دانلود مقاله ISI انگلیسی شماره 38752
ترجمه فارسی عنوان مقاله

فرار از دستگیری: دوزبانگی حواس پرتی از حافظه کاری را تعدیل می کند

عنوان انگلیسی
Escaping capture: Bilingualism modulates distraction from working memory
کد مقاله سال انتشار تعداد صفحات مقاله انگلیسی
38752 2012 14 صفحه PDF
منبع

Publisher : Elsevier - Science Direct (الزویر - ساینس دایرکت)

Journal : Cognition, Volume 122, Issue 1, January 2012, Pages 37–50

ترجمه کلمات کلیدی
دوزبانگی - ارشاد از توجه - حافظه کاری - سینگلتون
کلمات کلیدی انگلیسی
Bilingualism; Guidance of attention; Working memory; Singleton
پیش نمایش مقاله
پیش نمایش مقاله  فرار از دستگیری: دوزبانگی حواس پرتی از حافظه کاری  را تعدیل می کند

چکیده انگلیسی

Abstract We ask whether bilingualism aids cognitive control over the inadvertent guidance of visual attention from working memory and from bottom-up cueing. We compare highly-proficient Catalan–Spanish bilinguals with Spanish monolinguals in three visual search conditions. In the working memory (WM) condition, attention was driven in a top-down fashion by irrelevant objects held in WM. In the Identify condition, attention was driven in a bottom-up fashion by visual priming. In the Singleton condition, attention was driven in a bottom-up fashion by including a unique distracting object in the search array. The results showed that bilinguals were overall faster than monolinguals in the three conditions, replicating previous findings that bilinguals can be more efficient than monolinguals in the deployment of attention. Interestingly, bilinguals were less captured by irrelevant information held in WM but were equally affected by visual priming and unique singletons in the search displays. These observations suggest that bilingualism aids top-down WM-mediated guidance of attention, facilitating processes that keep separate representations in WM from representations that guide visual attention. In contrast, bottom-up attentional capture by salient yet unrelated input operates similarly in bilinguals and monolinguals.

مقدمه انگلیسی

. Introduction It has been repeatedly shown that bilingualism has an impact on cognitive control mechanisms required to resolve conflicting responses to stimuli – as in Stroop-like tasks (e.g. Bialystok and Martin, 2004, Carlson and Meltzoff, 2008, Costa et al., 2009, Costa et al., 2008, Hernández et al., 2010 and Martin-Rhee and Bialystok, 2008). This impact has been indexed by two effects. First, conflict effects produced by incongruent information are larger for monolinguals than for bilinguals (e.g., Bialystok et al., 2004, Bialystok et al., 2008, Costa et al., 2008, Costa et al., 2009, Hernández et al., 2010 and Hernández et al., 2010). Reduced conflict effects have been taken as evidence that bilingualism could benefit inhibitory mechanisms required to overcome conflicting information. Second, bilinguals are overall faster than monolinguals (e.g., Bialystok, 2006, Costa et al., 2008, Costa et al., 2009 and Martin-Rhee and Bialystok, 2008). This observation has been interpreted as evidence that bilinguals are more efficient at adjusting behaviour according to the current demands. Thus, in the case of tasks involving conflicting information, bilinguals can alternate more easily than monolinguals between trials that require conflict resolution and those that are free of conflict (see Costa et al., 2009). These bilingual advantages have been associated with the use of control mechanisms that prevent linguistic interference during bilingual language processing. Although the specific language control mechanisms at play are still under debate [inhibition of the unintended language (e.g. Green, 1998) vs. selection of the intended language (e.g. Costa, Santesteban, & Ivanova, 2006)], it has been argued that these mechanisms partially overlap, functionally and anatomically, with general cognitive control mechanisms (Abutalebi and Green, 2007 and Abutalebi and Green, 2008). As a result of this, bilinguals may engage general mechanisms of cognitive control (e.g., inhibitory processes, task monitoring) more frequently than monolinguals, giving rise to the bilingual advantage in control processes. Recent studies have extended the bilingual advantage to other cognitive control processes that do not seem to involve conflict resolution, such as dual modality monitoring (Bialystok, Craik, & Ruocco, 2006), reactive inhibition (Colzato et al., 2008), and task-switching (Prior & MacWhinney, 2010). These findings raise the question of the boundaries for the effects of bilingualism on general-domain cognitive control. In the present study we explored these boundary conditions by examining the effects of bilingualism on particular bottom-up and top-down factors that modulate visual search. In the remains of the Introduction, we provide a description of the top-down and bottom-up phenomena of attentional guidance, followed by our predictions on the impact that bilingualism would have on them. 1.1. Top-down and bottom-up factors guiding visual attention The ability to guide attention to a target object can be affected by distracting stimuli that are either highly salient relative to the other elements present (bottom-up guidance; e.g. Theeuwes, 1991, Theeuwes, 1992 and Theeuwes et al., 2003) or that match items held in working memory (WM) (top-down guidance; e.g. Downing, 2000, Soto et al., 2005 and Soto and Humphreys, 2009). For example, when driving your car, your attention can be captured by signals indicating certain directions with similar names to the one you have in mind (top-down guidance). However, your attention is also captured by salient stimuli such as the sound of an ambulance or the red lights indicating sudden braking of the car just in front of you (bottom-up guidance). Experimental studies of bottom-up guidance of attention have often used the phenomenon of singleton capture. As example of this task, and highly relevant for our study, Theeuwes et al. (2003) found that visual search is disrupted by the concurrent presence of distracting stimuli that are unique in some irrelevant dimensions (i.e. singletons). In these Singleton paradigms, participants are presented with a search display composed of coloured geometrical figures, each containing a line. All lines are totally straight (distracters), except for one that is slightly tilted towards the left or the right (the target). Participants’ task consists in looking for the tilted line and indicating its direction via button press (right or left), as fast as possible. That is, the shapes and colours of the geometrical figures are completely irrelevant for the task at hand, and participants only have to ignore them. The crucial aspect of this paradigm is the inclusion of so-called singletons, namely a figure that outstands among the rest in the display. Neither the shape nor the colour of this singleton figure is repeated in any other figure, which makes the singleton perceptually salient in the display search (see Fig. 1C in Section 2 for a schematic illustration of the Singleton version of our paradigm). The perceptual saliency of the singleton makes it hard for participants to prevent this item from capturing attention during visual search. In fact, performance is facilitated (Singleton benefit) when the target line falls within the singleton figure, and disrupted (Singleton cost) when the target line falls within any other figure of the display (relative to a neutral condition where no singleton figure is present in the search display). Display sequences for each condition A: Illustration of the display sequences ... Fig. 1. Display sequences for each condition A: Illustration of the display sequences used in each condition for each type of trial (valid, neutral, invalid). (A) Examples of the WM condition. Participants should respond “Different” to the memory-probe. (B) Examples of the Identify condition. Participants should withhold the response on catch trials. (C) Examples of the Singleton condition. Figure options Effects of top-down guidance have been demonstrated in studies examining how information actively maintained in WM affects attentional guidance. Soto et al. (2005) found that visual search performance is affected by the concurrent presence of distracters that match the contents of stimuli held (WM). As in Singleton paradigms, Soto et al.’s participants were instructed to look for the tilted line (target) among all the lines inside each coloured geometrical figure in the search display. Prior to be presented with the search display, however, participants had to memorize a coloured figure (the cue) and maintain it in WM for a memory test, which came immediately after the visual search task. That is, participants were first presented with a to-be-memorized cue. Then, the search display appeared and they had to search for the tilted line. Subsequently, participants were presented with a single figure and asked whether it matched both the colour and shape as the cue they were keeping in WM. The crucial manipulation is whether the target line falls within an irrelevant coloured figure that corresponds to the cue kept in WM. If the cue kept in WM contains the target line, performance is facilitated (a WM benefit); however if the cue does not contain the target line then performance is disrupted (a WM cost) (performance is in both cases compared against a neutral condition in which the cue held in WM is not present in the search display) (see Fig. 1A in Section 2 for a schematic illustration of the WM version of our paradigm). Importantly, these effects are very much reduced when the cue presented prior to the visual search display does not need to be kept in WM. In the so-called Identify paradigm, participants are presented with exactly the same cues and search displays as in the WM paradigm, but they are not asked to keep the cue in WM. For example, they may be asked to compare the colour and shape of two initial visual cues presented consecutively within a short period of time. The search task is carried out if the cues match (as in the WM condition), otherwise no response is made (e.g., Soto and Humphreys, 2009, Soto et al., 2007 and Soto et al., 2005) (see Fig. 1B in Section 2 for a schematic illustration of the Identify version of our paradigm). Since the effects of the cue are greater in the WM paradigm compared with the Identify paradigm, it can be concluded that there is top-down WM-based cueing in addition to any effects based on bottom-up visual priming (e.g., Hernández et al., 2010 and Soto et al., 2007). 1.2. A bilingual impact on guidance of attention? The main goal of our study is to explore to what extent bilingualism has an impact on the top-down and bottom-up control of attentional guidance. We hypothesize that bilingualism may have an impact on the top-down processes involved in guidance of attention. This hypothesis is based on the cognitive consequences of the extra processes that bilingualism exerts during language processing. There is agreement that the two languages of a bilingual are simultaneously activated, both during speech production (e.g., Colomé, 2001, Costa and Caramazza, 1999, Costa et al., 1999, Green, 1998, Hermans et al., 1998 and Kroll et al., 2006) and comprehension (e.g., Spivey and Marian, 1999 and Van Heuven et al., 1998). In the case of bilingual speech production, the utterance is conceptually-driven. That is, the speaker has to map a thought (i.e., the concept of table) onto the lexical items of the intended language. However, the concept may activate the lexical forms of more than one language (e.g., both table and mesa [the Spanish word for table]). This means that bilinguals need to override distraction from the irrelevant lexical form (e.g., mesa), in order to avoid miss-selection of the lexical item and a consequent language intrusion. A similar process is imposed by bilingual language comprehension. Although in this case language processing is initiated by perceptually-driven mechanisms (i.e., recognizing spoken or written words), the lexical forms of the two languages may again be simultaneously activated. That is, accessing a concept through a written or spoken word in either language may automatically activate the lexical form of the other language too. This means that the bilingual speaker needs to apply top-down mechanisms of control during language comprehension to focus on one set of lexical representations (i.e., language A) and not the other (e.g., language B). The simultaneous activation of the two lexical forms during both bilingual language production and comprehension leads us to hypothesize that bilingual speakers may need to develop strong mechanisms of top-down guidance to keep focused on the current goal (language A) while avoiding capture from the active representations of the unintended language (language B). This hypothesis proposes that, relative to monolinguals’, bilinguals’ language processing places extra demands on top-down control. In turn this may lead to bilingual participants having a generally better ability to modulate any top-down directing of attention, in our case to avoid distraction from irrelevant information. To test this hypothesis we compared the performance of bilinguals and monolinguals in a version of the WM paradigm of visual guidance of attention used by Soto et al. (2005) – (Experiment 1a – WM condition). If indeed bilinguals have a more efficient top-down mechanism of attentional guidance, then their search performance should be less affected by the contents of WM, when compared to monolingual individuals. Note though that, since many of the components of the EC system are not well-understood, it is still unclear which specific top-down WM-mediated mechanisms are used to guide attention. Thus, we cannot more precisely predict which particular top-down mechanisms of attentional control are affected by bilingualism. Even so, we note that several authors have suggested that one mechanism of attentional control may be to compartmentalize items in WM, so only the most relevant items (i.e., the search target and not the memory stimulus, in our paradigm) gain control of behaviour by accessing WM (see Olivers, 2009). This compartmentalization process may be modulated through frontal lobe structures (Soto, Humphreys, & Heinke, 2006). From this it can be predicted that search in bilingual participants would be more guided by the task-relevant target and less by irrelevant items in WM, compared with monolingual individuals. This hypothesis also makes interesting predictions about the two main effects observed in the visual search task (WM costs and benefits). The reduced effect of irrelevant information in WM on search in bilingual participants should, in principle, be indexed by a reduction in the magnitude of the WM cost and WM benefit for bilinguals as compared to monolinguals. That is, whatever the effects of the item held in WM (e.g., whether it helps or hinders target detection, on valid and invalid trials), the effect should be reduced in the case of bilinguals since their visual search performance would be less affected by the memory item. Importantly, however, a reduction of the WM benefit for bilinguals may be much more informative than a reduction of the WM cost. This is because the WM cost involves at least two different components: attentional capture from the cue held in WM, and redirection of attention to the correct figure where the target is located (see Soto et al., 2006, for neuropsychological evidence on the effects of frontal lobe damage on impaired redirection of attention on invalid trials). In contrast, the WM benefit is more likely to reflect the initial directing of attention, given that participants would not need to further reallocate their attention on valid cue trials, since the target is already inside the attended figure. These predictions make clear that one can find experimental situations in which the increased control of attention in bilingual individuals is translated into worse performance in the task – in this case with the benefits from valid cueing being reduced. Note that, there are other studies showing that better attentional processing in bilinguals may lead to a disadvantage under certain experimental conditions. Take the attentional blink (AB) paradigm. There is evidence that a deeper engagement of attention in the processing of a first target can result in a more pronounced AB for a second target (e.g., Olivers & Nieuwenhuis, 2005). Consistent with bilinguals showing deeper engagement of attention, Colzato et al. (2008) showed that bilingual participants had a larger AB than monolingual participants. To ensure that any results in the WM paradigm reflect WM and not bottom-up priming, we also compared the performance of bilinguals and monolinguals in a version of the Identify paradigm used by Soto et al. (2005) – (Experiment 1b – Identify condition). In this condition participants identified the initial coloured figure (the cue) presented on each trial, but they did not have to hold it in memory (see above). If the better attentional control of bilingual participants is shown primarily when attention is directed through WM, not when it is cued in a more bottom-up fashion, then we expect that the effects of the cue would be comparable between bilinguals and monolinguals in the Identify condition. This result would rule out the possibility that the effect of bilingualism in the WM condition is due to contrasting effects of visual priming between bilinguals and monolinguals. Before discussing the bottom-up processes, it is important to note that the aim of the current investigation is not to compare bilinguals and monolinguals in WM capacity, but rather to assess differences in WM-based control of attention. That is, we will examine whether bilingual performance in visual search is affected by the item held in WM to a lesser extent than monolingual performance. To asses whether bilingualism has any impact on the bottom-up guidance of attention we used a version of the attentional guidance paradigm described above (Theeuwes et al., 2003). This visual search task consists in finding the only tilted line present in a display. Distracters consisted in a “singleton” item – one of the geometrical coloured figures in the display that was unique in colour and shape relative to the other figures (see above). As with top-down cueing from WM, singleton costs and benefits can be distinguished according to whether the singleton contains the target line (tilted line; benefit) or a distracter line (a straight line; cost) – (relative to the neutral baseline). Following the same logic developed when drawing hypotheses about the bilingual effect in the WM condition, bilingual modulation of the Singleton benefit would clearly indicate that bilingualism has an impact on bottom-up attentional guidance. Bilingual modulation of the Singleton cost, however, could be attributed to bilingual advantages in either bottom-up attentional guidance or to more efficient attentional disengagement and redirection. Thus, similarly to what happens in the WM condition, Singleton benefits rather than Singleton costs are more likely to reflect the initial directing of attention. When assessing whether bilingualism has an effect on bottom-up processes, one could think that this issue is already known given that bilingualism has been found to have an effect in tasks where conflicting distracters are generally processed in a bottom-up fashion. However, we believe that such studies cannot be taken as definite indication that bilingualism affects bottom-up guidance of attention. Consider for example the flanker task, in which incongruent trials contain flanker arrows pointing in the opposite direction (distracters) to a central target (e.g., ←←→←←). These distracters lead to conflict by eliciting the opposite response (left) to the target (right). As a consequence, cognitive control mechanisms need to be put at play to suppress the conflicting information provided by the flankers. Although such conflict effects arise through bottom-up processing of the distracters, we cannot determine whether the reduced conflict effects shown by bilingual participants arise because bilinguals have reduced capture by the flankers (a bottom-up account) or because bilingual individuals can effect better control over the subsequent conflict (a top-down account). Alongside differences in the control of attention, bilinguals may also show general improvements in performance relative to monolingual participants. As described above, an overall advantage for bilinguals has been observed in Stroop-like tasks (e.g., Bialystok, 2006, Costa et al., 2008, Costa et al., 2009 and Martin-Rhee and Bialystok, 2008), and this has been interpreted in terms of a more efficient monitoring system (e.g., Costa et al., 2009). Monitoring can be thought to reflect the ability of the cognitive system to calibrate itself to deal with specific task-demands through appropriate adjustments (e.g., engaging inhibitory processes). For example, in a Stroop-like context, this self calibration can be achieved on the basis of an assessment of current demands by extracting an abstract index of conflict. On the basis of the computed trial-by-trial conflict, the monitoring system may trigger cognitive control processes (e.g., the prefrontal cortex in case of inhibition) to intervene and adjust processing according to the current demands (e.g., Botvinik, Braver, Barch, Carter, & Cohen, 2001). In this scenario, a bilingual advantage in overall speed may arise in any task engaging the monitoring system. Note that similar monitoring processes may be involved here, perhaps adjusting WM control according to cue validity. Whether the overall differences in processing speed, reflecting this monitoring process, modulate the effects of (both top-down and bottom-up) attentional capture can be assessed by evaluating how cue validity effects in bilingual and monolingual participants vary across processing time. For example, if differences in overall reaction time (RT) reflect changes in a general parameter, then the cue validity effects may be relatively independent of differences in RT distributions across participants. This was tested here (Costa et al., 2009). To recapitulate, we examined the ability to control attention in bilingual and monolingual participants under conditions of cueing from irrelevant stimuli in WM, bottom-up visual priming, and singleton distraction. We expected to find an overall RT difference between monolinguals and bilinguals in all the three paradigms (WM, Singleton and Identify). However, if bilingualism specifically modulates top-down attentional guidance,1 then differences in the magnitude of the cue validity effects (and especially in the benefits from valid cueing) will be present only in the WM paradigm.

نتیجه گیری انگلیسی

. Results Trials with errors were excluded from the RT analyses (no main effects or interactions were found in the analyses of errors; all Ps > .16). RTs that were greater than 2.5 SDs from the mean were also excluded. This last action led to the loss of fewer than 2.7% (bilinguals) and 2% (monolinguals) of the trials in the WM condition, 1.8% (bilinguals) and 2% (monolinguals) of the trials in the Identify condition, and 5% (bilinguals) and 3.2% (monolinguals) of the trials in the Singleton condition. Performance on memory-probe trials in the WM condition was high (95% correct recognition of the conjunction cue, both for the bilingual and the monolingual groups). Performance on catch trials of the Identify condition was high as well (98.6% and 97.5% correct withholding of response for bilinguals and monolinguals respectively). The errors for the search task are reported in Table 3. The mean RTs in the conditions are illustrated in Fig. 2, and the benefits and costs from the critical distracters, relative to the neutral conditions, are presented in Fig. 3. Table 3. Error rates (percent) as a function of Group, Condition and Type of Trial. Values in brackets refer to standard deviations. Condition Type of Trial Group Bilingual Monolingual WM Invalid 1.5 (0.41) 1.25 (0.39) Neutral 0.75 (0.34) 1 (0.32) Valid 0.87 (0.35) 1.12 (0.48) Identify Invalid 1.87 (0.56) 3.87 (1.06) Neutral 2.13 (0.73) 3.12 (0.68) Valid 1.87 (0.61) 2.25 (0.74) Singleton Invalid 2.5 (0.85) 1.63 (0.45) Neutral 1.5 (0.40) 1.75 (0.49) Valid 1.25 (0.39) 1.37 (0.44) Table options Response latencies as a function of Type of Trial (valid, neutral, invalid), ... Fig. 2. Response latencies as a function of Type of Trial (valid, neutral, invalid), Condition (WM, Identify, Singleton), and Group (Bilinguals, Monolinguals). Error bars represent standard errors. Figure options Magnitude of the costs and benefits (in ms) broken down by Condition (WM, ... Fig. 3. Magnitude of the costs and benefits (in ms) broken down by Condition (WM, Identify, Singleton) and Group (Bilinguals, Monolinguals). Cost: invalid – neutral trials. Benefit: valid – neutral trials (for the ease of reading the graph, benefits are re-presented in positive instead of negative values). Error bars represent standard errors. Figure options 4.1. WM condition The main effect of “Group” (F(1, 78) = 3.14, MSE = 108384.29, P = .08) approached significance; bilinguals (1015 ms) tended to be faster overall than monolinguals (1090 ms). A main effect of “Type of Trial” (F(2, 156) = 345, MSE = 7307.34, P = .0001) was observed. Planned comparisons showed slower performance for invalid than for neutral trials (a WM cost) and faster performance in the valid than the neutral trials (a WM benefit; overall RTs: invalid – 1236 ms, neutral – 1042 ms, valid – 881 ms; all Ps = .0001). Crucially, there was an interaction between “Type of Trial” and “Group” (F(2, 156) = 7.99, MSE = 7307.34, P = .0001) indicating differences in the magnitude of the effects between bilinguals and monolinguals. Further analyses assessed the magnitude of the WM benefits and costs separately. For the WM cost (invalid and neutral trials), there was a two-way interaction between “Type of Trial” and “Group” (F(1, 78) = 4.08, MSE = 6506.59, P = .046), reflecting a smaller WM cost for bilinguals (169 ms, SD = 97 ms) compared to monolinguals (220 ms, SD = 129 ms). A two-way interaction was also observed for the WM benefit (valid and neutral trials) (F(1, 78) = 6.28, MSE = 5081.84, P = .014); there was a smaller WM benefit for bilinguals (132 ms, SD = 83 ms) than for monolinguals (189 ms, SD = 115 ms). All in all, the results of the WM condition revealed that WM costs and benefits were reduced for bilinguals, suggesting that, when performing a visual search task, bilinguals are less affected by the contents of irrelevant information held in WM. 4.2. Identify condition The main effect of “Group” (F(1, 78) = 11.16, MSE = 88726.73, P = .001) was significant; bilinguals were 128 ms overall faster than monolinguals. There was also a main effect of “Type of Trial” (F(2, 156) = 23.98, MSE = 6193.23, P = .0001). Planned comparisons showed slower performance for invalid than for neutral trials, and faster performance in the valid than the neutral trials (mean RTs: invalid – 1003 ms, neutral – 967 ms, valid – 918 ms; all Ps = .0001). In addition the interaction between “Type of Trial” and “Group” was not significant (F(2, 156) = 1.17, MSE = 6193.23, P = .311), revealing that bilinguals and monolinguals showed similar visual priming from the cue (Priming benefit mean: 31 ms (SD = 80 ms) and 69 ms (SD = 139 ms) for bilinguals and monolinguals respectively; Priming cost mean: 46 ms (SD = 75 ms) and 26 ms (97 ms) for bilinguals and monolinguals respectively). To assess whether the differences in the WM condition have to do only with WM guidance of attention and not to visual priming effects from the mere presentation of the cue, additional analyses were conducted comparing the magnitude of the effects in both conditions. An ANOVA was performed on correct RTs with “Type of Trial” (invalid, neutral, valid) as a within-subjects factor. To assess performance against the WM condition, we included “Condition” as a between-subjects factor (contrasting the Identify condition here with data from the WM condition) along with “Group” (monolingual vs. bilingual). There was a reliable main effect of “Group” (F(1, 156) = 12.64, MSE = 98555.51, P = .0001) with bilinguals responding 102 ms faster overall than monolinguals. A main effect of “Condition” (F(1, 156) = 9.83, MSE = 98555.51, P = .002) was observed too; participants were 90 ms faster in the Identify than in the WM condition. This replicates prior studies ( Soto et al., 2005). There was also a main effect of “Type of Trial” (F(2, 312) = 287.71, MSE = 6750.28, P = .0001). The two-way interactions among “Condition” and “Type of Trial” (F(2, 312) = 108.22, MSE = 6750.28, P = .0001) and between “Type of Trial” and “Group” (F(2, 312) = 6.43, MSE = 6750.28, P = .002) were significant. Importantly, the three-way interaction among “Condition”, “Type of Trial” and “Group” (F(2, 312) = 3.29, MSE = 6750.28, P = .038) was also significant, indicating that any contrasting effects of cue validity across the bilingual and monolingual groups differed in the WM and Identify conditions. Note that the magnitude of the benefits and costs in the Identify condition were much smaller than in the WM condition (collapsing across the different participants the benefits were 49 ms and 161 ms in the Identify and WM conditions respectively, t(158) = 6.4, P = .0001; the comparable costs were 37 ms and 194 ms in the Identify and WM conditions respectively, t(158) = 9.75, P = .0001). These differences highlight that the benefits and costs in the WM condition are strongly related to the active maintenance of the cue in memory. 4.3. Singleton condition There was a reliable main effect of “Group” (F(1, 78) = 10.38, MSE = 44983.64, P = .002); bilinguals were 88 ms overall faster than monolinguals. A main effect of “Type of Trial” was also significant (F(2, 156) = 100.67, MSE = 3149.93, P = .0001). Planned comparisons showed that performance was slower for invalid than for neutral trials (there was a Singleton cost) while performance was faster for valid than for neutral trials (there was a Singleton benefit) (overall RTs: invalid – 887 ms, neutral – 843 ms, valid – 763 ms; all Ps = .0001). The two-way interaction between “Type of Trial” and “Group” did not approach significance (F < 1); the Singleton benefits and costs were equivalent for bilingual and monolingual participants (Singleton benefit mean: 70 ms (SD = 78 ms) and 91 ms (SD = 70 ms) for bilinguals and monolinguals respectively; Singleton cost mean: 48 ms (SD = 78 ms) and 40 ms (SD = 70 ms) for bilinguals and monolinguals respectively). 4.4. On overall speed and the magnitude of the cueing effects The above analyses reveal the following main results. First, bilinguals tended to be overall faster in all three conditions (WM, Identify, and Singleton). Second, the magnitude of the validity effects in the visual search task was significantly smaller for bilinguals than monolinguals, but only when the task involved holding the initial cue in WM. As advanced in the Introduction, the presence of overall differences in RTs between bilinguals and monolinguals introduces some difficulties in interpreting the origin of the differences in the magnitude of the effects. In this section, we present some further analyses trying to disentangle overall RTs and attentional cueing effects. However, note that differences in overall speed between monolinguals and bilinguals do not per se lead to differences in the magnitude of the effects. In fact, bilinguals were overall faster in all three conditions, but only in the WM condition (as predicted) there were differences in the magnitude of the effects. This dissociation already suggests some differences between the RT advantage and cueing. To assess the relations between overall RTs and cueing further, we selected sub-samples of the bilingual (n = 27; RT = 1066 ms) and monolingual (n = 30; RT = 1105) groups matched for their overall speed in the WM condition (t(55) = 1.11, P = .273). Note that this reduction of the sample may entail a loss of statistical power. In order to overcome this potential limitation and to maximize statistical power we combined the WM benefits and costs into a single ANOVA with “WM Effect” as a within-subjects factor with two levels (Cost, Benefit) in absolute values, and “Group” (bilinguals, monolinguals as a between-subjects factor. There was a reliable main effect of “Group” (F(1, 55) = 5.88, MSE = 9035.55, P = .019) and no trace of an interaction between “Group” and “WM Effects” (F < 1). This demonstrates that monolinguals showed both larger benefits and costs relative to bilinguals, and the increased size of the effects did not differ for benefits and costs, even when bilinguals and monolinguals were matched for overall speed. Finally, we assessed whether the magnitude of the WM effects was constant across the RT distributions for bilinguals and monolinguals. To do this, we considered all correct responses given by all participants in each group. Taken together there were 783, 779 and 764 correct responses given by the bilingual group to valid, neutral and invalid trials, respectively. As a whole, the monolingual group gave 787, 780, and 743 correct responses to valid, neutral and invalid trials, respectively. We calculated how many of these responses (in percentage terms) were given within different RTs intervals (size of the interval = 200 ms). In Fig. 4 we plot the percentage of responses for trials from which the WM benefit is calculated (i.e., valid and neutral). For both bilinguals (Panel A) and monolinguals (Panel B), the WM benefit is reflected in the higher proportion of responses to valid relative to neutral trials across RTs between 600 ms and 1000 ms, and a higher proportion of responses to neutral relative to valid trials across RTs between 1000 ms and 1800 ms. In addition, we have plotted the difference between the percentage of responses to valid and neutral trials for both bilinguals and monolinguals (Panel C) (this figure reflects the percentage of responses to neutral trials minus the percentage of responses to valid trials). Importantly, the peak of this difference occurs at overlapping RTs intervals (800 ms RT interval) in both groups, being about 5% higher for monolinguals than bilinguals. In Fig. 5 we plot responses for WM costs (i.e., invalid and neutral trials in the WM condition). For both bilinguals (Panel A) and monolinguals (Panel B), the WM cost is reflected in higher percentages of responses to neutral relative to invalid trials across RTs of 600–1100 ms, and higher proportions of responses to invalid relative to neutral trials across RTs between 1200 and 2400 ms. In addition, we plotted the difference between the percentage of responses to invalid and neutral trials for bilinguals and monolinguals (Panel C) (this figure reflects the percentage of responses to invalid trials minus the percentage of responses to neutral trials). This figure reveals that the largest difference between the magnitudes of the WM costs for bilinguals and monolinguals was just 5% and arose when there were short RTs (between 800 ms and 1000 ms). These analyses indicate that the WM benefits and costs occurred at similar RTs intervals for bilinguals and monolinguals, and that the difference between groups was not due for instance to few monolinguals being disproportionately slower than the rest of participants. Distribution of valid and neutral trials across reaction time intervals (ms) for ... Fig. 4. Distribution of valid and neutral trials across reaction time intervals (ms) for bilinguals (Panel A) and monolinguals (Panel B). Panel C: Difference between the distributions of valid and neutral trials broken down by group of participants (Bilinguals, Monolinguals). Figure options Distribution of invalid and neutral trials across reaction time intervals (ms) ... Fig. 5. Distribution of invalid and neutral trials across reaction time intervals (ms) for bilinguals (Panel A) and monolinguals (Panel B). Panel C: Difference between the distributions of invalid and neutral trials broken by group of participants (Bilinguals, Monolinguals). Figure options All of these effects suggest that overall speed cannot explain the differences between bilingual and monolinguals in the magnitude of the WM effects on attentional guidance. In addition to these extra analyses on overall speed, other results make it unlikely that the reduced WM benefits and costs of bilinguals relative to monolinguals in the WM conditions are due to their faster speed of response. First, the “Group” (bilingual vs. monolingual) by “Type of Trial” (valid, neutral, invalid) interaction reflecting a smaller benefit and cost for bilinguals relative to monolinguals is only present in the WM condition. Second, Fig. 2 shows equal RTs for bilinguals and monolinguals in the valid condition. If the observed differences between groups were to be related only to differences in speed of processing, we should have found such a speeded processing for bilinguals also in the valid trials of the WM condition. In this respect, we could still argue that bilinguals were not faster than monolinguals on valid trials in the WM condition because of a floor effect. However, this possibility is unlikely given that both bilinguals and monolinguals were faster on valid trials in the Singleton condition relative to comparable trials in the WM condition. In other words, data from the Singleton condition indicates that there was still room in the WM condition to find a difference between the groups on valid trials. Nevertheless, we admit that the bilingual advantage on overall speed is a constraint on interpretations of reduced effects in bilingual relative to monolingual participants. In the present case, an additional caveat is that arguments are based on a single data point (valid trials for bilinguals in the WM condition).