Nonverbal Switching Ability of Monolingual and Bilingual Children with and without Developmental Language Disorder

1. Introduction

In everyday life, children continuously need to adapt to changing situations. This requires cognitive switching: the ability to adapt behavior to new situations and switch flexibly between tasks or perspectives (Diamond 2013; Legare et al. 2018). Switching, sometimes referred to as cognitive flexibility or shifting, is one of the core components of executive functioning (EF), which also includes the updating of working memory and the inhibition of unwanted behavior (Miyake et al. 2000). Together, these EF components regulate one’s thoughts and actions, allow for complex skills such as the planning and execution of goal-oriented behavior (Barkley 2012; Diamond 2013), and are essential for school success and quality of life (Borella et al. 2010; Brown and Landgraf 2010).

Previous research has shown that EF may be positively affected by bilingualism (Hilchey and Klein 2011; Martin-Rhee and Bialystok 2008; van den Noort et al. 2019; Ware et al. 2020). In contrast, having a developmental language disorder (DLD), which is characterized by weak language ability without a known origin, may negatively impact EF (Aljahlan and Spaulding 2021; Kapa and Plante 2015; Pauls and Archibald 2016). Specifically for the ability to switch between tasks, the literature likewise suggests a pattern of differential effects of bilingualism and DLD. Bilinguals have been reported to outperform monolingual peers on tasks assessing switching ability (e.g., Barac and Bialystok 2012; Prior and MacWhinney 2010), although some studies do not find evidence for such a bilingual advantage (e.g., Paap et al. 2017; Timmermeister et al. 2020). Collapsing evidence from multiple studies, two recent meta-analyses report superior switching performance of bilingual participants (Gunnerud et al. 2020; Ware et al. 2020). Conversely, children with DLD have been suggested to have weaker switching ability than their typically developing (TD) peers in some studies (e.g., Blom et al. 2021; Farrant et al. 2012), but not in others (e.g., Henry et al. 2012). In line with these mixed findings, meta-analyses showed small but reliable effects of DLD on switching (Pauls and Archibald 2016) which were dependent on the type of task (Aljahlan and Spaulding 2021).

Thus, findings on the relationship between both bilingualism and switching ability as well as DLD and switching ability are mixed and have been the subject of discussion (see, Kapa and Plante 2015; Poarch and Krott 2019). To our knowledge, previous work has not addressed the interaction of bilingualism and DLD on switching ability. This is important, as it provides insight into the EF profile of children with DLD who grow up bilingually, which is a large and ever-growing group of children given increases in global mobility and migration (for Dutch statistics, see Centraal Bureau voor de Statistiek 2021). Moreover, it may reveal whether a bilingual benefit can be observed despite impaired language development, informing us about the mechanisms underlying such an advantage. In the present study, we aimed to further elucidate the effects of both bilingualism and DLD on nonverbal switching by adopting a four-group design. We focus on nonverbal switching because nonverbal tasks can inform us whether switching problems or benefits extend beyond the linguistic domain (see also Pauls and Archibald 2016). The four groups consisted of monolingual and bilingual children, both with and without DLD. Furthermore, we aimed to explore potential causes for differences in performance between groups, which may inform us about the origin of enhanced or weakened switching ability. To this end, we investigated the role of inhibition and attention skills in explaining group effects on switching ability. Inhibition and attention are both implicated in switching ability (Garon et al. 2008) and have been suggested to be strengthened in bilinguals (Ware et al. 2020) and weakened in children with DLD (Ebert et al. 2019; Pauls and Archibald 2016; Smolak et al. 2020), possibly driving group effects (see also Pauls and Archibald 2016).

1.1. Development and Assessment of Switching Ability

Although all core EF components start to develop during infancy and preschool years, switching ability is known to develop later than inhibition and working memory (Diamond 2013; Garon et al. 2008), following a protracted development through adolescence (Best and Miller 2010). Switching is considered the most complex EF component (Garon et al. 2008). It builds on inhibition and working memory but involves more than only the sum of the two. To be able to switch from one perspective to another, for example, a person needs to inhibit the first perspective and then load the new perspective into working memory (Diamond 2013; Dajani and Uddin 2015), but it also requires a person to reconfigure his/her responses according to the new situation (Dajani and Uddin 2015). Moreover, in addition to dependencies among the three core EF components, basic and early-developing attentional processes are proposed to underlie EF in general, including the development of switching (Garon et al. 2008; Rothbart and Posner 2001). As suggested by Kapa et al. (2017), these hierarchical relations between attention, working memory, inhibition, and switching have important implications for children with weaknesses or strengths in one of these components. Impaired or enhanced development in a lower-level component may have cascading effects on higher-level components. This may be key in explaining the observed effects of bilingualism and DLD on switching ability, which will be reviewed in Section 1.2 and Section 1.3.

Experimentally, switching has been probed with a multitude of measures, which, for example, differ depending on the frequency and regularity of switches, the presence of explicit cues indicating to switch, and/or the level at which switching takes place (i.e., switching between different task goals or between different rules while the task goal remains constant). An example of a task that has often been used with children is the Dimensional Card Change Sort (DCCS; Zelazo 2006), in which children are asked to sort cards according to one dimension (e.g., color) after which they need to switch to a second dimension (e.g., shape). Most 5-year-old children are able to correctly make the switch to the second sorting rule, while this is still too difficult at the age of 3 (Zelazo 2006). For the present study, including participants between 5 and 8 years old, we used a slightly more complex color–shape task-switching paradigm. Children had to indicate whether a stimulus corresponded to one of two colors or shapes through button presses where each of two buttons was associated with one feature of the task. The left button corresponded to a blue or square object, while the right button corresponded to an orange or triangular object. Comparable to the DCCS, the task started with two subsequent single-task blocks, in which children were presented with one task (i.e., sorting either on color or shape). In the current paradigm, these two single-task blocks were followed by a switching block in which children had to flexibly switch between the two sorting rules: if the task was to sort according to color, the stimulus “blue triangle” would require a left button press, while it would require a right button press when sorting according to shape. A cue indicated which sorting rule was active. Previous research with such a task-switching paradigm showed a decrease with age in the performance costs associated with switching between tasks (e.g., Huizinga and van der Molen 2011; see also Cragg and Chevalier 2012).

These performance costs are typically indexed by two outcome measures: mixing costs and switching costs. Mixing costs are the global performance costs associated with the presence of two different tasks (single-task/task-switch), as indicated by the performance difference between trials in the switching block that do not involve a switch (repeat trials) and trials in the single-task blocks. As argued by Braver et al. (2003), mixing costs reflect proactive control processes, including the use of sustained attention to keep multiple tasks activated. Switching costs are the local costs related to the necessity to switch from one task to the other and are measured by looking at the difference between trials in the switching block that involve a switch (switch trials) and those that do not (repeat trials). Switching costs may more closely reflect reactive control processes (Braver et al. 2003), which enable a successful response after the detection of interference and the reactivation of goals (Braver 2012). As such, switching costs reflect the inhibition of previous stimulus-response associations (i.e., response inhibition; Druey and Hübner 2008; Vandierendonck et al. 2010).

1.2. Nonverbal Switching in Bilingual Children

The use of more than one language requires bilinguals to switch between languages, which in turn involves the inhibition of the non-target language and the selection of the target language depending on the context (Green and Abutalebi 2013). This natural switching in bilingual language use is suggested to draw on domain-general EF, which is thereby continuously trained (Green 1998; Green and Abutalebi 2013). This training effect may explain why bilinguals have been shown to outperform monolinguals on a wide range of EF tasks, including verbal and nonverbal measures (Adesope et al. 2010; Bialystok and Craik 2010; Hilchey and Klein 2011; van den Noort et al. 2019; Ware et al. 2020). A bilingual EF benefit has been found in both adult and child populations, although the evidence appears stronger for older groups (Gunnerud et al. 2020; van den Noort et al. 2019; Ware et al. 2020). As numerous studies failed to find positive effects of bilingualism or found effects to be very small (de Bruin et al. 2015; Duñabeitia et al. 2014; Gathercole et al. 2014), open questions are whether EF differences between monolinguals and bilinguals hold for all bilinguals, including bilinguals with development disabilities, which factors modulate these differences, and what they mean in real life (Poarch and Krott 2019).

The focus of the present study is on nonverbal switching, a domain-general ability that may underlie switching between languages. In a comprehensive meta-analysis of 170 studies, using tasks tapping into different EF domains and including both children and adults, Ware et al. (2020) report superior performance of bilinguals as compared to monolinguals on two dependent variables of the task-switching paradigm. A bilingual benefit was found for accuracy on incongruent trials (corresponding to switch trials in the task-switching paradigm; Hedges’ g = 0.225) and on interference effects (corresponding to switching and/or mixing costs in the task-switching paradigm; Hedges’ g = 0.693). Accuracy on congruent trials as well as reaction times on congruent and incongruent trials did not reveal significant effects. Focusing only on children, another recent meta-analysis including 100 publications also supports stronger switching performance of bilinguals relative to monolinguals (Hedges’ g = 0.329; Gunnerud et al. 2020). Switching was the only EF domain for which effects remained significant after controlling for publication bias.

When we further zoom in on child participants, there are only a few studies that used a similar cued color–shape switching task as included in the present study. Barac and Bialystok (2012) assessed switching performance in 78 6-year-old children in three bilingual groups (Chinese-English, French-English, and Spanish-English) as well as in 26 English monolinguals. All three bilingual groups were found to perform similarly and exceeded the monolinguals as shown by smaller mixing costs. However, no differences between groups were found for switching costs. Antoniou et al. (2016) only looked at switching costs and also found a bilingual advantage, including 44 bilectal, 22 multilingual, and 25 monolingual children aged 6 to 9 years old from Cyprus and Greece. Finally, Timmermeister et al. (2020) likewise adopted a similar task design and assessed nonverbal switching ability in a sample of 27 5- to 8-year-old Turkish-Dutch bilinguals and 27 age-matched Dutch monolinguals. Their study resulted in comparable mixing and switching costs for bilinguals and monolinguals.

1.3. Nonverbal Switching in Children with DLD

Children with DLD have severe difficulties learning language, in the absence of an obvious cause, such as a hearing impairment or low intellectual functioning (Leonard 2014). While some argue that DLD is a domain-specific deficit that selectively impairs language acquisition (Rice and Wexler 1996; Van der Lely 2005), the current dominant view is that the language problems coincide and interact with broader impairments in cognitive and/or perceptual mechanisms, including attention and EF (Botting and Marshall 2017; Kapa and Plante 2015). Research indeed indicates that children with DLD are outperformed by TD peers on a wide range of EF measures (for reviews/meta-analyses, see Aljahlan and Spaulding 2021; Kapa and Plante 2015; Pauls and Archibald 2016; Vissers et al. 2015). Studies have, however, not revealed unequivocal results (Archibald and Gathercole 2006; Henry et al. 2012; Im-Bolter et al. 2006; Lukács et al. 2016; Noterdaeme et al. 2001). For example, several studies only found differences between children with DLD and TD peers on verbal EF tasks and not on nonverbal measures (Archibald and Gathercole 2006; Henry et al. 2012; Lukács et al. 2016; but see, Kapa et al. 2017). As of yet, the breadth and origin of the EF weaknesses of children with DLD remain unclear.

With respect to switching, two meta-analyses compile the evidence. A recent meta-analysis of 20 studies by Aljahlan and Spaulding (2021) demonstrated poorer performance of children with DLD in comparison with TD peers (Hedges’ g = −0.42). This corresponds to an earlier meta-analysis of 22 studies by Pauls and Archibald (2016), showing a small, but reliable, effect of DLD on switching (Hedges’ g = −0.27). Both meta-analyses investigated a number of potential moderating variables, and both found no moderating effect of age. In addition, the linguistic demand of the task (Pauls and Archibald 2016), the severity of DLD (Pauls and Archibald 2016), and the outcome measure (i.e., accuracy or reaction time; Aljahlan and Spaulding 2021) were not related to the poor switching performance of children with DLD. Aljahlan and Spaulding (2021) did find a significant moderating effect of the type of switching task, showing a significant negative effect of DLD on set-shifting tasks (Hedges’ g = −0.52) in contrast to alternating tasks (Hedges’ g = −0.18). The authors defined set-shifting tasks (e.g., DCCS) as tasks in which children were asked to sort stimuli according to more than one rule, but in which a novel rule was only introduced after a dominant response pattern was established, thus involving infrequent switches between rules. In contrast, alternating tasks were defined as tasks in which participants are familiarized with an alternating set of rules and are asked to respond in a consistent, predictable alternating pattern (e.g., following numbers and letters alternately in sequence, 1-A-2-B-3-C, in the Trail-Making Test). Aljahlan and Spaulding (2021) indicate that their findings may be explained by the extent to which a specific task taxes additional executive processes, such as inhibition and attention, which are also weak in children with DLD. Some set-shifting tasks may rely more extensively on such processes than alternating tasks. Important to note here is that the task-switching paradigm of the current study does not fall perfectly into either task category as defined by Aljahan and Spaulding, although it more closely fits the set-shifting category. To our knowledge, the task-switching paradigm has not yet been used in DLD research and is thus not directly comparable with previously used measures in that field.

The hypothesis that other cognitive processes play an important role in explaining the switching abilities of children with DLD is also discussed by Pauls and Archibald (2016). They speculate that the observed small effect of DLD on switching tasks may be driven by an inhibition deficit, as switching to another task is difficult when a previous task is not adequately suppressed (Cragg and Chevalier 2012; see also Im-Bolter 2003). The results of a study by Kapa et al. (2017) further support this hypothesis, showing a linear increase in the effect of DLD from working memory to inhibition to switching. As EF components become more complex and there is more need to coordinate lower-level skills, deficits of children with DLD seem to increase. Consequently, the relatively large switching deficit of the children with DLD may have been caused by more subtle weaknesses in lower-level components, including working memory and inhibition. This possibility has, however, not been directly investigated. In addition, the role of attentional processes, which are thought to be the lowest in the hierarchy (Garon et al. 2008), needs to be studied as well. Against their predictions, the children with DLD in the study of Kapa et al. (2017) showed the largest deficit in sustained attention. There is an increasing body of work reporting weak sustained attention skills of children with DLD (for a meta-analysis, see Ebert and Kohnert 2011; Ebert et al. 2019), which are, furthermore, found to be associated with children’s language skills (Boerma et al. 2017; Smolak et al. 2020). It is unknown whether sustained attention is also key in explaining the difficulties of children with DLD in nonlinguistic domains, such as nonverbal switching.

1.4. The Present Study

Previous research reported opposite effects of bilingualism and DLD on switching ability, respectively, being associated with enhanced (Ware et al. 2020) and weakened (Aljahlan and Spaulding 2021; Pauls and Archibald 2016) performance. However, both studies with bilinguals as well as studies including children with DLD have revealed mixed results. Moreover, the two fields have, as of yet, operated separately with respect to the investigation of switching, which means that it is unknown how switching ability is impacted by the interaction of bilingualism and DLD. It may be that the positive effects of bilingualism on switching mitigate the negative effects of DLD, as is shown by two studies on the switching abilities of bilingual children and children with Autism Spectrum Disorder (ASD), who also often have language and EF deficits (Gonzalez-Barrero and Nadig 2019; Peristeri et al. 2021). However, previous work on bilingual children and children with DLD has not found such mitigating effects in other EF domains. These studies, investigating working memory, inhibition (Boerma and Blom 2020), and attention (Ebert et al. 2019), suggest that the effect of bilingualism on these cognitive skills is not different for TD children or children with DLD and that, likewise, DLD does not impact monolingual and bilingual children in different ways. In both studies, children with DLD scored weaker than TD children, while no effect of bilingualism nor an interaction between bilingualism and DLD was found. A third study on three attention processes partly replicated these findings but also showed a bilingual advantage in orientation skills, which was restricted to the DLD group (Park et al. 2019). This may indicate that children with DLD can benefit more from bilingual experience than TD children. However, previous research on switching demonstrated that bilingual children with high language proficiency outperform those with low language proficiency (Iluz-Cohen and Armon-Lotem 2013), suggesting that the low language abilities of children with DLD could actually weaken a bilingual advantage. Although no clear pattern can thus be distilled from these studies, their findings, together with the results from the current study, have important implications for the growing number of bilingual children on clinical caseloads, and for furthering our understanding of the factors moderating EF strengths and weaknesses.

The first aim of the current study is to investigate the independent and interaction effects of bilingualism and DLD on nonverbal switching ability. Particularly in the literature on DLD, the large majority of switching tasks used have a high verbal load (Pauls and Archibald 2016). This may disadvantage children with DLD and blur results, although linguistic demand was not found to be a significant moderator of the effect of DLD on switching ability (Pauls and Archibald 2016). We used a nonverbal switching task to eliminate the effect of language proficiency as much as possible, providing insight into the breadth of switching problems or benefits. We included four groups of children, monolingual and bilingual children with and without DLD, between the ages of 5 and 8 years old.

The second aim of the present study is to better understand the origin of strong or poor nonverbal switching ability of, respectively, bilingual children and children with DLD. Previous work indicates that switching is a complex process, which builds on other lower-level EF components and on more basic attentional processes (Garon et al. 2008). As these may also be enhanced due to bilingualism (Ware et al. 2020) and weakened due to DLD (Pauls and Archibald 2016; Smolak et al. 2020), it is possible that such lower-level cognitive processes explain switching problems or benefits (see Kapa et al. 2017). Here, we included two measures of cognitive processes on which switching skills may rely: response inhibition and sustained attention. These measures were chosen, because our switching outcome variables, mixing and switching costs, have been suggested to reflect sustained attention and response inhibition, respectively (Braver 2012; Braver et al. 2003; Druey and Hübner 2008; Vandierendonck et al. 2010). Moreover, Pauls and Archibald (2016) specifically suggest that the inhibition of a previously established response is the aspect of switching tasks that may be most challenging for individuals with DLD (see also Im-Bolter 2003). We conducted mediation analyses to investigate the role of response inhibition and sustained attention in explaining the significant effects of bilingualism and DLD on nonverbal switching ability.

For our first research aim, we hypothesized that, in line with the above-mentioned research, bilingual children would outperform monolingual peers on our nonverbal switching task, while children with DLD would score lower than TD peers. Given the mixed findings in the literature, we also reckoned with the possibility of finding null results. With respect to the interaction of bilingualism and DLD, we can only rely on studies that have investigated other EF domains in bilingual children with DLD or on studies using a switching task with bilinguals with ASD, as described above. As no clear pattern emerged from these studies, we refrain from formulating specific hypotheses. With respect to the second aim of the current study, our hypothesis was that response inhibition and sustained attention would mediate group effects on switching ability. Specifically, sustained attention was thought to be associated with mixing costs and response inhibition with switching costs.

2. Materials and Methods

2.1. Participants

The data of the present study were collected in the context of a longitudinal research project investigating the linguistic and cognitive development of children growing up in the Netherlands. It included three waves of data collection, with intervals of one year. For the present study, the same sample of participants was used as in the previous work of (Boerma et al. 2017; Boerma and Blom 2020), consisting of monolingual TD children (MOTD), bilingual TD children (BITD), monolingual children with DLD (MODLD), and bilingual children with DLD (BIDLD). The four groups of children were a subsample of the full sample and were matched on a number of background variables (see

Table 1. Background characteristics of the four groups of participants.

			Chronological Age in Months
		Gender	Wave 1	Wave 2	Wave 3	Nonverbal Intelligence	Socio-Economic Status b
	N a	Girls/Boys	Mean (SD)	Mean (SD)	Mean (SD)	Mean (SD)	Median (Range)
MOTD	32	14/18	70.9 (7.0)	82.5 (6.9)	94.1 (6.9)	100.4 (11.9)	6.8 (2–9)
MODLD	32	8/24	71.4 (6.3)	82.8 (6.5)	94.6 (6.6)	97.5 (12.9)	5.0 (2–8.5)
BITD	32	17/15	71.3 (7.3)	83.0 (7.1)	94.8 (7.1)	95.8 (15.0)	5.4 (1–9)
BIDLD	32	10/22	72.4 (8.6)	83.0 (8.9)	94.7 (8.8)	94.7 (15.3)	5.8 (2–9)

Note. MOTD: monolingual typically developing; MODLD: monolingual language disorder; BITD: bilingual typically developing; BIDLD: bilingual language disorder. ^a All children participated in each wave. ^b Socio-economic status was indexed by parental education (based on the International Standard Classification of Education) and could range from 1 (no education) to 9 (university degree). A value of 5 corresponds to intermediate vocational education.

). As the BIDLD group was the smallest group, we matched one child from each other group to a child in the BIDLD group based on these background variables. There were no significant differences between the four groups in terms of chronological age (wave 1: F(3,124) = 0.25, p = 0.86, η_p² < 0.01; wave 2: F(3,124) = 0.03, p = 0.99, η_p² < 0.01; wave 3: F(3,124) = 0.07, p = 0.98, η_p² < 0.01), nonverbal intelligence (F(3,124) = 1.02, p = 0.39, η_p² = 0.02), and socio-economic status (H(3) = 5.5, p = 0.14). Nonverbal intelligence (NVIQ) was tested in wave 1 with the Wechsler Nonverbal-NL (Wechsler and Naglieri 2008) and socio-economic status (SES) was estimated by the average parental education level in wave 1. The groups of children with DLD included relatively many boys, which is expected given previously reported prevalence rates (e.g., Tomblin et al. 1997), although group differences did not reach significance (χ²(3, N = 128) = 6.4, p = 0.09).

All participants were born in the Netherlands and attended elementary school from the age of 4 years onward. Dutch was the language of instruction at all schools of the participating children. Children’s home language environment determined whether they were classified as monolingual or bilingual, using parental report through the Questionnaire for Parents of Bilingual Children (PaBiQ; Tuller 2015). Parents of monolingual participants were both native speakers of Dutch and spoke this language with their child. One or both parents of bilingual participants were native speakers of another language than Dutch and spoke this language with their child on a regular basis. The bilingual children were exposed to another language than Dutch at least 30% of the time before the age of 4, but they varied considerably in how much Dutch input they received. The group of bilingual TD children and the group of bilingual children with DLD were therefore matched on percentage of exposure to Dutch before the age of 4 years (BITD: Mean(SD) = 43.0 (8.3), BIDLD: Mean(SD) = 40.9 (11.1); F(1,61) = 0.68, p = 0.41, η_p² = 0.01) and percentage of exposure to Dutch in wave 1 (BITD: Mean(SD) = 50.9 (12.0), BIDLD: Mean(SD) = 45.2 (16.5); F(1,62) = 2.5, p = 0.12, η_p² = 0.04). Four-fifths of the bilingual participants came from the two largest immigrant groups in the Netherlands, coming from Turkish or Moroccan descent.

All participants in the current study had an NVIQ of 70 or above, did not have hearing problems or severe articulatory difficulties, and were not diagnosed with Autism Spectrum Disorder or Attention Deficit Hyperactivity Disorder. None of the TD participants were reported to have language problems and all attended regular elementary schools, through which they were recruited. The participants with DLD were recruited through Royal Dutch Kentalis or Royal Auris Group. In waves 1 and 2, they all received educational support (i.e., either special education or regular education with ambulatory care) for their language difficulties and thus met the official Dutch criteria which specify when a child is eligible for such support (Stichting Siméa 2014). This was established by clinicians before the start of data collection and meant that they scored at least 1.5 standard deviations (SD) below the mean on two out of four subscales of a standardized language assessment test battery. Alternatively, they scored at least 2 SD below the mean on an overall score of this test battery. For the bilingual children, language development was evaluated with a bilingual anamnesis and, if possible, assessment of both languages (Stichting Siméa 2016). In the bilingual anamnesis, parental information is obtained on whether language difficulties are present in both languages, the persistence of these difficulties, and the amount of language input in both languages. In wave 3, there were four monolingual and four bilingual children with DLD who did not qualify for educational support anymore. These children were not excluded, as DLD and language difficulties are known to be persistent (Scarborough and Dobrich 1990) and diagnostic categories have been shown to be unstable (Conti-Ramsden and Botting 1999).

The Peabody Picture Vocabulary Test (PPVT; Schlichting 2005) and the subtest Sentence Repetition from the Taaltoets Alle Kinderen (TAK; Verhoeven and Vermeer 2001) measured Dutch receptive vocabulary and grammatical skills, respectively (see

Table 2. Language skills of the four groups of participants (raw scores).

	PPVT			TAK Sentence Repetition
	Wave 1	Wave 2	Wave 3	Wave 1	Wave 2	Wave 3
	Mean (SD)	Mean (SD)	Mean (SD)	Mean (SD)	Mean (SD)	Mean (SD)
MOTD	86.6 (10.4)	98.5 (9.6)	103.8 (8.9)	28.7 (5.9)	32.0 (5.5)	34.3 (3.7)
MODLD	76.5 (9.2)	86.6 (11.9)	95.9 (10.1)	10.2 (5.7)	14.9 (6.1)	19.9 (7.2)
BITD	75.8 (10.5)	87.8 (11.5)	96.2 (12.4)	21.8 (7.0)	27.7 (7.1)	31.0 (6.3)
BIDLD	62.9 (13.3)	77.8 (15.5)	86.3 (14.0)	9.5 (5.8)	14.8 (7.5)	19.0 (8.7)

Note. MOTD: monolingual typically developing; MODLD: monolingual language disorder; BITD: bilingual typically developing; BIDLD: bilingual language disorder; PPVT: Peabody Picture Vocabulary Test (range from 0 to 204); TAK: Taaltoets Alle Kinderen (range from 0 to 40).

). There were significant group differences in each wave on both the PPVT (wave 1: F(1,122) = 24.7, p < 0.001, η_p² = 0.38; wave 2: F(1,123) = 15.1, p < 0.001, η_p² = 0.27; wave 3: F(1,124) = 12.4, p < 0.001, η_p² = 0.23) and the TAK (wave 1: F(1,122) = 74.9, p < 0.001, η_p² = 0.65; wave 2: F(1,123) = 57.3, p < 0.001, η_p² = 0.58; wave 3: F(1,124) = 42.6, p < 0.001, η_p² = 0.51). The MOTD and BITD group consistently outperformed the MODLD and BIDLD group, respectively. The BITD group scored lower on vocabulary than the MOTD group in waves 1 and 2, but caught up in wave 3. With respect to grammar, significant differences between the two TD groups only emerged in wave 1. The BIDLD group had weaker vocabulary skills than the MODLD group in each wave, but no differences between the two groups were found for grammatical skills.

2.2. Instruments

2.2.1. Nonverbal Switching

Nonverbal switching ability was measured in waves 2 and 3 with a cued color/shape switching task (Timmermeister et al. 2020). In this task, square- or triangular-shaped objects which were blue- or orange-colored were presented on a 15-inch laptop screen using E-Prime 2.0 (Schneider et al. 2002). At the top of the screen, a cue in the form of one of two cartoon faces (“Mr. Color” or “Mr. Shape”) appeared. Based on the cue, children had to identify either the color or the shape of the target object. The target object was presented in the middle of the screen, preceded by a fixation cross (350 ms) and a blank space (150 ms). The cue was presented 650 ms before the fixation cross and remained on the screen until the end of the trial. A blue square and orange triangle were also present during all trials, on respectively the left and right sides of the bottom of the screen (see Figure 1). Children responded by pressing one of the two fixed buttons on the left and right sides of the keyboard, which corresponded to those two objects. When “Mr. Color” appeared, the left button had to be pressed for blue objects and the right button for orange objects. When “Mr. Shape” appeared, the left button was for squared objects and the right for triangles. Children had a maximum of 7 s to respond to a trial.

The task started with two single-task blocks, both preceded by a practice phase (5 items each), in which the cue remained the same and children had to respond to either the color or shape of the target object throughout the block. The order of the two single-task blocks was counterbalanced, with some children starting with color and some children starting with shape. There were 28 trials in each of the two blocks. The single-task blocks were followed by a task-switching block, in which the cues for color and shape changed every two to five trials. The order of the trials was fixed but unpredictable. This block included a practice phase (8 items) and subsequently 56 trials, equally divided between color and shape. Trials in the task-switching block involved a cue switch in 25% of the trials (switch trials). In 75% of the trials in this block, there was no change of cue from color to shape or vice versa (repeat trials).

2.2.2. Response Inhibition and Sustained Attention

Response inhibition and sustained attention were measured in wave 1 with an integrated auditory and visual Continuous Performance Task (CPT; see Boerma et al. 2017) using E-Prime 2.0 (Schneider et al. 2002). Children saw (for 167 ms) or heard the number ‘1’ or the number ‘2’. In response to the number ‘1’, which was the target stimulus, children were instructed to press the space bar. When hearing or seeing the number ‘2’, which was the distractor, children had to refrain from responding. The task started with a practice phase in which 10 items were presented visually, followed by 10 items that were presented auditorily, and finally 10 items in which the mode of presentation was mixed. The test phase consisted of 168 trials. In the first and third part (impulsivity blocks), the ratio of targets versus distractors was five to one, respectively, whereas this was reversed in the second and fourth part (inattention blocks).

2.3. Procedure

The longitudinal project, within which the data for the current study was collected, was approved by The Standing Ethical Assessment Committee of the the Faculty of Social and Behavioral Sciences at Utrecht University. Parents of participating children signed a written informed consent form. Testing took place in a quiet room at a child’s school and was done individually by a trained researcher. Children completed several language and cognitive tasks in two test sessions. Instructions were in Dutch and could be repeated until a child understood the task. The CPT and NVIQ measures were administered in wave 1 and the nonverbal switching task was administered in waves 2 and 3. It was not feasible to administer the CPT and the switching task at all three waves, because the duration of the full test battery would have become too long. Other tasks were assessed at all three waves.

2.4. Data Preparation

2.4.1. Nonverbal Switching

We first looked at children’s accuracy scores in the single-task blocks to ensure that children understood the basic instructions, as this is a prerequisite to measuring switching ability. The probability of scoring at least 19 out of the 28 trials correct (i.e., 67.8%) due to chance is smaller than 5% (based on a binomial distribution). This cut-off thus gives us relative certainty that children with a score of at least 19 performed above chance and did not respond randomly. In wave 2, there were 17 children who scored below this cut-off (scoring 18 trials correct or lower) on either the color or shape block or on both, including ten children from the BIDLD group. In wave 3, there were four children who scored below this cut-off, including one child from each of the four groups. Given the large number of children who were not able to accurately complete the single-task blocks in wave 2, we decided to focus on the nonverbal switching data from wave 31. In all further analyses, the four children with poor accuracy scores on both or either one of the single-task blocks in wave 3 were excluded.

Accuracy. The first trial of the task-switching block was excluded, as this is not a repeat or a switch trial. Accuracy scores on the single-task blocks (Color: skewness = −1.36, kurtosis = 1.34; Shape: skewness = −1.07, kurtosis = 0.94; ColorandShape: skewness = −1.06, kurtosis = 0.82) and on the repeat trials of the task-switching block (skewness = −1.17, kurtosis = 0.996) were slightly skewed, which was not improved by transformation. Non-parametric test results confirmed parametric tests, which is why only the latter are reported. There were no significant differences between accuracy scores on the single-task blocks for color and shape, as indicated by a mixed ANOVA with Condition (color/shape) as within-subjects factor and Group as between-subjects factor. Results showed no significant effect of Condition (F(1,120) = 0.20, p = 0.65, η_p² = 0.00), Group (F(3,120) = 1.33, p = 0.27, η_p² = 0.03), nor interaction effect (F(3,120) = 1.30, p = 0.28, η_p² = 0.03). This gives reason to pool the color and shape conditions in the accuracy analyses of the task-switching block.

Reaction times. Mean reaction times (RT) were calculated using accurate responses and, like accuracy, the first trial of the task-switching block was excluded. Moreover, responses below 200 ms and responses of more than three SD above a child’s mean RT were excluded. Together, this meant that we excluded 15.9% of the data for the RT analyses (of which 13.7% were incorrect responses). All reaction time outcome measures were logtransformed, resulting in normally distributed variables, and logtransformed variables were used in the analyses. There were no significant differences between RTs in the single-task blocks for color and shape, as indicated by a mixed ANOVA with Condition as the within-subjects factor and Group as the between-subjects factor. Results showed no significant effect of Condition (F(1,120) = 2.09, p = 0.15, η_p² = 0.02), Group (F(3,120) = 1.11, p = 0.35, η_p² = 0.03), nor interaction effect (F(3,120) = 0.974, p = 0.41, η_p² = 0.02). We therefore pooled the color and shape conditions in the RT analyses of the task-switching block.

2.4.2. Response Inhibition and Sustained Attention

Response inhibition was indexed by the number of false alarms in response to distractors on the CPT. Sustained attention was measured with the number of hits in response to target stimuli on the CPT. Accurate responses to targets with an RT below 100 ms were not included (<1% of all trials). The number of false alarms in the impulsivity blocks was strongly correlated with the overall number of false alarms (r = 0.75, p < 0.001). Similarly, the number of hits in the inattention blocks was strongly correlated with the overall number of hits (r = 0.82, p < 0.001). We therefore used the overall number of false alarms and hits as outcome measures, as these are based on a larger number of trials and are more robust. The number of false alarms and the number of hits were not significantly correlated (r = −0.17, p = 0.06), and were thus considered to underlie separate cognitive processes.

2.5. Data Analysis

All statistical analyses were done with SPSS 24 (IBM Corp. 2016). The first analyses corresponded to the first aim of the study, investigating the effects of DLD and bilingualism on nonverbal switching ability. In line with previous research, we analyzed mixing costs and switching costs in both accuracy and RT (for similar analyses, see Prior and MacWhinney 2010). To analyze mixing costs, we ran a mixed ANOVA with accuracy scores and a similar model with RTs as the dependent variable. We investigated whether the groups of children responded differently to trials in the single-task blocks in comparison with repeat trials. Trial (trials in single-task blocks/repeat trials in task-switching block) was entered as the within-subjects factor and Language Group (monolingual/bilingual) and Impairment Status (TD/DLD) as between-subjects factors. The same was done to analyze switching costs, with the exception that the within-subjects factor Trial now included a comparison between the switch and repeat trials in the task-switching block. Analyses were subsequently conducted with NVIQ and SES as covariates. Moreover, we also ran the RT analyses only including the children who performed above chance in the task-switching block (scoring more than 34 of 56 trials correct; 60.7%).

In case of significant effects, we subsequently conducted mediation analyses to investigate the role of response inhibition (number of false alarms to distractor) and sustained attention (number of hits to target) in explaining the relation between Language Group/Impairment Status and switching. This corresponds to the second aim of the current study. The PROCESS macro for SPSS of Hayes (2017) was used for this purpose. One requirement of this mediation analysis is temporal precedence from the independent variable (X) to the mediator (M) to the dependent variable (Y). This requirement was met, since group distinction (X) was formed before the assessment of response inhibition and sustained attention (M) which were, in turn, assessed prior to the evaluation of switching ability (Y). To determine whether the effect of X on Y is significantly reduced due to M (i.e., the indirect or mediation effect), bootstrapped tests (5.000—bias-corrected) and confidence intervals were used, as these are more reliable than p-values. Meaningful mediation is assumed if zero is not included in the confidence intervals of the indirect effects.

3. Results

Table 3. Mean % accuracy scores of the four groups of children.

				Task-Switching Block
		Single-Task Blocks		Repeat Trials		Switch Trials
	N	Mean (SD)	Range	Mean (SD)	Range	Mean (SD)	Range
MOTD	31	94.5 (4.8)	84–100	85.8 (10.0)	53–100	77.8 (16.8)	31–100
MODLD	31	93.3 (5.7)	80–100	77.3 (16.2)	43–100	70.8 (14.2)	38–100
BITD	31	91.6 (6.0)	79–100	85.1 (10.0)	58–100	72.4 (11.7)	44–94
BIDLD	31	93.4 (6.8)	71–100	78.5 (17.4)	38–100	69.2 (15.7)	38–94

Note. Four children (one from each group) were excluded based on low accuracy scores on one or both single-task blocks.

and

Table 4. Reaction times (in milliseconds) of the four groups of children.

				Task-Switching Block
		Single-Task Blocks		Repeat Trials		Switch Trials
	N	Mean (SD)	Range	Mean (SD)	Range	Mean (SD)	Range
MOTD	31	805.7 (354.4)	555–1651	1179.7 (374.8)	614–2719	1330.6 (354.4)	836–2484
MODLD	31	822.8 (210.9)	543–1514	1233.4 (332.8)	693–1904	1572.7 (435.1)	649–2805
BITD	31	784.2 (207.5)	432–1170	1244.3 (367.1)	547–2043	1560.5 (525.6)	744–2884
BIDLD	31	881.1 (241.7)	544–1463	1361.5 (395.5)	756–2143	1705.8 (498.2)	597–2954

Note. Four children (one from each group) were excluded based on low accuracy scores on one or both single-task blocks.

present the mean accuracy scores and reaction times, respectively, of the four groups of children on the single-task blocks (color/shape), on the repeat trials of the task-switching block, and on the switch trials of the task-switching block.

3.1. Mixing Costs: Effects of DLD and Bilingualism

3.1.1. Accuracy

Results showed a significant effect of Trial (F(1,120) = 107.34, p < 0.001, η_p² = 0.47). Accuracy scores on the trials in the single-task blocks were significantly higher than scores on the repeat trials in the task-switching block. There was also a significant effect of Impairment Status (F(1,120) = 5.48, p = 0.021, η_p² = 0.04), showing that children with DLD scored lower than TD children. The main effect of Language Group was not significant (F(1,120) = 0.14, p = 0.71, η_p² = 0.00). Furthermore, there was a significant interaction effect of Trial*Impairment Status (F(1,120) = 12.38, p = 0.001, η_p² = 0.09), indicating that the effect of Trial type on children’s accuracy scores was different for the TD group in comparison with the DLD group. Other interaction effects were not significant (Language Group*Impairment Status: F(1,120) = 0.65, p = 0.42, η_p² = 0.01; Trial*Language Group: F(1,120) = 0.60, p = 0.44, η_p² = 0.01; Trial*Impairment Status*Language Group: F(1,120) = 0.05, p = 0.83, η_p² = 0.00). The results did not change when controlling for SES and NVIQ, with NVIQ being a significant covariate and SES not.

We conducted two univariate ANOVAs to unpack the significant interaction between Impairment Status and Trial. The first analysis with accuracy scores on trials in the single-task blocks as the dependent variable and Impairment Status as the fixed factor showed that there was no difference between children with DLD and TD children (F(1,122) = 0.07, p = 0.79, η_p² = 0.00). The second analysis did reveal a significant effect (F(1,122) = 9.3, p = 0.003, η_p² = 0.07), indicating that children with DLD scored significantly lower on the repeat trials of the task-switching block than TD children. The results did not change when controlling for SES and NVIQ, with NVIQ being a significant covariate and SES not.

3.1.2. Reaction Times

Results showed a significant effect of Trial (F(1,120) = 383.4, p < 0.001, η_p² = 0.76). RTs on the trials in the single-task blocks were significantly lower than RTs on the repeat trials in the task-switching block. Other effects were not significant (Impairment Status (F(1,120) = 2.58, p = 0.11, η_p² = 0.02; Language Group: F(1,120) = 0.95, p = 0.33, η_p² = 0.01; Language Group*Impairment Status: F(1,120) = 0.61, p = 0.44, η_p² = 0.01; Trial*Impairment Status: F(1,120) = 0.00, p = 0.97, η_p² = 0.00; Trial*Language Group (F(1,120) = 1.66, p = 0.20, η_p² = 0.01); Trial*Impairment Status*Language Group: F(1,120) = 0.34, p = 0.56, η_p² = 0.00). The RTs of the groups of children were thus not differently affected by the two types of trials. The results did not change when controlling for SES and NVIQ. Neither NVIQ nor SES were significant covariates. When excluding children with low accuracy scores in the task-switching block (below 60.7% correct; 1 MOTD, 6 MODLD, 1 BITD, 5 BIDLD) and controlling for NVIQ and SES, the main effect of Impairment Status became significant (F(1,105) = 4.07, p = 0.046, η_p² = 0.04). Children with DLD responded slower than TD children. Other results, including the absence of an interaction between Trial and Impairment Status, remained similar.

3.2. Switching Costs: Effects of DLD and Bilingualism

3.2.1. Accuracy

Results showed a significant effect of Trial (F(1,120) = 56.96, p < 0.001, η_p² = 0.32) and Impairment Status (F(1,120) = 7.84, p = 0.006, η_p² = 0.06). Accuracy scores on repeat trials were significantly higher than scores on switch trials, and children with DLD had lower accuracy scores than TD children. Other effects were not significant (Language Group: F(1,120) = 0.51, p = 0.48, η_p² = 0.00; Language Group*Impairment Status: F(1,120) = 0.42, p = 0.52, η_p² = 0.00; Trial*Impairment Status: F(1,120) = 0.99, p = 0.32, η_p² = 0.01; Trial*Language Group: F(1,120) = 2.48, p = 0.12, η_p² = 0.02; Trial*Impairment Status*Language Group: F(1,120) = 0.14, p = 0.71, η_p² = 0.00). This means that the accuracy scores of the groups of children were not differently affected by the two types of trials. The results did not change when controlling for SES and NVIQ, with NVIQ being a significant covariate and SES not.

3.2.2. Reaction Times

Results showed a significant effect of Trial (F(1,120) = 88.76, p < 0.001, η_p² = 0.43). RTs on repeat trials were significantly lower than RTs on switch trials. There was also a significant effect of Impairment Status (F(1,120) = 4.42, p = 0.04, η_p² = 0.04), showing that children with DLD responded slower than TD children. Other effects were not significant (Language Group: F(1,120) = 3.39, p = 0.07, η_p² = 0.03; Language Group*Impairment Status: F(1,120) = 0.01, p = 0.93, η_p² = 0.00; Trial*Impairment Status: F(1,120) = 1.92, p = 0.17, η_p² = 0.02; Trial*Language Group: F(1,120) = 0.73, p = 0.39, η_p² = 0.01; Trial*Impairment Status*Language Group: F(1,120) = 1.48, p = 0.23, η_p² = 0.01). The RTs of the groups of children were thus not differently affected by the two types of trials. The results did not change when controlling for SES and NVIQ. Neither NVIQ nor SES were significant covariates. When excluding children with low accuracy scores in the task-switching block (below 60,7% correct; 1 MOTD, 6 MODLD, 1 BITD, 5 BIDLD), the main effect of Language Group became significant (F(1,120) = 4.98, p = 0.03, η_p² = 0.04), but this significant effect disappeared when controlling for NVIQ and SES (F(1,120) = 3.35, p = 0.07, η_p² = 0.03). Other results remained similar.

3.3. Mediation Effects: Mixing Costs

To understand the higher mixing costs of children with DLD, we conducted a mediation analysis. We first examined the correlations between mixing costs (i.e., accuracy scores on the repeat trials of the task-switching block minus accuracy scores on the single-task blocks) and the number of false alarms and number of hits on the CPT, respectively measuring response inhibition and sustained attention. Mixing costs were significantly related to both response inhibition (r = 0.27, p = 0.002) and sustained attention (r = −0.33, p < 0.001). Subsequently, we investigated relationships between Impairment Status (X), mixing costs (Y), and the mediators response inhibition and sustained attention (M).

Figure 2 presents the results of the mediation analysis. The results indicate that response inhibition and sustained attention mediated the effect of Impairment Status on mixing costs. The effect of Impairment Status on mixing costs did not remain significant when response inhibition and sustained attention were controlled for. The indirect effect of Impairment Status on mixing costs through both mediators accounts for 45.9% of the total effect of this relation. The results of the individual mediators show that the index of mediation (the standardized indirect effect) was larger for sustained attention (β = 0.20, 95% CI [0.01, 0.44]) than response inhibition (β = 0.08, 95% CI [−0.001, 0.22]). Only sustained attention showed a meaningful mediation effect (next to the total mediation effect of the two mediators together).

4. Discussion

While bilingualism is associated with enhanced switching skills, having a developmental language disorder (DLD) may negatively impact switching ability. However, previous work on bilingualism and/or DLD in relation to switching has revealed mixed results. Furthermore, the fields of bilingualism and DLD have largely operated separately, and it is thus unknown if and how switching ability is affected by the interaction of the two. In the present study, we first aimed to investigate not only the independent effects of bilingualism and DLD on nonverbal switching ability but also the interaction effect. Second, we aimed to elucidate the origin of (potential) strong or poor performance on a nonverbal switching task of, respectively, bilingual children and children with DLD. We adopted a four-group design to study children’s nonverbal switching ability and used a longitudinal approach to investigate whether sustained attention and response inhibition skills mediated the possible effects of bilingualism and/or DLD on switching ability.

Our findings provide clear evidence for overall performance costs associated with switching on our cued color–shape switching task. However, these performance costs were not found to be different for monolingual and bilingual children, irrespective of whether NVIQ and SES were statistically controlled for. While this corresponds to other studies that did not find evidence for a bilingual benefit on switching (Paap et al. 2017; Timmermeister et al. 2020) and on EF in general (de Bruin et al. 2015; Duñabeitia et al. 2014; Gathercole et al. 2014), it is not in line with our hypothesis based on previous work which supported a bilingual switching advantage (for a meta-analysis on child studies, see Gunnerud et al. 2020). It hereby adds to the body of research that questions the robustness of such an advantage (Paap et al. 2018; Poarch and Krott 2019). Even between the studies using a similar task-switching paradigm with children of comparable ages, mixed findings are observed, in favor of (Antoniou et al. 2016; Barac and Bialystok 2012) and against a positive bilingual effect (Timmermeister et al. 2020; the present study). Unlike Antoniou et al. (2016), we did not control for language ability, which could thus be a factor in explaining the different findings. However, Barac and Bialystok (2012) did not control for language ability either, while Timmermeister and colleagues did, suggesting that other factors play a role in moderating the bilingual switching advantage (for a review on moderating factors of EF in general, see Marton 2016). It was beyond the scope of the current study to investigate such factors, but future work, both in the context of typical development as well as in the context of developmental disabilities, is recommended to do so to offer more clarity on the nature, breadth, and relevancy of cognitive advantages of bilingualism. Such research could consider the impact of home language maintenance and proficiency, as well as the relative balance between both languages and switching frequency (see, e.g., Kuzyk et al. 2020; Verhagen et al. 2020).

Our findings, furthermore, demonstrated higher mixing costs for children with DLD than for TD children. Children with DLD had lower accuracy scores on repeat trials of the task-switching block than TD children, whereas there were no differences between the groups on trials in the single-task blocks. No differences between the groups emerged in terms of switching costs. The higher mixing costs confirm earlier meta-analytical findings of poorer switching ability of children with DLD (Aljahlan and Spaulding 2021; Pauls and Archibald 2016) and thereby correspond to our hypothesis. Moreover, this finding highlights the view that problems of children with DLD extend beyond the linguistic domain (Botting and Marshall 2017; Kapa and Plante 2015). We did not find any interactions between bilingualism and DLD in our task-switching paradigm. Thus, bilingualism did not differently affect the switching skills of children with and without DLD, and DLD did not have a differential impact on monolinguals and bilinguals. As the first investigation of this interaction in the domain of switching, the current study supports results from previous work on other EF components (see Boerma and Blom 2020) and attention skills (Ebert et al. 2019).

Next to studying both the independent and interaction effects of bilingualism and DLD on nonverbal switching performance, the present study explored what could be driving the observed group effects. We performed a mediation analysis with response inhibition and sustained attention skills as mediators to further elucidate the higher mixing costs of the children with DLD on the nonverbal switching task. Response inhibition and sustained attention accounted for more than 45% of the relation between DLD and mixing costs. This finding supports our expectations and previous work which considers switching to be a complex EF component that builds on lower-level cognitive processes, such as response inhibition and sustained attention (Dajani and Uddin 2015; Garon et al. 2008).

Comparing the individual mediators, we found that the effect of DLD on mixing costs was particularly driven by children’s ability to sustain their attention. This agrees with our predictions based on Braver et al. (2003), as mixing costs are thought to reflect the ability to maintain one’s attention to keep multiple tasks active. Response inhibition was thought to more strongly reflect switching costs (Druey and Hübner 2008; Vandierendonck et al. 2010). Although response inhibition is also weak in children with DLD, sustained attention deficits may be more severe (Kapa et al. 2017), possibly explaining why children with DLD were found to have higher mixing costs than TD children while switching costs were not different between the groups. In addition, the important role of sustained attention in our data substantiates the hypothesis that attention processes are the foundation for more complex EF skills (Garon et al. 2008), such as switching, and suggests cascading effects for children with poorly developed attention skills, such as children with DLD. These cascading effects have previously been shown to affect the language skills of children with DLD (Boerma et al. 2017; Blom and Boerma 2016), and are now also shown to be implicated in their EF development.

These findings were observed in an experimental setting, and it is important to consider what their implications may be for the daily lives of children with DLD. We showed that it is not necessarily problematic for children with DLD to frequently switch between tasks, as reflected by their switching costs, which were not different from TD children. Instead, it seems particularly difficult for children with DLD to maintain their attention on multiple tasks and thereby accurately satisfy these tasks’ goals, as reflected by the relatively high mixing costs. It can be hypothesized that this difficulty affects, for example, academic achievement, including reading development. When learning to read, a child needs to be able to sustain his/her attention on both the phonological form of the word or text as well as on the meaning. Indeed, previous studies indicate that switching ability is implicated in literacy (Lubin et al. 2016; Yeniad et al. 2013), although this has not yet been confirmed in children with DLD. Future studies are needed to translate findings from experimental EF tasks to the real-life challenges, such as learning to read, of children with DLD. Moreover, these studies could investigate the relations with real-life language switching experience in bilinguals, both in terms of addressing speakers of different languages as well as code-switching among other bilingual speakers, and nonverbal switching ability as measured by an experimental task (see, e.g., Hartanto and Yang 2020). Such experience may be an important factor in explaining variability in nonverbal switching ability and, in turn, in explaining the presence or absence of a bilingual advantage.

In addition, future research is necessary to learn more about the development of switching ability in children with DLD. The current study was longitudinal, which enabled us to better understand the causal mechanisms of poor switching ability, but we could only use our switching data from one time point when children were 7 to 8 years old. One year earlier, the switching task was still too complex for many children, especially for the children with DLD. While this is in and of itself informative, as other studies with only TD children were able to use data from 6-year-old children with a similar task-switching paradigm (Barac and Bialystok 2012), it did not allow us to measure switching ability at this younger age nor switching development over time. Additionally, the complexity of the task complicated our analyses with reaction times, as a relatively large proportion of data points (i.e., incorrect responses) had to be excluded. Future longitudinal work with children with DLD needs to take into account that, particularly when tapping into such complex skills as switching, tasks should be age-appropriate for all participants, including the children with DLD. Moreover, in work with older children, the inclusion of more trials in the switching task could be considered to increase reliability.

In conclusion, our findings do not point to performance differences between monolingual and bilingual children on a nonverbal cued switching task and indicate that the effect of bilingualism on nonverbal switching is not different for children with and without DLD. The absence of a bilingual advantage in the current study is in line with previous work which indicates that such advantages are not robust and may depend on many moderating factors. Future research is necessary to study such factors, also in the context of developmental disabilities. Additionally, our findings show that the presence of two different tasks resulted in higher performance costs for children with DLD, either monolingual or bilingual, relative to TD children. Sustained attention was a driving factor behind this effect of DLD. These results strengthen the view that attention processes are essential for the development of complex skills, such as switching, and further highlight the importance of sustained attention in the development of children with DLD.