Bilingualism but not bidialectalism influences cognitive control among young adult Chinese–English bilinguals

Bilingualism and cognitive control

Bilingualism refers to the knowledge and use of two or more languages by an individual or a community. Relative to monolinguals, bilinguals have been found to have advantages in different components of cognitive control, for example, inhibition, conflict monitoring, and mental set shifting. Inhibition and conflict monitoring are usually measured by the Simon task and Flanker task (Bialystok et al., 2004; Costa et al., 2009; Privitera & Weekes, 2022), and mental set shifting as measured by a task switch paradigm or the Wisconsin Card Sorting Test (WCST) (Prior & Macwhinney, 2010; Xie & Dong, 2017). However, it is worthwhile to acknowledge that these tasks do not necessarily measure separate components in isolation. These cognitive control tasks might not correlate well because of problems such as low reliability, different strategy use, and task impurity (Friedman & Miyake, 2017). Even similar conflict tasks (e.g., ANT, Number Stroop) may differ in their ability to detect bilingual advantages (Xia et al., 2022).

The bilingual advantages have been thought to come from cognitive control training in bilingual language use. When bilinguals intend to use the target language, the non-target language is also activated in the brain (Hoshino & Thierry, 2011; Kroll et al., 2008). In order to succeed in preventing the interference between the two language systems, an inhibitory control system (Green, 1998) is adopted to focus on the target language and at the same time suppress the non-target language. As a result, the long-term practice of such language control enhances bilinguals’ general cognitive control abilities, so that bilinguals also perform better in non-linguistic executive tasks. It is also indicated that bilinguals have greater neural network efficiency relative to monolinguals, which relates to a greater proportion of white/gray matter (Abutalebi et al., 2015). This is thought to delay the onset of dementia among older adult bilinguals (Bialystok et al., 2007).

However, whether bilingualism enhances cognitive control remains highly debated. Previous studies have provided empirical evidence that bilinguals perform better in cognitive control tasks compared with their monolingual counterparts (e.g., Dash et al., 2019; Durand López, 2021; Pliatsikas & Luk, 2016). By contrast, other studies have failed to observe the expected advantages (e.g., M. Antoniou, 2019; Paap et al., 2018; Paap & Greenberg, 2013). Moreover, recent meta-analyses on the relationship between bilingualism and cognitive control have supported both positive (Grundy, 2020; van den Noort et al., 2019) and null results (Donnelly et al., 2019; Lehtonen et al., 2018; Lowe et al., 2021), further adding fuel to the controversy. There is also a hot discussion on the origins of the inconsistency of bilingual advantage. Some scholars propose that demographic factors should be taken into account when interpreting the results (Valian, 2015), whereas others commend that cognitive control is caused by differential interactive contexts (Green & Abutalebi, 2013) or by the lack of a suitable theoretical framework (Bialystok & Craik, 2022). All these discussions point to the idea that an individual’s cognitive control performance is a dynamic combination of multiple factors (Dash et al., 2022). Titone and Tiv (2023), based on the current situation of bilingual debate, offer a new framework, “Systems Framework of Bilingualism” and urge cognitive scientists and neuroscientists to embrace sociolinguistic and sociocultural experiences as their theoretical and empirical purview. Under this framework, what constrains cognition, behavior, and neuroplasticity of an individual should include different layers of contexts, that is, interpersonal, ecological, societal, and temporal, and these layers dynamically integrate to impose an influence on the individual. For example, at the interpersonal level, bilingual individuals may use one language (dialect or L1) with their parents and another language with their siblings (standard or L2). At the level of ecological dynamics, bilinguals may live in a neighborhood where people speak a homogeneous language (or a variation, dialect), which constrains their use of other known languages. At the societal level, the government may promote or allow only the use of the official language in a particular situation, such as in schools, offices, and courts. Following this framework, therefore, it is reasonable to speculate that the complexity and heterogeneity of a bilingual experience may incur different demonstrations of the bilingual effect under different layers of contexts (or a combination). Most previous studies, however, compared bilinguals and monolinguals in a more general sense but did not look into how a specific case of bilingual experience in a sociocultural context may influence cognitive control.

To speak a language, a bilingual should achieve a certain level of language proficiency regardless of the individual’s cognition, demographics, and how she or he interacts with others. Thus, language proficiency is an essential variable for investigating the bilingual advantage (Mishra, 2014). Without a sufficient level of language proficiency (e.g., L2 proficiency for late bilinguals whose L1 is dominant or native), or without a relatively high degree of bilingualism, it is not possible for bilingual advantage to emerge. Furthermore, as L2 proficiency improves, we might see that cognitive control changes accordingly (Wang et al., 2022). There is evidence of a significant relationship between higher L2 proficiency and better cognitive control. For example, relative to bilinguals or monolinguals with a lower L2 proficiency, bilinguals with a higher L2 were reported to have better inhibition or suppression as reflected in the Attentional Blink, Simon, and Flanker tasks (Khare et al., 2013; Kheder & Kaan, 2021; Privitera et al., 2022), better monitoring or attention control as reflected by faster reaction times (Costa et al., 2009; Luque & Morgan-Short, 2021; Mishra et al., 2012; Privitera et al., 2022; Vega-Mendoza et al., 2015; Xie & Pisano, 2019), or better shifting ability as reflected by the switching cost or WCST performance (Kheder & Kaan, 2021; Prior & Gollan, 2011; Xie & Dong, 2017). Some studies, however, reported null effects of language proficiency in enhancing cognitive control. For example, some studies showed no support of bilingual advantage when examining the role of language proficiency, but showed that the intelligence score was a better predictor of cognitive control (e.g., Diaz et al., 2021; Rosselli et al., 2016). Other studies suggest that specific language use experience (e.g., interpreting), rather than L2 proficiency, is the determinant of bilingual advantage in cognitive control (Becker et al., 2016; Dong & Xie, 2014; Verreyt et al., 2016). To sum up, in the research of bilingual advantage, the influence of L2 proficiency on cognitive control needs further verification and in what context L2 proficiency matters to the emergence of bilingual advantage remains an open question.

Bidialectalism and cognitive control

How language is used or demonstrated in the ecological, social context is various. According to the “Systems Framework of Bilingualism,” an individual may speak a dialect with parents but speak a standard language with peers, or the individual may speak the dialect with locals in the neighborhood but hear the standard language from the media. Dialect speakers recruit cognitive control mechanisms in similar ways as bilinguals (Kirk et al., 2018, 2022). Individuals who speak two dialects must also maintain separation between two linguistic systems. Similar to bilingual activation, the two dialects of a bidialectal speaker are activated simultaneously for use, so bidialectal speakers also need to control selection processes across two lexica, activating the target and inhibiting the non-target language (or its variation), and monitoring when to use the standard and when to use the regional dialect (Ross & Melinger, 2017). Therefore, it is possible that bidialectals, just like bilinguals, may have advantage in cognitive control.

The case of bidialectal language experience, or bidialectalism, refers to the ability of using two dialects of the same language. One of the crucial differences between a language and a dialect lies in mutual intelligibility (e.g., K. Antoniou et al., 2016; Francis, 2016; Lee-James & Washington, 2018; Tang & Van Heuven, 2009). Languages are not mutually intelligible and dialects are mutually intelligible in general sense. However, although some dialects are closely related varieties of the same language, they are not mutually intelligible. For example, for the Chinese language, the so-called dialects of the Chinese language in a sense are similar to separate languages (e.g., English vs. Spanish), although these variants share lexical cognates with different sounds (Hsu, 2021). The Chinese language can be generally divided into eight groups: Mandarin, Min, Yue, Wu, Hakka, Gan, Jin, and Xiang. These variations can be split into two branches: a Mandarin branch and a Southern branch (non-Mandarin groups). Within the Mandarin branch, varieties are often intelligible to each other to some extent. However, most varieties in the Southern branch are mutually unintelligible either to each other or to the Mandarin varieties (Tang & Van Heuven, 2009, 2015; Yan, 2006; Zhang & Zhang, 2010). Like distinct but related languages, there are great differences in orthography, pronunciation, vocabulary, and grammar between these different dialects, and some dialects are speech unintelligible (Zhang & Zhang, 2010). As many Western scholars have observed, cross-dialectal communication in China is often seriously hampered by intelligibility problems. Chinese varieties are really more like discrete languages than dialects of the same language (Bruche-Schultz, 1997; Li, 2006). In complex dialectal regions, dialect can be further divided into several sub-dialect regions within each dialect, as in the case of speakers of subdialects of Min in Fujian province. Likewise, in Gan dialect region, it can be further divided into different areas: Nanchang region, Ji’an region, Fuzhou region, Shangrao region, Hakka region, and so on (Sagart, 2002; Xu et al., 2018). Therefore, Chinese dialects, from Cantonese through Hakka to Mandarin, vary as much as the Germanic languages such as English, German, and Swedish (Erbaugh, 2001).

However, whether bidialectal experience is to enhance cognitive control receives little attention. Stemming from the sociocultural constraints on using the dialects in different contexts, bidialectal speakers theoretically share a similar burden to bilingual speakers (Ross & Melinger, 2017). Only a few studies have compared the differences between bidialectals and monolinguals on cognitive control. Among these studies, some have failed to capture the bidialectal advantage, such as in Alrwaita, Meteyard, Houston-Price, & Pliatsikas (2022), where Arabic diglossic young adults and English monolinguals were compared on cognitive control tasks measuring inhibition and switching; in Kirk et al. (2014), where no evidence was found for reduced Simon cost among elderly bilinguals and bidialectals; and in Ross and Melinger (2017), where bidialectal and monolingual children were compared on the Simon and Flanker tasks. In Wu et al. (2016), young adult Chinese Mandarin monolinguals and Mandarin-Southern-Min bidialectals were compared on the Flanker task, and the results revealed no behavioral differences between groups. Similarly, Privitera and Zhou (2022) investigated whether differences in self-reported dialect proficiency impacted on the Simon task performance among a heterogeneous sample of Mandarin–English bilingual, bidialectal young adults, and their results showed no cognitive advantage. However, several studies have partially identified the bidialectal advantage. K. Antoniou et al. (2016), for example, compared bidialectal children, multilingual/bilingual children, and monolingual children on different components of cognitive control (working memory, inhibition, and shifting), and they found that both multilingual and bidialectal children exhibited an advantage over monolinguals in working memory and inhibition. The study by K. Antoniou and colleagues (2016) was limited by the fact that their bilingual/multilingual sample was also bidialectal, so it is not clear whether the advantage comes from bidialectalism or bilingualism, or both. In a later study, Hsu (2021) compared older Mandarin monolingual and Minnan–Mandarin/Hakka–Mandarin bidialectal adults on multiple nonverbal and verbal cognitive tasks via two experiments, and they found that the bidialectal advantage was found in the tasks of nonverbal Stroop Color-word, verbal Stroop Color-word, and Stroop day–night. However, these tasks are still relevant to the linguistic domain as these tasks used verbal stimuli. In the other three nonverbal tasks (Flanker, Simon Color and Space, and Spatial 1-back task), nevertheless, no bidialectal advantage was found. Kirk et al. (2022) examined language control in regional dialect speakers and found typical switch cost asymmetry and mix cost through a language-switching task, which suggests a similar pattern of unbalanced bilinguals during lexical access. In short, these limited but inconsistent studies indicate that whether bidialectalism (which resembles bilingualism) enhances non-linguistic cognitive control needs further exploration.

The current study

Based on our review above, both bilingualism and bidialectalism may have an effect on cognitive control. As presented in the “Systems Framework of Bilingualism,” it is possible that bilingualism and bidialectalism are intermingled with each other in ecological language use context. For example, for individuals from China, some people acquire local dialect in their early years as their native language, and then learn Mandarin in kindergarten or school, but others acquire Mandarin as their native language, and then formally learn English as a second/foreign language in school. It is also common that some people acquire the local dialect first, and then learn Mandarin and English in school, making them an integration of both bilingualism and bidialectalism.

In China, the overall dialect use in daily life is varied. In public places, people usually speak Mandarin (80.49%). It is reported that in general, 29.42% of the population speak Mandarin as the only native language, whereas 31.71% of the population speak a dialect as the only native language, and 35.21% of population take both dialect and Mandarin as native languages. Bidialectal speakers usually speak Mandarin with young people in their daily life (48.48%), but they speak a dialect when they interact with their parents and grandparents (62.80%–62.20%). ¹ However, how these different bilingual/bidialectal experiences may influence cognitive control, respectively, and how these two variables may interact in incurring such effect remains to be scrutinized.

Therefore, by taking both bilingualism and bidialectalism into consideration, the current study aims to investigate how these different bilingual experiences may incur a differential effect on cognitive control, and whether there is an interaction effect. First, we compared two groups of Chinese–English bilinguals who differed in L2 (English) proficiency, with participants’ dialectal background taken into account, aiming to verify the bilingual advantage. Second, the two groups differing in L2 proficiency were further divided into groups based on the dialect proficiency, so that both variables of L2 proficiency and dialect proficiency were examined. Furthermore, such design allows us to examine whether there is interaction between the two. As cognitive control is modulated by a series of linguistic and demographic variables, our studies strictly controlled all these confounding variables, for example, language background, age, intelligence quotient (IQ), socioeconomic status (SES), and so on.

Participants

Seventy young adult Chinese–English bilinguals (mean age = 18.79 years, SD = 0.75, range = 17–21) voluntarily took part in the study either for monetary compensation or for course credits. All the participants were right-handed, reported normal or corrected to normal vision, and had no speech or hearing disorders. Participants gave informed consent and their rights were protected in accord with the ethical standards of the Academic Committee of Jiangxi Normal University. The participants were divided into two groups according to their tested English scores and self-reported L2 proficiency, and therefore we created two groups differing in L2 proficiency. The High L2 Proficiency Group (n = 37) was composed of English major students from Jiangxi Normal University in China, who reported higher self-rated L2 proficiency and achieved higher English (L2) scores in the College Entrance Examination (mean score = 129.61 out of 150, SD = 5.85, range = 118–140). The Low L2 Proficiency Group (n = 33) were non-English-major students from Nanchang Institute of Technology in China, who reported much lower self-rated L2 proficiency and achieved a significantly lower English (L2) score in the College Entrance Examination (mean score = 54.24 out of 150, SD = 16.87, range = 21–92). Furthermore, as English major students, the high L2 proficiency group spent more time in L2 learning and used the L2 more extensively than the low L2 proficiency group. According to their class schedules, English-major students took around 16 hours of in-class English learning and using per week, whereas the non-English-major students took only about 3 hours of in-class English learning and using per week.

Materials and procedure

Data trimming

In the Flanker task, data from erroneous responses were excluded, and data with RTs above three standard deviations (SDs) of the overall mean for each subject in each condition in the task were eliminated, accounting for 2.55% of the total. In the WCST, one participant was removed due to missing data in the WCST. Data with RTs above three SDs of the overall mean for each subject were excluded. Finally, four participants’ data were excluded in order to match the confounding variables such as SES, IQ, and dialect proficiency between groups, with 33 participants left in the low L2 group and 32 participants in the high L2 group.

Demographic and linguistic background

Details of the background information are presented in Table 1, including self-rated L1 and L2 proficiency, self-rated dialect proficiency, tested L2 score in College Entrance Examination, and demographic factors such as age, years of education, IQ, and SES. As Table 1 shows, the independent sample t-test results showed that there were no group differences between the two groups in age, education, SES, IQ, dialect proficiency, and Chinese proficiency (L1) (ps > .108). However, the two groups differed significantly in (self-rated) L2 proficiency, t (63) = –21.827, p < .001, and English (L2) score in the College Entrance Examination, t (63) = –20.610, p < .001. Therefore, if there are differences in the cognitive control tasks performance between groups, we could reasonably attribute the differences to L2 proficiency variances.

OPEN IN VIEWER

Table 1.

Linguistic and demographic background of L2 proficiency groups.

	Low L2 group (n = 33)		High L2 group (n = 32)		t-testp value
	M	SD	M	SD
Age	18.9	0.7	18.7	0.8	.232
Education	12.8	0.6	13.0	0.4	.108
SES (1–5)	2.5	1.0	2.5	1.1	.999
IQ (0–72)	64.2	4.1	64.5	5.3	.768
Dialect proficiency (1–20)	14.7	5.1	15.0	4.5	.821
Chinese (L1) proficiency (1–40)	34.2	6.3	35.3	3.1	.417
English (L2) proficiency (1–40)	6.6	3.3	25.4	3.7	.000
English Examination (0–150)	54.2	16.9	127.2	11.2	.000

Note. SD: standard deviation; SES: socioeconomic status; IQ: intelligence quotient; bold values indicate that group difference is significant at .05 level.

The participants reported different dialect backgrounds or different dialect variations in the same dialect regions. Of all bidialectals, 52.5% speak the Jiangxi (Gan) dialect, 15% the Shandong dialect, 15% the Henan dialect, 5% the Hunan dialect, 7.5% the Anhui dialect, 2.5% the Zhejiang dialect, and 2.5% the Shanxi dialect. All participants reported to have acquired dialects in early childhood from their parents and tended to use dialects at home and in the local bidialectal community. In addition, correlation between self-reported L2 proficiency and English (L2) examination was high, R = .882, p < .001, but correlation between L1 and dialect proficiency is not significant, R =–.038, p = .756, indicating that different things were measured.

Flanker task

The data for the Flanker task performance between groups are presented in Table 2. In order to examine whether there are differences across conditions and between groups, we conducted a repeated-measures analysis of variance (ANOVA) with Group (two participant groups) as between-subject variables and Condition (congruent, neutral, and incongruent) as within-subject variables. This analysis revealed a significant main effect of Condition, F (2, 126) = 225.381, p < .001, η² = .782, but there was no Condition and Group interaction, F (2, 126) = 0.393, p = .676, η² = .006, indicating that there were differences among the three conditions of the task but these differences were similar across the groups. Specifically, planned comparisons showed that all participants responded more quickly in the congruent condition (508 ms) than in the neutral (521 ms), F(1, 63) = 27.104, p < .001, η² = .301, and the incongruent conditions (565 ms), F(1, 63) = 382.958, p < .001, η² = .859. Participants also responded more quickly in the neutral condition (521 ms) than in the incongruent condition (565 ms), F(1, 63) = 226.779, p < .001, η² = .783.

OPEN IN VIEWER

Table 2.

Task performances between L2 proficiency groups.

	Low L2 group (n = 33)		High L2 group (n = 32)		t-testp value
	M	SD	M	SD
Flanker task
Congruent condition	540	93	474	61	.001
Neutral condition	551	86	490	57	.002
Incongruent condition	596	90	532	56	.001
Flanker effect	56	22	58	24	.817
WCST
RTs	2,084	979	1,501	493	.004
Completed category	7.8	3.3	6.8	2.9	.213
Overall errors	50.6	15.8	56.2	12.7	.122
Perseverative errors	29.0	15.5	34.6	13.5	.124
Previous category errors	13.1	13.1	13.9	8.0	.781

Note. SD: standard deviation; WCST: Wisconsin Card Sorting Test; RT: response time; bold values indicate that group difference is significant at .05 level.

However, what is more important to this study is that the main effect of group was significant, F (1, 63) = 11.534, p = .001, η² = .155. Independent-samples t-test showed that there were no group differences on the Flanker effect (p = .817), but the high L2 proficiency group was faster than the low L2 proficiency group in all the three conditions: congruent t (63) = 3.328, p = .001, Cohen’s d = .839; neutral t (63) = 3.313, p = .002, Cohen’s d = .836; incongruent t (63) = 3.425, p = .001, Cohen’s d = .854. (ps < .01), indicating that greater L2 proficiency is significantly related to faster RTs in the Flanker task, reflecting conflict monitoring advantage in cognitive control.

WCST

Data of the two groups’ performances in WCST are presented in Table 2. Independent-samples t-test was conducted to compare the differences between the two groups. The results showed that there were no significant group differences on completed category, overall errors, perseverative errors, or previous category errors (ps > .122), while the group difference on RTs was significant, t (63) = 3.364, p = .004, Cohen’s d = .752. Specifically, the group with high L2 proficiency performed much faster in the WCST than the group with low L2 proficiency, reflecting better ability of mental set shifting.

Interaction effect of bilingualism and bidialectalism

As mentioned above, in our study we matched the dialect background between the two bilingual groups, who differed in L2 proficiency. However, this does not rule out the potential effect of bidialectalism and whether there is interaction between bilingualism and bidialectalism. Therefore, we take the bilingualism and bidialectalism as two independent factors in our further data analyses. The participants were divided into two groups according to their dialect level. The mean of low dialect group (29) was 10.5 (SD = 3.8), and the mean of high dialect group (36) was 18.3 (SD = 1.5). The two groups differed significantly in the dialect proficiency, t (63) = −11.228, p < .001. Therefore, the participants were categorized into four groups: high dialect proficiency group with low L2, high dialect proficiency group with high L2, low dialect proficiency group with low L2, and low dialect proficiency group with high L2. The Flanker task and the WCST performances across groups are shown in Table 3.

OPEN IN VIEWER

Table 3.

Task performances across groups by L2 and dialect proficiency.

	Low dialect group		High dialect group
	Low L2 (n = 14)	High L2 (n = 15)	Low L2 (n = 19)	High L2 (17)
Flanker task
Congruent	533 (101)	451 (50)	545 (89)	496 (64)
Neutral	542 (92)	470 (45)	558 (83)	509 (63)
Incongruent	588 (102)	508 (47)	602 (83)	554 (57)
Flanker effect	55 (24)	57 (22)	57 (22)	58 (27)
WCST
RT	2,047 (629)	1,459 (587)	2,111 (1,189)	1,538 (406)
Completed categories	7.28 (3.36)	6.53 (3.39)	8.16 (3.32)	7.06 (2.51)
Overall errors	52.64 (16.2)	57.60 (13.28)	49.16 (15.81)	55.00 (12.4)
Perseverative errors	30.8 (16.3)	35.3 (14.3)	27.6 (15.2)	34.1 (13.2)
Previous category errors	16.1 (17.0)	14.0 (8.2)	11.0 (9.3)	13.8 (8.1)

Note. WCST: Wisconsin Card Sorting Test; RT: response time.

To examine the effect of bilingualism and bidialectalism and their interaction effect on cognitive control, we conducted two-factor general linear model ANOVA analyses. The results of the Flanker task performance revealed a significant effect of L2 proficiency in all three conditions [congruent F (1, 61) = 11.228, p = .001; neutral F (1, 61) = 10.946, p = .002; incongruent F (1, 61) = 11.782, p = .001]. However, there was no bidialectalism effect on any indicator of the Flanker task (ps > .110). Moreover, there was no interaction between bilingualism and bidialectalism (ps > .402). The results of the WCST performance showed that L2 proficiency had effect only on the RTs of the WCST performance, F (1, 61) = 8.649, p = .005. However, there was no effect on other indicators (ps > .141). In addition, there was no effect of bidialectalism on any indicators of the WCST performance (ps > .379).

In order to further verify the roles of L1 proficiency, L2 proficiency (scores of English examination), dialect proficiency, and demographic factors in cognitive control, we conducted multiple step-wise regression analyses by taking those factors as continuous variables. The results showed that L2 proficiency and SES significantly predicted RTs of congruent and neutral conditions in Flanker task [congruent: R = .495, adjusted R² = .221, F(2, 62) = 10.083, p < .001; neutral: R = .501, adjusted R² = .226, F (2, 62) = 10.368, p < .001], and only L2 proficiency significantly predicted RTs of incongruent condition in Flanker task and RTs in WCST [incongruent: R = .460, adjusted R² = .199, F(1, 63) = 16.926, p < .001; WCST: R = .307, adjusted R² = .080, F (1, 63) = 6.541, p = .013]. All other variables were excluded in the model. These results reveal that bilingualism (indicated by L2 proficiency) contributed significantly to cognitive control, whereas bidialectalism (indicated by dialect proficiency) did not.

L2 proficiency and cognitive control

The results in the current study are mixed. The result that the high L2 proficiency group performed faster than the low L2 proficiency group in the Flanker task is consistent with some previous studies that compared bilinguals differing in L2 proficiency (i.e., Kheder & Kaan, 2021; Mishra et al., 2012; Xie & Pisano, 2019). However, in the current study, the high L2 proficiency group did not outperform the low L2 proficiency group in inhibition (reflected by the Flanker effect—RT differences between congruent and incongruent trials), which is not consistent with Green’s (1998) Inhibitory Control Hypothesis. The hypothesis suggests that bilinguals are better at inhibiting irrelevant information or responses and, thus, have better inhibition or conflict resolution skills than monolinguals, which has been proved by some studies showing that bilinguals have better performance in tasks requiring inhibitory control than monolinguals (e.g., Bialystok et al., 2004; Crivello et al., 2016). However, according to Hilchey and colleagues, it is concluded that a bilingual advantage in inhibition is scarce (Hilchey et al., 2015; Hilchey & Klein, 2011).

This null advantage of inhibition may be related to the lack of theoretical construct of “inhibitory control.” Some scholars have proposed that inhibition is one of the core inseparable components of other aspects (Miyake & Friedman, 2012), so it is difficult to find inhibition advantage. Or, bilingual advantage may be better explained in “attentional control” framework than “inhibitory control” model in that the attentional control provides a more satisfactory account for incoherent findings that cannot reasonably be attributed to inhibition. It is suggested that group differences will not emerge until the attentional demands of a task exceed the control abilities of the tested groups, regardless of the cognitive control components involved (Bialystok, 2017; Bialystok & Craik, 2022). In addition, for the Flanker task, it is possible that the age of our sample prevented inhibitory control advantages from being identified, as they were young adults whose cognitive control ability was at peak.

In the WCST, there were group differences across the two groups only in the global RTs, but not in completed category, overall errors, perseverative errors, or previous category errors, which partially provides the evidence that higher L2 proficiency contributes to mental set shifting, although some scholars suggested that the processing advantage in WCST is related to monitoring (Czapka & Festman, 2021). However, previous research showed that bilinguals performed better than monolinguals in the aspect of switching in a similar Dimensional Card Sorting Task (i.e., Bialystok & Martin, 2004; Carlson & Meltzoff, 2008). Specifically, some studies showed that bilinguals who switched languages (e.g., interpreting) more often had a higher ability in mental set shifting than those who switched less (i.e., Bialystok & Poarch, 2014; Dong & Xie, 2014; Xie & Antolovic, 2022; Yudes et al., 2011). In the current study, our participants did not have intensive language switching experience such as interpreting, although there are situations in which bilinguals switch from Chinese to English or vice versa in class.

Dialectal proficiency and cognitive control

In the current study, we failed to capture the bidialectal effect, although this study expands the perspective from L2 proficiency to dialect proficiency. The finding is consistent with some previous studies that compared the bidialectals and the monolinguals (e.g., Blom et al., 2017; Kirk et al., 2014; Ross & Melinger, 2017). The null effect is similar to the findings of the study by Privitera and Zhou (2022), which investigated the bidialectal effect among Mandarin–English bilingual/bidialectal young adults through Simon task. However, in Hsu (2021), Mandarin monolinguals and Minnan–Mandarin/Hakka–Mandarin bidialectals were compared on the performance of nonverbal and verbal cognitive tasks. The bidialectal advantage was found in one nonverbal task (Stroop Color and Word) and two verbal task (Stroop Color and Word and Stroop day–night), but in their study the sample consisted of older adults, whereas in our study the sample consisted of young adults.

Whether or not bidialectals have cognitive advantage may depend on how they use the language in a particular context. It is suggested that bidialectals do not have advantage in cognitive control because they usually do not have as much intensive switching experience in the conversational context as bilinguals (Alrwaita, Houston-Price, & Pliatsikas, 2022; Blom et al., 2017), which is actually coherent with the “Systems Framework of Bilingualism.” Individuals with two dialects typically use one dialect in one context (e.g., at work) and another in another context (e.g., at home) (Alrwaita, Houston-Price, & Pliatsikas, 2022), which is a typical single language context (Green & Abutalebi, 2013), where cognitive control is least needed. Our bidialectal participants who speak a dialect and Chinese Mandarin also belong to this category, as they reported using dialect only when they talk to their parents and grandparents, whereas in school they usually speak Mandarin. This conversational context limited their exposure to dialect and limited their switching experience between dialect and Mandarin. We speculate that this particular ecological, social, and cultural conversational context might have restrained the emergence of cognitive advantage.

Several other factors might also explain the null effect reported here. First, although our study adopted reliable widely used tasks measuring core aspects of cognitive control, for example, Flanker for inhibition (Privitera & Weekes, 2022), the adoption of only two tasks (i.e., Flanker, WCST) cannot show us the full picture of potential cognitive advantage in bidialectals, as cognitive control is a complex mental process comprising different aspects. Moreover, it is suggested that bidialectal advantages are more likely to be found in some tasks involving more attention and inhibitory control and had a higher or intermediate level of task difficulty (Hsu, 2021). Second, our participants’ dialect variation from Mandarin was not great enough for the emergence of advantage. As suggested by Ross and Melinger (2017), the more different the regional dialect is from the standard dialect, the more effort is needed to prevent interference, and the greater the demands on executive function abilities. As reported in the participants description, most of the participants (52.5%) came from the same dialect background, Jiangxi dialect, which is relatively similar to Mandarin if compared with Cantonese. In Hsu (2021), however, bidialectal advantage was observed probably because the bidialectals spoke dialects (i.e., Mandarin, Minnan, Hakka) that varied greatly. Finally, it is to note that we did not assess participants’ differences in video game or musical instrument experience given their known impact on cognitive control performance (Bermudez et al., 2009; Best, 2010), so we recognize this as a limitation of the study.

For the interaction effect between dialect proficiency and L2 proficiency, which is rarely mentioned in previous studies, the result that interaction effect was not significant for both tasks indicates that the contribution of L2 proficiency to cognitive control was not influenced by dialect proficiency. It is still possible, as we have discussed above, that bidialectal advantage may be observed if participants have a higher level of dialect proficiency or speak a more distinct dialect from the standard variation (e.g., for Chinese such as Cantonese, Hakka, and Min dialect). If both bilingual and bidialectal advantages are observed, it is more likely to find their interaction effect. Therefore, we suggest more future studies to take these discrepancies of bidialectalism/bilingualism into account when studying the cognitive advantage.