THEORETICAL FOUNDATIONS OF THE BILINGUAL ADVANTAGES VIEW

Some underlying assumptions of the bilingual advantages view are that the neurocognitive operations that support moment-by-moment language processing differ between bilinguals and monolinguals and that these differences accrue over the life span to create significant neuroplastic change in the bilingual brain. Whether such differences exist, and whether they are quantitative, qualitative, or both, remain open questions that are part and parcel of the discussion regarding the notion of bilingual advantages. It is thus instructive to consider the ways in which the bilingual experience could differ from the monolingual experience and whether such differences could potentially sculpt the neurocognitive substrate of executive control functions generally.

Perhaps the most obvious way the bilingual experience differs from the monolingual experience is in terms of the automatic and simultaneous activation of multiple linguistic forms across virtually all levels of language. When bilinguals bring any idea to mind and retrieve linguistic labels or grammatical frames with which to verbalize it, or when they hear or read any linguistic stimulus presented to them, there is always a risk of experiencing some form of linguistic or conceptual ambiguity that must be managed. Of note, the degree of automatic cross-language activation depends on many factors, including the kind of language task in which one is currently engaged, the relative degree of first language (L1) and second language (L2) knowledge and proficiency, cross-language or within-language cues from the present context, and the differences between the L1 and L2 in question (Blumenfeld & Marian, Reference Blumenfeld and Marian2011; Dijkstra, Miwa, Brummelhuis, Sappelli, & Baayen, Reference Dijkstra, Miwa, Brummelhuis, Sappelli and Baayen2010; Kroll, Dussias, Bogulski, & Valdes Kroff, Reference Kroll, Dussias, Bogulski, Valdes Kroff and Ross2012; Libben & Titone, Reference Libben and Titone2009; Marian & Spivey, Reference Marian, Spivey, Hahn and Stones1999; Mercier, Pivneva, & Titone, Reference Mercier, Sudarshan, Pivneva, Baum and Titone2014; Schwartz & Kroll, Reference Schwartz and Kroll2006; Titone, Libben, Mercier, Whitford, & Pivneva, Reference Titone, Libben, Mercier, Whitford and Pivneva2011; Van Assche, Duyck, & Hartsuiker, Reference Van Assche, Duyck and Hartsuiker2012; van Hell & de Groot, Reference van Hell and de Groot2008; van Hell & Tanner, Reference van Hell and Tanner2012). Thus, bilinguals arguably experience greater executive control demands because of the need to inhibit cross-language linguistic representations and to selectively attend to target language linguistic representations. This form of executive control has been referred to in the bilingualism literature as “local” inhibition (de Groot & Christoffels, Reference Christoffels2006). Numerous studies show evidence of local cross-language activation and inhibition during language production and comprehension (Blumenfeld & Marian, Reference Blumenfeld and Marian2011; Christoffels, Firk, & Schiller, Reference Christoffels, Firk and Schiller2007; Dijkstra, Reference Dijkstra, Kroll and de Groot2005; Green, Reference Green2011; Guo, Liu, Misra, & Kroll, Reference Guo, Liu, Misra and Kroll2011; Kroll et al., Reference Kroll, Bobb, Misra and Guo2008; Macizo, Bajo, & Martin, Reference Macizo, Bajo and Martin2010; Martin, Macizo, & Bajo, Reference Martin, Macizo and Bajo2010; Misra, Guo, Bobb, & Kroll, Reference Misra, Guo, Bobb and Kroll2012; Pivneva, Mercier, & Titone, Reference Pivneva, Mercier and Titone2014).

A second way in which the bilingual experience can differ from the monolingual experience is in the need to manage activation of two or more language systems in anticipation of a specific upcoming communicative demand. For example, if bilingual individuals speak English at home and French at work, they may use substantial executive control resources to downregulate activation of English generally (and upregulate French) the moment they enter their French-speaking workplace. Similarly, they may then use executive control resources to downregulate activation of French generally (and upregulate English) once they return home to their English-speaking families. This form of executive control has been referred to as “global” inhibition in the bilingualism literature (de Groot & Christoffels, Reference Christoffels2006), and several studies show evidence of global inhibition, particularly during language production (Abutalebi, Tettamanti, & Perani, Reference Abutalebi, Tettamanti and Perani2009; Green, Reference Green1998; Guo et al., Reference Guo, Liu, Misra and Kroll2011; Kroll et al., Reference Kroll, Bobb, Misra and Guo2008; Meuter & Allport, Reference Meuter and Allport1999; Misra et al., Reference Misra, Guo, Bobb and Kroll2012; Pivneva, Palmer, & Titone, Reference Pivneva, Palmer and Titone2012; Von Studnitz & Green, Reference Von Studnitz and Green2002). For example, the bilingualism literature shows costs during language production that are associated with switching between first and second language blocks, suggesting that there is some cognitive overhead associated with tuning into one global language system over another (Campbell, Reference Campbell2005; Costa & Santesteban, Reference Costa and Santesteban2004; Meuter & Allport, Reference Meuter and Allport1999; Misra et al., Reference Misra, Guo, Bobb and Kroll2012). It is even more interesting that these global switch costs are asymmetric in that performance is significantly more impaired when bilinguals engage in an L1 speaking block that follows an L2 speaking block, compared to the reverse situation, when an L2 speaking block follows an L1 speaking block (e.g., Meuter & Allport, Reference Meuter and Allport1999). Such asymmetries suggest that if bilinguals first speak in an L2-only block, they recruit executive control to globally downregulate activation of their L1 to maintain L2 fluency. Consequently, when they switch to an L1 speaking block, performance is lower than it would have been had the prior L2 block not been encountered. In contrast, when bilinguals first speak in an L1 block, there is less of a need to recruit executive control to globally inhibit the L2, as the L2 is less entrenched to begin with compared to the L1. Thus, when they switch to the L2, performance is the same as it would have been had the L1 block not come first. The existence of asymmetric switching costs of this kind provide compelling evidence for the bilingual capacity of downregulating an entire language as a function of task demands (for consistent evidence in a passive listening comprehension task, see also Mercier, Pivneva, & Titone, in press). Such asymmetries may be indicative of switching between an overlearned and less practiced task generally, regardless of whether the task is linguistically oriented (Kiesel et al., Reference Kiesel, Steinhauser, Wendt, Falkenstein, Jost and Philipp2010).

Thus, the bilingual experience with language potentially differs from the monolingual experience in at least two ways: the need to use executive control to manage local cross-language activation that arises during language processing, and the need to use executive control to manage activation of an entire language system globally, presumably with the aim of preempting future anticipated demands of local cross-language activation. There are other variations of this basic distinction. For example, de Groot (Reference de Groot2011, p. 280) provides a comprehensive overview of the several theoretical dimensions along which bilingual language processing models vary that are relevant to notions of bilingual language control, two of which are key for our purposes. The first, scope, pertains to whether executive control extends to specific representations within a particular language (i.e., single words) versus all elements of a particular language (i.e., the entire language system). The second, direction, pertains to whether control processes are initiated after encountering or generating a particular linguistic item (i.e., reactively) or prior to encountering a particular linguistic item (i.e., proactively). In our view, these two dimensions may not be completely orthogonal. That is, executive control processes that are more likely to target individual words might tend to occur reactively, whereas executive control processes that keep a whole language downregulated given specific communicative demands would tend to occur proactively. However, it may be possible to isolate all four combinations of these two dimensions within particular experimental tasks (reviewed in de Groot, Reference de Groot2011).

It is interesting that the ways that bilingualism researchers have discussed language control have analogues in the nonlinguistic executive control literature. For example, the distinction between global and local executive control fits with recent work in the general executive control literature (Braver, Reference Braver2012). According to the dual mechanisms of control framework, proactive control maintains goal-relevant information in anticipation of future demands, for example, when bilinguals globally downregulate knowledge of one language and shift attention to another language. In contrast, reactive control acts as a late correction mechanism for high competition stimuli encountered in real time, for example, when bilinguals locally downregulate semantically incompatible meanings of interlingual homographs when reading or nontarget object labels when speaking. This distinction is somewhat different from other executive control frameworks, which distinguish between a common executive control capacity (i.e., the unity of executive function), and distinct subsystems of switching and context updating (i.e., the diversity of executive function; Miyake & Friedman, Reference Miyake and Friedman2012).

Regardless of the exact executive control framework to which one subscribes, the distinctions between proactive/global versus reactive/local control during bilingual language processing fit with an influential model of bilingual language production (which could conceivably be extended to comprehension) known as the inhibitory control (IC) model. According to the IC model, language production is framed as a communicative action that is analogous to nonlinguistic physical actions (Abutalebi & Green, Reference Abutalebi and Green2008; Green, Reference Green1998), in that it can have goals that are routine or nonroutine (Shallice & Burgess, Reference Shallice and Burgess1996). Within this view, L1 language production would be a routine communicative action whereas L2 language production would be nonroutine. Moreover, language production requires inhibitory control at two levels. At the language schema level, inhibitory control modulates the relative activation and inhibition of L1 and L2 generally, for example, whether one is about to communicate with a monolingual speaker of one's L1 where the L2 should optimally be suppressed versus engaging in simultaneous translation where the L1 and L2 must remain simultaneously engaged. At the word selection level, inhibitory control fine-tunes the relative activation and inhibition of specific words within each language.

Consistent with the idea that language control and nonlinguistic executive control engage shared neural systems, general executive control functions and executive control functions used during bilingual language processing (language processing generally) recruit similar brain networks. For example, Abutalebi and Green (Reference Abutalebi and Green2007) extended the IC model to incorporate neurocognitive evidence about bilingual language production. Here, they identified a network of cortical regions involving the prefrontal cortex (PFC), inferior parietal, and anterior cingulate cortex (ACC) and subcortical structures (basal ganglia, the head of the caudate nucleus in particular) that modulate competition between L1 and L2 knowledge activation during bilingual language production. It is interesting that the PFC (inferior and lateral regions) and ACC are also involved in nonlinguistic general executive control tasks in a comparable manner (e.g., Braver, Reference Braver2012). Common to both views, activation of the PFC is associated with the exertion of executive control (both reactive and proactive, across PFC subregions according to the dual mechanisms of control literature), whereas the ACC generates error monitoring signals that signal the need for reactive control and may subsequently trigger the application of proactive control to resist future errors.

Abutalebi and Green (Reference Abutalebi and Green2007) also noted the implications of this model for the role of L2 proficiency in bilingual language control. When L2 proficiency is low, L2 language production is more controlled and less automatic (see also Favreau & Segalowitz, Reference Favreau and Segalowitz1983; Segalowitz, Reference Segalowitz2010; Segalowitz & Hulstijn, Reference Segalowitz, Hulstijn, Kroll and de Groot2005), thus requiring inhibitory control (prefrontal function, in particular; see also Petrides, Reference Petrides1998). However, when L2 proficiency is high, L2 production is automatic and less dependent on inhibitory control, although L1 production effort might instead increase due to a collateral weakening of the links between word forms and concepts in the L1 (Bialystok, Reference Bialystok2001; Bialystok, Luk, Peets, & Yang, Reference Bialystok, Luk, Peets and Yang2010; Gollan et al., Reference Gollan, Montoya, Cera and Sandoval2008; Gollan, Slattery, et al., Reference Gollan, Salmon, Montoya and Galasko2011; Ivanova & Costa, Reference Ivanova and Costa2008; Michael & Gollan, Reference Michael, Gollan, Kroll and de Groot2005; Whitford & Titone, Reference Whitford and Titone2012) or to an increased likelihood of intrusions of L2 knowledge onto L1 processing.

To summarize thus far, several important theoretical views of bilingualism presume that cross-language activation generated during in the moment bilingual language processing creates executive control demands that differ from those experienced by monolinguals, though there is ongoing debate about whether such differences are quantitative (i.e., just more of what monolinguals normally experience), qualitative (i.e., fundamentally different from what monolinguals normally experience), or both. However, the bilingual advantages view takes this link a step further by stating that these in the moment bilingual experiences collectively lead to enduring changes in the minds and brains of bilinguals over developmental time. Thus, we now turn to some of the empirical research that addresses this idea with respect to bilingual children and young adults.

EMPIRICAL FOUNDATIONS OF THE BILINGUAL ADVANTAGES VIEW

One of the first papers to speak to the issue of bilingual advantages was a landmark study by Peal and Lambert (Reference Peal and Lambert1962). Prior to this study, the scientific research emphasized the notion that bilinguals had substantial disadvantages with respect to monolinguals in terms of language proficiency and general intellectual function (Peal & Lambert, Reference Peal and Lambert1962). The impression from this early literature was that being bilingual was highly undesirable and to be avoided. Peal and Lambert contradicted the negative view of bilingualism by showing that bilingual children actually performed better than monolingual children on a battery of verbal and nonverbal IQ tests and language proficiency and language attitude tests when methodological confounds of prior work were controlled (e.g., socioeconomic status, quality of schooling). This finding was among the first to highlight the notion that the bilingual experience creates opportunities rather than liabilities for children to engage in mental flexibility, including the ability to switch between not only languages but also cultures. Other research followed from this work showing that the experience of being bilingual can lead to other targeted advantages in general cognitive capacities that are presumably required of being bilingual, including metalinguistic awareness, linguistic rule learning, and executive control, which is the focus of this paper (Adesope, Lavin, Thompson, & Ungerleider, Reference Adesope, Lavin, Thompson and Ungerleider2010; Ben-Zeev, Reference Ben-Zeev1977; Bialystok, Reference Bialystok1986, Reference Bialystok1988; Bialystok et al., Reference Bialystok, Craik, Klein and Viswanathan2004; Galambos & Goldin-Meadow, Reference Galambos and Goldin-Meadow1990; Galambos & Hakuta, Reference Galambos and Hakuta1988; Salvatierra & Rosselli, Reference Salvatierra and Rosselli2011).

Bialystok and colleagues initiated the recent surge of interest regarding this question by building upon the original results of Peal and Lambert in children and extending this work to new bilingual groups, such as younger and older adults, and pathological populations (reviewed in Bialystok et al., Reference Bialystok, Craik and Luk2012). Bialystok initially investigated one particular area of bilingual advantages that had received a great deal of early attention: metalinguistic awareness (e.g., Bialystok, Reference Bialystok1988). By conducting experiments that were designed to parse the exact process by which bilingual children performed better on metalinguistic tasks, Bialystok and colleagues concluded that it had less to do with the availability of linguistic knowledge and more to do with an increased ability to selectively attend to competing linguistic cues or constraints (Bialystok & Majumder, Reference Bialystok and Majumder1998; Martin-Rhee & Bialystok, Reference Martin-Rhee and Bialystok2008). This realization thus pushed the locus of the bilingual advantages effect to more domain-general aspects of cognition, such as selective attention. In pursuing this hypothesis, Bialystok and colleagues, along with other research groups, observed bilingual advantages in children across a variety of tasks that arguably tap into executive control, similar in spirit to the original work by Peal and Lambert. As reviewed in Bialystok et al. (Reference Bialystok, Craik and Luk2012), these include block design from the Wechsler Intelligence Scale for Children, the flanker task, and the Simon task, as well as executive control tasks that are less lab-based and more similar to real-world experience (e.g., Noelting's juice task, theory of mind tasks; Bialystok & Majumder, Reference Bialystok and Majumder1998; Costa, Hernández, & Sebastián-Gallés, Reference Costa, Hernández and Sebastián-Gallés2008; Goetz, Reference Goetz2003; Kovács, Reference Kovács2009; Martin-Rhee & Bialystok, Reference Martin-Rhee and Bialystok2008; Yang, Yang, & Lust, Reference Yang, Yang and Lust2011). Bialystok and colleagues replicated these findings in bilingual younger adults and in bilingual older adults, which we discuss in more detail in later sections (Bialystok et al., Reference Bialystok, Craik, Klein and Viswanathan2004, Reference Bialystok, Craik and Luk2008; Colzato et al., Reference Colzato, Bajo, van den Wildenberg, Paolieri, Nieuwenhuis and Heij2008; Hilchey & Klein, Reference Hilchey and Klein2011; Prior & MacWhinney, Reference Prior and MacWhinney2010; Rubio-Fernandez & Glucksberg, Reference Rubio-Fernandez and Glucksberg2012).

However, subsequent work questioned such findings. With respect to studies of children, at least one study failed to find significant bilingual advantages on one measure of executive control in children: the Simon task (Morton & Harper, Reference Morton and Harper2007). This led the authors to conclude that prior findings of bilingual advantages arose because of potential confounds with bilingual status such as socioeconomic status or immigration status, problems that were claimed to be controlled in the study that produced null group results. However, socioeconomic status was argued to be controlled in prior reports (Bialystok, Reference Bialystok2009), as all the children tested were recruited deliberately from upper middle class school districts, thus making it likely that socioeconomic status was not a factor. Bialystok also argued that the null effect observed by Morton and Harper may have arisen because of a lack of power for detecting such an effect attributable to large response time variability combined with a relatively small sample size in their study (n = 17; Bialystok, Reference Bialystok2009; see also Kroll & Bialystok, Reference Kroll and Bialystok2013). Although reports of null results can be important empirical anchor points (i.e., to avoid “file drawer problems” in scientific reporting), there are unfortunately many uninteresting roads that can lead to null results, such as lack of power, differences across particular tasks or participant samples, and choice of particular tasks. From our view, while a failure to replicate a group effect is certainly inconsistent with a theory that predicts a group effect, the ability to attribute the original group effect to a particular confounding variable remains a hypothesis until such an experiment is undertaken showing affirmatively that the effect of interest is present only under conditions where the confound is present and absent under conditions when the confound is not present (e.g., crossing bilingual status with socioeconomic status and showing a bilingual advantage for one group but not the other).

With respect to studies of younger adults, similar questions have arisen in that two notable papers have questioned different assumptions of the bilingual advantages view. First, a comprehensive review of the literature on this topic failed to find any bilingual advantage with respect to inhibitory control (Hilchey & Klein, Reference Hilchey and Klein2011), although they did note bilingual advantages in reaction times generally among bilinguals across both conflict and nonconflict conditions of various executive control tasks. Second, a recent empirical paper also raised questions about the validity of bilingual advantages among young adults (Paap & Greenberg, Reference Paap and Greenberg2013). However, as Paap and Greenberg note, it is possible that their study, as well as other bilingual advantage studies, inadvertently included unidentified sources of variability that led to particular experimental outcomes, a situation which is particularly problematic in the context of null group findings. For example, a later empirical paper coauthored by Hilchey and Klein (Reference Hilchey and Klein2011; Misra et al., Reference Misra, Guo, Bobb and Kroll2012) found that bilinguals were advantaged over monolinguals on a task that assessed inhibition of return, which reflects the ability to disengage attention from a task-irrelevant peripheral cue. Of note, they found that bilinguals who had higher L2 proficiency were even more advantaged than bilinguals with lower L2 proficiency. This highlights two important potential sources of variability in such studies: the kinds of tasks used and their sensitivity for assessing executive control and the kinds of bilinguals tested and how they differ qualitatively and quantitatively (see also Bialystok, Reference Bialystok2006; Costa, Hernández, Costa-Faidella, & Sebastián-Gallés, Reference Costa, Hernández, Costa-Faidella and Sebastián-Gallés2009; Costa et al., Reference Costa, Hernández and Sebastián-Gallés2008; Hernández, Costa, Fuentes, Vivas, & Sebastián-Gallés, Reference Hernández, Costa, Fuentes, Vivas and Sebastián-Gallés2010). For example, Costa and colleagues found that the bilingual advantage is specific to executive control tasks that are maximally demanding (e.g., Costa et al., Reference Costa, Hernández, Costa-Faidella and Sebastián-Gallés2009), and Kroll and Bialystok (Reference Kroll and Bialystok2013) have argued that bilingual advantages might be more about general mental flexibility than any specific cognitive component.

Thus, several sources of unidentified variability may have contributed to the null group results reported by Paap and Greenberg (Reference Paap and Greenberg2013). For example, as they mention themselves, the bilinguals tested in their study came from a very wide array of language and cultural backgrounds and were immersed in an English language university context where the bilingual experience could have been more of the subtractive versus additive type (see Peal & Lambert, Reference Peal and Lambert1962, for a discussion of how these two types of bilingualism might lead to different cognitive outcomes). Finally, degree of bilingualism was only assessed using a single self-report measure consisting of a 10-point scale, where people endorsed global qualitative statements about their bilingualism that may have been difficult to distinguish and that may have overlooked crucial aspects of the bilingual experience that are relevant for the recruitment of executive control (e.g., daily language switching). The nature of this single questionnaire is worth considering (i.e., Beginner, know some words and basic grammar; advanced beginner, can converse with a native speaker only on some topics and with quite a bit of difficulty; intermediate, can converse with a native speaker on most everyday topics, but with some difficulty; advanced intermediate, can converse with little difficulty with a native speaker on most everyday topics, but with less fluency than a native speaker; near fluency, almost as good as a typical native speaker on both everyday topics and specialized topics I know about; fluent, as good as a typical native speaker; super fluency, better than a typical native speaker). Although such categories are descriptively rich, it is unclear whether they can capture nuanced differences among bilinguals in the way that multiple languages are used while speaking, reading, listening, or historically. For example, there is evidence suggesting that only certain classes of bilingual behavior, such as switching, are related to individual differences in executive control (Festman & Braun, Reference Festman and Braun2012; Festman et al., Reference Festman, Rodriguez-Fornells and Munte2010; Festman & Munte, Reference Festman and Munte2012; Prior & Gollan, Reference Prior and Gollan2011).

Thus, the field of bilingualism is likely to benefit by more thoroughly characterizing these different sources of variability. However, it is possible that even under the best of circumstances, assessing the nature of the relationship between executive control and bilingualism using only cognitive tasks may lack the nuance necessary to identify such a relationship definitively. One means of examining this relationship more closely is by searching for neurofunctional overlap or neuroplastic changes in one domain that are associated with changes in the other. Thus, in the next section, we consider some of the relatively small number of studies that have addressed the claims regarding a bilingual advantage using functional and structural neuroimaging methods.

NEUROFUNCTIONAL EFFECTS OF BILINGUALISM

Before reviewing the studies on bilingualism in particular, it is important to provide some background regarding neuroplastic changes across the life span more generally. From before birth through adolescence, there is evidence of rapid growth of the brain in terms of synaptogenesis, neurogenesis, and neuronal migration, which yields increases in both gray and white matter density (for reviews, see Stiles & Jernigan, Reference Stiles and Jernigan2010; Tau & Peterson, Reference Tau and Peterson2010). Simultaneously and subsequently, substantial synaptic pruning occurs, which can be associated with automatization of behavior and efficiency of function. There is also experience-dependent plasticity that occurs throughout the life span, inducing both short- and long-term changes in neural circuitry (e.g., Aydin et al., Reference Aydin, Ucar, Oguz, Okur, Agayev and Una2007; Luders, Toga, Lepore, & Gaser, Reference Luders, Toga, Lepore and Gaser2009; I. Park et al., Reference Park, Lee, Han, Lee, Lee and Park2009). Advanced aging is generally associated with reductions in gray and white matter density (atrophy; e.g., Kennedy & Raz, Reference Kennedy and Raz2009), with the frontal lobes most susceptible to deterioration (McGinnis, Brickhouse, Pascual, & Dickerson, Reference McGinnis, Brickhouse, Pascual and Dickerson2011; Raz, Reference Raz, Craik and Salthouse2000; Raz et al., Reference Raz, Lindenberger, Rodrigue, Kennedy, Head and Williamson2005; West, Reference West1996). The structural declines may be associated with increases in functional activation (overactivation) in other (linked) regions, including reduced suppression of the so-called default network (i.e., the network activated during the resting state) during cognitively demanding tasks (Grady, McIntosh, & Craik, Reference Grady, McIntosh and Craik2003; Grady, Springer, Hongwanishkul, McIntosh, & Winocur, Reference Grady, Springer, Hongwanishkul, McIntosh and Winocur2006; Park & Bischof, Reference Park, Bischof, Schaie and Willis2011; Park, Carp, Hebrank, Park, & Polk, Reference Park, Carp, Hebrank, Park and Polk2010; Park, Polk, Hebrank, & Jenkins, Reference Park, Polk, Hebrank and Jenkins2010), along with evidence of decreased connectivity within the resting state network in older adults (e.g., Grady et al., Reference Grady, Protzner, Kovacevic, Strother, Afshin-Pour and Wojtowicz2010; Park, Carp, et al., Reference Park, Carp, Hebrank, Park and Polk2010). Nonetheless, recent investigations have shown that despite age-related neuronal decline the aging brain is still sensitive to experience-dependent plasticity (Bavelier, Levi, Li, Dan, & Hensch, Reference Bavelier, Levi, Li, Dan and Hensch2010; Park & Bischof, Reference Park, Bischof, Schaie and Willis2011).

Of particular relevance to the current discussion, evidence has been reported suggestive of changes in gray matter density and volume in specific brain regions associated with the development of certain skills, for example, musical performance (Elbert, Pantev, Wienbruch, Rockstroh, & Taub, Reference Elbert, Pantev, Wienbruch, Rockstroh and Taub1995; Herholz & Zatorre, Reference Herholz and Zatorre2012), juggling (Draganski et al., Reference Draganski, Gaser, Busch, Schuierer, Bogdahn and May2004), navigation (Maguire et al., Reference Maguire, Gadian, Johnsrude, Good, Ashburner and Frackowiak2000), and acquisition of a second language (Berken et al., Reference Berken, Mok, Chen, Gracco, Baum and Klein2012; Klein et al., Reference Klein, Berken, Chen, Gracco, Baum and Mok2012; Mechelli et al., Reference Mechelli, Crinion, Noppeney, O'Doherty, Ashburner and Frackowiak2004). Recent data have also reported improved white matter integrity in bilingual older adults relative to monolinguals using diffusion tensor imaging (Luk, Bialystok, Craik, & Grady, Reference Luk, Bialystok, Craik and Grady2011; but see Mohades et al., Reference Mohades, Struys, Van Schuerbeek, Mondt, van de Craen and Luypaert2012, for diffusion tensor imaging findings in children; and Cummine & Boliek, Reference Cummine and Boliek2013, for inconsistent findings in young adults). The findings of these studies provide a neurostructural basis to support the notion of a cognitive advantage associated with bilingualism.

With regard to functional neuroimaging studies of bilinguals, numerous studies have focused on language switching, which has been shown to activate dorsolateral prefrontal cortex as well as portions of the left inferior frontal gyrus, bilateral temporal lobes, and bilateral caudate nuclei (Abutalebi et al., Reference Abutalebi, Annoni, Zimine, Pegna, Seghier and Lee-Jahnke2008; Guo et al., Reference Guo, Liu, Misra and Kroll2011; Hernandez, Reference Hernandez2009; Wang, Kuhl, Chen, & Dong, Reference Wang, Kuhl, Chen and Dong2009; see also Crinion et al., Reference Crinion, Turner, Grogan, Hanakawa, Noppeney and Devlin2006; Kim, Relkin, Lee, & Hirsch, Reference Kim, Relkin, Lee and Hirsch1997; Wang, Xue, Chen, Xue, & Dong, Reference Wang, Xue, Chen, Xue and Dong2007). Most of the frontal regions activated have also been independently implicated in general executive control, supporting a connection between the domains (e.g., Abutalebi & Green, Reference Abutalebi and Green2008; Hedden & Gabrieli, Reference Hedden and Gabrieli2010; see also Bialystok et al., Reference Bialystok, Craik, Grady, Chau, Ishii and Gunji2005; Luk, Anderson, Craik, Grady & Bialystok, Reference Luk, Anderson, Craik, Grady and Bialystok2010). It is interesting that, even at subcortical levels, there is evidence to suggest that bilingualism yields changes in processing. In particular, bilingual individuals display a more accurate frequency-following response in brain stem auditory evoked potentials recording (e.g., Krizman, Marian, Shook, Skoe, & Kraus, Reference Krizman, Marian, Shook, Skoe and Kraus2012). Such effects have been demonstrated for trained musicians as well (Kraus & Chandrasekaran, Reference Kraus and Chandrasekaran2010), suggesting a more general benefit associated with specific types of auditory experience.

However, despite the application of sophisticated brain imaging technology, studies on the representation of L1 and L2 are still needed (Chee, Reference Chee2006; Dehaene et al., Reference Dehaene, Dupoux, Mehler, Cohen, Paulesu and Perani1997; Kim et al., Reference Kim, Relkin, Lee and Hirsch1997; Klein, Milner, Zatorre, Meyer, & Evans, Reference Klein, Milner, Zatorre, Meyer and Evans1995; Klein et al., Reference Klein, Zatorre, Chen, Milner, Crane and Belin2006). Although such variables as age of L2 acquisition (e.g., Hernandez & Li, Reference Hernandez and Li2007; Kim et al., Reference Kim, Relkin, Lee and Hirsch1997) and the proficiency of the participants have been explored to some extent (Chee, Soon, Lee, & Pallier, Reference Chee, Soon, Lee and Pallier2004; Perani et al., Reference Perani, Paulesu, Galles, Dupoux, Dehaene and Bettinardi1998), few studies have taken into consideration the processing and production characteristics of individual subjects or subject groups. Moreover, although numerous investigations have explored neurofunctional patterns in bilinguals with different ages of acquisition (e.g., Wartenburger et al., Reference Wartenburger, Heekeren, Abutalebi, Cappa, Villringer and Perani2003), very few have focused on potential alterations in brain structure as a function of bilingualism (cf. Berken et al., Reference Berken, Mok, Chen, Gracco, Baum and Klein2012; Klein et al., Reference Klein, Berken, Chen, Gracco, Baum and Mok2012; Mechelli et al., Reference Mechelli, Crinion, Noppeney, O'Doherty, Ashburner and Frackowiak2004).

AGING, COGNITION, AND LANGUAGE

To better understand the potential links between bilingualism and executive control functions in older adults, it is important to consider broader issues pertaining to aging, cognition and language generally. Over the past 30 years, a great deal of work has investigated the cognitive and linguistic changes associated with normal healthy aging. As already noted, it is well known that, as we age, our brains change in terms of both structure and function (e.g., Hedden & Gabrieli, Reference Hedden and Gabrieli2004). How those brain changes map onto cognitive performance and how observable behavioral changes map onto neural changes remain areas of intense investigation. Several lines of evidence have demonstrated that aging is generally associated with decreased processing speed (e.g., Salthouse, Reference Salthouse1986, Reference Salthouse1996), reduced sensory acuity (Schneider et al., Reference Schneider, Daneman and Pichora-Fuller2002, Reference Schneider, Li and Daneman2007; Schneider, Pichora-Fuller, Kowalchuk, & Lamb, Reference Schneider, Pichora-Fuller, Kowalchuk and Lamb1994; Schneider, Speranza, & Pichora-Fuller, Reference Schneider, Speranza and Pichora-Fuller1998), and reductions in working memory and other executive control functions, such as inhibition (Burke, Reference Burke1997; Darowski et al., Reference Darowski, Helder, Zacks, Hasher and Hambrick2008; Hasher et al., Reference Hasher, Lustig and Zacks2007; Hedden & Gabrieli, Reference Hedden and Gabrieli2004; Salthouse & Meinz, Reference Salthouse and Meinz1995). It is not surprising that language processes that rely on executive control, such as the resolution of linguistic competition during spoken and written comprehension (and production), are especially vulnerable for older adults (Abada et al., Reference Abada, Baum and Titone2008; Copeland & Radvansky, Reference Copeland and Radvansky2007; Kjelgaard et al., Reference Kjelgaard, Titone and Wingfield1999; Lee & Federmeier, Reference Lee and Federmeier2011, Reference Lee and Federmeier2012; May et al., Reference May, Zacks, Hasher and Multhaup1999; Meyer & Federmeier, Reference Meyer and Federmeier2010; Peelle et al., Reference Peelle, Troiani, Wingfield and Grossman2010; Robert & Mathey, Reference Robert and Mathey2007; Titone et al., Reference Titone, Koh, Kjelgaard, Bruce, Speer and Wingfield2006; Wright & Newhoff, Reference Wright and Newhoff2002).

To illustrate the interplay between aging, cognition, and language, we briefly consider several studies of spoken word recognition that have compared older and younger adults in terms of monolingual language processing (though it is often unknown whether prior studies in the “monolingual” literature have included participants who also knew other languages). As is well known, most theories of spoken word recognition posit some form of competition between potential word candidates (Luce & Pisoni, Reference Luce and Pisoni1998; McClelland & Elman, Reference McClelland and Elman1986; Vitevitch, Luce, Pisoni, & Auer, Reference Vitevitch, Luce, Pisoni and Auer1999; Zhuang, Randall, Stamatakis, Marslen-Wilson, & Tyler, Reference Zhuang, Randall, Stamatakis, Marslen-Wilson and Tyler2011). For example, when people hear a target word (e.g., cap), they immediately activate other words that share a spoken word onset (e.g., can, cat) or that are phonologically similar to the target word in other ways (e.g., gap, tap; Allopenna, Magnuson, & Tanenhaus, Reference Allopenna, Magnuson and Tanenhaus1998; Connine, Blasko, & Titone, Reference Connine, Blasko and Titone1993; Connine, Blasko, & Wang, Reference Connine, Blasko and Wang1994; Connine, Titone, & Blasko, Reference Connine, Titone and Blasko1991; Connine, Titone, Deelman, & Blasko, Reference Connine, Titone, Deelman and Blasko1997). Listeners must thus inhibit partially activated competitors (e.g., can) while simultaneously enhancing the activation of intended word (cap) for comprehension to be successful.

These processes appear to be less efficient in older relative to younger adults, particularly in the context of increased task demands associated with recognition of lower frequency words (e.g., Revill & Spieler, Reference Revill and Spieler2012), words from higher density lexical neighborhoods (e.g., Taler, Aaron, Steinmetz, & Pisoni, Reference Taler, Aaron, Steinmetz and Pisoni2010), words with reduced contextual support (e.g., Sommers & Danielson, Reference Sommers and Danielson1999; Wingfield, Aberdeen, & Stine, Reference Wingfield, Aberdeen and Stine1991), and words presented in the context of noise (e.g., Ben-David et al., Reference Ben-David, Chambers, Daneman, Pichora-Fuller, Reingold and Schneider2011; Taler et al., Reference Taler, Aaron, Steinmetz and Pisoni2010). Of note, while changes in peripheral auditory sensitivity and associated speech discrimination difficulties may certainly exacerbate such age-related impairments (Pichora-Fuller, Reference Pichora-Fuller2003a, Reference Pichora-Fuller2003b; Sommers & Danielson, Reference Sommers and Danielson1999; Tremblay, Piskosz, & Souza, Reference Tremblay, Piskosz and Souza2003; Wingfield et al., Reference Wingfield, Aberdeen and Stine1991), impaired spoken word processing in older adults is observed even when participants are carefully screened for peripheral hearing ability.

For instance, Sommers and Danielson (Reference Sommers and Danielson1999; see also Sommers, Reference Sommers1996) examined spoken word identification in healthy older and younger adults and whether the availability of supportive semantic context would alleviate any age-related impairments. They compared older and younger adults’ recognition of words with large and small numbers of competitors (hard vs. easy, respectively, defined in terms of neighborhood density and frequency) in isolation and in the context of low- and high-predictable sentence contexts, at two different signal to noise ratios. The authors found that older adults were impaired relative to younger adults in the recognition of high-density hard words in isolation; however, they reaped greater benefit from the addition of supportive sentential context. In a second experiment, the investigators correlated performance on inhibitory skills (as measured by a speeded classification task and an auditory Stroop task) with word recognition performance in the older adults and found a significant relationship between inhibitory ability and hard word recognition. Based on these results, they concluded that impaired inhibition of lexical competitors contributes to age-related word recognition deficits (see also Taler et al., Reference Taler, Aaron, Steinmetz and Pisoni2010).

More recent investigations have confirmed a link between impaired inhibitory capacity and reduced word recognition performance. For instance, Ben-David et al. (Reference Ben-David, Chambers, Daneman, Pichora-Fuller, Reingold and Schneider2011) used eye-tracking (i.e., the visual world paradigm) to examine auditory word recognition in older and younger adults. In this paradigm, people hear spoken words while they view pictures on a computer screen that bear some relation to the spoken words they are hearing. For example, people might hear the word beaker, and see displays that contain a picture of a beaker (the target), unrelated control words (table), and different kinds of competitors, such as a word-onset competitor (beetle) or a word-rhyme competitor (speaker). Based on an analysis of which pictures people look at as they hear the spoken word, it is possible to evaluate which lexical candidates were partially activated in memory as the spoken word unfolded acoustically. It is interesting that Ben-David et al. found that older adults showed the same pattern of looks to word-onset competitor pictures, suggesting equivalent competition for these lexical items. However, they showed greater looks to rhyme pictures, suggesting greater competition from such candidate words (Ben-David et al., Reference Ben-David, Chambers, Daneman, Pichora-Fuller, Reingold and Schneider2011; for similar findings related to word frequency, see Revill & Spieler, Reference Revill and Spieler2012).

It is, of course, crucial to note that the findings reviewed above refer to group patterns and, in many instances, studies of older adults assume that older adults exhibit declines in cognitive functions and do not test those functions explicitly. It is self-evident that there is a great deal of interindividual heterogeneity in patterns of cognitive decline and resilience, as well as a lack of uniformity in age-related changes across cognitive and linguistic domains. Many factors undoubtedly contribute to this variability, including biological and environmental ones; neuroplasticity and compensatory abilities also vary across individuals, contributing further to the heterogeneity seen. For example, structural magnetic resonance imaging analyses have shown differences across different areas of the brain in the rate at which gray matter atrophies with increasing age (Good et al., Reference Good, Johnsrude, Ashburner, Henson, Friston and Frackowiak2001; Ohnishi, Matsuda, Tabira, Asada, & Uno, Reference Ohnishi, Matsuda, Tabira, Asada and Uno2001; Resnick, Pham, Kraut, Zonderman, & Davatzikos, Reference Resnick, Pham, Kraut, Zonderman and Davatzikos2003; Sowell et al., Reference Sowell, Peterson, Thompson, Welcome, Henkenius and Toga2003). Some functional neuroimaging investigations have demonstrated more bilateral activation in older relative to younger adults, particularly in those who perform well on the tasks under examination, possibly indicative of compensatory reorganization of function (Cabeza, Anderson, Locantore, & McIntosh, Reference Cabeza, Anderson, Locantore and McIntosh2002; Grady, Bernstein, Beig, & Siegenthaler, Reference Grady, Bernstein, Beig and Siegenthaler2002; Reuter-Lorenz et al., Reference Reuter-Lorenz, Jonides, Smith, Hartley, Miller and Marshuetz2000; but cf. Logan, Sanders, Snyder, Morris, & Buckner, Reference Logan, Sanders, Snyder, Morris and Buckner2002).

As an example of one cognitive or language-based factor that may influence the variability in the effects of age-related changes, the so-called Nun Study, a longitudinal investigation of aging and dementia, reported a strong negative relationship between language complexity in autobiographical essays completed in young adulthood (as measured by content or idea density) and the development of dementia later in life (Riley, Snowdon, Desrosiers, & Markesbery, Reference Riley, Snowdon, Desrosiers and Markesbery2005; Snowdon, Kemper, & Mortimer, Reference Snowdon, Kemper and Mortimer1996; Tyas, Snowdon, Desrosiers, Riley, & Markesbery, Reference Tyas, Snowdon, Desrosiers, Riley and Markesbery2007). In a follow-up investigation, Farias and colleagues examined idea density in a group of older individuals and found that, even when measured in an aging population, idea density continued to predict later cognitive decline (Farias et al., Reference Farias, Chand, Bonnici, Baynes, Harvey and Mungas2012). These findings suggest that greater linguistic abilities in early life confer resilience to age-related cognitive decline. In a related investigation, Iacono and colleagues examined the autopsied brains of four groups of subjects from whom cognitive measures had been obtained shortly before death: a group of asymptomatic Alzheimer disease (AD) participants, a group of patients with mild cognitive impairment (MCI), a group of patients with AD, and a normal control group free from cognitive impairment (Iacono et al., Reference Iacono, Markesbery, Gross, Pletnikova, Rudow and Zandi2009; most of the participants had also been in the Nun Study). The investigators reported neuronal hypertrophy in the asymptomatic AD group compared to both the MCI and control groups. The asymptomatic AD group (as well as the control group) also demonstrated higher idea density scores relative to the MCI and AD groups. Taken together, the findings support previous results indicating that stronger language skills may confer some form of cognitive reserve and that neuronal hypertrophy may reflect a compensatory response that helps to preclude cognitive impairment despite the presence of AD pathology in the brain (Iacono et al., Reference Iacono, Markesbery, Gross, Pletnikova, Rudow and Zandi2009; see also Rentz et al., Reference Rentz, Locascio, Becker, Moran, Eng and Buckner2010). Extending this conclusion, one might argue that bilingualism inherently represents a type of advanced language ability, and thus one might predict that bilingualism, too, would confer cognitive reserve. It is also possible that there are inherent differences in the brains of individuals who are better at language in young adulthood and have greater neurocognitive capacities in later life, a key point to which we return later.

BILINGUAL LANGUAGE PROCESSING IN HEALTHY OLDER ADULTS

We now turn to the relatively small number of studies that specifically investigate bilingualism in healthy older adults. Like the studies of bilingual younger adults reviewed above, this work tends to comprise two distinct but related types. The first type includes studies of language processing performance in bilingual older adults. This work typically examines bilingual older adults in relation to bilingual younger adults and thus addresses the basic issue of whether there are general effects of age in the ability of bilinguals to manage cross-language activation during bilingual language processing. The second type includes studies of executive control performance in bilingual older adults. This work typically examines bilingual older adults in relation to monolingual older adults completing a variety of executive control tasks and thus addresses the issue of whether being bilingual confers executive control advantages. Thus, the former assesses only language processing, or the link between bilingual language processing and executive control functions locally, in the moment, as language processes occur. In contrast, the latter assesses whether such local dependencies (which are often presumed in this literature by using language history as a proxy) lead to enduring changes in executive control generally.

With respect to lexical processing in bilinguals, investigations with young adults have demonstrated activation of candidates from both languages simultaneously (termed nonselective access), particularly if the target language is the listener's L2 (e.g., Blumenfeld & Marian, Reference Blumenfeld and Marian2007; Broersma & Cutler, Reference Broersma and Cutler2011; Canseco-Gonzalez et al., Reference Canseco-Gonzalez, Brehm, Brick, Brown-Schmidt, Fischer and Wagner2010; Marian & Spivey, Reference Marian and Spivey2003a, Reference Marian and Spivey2003b). The coactivation of candidates from multiple languages increases the need for suppression of competitors (Dijkstra, Van Jaarsveld, & Ten Brinke, Reference Dijkstra, Van Jaarsveld and Ten Brinke1998; Green, Reference Green1998). To date, only a limited number of studies have explored word recognition in older adult bilinguals. In one recent investigation of written language processing, Kousaie and Phillips (Reference Kousaie and Phillips2010) used reaction time and electrophysiological measures (specifically the N400, which is thought to reflect the ease of conceptual integration) to examine lexical priming effects for interlingual homographs (words that share spelling but not meaning across two languages; e.g., for French/English, the string coin which means corner in French) in younger and older adult bilinguals (Kousaie & Phillips, Reference Kousaie and Phillips2010). They used a word triplet paradigm common in research on unilingual processing of ambiguous words (Hagoort, Reference Hagoort1989; Milberg, Blumstein, & Dworetzky, Reference Milberg, Blumstein and Dworetzky1987) and hypothesized that younger adults would demonstrate priming effects irrespective of whether the prime and target appeared in the same language (i.e., the triplets “shoe-coin-money” and “soulier-coin-money” should both yield shorter reaction times and reduced N400 amplitudes; Kousaie & Phillips, Reference Kousaie and Phillips2010, p. 28). In contrast, because older adults have frequently been shown to be more sensitive to context (e.g., Wingfield & Tun, Reference Wingfield and Tun2007; Wingfield, Tun, McCoy, Stewart, & Cox, Reference Wingfield, Tun, McCoy, Stewart and Cox2006), older adults were predicted to only demonstrate priming in the language-consistent conditions. Both reaction time and event-related potential results supported their predictions, with younger adults exhibiting priming in both language-consistent and language-inconsistent contexts (i.e., nonselective access) and older adults only displaying priming in the language-consistent contexts, indicating their increased reliance on contextual information to facilitate language processing (Kousaie & Phillips, Reference Kousaie and Phillips2010).

A more recent investigation, focused on spoken language processing (Mercier, Sudarshan, Pivneva, Baum, & Titone, Reference Mercier, Sudarshan, Pivneva, Baum and Titone2014), made use of eye-tracking in the visual world paradigm to examine the level of both within-language and cross-language competition for lexical activation in older and younger French–English bilingual adults. Participants were presented with an auditory word (in English) along with a set of four pictures, which either included a within-language word onset competitor or a cross-language word onset competitor, as well as two unrelated distractors. Eye movement data revealed that older adult bilinguals exhibited greater within- and possibly cross-language competition, particularly for participants whose proficiency in English (as an L2) was relatively low, in keeping with the IC model (Abutalebi & Green, Reference Abutalebi and Green2007; Green, Reference Green1998).

With respect to language production, most studies have suggested a disadvantage of bilingualism, particularly in naming tasks. For example, young adult bilinguals name pictures more slowly than monolinguals, have lower accuracy rates on naming tests, and produce fewer words in category fluency tasks (e.g., Gollan, Fennema-Notestine, Montoya, & Jernigan, Reference Gollan, Fennema-Notestine, Montoya and Jernigan2007; Gollan, Montoya, Fennema-Notestine, & Morris, Reference Gollan, Montoya, Fennema-Notestine and Morris2005; Gollan, Montoya, & Werner, Reference Gollan, Montoya and Werner2002; Kohnert, Hernandez, & Bates, Reference Kohnert, Hernandez and Bates1998; Roberts, Garcia, Desrochers, & Hernandez, Reference Roberts, Garcia, Desrochers and Hernandez2002). According to Gollan and colleagues (Reference Gollan, Montoya and Werner2002; Gollan & Silverberg, Reference Gollan and Silverberg2001), these disadvantages are presumably linked to lower frequencies of use of any given word due to the use of a larger number of different words by bilinguals across two languages: the so-called weaker links, or frequency-lag, hypothesis (Gollan et al., Reference Gollan, Montoya and Werner2002; Gollan & Silverberg, Reference Gollan and Silverberg2001). Moreover, it has been hypothesized that this bilingual disadvantage in production should be reduced in older adults because they have had more time to make frequent use of all words in both languages (e.g., Gollan et al., Reference Gollan, Montoya, Cera and Sandoval2008). Alternatively, if the bilingual disadvantage is instead due to cross-language competition for production, then the disadvantage should increase in older adults who may be unable to effectively inhibit competitors and control language activation (e.g., Hernandez & Kohnert, Reference Hernandez and Kohnert1999). In one study that directly addressed the naming disadvantage and its purported explanations in older adults, Gollan and colleagues (Reference Gollan, Montoya, Cera and Sandoval2008) found support for the weaker links hypothesis by demonstrating that, while both younger and older bilingual adults exhibited larger effects of frequency than did monolinguals, effects of language dominance for low frequency words (with the nondominant language yielding larger frequency effects than the dominant language) were reduced in older adults compared to younger adults. The authors contend that these findings support the weaker links hypothesis because the older adults had greater opportunity to use the low frequency words in their nondominant language over a lifetime and thus were less affected by their (low) frequency (Gollan et al., Reference Gollan, Montoya, Cera and Sandoval2008).

Turning to studies of executive functions in older adult bilinguals, by far the most data supporting the notion of a bilingual advantage in aging has come from investigations of specific executive functions in bilingual relative to monolingual groups. In a series of investigations based in part on Green's (Reference Green1998) IC model of bilingual processing, Bialystok and colleagues (e.g., Bialystok, Reference Bialystok2006; Bialystok et al., Reference Bialystok, Craik, Grady, Chau, Ishii and Gunji2005; Bialystok et al., Reference Bialystok, Craik, Klein and Viswanathan2004) have examined the performance of older (and younger) adults on tasks that tap executive control. In the majority of such studies, the investigators have compared monolingual and (relatively heterogeneous) bilingual participants with respect to their performance on tasks incorporating conflict conditions. As an example, in one of their earlier studies with older adults, Bialystok and colleagues (Reference Bialystok, Craik, Klein and Viswanathan2004) made use of the Simon task (Simon, Reference Simon1969) in which participants are required to learn an association of a colored square, for example, with a specific key press on the left or right side of a keyboard or pad. The color may appear on the same side as the associated key (a congruent condition) or on the opposite side (a conflict condition). Normal, monolingual young adults demonstrate increased response times in the conflict condition due to the cost associated with inhibiting the “misleading” cue (i.e., the Simon effect). Older adults typically show an increased Simon effect relative to their younger peers. Both young and older bilinguals exhibit a reduction in the magnitude of the Simon effect compared to monolingual participants, with the reduction even greater for the older bilingual individuals (e.g., Bialystok et al., Reference Bialystok, Craik, Klein and Viswanathan2004). These findings have been interpreted to suggest that bilingualism confers resistance to age-related cognitive decline, at least with regard to inhibition of irrelevant or misleading cues (Bialystok et al., Reference Bialystok, Craik, Klein and Viswanathan2004; see also, Bialystok, Reference Bialystok2006; Bialystok et al., Reference Bialystok, Craik, Grady, Chau, Ishii and Gunji2005; but cf. Hilchey & Klein, Reference Hilchey and Klein2011). As already discussed, this protective advantage has been attributed to the frequent need, on the part of bilingual speakers, to switch between languages and thus to exercise inhibitory control (Festman et al., Reference Festman, Rodriguez-Fornells and Munte2010; Prior & Gollan, Reference Prior and Gollan2011).

In another investigation that reported cognitive reserve associated with bilingualism, Kavé, Eyal, Shorek, and Cohen-Mansfield (Reference Kavé, Eyal, Shorek and Cohen-Mansfield2008) conducted a longitudinal study of a large group of older individuals. Participants who were bilingual, trilingual, or multilingual (according to self-report) were tested on two cognitive screening measures at three points in time over the course of 12 years. The findings demonstrated a strong relationship of number of languages spoken with cognitive test results across the three test intervals (Kavé et al., Reference Kavé, Eyal, Shorek and Cohen-Mansfield2008), again suggesting that proficiency in more than one language provides cognitive benefits to individuals. Similar beneficial effects of number of languages were also found in pathological populations to be reviewed later (i.e., Chertkow et al., Reference Chertkow, Whitehead, Phillips, Wolfson, Atherton and Bergman2010).

In contrast to these investigations, Kousaie and Phillips failed to find an advantage over monolingual peers in a group of older bilinguals on a Stroop interference task (Kousaie & Phillips, Reference Kousaie and Phillips2012). The authors were careful to control important variables with respect to their participants, including proficiency, age of acquisition, and sociocultural factors such as immigrant status. Kousaie and Phillips (Reference Kousaie and Phillips2012) were fortunate to have had access to groups of highly proficient nonimmigrant bilinguals, rare in most previous studies of older bilingual populations. Given the absence of a bilingual advantage in their experiment, the authors raise questions concerning the degree to which sociocultural factors might have contributed to previous demonstrations of such an advantage (Kousaie & Phillips, Reference Kousaie and Phillips2012).

The majority of investigations reviewed thus far have focused on behavioral measures of cognitive function as a window into neuroplastic changes associated with bilingualism; only a handful have as yet directly explored neuroanatomical and neurofunctional patterns in older adult bilinguals. As alluded to in a previous section, one particularly interesting investigation examined both structural (via diffusion tensor imaging) and functional (resting-state) connectivity in older adult bilinguals (Luk et al., Reference Luk, Bialystok, Craik and Grady2011; but cf. Cummine & Boliek, Reference Cummine and Boliek2013). The findings revealed greater white matter integrity (as measured by fractional anisotropy) mainly in portions of the corpus callosum and the superior longitudinal fasciculi (bilaterally) in older bilinguals relative to their monolingual peers. In their investigation of patterns of resting state connectivity (often reduced in older individuals; Grady et al., Reference Grady, Protzner, Kovacevic, Strother, Afshin-Pour and Wojtowicz2010; Park et al., Reference Park, Polk, Hebrank and Jenkins2010), increased anterior–posterior connectivity emerged in the group of older bilinguals compared to monolinguals (Luk et al., Reference Luk, Bialystok, Craik and Grady2011). These provocative findings suggest a neuroanatomical basis for the cognitive reserve attributed to bilingualism in the aging population.

In another recent investigation, Gold, Kim, Johnson, Kryscio, and Smith (Reference Gold, Kim, Johnson, Kryscio and Smith2013) made use of a perceptual task-switching paradigm in a functional magnetic resonance imaging study of younger and older adult bilingual and monolingual participants. They found that older and younger bilingual individuals (from a diverse set of language and sociocultural backgrounds and primarily immigrants to the United States) performed better than the monolinguals and that the commonly occurring age-related increase in neural activation was reduced in the bilingual older adults compared to their monolingual peers. The authors interpret these findings as supportive of the view that bilingualism serves as a buffer against cognitive and neural decline associated with aging (Gold et al., Reference Gold, Kim, Johnson, Kryscio and Smith2013). However, bear in mind the caveats raised by the Kousaie and Phillips (Reference Kousaie and Phillips2012) study regarding drawing conclusions without carefully controlling for sociocultural factors. Nonetheless, the findings of these few investigations are certainly suggestive of a neuroplastic effect of bilingualism on age-related cognitive decline (for data on young adults, see also Berken et al., Reference Berken, Mok, Chen, Gracco, Baum and Klein2012; Klein et al., Reference Klein, Berken, Chen, Gracco, Baum and Mok2012; Mechelli et al., Reference Mechelli, Crinion, Noppeney, O'Doherty, Ashburner and Frackowiak2004). As noted earlier, we believe that these types of investigations will prove a particularly fruitful means of advancing our understanding of the relationship between bilingualism, executive control, and aging.

BILINGUALISM AND PATHOLOGICAL AGING

In this section, we turn to studies of pathological aging that are relevant to the potential links among bilingualism, executive control, and aging. The purported increase in cognitive reserve associated with bilingualism has led investigators to examine its potential clinical significance for the development and (degenerative) progress of dementia. In a now landmark study, Bialystok, Craik, and Freedman (Reference Bialystok, Craik and Freedman2007) hypothesized that the increased cognitive reserve associated with bilingualism has the potential to delay the onset of dementia. To test this hypothesis, they conducted a retrospective analysis of patients who had been referred to a memory clinic in the Toronto area for potential diagnosis of dementia. Within that pool, the investigators identified a subset of participants who had spent the majority of their lives (at least since early adulthood) regularly using two languages. In comparing this subgroup with monolingual 4 years on average in the bilingual group but no difference in the rate of decline subsequent to diagnosis (Bialystok et al., Reference Bialystok, Craik and Freedman2007). Based on these data, the authors concluded that bilingualism helps individuals to tolerate the disease, though it does not fundamentally alter the pathological brain process (Bialystok et al., Reference Bialystok, Craik and Freedman2007; see also Craik, Bialystok, & Freedman, Reference Bialystok and Craik2010).

As would be expected, given the exciting nature of such findings and their potential clinical significance, numerous researchers are now investigating the generality and reliability of the reported bilingualism effect. In one subsequent study, Chertkow et al. (Reference Chertkow, Whitehead, Phillips, Wolfson, Atherton and Bergman2010) questioned whether the bilingual advantage in the original study arose for one of several reasons, which included the following: the inclusion of patients with a mixed group of dementias, using a somewhat subjective measure of age of onset, and potential sociocultural confounds of the bilingual and monolingual groups (i.e., bilingual immigrants vs. unilingual native Canadians). In a new investigation, they therefore compared groups of monolinguals, bilinguals, and multilingual speakers (each of which included a subgroup of immigrants to Montreal) who spoke English and French. Their comparisons revealed no effect of immigrant status on age of diagnosis of dementia; however, there was a small but statistically significant effect of number of languages spoken on age of diagnosis (Chertkow et al., Reference Chertkow, Whitehead, Phillips, Wolfson, Atherton and Bergman2010). In comparing only the nonimmigrant subgroups of English–French bilinguals with groups of English and French monolinguals (who thus had had similar life experiences), the authors reported that, in contrast to Bialystok et al.'s (Reference Bialystok, Craik and Freedman2007) findings, the monolingual groups were diagnosed at later ages than the bilingual group. Based on these and other analyses of their relatively large sample, Chertkow et al. (Reference Chertkow, Whitehead, Phillips, Wolfson, Atherton and Bergman2010) concluded that there was limited support for bilingualism (i.e., knowledge of only two languages) providing a protective advantage against the onset of dementia but suggesting that knowledge of more than two languages (i.e., multilingualism) may confer some measure of cognitive reserve.

In a similar vein, Gollan, Salmon, et al. (Reference Gollan, Salmon, Montoya and Galasko2011) found that second language proficiency affected age of diagnosis for dementia only in individuals with lower degrees of education; the authors interpreted their findings as indicative of the potential of bilingualism to increase cognitive reserve but only in individuals who may not already have achieved their maximum potential as a result of other factors (e.g., high levels of education; Gollan, Slattery, et al., Reference Gollan, Slattery, Goldenberg, Van Assche, Duyck and Rayner2011; see also Sanders, Hall, Katz, & Lipton, Reference Sanders, Hall, Katz and Lipton2012). However, one problem with all of these studies, acknowledged by the authors themselves, is that they are cross-sectional in nature, rendering conclusions of a longitudinal nature difficult. It may also be important to examine population-level prevalence and incidence statistics, as well as ages of onset of dementia/diagnosis in countries outside of North America where bilingualism is the norm. Population-based studies may make it possible to statistically control for numerous factors that have the potential to also influence cognitive capacity, including personal and social factors (education, availability of health services, stigma associated with diagnosis, etc.).

BRINGING IT TOGETHER

In this selective review focused on bilingualism and executive control in the aging population, we have considered the foundations of the bilingual advantages perspective as well as the data that have been gathered in support thereof. It is clear from our discussion of the available behavioral data that there are numerous tantalizing findings that address the relationship between bilingualism and cognitive control; however, variability in study outcomes is also prevalent. The functional neuroimaging data also reveal some inconsistencies and interpretive dilemmas. While such measures of brain function may be considered somewhat more directly representative of language and cognitive processing, they are only as good as the tasks and participant populations included. In our view, investigations incorporating structural neuroimaging seem to hold more hope of reflecting true and lasting effects of bilingualism on the brain. In what follows, we turn to a consideration of how the field may begin to come to grips with the heterogeneity across individuals, languages, tasks and paradigms, in an effort to advance our understanding of language and cognitive processing in bilingual older adults. We focus on three main themes that we believe will lead to advances in our understanding of bilingualism and aging, and neuroplasticity generally.

• Theme 1: Bilinguals differ in ways that matter: let us embrace this variability.

Perhaps the single most important concern that plagues the vast majority of studies of bilingual speakers, whether younger or older, is the limited control or the systematic exploration of the language-learning characteristics and language proficiency of participants. Many of the existing studies on L2 acquisition and processing have been confounded by the tendency to include, in a single-subject group, individuals whose language learning characteristics and histories differ significantly (e.g., degree of fluency, early vs. late age of L2 acquisition, native language, country of origin, ratio of L1/L2 usage in daily life, community norms with respect to language switching, subtractive vs. additive bilingual situations) without explicitly investigating the import of such variables and focusing exclusively on group-level comparisons. Thus, participant-related heterogeneity both within and across studies makes it extremely difficult to draw convincing conclusions, and even more important, it prevents us from capitalizing on systematic variability among bilinguals that is likely related to the use of executive control during language processing (Green, Reference Green1998, Reference Green2011; Thierry & Wu, Reference Thierry and Wu2010). Such differences among bilinguals are almost completely lost for approaches that emphasize group differences alone.

For example, Green (Reference Green2011) describes an array of “behavioral ecologies” that are possible for bilinguals, which are self-evident to anyone who is bilingual or lives in a highly bilingual community. Given that every behavioral ecology will recruit language and executive control in a different way, it is essential for the field to move beyond simplistic global group comparisons (e.g., monolingual vs. bilingual) and move toward a more nuanced understanding of how the specific ways that bilinguals control linguistic knowledge might relate to specific executive control processes. Consider an informal sampling of the kinds of bilingual behaviors one can encounter in a highly bilingual city like Montreal. At one extreme, many people are clear simultaneous bilinguals, having learned both English and French from the moment they are born, often to parents of mixed or equally simultaneous linguistic backgrounds. For these people, the behavioral ecology is one of high language integration and normative mixing, where everyone in a family may speak two (or more) languages, and what is nonnormative (and thus potentially demanding with respect to executive control recruitment) is single language use where one of two highly proficient languages (i.e., two pseudo-L1s, though in practice one is usually dominant) must be unnaturally suppressed. At the other extreme, many people are clear sequential bilinguals, having learned a single language in the home and another in a different social context but then having to function in an exclusive other-language or highly mixed language contexts for substantial portions of their day (such is the case for the children of both authors). For these people, the behavioral ecology is one of normative linguistic exclusivity usually of the L1, and thus what is nonnormative (and thus potentially highly demanding with respect to executive control recruitment) is the use of L2 generally and the need to suppress accidental code mixes of one language into the other, mostly from the L1 to L2 direction.

In addition to the possible extremes of sequential versus simultaneous bilingualism, there are many points in between. They include bilinguals who are better at speaking/listening their L2 versus reading/writing and vice versa; bilinguals who fluently carry on conversations that are highly code-mixed within sentences, in an intentional rather than unintentional way; those who carry on conversations where one person speaks consistently in English while the other consistently responds in French; and bilinguals who know more than two languages and must integrate knowledge of the third language with knowledge of the other two. The demands for bilinguals along this entire continuum of language use can be very different in communities as a function of sociocultural factors, such as whether the two known languages are equally high status or where one language is higher status than another (as originally indicated by Peal & Lambert, Reference Peal and Lambert1962). The incentive to appear nativelike and to maintain language exclusivity of the high-status language might recruit executive control differently than in the case where two languages are equally valued. Moreover, such differences in language status can operate at a societal or individual level (i.e., whether a given child must speak one language at a particular school in order to be accepted by the majority of his or her peers).

Thus, with respect to language variables alone, there are clear quantitative and qualitative differences within the bilingual experience that can have major implications for executive control. In addition, while quantitative differences among bilinguals (e.g., increased overall L2 exposure of global L2 proficiency) are likely correlated with particular qualitative spheres of bilingual use (e.g., the likelihood of participating in highly mixed bilingual interactions), these two dimensions of bilingualism may not be completely independent. Thus, there could be two individuals who are matched on global L2 ability or exposure but differ dramatically in how their exposure/usage is distributed over the course of a day; some may consistently find themselves in mixed linguistic environments where all people code-switch intentionally and fluently, whereas others may consistently find themselves in compartmentalized linguistic environments, where they spend part of the day in their L1 and part of the day in their L2. Such differences have potential implications for the kinds of language control operations bilinguals will engage in and thus which executive control systems become “exercised” over the long term.

There are also other key sources of systematic variability that hinder coarse group comparison approaches. As explicitly noted by virtually all players in the literature, including Bialystok and colleagues, bilingualism is not the only road to neuroplastic changes to executive control networks within the brain or to increased cognitive reserve (Bialystok et al., Reference Bialystok, Craik and Luk2012). Thus, it can easily be the case that detecting any positive influence of bilingualism will be more difficult in populations who already benefit from other sources of enrichment (e.g., high socioeconomic status and all the wide-ranging advantages that come with it, along with many other daily living advantages). Equally important is that monolinguals differ as well, with some showing “non-native-like” language performance in their one and only known language (Pakulak & Neville, Reference Pakulak and Neville2010). Thus, if one must do group comparisons, it is important to characterize monolingual variability before understanding the effects of adding a second (or third) language. We know from the monolingual literature that executive control is important for many aspects of within-language ambiguity, thus an important question for future work is to disentangle exactly where bilinguals differ qualitatively rather than only quantitatively from the monolingual case.

Characterizing and accounting for participant variability becomes somewhat more complex with respect to bilingual older adults. One must also be cautious about numerous other participant-related factors, including sensoriperceptual declines, changes in speed of processing and motor response times, psycho- and neurotropic medication use, and more. In addition, one must also think concretely about the ways in which bilingual older adults necessarily differ from any bilingual younger adult control group. For example, is it enough to simply control overall L2 proficiency across an older and younger bilingual group? Maybe not. There may be fundamental differences in the bilingual experience between older and younger adults that pertain to how members of each group experienced their bilingualism in a larger societal or historical context. Turning again to examples from the authors’ own geographic region (Montreal), attitudes toward bilingualism and the use of English versus French have changed enormously over the lifetime of any bilingual older adult we can now recruit into the lab, ranging from almost total linguistic balkanization, where most people spoke English and French was marginalized (prior to the Quiet Revolution), to societal changes that led to an almost opposite situation where French is the language of government and English is the minority language with bilingualism on the rise (after the Quiet Revolution), to today where, at least in Montreal, almost all people are fluent in both French and English and bilingualism is commonplace. Thus, by definition, any older adult recruited into our lab will have sampled every sociocultural point along this timeline, although their formative language learning experience may be specific to a particular era, perhaps corresponding to the time they were in school. In contrast, any younger adult recruited into a lab will have only sampled one endpoint along this timeline, by definition. Thus, to the extent that differences in the bilingual experience relate to sociocultural factors, it will be impossible to perfectly match older and younger bilinguals within certain bilingual communities.

How are we to reasonably handle all of this variability? At the very least, we must adequately characterize it both in terms of self-report and objective measures where possible (and to do the same for item-level variability in our stimulus materials in equally robust and thorough ways). Then, we ought to use this information in evaluating the outcome of our experiments (e.g., did our effect of interest vary as a function of L2 ability or some other dimension). While many studies examine such factors, many do not, and such information is highly valuable. As a second step, the field at large may greatly benefit from large-scale, epidemiological approaches that systematically identify the component ways that the bilingual experience varies globally and then statistically reduce this undoubtedly high dimensional space to a smaller number of core dimensions. This may require efforts that span laboratories and geographic locations. If such cross-laboratory approaches are not feasible in the short term, similar approaches could occur within the context of single studies that have sufficiently large sample sizes (for the pitfalls of small sample sizes in making bilingual vs. monolingual comparisons, see Kroll & Bialystok, Reference Kroll and Bialystok2013).

As a third step, it may be time to reconsider our traditional statistical approaches, such as repeated measures analysis of variance, so that our research questions and designs are not limited by our statistical tools. To this end, an increasing number of researchers within psycholinguistics are using regression-style or other multivariate approaches that do not force experimenters to compress natural variability into two or three discrete categories. Thus, many language researchers are now turning to linear mixed effects regression modeling in studies of bilingualism (Dijkstra et al., Reference Dijkstra, Miwa, Brummelhuis, Sappelli and Baayen2010; Gollan & Goldrick, Reference Gollan and Goldrick2012; Pivneva et al., Reference Pivneva, Palmer and Titone2012; Van Assche, Duyck, Hartsuiker, & Diependaele, Reference Van Assche, Duyck, Hartsuiker and Diependaele2009; Whitford & Titone, Reference Whitford and Titone2012), which while not entirely straightforward (Baayen, Reference Baayen2008; Barr, Levy, Scheepers, & Tily, Reference Barr, Levy, Scheepers and Tily2013) allows researchers to examine how both participant-level and item-level variability simultaneously affect dependent variables of interest. Similar approaches have also been used within the neuroimaging community as attested by advances in statistical techniques such as partial least squares analyses (McIntosh, Bookstein, Haxby, & Grady, Reference McIntosh, Bookstein, Haxby and Grady1996). We believe that such approaches, which ultimately change the way we can think about our experimental designs, will be absolutely crucial for characterizing how bilingualism alters the structure and function of the brain. They will be especially vital for understanding bilingual older adults, who vary within and between groups in other key ways as well.

• Theme 2: Do bilinguals have advantages because they had them to begin with? Let us better distinguish cause from consequence.

One common argument regarding questions concerning the effects of bilingualism on the brain, is that people don't generally choose to be bilingual but rather they are born to families whose geographic locations or personal circumstances put them into a bilingual situation (e.g., Bialystok et al., Reference Bialystok, Craik and Luk2012). Given this view, any correlation one sees between being bilingual and other cognitive capacities implies a degree of causality that may not be considered for other types of overlearned skills (e.g., learning to play a musical instrument, training to become an elite athlete). The vast majority of work linking bilingualism to domain-general cognition is correlational, relying on inherent differences among bilinguals the moment they walk through the laboratory door.

There may be truth to this assumption. Children do not actively choose or self-select to be raised in a bilingual context. They are either born to such a context or they find themselves in such a context due to the choices of their parents. Many children raised in exclusively English-speaking Montreal households (including those of the authors and many other Quebec immigrants) are required to attend French school by law. Thus, these children have the great fortune of being immersed in a bilingual social context and will undoubtedly become bilingual. If they reap any cognitive benefits from being bilingual, it is not likely that their brains were already constructed that way ahead of time. However, this may not be the whole story, and assumptions regarding preexisting capacities and causality might deserve greater empirical scrutiny. For example, although a child may be serendipitously placed into a bilingual context and may become functionally bilingual, it remains possible that the kind of bilingual that they become may vary as a function of preexisting cognitive capacities, which in turn influence the social experience they later seek. Thus, a child with an outstanding working memory or executive control capacity may be more likely to put herself or himself into communicative situations that involve intensive L2 exposure or language mixing (e.g., choosing to spend equal or more amounts of time with L2-speaking peers in the playground vs. L1-speaking peers). It is conceivable that small, local social decisions of this sort could compound over time such that people who end up being better bilinguals in older adulthood were actually more cognitively flexible to begin with. Such a possibility is suggested by the Nun Study previously described (e.g., Snowdon et al., Reference Snowdon, Kemper and Mortimer1996). Here, the women who had greater cognitive resiliency in older adulthood were exactly those women who had had greater language skills in young adulthood.

Thus, studies that move beyond correlational approaches and that directly test the causes and consequences of bilingual proficiency and style are essential. These are of course difficult to do, as they would involve longitudinal studies over large time scales, or more experimentally oriented studies of the effects of second language training on cognition, or vice versa. Although many longitudinal studies exist regarding bilingual acquisition (Genesee, Reference Genesee2009, Reference Genesee, Garcia and Frede2010), these are typically time limited, spanning only a few years during childhood at most. Until funding agencies award research grants with multidecade terms, within-participant longitudinal studies of the kind necessary may not be feasible. However, studies that investigate the causes and consequences of bilingualism in other experimental ways, such as the impact of language training on cognitive control, or more recently, the impact of executive control training on language, may be of great use.

Some studies have shown training-related transfer from L2 learning to executive control (Martensson et al., Reference Martensson, Eriksson, Bodammer, Lindgren, Johansson and Nyberg2012); however, testing the question in this direction requires intense L2 learning situations and unusually motivated, self-selected learners who may already differ in cognitive control. Other work has looked at the effects of formal instruction in simultaneous interpretation (Christoffels, Reference Christoffels2006; Christoffels & de Groot, Reference Christoffels and de Groot2004; de Groot & Christoffels, Reference Christoffels2006; Elmer, Meyer, & Jancke, Reference Elmer, Meyer and Jancke2010; A. Green, Nicholson, Vaid, White, & Steiner, Reference Green, Nicholson, Vaid, White and Steiner1990; Macizo & Bajo, Reference Macizo and Bajo2006) and could presumably follow people throughout training to assess whether the ability to switch between two languages while engaging in simultaneous interpretation would enhance domain-general aspects of nonlinguistic cognitive control. However, here again, there is a possible issue of self-selection in that the individuals who survive intensive training in simultaneous interpretation may have had excellent executive control to begin with. However, if bilingualism and domain-general cognition exercise the same neurocognitive substrate, transfer should also occur from the direction of executive control to language processing. Consistent with this idea, recent work suggests transfer from conflict resolution training to L1 syntactic ambiguity resolution (Hussey & Novick, Reference Hussey and Novick2012). We believe that more studies of this kind, which involve true experimental manipulations, will be crucial for working out the many issues raised by the notion of bilingual advantages in both younger and older adults.

• Theme 3: Is it really “bilingual advantages” we are after? Let us reframe the issue to address life-long neuroplasticity.

In recent years, a great deal of ink has been spilled, and perhaps many voices may have gone hoarse, debating the existence or nonexistence of bilingual advantages in executive control in older adults and in bilinguals generally. As alluded to previously in this paper and by others, this debate is situated within a particular sociocultural and scientific context. In a sociocultural sense, in the past, and in many different parts of the world even today, being bilingual was seen as a cognitive liability. Such a viewpoint has serious implications for how parents decide to raise their children and for how societies make decisions concerning resource allocation to bilingual education (i.e., whether it should be done at all, if so, at what age should it be introduced). Prior to about 10–15 years ago, the scientific study of bilingualism was seen as marginal to mainstream concerns about the psychology and cognitive neuroscience of language and as a specialty area. In recent years, times have changed, and rightly so given the number of people worldwide who speak more than one language, which is often claimed to exceed the number of people worldwide who speak only one language. This shift in both sociocultural and scientific attitudes about bilingualism has arisen from some of the important empirical and theoretical work beginning with Peal and Lambert (Reference Peal and Lambert1962) and continuing through the current research era that is intensively focused on whether bilingualism leads to cognitive benefits generally. Thus, from a sociocultural and scientific standpoint, the search for bilingual advantages has done a great service to the field at large by shining a light on the remarkable capacities of bilinguals and how these capacities relate to more general cognitive capacities.

Of course, the issues raised here go beyond mere debate because there is actual empirical evidence to be considered. As we have seen, the question of whether bilinguals are advantaged compared to monolinguals is hard to answer with a simple “yes” or “no,” given existing empirical evidence. Many studies show bilingual advantages across the life span, particularly in children, older adults, and during pathological aging, and the findings of many of these studies are quite compelling. Other studies show no advantage or, at least, differences in findings. Consequently, the clearest answer thus far involves the somewhat unsatisfying and all too common “it depends” response, where there are many candidate differences across studies, including the nature of the tasks; differences among bilinguals, monolinguals, or between groups; and the like. There is also the general difficulty associated with the interpretation of null effects: did they arise because there truly is no difference, or was there a lack of power, or an odd sampling issue?

However, we believe that the larger problem is that if one asks a simple question about a complex phenomenon, one is likely to get a simple (and unsatisfying) answer. Thus, if we ask different and potentially more sophisticated questions about bilingualism, as many people are now doing, we may come upon more interesting and clearer answers, or at the very least fruitful directions for future work.

For example, what follows are open questions about which we are most intrigued. Which aspects of using more than one language induce neuroplastic changes in the human brain? Are there parallels in monolingual language processing, or are there certain things that bilinguals do that have no parallel with monolinguals? Are differences with respect to language processing among bilinguals and compared to monolinguals qualitative or quantitative in nature? Are these differences and their relation to executive control static over the life span, or do they vary in systematic ways? Are there fundamental parallels to nonlinguistic behaviors, such as musical expertise or other complex motoric behaviors? What can the different aspects of bilingual language function tell us about executive control generally? Are certain bilingual behaviors, such as mixing, always more taxing cognitively, or does it depend on the way in which a given bilingual individual acquired knowledge of his or her multiple languages or on differences between the structures of the languages in question? How do multiple contributors to cognitive reserve, including bilingualism, accumulate within individuals? Does being bilingual have less of an impact if someone is already engaged in high-cognitive reserve activities? One senses that questions arising from trying to understand failures to replicate a bilingual advantage may be potentially more interesting than clear demonstrations of the effect itself! Thus, from the standpoint of generating new lines of research, the bilingual advantages hypothesis has been an unmitigated success. Let us build upon this success by moving into the next phase of inquiry, which is to dig deeper regarding the specific points of contact within and across language and general executive control domains. Let us move beyond “yes” or “no” questions.

FINAL REMARKS

At the outset of this paper, we noted that with advancing age comes a conflict between a vast amount of knowledge and experience accumulated over many years and the natural structural and functional brain changes that occur in later life, tending to lead to decrements in cognitive capacities. Throughout the discussion we have highlighted investigations that suggest that experience-dependent plasticity yields both short- and long-term changes. In particular, following Bialystok and others (Bialystok & Craik, Reference Bialystok and Craik2010; Bialystok et al., Reference Bialystok, Craik, Klein and Viswanathan2004, Reference Bialystok, Craik and Freedman2007), lifelong bilingualism (among other factors) may contribute to the development of cognitive reserve and thus improve cognitive and linguistic processing efficiency in older adults. One important question that has not been addressed in the bilingualism literature, in contrast to other domains focused on experience-dependent plasticity in aging (e.g., exercise, see Voss, Nagamatsu, Liu-Ambrose, & Kramer, Reference Voss, Nagamatsu, Liu-Ambrose and Kramer2011; cognitive training, see Lustig, Shah, Seidler, & Reuter-Lorenz, Reference Lustig, Shah, Seidler and Reuter-Lorenz2009; Noack, Lovden, Schmiedek, & Lindenberger, Reference Noack, Lovden, Schmiedek and Lindenberger2009) is how one might translate these findings to recommendations for individuals or social policies. Should we suggest that, upon retirement, everyone should learn a second language (much like recommending an active and engaged lifestyle in older adulthood is now touted in the popular media)? Is it too late at that point for neurocognitive benefits to accrue? Should we instead recommend that multiple languages be taught from the earliest possible ages, providing the greatest potential for neuroplastic effects? Based on the relatively limited data collected to date, it is clearly too early to draw definitive conclusions or recommend large-scale changes to social policy regarding the consequences of bilingualism. Nonetheless, the increasing number of studies demonstrating both structural and functional neural alterations associated with bilingualism hold great promise that in the near future we may be able to identify those aspects of language learning that are most crucial to the development of cognitive reserve and thus have the greatest potential to influence quality of life for older adults.

In the literature on the influence of exercise on neuroplastic changes in aging, research has suggested that not only does exercise induce the growth of nerve cells and blood vessels, it appears to increase production of important chemicals in the brain, including brain-derived neurotrophic factor, which may be important in the survival and repair of neural tissue (e.g., Voss et al., Reference Voss, Nagamatsu, Liu-Ambrose and Kramer2011). Similarly, in the domain of cognitive training, of particular interest are investigations demonstrating transfer effects from the trained domain to other aspects of cognition (e.g., Lustig et al., Reference Lustig, Shah, Seidler and Reuter-Lorenz2009; Noack et al., Reference Noack, Lovden, Schmiedek and Lindenberger2009) and associated underlying neural changes (e.g., Jones et al., Reference Jones, Nyberg, Sandblom, Stigsdotter Neely, Ingvar and Petersson2006; Kelly & Garavan, Reference Kelly and Garavan2005; Klingberg, Reference Klingberg2010). It is important to note that investigators in these domains are beginning to highlight the crucial importance of examining individual differences in the effects of training in order to truly understand the underlying mechanisms and explain why certain individuals benefit more from specific training than do others (Garrett, MacDonald, & Craik, Reference Garrett, MacDonald and Craik2012). As research in the field of bilingualism and aging continues to grow, we suggest that more interdisciplinary studies are needed that can begin to tackle the mechanisms underlying the neuroanatomical and neurophysiological changes associated with being bilingual.