4.1 IS

Most of the studies reviewed in this section approach IS from a ‘generic or L1 perspective’, that is, they usually do not specify the particular language(s) in which IS is carried on, indirectly assuming IS is conducted—as if by default—in one's native language. In empirical studies, this L1 is frequently not specified, nor is the participants’ knowledge of other languages taken into consideration. This view thus either implicitly adheres to a monolingual perspective or disregards the participants’ bi- or multilingualism as irrelevant to the study of IS. It should be mentioned, however, that some reviews of the literature do recognize variation in language experience (Alderson-Day & Fernyhough Reference Alderson-Day and Fernyhough2015a) or bilingualism (Morin Reference Morin and Ramachandran2012) as topics of interest related to IS research.

Research on IS from a generic-L1 perspective has been intensively produced in recent years, including thorough reviews or syntheses of the literature (e.g., Brown Reference Brown2009; Morin Reference Morin and Ramachandran2012; Perrone-Bertolotti et al. Reference Perrone-Bertolotti, Rapin, Lachaux, Baciu and Loevenbruck2014; Alderson-Day & Fernyhough Reference Alderson-Day and Fernyhough2015a) as well as empirical studies. This research has a predominantly ‘adult’ or ‘fully developed’ IS focus, insofar as studies using child populations usually fall under the realm of ‘PS’ or ‘developmental IS’ research (to be discussed below).

One area of studies centers on features of IS that had been underexplored in previous research, such as functions, contents, and varieties of IS. A study by Morin, Uttl & Hamper (Reference Morin, Uttl and Hamper2011), for example, showed that the contents of IS were predominantly self-referential; i.e., IS dealt with thoughts about the respondents’ own emotions, physical appearance, relationships, and the like. In addition, the most frequently reported functions were self-regulation (planning, controlling emotions), memory (remembering, rehearsing what to say), and problem-solving. DaSilveira & Gomes (Reference DaSilveira and Gomes2012) uncovered an interesting relationship between the contents of IS and problem-solving: when IS was evenly distributed to address the problem and to address the self, task performance tended to be successful; however, performance was affected negatively by increased verbalizations about the self or unrelated to the task and by ruminative inner dialog (consisting of redundant or repetitive thinking). Various specific properties of IS—dialogicality, condensed or expanded form, evaluative and motivational functions, and hearing other people's voices in the head—were explored by McCarthy-Jones & Fernyhough (Reference McCarthy-Jones and Fernyhough2011), who found the dialogical variety of IS to be much more frequent than the condensed one. Based on college-student population data, the three studies mentioned above, however, provide a very particular profile of IS. Studies to be conducted with more diverse populations may offer differential results.

Another area of interest relates to the VWM model, which views IS as a dual-natured phenomenon consisting of an articulatory (inner voice or inner speaking) and an auditory (inner ear or inner hearing) component, both part of the phonological loop system (Buchsbaum Reference Buchsbaum2013). The key role of articulatory IS in VWM was confirmed in studies showing the detrimental effects of tampering with covert articulation during silent verbal tasks (Durisko Reference Durisko2006; Aguilar & Aguilar Reference Aguilar and Aguilar2008). A study by Scott (Reference Scott2013) on the auditory component of IS, in turn, argued that the experience of hearing the sound of one's IS is due to a brain mechanism known as corollary discharge. The involvement of auditory speech imagery in silent reading or writing is actually an area which has shown conflicting results. Filik & Barber (Reference Filik and Barber2011), for example, demonstrated that auditory imagery during silent reading is influenced by the reader's familiarity with features of the author's voice or, if this is unknown, by the reader's own regional accent. Perrone-Bertolotti et al.’s (Reference Perrone-Bertolotti, Rapin, Lachaux, Baciu and Loevenbruck2014) review, however, shows that IS during reading or writing may not always involve auditory imagery and may in fact adopt non-phonological modalities, such as inner signing, graphic imagery, and abstract linguistic representations.

In the 2005–2015 timespan, research on IS has also been actively pursued regarding special populations, that is, individuals with some type of psychological, social, or linguistic condition or disorder. Several studies have dealt with IS in the form of pathological voice hallucinations (Allen, Aleman & McGuire Reference Allen, Aleman and McGuire2007; Simons et al. Reference Simons, Tracy, Sanghera, O'Daly, Gilleen, Dominguez, Krabbendam and Shergill2010). Research has focused on dysfunctional self-talk among individuals affected by test anxiety, anorexia, depression, or insomnia (Morin Reference Morin and Ramachandran2012). Investigations have also targeted populations affected by communicative problems or impairments, such as autism, deafness, and aphasia (Perrone-Bertolotti et al. Reference Perrone-Bertolotti, Rapin, Lachaux, Baciu and Loevenbruck2014; Alderson-Day & Fernyhough Reference Alderson-Day and Fernyhough2015a). Because of space limitations, only one study will be mentioned to exemplify the important implications about the nature of IS that can be derived from these specialized areas of IS research, namely, the study by Williams, Bowler & Jarrold (Reference Williams, Bowler and Jarrold2012), which showed deficits in dialogic—but not monologic—IS among autistic individuals, deficits also associated to the severity of their communication impairments. The study points to the importance of early-age interpersonal interactions in the development of intrapersonal dialogic communication.

Finally, intensive work has been produced in recent years on the cerebral correlates of IS, with an abundance of studies relying on neuroimaging technology. Overall, results from studies on healthy adult individuals reveal cerebral IS activity to be extremely complex and dynamic, involving multiple neural regions and networks of activation depending on the particular language task being carried out and the processes involved (Price Reference Price2012; Alderson-Day & Fernyhough Reference Alderson-Day and Fernyhough2015a). Areas that have consistently shown to be the site of IS activity during silent verbal rehearsal or repetition are the left inferior frontal gyrus and the left supramarginal gyrus (Geva et al. Reference Geva, Jones, Crinion, Price, Baron and Warburton2011; Morin Reference Morin and Ramachandran2012). Studies reviewed by Perrone-Bertolotti et al. (Reference Perrone-Bertolotti, Rapin, Lachaux, Baciu and Loevenbruck2014) also show that, although IS and external speech share certain brain regions typically associated with language processing (such as Broca's and Wernicke's areas), other distinct areas and networks of activation have been detected for covert and overt speech. Researchers concur that IS cannot be simply described as overt speech minus the articulatory motor component; however, ‘the level at which the speech production process is interrupted (abstract linguistic representation vs. articulatory representation) is still debated’ (Perrone-Bertolotti et al. Reference Perrone-Bertolotti, Rapin, Lachaux, Baciu and Loevenbruck2014: 227; see also Geva et al. Reference Geva, Jones, Crinion, Price, Baron and Warburton2011).

4.2 PS

In the period 2005–2015, research on the covert uses of language from a generic-L1 perspective has included looking into the PS phenomenon, predominantly as part of children's IS development and to a lesser extent as manifested among adults. Recent research on child PS has been able to confirm its use for a variety of functions or purposes, such as problem-solving (Damianova, Lucas & Sullivan Reference Damianova, Lucas and Sullivan2012; Alarcón-Rubio, Sánchez-Medina & Prieto-García Reference Alarcón-Rubio, Sánchez-Medina and Prieto-García2014; McGonigle-Chalmers, Slater & Smith Reference McGonigle-Chalmers, Slater and Smith2014), planning and numeracy tasks (Lidstone, Meins & Fernyhough Reference Lidstone, Meins and Fernyhough2010, Reference Lidstone, Meins and Fernyhough2011), referential communication (San Martín, Boada & Feigenbaum Reference San Martín, Boada and Feigenbaum2011), free play (Davis, Meins & Fernyhough Reference Davis, Meins and Fernyhough2013), and regulating emotions (Day & Smith Reference Day and Smith2013). Lately, interest has also gathered on social aspects of PS, such as its role in social interactive discourse (San Martín et al. Reference San Martín, Boada and Feigenbaum2011) or the stimulating effect of potential listeners or interactants (Damianova et al. Reference Damianova, Lucas and Sullivan2012; McGonigle-Chalmers et al. Reference McGonigle-Chalmers, Slater and Smith2014).

Most of the current studies on PS among children provide support for Vygotsky's (Reference Vygotsky1986) general hypothesis that vocalized PS emerges in early childhood, peaks right before school age, and fades off in the school years as children learn to regulate their mental processes through silent IS (Lidstone et al. Reference Lidstone, Meins and Fernyhough2011; San Martín et al. Reference San Martín, Boada and Feigenbaum2011; Damianova et al. Reference Damianova, Lucas and Sullivan2012; Alarcón-Rubio et al. Reference Alarcón-Rubio, Sánchez-Medina and Prieto-García2014). Lidstone et al. Reference Lidstone, Meins and Fernyhough2011 investigated the occurrence of PS in middle childhood (7–10 years old), an age when children are typically transitioning from fully externalized PS to inaudible IS. To measure levels of internalization, the researchers utilized the following scale, Level 5 representing the child's developing ability to mediate cognition through covert, rather than overt, speech.

The pattern of development suggested by this scale, sometimes described as an inverted U-shaped pattern of internalization, is moderated by findings showing that audible PS does not totally disappear with age but persists among older children, adolescents, and adults as a useful non-primary form of verbal self-mediation. According to Duncan & Tarulli, however, the persistence of PS after childhood does not alter the notion of childhood as a period of internalization, ‘in the sense that it may be a time when children form the capacity for covert verbal thinking’ (Reference Duncan, Tarulli, Winsler, Fernyhough and Montero2009: 176).

4.3 Implications of generic-L1 inner/private speech research for covert use of the L2

Late research on IS/PS conducted from a generic-L1 perspective cannot but confirm the essential role that language—any language—has as an instrument of thought. This research also suggests the covert use of language fulfills a wide range of functions, mostly self-regulatory and self-referential. If IS and PS respond to the basic intellectual needs of an individual, a similar role might then be expected from L2 IS/PS. But, how specific will use of an L2 be for particular intramental functions? Are there social and affective roles for L2 PS, as in L1 PS? Whereas command in the L1 among monolinguals represents the availability of a very efficient psychological tool, to what extent can full or partial knowledge of another language or multiple languages facilitate or obstruct the intramental life of an individual? Research on generic-L1 IS suggests that the dialogic form of IS is prevalent (at least among certain populations) and that too much ruminative IS can negatively affect problem-solving. Are these findings replicated in L2 IS? Strong evidence for the role of PS as a genetic precursor for IS, its persistence in adulthood, and the definite evolution towards full covertness of speech-for-the-self suggest similar patterns will occur in L2/LX learning. It is also clear from generic-L1 IS research that covert linguistic activity involves multiple areas and complex patterns of brain activation as well as articulatory and auditory representations, the extent and manner of which are still partially understood. Going into these issues from an L2 IS perspective may yield interesting similarities and differences. Recent studies on L2 IS/PS discussed in the next section will address many of these concerns, yet the full potential of generic-L1 research in terms of its implications for L2 IS/PS remains to be explored.

5.1 Empirical research on L2 private speech

In the past decade, a significant amount of empirical research has focused on the overt manifestations of self-directed speech among L2 learners and users. This line of research is of interest for two main reasons: (1) it offers evidence of the functions and the extent to which an L2 is used intramentally, and (2) it provides insights into the process of L2 internalization and development of L2 IS.

As in past research, new studies have called attention to L2 PS for self-regulation among L2 learners or users while performing challenging cognitive tasks. For example, in a study conducted with a group of 30 Spanish-English bilinguals (Spanish-dominant, English-dominant, and balanced), Jiménez-Jiménez (Reference Jiménez-Jiménez2015) found massive use of PS in both languages for self-regulation as the participants worked independently in solving question-like problems in Spanish. Participants who were more dominant in one of the languages, however, tended to use this language more frequently for spontaneous verbalized reasoning. The balanced bilinguals made use of L1 and L2 in their PS, with no clear pattern of preference for either language.

Whereas the above study provides evidence of PS for purposes of self-regulation in solitary conditions, several studies have revealed the occurrence and functions of self-directed talk in social and collaborative L2 scenarios (Smith Reference Smith2007; Yoshida Reference Yoshida2009; Sönmez Reference Sönmez2011; Steinbach-Kohler & Thorne Reference Steinbach-Kohler, Thorne, Hall, Hellermann, Doehler and Olsher2011; Hauser Reference Hauser2015). PS of this type, in addition to being a self-regulatory mechanism, appears to have an important social dimension and other-regulating functions. For example, in a study involving school children interactively playing board games especially designed to learn English (the L2), Smith (Reference Smith2007) found that although PS was primarily produced for intramental purposes, it was also attended to by peers as if it were social in nature. Indeed, Smith showed that the naturally occurring PS not only helped the children self-regulate their thinking in and about the L2, it also had intermental functions, such as calling peers’ attention to and encouraging verbalization on certain aspects of the joint activity. In Yoshida's (Reference Yoshida2009) study, the Japanese learners’ PS served important social and affective functions in that it helped learners maintain involvement in classroom interaction and express affection through covert participation. In Steinbach-Kohler & Thorne, it was also shown that the self-directed talk of three L1 Swiss German adolescents working collectively on an L2 French pedagogical task worked as a ‘publicly available resource’ (Reference Steinbach-Kohler, Thorne, Hall, Hellermann, Doehler and Olsher2011: 71), useful for the maintenance of intersubjectivity and joint problem-solving. Lastly, two other studies point out the social functions—in addition to the intramental ones—of L2 PS produced in interactive contexts: (a) a group of non-native speakers of English (Turkish L1) socially interacting over a word game with a group of native English speakers (Sönmez Reference Sönmez2011) and (b) a non-native speaker of English (Japanese L1) practicing English in conversation with a native English speaker (Hauser Reference Hauser2015).

Recent studies have shown that externalized PS functions not just as a self-regulatory mechanism aiding problem-solving but as a tool in learning and internalizing the L2. This research confirms the occurrence of previously documented PS behaviors such as repetitive and transformative imitation, rehearsal for memorization or before public speaking, vicarious response, language play and manipulation, translation into the L1, and displays of metalinguistic awareness (Guerrero Reference Guerrero and Chapelle2012b, Reference Guerrero and Commander2013). These forms of PS appear to have an internalizing role, providing learners the opportunity to experiment with and appropriate elements of the L2, obtain meaning from L2 affordances, reinforce existing knowledge, and creatively extend the L2 to possible or imaginary scenarios. In Vygotskyan SCT, overt PS, that is, speech that in form and function has lost its social-communicative nature but is still observable, as it occurs among L2 learners, is a reflection of the L2 internalization process, representing an early stage in the development of L2 IS (Guerrero Reference Guerrero2005).

A recent study (Yi Reference Yi2010) exemplifies the internalizing role of overt L2 PS and developmental progression towards IS. In Yi's study, three Korean preschoolers were observed as they learned English in a US classroom over the course of one year. The children's language learning process started with active observation of L2 use, progressed to overt self-directed vocalizations of the L2, and gradually developed into social speech as exposure to and interactions in English continued. PS verbalizations were used mainly to learn new words and self-direct during solitary play. As verbal communication increased, PS became non-observable, turning inwards as silent IS: ‘The ELLs [English language learners] no longer used private speech when they internalized the learning and it became inner speech’ (2010: 167).

PS for language learning has also been documented among preschoolers in English as a foreign language (EFL) classrooms. In Wang & Hyun (Reference Wang and Hyun2009), for example, children in Taiwan, fluent in Mandarin Chinese, used PS to memorize English words, practice pronunciation, formulate sentences, rehearse for future production, and perform several language learning tasks. Sometimes, translation into the L1 was used to facilitate comprehension. In the following example, a child is shown using several PS strategies to facilitate learning of the English word foot, from meaning-making drawing and code-switching to repetition for memorization.

This type of PS is not restricted to small children. There is recent robust evidence that older children and adolescents (Smith Reference Smith2007; Fallahchai Reference Fallahchai2011; Steinbach-Kohler & Thorne Reference Steinbach-Kohler, Thorne, Hall, Hellermann, Doehler and Olsher2011) as well as adults (Borer Reference Borer2007; Yoshida Reference Yoshida2009; Sönmez Reference Sönmez2011; Abadikhah & Khorshidi Reference Abadikhah and Khorshidi2013; Stafford Reference Stafford2013) use self-directed talk to learn and internalize a language in classroom and social environments. Some studies have demonstrated the beneficial effects of using PS in the form of rehearsal to learn FL vocabulary (Borer Reference Borer2007; Fallahchai Reference Fallahchai2011). Borer found that repetition was the most frequent form of self-vocalization to retain new words; however, it was elaborative rehearsal involving deep processing, not simple repetition, that enhanced vocabulary retention. Deep processing through manipulation and generation signaled ‘movement toward internalization’ (Reference Borer2007: 278).

In Yoshida's (Reference Yoshida2009) study, spontaneous repetition was also the most common type of PS among six (English and/or Chinese L1) college learners of Japanese. Other functions of PS in Yoshida's study, in addition to repetition, were vicarious response to the teacher's questions; manipulating sounds, forms, or structures; reading aloud; commenting on new words or forms; and practicing task answers. The following excerpt from Yoshida reveals an instance in which subvocal repetition of a teacher's model turns into transformative imitation; that is, the learner softly repeats the input word chokoreito extending it to an utterance of her own creation.

Further evidence on the nature of L2 PS for internalizing purposes comes from Lee (Reference Lee2006, Reference Lee2008), who focused on the PS produced by seven bilingual (Korean-L1, advanced English proficiency) students enrolled at a US university as they were preparing alone for a biology exam. Although their goal was to study for the exam, the activity also turned into a language learning task, with PS as an important mediational tool. Their PS had several main functions: establishing meanings; mental rehearsal for memorization; monitoring, planning, and self-motivation; and expressing emotions. A notable finding in Lee's research was that the PS of the participants was produced not only orally but also in the form of written notes and drawings as well as communicative and self-regulatory gestures (see also Yoshida Reference Yoshida2009, for PS as note-taking, and Steinbach-Kohler & Thorne Reference Steinbach-Kohler, Thorne, Hall, Hellermann, Doehler and Olsher2011, for private gesturing).

5.2 Empirical research on L2 IS

Several recent data-based studies on the covert uses of an L2 to mediate verbal thinking, self-regulation, and self-communication will be reviewed in this section. Most of these studies are concerned with the question ‘What language do L2 speakers, bilinguals, or multilinguals use for thinking?’ This body of research throws light on whether and to what extent an LX is used in IS as well as the variables that affect its use and development.

Important information on the language of choice for IS among multilinguals has been offered by Dewaele (Reference Dewaele2015) in a series of studies based on data gathered through the Bilingualism and Emotions Questionnaire (BEQ) (Dewaele-Pavlenko Reference Dewaele and Pavlenko2001–Reference Slobin, Gentner and Goldin-Meadow2003). This web-administered questionnaire, which focused on multilinguals’ expression of emotions, included two questions on the silent use of language: ‘Which language(s) do you use for mental calculations/arithmetic?’ and ‘If you form sentences silently (inner speech), what language do you typically use?’ Results of the BEQ based on responses from 1,579 multilinguals representing 77 different L1s and up to four additional languages (L2, L3, L4, and L5) showed an overall preference for using the L1 for IS (Dewaele Reference Dewaele2015). More than a third (37.9%) of the overall BEQ respondents, however, indicated using their L2 ‘frequently’ or ‘all the time’ for mental calculations and more than half (56.6%) using the L2 ‘frequently’ or ‘all the time’ for IS (Dewaele Reference Dewaele2009). L3s, L4s, and L5s were used decreasingly less for these cognitive functions.

Dewaele's (Reference Dewaele2015) research also showed that early acquirers, i.e., people who had started learning their L2 and L3 at ages 3 to 7, had higher rates of use of IS in these languages than late acquirers. Age of onset of acquisition was thus found to be a predictor of L2 and L3 use as language for IS and mental calculations. The strongest predictor of choice of language for IS, however, was frequency of use in everyday interactions. Other variables that appeared to predict whether an L2 or L3 would become the medium for IS were proficiency and socialization in the language, size of network of interactants, context of acquisition, and number of learned languages. Context of acquisition was indeed a significant predictor for the language of mental calculations, suggesting that the type of setting—whether naturalistic, instructed, or mixture of both—in which performing mental calculations is learned will have an impact on the language chosen to perform those operations in adult life. An interesting result regarding the gender variable (one that is seldom tested in IS studies) was that females reported using the L2 significantly more often than males for IS. Finally, another remarkable result was the multilinguals’ choice of the L1 as the preferred language of ‘emotional inner speech’ (internal use of language to express deepest feelings). Moreover, Dewaele found that IS use in general was greater than emotional IS in any LX, suggesting ‘that it takes a while before an LX becomes internalized to the point of becoming a multilingual's “language of the heart”’ (Dewaele Reference Dewaele2015: 15).

Some studies have replicated, at least partially, Dewaele's research on the question of multilinguals’ preference of language for IS. Hammer (cited in Dewaele Reference Dewaele2015) found that higher levels of acculturation and greater size of the L2 speaking social network were correlated with higher use of the L2 for IS among Polish L1-English L2 bilinguals residing in the UK. Ewert (Reference Ewert, Arabski and Wojtaszek2010) also explored language preferences for internal and external uses. The sample consisted of 17 Polish L1 multilinguals enrolled in a combined Russian/English degree program at a Polish university. All participants used these three languages frequently for a variety of academic and social purposes. Responses to a self-report questionnaire indicated that Polish, the L1, was the preferred language for internal functions, such as performing mental calculations, making private notes, praying, dreaming, swearing, and formulating IS sentences. The two other languages were used sometimes or rarely for these functions, except for English, which was used frequently or sometimes for IS sentences and swearing. An explanation for this phenomenon might be that swear words sound much stronger in the L1 than in the L2; perhaps participants were trying to lessen the emotional charge of these words by rendering them in the L2. Also, interestingly, private prayer was for most participants conducted exclusively in the L1. Overall, the findings reflect that preference for a particular language for internal purposes is consistent with how dominant this language is in the life of the individual. Ewert's study additionally revealed (through questions not derived from the BEQ) that mental code-switching was done sometimes or frequently for talking to oneself and taking private notes but not so much for arithmetic tasks and that mental translation, i.e., formulating thoughts in one language and expressing them in another when writing or speaking, was done rarely.

Using an instrument of his own, Schrauf (Reference Schrauf2009) continued the investigation of L2 IS among bilingual immigrants (Larsen et al. Reference Larsen, Schrauf, Fromholt and Rubin2002) by focusing particularly on a cohort (N = 60) of Spanish L1 Puerto Rican elders who had migrated to the US in their twenties. In this study, the participants self-rated their frequency of use of English and Spanish in social and psychological domains, this latter domain including cognitive and affective functions of internal self-directed speech. Results showed that the immigrants with higher levels of L2 proficiency made greater use of the L2 not only in the social but also in the internal domain, for both cognitive (thinking, talking to themselves, and counting) and affective (expressing feelings, swearing, and dreaming) functions. However, no predominant shift to the L2 was found in any social or psychological domain, even among the most fluent L2 users. The study suggests that it is possible for long-time immigrants living in ethnically and linguistically concentrated communities to keep adherence to the L1 despite development of high proficiency in the L2.

An insightful and original perspective on the issue of language shift in the IS of multilinguals comes from Pavlenko's (Reference Pavlenko2011a, Reference Pavlenko2014) analyses of autobiographical narratives (journals, memoirs) of individuals, many of them well-known writers and scholars, who acquired LXs as part of their migration to linguistic communities different from their native ones. Testimonies from such individuals as Mary Antin, Veronica Zhengdao Ye, Gerda Lerner, Eva Hoffman, Ariel Dorfman, Kyoko Mori, Tzvetan Todorov, and others show that being able to think in a new language involves developing new concepts and perspectives and frequently a new sense of self. Based on these testimonies, Pavlenko has identified several turning points in an individual's journey into a new mode of verbal thinking:

1. 1. Upon arriving into a new linguistic community, the migrant continues to use the L1 as a medium of thought.

2. 2. After some time, there is a loss of IS, as the L1 begins to fade as a functional medium of thought and the L2 has not been sufficiently developed yet. This is usually a very difficult and frustrating time as the person is ‘caught between languages’ and unable ‘to render one's thoughts in either’ (Pavlenko Reference Pavlenko2011a: 17).

3. 3. With time and intensive involvement in the new community, the L2 is internalized, becoming the dominant vehicle for thinking, a process that shapes the individual's IS into a new way of perceiving the world and self-communicating.

4. 4. Frequently, loss of the L1 as a meaningful mechanism for the expression of thoughts and emotions entails the inability to translate one's thoughts from one language to another.

5. 5. Finally comes the feeling of being a different self in the new language and, for some, a sense of accomplishment in finding a new internal voice: ‘It took several years before I began to think in English. It was exciting when it actually happened and it made a qualitative difference in the way I lived,’ wrote Gerda Lerner, a German-English translingual (cited in Pavlenko Reference Pavlenko2014: 271).

A few empirical studies provide additional insights on language learners’ preferences of language for IS. Choi & Lantolf's (Reference Choi and Lantolf2008), for example, examined the ‘thinking for speaking’ (Slobin Reference Slobin, Gentner and Goldin-Meadow2003) patterns of four advanced Korean and English L2 learners to observe shifts from L1 to L2 in the use of hand gestures while speaking in the L2. Because gesturing when speaking reflects common underlying mental processes and semiotic systems, gestures may also offer clues about choices of language for IS. In Choi & Lantolf's study, some evidence of a shift from L1 to L2 was found in gestures indicating the ‘path’ of a motion event, thus corroborating earlier results by Negueruela et al. (Reference Negueruela, Lantolf, Jordan and Gelabert2004) and Stam (Reference Stam2006); however, very little evidence of such a shift was obtained for gestures involving ‘manner’ of motion, a shift which possibly also involves a change in conceptual meanings. Previous results for manner of motion had been conflicting: Negueruela et al.’s and Stam's studies showing no switch, Özyürek (Reference Özyürek, Skarabela, Fish and Do2002) providing positive evidence of a shift among very advanced Turkish L1-English L2 speakers. In a recent study by Gullberg (Reference Gullberg, Bohnemeyer and Pederson2011), however, advanced speakers of Dutch and French were able to use gestures that were linguistically appropriate to each language to indicate object placement, suggesting that a linguistic shift in thinking for speaking had taken place. It is clear that much further research is needed on gesture-speech processes that might throw light on the nature of bi- and multilinguals’ language of IS, particularly as it involves the adoption of new (L2) conceptual meanings.

Another study which looked into language preference for IS was Gabryś-Barker (Reference Gabryś-Barker2006, Reference Gabryś-Barker, Piasecka, Adams-Tukiendorf and Wilk2015). The researcher observed how 48 trilinguals (Portuguese L1, English L2, German L3) activated the different languages in two translation tasks (from L1 to L3 and from L2 to L3). By thinking aloud (i.e., verbalizing IS through PS) while performing the tasks, the participants gave hints as to what languages were activated at the levels of conceptualization, formulation, or articulation. It was found that language processing of the tasks was greatly influenced by the language of ‘input’ (i.e., text to be translated). In general, both the input (L1 or L2, depending on the task) and ‘output’ (L3) languages were predominantly selected at all levels of activation. Specifically, however, different processing sequences called for different languages to be selected; for example, during the task itself, L2 and L3 were more frequent than L1; when a problem occurred, L1 and L3 were activated; all languages were activated at the level of articulation and evaluation. According to Gabryś-Barker, these complex results may be accounted for by individual factors, such as the participants’ learning history, level of competence in L2 and L3, and approach to the task. This study shows that the IS produced by multilinguals in performing non-immediate tasks, such as translation, involves very complex patterns of language activation, subject to great individual variation and difficult to generalize.

In another study, Gabryś-Barker asked 26 multilinguals (same languages as above) to write self-reflective narratives on their preference for a language of thought. Quantitative analysis of the data indicated overall preference for the L1, frequent L2 activation among highly proficient L2 users, and little or no use of the less dominant L3 by the participants. Qualitative analysis revealed particular language choice or activation was determined by various reasons. Linguistic reasons were ‘high level of language proficiency, perceived language economy, code-switching, and communication with NS [native speakers] and speakers of other languages’ whereas non-linguistic reasons were ‘exposure and immersion (media, music, film, studies and work), and the affective dimension (attitude to a language, confidence in one's ability, emotional states)’ (Gabryś-Barker Reference Gabryś-Barker, Pawlak and Aronin2014: 193). The following sample from Gabryś-Barker's data demonstrates the complex variables at work for a multilingual's choice (or rejection) of a language for thought, from effect of exposure to an L2 to feelings of dislike for an L3:

Also related to the issue of language selection for IS among L2 users, two recent studies investigated the types of language used for thinking and their effect on the production of collocations by EFL learners, L1 Mandarin Chinese (Wang & Shih Reference Wang and Shih2011) and Persian (Nejadansari & Alijanian Reference Nejadansari and Alijanian2012). In both studies learners were asked about the language they used for thinking as they performed an oral production task. In both studies most of the learners reported using a combination of the L2 and the L1, with significantly fewer learners using mostly L2 or mostly L1. (In Wang & Shih, although some participants were knowledgeable of other languages, none reported using any of these for thinking.) Results of these studies also demonstrated that inner use of the L2 combined with the L1 was more beneficial to the EFL learners, as it led to significantly greater accuracy in the production of collocations than use of the L2 or L1 alone. In both studies, it was speculated that, if the learners had relied mostly on their L1, this would have led to direct translation and greater L1 interference.

Two empirical studies have dealt with other aspects of L2 IS. One is Shigematsu's (Reference Shigematsu2010) dissertation research on the concept of inner voice and its development among L2 learners. Shigematsu defined inner voice as the conscious ‘use of the L2 for thinking as a way of helping one mediate the ways of the L2 and culture’ (Reference Shigematsu2010: 2). Self-reports from five L2 learners of English and Japanese revealed that an L2 inner voice tends to develop with increased proficiency and exposure to naturalistic L2 settings. The other study (Kato Reference Kato2009) highlights the role of IS in reading L2 texts. Data from 64 L1 Japanese English as a second language (ESL) participants showed that, when the ability to use IS (covert articulation) was suppressed, reading comprehension was affected negatively. The effect was greater among learners with lower reading proficiency levels, suggesting that less proficient readers rely more on covert phonological activation and articulation during reading than more advanced readers who have developed the ability to derive meaning directly from visual orthographic symbols (bypassing the phonological route). Kato's research suggests that helping learners develop less dependence on covert expanded articulation and greater reliance on direct orthographic processing may contribute to increased L2 reading proficiency.

Additional light into L2 covert speech processes comes from the intensive neuroimaging research carried out in the last decade. Although most brain-imaging studies involving non-native speakers (NNSs), L2/FL learners, bi- and multilinguals do not use the term ‘IS’, they do have much to say about the impact of learning additional languages on brain areas of activation and covert speech processes (e.g., Kovelman, Baker & Petitto Reference Kovelman, Baker and Petitto2008; Grogan et al. Reference Grogan, Parker Jones, Ali, Crinion, Orabona, Mechias, Ramsden, Green and Price2012; Klein et al. Reference Klein, Mok, Chen and Watkins2014; Stein et al. Reference Stein, Winkler, Kaiser and Dierks2014; Wattendorf et al. Reference Wattendorf, Festman, Westermann, Keil, Zappatore, Franceschini, Luedi, Radue, Münte, Rager and Nitsch2014). Because this area of literature is too vast to review here in depth, the two studies that are mentioned next offer glimpses of the implications that can be derived—and remain to be made—on the nature of L2 IS from neuroimaging research. In a study of lexical efficiency and brain structure correlates, for example, Grogan et al. (Reference Grogan, Parker Jones, Ali, Crinion, Orabona, Mechias, Ramsden, Green and Price2012) found that native speakers of two or more languages (multilinguals) had higher gray matter density than NNSs of only one language (bilinguals), differences that may be attributed to increased covert use of lexical processing regions among multilinguals. Another study (Kovelman et al. Reference Kovelman, Baker and Petitto2008), showed that, although both bilingual and monolingual participants displayed activity in classic language areas (such as the left inferior frontal cortex) during a silent syntactic processing task, the bilingual group had a higher activation level while processing the L2. The study suggests the possibility of a different neural signature (pattern of neural activity) among bilinguals. Much information pertinent to L2 IS can result from neuroimaging research, yet the landscape that is arising is extremely complex and numerous variables—number and type of languages known, age and context of acquisition, proficiency level, nature of task involved, to name a few—will have to be taken into account to fully understand the impact of learning and using an L2 on brain structure and activity.

6.1 Collecting private speech data

Externalized PS has the advantage that it can be visibly and audibly observed and recorded although it is not always easy to capture intelligibly or distinguish from social-communicative speech. Most of the studies based on PS data reviewed here use a combination of audio and video recording and sometimes, additionally, observation notes, especially when it is impractical to utilize cameras or microphones, or simply to complement and corroborate video and audio recordings (e.g., Yi Reference Yi2010). Still, a caveat in most PS studies is that there is the possibility for additional L2 PS to have occurred but not been observed or recorded.

Quality of recording and subsequent transcribing as well as description of context is crucial in establishing reliability of PS data for analysis. In particular, identifying and categorizing self-directed speech are extremely dependent on reliable data collection. Most of the observation and category systems utilized in the studies reviewed entail differentiating among a nuanced variety of self-addressed vocalizations. Usually, PS categories range in degrees of covertness between audible and visible to inaudible or in a very low voice. Establishing degrees of covertness through reliable categories is especially crucial when stages of internalization are considered (e.g., Lidstone et al. Reference Lidstone, Meins and Fernyhough2011). As recent studies also show (Smith Reference Smith2007; San Martín et al. Reference San Martín, Boada and Feigenbaum2011; Steinbach-Kohler & Thorne Reference Steinbach-Kohler, Thorne, Hall, Hellermann, Doehler and Olsher2011), reliable recordings and context description are essential in determining the social aspects and impact of PS verbalizations embedded in interactive discourse or in the presence of others. Identification and categorization of PS are frequently corroborated by the stimulus response (SR) technique, whereby participants view/hear recordings of their own previous verbalizations and make retrospective comments on these. SR interviews, as in Lee (Reference Lee2006), Borer (Reference Borer2007), Scott & de la Fuente (Reference Scott and de La Fuente2008), and Yoshida (Reference Yoshida2009), contribute to better data interpretation. In Yoshida, for example, the learners’ comments produced in the SR sessions helped the researcher more accurately identify whether an utterance was an instance of PS or addressed to someone else (2009: 140).

6.2 Collecting IS data

Because of its inherently covert nature, silent IS, unlike audible and/or visible PS, has traditionally relied on indirect methods of data collection. Vygotsky's genetic method, i.e., the study of IS through its overt developmental precursor, PS, is an indirect approach that continues to be used in studies anchored in the internalization hypothesis. It is essential in these studies to be able to document not only the gradual evolution in time towards covertness through levels (as mentioned above) but also to identify IS activity through reliable ‘observable’ indicators. In some longitudinal and cross-sectional studies in L1 (e.g., San Martín et al. Reference San Martín, Boada and Feigenbaum2011; Alarcón-Rubio et al. Reference Alarcón-Rubio, Sánchez-Medina and Prieto-García2014) and L2 (Yi Reference Yi2010), IS is presumed to take place when silence takes over and no audible and visible signs are produced during cognitive activity. In addition, San Martín et al. have listed some ‘rigorous’ criteria to help infer IS from silence: (a) silence or pause persists for two seconds or longer, (b) silence is accompanied by inaudible mutterings or lip movements, (c) facial expression indicates concentration in task, and (d) there are no communicative gestures aimed at interlocutor (Reference San Martín, Boada and Feigenbaum2011: 222). Although these criteria seem reasonable, further research would be desirable to test their usefulness and purported rigorousness.

For the most part, recent studies on IS (L1 and L2) rely on self-report methodologies—questionnaires (or question prompts), interviews, autobiographical narratives, thought sampling, and think-aloud—to collect data. Some of the questionnaires used in late IS research are instruments that had previously been created to measure various aspects of inner life and mental states (DaSilveira & Gomes Reference DaSilveira and Gomes2012; Zimmerman & Brugger Reference Zimmerman and Brugger2013). A questionnaire which was especially designed to assess the experience of IS from a generic-L1 perspective is the Varieties of Inner Speech Questionnaire (VISQ) (McCarthy-Jones & Fernyhough Reference McCarthy-Jones and Fernyhough2011). In addition, several question-based instruments have specifically explored IS among L2 learners, bilinguals, and multilinguals. Worth mentioning are the BEQ (Dewaele-Pavlenko Reference Dewaele and Pavlenko2001–Reference Dewaele and Pavlenko2003), which included two questions on the silent use of language and contributed to gather massive data on multilinguals’ IS, and Ewert's (Reference Ewert, Arabski and Wojtaszek2010) adaptation of the BEQ to be used with multilinguals on a smaller scale. Also within the questionnaire modality are Schrauf's (Reference Schrauf2009) Language Background Questionnaire probing bilinguals on language usage in various domains, including internal speech, and Wang & Shih's (Reference Wang and Shih2011) questionnaire (adapted from Guerrero Reference Guerrero1994) exploring EFL learners’ choice of language for thought during an oral production task. A few other studies have used question-like prompts to find out L2 learners’ use of language for thinking (Nejadansari & Alijanian Reference Nejadansari and Alijanian2012) and an L2 inner voice (Shigematsu Reference Shigematsu2010). Questionnaire-type instruments have proved useful in providing insights on certain aspects of IS use, such as language choice, contents, and functions, and have generally yielded high statistical reliability; however, they have been criticized for lack of convergent validity, i.e., high correlations among them (Uttl, Morin & Hamper Reference Uttl, Morin and Hamper2011); lack of phenomenological fidelity, i.e., faithfully representing the phenomenon of inner speaking as a subjective experience (Hurlburt, Heavey & Kelsey Reference Hurlburt, Heavey and Kelsey2013); and respondents’ inability to retrospect accurately on formal features of IS, such as condensed or expanded structure (McCarthy-Jones & Fernyhough Reference McCarthy-Jones and Fernyhough2011).

Interviewing participants is another methodology that has been lately employed to collect data on IS. Usually, interviews are used in the SR modality, that is, as stimulus for participants to comment on, corroborate, or expand the information gathered through other instruments. DaSilveira & Gomes (Reference DaSilveira and Gomes2012), for example, interviewed subjects after they had answered an IS questionnaire and performed a think-aloud task in order to find out about their experience in responding to these instruments and in having an inner dialog. Shigematsu (Reference Shigematsu2010), in a study of L2 inner voice development, conducted in-depth interviews so that subjects could elaborate on responses previously obtained through online prompts.

A very productive means of gathering data in IS research is the examination of autobiographical narratives in search for first-person accounts of IS experiences. Pavlenko (Reference Pavlenko2011a, Reference Pavlenko2014) has applied this methodology to the analysis of memoirs of bi- and multilingual users, predominantly famous writers. Recognizing the limited use of autobiographical narratives in the study of linguistic development and self-reconstruction in the past, Pavlenko recommends expanding future research to include self-reports from various types of ‘bilingual speakers, including simultaneous bilinguals, multilinguals, and language attriters’ (Reference Pavlenko2014: 251). Gabryś-Barker's (Reference Gabryś-Barker, Pawlak and Aronin2014) study, based on the analysis of texts written in response to the prompt ‘Language(s) of our thoughts’, applied the self-narrative technique to multilingual L2 learners/users. No recent study, however, has deliberately exploited the diary form of the self-report narrative technique (such as in Guerrero Reference Guerrero2004) to collect longitudinal data on L2 learners’ personal experiences in developing L2 IS.

The thought sampling technique, which literally aims at obtaining samples of people's thoughts in response to a prompt, has been the method of preference for some researchers in recent studies on IS. In Morin et al. (Reference Morin, Uttl and Hamper2011), the technique was used as ‘thought-listing’, i.e., asking a large group of college students to list, retrospectively, what they usually said to themselves when engaged in IS. As Morin & Uttl (Reference Morin and Uttl2013) point out, this method can cause recall errors: ‘There is the possibility that participants may forget some actual IS occurrences or may “recall” instances that never occurred’ (Reference Morin and Uttl2013: 6–7). Rather than using this retrospective and reconstructive procedure, the researchers employed cell phones to send a prompt to participants asking them to respond immediately by text message on the contents of their current IS (Morin & Uttl Reference Morin and Uttl2013). Thought sampling by text-messaging is a modernized version of descriptive experience sampling, a procedure in which participants carry a beeper and, at the sound of a random beep, jot down notes on whatever inner experience they are having (Hurlburt et al. Reference Hurlburt, Heavey and Kelsey2013).

Introspective techniques, unlike the retrospective ones mentioned so far, elicit verbalization of thought processes as they are taking place (or immediately after) and are thus less likely to suffer from memory limitations. Thinking aloud is an introspective procedure that has also been used lately in gathering IS data (Gabryś-Barker Reference Gabryś-Barker2006; DaSilveira & Gomes Reference DaSilveira and Gomes2012; Gabryś-Barker Reference Gabryś-Barker, Piasecka, Adams-Tukiendorf and Wilk2015). For the technique to be successful, training is usually necessary so that subjects learn to say out loud whatever is on their minds while performing an action. Think-alouds, however, should not be taken as exact, direct verbalizations of silent thinking. Think-alouds are close records of IS, but they are ‘externalizations’ of IS, i.e., internal speech that has most likely suffered structural changes in its transition to overt speech (Guerrero Reference Guerrero2005: 102–103).

A review of the self-report methodologies adopted in late studies on IS reveals strengths and weaknesses, some of which had already been pointed out in the past (Guerrero Reference Guerrero2005: 104–108). To summarize these in brief, self-reports are valuable sources of insights and information on people's perceptions of their own IS, particularly as it pertains to its contents and functions (though not so much on its structural features), and useful in making generalizations when large samples are used. Self-reports, however, are limited in providing direct and precise records of actual IS activity, a phenomenon which is non-observable to outsiders and often inaccessible to self-control and attention. The more retrospective a personal account of IS, i.e., the further removed in time from the actual verbalizing event, the more likely it may not be remembered well or may be reconstructed and misreported. Even immediate introspective verbalizations of IS are prone, as was pointed out above, to modifications of the actual verbal thinking phenomenon. It is very important for researchers to be aware of these shortcomings and to not only follow guidelines that may safeguard against distorted or unwarranted conclusions but also clearly report and explain step-by-step procedures so that readers may assess on their own the relative merits and usefulness of the research. Triangulation—i.e., using multiple approaches in gathering data—of self-report methodologies is recommended in the study of IS, as for example in DaSilveira & Gomes (Reference DaSilveira and Gomes2012). McCarthy-Jones & Fernyhough (Reference McCarthy-Jones and Fernyhough2011) point out that one of the limitations of the instrument they developed, VISQ, is that participants may be unable to accurately retrospect on the subjective qualities of IS; thus, the authors recommend supplementing results of the VISQ with more ecologically valid methodologies, such as descriptive experience sampling (see study employing this joint methodology: Alderson-Day & Fernyhough Reference Alderson-Day and Fernyhough2015b).

Aside from self-reporting, a few other methodological strategies have been used lately in PS/IS research. One of these is the measurement of covert verbal responses through speech interference (frequently including comparisons between silent, subvocal, and overt speech). Techniques such as having the tongue clamped between the teeth, concurrent finger or foot tapping, and concurrent word repetition, which might interfere with normal verbal production while performing a silent task, have contributed to throw light on the role and extent of phonological articulation in covert speech production (Durisko Reference Durisko2006; Aguilar & Aguilar Reference Aguilar and Aguilar2008; Kato Reference Kato2009). Other psycholinguistic methodologies include eye-tracking assessment to obtain information on the pronunciation features of IS during reading (Filik & Barber Reference Filik and Barber2011) and experimenting with speech imagery to collect data on the brain mechanism responsible for an inner voice (Scott Reference Scott2013).

Lastly, brain scanning techniques, most prominently neuroimaging (as in fMRI or PET), have been intensively exploited in the past decade in the study of brain areas and processes associated with IS activity. (See Girbau Reference Girbau2007 for a detailed overview of the evolution of these techniques in IS research.) To their advantage, brain scanning and imaging techniques provide, to date, the most ‘direct’ method of observation within IS research. The images produced through these technologies make it is possible to ‘see’ brain structures and pathways of activation implicated in verbal thinking. In the case of research involving speakers of more than one language, in particular, brain scanning techniques offer invaluable information about anatomical and procedural differences related to language processing. Data-gathering through these technologies, however, is not without weaknesses. One of these is lack of ecological validity—neuroimaging experiments can only be conducted in highly controlled laboratory conditions. As Morin points out, brain-imaging studies are restricted to dealing with artificially elicited IS, mostly of the mental recitation type: ‘Naturally occurring inner speech, with its condensed and dialogic qualities, has never been imaged’ (Reference Morin and Ramachandran2012: 440). Another drawback is the danger of reduced internal validity—it is difficult to establish with complete certainty that the person whose brain is being scanned is mentally engaged in the IS task as instructed. (For further discussion of neuroimaging strengths and shortcomings in the study of IS, see Alderson-Day & Fernyhough Reference Alderson-Day and Fernyhough2015a.)

Finally, one crucial limitation affecting many PS/IS studies is failure in collecting detailed information on the language background of the participants. Except for studies specifically involving L2, bi-, or multilinguals (which generally do include information about multiple languages learned or spoken), most studies do not go beyond reporting the native language category of the participants. As a rule, the language(s) known by the inner speaker, the language(s) used in the IS under observation, or the language(s) spoken or scripted in the external context are not issues to be taken into consideration or variables to be included in the research design. It is an important gap because assuming (often incorrectly!) monolingualism or disregarding people's knowledge of more than one language leads to failure in investigating how multiple-language knowledge may affect people's thinking processes. This is a problem particularly blatant in self-reports, where participants are frequently not asked what language they use for covert self-verbalizations. In Morin et al. (Reference Morin, Uttl and Hamper2011), 380 college students were asked to list, retrospectively, what they usually said to themselves when engaged in IS. The authors do mention that 89% of the participants were English-L1 speakers, but not whether they knew other languages. Nor was it probed what language the English-L1 and English-L2 (11%) participants used in their IS verbalizations, and therefore it remains unknown whether another language was used by participants in their thinking or to what extent L2 IS was used or for what purposes. (See also Dewaele Reference Dewaele2015 on this issue.)

7.1 Providing learners with opportunities for social engagement with the L2

As Guerrero put it, ‘inner speech is derived from external speech and cannot happen in the absence of exposure to and participation in the particular discursive practices of a social group’ (Reference Guerrero2005: 191). The importance of creating socially stimulating conditions in classrooms and providing learners with plenty of opportunities for social interaction in the L2—for external-to-internal movement to occur—is confirmed by studies indicating socially related variables (frequency of L2 use in daily interactions, size of social networks, naturalistic contexts of acquisition) as the strongest predictors of L2 (or LX) IS use among bi- and multilinguals (Shigematsu Reference Shigematsu2010; Pavlenko Reference Pavlenko2014; Dewaele Reference Dewaele2015). Socially interactive settings (classrooms, playground) and activities approximating conditions of naturalistic L2 use, including learner–learner and teacher–learner interactions, may also foster self-directed use of the L2 (Smith Reference Smith2007; Steinbach-Kohler & Thorne Reference Steinbach-Kohler, Thorne, Hall, Hellermann, Doehler and Olsher2011). Use of the L2 in social interactions may prove particularly essential in the development of the dialogic (i.e., conversation-like) form of IS as suggested by studies on the effects of social speech impairment in L1 (Williams et al. Reference Williams, Bowler and Jarrold2012; Alderson-Day & Fernyhough Reference Alderson-Day and Fernyhough2015a) and on the evolution of an internal L2 voice (Shigematsu Reference Shigematsu2010), just as the social presence of interacting or non-interacting others may have a ‘stimulating’ effect on PS occurrence (San Martín Reference San Martín, Boada and Feigenbaum2011; Damianova et al. Reference Damianova, Lucas and Sullivan2012; McGonigle-Chalmers et al. Reference McGonigle-Chalmers, Slater and Smith2014).

7.2 Fostering internalization and externalization of the L2

Although the capacity to think through a new language owes much to factors associated with naturalistic conditions of acquisition, pedagogical intervention can also greatly contribute to foster internalization and externalization of the L2. Actually, in Dewaele (Reference Dewaele2015), it was ‘mixed context’ of acquisition (i.e., a combination of naturalistic learning and formal instruction) which accounted for highest L2 use in IS. Several instructional methods and learner strategies seem to encourage internalization and externalization of the L2. First and foremost is learners’ use of PS in academic settings. The literature on PS among L2 learners/users shows quite convincingly that it is widespread as a learning mechanism or as a self-regulating tool in solving problems in the L2. Researchers recommend that teachers pay close attention to learners’ production of PS, acknowledge it, and certainly not inhibit it (Lee Reference Lee2006; Smith Reference Smith2007; Yoshida Reference Yoshida2009; Day & Smith Reference Day and Smith2013; Clark Reference Clark2005). Yoshida derived the following implication from her PS study among college learners of Japanese:

What can teachers do to deliberately stimulate internalization and development of IS in the L2? Recent studies suggest certain instructional methods and strategies may be particularly favorable. One of these is the integration of subvocal repetition or rehearsal in pedagogical activities. Borer (Reference Borer2007), for example, implemented learners’ use of self-vocalization in the learning of L2 vocabulary, finding positive effects of deep-processing rehearsal on word retention, both in solitary or collaborative conditions. Specifically, Borer recommends exposing learners to new words alone (rather than in peer groups) so they can have better control over their private rehearsal process. In Borer's study, ‘by selectively rehearsing linguistic features such as phonology, spelling, or meaning for future use, learners began the transformation from external to inner speech’ (Reference Borer2007: 287). Fallahchai (Reference Fallahchai2011) similarly found that training learners in phonological awareness and performing vocal and subvocal rehearsal contributed to successful learning of FL words. Positive effects on vocabulary learning may be attributed, according to Fallahchai, to the learners’ efforts at vocal and subvocal articulation of new words and their increased ability to hear and modify their own vocal production through PS and IS.

Shadowing and summarizing are two other classroom techniques that rely on repetition and rehearsal and have great potential for enhancing IS processes, such as inward auditory reproduction, subvocal phonological articulation, retention in memory, semantic condensation, and verbal restructuring. Guerrero & Commander (Reference Guerrero and Commander2013) integrated shadowing and summarizing into a technique labeled ‘shadow-reading’, which required learners working in pairs to read aloud and shadow parts of a story to each other, then orally summarize the story, and finally produce a written retelling. The shadowing, which consisted of repeating segments of a text in three modes—out loud, low voice, and silent (subvocal)—provided repeated opportunities for internalization during the stages of self-directed low-voice and silent repetition. Summarizing and retelling, in turn, provided affordances for meaningful imitation, internalization, transformation, and externalization of L2 models.

Dialog journal writing is another technique recommended by Mahn (Reference Mahn2008) to foster the IS process among L2 learners. As Mahn proposes, learners write freely in their journals without worrying about making mistakes; rather, the emphasis is on meaning-making when writing in the L2. Dialog journals encourage learners to draw from their ‘systems of meaning’ (Vygotsky's notion of the internalized conceptual system on which IS relies): ‘When students focus on meaning and are relieved of the pressure of producing error-free writing, they increasingly rely on inner speech and verbal thinking’ (Reference Mahn2008: 122). The technique, the author claims, improves learners’ fluency in writing, stimulates efforts to develop mastery over the structure of language, and encourages use of IS to express thoughts in meaningful ways.

7.3 Developing a conceptual foundation in the L2

The capacity to engage in L2 IS entails changing the L1 conceptual foundation to accommodate new concepts expressed linguistically in the L2. Negueruela (Reference Negueruela2008; Negueruela-Azarola, García & Buescher Reference Negueruela-Azarola, García, Buescher and Markee2015), who conceives L2 development as a conceptual process, argues that instruction should aim at promoting the internalization of new conceptual categories through properly organized, meaningful, and awareness-raising pedagogical tasks. Instructional applications of this concept-based approach have mostly focused on the teaching of new categories in areas related to grammar (e.g., verbal aspect, modality, voice), writing genres, and academic discourse (see studies in Lantolf & Poehner Reference Lantolf and Poehner2008). Negueruela & Lantolf (Reference Negueruela, Lantolf, Salaberry and Lafford2006), for example, found that Spanish learners showed greater coherence and accuracy in using preterit and imperfect forms after engaging in conscious reflection and private verbalization on the concept of verbal aspect. Similarly, Escandón & Sanz (Reference Escandón and Sanz2011) obtained favorable results when utilizing a ‘bottom-up move’ (based on Galperin's external-to-internal methodological approach to concept formation) to teach learners the notions of Spanish gender and number agreement. In general, however, this review reveals that the potential for the application of a deliberate and systematic conceptually-based approach to L2 instruction remains largely untapped.

7.4 Creating awareness about inner and private speech

The final implication derived from the studies reviewed is the need to raise awareness about IS and PS among educators and learners. Several researchers advocate for pedagogical approaches that include the explicit and deliberate stimulation of PS and IS development in the L2 (Clark Reference Clark2005; Smith Reference Smith2007; Yoshida Reference Yoshida2009; Little Reference Little2010; Wang & Shih Reference Wang and Shih2011). Little, for example, argues that ‘developing the learner's capacity for L2 inner speech should be an explicit goal of L2 pedagogy’ (Reference Little2010: 28). It is necessary, however, to first become aware of the occurrence and importance of PS and IS in L2 education.

Research indicates that PS often goes unnoticed (Clark Reference Clark2005) or is discouraged (Manfra & Winsler Reference Manfra and Winsler2006) in classrooms. Paying attention to covert speech processes, however, can be highly revelatory to teachers. Clark believes there is little understanding of the significance of L2 PS in early childhood education and claims for a pedagogy that recognizes that ‘private speech is meaningful, . . . can be revealing about the child's use of language, . . . can provide insight into their thinking and their needs, and . . . can provide clues about the effects of the environment on their learning’ (Reference Clark2005: 217). Smith observed: ‘The private speech produced during a learning activity may reveal what pupils have already learned, and how they are using language as a tool in the process of learning’ (Reference Smith2007: 254). Day & Smith (Reference Day and Smith2013) also point out that PS has an important emotion-regulating function and that, by paying attention to children's PS, teachers may help learners who express negative emotions cope better when performing cognitively frustrating or difficult tasks. Importantly, research (Clark Reference Clark2005; Pavlenko Reference Pavlenko2014) suggests that in the case of translingual learners who become immersed in a new L2 community, by paying attention to their PS, teachers may help learners cope with the problems they undergo in the process of attriting from an L1 inner voice and acquiring a new one mediated by the L2.

Research also proposes that learners be made aware of their own PS and IS processes in learning the L2 and of the role and benefits of PS and IS in conducting cognitive-verbal tasks in the L2 (reading, writing, mathematical operations, problem-solving, retaining in memory, planning, etc.). Although many covert speech processes take place at an unconscious level, learners can take an active role in those phases of IS that are within their control: rehearsal and repetition, switching from one language to another, deliberately engaging in L2 self-talk, and the like. Even though awareness of intrapersonal speech is considered to be a rather late higher-level cognitive development among children, Manfra & Winsler found that even very young children (preschoolers) understand that self-talk can be a useful tool in learning (Reference Manfra and Winsler2006: 547). Wang & Shih (Reference Wang and Shih2011) believe learners can be offered metacognitive training into using L1 and L2 for language support in IS. The implication from research is thus that learners’ awareness and deliberate covert use of the L2 through PS and IS may have positive effects on internalization and externalization of the L2.