A critical review of research relating to the learning, use and effects of additional and multiple languages in later life

2. Constructing the ‘third age’ and third-age identities

What does the label ‘third age’ precisely include? When does the participant sample of a study qualify as consisting of ‘third-age’ or senior learners, and how flexible is this label? Age is strongly related to a number – that is, the number of years that have passed since one's birth. Furthermore, in many societies, retirement is still promoted as a shift from full-time employment to full-time leisure around age 62 or 65 (Degnen, Reference Degnen2007; Moen, Reference Moen, Carr and Komp2001). Some scholars try to attribute a certain number of years to the third age, while others state that the third age begins and ends at given ages. These efforts constitute ‘age-grading’ (Oxford, Reference Oxford and Gabrys-Barker2018). Lawrence-Lightfoot (Reference Lawrence-Lightfoot2009) identified ‘the third chapter’ (roughly equivalent to the third age) as lasting from age 50 to 75; whereas for Smith (Reference Smith2000) the third age extends from 70 to 84. In Goral (Reference Goral, De Houwer and Ortega2019) the definition of ‘older’ refers to adults who are 60 years and older. The eighth and final stage of Erikson's (Reference Erikson1982) model of psychosocial development, which is characterised by dichotomies of intimacy versus isolation, generativity versus stagnation and integrity versus despair, identified as starting around 65, might purportedly involve a crisis of identity or despair and might include the third age. Since theorists typically view the third age as a time before any serious disability arises for a given person (Oxford, Reference Oxford and Gabrys-Barker2018), it has been hypothesized third age (as a state) could be ‘prolonged’ through language learning as an ‘anti-ageing activity’ (Ryan & Dörnyei, Reference Ryan, Dörnyei and Berndt2013, p. 93; see our discussion below).

As Oxford (Reference Oxford and Gabrys-Barker2018) has noted, reliance on chronological or calendar age is problematic in many respects. Although age is used to describe individuals, its nature and influence do not originate solely within the individual; at any point in life, the import of age derives to a large extent from how life is socially organised (Gullette, Reference Gullette2003). In other words, although everyone experiences ageing in his/her unique way, people are always ‘aged by culture’ in the sense that cultural ageing discourses delimit experiences (Isopahkala-Bouret, Reference Isopahkala-Bouret2015). Treating age as a numerical attribute of the person ignores considerations concerning the social utilization of age as a categorisation and identification device (Rughiniș & Humă, Reference Rughiniș and Humă2015). Finally, the length of the third age is difficult to ‘set in stone’, as social, medical and scientific advances make possible an ever-increasing number of third-age years (Oxford, Reference Oxford and Gabrys-Barker2018).

The cultural and social contexts of ageing have changed greatly over the last two decades and ageing experiences have become more differentiated (see e.g. Biggs, Reference Biggs2005). The range of descriptions of ageing has expanded to the extent that ‘old age’ as we knew it is now joined by ‘third age’ and its associated array of new possibilities. There have been few attempts, however, precisely to understand the denomination ‘third age’ (and indeed ‘fourth age’), let alone to theorise around them. In Laslett's seminal book on the third age, A Fresh Map of Life, he defined third-agers as ‘active seniors with free time’, depicting the fourth age as a period of decline and decrepitude that particularly affected those who lived beyond their mid-eighties (Laslett, Reference Laslett1989, p. 41). The third age is typically viewed as a time before any serious disability arises for a given person, a period of subjective well-being, while the ‘fourth age’ implies the presence of disability (Carr & Komp, Reference Carr and Komp2011). Sometimes referred to as the ‘young-old age’ or even ‘the golden age of adulthood’ (Barnes, Reference Barnes2011, p. 1), part of the definition of the third age is thus its exclusion of ‘old age’ and ‘agedness’ (Gilleard & Higgs, Reference Gilleard and Higgs2010). Oxford (Reference Oxford and Gabrys-Barker2018) uses the term ‘third-agers’ to refer to ‘healthy, motivated individuals no longer working full-time’, as compared to the fourth age, which is termed ‘old-old’ age or ‘old age’. Gilleard and Higgs (Reference Gilleard and Higgs2010) argue that the fourth age is not about non-participation in the cultural field of the third age. As a space ‘where choice, autonomy, self-expression, and pleasure collapse into a silent negativity’ (p. 126), the fourth age does not necessarily represent a particular age cohort or distinct phase of life; rather, it is defined as a metaphorical ‘black hole of ageing’ (p. 125) or the ‘othering of old age’ (p. 126).

The identities offered to older adults are becoming more diverse. Isopahkala-Bouret (Reference Isopahkala-Bouret2015) suggests that people who are now in their 60s have reached middle age in ways that are quite unlike those of their parents or grandparents. For present-day 60-year-olds ageing is no longer defined only by standardized life-course trajectories. Rather, much of the cultural fabric that makes the so-called ‘third age’ can be traced to the identity politics of the 1960s and its emphasis upon choice, autonomy, self-expression and pleasure (Gilleard & Higgs, Reference Gilleard and Higgs2002, Reference Gilleard, Higgs, Rees Jones, Higgs and Ekerdt2009). As fixed identities cease to be ascribed to age, a self-conscious compromise is required between intention, external pressures, and one's own interior world (Biggs, Reference Biggs2005, p. 118). Older adults may increasingly experience multiple and manifold pressures on ageing identities that were previously experienced as stable and predictable (Biggs, Reference Biggs2005).

Thus, there has recently been a ‘cultural turn’ in gerontology (Biggs, Reference Biggs2005, p. 119); adult ageing is perceived as becoming increasingly homogeneous in terms of age boundaries. How far are adults the same regardless of age and how far does later life have its own distinctive qualities? Originally, ‘distinctivist’ trends could be identified in work (e.g. Cole, Reference Cole1992) which argued for discernible differences between later life and other parts of adult life. The argument that life-course categories are becoming increasingly indistinct was initiated by Featherstone and Hepworth (Reference Featherstone and Hepworth1983, Reference Featherstone, Hepworth and Bytheway1989, Reference Featherstone, Hepworth, Featherstone, Hepworth and Turner1991, Reference Featherstone, Hepworth, Featherstone and Wernick1995) and has been followed through in the cultural sociology of Gilleard and Higgs (Reference Gilleard and Higgs2002, Reference Gilleard, Higgs, Rees Jones, Higgs and Ekerdt2009, Reference Gilleard and Higgs2010, Reference Gilleard and Higgs2013). Moen (Reference Moen, Carr and Komp2001), likewise, calls for the consideration of alternative, more flexible life paths. According to Biggs (Reference Biggs2005) the possibility of virtual identities has multiplied the options open to older people no longer bound by social and biological reference points. Biggs speculates that the suggestion that adult ageing is becoming increasingly homogeneous means that soon there may be little reason to study late life separately from other parts of the life-course. If, however, later life is seen as subject to its own specific priorities, then the ability to interchange older with younger age groups becomes questionable.

In sum, chronological age does not determine the positioning of L2 learners in later life: age is part of a complex web of social distinctions such as gender, socio-economic background and psychological and individual factors that intersect in the construction of a learner's relative status and opportunities.

3. What older learners bring to the learning of additional languages

Growing evidence from research in education, psycholinguistics, cognitive science and neurolinguistics challenges the view that the age factor as the non-plus-ultra predictor of L2 learning outcome, arguing, for example, that even young adult L2 beginners can reach native-like L2 proficiency (see e.g. Moyer, Reference Moyer2004; Birdsong, Reference Birdsong2006; Muñoz & Singleton, Reference Muñoz and Singleton2011; Kinsella & Singleton, Reference Kinsella and Singleton2014). However, there is broad acceptance of the notion of multifactorial differences between child and adult L2 acquisition, and likewise there are also claimed to be reasons why younger and older adults should be treated differently in the SLA context. First, there are structural and functional changes in the ageing brain between younger and older adulthood, and there is research which suggests that such changes influence cognitive functioning and the brain's ability to learn. Second, there is a pronounced degree of variability among older adults, relating to variation in lifestyle factors and measure of involvement in cognitively stimulating activities (e.g. engagement with additional languages) in later life, and this variation has been shown to modulate cognitive performance. Third, third-agers’ attitudes towards L2 acquisition can be expected to vary considerably, not only between younger and older adults but also actually among older adults. Reasons for this may stem from differences in employment status, amount of free time available, changes that have occurred in the political situation, personal interests, personality traits or even social stigma.

4. How older adults may benefit from L2 learning

We turn now to the question of how good additional language learning is for older learners. As mentioned above, L2 learning has been hypothesised to contribute to healthy and active ageing, to be a strategy in later adulthood to maintain and enhance ‘healthy’ ageing – since it is a cognitively challenging – to promote neural plasticity and foster social interaction, individual mobility and autonomy (e.g. Rönnlund, Bäckman, & Nilsson, Reference Rönnlund, Bäckman and Nilsson2005; Park & Reuter-Lorenz, Reference Park and Reuter-Lorenz2009; Antoniou et al., Reference Antoniou, Gunasekera and Wong2013; Green, Reference Green and Gabrys-Barker2018).

However, while long-term active use of two languages seems to be neuroprotective, we lack studies of the potential neuroprotective effects of the learning and use of an L2 later in life. There are only a few existing studies investigating the relationship between SLA in older adults and the effects of such L2 training on cognition. Moreover, such research has to be relatively complex in its design in terms of sample size, number of measurements per participant, types of control groups, intensity of L2 training, suitability of L2 teaching methods, theoretical and statistical model formulation and interactions between the predictors under investigation. As Ramos, García, Casaponsa, and Duñabeitia (Reference Ramos, García, Casaponsa and Duñabeitia2017, p. 46) point out, ‘establishing a direct link between a specific training in language control or in language acquisition and a consequent improvement of non-linguistic control may not be an easy endeavour’.

We refer below to the results of (1) intervention studies focusing on the effects of L2 learning on older adults who begin the relevant language learning process in old age versus studies that involve participants who have already had years of linguistic experience with an L2; (2) longitudinal studies versus studies that present snapshots of states; and (3) intervention studies conducted in naturalistic settings versus school contexts.

First results of such intervention studies suggest that

1. (a) language learning supports non-auditory neural plasticity in cognitive-related brain areas (frontal lobe) (Schlegel et al., Reference Schlegel, Rudelson and Tse2012; Stein et al., Reference Stein, Federspiel, Koenig, Wirth, Strik, Wiest and Dierks2012);

2. (b) cognitive fitness constitutes a major factor explaining the variance in FL development as a function of L2 training in a school context – see e.g. Kliesch, Giroud, Pfenninger, and Meyer (Reference Kliesch, Giroud, Pfenninger, Meyer and Gabrys-Barker2018), who found a significant relationship between participants’ L2 growth and verbal fluency and working memory;

3. (c) participants’ cognitive capacities seem to be more predictive of L2 progress in later life than age in naturalistic settings; i.e. only older adults with the largest working memory capacity show significant gains in L2 learning (see e.g. Mackey & Sachs, Reference Mackey and Sachs2012);

4. (d) even a short period of intensive language learning can modulate attentional functions (e.g. attention switching) from which both older and younger learners can benefit (Bak, Long, Vega-Mendoza, & Sorace, Reference Bak, Long, Vega-Mendoza and Sorace2016); such cognitive benefits have also been found in delayed post-tests after six months of intensive FL learning via computer-based language training software and computer games (Wong et al., Reference Wong, Ou, Pang, Zhang, Tse, Lam and Antoniou2019) – although the opposite has also been reported (see e.g. Berggren et al., Reference Berggren, Nilsson, Brehmer, Schmiedek and Lövdén2018; Ramos et al., Reference Ramos, García, Casaponsa and Duñabeitia2017);

5. (e) there are advantages in additional (instructed) language learning for late bilinguals compared to monolinguals regardless of type of instruction (Cox & Sanz, Reference Cox and Sanz2015; Cox, Reference Cox2017);

6. (f) effects measured during sentence comprehension using electroencephalography (EEG) become more similar between native speakers and adult L2 learners as a function of proficiency (Dowens, Vergara, Barber, & Carreiras, Reference Dowens, Vergara, Barber and Carreiras2010; Soskey, Holcomb, & Midgley, Reference Soskey, Holcomb and Midgley2016); furthermore, for older learners who make more L2 gains, switching between the L1 and the L2 is less effortful than for those with lower L2 proficiency after training (Van Der Meij, Cuetos, Carreiras, & Barber, Reference Van Der Meij, Cuetos, Carreiras and Barber2011; Prat, Yamasaki, Kluender, & Stocco, Reference Prat, Yamasaki, Kluender and Stocco2016; Kliesch et al., Reference Kliesch, Giroud, Pfenninger, Meyer and Gabrys-Barker2018);

7. (g) although some implicit forms of learning – particularly those involving the acquisition of subtle probabilistic sequential regularities – show age-related declines, older adults are capable of statistical learning of the distributional information underlying grammatical form class categories, following brief auditory exposure to a subset of possible strings of the language (Schwab et al., Reference Schwab, Schuler, Stillman, Newport, Howard and Howard2016);

8. (h) despite equivalent performance, distinct neural networks characterise the ceiling of young and elderly French-speaking participants trained by means of a computerised Spanish word program (Marcotte & Ansaldo, Reference Marcotte and Ansaldo2014);

9. (i) the degree of difference in learning success between older individuals is predicted by various factors that may modulate benefit, such as prior knowledge of more than one language, which results in more successful L2 development (Pfenninger & Polz, Reference Pfenninger and Polz2018), language typology, bi- versus multilingualism, and age of acquisition (Antoniou & Wright, Reference Antoniou and Wright2017);

10. (j) learner investment is likely to interact with the cognitive capacities of older learners, such that high cognitive demand leads to exhaustion, which may increase perceptions of low learning abilities and tiredness and reduce motivation (Ware et al., Reference Ware, Damnee, Djabelkhir, Cristancho, Wu, Benovici and Rigaud2017);

11. (k) L2 learning plays a key role inasmuch as it addresses socio-affective issues – i.e. by preventing isolation, fostering linguistic flexibility, self-esteem and autonomy (Antoniou et al., Reference Antoniou, Gunasekera and Wong2013; Pot et al., Reference Pot, Keijzer, de Bot and Gabrys-Barker2018a).

One of the few behavioural studies investigating the relationship between L2 learning in older adults and cognition which made use of L2 training based on communicative tasks and interactional feedback is Mackey and Sachs's (Reference Mackey and Sachs2012) study of the effect of inter-individual differences in cognitive capacities on L2 learning success in older adults. Their research involved nine Spanish-speaking participants aged 65–89 (M = 72, SD = 8.32) who had lived in the US for an average of 17 years, and so were not yet in the third-age category when they started learning the L2. They engaged in a variety of communicative tasks five times over a five-week period. The results of three post-tests reveal that participants’ L1 listening-span capacities were more predictive of their L2 progress than age. However, Mackey and Sachs's findings are of restricted generalizability, since (1) the training focused only on the acquisition of one particular linguistic feature and is therefore not representative of L2 acquisition in general; (2) although the training was purely auditory, hearing skills were not controlled for, which means that differences in L2 improvement could simply be due to differences in sensory acuity.

Schlegel et al. (Reference Schlegel, Rudelson and Tse2012) followed young adults, assessing white matter structural plasticity using diffusion tensor imaging during intensive L2 training in Modern Standard Chinese and compared them to controls who were not participating in any L2 programme. Compared to the controls, the L2 trainees exhibited higher integrity in language-relevant white matter connections between auditory and frontal brain regions in the left and right hemisphere as a function of the training. Furthermore, the increase of the integrity in the white matter connections correlated with the individual increase of the grades achieved in the L2 class. Investigating grey matter structural plasticity, Stein et al.’s (Reference Stein, Federspiel, Koenig, Wirth, Strik, Wiest and Dierks2012) study pointed to a similar correlation between the increase in cortical thickness in the left inferior gyrus and the increase in proficiency of L2 acquisition, which was independent of the absolute L2 proficiency level. Such correlations show that it is mainly inter-individual differences of the proficiency increase that are reflected in brain plasticity rather than the initial or resulting PLs. Thus, there is conclusive evidence for substantial structural and functional neural plasticity in adulthood as a function of inter-individual trajectories of learning during L2 acquisition.

In Kliesch et al. (Reference Kliesch, Giroud, Pfenninger, Meyer and Gabrys-Barker2018), an EEG experiment was conducted after an intensive English course (four hours a day, i.e. 60 lessons over the course of three weeks) with ten German-speaking ‘young-olds’ (monolingual Swiss- or High-German native speakers, mean age = 68.20, SD = 2.44, range = 65–74) who were controlled for musical training, hearing ability, use of English (beginners), amount of time spent in an English-speaking context, neurological, psychiatric or language disorders, IQ and physical activity. One of the goals of this study was to investigate participants’ processing of L2 sentences and learn about their ability to (1) switch between the new L2 and their native language and (2) predict meaningful sentence endings in the L2. The results showed a typical N400 component with a centro-parietal distribution in the time window between 300 and 450 ms with larger amplitudes in the switch conditions compared to the non-switch conditions. In other words, the EEG data indicated that for participants with higher L2 knowledge, switching between English and German was less effortful than for those with lower L2 proficiency. This replicates the findings of Van Der Meij et al. (Reference Van Der Meij, Cuetos, Carreiras and Barber2011), who also found an effect of switch on N400 amplitude.

To be noted, however, is that the opposite has also been found – e.g. in Berggren et al.’s (Reference Berggren, Nilsson, Brehmer, Schmiedek and Lövdén2018) study with a sample of 160 healthy older participants aged 65–75 years (Swedish speakers with no prior knowledge of any Romance language – although it is not clear if there were any bilinguals among them), who were randomised to 11 weeks of either language learning or relaxation training. Participants in the language learning condition met twice per week over 11 weeks for 2.5 hours each class. Cognitive performance was assessed two weeks before (pretest) and one week after the intervention (post-test), including tests of spatial intelligence, verbal intelligence, working memory, long-term associative memory and item memory. Language gains were measured through ‘a vocabulary test consisting of 110 words randomly sampled from the chapters completed during the course’ i.e. a translation task. Although the participants were able to acquire ‘a basic Italian vocabulary’, the authors argued that ‘this is not due to either poor measurement, low course intensity, or low statistical power, but that language learning in older age is likely to have no or trivially small effects on cognitive abilities’ (p. 1). The authors’ focus on vocabulary, however, is regrettable, considering that in many immigrant studies (e.g. Clahsen & Felser, Reference Clahsen and Felser2006; McDonald, Reference McDonald2006, Reference McDonald2008; DeKeyser, Alfi-Shabtay, & Ravid, Reference DeKeyser, Alfi-Shabtay and Ravid2010; Granena & Long, Reference Granena and Long2013) it is pointed out that L2 morpho-syntax and phonology seems to be more vulnerable to processing difficulties than L2 lexico-semantics and more susceptible to age. Such difficulties have been linked to resource limitations that might lead to the inability (a) to access and retrieve stored L2 knowledge (semantically related difficulties) and/or (b) to detect phonological discriminations in the input (phonologically related difficulties), similar to the difficulties of native speakers under specific types of stress manipulation (McDonald & Roussel, Reference McDonald and Roussel2010; Pfenninger, Reference Pfenninger2011). However, recent accounts predict starting age effects on lexical behaviour (see e.g. Hernandez, Li, & MacWhinney, Reference Hernandez, Li and MacWhinney2005; Hernandez & Li, Reference Hernandez and Li2007; Hernandez, Reference Hernandez2013; Li & Grant, Reference Li and Grant2015; Bylund, Abrahamsson, Hyltenstamb, & Norrman, Reference Bylund, Abrahamsson, Hyltenstamb and Norrman2019; Hernandez et al., Reference Hernandez, Claussenius-Kalman, Ronderos, Castilla-Earls, Sun, Weiss and Youngin press).

Ramos et al. (Reference Ramos, García, Casaponsa and Duñabeitia2017) also reported no benefits after an eight-month-long language training for 26 Spanish monolingual elderly learners of Basque in the Basque country (Spain) who were able to use the L2 beyond the classroom. Hypothesising that L2 training might boost switching between mental sets and increase inhibition ability, the authors failed to find a significant difference in switch costs between their experimental and control group (Spanish monolinguals who received no L2 instruction). Pot et al. (Reference Pot, Keijzer, de Bot and Gabrys-Barker2018a) suggest that this might have to do with the fact that Ramos et al. (Reference Ramos, García, Casaponsa and Duñabeitia2017) merely looked into improvements with respect to switching ability as opposed to, e.g. inhibitory control or attention, as tested in Bak et al. (Reference Bak, Long, Vega-Mendoza and Sorace2016). Pot et al. further point out that Ramos and colleagues made no reference to social engagement benefits, even though cognitive engagement and stimulation, especially through language learning, may have effects in the social or biological domain of ageing. Pfenninger and Polz (Reference Pfenninger and Polz2018) found that personal motivation, self-study effort, learning progress and social aspects appeared to be co-dependent at least for some participants; it thus makes sense to include (psychometric) tests that probe the learners’ environment, history and personality. A similar picture emerged in Ware et al. (Reference Ware, Damnee, Djabelkhir, Cristancho, Wu, Benovici and Rigaud2017), who conducted a four-month pilot study (16 two-hour sessions) with 14 French participants: five beginners, five intermediate, and four advanced learners of L2 English. Three measures – semi-structured qualitative interviews, the Montreal Cognitive Assessment (MoCA) test and the University of California Loneliness Assessment (UCLA) scale – were administered before and after a technology-based intervention to measure cognitive level and feelings of loneliness or isolation. Their qualitative analyses shed light on the social benefits of L2 learning in the third age: most of the participants found the programme to be stimulating and enjoyable and were eager to learn more, motivated by the feeling that that their engagement in educational activities allowed them to stay involved with the outer world.

This is in line with Erikson's (Reference Erikson1982) model of psychosocial development. Lifelong education offers seniors new opportunities for overcoming a state of isolation, stagnation and despair by offering contact with other seniors (and not only them), involvement in developing one's knowledge and skills (e.g. language skills) and more than anything else, it can give a new aim to life (see also Gabrys-Barker, Reference Gabrys-Barker2018). Ybarra et al. (Reference Ybarra, Burnstein, Winkielman, Keller, Manis, Chan and Rodriguez2008, p. 257) point out that ‘social interaction and relationships not only sharpen our knowledge and social skills, but also strengthen the cognitive processes that underlie those skills’. Cattan, White, Bond, and Learmouth (Reference Cattan, White, Bond and Learmouth2005) synthesised 30 quantitative outcome studies, 12 qualitative as well as evaluation studies and surveys involving interventions designed to prevent isolation and loneliness among older adults. Interventions judged to be effective included interventions with focused educational input and those enabling ‘some level of participant control’ (p. 57; see also Shapira, Barak, & Gal, Reference Shapira, Barak and Gal2007). Swain and Lapkin (Reference Swain and Lapkin2011) emphasised that language plays a vital role in this as it mediates cognitive processes such as attention, recall, and knowledge creation, each being important aspects of cognitive enhancement. What is more, enhanced self-esteem experienced by participants in intervention studies might be a result of their interaction with the researchers.

Keijzer's project ‘VIDI’ proceeds from the assumption that no two bilinguals are the same and bilingualism effects are therefore not universally reported, leading to questions regarding its validity (see Paap et al., Reference Paap, Johnson and Sawi2015). She proposes a new impetus to the field of bilingualism but also its applicability in healthy ageing, through:

1. (1) an epidemiological study that relates individual differences in bilingual experiences to healthy ageing outcomes in a sample of more than 12,000 65+ North Netherlanders. Differences include languages or dialects spoken, age at acquisition, intensity of use, and language attitudes;

2. (2) introducing an FL learning experience to a functionally monolingual group of healthy seniors and those diagnosed with Mild Cognitive Impairment or Late Life Depression over the course of six months.

This two-step design is expected to shed new light on two important questions: the nature of the BA in old age and how learning a new language after the age of 65 enhances cognitive flexibility and well-being levels in healthy and non-healthy elderly people.

Recruiting older Turkish women (first-generation migrants) in the Netherlands, Pot (Reference Potin prep.) analyses multilingual ageing from two perspectives: the advantages that present themselves to multilingual elderly such as Frisian and bidialectal speakers in the North of the Netherlands but also the detrimental health and social effects of growing old in an environment where the dominant language is not your mother tongue. In another current Ph.D. project in Groningen, van der Ploeg explores what the best way is to teach older adults a new language. Partly on the basis of a project tapping older adults’ preferences regarding third age language learning, an implicit and explicit language condition were created and their effectiveness assessed.

Finally, in a project in Germany, Grossmann, Teichmann, and Meyer (Reference Grossmann, Teichmann and Meyerin prep.) are currently investigating the effects of L2 learning on cognition in older adulthood, dividing 60 native German-speaking healthy older participants aged 65–80 years into a training group and a control group. Language learners attend English classes for beginners for 1.5 hours a day, five days a week for three weeks. Changes in executive functions, attention, and episodic memory are determined through assessment before and after participation in the classes. Participants in the control group will also be assessed twice within the same time frame but will attend language classes only after the second measurement so as to determine when first cognitive changes occur. Nine months after the training a post-test will be conducted.

All of these studies are product-oriented rather than process-oriented and therefore risk ‘decontextualizing, segregating, and atemporalizing’ language development according to Larsen-Freeman and Cameron (Reference Larsen-Freeman and Cameron2008, p. 252). However, our contemporary understanding of SLA views it as a complex, dynamic, ecologically situated, multivariate phenomenon – hence the impact of theories like chaos/complexity theory (Larsen-Freeman & Cameron, Reference Larsen-Freeman and Cameron2008) and dynamic systems theory (de Bot, Lowie, & Verspoor, Reference de Bot, Lowie and Verspoor2007; Lowie, Reference Lowie, Ortega and Han2017). These characteristics of language development can make it extremely difficult to adopt traditionally formulated, linearly framed research methods (Lorenzo, Moore, & Casal, Reference Lorenzo, Moore and Casal2011). This makes the non-reductionist, ecological, systems view afforded by dynamics and complexity theories – henceforth referred to as Complex Dynamic System Theory (CDST) (Larsen-Freeman, Reference Larsen-Freeman, Ortega and Han2017; Pfenninger & Neuser, Reference Pfenninger, Neuser, Rose and McKinleyunder review). CDST is a meta-theory of change which describes relationships among embedded subsystems that cannot be assumed to be linear, based on the idea that intra-learner variability is a necessary condition for development (de Bot et al., Reference de Bot, Lowie and Verspoor2007). This approach fundamentally questions the feasibility of investigating cause–effect relations, the traditional basis of generalisable theories. Since it is unlikely that a single cause (or a handful of independent variables) will give rise (in a linear fashion) to such a complex event as L2 learning, the suggestion is to focus on ‘tendencies, patterns and contingencies’ rather than simple cause–effect explanations (de Bot & Larsen-Freeman, Reference de Bot, Larsen-Freeman, Verspoor, de Bot and Lowie2011, p. 23). CDST also requires us to revisit the idea of predicting behaviour, considering that no two situations can be similar enough to produce the same behaviour (Larsen-Freeman & Cameron, Reference Larsen-Freeman and Cameron2008). In sum, in a complex world, we lose predictability; the nature of explanation changes; and cause and effect work differently (Larsen-Freeman & Cameron, Reference Larsen-Freeman and Cameron2008, p. 72). Along those lines de Bot and Makoni (Reference de Bot and Makoni2005) propose a dynamic model of language use in ageing, in which ageing is regarded as a system that develops under the influence of changes within the individual (physical and psychological) and external changes relating to attitudes towards ageing in society, and an individual's perception of this.

Such a process-oriented approach is currently used to analyse data from a longitudinal study with dense measurements called ‘Dynamic analyses of language learning in the third age (DYNAGE3)’ at the Universities of Salzburg and Zurich (Benisowitsch, Reference Benisowitschin prep.; Haselsberger, Reference Haselsbergerin prep., Kliesch, Reference Klieschin prep.). Some of the key questions asked are (1) when L2 development is statistically significantly increasing (or decreasing), and (2) whether the set of predictors in question (age, cognitive fitness, bilingualism, affect) have a significant effect on the trajectories under investigation. The data set contains 80 learning trajectories (i.e. 40 participants on a mono/bilingual continuum in the experimental group, 20 participants in the active control group, and 20 participants in the passive control group), each of them represented by 30 measurements taken at equal intervals over one year, amounting to 2,400 data points per participant and test. Participants are tested on a range of cognitive, linguistic, neurofunctional, and socio-affective parameters, with the aim of identifying factors that facilitate successful learning and help explain inter-individual differences that are age-independent. In such a study it is critical to include a passive control group as well as an active control group engaging in another type of training that does not involve the use of language (see e.g. Basak, Boot, Voss, & Kramer, Reference Basak, Boot, Voss and Kramer2008; see also Hargreaves, Pexman, Zdrazilova, & Sargious, Reference Hargreaves, Pexman, Zdrazilova and Sargious2012). Analyses such as these reveal learners’ individuality in the FL learning process and enable applied linguists to identify consistent patterns of individual differences as well as phases with significant increase/decline – and the factors that cause them. A particular emphasis will be on the role of social participation, a feeling of belonging and meaning, that is facilitated through language learning. Is there any evidence that actual use of the learned L2 makes a contribution, whether direct or indirect, to the highly desirable state of social well-being?

5. Teaching additional languages to third age learners

As a consequence of current demographic changes, third-age universities are flourishing throughout Europe (e.g. ‘Uni 55plus’ in Austria, ‘Senioren-Universität’ in Switzerland, ‘Universités du troisième Age/Tous Age/du Temps Libre’ in France, ‘Uniwersytet Trzeciego Wieku’ in Poland). Furthermore, the last decade of the twentieth century saw the emergence of ‘critical educational gerontology (CEG)’, a field which ‘conceives older adults as in control of their thinking and learning, and more importantly, as capable of further development’ (Formosa, 2002; quoted in Ramírez Gómez, Reference Isopahkala-Bouret2016: 96). CEG addresses the relations between ageing, economic conditions, and the activity of the state in regard to educational policy vis-à-vis older adults – education that liberates rather than oppresses (see e.g. Warnes, Friedrich, Kellaher, & Torres, Reference Warnes, Friedrich, Kellaher and Torres2004; Warnes & Williams, Reference Warnes and Williams2006; Ramírez Gómez, Reference Ramírez Gómez2015). An increasing body of research focuses on the extent of older adult success in what and how they learn (Hartshorne & Germine, Reference Hartshorne and Germine2015; Amer, Campbell, & Hasher, Reference Amer, Campbell and Hasher2016), rather than possible failure to learn as much or as quickly as young adults. ‘Best practice’ in older adult language education has developed little since the 1980s, when Joiner (Reference Joiner1981) devised an optimal, goal-driven learning programme with individualised, self-paced instruction at its core, and the inclusion of real-world materials to link learning to practice and boost motivation.

According to Knowles, Holton, and Swanson (Reference Knowles, Holton and Swanson2011), there are three sets of variables especially worth considering in adult education:

• • Core learning principles: the learner's self-concept, prior experience, readiness to learn, orientation to learning and motivation, and learners’ need to know why they are learning what they are learning.

• • Individual and situational differences, which include subject-matter differences (e.g. the level of complexity of the content), situational differences (e.g. location, number of learners in the class) and individual differences (e.g. personality, learner background).

• • Goals and purposes for learning, which help coordinate the effort and direction of the learning process; according to Beder (Reference Beder, Merriam and Cunningham1989), these may have an individual, institutional or societal nature.

According to certain scholars (e.g. Grotek, Reference Grotek and Gabryś-Barker2018) benefits may be observed only if participants feel ‘safe’ and at ease in the learning environment. This is claimed to be achievable via small (homogeneous) groups, a positive stance towards participants’ L1, and the use of non-threatening material in the form of relatable, restricted content, visual cues and an absence of digital materials (despite the learning benefits claimed for these) (Schleppegrell, Reference Schleppegrell1987) – see also the main principles of ‘critical foreign language geragogy’ outlined in Ramírez Gómez (Reference Ramírez Gómez2016, p. 108). Several authors (e.g. Oxford, Reference Oxford and Gabrys-Barker2018; Pawlak, Derenowski, & Mystkowska-Wiertelak, Reference Pawlak, Derenowski, Mystkowska-Wiertelak and Gabrys-Barker2018; Piechurska-Kuciel & Szyszka, Reference Piechurska-Kuciel, Szyszka and Gabrys-Barker2018) recommend conducting strategy instruction, woven into language teaching, and involving activities from ‘sage-ing’, i.e. a ‘late-life … process that enables older people to become spiritually radiant, physically vital, and socially responsible ‘elders of the tribe’’, who are full of ‘meaning and purpose’ (Schachter-Shalomi & Miller, Reference Schachter-Shalomi and Miller1995, pp. 5–6), mindfulness (see also Tolle, Reference Tolle2004; Williams & Penman, Reference Williams and Penman2012; Kabat-Zinn, Reference Kabat-Zinn2013), and the selection, optimisation, and compensation process (see also Derenowski, Reference Derenowski and Gabryś-Barker2018).

Third-agers’ prior experience of learning and mental models provides the resources which may facilitate their coping with a new learning situation in later life (Knowles et al., Reference Knowles, Holton and Swanson2011). Park et al. (Reference Park, Lodi-Smith, Drew, Haber, Hebrank, Bischof and Aamodt2014) showed that in older adults, the beneficial effect of productive engagement in demanding tasks and involvement in a social setting is highest if these are combined, accordingly underlining the importance of social exchange in cognitive training interventions with older learners. Thus, the immediate feedback provided by an instructor and the interaction between instructor and learner may also play an essential and beneficial role for both the learners’ motivation and the learning outcome.

Oxford (Reference Oxford and Gabrys-Barker2018) emphasises that any differentiation of language instruction for teaching third-agers should take into account interests and desires; adjust instruction to learners’ sensory preferences and abilities; provide instruction that matches learners’ cognitive levels, cognitive styles, and PLs in the language being learned; and tune instruction to the level and types of learning strategy in use. According to Chamot (Reference Chamot, Oxford and Amerstorfer2017) differentiation also includes paying attention to diversity in personality factors, motivation and willingness. Learners of all ages benefit from seeing that they have options in respect of how to express themselves: ‘every time they use language, they are making choices, and by so doing, negotiating their identities’ (Larsen-Freeman, Reference Larsen-Freeman, Ortega and Han2017, p. 28).

6. Extending the research agenda

One of the challenges in this line of research is the prevalent ‘deficit view’ inadvertently included in the representation of research findings: evaluating older language learners against ‘young learner’ competence, i.e. setting the criterion for ‘success’ in L2 as the level of proficiency associated with young adults; participating in a rhetoric of age as an internal causal factor, thereby de-socialising age (Rughiniș & Humă, Reference Rughiniș and Humă2015); estimating ageing effects in the aggregate, i.e. grouping third-agers into cohorts according to their age, thus dismissing people's agency in shaping their lives, and the role of social variation and change (Rughiniș & Humă, Reference Rughiniș and Humă2015); using patronising teaching materials (e.g. explicitly labelled ‘adult learner’ course books); using tasks for the comparison of age groups that compare performance between groups whose experience varies; speaking about the third-age learner as a failure for being unlike the younger learner and about lifelong learning ‘as a practice to battle against old age’ (Isopahkala-Bouret, Reference Isopahkala-Bouret2015, p. 2); and finally, anticipating, through the statistical parlance about ‘age effects’, matter-of-factly representing them as underlying deficient behaviour (Rughiniș & Humă, Reference Rughiniș and Humă2015).

The main question is, to use Rughiniș and Humă’s (Reference Rughiniș and Humă2015, p. 147) words: ‘how can one avoid methodological reductionism in interpreting findings, so that the “age effect” is not implicitly rendered synonymous with depreciation?’ Getting rid of stereotypes in research design, task design and data analysis is a fundamental step towards understanding age-related changes and age parameters in SLA, and, by extension, encouraging older language learners and creating more positive ageing identities not only at an individual level but at a societal level. Focusing on within-learner generalisability – rather than between-learner generalisations – might be a step in the right direction, considering that (1) there is particularly great inter-variability and heterogeneity among older adults, which makes statistical generalisations across subjects almost impossible, and (2) L2 learning is a complex, non-linear process. Dynamic systems perspectives (see Ortega & Han, Reference Ortega and Han2017) on third-age learning are particularly appealing if we want to make a distinction between chronological age and social or ‘contextual’ age; the former is understood to be a ‘predictable, even ineluctable progression along an incremental scale,’ while the latter is ‘a far less predictable ebbing and flowing, reflecting the arrival and passing of particular somatic, experiential and emotional circumstances’ (Coupland, Coupland, & Giles, Reference Coupland, Coupland and Giles1991, p. 140; see also Divita, Reference Divita2014). In contrast to the intervention studies mentioned above, in which L2 learners in later life were tested only twice or thrice (usually at the beginning and end of L2 training), it is thus vital to make use of denser time serial measurements in order to focus on micro-development and idiodynamic trajectories, capturing and capitalising on individuality in the L2 learning process, homing in on variability, non-linear development, and progression in iterative steps (see Singleton & Pfenninger, Reference Singleton and Pfenninger2019, for specific research agendas). Data from such iterated investigation of third-age learning will make a substantial contribution to research on late L2 acquisition as well as on cognitive aspects of healthy and active ageing. That way age ceases to be a ‘predictor’ and instead becomes a correlate, in which typical ages are described by typical situations and configurations.

Another issue with the conceptualisation of third age is the prioritising of the concept of ‘successful ageing’ over the idea of ‘harmonious ageing’ in age-based discourse. In Liang and Luo's (Reference Liang and Luo2012) theory critique, the discourse of successful ageing is identified as problematic with respect to four dimensions: first, successful ageing is ageist in nature, and it produces a disharmony between body and mind. Second, the capitalist and consumerist components of successful ageing are under-addressed (successful ageing is often used as the trademark for selling the idea of ‘lifestyles’). Third, successful ageing is a discourse developed on the basis of Western (specifically, American) values and may not readily apply to other cultures. Fourth, successful ageing overlooks the deeper concern of quality experience (e.g. it overemphasises frequency of activity participation). For instance, Carlson et al. (Reference Carlson, Parisi, Xia, Xue, Rebok, Bandeen-Roche and Fried2012) found that participation in a variety of lifestyle activities is more predictive than frequency or level of cognitive challenge for significant reductions in risk of impairment on measures sensitive to cognitive ageing and risk for dementia.

Furthermore, knowledge about identity construction and the subjectivity of language acquisition and use in later life remains curiously thin. How do the interior and external logics of adult ageing compare, how is identity performed in different acquisitional settings, and what implications does this have for L2 acquisition and L2 pedagogy? How do particular approaches interact with age-identity? We need more studies that are distinguished by their focus on contingency and diversity rather than their focus on difference, studies that are based on encouraging third-age learners to develop and express their own ‘voice’ and to articulate in their own words what learning an additional language means to them. One way to resist age stereotyping is to let third-agers define for themselves the category of ‘ageing learner’ or ‘older learner’ and make sense of their ageing selves; lifelong learning clearly plays a part in new ageing identities. In Gilleard and Higgs’ (Reference Gilleard and Higgs2013, p. 20) words, ‘these new ways of narrating and performing age have emphasised desire, potential and agency over need, vulnerability and limitation’. Along those lines, Eckert (Reference Eckert and Coulmas1998, p. 165) recommends a shift in focus ‘away from chronological age and towards life experiences that give age meaning’ (Reference Eckert and Coulmas1998, p. 165) – which explains the need for single-case studies and case-series studies, i.e. qualitative approaches.

Related to this, there is also the question of the extent to which research studies view third-age learners as human subjects of study, and to what extent they view them as co-participants, mobilising them to co-operate in the data collection process and/or in the development of teaching methodologies, thus contributing to the growing tradition of Citizen Science. Biggs (Reference Biggs2005, p. 124) suggests that ‘the use of older researchers to research older people may go some way to reduce biases associated with the unequal distribution of power between researchers sponsored by state organizations and those who use their services.’

Since age-identities are managed as much in research situations as in any other (Biggs, Reference Biggs2005), it is also of utmost importance that younger researchers be sensitised to the dangers of imbalances in generational power, which can lead to biased data. Younger researchers cannot assume their questions will be perceived as neutral and that they are protected by a research persona; e.g. the identity presented to the researcher may only partially reflect the experience of ageing or the perspective of the older adult.

From a quantitative perspective, an important question concerns the methodologies needed to study complex dynamic social systems. By working with age as a ‘socio-demographic variable’, quantitative researchers often convert it (inadvertently) into an exclusively quasi-biological feature (see Rughiniș & Humă, Reference Rughiniș and Humă2015), which does not do justice to the socio-affective and contextual nature of the age factor. Accordingly, statistical methods are necessary which are able adequately to analyse change and account for variability and auto-correlation in linear as well as non-linear patterns (i.e. the short-term and long-term developmental trajectories of older adults). Traditional ANOVA-based analyses, which run on averaged data and thus aggregate participants, define ageing as a uniform process across cohorts, thus encouraging a biological interpretation. In such analyses, it is cohort that is of principal interest for researchers, and not ageing, that is, changes in cohorts’ behaviour while they progress through life. As Rughiniș and Humă (Reference Rughiniș and Humă2015) rightly point out, cohort and ageing represent two distinctive (and potentially contrary) sources of variability. Furthermore, sample-based research does not tell us about individuals. This is where the generalised (mixed-effects) regression framework comes into its own, including generalised additive (mixed) models (GAM(M)s), which represent an important statistical development of the last 45 years and provide an extremely useful set of tools for the dynamic analysis of linguistic data within the field of age-related research.

Finally, the findings presented in the intervention studies outlined above are preliminary at this point and the conclusions drawn from them must be considered against the methodological constraints inherent in their design. However, such preliminary results, while requiring sensible critique, are encouraging and seem to warrant more research with larger sample sizes and different methodologies to obtain a clearer and more detailed picture.