DLD group were recruited and tested in their preschools and day cares. The remaining 32 participants with DLD were recruited and tested at a summer program for children with language impairments conducted by The University of Arizona. Each participant was tested individually over approximately ten 30-min sessions that occurred over approximately 2 weeks. Testing occurred 5 days a week, and participants continued to complete tests until they had finished the battery. Some children required longer than 2 weeks to complete the 10 sessions due to absences. The 2-day word learning tasks were presented in a fixed order with Day 1 procedures preceding Day 2, but other tasks were completed in a pseudorandom order such that participants continued completing one or two tasks per day until they had completed all study-related testing. The order of tasks varied between children and was contingent on which tasks a child had already completed and which tasks were available for use at the time when that particular participant was being tested (i.e., if both copies of a standardized test were in use, the child did not complete that measure at that time).

Executive Function Tasks

Following integrative framework models of executive function (Garon et al., 2008; Miyake et al., 2000), multiple executive function tasks were employed to test the executive function components of short-term memory, working memory (updating), inhibition, and shifting as well as sustained selective attention. Although tasks were selected to tax certain executive function components and are categorized here for the purpose of description based on the executive component they are designed to measure, it is likely the case that these measures required children to use multiple executive function skills (i.e., task impurity). The executive function measures used here are generally categorized as appropriate for assessing the targeted skills in preschoolers (Carlson, 2005; Garon et al., 2008) and have been successfully modified for use with preschoolers with DLD (Kapa et al., 2017).

Sustained selective attention was measured using a computerized continuous performance task (CPT) in which children responded with a button press to a recurring target phrase (“It's a frog.”) and withheld responses to nontarget phrases (e.g., “It's a rock.”). Participants were introduced to the target phrase during the task instructions and 15 practice trials that included 12 target phrase trials. Throughout the task, participants received positive reinforcement for responses to the target phrase (a short animation) and saw a red X if they pressed the button after hearing a nontarget phrase. Four “reinforcers” were presented throughout the task, during which participants saw an animation to keep them engaged and were reminded of the task goals and target phrase. The same target phrase was used throughout the task, whereas unique nontarget phrases that varied in their final word were presented for each trial on which participants should withhold responding. Thus, in order to succeed on the task, participants had to maintain attention to the target stimulus (selective attention) across the duration of the task (sustained attention). The task consisted of 60 test trials (20 target items and 40 nontargets) in random order. Attentional demands were increased by the inclusion of a white noise mask that degraded the sound signal for half of all target and nontarget items (see Spaulding et al., 2008, for details). Participants' overall proportion of correct responses was included in analyses with scores ranging from 0 to 1.

Short-term memory was assessed using a forward word span task, and working memory was assessed using a reverse word span task. Participants repeated lists of common, monosyllabic English words (e.g., cat, sun, spoon) after the experimenter, in the same (e.g., cat, sun, spoon) or reverse (e.g., spoon, sun, cat) order, depending on the task (i.e., forward or reverse word span). Words included in the span tasks were expected to be familiar to young children based on the relatively young age of acquisition (AoA), high imageability, and high frequency of the words. The average AoA was 3.73 years (Kuperman et al., 2012). The average word imageability rating from a scale of 1 (low) to 7 (high) was 6.57 (Cortese & Fugett, 2004). Finally, the average log₁₀ frequency of the span words was 3.19 (from a total child corpus of 1,028,417 English tokens), which is higher than the corpus average of 2.04 (Storkel & Hoover, 2010).

In the forward span task, children were told to repeat the words exactly as the examiner had said them, after the examiner finished speaking. For the reverse span, children were told to follow the same instructions but to repeat the words “the silly, backwards way” (i.e., in reverse order). In both span tasks, the words were presented to the participants at a rate of one word per second. The forward span task only required storage and repetition of words, which requires short-term memory. The reverse span task also requires manipulation of the word lists to reverse the order before repetition, so this is an index of working memory.

Prior to test trials, participants were trained to complete each version of the task by completing six practice trials. Half of the practice trials included picture cards that illustrated the words and order of representation to reduce memory demands and visually demonstrate the forward or reverse word order. During the practice trials, children received feedback from the examiner, and practice trials were repeated if needed. After practice, testing began with two 2-word lists and continued with two lists at each subsequent length (three-word, four-word, etc.) until participants made errors on both lists of the same length or until they completed the final, six-word list. Participants completed the forward span task, immediately followed by the reverse span task. The forward span score (i.e., short-term memory) was the number of lists correctly repeated in the same order presented by the experimenter, and the reverse span score (i.e., working memory) was the number of lists correctly repeated in reverse order. Scores for the forward and reverse span tasks ranged from a minimum of 0 to a maximum score of 12.

We measured inhibition using a computerized test based on the day–night task (Diamond et al., 2002). Children heard the words “day” and “night” and pressed a button labeled with the opposite image (i.e., moon or sun). The task included 30 trials, half “day” and half “night.” Participants were instructed to respond as quickly as possible while maintaining accuracy. The trials were presented in a random order, with the limitation of no more than four instances of the same item (day or night) presented in succession. Participants' proportion of correct responses (range: 0–1) was calculated and included in analyses.

Attentional shifting abilities were tested via the Dimensional Change Card Sort (DCCS) task (see Zelazo, 2006, for details). Two target boxes were provided with bivalent picture cards attached (e.g., a green boat and a yellow dog). Participants sorted test cards with the same features (e.g., green dogs and yellow boats) into the boxes. During the preswitch phase, participants sorted based on one dimension (e.g., shape) and, after six cards, switched to sort based on the other dimension (e.g., color) during the postswitch sorting phase. The total number of cards correctly sorted (maximum = 12) served as participants' DCCS score.

Word Learning Paradigm

Participants were familiarized with and tested on 10 novel pseudoword nouns, half of which referred to familiar and namable objects (e.g., ball) and half of which referred to objects without known labels from the Novel Object and Unusual Name Database (Horst & Hout, 2016). We did not explicitly test whether participants knew the labels for the familiar objects because doing so may reinforce the known label within the study context and make the child's task of subsequently relabeling the object with a pseudoword even more challenging. However, the familiar object labels were all monosyllabic (e.g., kite, block, duck, car, ball), early acquired (average AoA = 3.83; Kuperman et al., 2012), highly imageable (average imageability rating = 6.44; Cortese & Fugett, 2004), and frequently produced by children based on child corpus analysis (average log₁₀ frequency = 3.35; Storkel & Hoover, 2010). Therefore, it is likely that all of the children knew the familiar object labels and were engaged in the task of relabeling the objects during the pseudoword learning task. Familiar objects were further selected to have similar perceptual features as the unknown objects, which were all colorful toys.

All novel word stimuli contained English consonants that are acquired relatively early (/k/, /b/, /g/, /p/, /w/, /t/, /m/, /n/, and /d/), with a median age of customary articulation of 24 months or younger (Sander, 1972). Pseudoword stimuli were derived from previous child word learning studies (Alt & Plante, 2006; Alt & Suddarth, 2012; Weismer & Hesketh, 1998; Kan & Kohnert, 2008). See Table 2 for a list of words and referents.

Table 2.

Pseudowords transcribed in International Phonetic Alphabet (IPA) and corresponding referents.

Pseudoword (IPA)	Referent
/bæbɪn/	duck
/bɪm/	kite
/gɪp/	block
/kub/	car
/nokɛn/	ball
/kɪdɪt/	Unfamiliar
/mæbɛp/	Unfamiliar
/pɛb/	Unfamiliar
/tæm/	Unfamiliar
/wæb/	Unfamiliar

The word learning paradigm was conducted across 2 consecutive days. On Day 1, participants were introduced to the stimulus words and their referents by repeating each word after an experimenter while viewing a corresponding picture. During this time, participants' repetitions were audio-recorded to capture individuals' systematic articulation errors for subsequent scoring. This ensured that children who produced articulation errors during this phase (e.g., repeated /kub/ as /tub/) would not have this same error counted against them during word learning testing. After initial exposure, participants viewed a familiarization video in which they were exposed to each word-referent pairing 10 times in a random order. During familiarization, words were embedded as the final word in phrases (e.g., “Where's the tam?” and “I like the peb.”) that were recorded by five female speakers of American English. Speaker variability was used to enhance novel word production outcomes (Richtsmeier et al., 2009). Each phrase was recorded and presented in its entirety to maintain natural speaking rate, prosody, and coarticulation.

Expressive knowledge of the novel words was assessed via a production task in which they were shown pictures of each referent and asked to provide the new label for each. Participants completed 10 trials, and their responses were written by the experimenter and audio-recorded for reliability scoring. Responses were scored as correct if they contained one or zero phoneme errors (deletion, addition, substitution) and incorrect if the participant provided no response or if they included more than one phoneme error. Children's responses were scored based on their productions from the initial exposure phase (e.g., if a child had repeated /kub/ as [tup] during initial exposure, a response of [tup] during the production test would be scored as correct). If participants produced the English label for a familiar object (e.g., “kite”), they were reminded to provide the new label, and their subsequent response was scored.

Participants' ability to recognize the correct phonological forms of the novel words was assessed using a three-choice test (Gordon & McGregor, 2014) in which participants were shown a picture of a referent and given a choice of three possible labels. The labels included the target (e.g., /pɛb/), a word that differed from the target by one phoneme (e.g., /pæb/), and a word with no shared phonemes (e.g., /blɪk/). Each label was said aloud as the experimenter touched one of three identical dots in a row. Participants responded by selecting the dot corresponding with the label they perceived as accurate. If participants provided a verbal response of the target or foil word that was incongruent with the dot that was selected, their verbal response was scored. For example, if the child produced the correct target word while choosing the wrong dot, the response was correct. In contrast, if a participant produced an incorrect foil while touching the correct dot, the response was scored as incorrect. The task included 10 trials, and responses were scored as correct (target) or incorrect (nontargets).

Novel word comprehension was assessed via a picture selection task in which participants heard each word and chose which of four pictures represented the referent. The visual stimuli included the target, two foils that were other named objects in the task, and a third novel foil image that was not included in familiarization.

Day 2 of the word learning paradigm began with the comprehension picture selection task to assess retention of learning from Day 1. Afterward, participants viewed the familiarization video and completed the production test and phonological recognition test from following the same procedures described above for Day 1. The Day 2 comprehension task differed from Day 1 in that it included two trials for each pseudoword—one with the same image used in familiarization and one with the item in a different color. The total number of trials in this task was 20. Neither group differed in their responses to trained versus novel versions of the referents, so we only included the 10 trained referent responses in our analyses here in order to make a clearer comparison between receptive knowledge on Day 1 and Day 2.

Participants' expressive, recognition, and receptive knowledge of the pseudowords was tested in a fixed order with participants completing the production test first, followed by the phonological recognition task second, and then the comprehension test. The order of task presentation was fixed in order to hold the number of times participants heard each pseudoword constant. Following this procedure, participants had heard each word 11 times (i.e., once during repetition and 10 times during familiarization) when they completed the production and recognition tasks on Day 1. After hearing each word said again during the recognition task, each word had been presented 12 times when the participants completed the comprehension measure where they heard each word a 13th time. On Day 2, participants had heard each word 24 times when completing the production and recognition tasks (i.e., 13 presentations on Day 1, one presentation in the retention comprehension task, and 10 times during familiarization). Finally, after the recognition task on Day 2, participants had heard each word 25 times before completing the last comprehension test, during which participants heard each word an additional 2 times.

Scoring Reliability

Production task data from 39% of tests (eight TD Day 1, eight TD Day 2, eight DLD Day 1, eight DLD Day 2) were scored offline from audio recordings by a rater (the second author) who neither collected nor scored the online data. Scoring agreement ranged from 90% to 100%, with an average interrater agreement of 96.6%.

Analyses

A mixed analysis of variance (ANOVA) was conducted to compare word learning performance on the three outcome measures (production, comprehension, and phonological recognition) between the two groups across the 2 days of testing in order to confirm group differences in word learning and to determine whether word learning scores could be collapsed across day and/or task in our regression analyses. A mixed ANOVA was also used to test whether referent familiarity affected word learning for either of the participant groups. Subsequent t tests were used to probe main effects and interactions. Before running the repeated-measures ANOVA, we confirmed normal distribution for each combination of factors (Group × Day × Task) via visual inspection of histogram plots and acceptable skew and kurtosis values. Homogeneity of variance for all word learning variables was established through inspection of residual scatter plots.

Hierarchical linear regression analyses were used to assess whether executive function measures accounted for a significant amount of variance in participants' word learning outcome scores. Although Weiland et al. (2014) created a unitary, composite executive function variable in their analyses, we did not do so in our analyses. The theoretical model of a unitary executive function in early childhood has been empirically supported for TD children (e.g., Wiebe et al., 2008) but has not been tested in children with DLD. Furthermore, in the current data, executive function measures were significantly positively correlated with each other for children with typical language, but there were no significant correlations between the measures in our group of children with DLD, which suggests that combining the measures into a single variable is inappropriate in this case. Instead, each executive function component was included in our regression analyses as a separate predictor.

Before running regression models, scatter plots were visually inspected to confirm linear relationships between the dependent and independent variables. To test the assumption of constant variance, residuals were plotted against predicted dependent variable values and visually inspected. The variance inflation factor was calculated for each continuous predictor variable in the model to determine the extent to which multicollinearity among predictors inflated the variances of the regression coefficients in the model. Based on a maximum value of 10, a widely accepted cutoff for variance inflation factor (Kutner et al., 2004), none of our predictors had problematic multicollinearity.

Results

Group Comparisons

Nonparametric Mann–Whitney U tests were used for independent group comparisons of executive function measures, which were nonnormally distributed. Bonferroni corrections were made for multiple comparisons (adjusted α = .01). The TD group outperformed the DLD group on every task: CPT measure of sustained selection attention, U = 579.0, p = .01; forward word span measure of short-term memory, U = 465.5, p < .001; reverse word span measure of working memory, U = 369.0, p < .001; day–night measure of inhibition, U = 468.0, p = .001; and DCCS measure of shifting, U = 386, p < .001 (see Table 1).

To confirm the expected pattern of a word learning advantage for TD children and to determine whether testing days and/or tasks could be combined in outcome in our regression analyses, a Group (DLD, TD) × Task (production, comprehension, recognition) × Day (Day 1, Day 2) mixed ANOVA was used to analyze between- and within-group effects on the word learning task (see Table 3).

Table 3.

Results of Group × Task × Day mixed analysis of variance.

Variable	F	df	p	ηp 2
Group	28.63	1, 80	.000	.264
Task	350.03	1, 80	.000	.814
Day	76.50	1, 80	.000	.395
Group × Task	6.58	2, 80	.002	.076
Group × Day	1.59	1, 80	.300	.013
Task × Day	1.37	2, 80	.395	.012
Group × Task × Day	4.43	2, 80	.051	.036

Based on paired-samples t tests, scores were significantly higher on Day 2 tests compared to Day 1 tests, t(81) = 7.22, p < .001, d = 0.80. However, the interaction between group and day was not significant, indicating that the difference between groups did not increase with time. This allowed us to combine participants' scores from Day 1 and Day 2 in our regression analyses.

The main effect of task results from significantly lower production task scores compared to both comprehension, t(81) = −18.54, p < .001, d = 2.05, and recognition, t(81) = −25.83, p < .001, d = 2.85, task scores. Comprehension scores were significantly lower than recognition scores, t(81) = −4.64, p < .001, d = 0.51. Figure 1 displays the mean number of items correct for the two groups on each word learning task across the 2 days. Because task scores significantly differed, we used separate regression analyses to predict performance on each task as a separate outcome.

Figure 1.

Mean items correct for participants with DLD versus TD participants on novel word learning production, recognition, and comprehension tasks on Days 1 and 2. Bars represent standard error of the mean. DLD = developmental language disorder; TD = typically developing.

The difference between word learning task scores is not surprising given the variability in chance level performance between the comprehension (chance = 0.25 or 5/20), phonological recognition (chance = 0.33 or 6.7/20), and production (no chance level) tasks. DLD group means (standard deviations) were 7.76 (3.70) for the receptive task and 10.22 (3.41) for phonological recognition. TD group means (standard deviations) were 12.44 (4.54) for the receptive task and 13.61 (4.12) for phonological recognition. Both the DLD and TD groups exceeded chance performance on the comprehension and phonological recognition tests (all ts > 4, ps < .001).

In addition to a significant main effect of task, the Group × Task interaction was also significant. This interaction was probed and indicated that the groups differed significantly on all word learning outcome measures, but the largest group differences were on the comprehension task, t(81) = −5.12, p < .001, d = 1.13, followed by the production task, t(81) = −4.43, p < .001, d = 0.98, and then the recognition task, t(81) = −4.06, p < .001, d = 0.90.

Relations Between Variables

Because there were no significant Task × Day or Group × Day interactions in the mixed ANOVA, participants' scores were summed from Day 1 and Day 2 to create single scores for production, phonological recognition, and comprehension, which were then used for subsequent correlational and regression analyses. Spearman correlations were calculated separately for the DLD and TD groups between participants' word learning outcomes and their maternal education, KABC-II, PPVT-4, SPELT-P2, CPT, forward and reverse word span, day–night, and DCCS performance (see Tables 4 and 5).

Table 4.

Spearman correlations between demographics, standardized tests, executive function tasks, and word learning tasks for the group with developmental language disorder.

	Variable	1	2	3	4	5	6	7	8	9	10	11
1.	Maternal education
2.	KABC-II	.26
3.	SPELT-P:2	−.01	.13
4.	PPVT-4	.14	.35*	.43**
5.	CPT	.28	.04	.45**	.29
6.	Forward word span	.13	.34*	.09	.11	.16
7.	Reverse word span	−.05	.29	−.08	.08	.18	.29
8.	Day–night	.16	.29	.26	.39*	.06	.03	.03
9.	DCCS	.12	.41**	.37*	.31	.20	.09	.31	.20
10	Production task	.18	.17	−.13	−.06	−.01	.25	−.11	.23	−.07
11.	Phonological recognition task	−.08	.30	−.13	.20	.12	.11	.05	.29	.21	.45**
12.	Comprehension task	.09	.25	.20	.16	.24	−.002	−.15	.14	.15	.22	.43**

Note. KABC-II = Kaufman Assessment Battery for Children–Second Edition; SPELT-P:2 = Structured Photographic Expressive Language Test–Preschool: Second Edition; PPVT-4 = Peabody Picture Vocabulary Test–Fourth Edition; CPT = continuous performance task; DCCS = Dimensional Change Card Sort.

^* p < .05.

^** p < .01.

Table 5.

Spearman product correlations between demographics, standardized tests, executive function tasks, and word learning tasks for the typically developing group.

	Variable	1	2	3	4	5	6	7	8	9	10	11
1.	Maternal education
2.	KABC-II	.35*
3.	SPELT-P:2	.26	.41**
4.	PPVT-4	.33*	.46**	.43**
5.	CPT	.14	.19	.24	.03
6.	Forward word span	.22	.49**	.51**	.44**	.02
7.	Reverse word span	.25	.44**	.46**	.33*	.49**	.34*
8.	Day–night	.45**	.45*	.33*	.61**	.16	.39*	.49**
9.	DCCS	.11	.03	−.09	.27	.12	−.22	.11	.22
10.	Production task	.02	−.02	.43**	.22	.10	.31*	.23	.29	.13
11.	Phonological recognition task	.34*	.35*	.51*	.50**	.24	.53**	.41**	.55**	.06	.59**
12.	Comprehension task	.21	.29	.45**	.41**	.11	.45**	.26	.40*	.24	.69**	.71**

Note. KABC-II = Kaufman Assessment Battery for Children–Second Edition; SPELT-P:2 = Structured Photographic Expressive Language Test–Preschool: Second Edition; PPVT-4 = Peabody Picture Vocabulary Test–Fourth Edition; CPT = continuous performance task; DCCS = Dimensional Change Card Sort.

^* p < .05.

^** p < .01.

Among the participants with DLD, significant positive correlations were found between the novel word phonological recognition task and the production and comprehension tasks, but no significant relations were found between word learning and demographic, standardized, or executive function variables. In contrast, in the TD group, the production task was positively correlated with SPELT-P:2 and forward word span scores, phonological recognition performance was positively correlated with all variables except CPT and DCCS, and comprehension significantly correlated with SPELT-P:2, PPVT-4, forward word span, and day–night. The word learning measures were also significantly positively correlated with each other.

A series of hierarchical linear regression analyses were used to identify variables that account for variance in word learning outcomes. Predictors of theoretical interest included the five executive function measures, nonverbal IQ (KABC-II), receptive vocabulary (PPVT-4), and SES (maternal education). However, maternal education did not significantly differ between the two groups, and DCCS and CPT were not correlated with word learning outcomes for either group, so they were not included as predictors in the regression models.

Model 1 includes group (DLD vs. TD) as the predictor of word learning production, phonological recognition, and comprehension. In each case, group membership accounts for a significant amount of variance in word learning outcomes (see Table 6). Model 2 includes group in Step 1 and three executive function measures—forward word span, backward word span, and day–night—in Step 2. The addition of the executive function predictors significantly increased the R ² value for all word learning measures, indicating that executive function explains additional variance above what can be accounted for by language group alone. Day–night was a significant predictor of all word learning outcomes, and forward word span also accounted for a significant amount of variance in novel word comprehension scores. The analyses in Model 2 were also conducted including Group × Executive Function interaction terms to assess whether the relationship between executive function and word learning differed as a function of language status. The interaction terms were nonsignificant in all models, and their inclusion did not significantly increase R ² .

Table 6.

Regression models with executive function tasks as predictors of each word learning outcome measure.

Model 1	Word learning outcome measure
	Production				Phonological recognition				Comprehension
		ΔR 2	R 2	p		ΔR 2	R 2	p		ΔR2	R2	p
Predictors		.197	.197	.000		.171	.171	.000		.247	.247	.000
	B	β	t	p	B	β	t	p	B	β	t	p	VIF
Group	−1.98	-.44	−4.43	.000	−3.39	−.41	−4.06	.000	−4.68	−.50	−5.12	.000	1.00
Model 2		ΔR2	R2	p		ΔR2	R2	p		ΔR2	R2	p
Predictors		.098	.296	.017		.187	.358	.000		.104	.350	.009
	B	β	t	p	B	β	t	p	B	β	t	p	VIF
Group	−1.39	−.31	−2.78	.007	−1.57	−.19	−1.79	.077	−3.48	−.37	−3.42	.001	1.37
Forward word	0.26	.17	1.57	.121	0.38	.13	1.27	.208	0.79	.24	2.31	.023	1.32
Backward word	−0.06	−.04	−0.31	.761	0.36	.12	1.01	.316	−0.27	−.07	−0.64	.523	1.56
Day–night	2.07	.27	2.52	.014	5.13	.36	3.56	.001	3.74	.23	2.24	.028	1.23

Note. VIF = variance inflation factor.

Model 3 included group in Step 1 and KABC-II and PPVT-4 as predictors in Step 2, resulting in a significant increase in R ² for word learning comprehension and phonological recognition compared to Model 1. PPVT-4 was a significant predictor of phonological recognition and comprehension. KABC-II was not a significant predictor for any of the word learning outcomes.

Finally, Model 4 included group in Step 1, KABC-II and PPVT-4 in Step 2, and the executive function variables in Step 3. Compared to Model 3 regressions, the inclusion of executive function significantly increased R ² for the expressive and phonological recognition tasks. In both cases, day–night remained a significant predictor of task performance after accounting for group membership, nonverbal IQ, and receptive vocabulary (see Table 7 for Models 3 and 4).

Table 7.

Regression models with group, nonverbal IQ, receptive vocabulary, and executive function tasks as predictors of each word learning outcome measure.

Model 3	Word learning outcome measure
	Production				Phonological recognition				Comprehension
		ΔR2	R2	p		ΔR2	R2	p		ΔR2	R2	p
Predictors		.02	.218	366		.12	.291	.002		.088	.334	.008
	B	β	t	p	B	β	t	p	B	β	t	p	VIF
Group	−1.52	−.34	−2.76	.007	−1.33	−.16	−1.38	.17	−2.66	−.28	−2.48	.02	1.52
KABC-II	0.01	.06	0.44	.66	0.05	.17	1.40	.17	0.05	.14	1.19	.24	1.53
PPVT-4	0.02	.14	1.06	.30	0.08	.31	2.49	.02	0.08	.27	2.22	.03	1.75
Model 4		ΔR2	R2	p		ΔR2	R2	p		ΔR2	R2	p
Predictors		.08	.298	.04		.08	.371	.03		.04	.374	.20
	B	β	t	p	B	β	t	p	B	β	t	p	VIF
Group	−1.48	−.33	−2.67	.009	−1.07	−.13	−1.10	.27	−2.74	−.29	−2.48	.02	1.65
KABC-II	−0.008	−.04	−0.32	.75	0.01	.04	0.33	.74	0.04	.09	0.75	.46	1.91
PPVT-4	−0.005	−.04	−0.25	.80	0.04	.15	1.11	.27	0.05	.17	1.25	.21	2.17
Forward word	0.29	.19	1.62	.11	0.28	.10	0.90	.37	0.63	.20	1.78	.08	1.44
Backward word	−0.04	−.02	−0.18	.86	0.32	.10	0.54	.40	−0.37	−.10	−0.85	.40	1.74
Day–night	2.24	.29	2.39	.02	4.17	.29	2.55	.01	2.34	.14	1.25	.22	1.57

Note. VIF = variance inflation factor; KABC-II = Kaufman Assessment Battery for Children–Second Edition; PPVT-4 = Peabody Picture Vocabulary Test–Fourth Edition.

Word Referent Properties

To address our question regarding the role of referent familiarity in word learning, a Group (DLD, TD) × Task (production, comprehension, recognition) × Referent Familiarity (familiar, unfamiliar) repeated-measures ANOVA was conducted. The results indicated that there was a significant main effect of referent familiarity, F(1, 80) = 16.67, p < .001, η_p ² = .17, with higher test scores for words with unfamiliar referents relative to words with familiar referents. A significant Task × Referent Familiarity interaction, F(2, 80) = 13.97, p < .001, η_p ² = .15, was driven by the fact that significant familiarity effects were limited to the production task, t(81) = −7.60, p < .001. However, there were no interactions between group and familiarity, which indicates that participants with and without DLD were less accurate in producing pseudowords with familiar referents.

Discussion

The first goal of this research was to assess the relationship between executive function and word learning among young participants with and without DLD. This work was motivated by previous research reporting word learning (Kan & Windsor, 2010) and executive function deficits (Vissers et al., 2015) among young children with DLD and the established relationship between executive function and vocabulary development in TD preschoolers (Weiland et al., 2014). In the current study, we confirmed group differences in novel word learning and executive function between preschoolers with DLD compared to their TD peers. Additionally, we found a positive relationship between executive function and word learning, with executive function performance accounting for a significant amount of variance in children's word learning outcomes over and above the variance accounted for by DLD versus TD status. Specifically, inhibition and short-term memory were found to be significant predictors of word learning in our regression models.

We also considered the effects of referent familiarity on word learning. Here, we expected learning pseudoword labels for familiar referents to be more difficult due to additional cognitive demands associated with relabeling a familiar referent, which we expected to be especially problematic for participants with DLD. Familiar objects were more difficult, but this was equally true for both groups on the production task where there was an advantage for pseudowords with unfamiliar referents. The effect of familiarity on production is likely because the known label for a familiar referent was a strong semantic competitor, making production of the novel word more difficult. This pattern also suggests that children with DLD may be using mutual exclusivity to some extent to account for their relative strength in producing pseudowords paired with novel objects.

Significant correlations were found between several of our study variables, although relationships differed for participants with DLD versus TD participants. Of particular interest are variables related to word learning outcomes. None of our other study measures were significantly related to word learning among the participants with DLD, but nonverbal IQ, expressive language, receptive vocabulary, and SES were positively correlated with outcomes for participants with typical language abilities. The absence of correlation between receptive vocabulary and word learning outcomes for children with DLD has been previously reported (Gray, 2003, 2006; Nash & Donaldson, 2005; Rice et al., 1990), which may reflect the difference between crystallized word knowledge captured by vocabulary tests versus the initial process of word-referent mapping that we measured in our learning paradigm. In contrast, we found a significant correlation between receptive vocabulary and word learning in our TD group, which has also been previously reported among children with DLD (Alt et al., 2004; Kan & Windsor, 2010).

Based on the model presented by Weiland et al. (2014), we predicted that executive function performance would account for a significant amount of variance in word learning outcomes, which was substantiated by our results. However, in this previous research, a composite executive function score was used, so it was less clear which specific executive function components would relate to word learning when they were tested independently. We analyzed data keeping the executive function measures separate because they were uncorrelated in our DLD group, and therefore, a combined score would not accurately represent their performance. In a series of linear regression analyses, after accounting for variance in word learning explained by DLD versus TD group membership, day–night performance (i.e., inhibition) remained a significant predictor of novel word production, phonological recognition, and comprehension, and forward word span (i.e., short-term memory) was a significant predictor of novel word comprehension. The addition of the executive function predictors accounted for an additional 9.8% of the variance in children's production task performance, 18.7% of the variance in phonological recognition scores, and 10.4% of the variance in nonword comprehension.

The relationship between inhibition and word learning is consistent with models of lexical access in which accessing a target word requires inhibition of phonological and/or semantic competitors (e.g., TRACE model; McClelland & Elman, 1986). In such models, an integral aspect of lexical access is one's ability to inhibit retrieval of similar nontarget items. Within the context of a novel word-learning paradigm, participants are dealing with competitors within their existing lexicon. For example, /bɪm/ likely has semantic (e.g., “kite”) and phonological (e.g., “bin”) competitors in the child's lexicon. Additionally, participants are faced with competition between the novel words that are being learned simultaneously. Furthermore, it is worth noting that many language assessments, including those used in this study, place explicit demands on inhibition by having children make a forced choice among foils, which act as competitors that must be inhibited. This approach may increase executive function demands beyond what is necessary for naturalistic language acquisition/use (Frizelle et al., 2017). This potential issue is present across many standardized language assessments that rely on similar forced-choice responses and may in turn underestimate a child's language abilities due to the increased executive function demands of the assessment.

In addition to inhibition, participants' short-term memory, as assessed by a forward word span task, accounted for significant variance in novel word comprehension. This is consistent with prior research linking short-term memory and word learning (Alt & Plante, 2006; Gathercole, 2006) and likely relates to participants' ability to hold both the phonological form and visual referent in mind to create an association between the two. This mapping between word and referent is essential to selecting the correct referent for each word in the comprehension task.

Alternatively, the same cognitive explanation offered by Weiland et al. (2014) as the possible underlying mechanism of the relationship between executive function and vocabulary learning may apply to our findings. As such, individuals with stronger executive function are better equipped to learn from the stimuli in their environment. In this case, the relationship between executive function and word learning may be explained by the fact that children who were better able to ignore distractions to attend to the visual/auditory input during familiarization and who were able to hold the new information in memory were better word learners. This explanation suggests a broader link between executive function and learning that may not be specific to language learning.

Although these theoretical relationships fit within the directional model provided by Weiland et al. (2014), in our current data, we can neither differentiate between the possible mechanisms by which executive function may relate to word learning nor confirm the directionality of this relationship. Indeed, it may be the case that the relationship between the variables is because children who are better word learners have stronger language abilities, which in turn improves their ability to mediate their behavior on our executive function measures (e.g., Vygotsky & Kozulin, 1986).

Finally, we addressed the possible confounding factors of IQ and receptive vocabulary, which differed significantly between our groups, by including these in our regression models. None of the executive function predictors were significant in models of comprehension after including these variables, but inhibition still accounted for a significant amount of variance in novel word production and phonological recognition performance over and above these other variables. This suggests that inhibition accounts for unique variance in word learning such that children with better inhibition abilities are also more successful word learners. This relationship may be especially true in the current data where half of the pseudowords referred to known objects that required participants to inhibit the conventional label to correctly identify/use the pseudoword. Although the need to relabel known objects is not characteristic of all word learning paradigms used for research, it does occur in naturalistic word learning outside the laboratory.

Limitations and Future Research

The present findings provide an interesting picture of the relationship between executive function and word learning in preschoolers with and without DLD, but there are some limitations worth noting. The design of this research allowed us to document relationships between performance on executive function tasks and word learning outcomes; however, we cannot establish causality or directionality. Subsequent investigations may employ experimental designs by utilizing interventions targeting word learning and/or executive function and assessing the effects of intervention on both trained and untrained skills. Without such data, we cannot determine whether language abilities are affecting executive function, whether executive function is affecting language, or whether language and executive function deficits simply co-occur in children with DLD. Furthermore, all of our executive function tasks were verbal in that they included verbal instructions and stimuli, and our span task required verbal responses. This precludes drawing conclusions about the potential relationship between domain-general executive function and word learning. Instead, we can conclude that executive function measured using verbal tasks is related to word learning, which is also a verbal task.

Additionally, the relationships between executive function and word learning we established here may be restricted to preschool-age participants. In other words, these relationships may differ and/or be nonexistent at different points in the life span. A related issue is that we cannot use the current data to address the development of either executive function or word learning over time because all our participants are of the same age. Future research addressing similar questions using cross-sectional or longitudinal designs will help elucidate how executive function and word learning may be related at different points in development and can address how change in one variable may be associated with change in the other variable.

Conclusions

As expected, participants with DLD performed significantly worse on measures of novel word learning and executive function relative to age-matched, TD participants. Despite the group differences in word learning outcomes, children in both groups showed evidence of word learning and were equally affected by referent familiarity as both groups showed lower production accuracy for words with familiar referents. Children's inhibition abilities accounted for a significant amount of variance in all word learning outcomes over and above group membership, and short-term memory was a significant predictor of word comprehension. Inhibition remained a significant predictor of production and recognition even after controlling for participants' IQ and receptive vocabulary abilities. These findings indicate that executive function is related to word learning performance in children with and without DLD and accounts for unique variance in children's word learning outcomes. Establishing a relationship between these variables provides a foundation for future work assessing the directionality of the relationship between executive function and language in children with DLD and has clinical implications related to the possible utility of interventions that are designed to improve word learning and/or executive functioning in this population.

Funding Statement

This work was supported by the National Institute on Deafness and Other Communication Disorders (Grant F32DC014188, awarded to L. Kapa).