2. Theoretical foundations

Automatization has been a recurring concept over several decades in language teaching and learning. The emergence of automatization as a research topic in SLA can be traced back to the 1980s. Following the rise of the Monitor Model (e.g., Krashen, Reference Krashen1985), the information processing approach was introduced as a complementary or alternative cognitive theory in SLA (e.g., McLaughlin et al., Reference McLaughlin, Rossman and McLeod1983). Drawing on findings from cognitive psychology, L2 learning was seen as the acquisition of complex skills that draw on automatic and controlled processes (Shiffrin & Schneider, Reference Shiffrin and Schneider1977). As successful L2 use hinges on the coordination of linguistic skills in an integrated manner, automatization compensates for a learner’s limited attentional capacity by freeing up cognitive resources; for instance, by automatizing lower-level skills such as lexical retrieval for speaking (e.g., Levelt, Reference Levelt1978) and word recognition for reading skills (e.g., McLeod & McLaughlin, Reference McLeod and McLaughlin1986).

Information processing theory laid the groundwork for exploring the learner’s transition from controlled processing, a slow and effortful stage, to automatic processing, characterized by its speed, effortlessness, and minimal requirement for conscious attention. Building on these cognitive principles, researchers examined to what extent learning theories (e.g., Anderson et al., Reference Anderson, Bothell, Byrne, Douglass, Lebiere and Qin2004; Logan, Reference Logan1988), originally developed for general cognitive skills, could be applied to SLA. Since the 1990s, empirical studies have tested key theoretical assumptions of skill acquisition theory (e.g., DeKeyser, Reference DeKeyser1997*; Robinson, Reference Robinson1997*), revealing that L2 skill learning can be accounted for by learning principles governing the acquisition of other general skills such as music, sports, algebra, chess, computer programming, and flying an aircraft (see Tenison & Anderson, Reference Tenison and Anderson2016, for a neuroimaging study).

Underlying mechanisms of automatization have been debated over decades in the fields of cognitive psychology and SLA research. A seminal study by Norman Segalowitz and his colleagues proposed a fine distinction between simple speed-up and genuine automatization in L2 processing (Segalowitz & Segalowitz, Reference Segalowitz and Segalowitz1993*Footnote ¹). Consider how L2 learners process grammar when forming sentences: they might first translate from L1, then reorder words to match L2 structure, and finally apply grammar rules. A mere speed-up means executing these same steps faster (e.g., reducing time from 900 to 500 milliseconds), analogous to a student solving multiplication problems by performing addition steps more rapidly (e.g., calculating 5 × 4 as 4 + 4 + 4 + 4 + 4) and a beginning typist locating keys more quickly while still looking at the keyboard. In contrast, genuine automatization represents a fundamental restructuring of these processes. Just as students progress from counting-based calculation to instant recognition of multiplication facts (5 × 4 = 20) and typists develop fluid movements without looking at keys, L2 learners develop the ability to directly access target language structures without relying on L1 translation.

While extended practice triggers both speed-up and restructuring, Norman Segalowitz argued that only restructuring leads to true automaticity (Segalowitz & Segalowitz, Reference Segalowitz and Segalowitz1993). To measure this distinction, he proposed using the coefficient of variation (CV) – calculated by dividing the standard deviation of response times (RT) of an individual by their own mean RT. Coefficient of variation provides insights into processing variability, with lower CV values indicating more stable and consistent processing thought to characterize automatic performance. Higher CV values suggest more variable processing typically associated with controlled, non-automatic processing.

Since then, the utility and validity of CV as a measure of automatization, alongside with RT, has been subjected to intensive empirical investigations. Several limitations have been raised: CV measurements assume accurate knowledge representation and may not capture development when new content is being learned simultaneously (Hulstijn et al., Reference Hulstijn, Van Gelderen and Schoonen2009*). Hence, CV may be only useful in investigating automatization in controlled laboratory settings where participants without prior knowledge are trained to learn a finite set of materials (e.g., words), and it may not be useful in real learning situations where learners simultaneously develop both (new, declarative) knowledge (e.g., accuracy) and efficient skills to retrieve existing knowledge.

With the methodological caveats of using CV in mind, automaticity can be captured more broadly, at least for practical purposes (e.g., using it as a predictor of proficiency), through a combination of several indices, such as speed, stability, ballistic (unstoppable) processing, resistance to interference, levels of consciousness, and so on (DeKeyser, Reference DeKeyser and Robinson2001; Segalowitz, Reference Segalowitz, Doughty and Long2003). Some research revealed that RT (processing speed) predicted L2 proficiency more strongly than CV (processing stability) (e.g., Suzuki & Sunada, Reference Suzuki and Sunada2018*), whereas other research found the opposite pattern (e.g., Zhang & Yang, Reference Zhang and Yang2023). This discrepancy may reflect different stages of L2 development: it appears that RT is a stronger predictor in earlier stages when learners are still developing rudimentary processing speed, while CV becomes more influential in advanced stages when processing stability becomes more crucial after declarative knowledge is established.

Building on these theoretical and methodological developments, researchers have expanded their toolkit for measuring automaticity. Recent methodological syntheses have revealed numerous experimental tasks, instruments, and objective measures to capture different aspects of fluent comprehension and production (sub)skills and knowledge (Suzuki & Elgort, Reference Suzuki, Elgort and Suzuki2023; S. Suzuki & Révész, Reference Suzuki, Révész and Suzuki2023). By employing various psycholinguistic tasks yielding different measures (e.g., accuracy, RT, CV, neural responses), researchers have delved deeper into the cognitive mechanisms underlying language acquisition and usage, offering a more nuanced understanding of how automaticity develops in the context of SLA. This approach has not only provided valuable insights into learners’ progression towards fluency but has also helped in understanding the outcomes of L2 practice aimed at facilitating automatization.

3. Automatization and knowledge and skills

In tandem with the growing interest in information processing theory during the 1980s and 1990s, there was burgeoning exploration into the realm of explicit and implicit learning. Explicit learning refers to conscious learning processes, whereas implicit learning concerns incidental learning without awareness (DeKeyser, Reference DeKeyser, Doughty and Long2003; Ellis, Reference Ellis1994; Rebuschat, Reference Rebuschat2015). This demarcation of learning processes can be considered as an extension of the distinction Krashen (1985) made between acquisition and learning.Footnote ²

One approach to understanding the roles of explicit and implicit learning in SLA is skill acquisition theory (e.g., DeKeyser, Reference DeKeyser, VanPatten, Keating and Wulff2020; Suzuki, Reference Suzuki2023). Since “skill” is central to both this theory and the automatization research reviewed thus far, it is essential to clarify the distinction between skill and knowledge in L2 learning. While knowledge refers to the mental representations of language forms and rules, a skill represents the learner’s ability to draw upon linguistic knowledge (stored in memory) to perform various linguistic tasks such as listening, reading, speaking, and writing (DeKeyser, Reference DeKeyser, Loewen and Sato2017). In the technical terminology of skill acquisition theory, these components are formalized as declarative and procedural knowledge. Declarative knowledge reflects one’s conceptual understanding of facts and rules, while procedural knowledge represents the ability to use these rules in performance. Within the framework of skill acquisition theory, procedural knowledge corresponds directly to the definition of skill described above – the ability to perform linguistic tasks by drawing upon knowledge stored in memory. For instance, L2 learners typically begin with declarative knowledge (such as understanding grammar rules) and, through practice, develop procedural knowledge (the ability to use these rules in speech). This progression from declarative to procedural knowledge forms the foundation of automatization, and researchers have extensively studied how initial declarative knowledge influences both proceduralization and eventual automatization of language skills (e.g., McManus & Marsden, Reference McManus and Marsden2019*; Sato & McDonough, Reference Sato and McDonough2019*).

The distinction between declarative and procedural knowledge in skill acquisition theory often aligns with the explicit-implicit learning dichotomy in SLA research. Based on the historical evolution of SLA research commencing from the Monitor Model, the learning stages of declarative-procedural-automatization are typically associated with explicit, rather than implicit, learning and knowledge (e.g., DeKeyser, Reference DeKeyser, VanPatten, Keating and Wulff2020). Therefore, the end product of automatization is often called automatized explicit knowledge – linguistic knowledge that can be deployed quickly albeit with some level of conscious access and mental effort (Suzuki, Reference Suzuki2017). While both automatized explicit knowledge and implicit knowledge enable rapid processing, they differ fundamentally in terms of awareness: automatized explicit knowledge, which typically develops through explicit learning of grammar rules followed by extensive practice (e.g., consciously accessing subject-verb agreement rules during speaking until the rules can be applied effortlessly), remains conscious and accessible even when highly practiced, whereas implicit knowledge, which typically develops through extensive exposure to input (e.g., developing intuition about article usage through extensive reading), is characterized as capacity to use linguistic rules without awareness. However, these two types of knowledge are functionally equivalent in everyday language use (DeKeyser, Reference DeKeyser, Doughty and Long2003; Suzuki, Reference Suzuki2017), as both support fluent communication. Indeed, recent neuroimaging research indicates that L2 speakers dynamically recruit both automatized explicit and implicit knowledge in complementary ways for accurate and fluent speech (Suzuki et al., Reference Suzuki, Jeong, Cui, Okamoto, Kawashima and Sugiura2023).

Furthermore, from a neurocognitive perspective, recent research has also demonstrated that individual differences in long-term memory – specifically declarative and procedural memory – are implicated in automatization (Buffington & Morgan-Short, Reference Buffington, Morgan-Short, Wen, Skehan, Biedroń, Li and Sparks2019; Skehan, Reference Skehan, Wen, Skehan, Biedroń, Li and Sparks2019). Learners with greater capacities in these memory systems typically advance more rapidly from effortful to fluent language production. A key question is how specific types of long-term memory differentially influence various stages of L2 learning across linguistic domains (e.g., syntax, lexis, phonology) and among diverse learner populations.

Fluency is another construct often associated with automatization (Tavakoli, Reference Tavakoli, Wen and Ahmadian2019). As a multifaceted concept extensively studied in SLA research, fluency encompasses multiple dimensions. Segalowitz (Reference Segalowitz2010, Reference Segalowitz2016) proposes three dimensions of fluency: perceived, utterance, and cognitive. Perceived fluency relates to subjective assessments of speech, while utterance fluency involves measurable speech characteristics such as speed, breakdown (pausing), and self-repair. Cognitive fluency, a cornerstone of fluent language use, entails the efficient integration of cognitive processes for producing fluent speech and is closely linked to automaticity in language processing. This cognitive dimension underpins utterance fluency and supports the efficient production of fluent utterances.

4. Practical applications and developments

The study of automatization has not only advanced our theoretical understanding of L2 acquisition but also influenced language teaching methodologies. This influence is particularly evident in foreign language contexts, where learners have limited exposure to the target language outside the classroom. In such settings, educators must carefully adapt their pedagogical strategies to support the protracted and gradual process of automatization.

The evolution of automatization research in SLA has paralleled significant shifts in language teaching approaches. The field of language teaching witnessed the emergence of communicative language teaching (CLT) as a response to earlier methods such as audiolingualism. During the same period, information processing theory gained prominence in L2 learning research during the 1980s and 1990s. While CLT successfully prioritized authentic language use in the classroom, it initially underestimated the importance of systematic practice in developing automaticity. Recognizing this limitation, Gatbonton and Segalowitz (Gatbonton & Segalowitz, Reference Gatbonton and Segalowitz1988, Reference Gatbonton and Segalowitz2005) proposed a framework that integrated automaticity development within the CLT approach. Their methodology emphasizes the mastery of naturally occurring utterances in communicative situations (e.g., idiomatic expressions, functional language for requesting, questioning, and describing) rather than focusing on abstract structures. This approach provides learners with abundant opportunities for meaningful repetition and practice, thereby fostering automaticity without resorting to the mechanical nature of traditional drills.

The turn of the century marked a shift towards a more nuanced understanding of deliberate and systematic practice and its role in automatization in L2 learning (DeKeyser, Reference DeKeyser2007; Jones, Reference Jones2018; Lyster & Sato, Reference Lyster, Sato and Mayo2013; Suzuki, Reference Suzuki2023). This period coincided with a significant expansion in L2 practice research, including influential studies informed by cognitive psychology, as documented in a recent synthesis by Maie and Godfroid (Reference Maie, Godfroid and Suzuki2023) which revealed the exponential growth of such studies in the twenty-first century. DeKeyser and his colleagues emphasized that mere exposure to language is insufficient for automatization; rather, the concept of deliberate and systematic practice, involving focused, goal-directed activities, is integral to the automatization of language skills (Suzuki et al., Reference Suzuki, Nakata and DeKeyser2019). Such goal-directed activities can be simple real-world tasks that incorporate repetition combined with increasing complexity. The judicious sequencing of these tasks leads learners to repeatedly use the same structures and phrases, providing high-quality practice without the disadvantages of drills. In this sense, the concept of automatization is not incompatible at all with more contemporary teaching methodologies such as task-based language teaching (TBLT).

Recently, task repetition research in fluency development within TBLT has incorporated the concept of automaticity (Bygate, Reference Bygate2018). This approach recognizes the tight link between fluency and automaticity development, emphasizing the importance of repeated engagement in communicative tasks for transitioning from controlled to automatic processing of lexico-grammatical structures (e.g., De Jong & Tillman, Reference De Jong, Tillman and Bygate2018). Understanding how language teaching approaches, such as TBLT, can promote automatization across various L2 skills is crucial for effective language teaching (DeKeyser, Reference DeKeyser and Bygate2018; Lambert, Reference Lambert and Suzuki2023).

In summary, this introduction has traced the evolution of automatization research in SLA over the past four decades. From its roots in information processing theory to the current prominence of skill acquisition theory, our understanding of automatization has grown increasingly sophisticated. This progression reflects advancements in cognitive psychology, shaping our understanding of how practice, repetition, and cognitive processing contribute to the automatization of L2 skills. Key developments include:

1. 1. The transition from the earlier Monitor Model to cognitive approaches in SLA.

2. 2. The exploration of explicit and implicit learning processes in relation to automaticity.

3. 3. The refinement of concepts such as fluency and its nature in relation to L2 skill development.

4. 4. The development of methodologies to measure and assess automaticity in L2 processing.

5. 5. The integration of automatization principles into language teaching methodologies such as CLT and TBLT.

This evolving understanding has significant implications for both research and practice in SLA. As we continue to refine our understanding of automatization in L2 learning, this research trajectory promises to yield further insights that will enhance our ability to foster successful language acquisition in diverse learning contexts.

The purpose of this timeline is to review major empirical research on automatization in L2 learning conducted over the last 40 years and to illustrate how the field has come to better understand its mechanisms and implications for language teaching. Due to space limitations, this timeline focuses on selected pivotal studies published in English as book chapters and as articles in leading academic journals. The criteria for selection included the citation frequency and contribution of novel insights that have significantly shaped subsequent research in the field. Additionally, we sometimes prioritize the relevance to language teaching and learning over the technicality often involved in psycholinguistic research for the interests of readers of Language Teaching. This curated approach ensures that the timeline reflects key developments and turning points in the understanding of automatization in L2 learning.

We highlight and categorize the main themes in the following ways throughout the timeline:

1. A. Research focus

   1. 1. Mechanism (testing theory of automatization)

   2. 2. Skill development (investigating how automatization relates to proficiency development)

   3. 3. Pedagogy (intervention for promoting automatization)

   4. 4. Method (methodological refinement of research on automatization)

   5. 5. Individual difference (investigating individual differences factors in automatization)

2. B. Contexts

   1. 1. Lab

   2. 2. Classroom

3. C. Study design

   1. 1. Observational (no control group)

   2. 2. Experimental

4. D. Linguistic domains

   1. 1. Phonology (PHO)

   2. 2. Lexis (LEX)

   3. 3. Morphosyntax (MSYN)

   4. 4. Pragmatics (PRAG)

5. E. Outcome measurements

   1. 1. Accuracy (ACC)

   2. 2. Fluency (FLU)

   3. 3. Response time (RT)

   4. 4. Coefficient of variance (CV)

   5. 5. Eye-tracking (EYE)

   6. 6. Event-related potential (ERP)

   7. 7. Functional magnetic resonance imaging (fMRI)