A diffusion model approach to analysing the bilingual advantage for the Flanker task: The role of attentional control processes

https://doi.org/10.1016/j.jneuroling.2016.08.002Get rights and content

Highlights

  • The study investigates the bilingual advantage on a flanker task.

  • RT distributions are analysed with the diffusion model.

  • There is an interaction between bilingualism and trial type on non-decision time.

  • Attentional processing is implicated in the bilingual advantage.

Abstract

Elderly bilingual speakers exhibit a response time (RT) advantage on tests of executive function such as the Flanker task. There is, however, a lack of consensus regarding the cognitive mechanisms underlying this bilingual advantage. We analysed Flanker task performance from elderly bilingual (N = 29, age range = 55–75) and monolingual (N = 27, age range = 53–75) speakers using Ratcliff's (1978) diffusion model, which conceptualizes decision-making as a stochastic evidence accumulation process governed by parameters with empirically validated psychological interpretations. These parameters were analysed to investigate differences in cognitive processing between bilingual and monolingual groups in flanker RT performance. A bilingual advantage on decision making onset (the non-decision time parameter) was observed. Non-decision time was shorter on incongruent flanker trials for bilingual speakers but other parameters relating to quality of evidence (drift rate) and decision criterion (boundary separation) did not differ between groups. We interpret this non-decision time cost as reflecting a process of attentional ‘filtering out’ of distracting information. We therefore contend that lifelong bilingual language experience generates enhanced attentional control for seniors.

Introduction

Bilinguals may have an advantage over monolinguals on tasks that require attention, memory and executive control (see Valian, 2015; for a review). This bilingual advantage manifests as differences in response time (RT) on tasks that require these cognitive processes. For example, bilinguals typically exhibit shorter mean RTs than monolinguals on tasks that generate response conflict such as the Flanker (Abutalebi et al., 2012, Costa et al., 2009, Luk et al., 2011), the Simon (Bialystok et al., 2004, Linck et al., 2008), the Stroop (Bialystok et al., 2008, Heidlmayr et al., 2014) and other related tasks (see Bialystok, Craik, & Luk, 2012 for a review; see also Bialystok & Barac, 2012). These effects tend to be more robust in children and seniors although they are not exclusive to these groups of participants (Zhou & Krott, 2015).

Theoretical accounts of the bilingual advantage have not reached consensus (Hilchey & Klein, 2011). One popular account of the advantage is that bilingual speech engages cognitive inhibition in order to prevent interference from non-target language items during discourse, thus making bilinguals experts in cognitive inhibition (Kroll & Bialystok, 2013). Some studies however, report no differences between bilinguals and monolingual peers in tasks that putatively require cognitive inhibition such as verbal and non-verbal Stroop tasks (Duñabeitia et al., 2014), the Attentional Network Test (Antón et al., 2014), the Wisconsin Card Sorting test, and tests of metalinguistic judgments (Gathercole et al., 2014). Inconsistent findings have led some to question the veracity of the bilingual advantage (Paap, Johnson, & Sawi, 2015), and is the basis of a lively debate that is current in the field of bilingual language processing (compare discussion between Bialystok, Kroll, Green, MacWhinney, & Craik, 2015, and de Bruin, Treccani, & Della Sala, 2015).

Inconsistent results across studies may be due to a range of factors. Prior and Gollan (2011) argue that variability in the language background of bilingual speakers makes the bilingual advantage elusive. For example, control over inhibitory processes involved in suppression of a non-target language(s) will vary according to whether one language is more dominant, as this would generate less inhibitory control compared to bilinguals who constantly switch between their languages in daily life (e.g. Green & Abutalebi, 2013). Valian (2015) meanwhile argues the bilingual advantage depends on age and notes that it is more common in seniors (Bialystok et al., 2004, Zahodne et al., 2014). She suggests that bilingualism is one of many ‘cognitively engaging’ activities which may provide a cognitive advantage; seniors in particular may be less likely to participate in many other activities of this kind, and thus the benefits of bilingualism appear more salient in older populations. The issue of when and how the bilingual advantage emerges is of considerable theoretical importance, but a deeper problem is the lack of a coherent account of the cognitive processes leading to the bilingual advantage. The behavioural studies used in the bilingual advantage literature typically utilize tasks that necessarily involve different cognitive processes to varying degrees (e.g. monitoring, inhibition, shifting; Miyake et al., 2000, Friedman et al., 2008, Miyake and Friedman, 2012) thus making it hard to identify the exact process(es) which drive the advantage. Indeed, Paap and Greenberg (2013) demonstrated that performance on the tasks that elicit the bilingual advantage do not correlate with each other, highlighting the multidimensional nature of the cognitive processing involved in these tasks, and the difficulty in identifying a common underlying factor for the bilingual advantage.

The difficulty in isolating specific cognitive processes from multidimensional tasks such as the flanker is further corroborated by the current methods of data analysis utilized in the bilingual advantage literature. Typically, the tasks used to interrogate the bilingual advantage generate RT data and errors, and analysis is often limited to the statistical comparison of mean RTs. While this provides a good description of overall performance across groups, it does not, by itself, identify the cognitive processes that drive group differences. Consequently, the cognitive processes that underlie the bilingual advantage remain largely underspecified (though see Costa et al. 2009), most critically within the population of bilingual seniors.

Abutalebi et al. (2015) reported an RT analysis of performance of elderly bilingual and monolingual speakers on an arrowed version of the Flanker task (Fan, McCandliss, Sommer, Raz, & Posner, 2002) and found a significant bilingual advantage in terms of faster mean RT. Furthermore, they utilized an ex-Gaussian distribution function to describe precise aspects of the shape of RT distributions where bilinguals and monolinguals differed. The ex-Gaussian distribution is a convolution of Gaussian and exponential distributions, and is defined by three parameters: μ, σ, and τ, which correspond, respectively, to the mean and variance of the Gaussian component, and the rate of the exponential component. The shape of the ex-Gaussian bears many of the characteristics of empirical RT distribution data (e.g., a steep “leading edge” and a positively skewed tail). Such correspondence permits efficient summary of RT distribution data via the parameters of the ex-Gaussian, which can reveal subtle differences in performance that can be obscured by analysis of mean RT (Heathcote et al., 1991, Luce, 1986). Abutalebi et al. (2015) found that the advantage on Flanker task performance for bilingual seniors is characterized by differences in the μ parameter on congruent trials and τ parameter for incongruent trials.

Although ex-Gaussian analyses are useful for identifying how RT distribution data differ across groups (e.g., degree of right skew), they do not readily detect the cognitive processes responsible for generating these differences (Matzke & Wagenmakers, 2009). An alternative method of analysing RT data that does permit interpretations about underlying cognitive processes involves fitting the RT distributions with a cognitive model of choice-RT, such as the diffusion model (Ratcliff and McKoon, 2008, Ratcliff, 1978). The diffusion model conceptualizes decision-making in terms of a stochastic evidence accumulation process. According to the model, decisions are driven by sequentially sampling stimulus information with each sample providing a quantity of evidence favouring one response alternative over another. Information from successive samples is accumulated over time until the process reaches a pre-defined evidence threshold set by the decision-maker, after which the appropriate response is initiated (see Fig. 1).

Choice behaviour in the diffusion model is governed by four key parameters: drift rate, boundary separation, starting point and non-decision time. Critically, for present purposes, each parameter can be associated with empirically validated psychological processes that are related to the decision-making process itself (Voss, Rothermund, & Voss, 2004). Thinking of decision-making as a stochastic accumulation of evidence for one decision or another, the starting point parameter reflects the point at which this evidence accumulation process begins, and is affected by the decision-maker's biases. Boundary separation meanwhile determines the amount of evidence that must be accumulated before a response is initiated and is determined by the cautiousness of the decision-maker. Drift rate reflects the speed of the decision process toward a decision boundary, and is affected by how strongly the sampled information provides evidence for one response or another. Finally, non-decision time encompasses the processes which are not involved in the evidence accumulation process, but also take time, including the perceptual encoding of the stimulus and the motor execution of task response. A summary of the main diffusion model parameters is presented in Table 1 below.

The diffusion model provides a complete account of choice RT data on a variety of experimental tasks (Ratcliff and McKoon, 2008, Ratcliff and Smith, 2004) and is therefore a useful alternative tool to analyse RT distribution data from bilinguals and monolinguals. Importantly, because each parameter can be mapped onto a specific decision mechanism, fitting the model to data permits inferences about the underlying cognitive processes that are affected by experimental manipulations and, how those processes might differ across participant groups. With this in mind, our primary goal is to identify the cognitive processes underlying the bilingual advantage in the RT data reported by Abutalebi et al. (2015) using the diffusion model. By comparing parameters across flanker conditions for monolinguals and bilinguals, we can characterize the bilingual advantage more precisely in terms of underlying cognitive mechanisms. It is important to note that our results are relevant to the Flanker task and seniors only, and may not generalize to other conflict tasks or age groups. While this may limit the impact of the analysis, the findings will still be relevant for isolating one or more psychological processes associated with the bilingual advantage in the Flanker task for seniors.

A limitation of the ex-Gaussian distribution is that its parameters do not uniquely correspond to any of the decision mechanisms defined by the diffusion model (Matzke & Wagenmakers, 2009). It is therefore difficult to make predictions about variation in diffusion model parameters for bilingual seniors based on results from Abutalebi et al. (2015). Fortunately however, the diffusion model has been used to investigate Flanker task performance in monolingual adults, which can inform the predictions of the present study. In particular, White, Ratcliff, and Starns (2011) attribute the flanker effect to an attentional filtering process wherein flanker information is initially processed, but has its influence over the decision process decrease with time. White and colleagues were particularly concerned with the temporal dynamics of this filtering process, and thus necessitated more complex versions of the diffusion model that allowed drift rate to take on different values over the course of an individual trial.

In the present study we use the simpler, standard version of the diffusion model as we are primarily interested in the general effects of attentional selection (as they pertain to bilingualism) whilst remaining impartial to differing accounts of the underlying selection dynamics. In the standard version of the diffusion model the time course of attentional selection is reflected in the non-decision time parameter, which is sensitive to the time course of processes that are separate from evidence accumulation, such as orienting attention to a target stimulus (e.g., Sewell, Lilburn, & Smith, 2016). Moreover, models that incorporate time-varying drift produce the same effects as allowing non-decision time to vary, making conclusions drawn from one set of assumptions relatable to those drawn from the other (Smith, Ratcliff, & Sewell, 2014). We thus expect non-decision time to vary along flanker trial type (congruent vs incongruent).

We also note however, that attentional focus is not assumed to be absolute in its selectivity (Eriksen & James, 1986), and thus even when attention is focused on a target, evidence accumulation is adversely influenced by conflicting information from distractors on incongruent trials. The time varying drift diffusion models of White et al. (2011) also take this into account, as overall drift rate in their models are determined by a combination of flanker and central target information. It is thus also expected that in our model, drift rate will vary along trial type.

With regards to bilingualism, no previous study has investigated the bilingual advantage in the context of the diffusion model. The model does however, overlap with the bilingual advantage literature, as the predictions of the model are sensitive to the effects of inhibitory processes, a popular interpretation of the bilingual advantage drawn from theories of bilingual language processing (Bialystok, Martin, & Viswanathan, 2005; Kroll & Bialystok, 2013). As we note above, it is often assumed that bilinguals inhibit the non-target language in their daily discourse in order to minimize the interference from it (Green, 1998). The exact mechanistic definition of inhibition in the context of the bilingual advantage is not well specified however (Colzato et al., 2008, Paap and Greenberg, 2013) and we thus reasoned that inhibition could affect diffusion model parameters in at least one of two ways. The first possibility is that inhibition could reduce interference from incongruent flanker information, preventing it from affecting evidence accumulation during decision-making. If the strength of inhibitory processing differs between monolingual and bilingual speakers, this would cause a difference in drift rates between groups. The second possibility is that inhibition could enhance the efficiency with which attention selectively focuses on target information—without necessarily reducing interference. If bilinguals differ from monolinguals in this aspect of inhibitory processing, bilingual speakers would be faster to attend to target items and proceed to the evidence accumulation stage, leading to a difference in non-decision time. Either outcome can inform our understanding of the bilingual advantage in the context of the Flanker task. For example, a drift, but no non-decision time effect would imply that bilinguals actively suppress conflicting information provided by flanker stimuli even though bilinguals and monolinguals are equally affected attentionally by flankers.

We thus expect the congruency effect (congruent trials faster than incongruent trials) to manifest in the drift and non-decision time parameters for both groups (cf. White et al., 2011). We also predict that there should be either drift or non-decision time (or both) interaction effects (trial type × group) depending on the nature of the bilingual advantage on the Flanker task.

Section snippets

Participants

Twenty-nine healthy bilinguals (13 males and 16 females; mean age = 63.4; standard deviation [SD] = 5.8) and twenty-seven healthy monolinguals (13 males and 14 females; mean age = 61.9; SD = 6.4) participated in the study. The bilingual participants were tested in Hong Kong while the monolingual group was tested in Milan, Italy. Monolingual controls were recruited from Milan due to the low prevalence of, and difficulty recruiting, monolingual speakers in Hong Kong. Monolinguals in Hong Kong

Diffusion model fitting

Diffusion model parameters (drift rate, non-decision time, boundary separation and starting point) were estimated by individually fitting each participant's RT distribution data. Three RT distributions (congruent, incongruent and neutral trials) were simultaneously fit per participant. Unique drift rate and non-decision time parameters were estimated for each of the flanker conditions. As there was no manipulation of reward for different response types we assumed an unbiased decision process

Discussion

The aim of the present study was to identify specific cognitive processes that underlie the bilingual advantage on the Flanker task for bilingual seniors. To this end, monolingual and bilingual RT distribution data from Abutalebi et al. (2015) was re-analysed using Ratcliff's diffusion model (Ratcliff and McKoon, 2008, Ratcliff, 1978).

Conclusion

The current analysis complements and extends the results of Abutalebi et al. (2015) by providing a cognitive process account for the bilingual advantage observed in the Flanker task in healthy elderly individuals. The diffusion model fits implicate attention as a key factor in the bilingual advantage, suggesting that some aspect of bilingualism may improve people's ability to orient or ‘zoom’ attention, particularly when there is distracting flanking information. The mechanistic account of the

Acknowledgements

This research was supported by the GRF grant 754412 awarded by the Research Grants Council of Hong Kong and Seed Grants from the University of Hong Kong. Dr. David Sewell is supported by an Australian Research Council Discovery Early Career Researcher Award (DE140100772).

References (56)

  • P.L. Smith et al.

    Modeling perceptual discrimination in dynamic noise: Time-changed diffusion and release from inhibition

    Journal of Mathematical Psychology

    (2014)
  • C.N. White et al.

    Diffusion models of the flanker task: Discrete versus gradual attentional selection

    Cognitive psychology

    (2011)
  • L. Zou et al.

    Structural plasticity of the left caudate in bimodal bilinguals

    Cortex

    (2012)
  • J. Abutalebi et al.

    Bilingualism tunes the anterior cingulate cortex for conflict monitoring

    Cerebral Cortex

    (2012)
  • S. Alladi et al.

    Bilingualism delays age at onset of dementia, independent of education and immigration status

    Neurology

    (2013)
  • E. Antón et al.

    Is there a bilingual advantage in the ANT task? Evidence from children

    Frontiers in psychology

    (2014)
  • E. Bialystok et al.

    Bilingualism, aging, and cognitive control: Evidence from the Simon task

    Psychology and aging

    (2004)
  • E. Bialystok et al.

    Cognitive control and lexical access in younger and older bilinguals

    Journal of Experimental Psychology: Learning, memory, and cognition

    (2008)
  • E. Bialystok et al.

    Publication bias and the validity of evidence What's the connection?

    Psychological science

    (2015)
  • E. Bialystok et al.

    Bilingualism across the lifespan: The rise and fall of inhibitory control

    International Journal of Bilingualism

    (2005)
  • A. de Bruin et al.

    Cognitive advantage in bilingualism an example of publication bias?

    Psychological science

    (2015)
  • L.S. Colzato et al.

    How does bilingualism improve executive control? A comparison of active and reactive inhibition mechanisms

    Journal of Experimental Psychology: Learning, Memory, and Cognition

    (2008)
  • F.I. Craik et al.

    Delaying the onset of Alzheimer disease Bilingualism as a form of cognitive reserve

    Neurology

    (2010)
  • J.A. Duñabeitia et al.

    The inhibitory advantage in bilingual children revisited: Myth or reality?

    Experimental Psychology

    (2014)
  • C.W. Eriksen et al.

    Visual attention within and around the field of focal attention: A zoom lens model

    Perception & Psychophysics

    (1986)
  • C.W. Eriksen et al.

    Information processing in visual search: A continuous flow conception and experimental results

    Perception & Psychophysics

    (1979)
  • J. Fan et al.

    Testing the efficiency and independence of attentional networks

    Journal of cognitive neuroscience

    (2002)
  • N.P. Friedman et al.

    Individual differences in executive functions are almost entirely genetic in origin

    Journal of Experimental Psychology: General

    (2008)
  • Cited by (30)

    • Linking early-life bilingualism and cognitive advantage in older adulthood

      2023, Neurobiology of Aging
      Citation Excerpt :

      Here, since our study sample included only older adults, our findings may point to a general advantage of early bilingualism against cognitive decline, without specific effects for single cognitive domains. This would be in line with the view that bilingualism contributes to one's cognitive reserve (Perani and Abutalebi, 2015; Perani et al., 2017). Importantly, the observed between-group differences are unlikely to be due to confounding variables such as years of education, premorbid intelligence, and SES since we corrected for these variables in the ANCOVA models and since our results were mostly stable after matching the samples for these confounds.

    • Task-specific bilingual effects in Mandarin-English speaking high school students in China

      2022, Current Research in Behavioral Sciences
      Citation Excerpt :

      We observed some unexpected findings that are not consistent with a bilingual advantage. First, there was no evidence of a bilingual inhibitory control advantage on the flanker task, although there was evidence of improved monitoring or ‘zooming’ of attention (Ong et al., 2017). Furthermore, performance on the Simon task provided only partial support for improved monitoring (i.e., improved performance on neutral and incongruent trials).

    • Diffusing the bilingual lexicon: Task-based and lexical components of language switch costs

      2019, Cognitive Psychology
      Citation Excerpt :

      For example, the processes that affect the onset of decision making can be differentiated from factors that determine the efficiency of the decision making process itself (Ong, Sewell, Weekes, McKague, & Abutalebi, 2017; Ratcliff, Thapar, & McKoon, 2004; Sewell, Lilburn, & Smith, 2016; Soares et al., 2018). In bilingual speakers, Ong et al. (2017) and Soares et al. (2018) reported a dissociation between performance on a binary (Flanker) task using the DM revealing that the non-decision time parameter had a privileged status - independent of other DM parameters - that was diagnostic of visual attention and also evident in brain structure. The present study is designed to analyse the entire distribution of RTs in a standard GLD task (Experiment 1) and in an alternating runs GLD task (Experiment 2) using a model of choice-RT, the DM (Ratcliff, 1978).

    • Strategic attention and decision control support prospective memory in a complex dual-task environment

      2019, Cognition
      Citation Excerpt :

      Average ongoing task and PM thresholds were lowest in the speed-important group, which also had the fastest ongoing task RTs and lowest frequency of non-response misses. Non-decision time was also shortest in the speed-important group, which is consistent with previous modeling showing reduced non-decision time in response to heightened time pressure (e.g., Dambacher & Hübner, 2015) and is typically assumed to reflect a combination of faster motor response execution and more efficient pre-decisional attentional selection (e.g., Ong, Sewell, Weekes, McKague, & Abutalebi, 2017; Ratcliff & Smith, 2010; Voss et al., 2010). Non-decision time was longest in the PM-important group, which may reflect additional processing related to the PM task that is not part of the evidence accumulation process, such as double-checking or pre-decisional visual search.

    View all citing articles on Scopus
    View full text