1.1 Some Background Terms

A key background term that is typically left in the background is language. It can be defined in many different ways, from many different perspectives. Perhaps the most important dividing line falls between cognitive and social/cultural perspectives. Language can be seen as something in an individual’s head or it can be seen as a feature of social relations and the cultures within which they are embedded. It is undeniably both and therefore can be and must be studied from both perspectives, with the ultimate goal of establishing a unified explanation.

The concept of input belongs primarily to the cognitive side of this divide.Footnote ¹ Cognitive theory owes a great deal to the computer metaphor, seeing the mind as an information processor. It receives information through the senses, processes and stores the information, retrieves that information when needed, and uses it to produce output in the form of actions, including speech. In this metaphor, input is the information taken in by the computer, or perhaps the process of taking it in. The nature of this “taking-in” process represents the fundamental issue for the study of input.

From a cognitive perspective, looking at what is inside our heads, language is the interaction among a number of distinct types of knowledge and ability. These types include, at least, phonological (linguistic sounds), morphological (the structure of words), syntactic (the ways that phrases and sentences are constructed), semantic (the meanings of linguistic units), pragmatic (meaning in social context), articulatory (the muscle movements that produce speech), and orthographic (written forms). This non-monolithic character of language will assume some importance in Section 4.

A second language shares this character, as it also includes sounds, structured sentences, meanings, and physical control of speech. The cognitive conception of a second language is well captured in the term interlanguage (Reference SelinkerSelinker, 1972). It refers to the knowledge of L2 learners, which is to say the state of their underlying linguistic system at any given point. The term has the advantage of being theory neutral, its main claim being simply that there is a system, not just scraps of information or modifications of the first language. To the extent that a given learner can use the language fluently and competently, the interlanguage is what makes this use possible.

The nature of the system has always been a source of dispute, and ideas have shifted over the years (see Reference Tarone, Han and TaroneTarone, 2014, for a useful summary). Contrary to Reference SelinkerSelinker’s (1972) original conception, interlanguage has generally come to be seen as a genuine language. Partly for this reason, universal grammar (UG) theorists (see 2.6) have embraced the concept, as it fits well with their overall conception of second language and second language learning and provides a useful way to frame major issues. The dynamic nature of the system has also come to the fore, though the details are again a theoretical issue. One important aspect is variability – not just over time, but also from one context or task to another. Research, from various perspectives, has also been dedicated to the discovery of more or less predictable ways in which interlanguage develops, on its own terms, not simply reflecting the character of the L1 or the target language. The L1 was originally given a prominent position. This role was subsequently challenged, but now it is once more recognized as a significant factor, though theoretical divergence occurs on exactly how it affects the interlanguage. Not withstanding the many disagreements, the idea of interlanguage as a developing system has become a key part of SLA and is likely to remain so.

If the interlanguage is a system, what about language-related knowledge that is not an integral part of that system? Such knowledge clearly does exist. An English learner who has been taught a given grammar rule might never manage to incorporate that rule in the interlanguage or use it in speech or writing but still be aware of it and be able to talk about it. A Mandarin learner who has little or no ability to produce the tones of the language or hear them in speech might still know that the language has four tones and be able to accurately describe them. Linguistics students may have mastered concepts like the Theta Criterion or feature interpretability, but it is unlikely that this knowledge will ever become part of their interlanguage for any second language they are learning.

This language-related knowledge has been referred to with a variety of terms, with somewhat varying meanings, notably including metalinguistic knowledge, knowledge about language (as opposed to knowledge of language), learned knowledge (as opposed to acquired knowledge), and Learned Linguistic Knowledge (as opposed to competence). It has also been associated with the notions of explicit knowledge and declarative knowledge, though considerable caution is required here. I will refer to it herein as “metalinguistic.”

1.2 Input

Many definitions have been offered for the term “input.” It is commonly understood in a traditional, intuitive way, partly reflecting the computer metaphor, and this makes a good starting point. But researchers have been aware for some time that considerable complexity lurks behind this neat metaphor. To understand input, we need to get beyond the surface, considering how it fits into the cognitive system as a whole.

1.3 Key Concepts Taken from Psychology

The importance of input for SLA lies in the contribution it makes to learning, and so learning can also be considered a key term in this area. Learning has always been a central concern of psychology, with ideas about it changing considerably over the years. A good contemporary account comes from Reference DehaeneDehaene (2020), who defines learning like this: “to learn is to form an internal model of the external world” (p. 5). He stresses that this model building is based on innate constraints – “Learning … always starts from a set of a priori hypotheses, which are projected onto the incoming data” (p. 26). Learning a language means forming a number of such models for the different components of language, based on innate constraints. The “incoming data” can be called input.

In SLA, the term “learning” is often used quite broadly, to mean simply positive changes in memory or ability. In this sense it is interchangeable with its companion term, acquisition, though a distinction is sometimes drawn between the two. Krashen in particular distinguished between a natural, unconscious process of acquisition and a more limited, conscious process of learning. When the process is viewed as natural and spontaneous development, the term growth is sometimes favored. The most neutral term is development.

L2 theory has conceptualized learning in a variety of ways. In UG approaches, for example, it means setting innate parameters, a process that is based on innate principles as well as existing settings (possibly the L1 settings) and of course input. Traditional skill-based approaches see learning as explicit study followed by proceduralization of knowledge and automatization of the resulting rules. In usage-based approaches, the heart of learning is statistical tallying of items in the input. Regardless of approach, a distinction is commonly drawn between implicit (unconscious) and explicit (conscious) learning. Increasingly important as well is the distinction between declarative and procedural learning, which is often seen now as a preferred alternative to dividing learning into implicit and explicit types (see Section 4).

The term incidental learning is sometimes associated with implicit learning but is probably better seen as the absence of an intention to learn (Reference SchmidtSchmidt, 1990). If a learner is only concerned with getting the meaning from input but in the process becomes aware of some aspect of form, learning that results from this encounter would be incidental but explicit, not implicit.

Memory can be thought of simply as what is learned, perhaps as the internal models of the world referred to in Dehaene’s definition of learning. And this simple view is often good enough for practical purposes. A serious interest in theory introduces a great many complications. Maybe for this reason, psychologists’ presentations of memory often do not include a general definition of the term, only of individual types of memory (e.g., Reference Baddeley, Eysenck and AndersonBaddeley, Eysenck, & Anderson, 2020). Alternatively, they offer lengthy discussion of complications inherent in the concept (e.g., Reference Tulving, Tulving and CraikTulving, 2000).

One important complication is that memory comes in a number of varieties. The most widely accepted categorization scheme starts by distinguishing long-term (LTM) from short-term memory (STM). Long-term consists of procedural memory, which is the knowledge underlying skills of all sorts, and declarative memory, which is then split into memories of events (episodic memory) and memory of facts (semantic memory). In SLA, the term knowledge typically covers the things that might otherwise be called long-term memory, as seen in the common terms implicit/explicit knowledge, procedural/declarative knowledge and, more generally, knowledge of language.

To these can be added working memory, a variant on the traditional idea of short-term memory emphasizing its active nature – we hold things in STM in order to use them. Increasingly important in SLA research is variation among learners in their working memory capacity, particularly how such variation correlates with success in various aspects of learning.

In current conceptions (see Reference Baddeley, Hitch, Allen, Logie, Camos and CowanBaddeley, Hitch, & Allen, 2021; Reference D’Esposito and PostleD’Esposito & Postle, 2015), working memory can almost be equated with attention – the things we are paying attention to are “in” working memory. The concept of attention has a long and very rich history in cognitive and neural research (see Reference CohenCohen, 2014; Reference Nobre, Mesulam, Kastner and NobreNobre & Mesulam, 2014), especially in the context of perception, and so, not surprisingly, has received extensive application in SLA. The logic of the application is straightforward: We are most likely to remember something, and to remember it most clearly and strongly, if we pay attention to it, so we might expect learners’ attention to input and to given aspects of it to be important for language learning. Attention has been closely associated with consciousness, in the cognitive literature and in SLA, though it would be a mistake to equate the two (Reference Koch and TsuchiyaKoch & Tsuchiya, 2012). Serious study of attention, like serious study of memory, takes us beyond the intuitive notion that we normally assume and introduces some complexities, to which I will return later.

Another important concept is consciousness (or awareness, often used as essentially synonymous). The term has been used in many different ways in different areas, but the meaning that is relevant here is the experiential one: We are conscious when we are in a normal waking state and unconscious when we are asleep or in a coma; we are conscious of something when that something is part of our immediate experience. The role of consciousness, in this sense, has always been a key issue in second language learning, if often implicitly. To what extent is learning a conscious process? Differing answers have inspired (or perhaps justified) differing approaches to teaching.

A number of common terms are related to consciousness. Reference KrashenKrashen’s (1981, Reference Krashen1982) conscious learning and unconscious acquisition are foundational concepts in the field, though the use of the terms has faded. The preferred terms now are explicit (conscious) and implicit (unconscious), reflecting prominent work in cognitive psychology (see Reference Reber and AllenReber & Allen, 2022). The terms distinguish both two types of learning and two types of knowledge. Especially important in the context of input are Reference SchmidtSchmidt’s (1990) notions of noticing and noticing the gap (Reference Schmidt, Frota and DaySchmidt & Frota, 1986). Noticing, for Schmidt, referred to a conscious registration of something present in the input, without regard to whether the learner knows what that something is or what it means. The term is commonly used with a much broader, ordinary language meaning, though. Noticing the gap is awareness that something in the input is not consistent with the current state of the interlanguage, though again the term is often used loosely. Prominent in the literature is Reference SchmidtSchmidt’s (1990) Noticing Hypothesis, which states that awareness of features of the input – “noticing” – is necessary for their acquisition.

Perception, the way that information from the senses is processed and interpreted, has not received a great deal of explicit attention in SLA. But input processing is itself a form of perception and so the concept deserves recognition here. I will consider it in some detail in Section 4 (for introductions to perception, see Reference MatherMather, 2011; Reference Wolfe, Kluender and LeviWolfe et al., 2018).

Other significant concepts in the area of input involve its emotional, or affective, aspect. It is generally agreed that language learning and language use are influenced by emotions such as anxiety, pride, and embarrassment, as is expected given the nature of perception, learning, and memory in general. Reference KrashenKrashen (1981, Reference Krashen1982; Reference Dulay, Burt and KrashenDulay, Burt, & Krashen, 1982) captured the relation in his affective filter, the idea being that negative emotion prevents input from reaching its destination, that is from being used in acquisition. Reference Sharwood SmithSharwood Smith (1996) briefly considered a possible positive role of affect, suggesting that in order to be useful for learning, input requires an affective “validation.” In the Modular Cognition Framework (Reference Truscott and Sharwood SmithTruscott & Sharwood Smith, 2019) this led to the notion of value assigned to instances of input and therefore to the target language and aspects of it (see Section 4). Dulay, Burt, and Reference KrashenKrashen’s (1982) speaker models and Reference Beebe, Gass and MaddenBeebe’s (1985) notion of learner preferences also address the relation between affect and input. The idea, in each case, is that learners attend specifically to the speech of people that they choose as their models and not to that of others, in effect selecting the input that they want to take in.

1.4 Key Concepts Associated with Instruction

A number of relevant terms are associated primarily with language instruction. One group involves the increasingly popular approach of using the target language, wholly or in part, to teach content. Such instruction is not intended specifically to provide extensive input, but this feature tends to dominate, if only because the classroom setting is more conducive to listening experience than to speaking experience. This type of instruction includes immersion, content and language integrated learning, bilingual education, content-based instruction, content-based language teaching, dual language education, and English medium instruction. The ideas and practices overlap greatly, a single type is commonly divided into several sub-types, and the terms themselves can have different meanings to different people (see, for example, Cenon, Genesee, & Gorter, 2014). The details of this confusing picture are peripheral to a discussion of input and so I will not pursue them here.

Instruction often includes efforts to provide learners with as much input as possible, spawning a number of common terms. Extensive reading is what its name suggests. Variant terms, sometimes involving limited adjustments, include free voluntary reading, sustained silent reading, guided self-selected reading, and pleasure reading, the latter term bringing out the emphasis commonly placed by proponents on the importance of learners’ enjoyment. Narrow reading focuses on works dealing with a particular topic or by a particular author. Paralleling extensive and narrow reading are extensive listening and narrow listening. To these can be added extensive viewing, which takes advantage of the visual clues provided by videos, for example. Its natural companion term is narrow viewing.

Reference LongLong (1983) argued that input is most effective when it occurs in the context of interaction. On this idea, instruction should emphasize two-way communication, in which negotiation between the participants results in modification of “the interactional structure of conversation.” Intimately associated with interaction is focus on form, introduced by Reference Long, de Bot, Ginsberg and KramschLong (1991). He defined it in part by contrast with focus on forms. The latter is essentially traditional grammar instruction, especially that following a grammatical syllabus. In focus on form, in contrast, activities are meaning-focused but with aspects of form explicitly addressed in the context of those activities. This can be seen especially now in task-based instruction. Input is a natural part of such instruction and the interventions are likely to be focused on aspects of the input, but it is only one part of the instruction.

Discussion of this topic requires some caution because “focus on form” has not always been used with the meaning that Long gave it. After Long’s proposal, the term became very popular, to the extent that it was sometimes applied to instruction and research that looked more like focus on forms. That said, the idea has generated a tremendous amount of research and will no doubt continue to do so. The looser term form focus is also in common use.

Intimately associated with interaction, as well as focus on form, is corrective feedback. It is a prominent research topic in both written and oral contexts. Particularly noteworthy in this context are recasts, which might well be considered a form of input.

Finally, ideas about output have acquired a prominent place in SLA, notably in Reference Swain, Gass and MaddenSwain’s (1985, Reference Swain and Hinkel2005) Output Hypothesis, which holds that comprehensible input must be supplemented by output practice.

Reference Sharwood SmithSharwood Smith (1981) introduced the idea of consciousness raising, in which learners discover features of the language guided by the teacher in different ways and to different degrees. To avoid the implication that changes in the learner’s state of mind (consciousness) are an essential feature of the process, he later offered input enhancement as an alternative term (Reference Sharwood SmithSharwood Smith, 1993), highlighting the fact that it is an external adjustment, which might or might not be reflected in the learner’s mind. This term has since become an institution in the field. As it is usually understood, input enhancement is about making selected aspects of input more salient in the hope that this will facilitate acquisition of those aspects (see Reference Sharwood SmithSharwood Smith, 1991, Reference Sharwood Smith1993).Footnote ² A teacher or textbook writer might place past tense endings in bold, for example, to draw learners’ attention to those forms. The oral equivalent would be to stress the forms.

An alternative to making a feature especially salient in a given instance is to provide input that contains many instances of that feature. A text might be written to include repeated use of passive forms, for example. This idea of an input flood (Reference Trahey and WhiteTrahey & White, 1993) was included in early notions of consciousness raising (Reference Rutherford and Sharwood SmithRutherford & Sharwood Smith, 1985), though it was only noted briefly and without the name. It might be considered a form of input enhancement, and is sometimes used in combination with more prototypical varieties.

The notion of salience is crucial here, and also has more general significance for SLA. Learners’ success in acquiring a linguistic element might be affected by whether that element is especially salient or especially nonsalient in the input. The term has most often been used in an intuitive sense, making serious study difficult, but efforts have been made to deal with it in a more solid and scientific manner (see Reference Gass, Spinner and BehneyGass, Spinner, & Behney, 2018).

2.1 Maximizing General Exposure to (Comprehensible) Input

If input, and comprehension of input, is the heart of learning, then a natural goal for instruction is to maximize the learners’ exposure to the language, in ways that facilitate comprehension. Some approaches to teaching adopt this as their primary goal (Reference AsherAsher, 2012; Reference Krashen and TerrellKrashen & Terrell, 1988; Reference Mason and KrashenMason & Krashen, 2020b; Reference WinitzWinitz, 1981, Reference Winitz2020). Not surprisingly, a significant amount of relevant research exists.

2.2 Input Enhancement and Input Flood

Input enhancement, again, is about making selected aspects of input more salient in the hope that this will facilitate acquisition of those aspects. The idea of enhancing input has generated a considerable amount of research (for review, see Reference BenatiBenati, 2016; Reference GascoigneGascoigne, 2006; Reference Han, Park and CombsHan, Park, & Combs, 2008; Reference Lee and HuangLee & Huang, 2008; Reference Nassaji, Loewen and SatoNassaji, 2017; Reference Pellicer-Sánchez, Boers, Siyanova-Chanturia and Pellicer-SánchezPellicer-Sánchez & Boers, 2019). Results have been inconsistent and not particularly impressive overall. Inconsistent results are commonly attributed, at least to a large extent, to differing methodologies, with unclear implications. Making sense of the findings probably also requires us to place them within a broader understanding of the mind, which is to say within a general theoretical framework (Reference Sharwood Smith and TruscottSharwood Smith & Truscott, 2014a). If we want to understand how adjustments in learners’ input impact their learning, we first need a clearer idea of how input is processed. Another limitation of the research is that the focus is almost always on the strictly linguistic aspects of the input. The focus could be expanded to include enhancement of other aspects, such as affect and context.

The alternative to standard input enhancement is to provide input that contains many instances of a selected feature, like simple past tense forms – an input flood. This approach has received much less attention in the research, but a number of studies have been done, a typical conclusion being that the flood can, inconsistently, help learners produce correct instances of the language but does not help them avoid incorrect instances (for review of the research, see Reference BenatiBenati, 2016; Reference Nassaji, Loewen and SatoNassaji, 2017; Reference Pellicer-Sánchez, Boers, Siyanova-Chanturia and Pellicer-SánchezPellicer-Sánchez & Boers, 2019).

2.3 Noticing and Noticing the Gap

The terms “noticing” and “noticing the gap,” ubiquitous in the literature, embody the idea that consciousness has a central role in second language learning. Paradoxically, their origin lies in Reference Krashen, Gass and SelinkerKrashen’s (1983) discussion of the way that input is used, unconsciously, in acquisition. He suggested that learners “notice (at a subconscious level)” both elements of the input and any “gap” that exists between those elements and the current state of their knowledge. This comparison process leads to acquisition. To this picture Reference Schmidt, Frota and DaySchmidt and Frota (1986) added, contrary to Krashen’s thinking, the hypothesis that the noticing is necessarily conscious – learners have to be aware of both the linguistic elements and the contrasts in order to benefit from their presence. Reference SchmidtSchmidt (1990) took the idea further, developing a more technical notion of noticing, which served as the basis for his Noticing Hypothesis – in order to benefit from input, learners must be aware of the relevant aspects of form in the input.

Noticing and the noticing hypothesis are referred to in a great deal of research on language instruction. Most of the work does not study them as such, though; the authors present them as background for the study or appeal to them as possible ways to understand the findings. A frequent but typically unrecognized problem in these references is that the meaning assigned to the word “noticing” shifts (see Reference Truscott and Sharwood SmithTruscott & Sharwood Smith, 2011). Reference SchmidtSchmidt (1990) was proposing a technical notion, with a relatively narrow meaning that excluded much, and probably most, of the things that researchers are interested in. Not surprisingly, then, there is a strong tendency in the literature to fall back on the ordinary language meaning of the word – awareness of something (anything). This research generally falls in the category of form focus, which I will briefly describe herein and then return to in Section 5.

A number of studies have specifically targeted the noticing hypothesis, usually with a recognition of the relatively narrow scope of “noticing.” This means bringing in the accompanying notion of “awareness at the level of understanding” (see especially Schmidt, 1995), a type of awareness that was excluded from the hypothesis but is important in the context of second language instruction. An important addition to this research area is the application of eye-tracking methodology to determine what is or is not noticed (see Reference GodfroidGodfroid, 2020). Several reviews of noticing research are available (Reference Gass, Behney and PlonskyGass, Behney, & Plonsky, 2020; Reference LeowLeow, 2015; Reference LoewenLoewen, 2020; Reference Loewen and SatoLoewen & Sato, 2017), reporting mixed findings. The issues involved in noticing and the research exploring it are large and complex, and I will not go into them here (for critical reviews, see Reference ParadisParadis, 2004, Reference Paradis2009; Reference TruscottTruscott, 1998; Reference Truscott and Sharwood SmithTruscott & Sharwood Smith, 2011; Reference VanPattenVanPatten, 1994, Reference VanPatten2015).

2.4 Implicit Knowledge and Implicit Learning

The notion of implicit (unconscious) knowledge/learning comes from cognitive psychology (e.g., Reference ReberReber, 1993), particularly from experiments designed to test the possibility that people can learn things without being aware of what they learned or of the fact that they learned it – in other words, the possibility of implicit knowledge and implicit learning. In one design, participants see a row of lights in front of them flashing on and off in patterns that appear random but actually follow a pattern, of which they were not told. Their task is to respond to a given light going on by pressing the button corresponding to that light as quickly as possible. It was found that their reaction times gradually decreased, as if they were correctly anticipating the lights, but no conscious knowledge of the underlying pattern could be found by the experimenters, apparently indicating that they had acquired implicit knowledge, through implicit learning.

The other standard paradigm for implicit learning research, which is perhaps more relevant here, uses small artificial grammars, mimicking natural language grammar. Learners are presented with large numbers of “sentences” that can be produced by the grammar and then tested on how well they can judge if other instances, not previously shown to them, can or cannot be produced by that grammar. If they show some ability to do so, as they often do, and do not show any signs of conscious knowledge of how they are doing it, we have evidence of implicit learning and implicit knowledge.

The implicit learning literature is enormous (see especially Reference Reber and AllenReber & Allen, 2022), and it has established beyond any reasonable doubt that implicit knowledge and learning are real and important. It is commonly conceptualized in terms of statistical learning, involving the frequencies of individual items and of the associations between them (see Reference Rebuschat, Reber and AllenRebuschat, 2022, for relevant discussion). This conception is developed especially in usage-based approaches (e.g., Reference Ellis, Wulff, VanPatten, Keating and WulffN. Ellis & Wulff, 2020). Unconscious knowledge and learning need not be seen in these terms, though. Universal grammar theorists, for example, typically take a very different view of unconscious knowledge and learning (see discussion).

The importance of implicit knowledge and learning for SLA has come to be widely recognized (see Reference Lichtman and VanPattenLichtman & VanPatten, 2021; also the various papers in Reference EllisN. Ellis, 1994; Reference RebuschatRebuschat, 2015; Reference Sanz and LeowSanz & Leow, 2011; Reference VanPatten, Keating and WulffVanPatten, Keating, & Wulff, 2020). The terms “implicit” and “explicit” now appear very widely in the research, notably in work on the effects of formal instruction. Researchers are interested in the effects of pedagogical interventions on the development of each type of knowledge and in the possible interactions of the two. For this purpose there is a need for means of distinguishing them in practice, a challenge that was taken up by Reference EllisR. Ellis (2005; see also Reference Ellis, Loewen and ElderR. Ellis et al., 2009). The criteria he offered are widely applied in research. But in this work, the role of input is at best a peripheral concern; its effects are tangled up with the effects of explicit instruction and other factors.Footnote ³ Another limitation of this work is that when implicit knowledge is acquired we do not know how it was acquired – implicitly or explicitly or through some combination of the two.

A substantial body of SLA research more directly addresses implicit learning and implicit knowledge, by studying them in controlled conditions (e.g., Reference Brooks and KempeBrooks & Kempe, 2013; Reference DeKeyserDeKeyser, 1995; Reference Hama and LeowHama & Leow, 2010; Reference Leung and WilliamsLeung & Williams, 2011; Reference Rogers, Révész and RebuschatRogers, Révész, & Rebuschat, 2016; Reference WilliamsWilliams, 2005). This work has made it reasonably clear that implicit learning does occur at least sometimes, though results are inconsistent. Perhaps more importantly, research of this type encounters serious issues of ecological validity, as it is difficult to pursue the questions in realistic contexts with realistic input and realistic measures of learning. We know that language-related knowledge can be implicitly acquired in laboratory settings, but what this tells us about actual language learning is open to dispute.

A significant issue for the study of implicit learning is whether it is influenced by explicit knowledge. We should expect there to be at least some influence of this sort, because of the potential of explicit knowledge for making input more comprehensible and its possible use by a learner to create virtual input (see Section 1). But the broader question is difficult to directly study. The problem of distinguishing implicit from explicit knowledge becomes severe when the latter has been automatized through extensive use and so is now used quickly and effortlessly and with little awareness. Reference DeKeyser, Loewen and SatoDeKeyser (2017; Reference Suzuki and DeKeyserSuzuki & DeKeyser, 2017) has been especially concerned with this problem, as automatized explicit knowledge has a central place in his theory.

Reference Suzuki and DeKeyserSuzuki and DeKeyser (2017) sought to empirically test the possibility that automatized explicit knowledge contributes to implicit knowledge. The measures that were used for automatized explicit knowledge are a possible issue in this study. The tests were timed grammaticality judgments, both written and aural, and a timed fill-in-the-blank task. The problem is that these tasks readily lend themselves to the use of implicit knowledge – people can easily perform them in their native language, for example, whether or not they have explicitly studied grammar. While the authors may be right that the tasks encouraged learners to focus on form and so encouraged the use of automatized explicit knowledge, a claim that implicit knowledge played no role in them, or that it had only a negligible role, seems doubtful. If the “explicit” tasks allowed even a fairly small role for implicit knowledge, then the weak relation that was found can be readily interpreted as a relation not between implicit knowledge and automatized explicit knowledge but rather between implicit knowledge and implicit knowledge. This is not, however, reason to close the book on research of this sort – challenges can be overcome.

2.5 Attention

Attention is an important topic in SLA, and the term appears throughout the literature. It is usually not treated as a distinct branch of its own, though, as most research on attention in SLA can also be classified as work on consciousness, noticing, implicit learning, input enhancement, focus on form, and possibly other topics as well.

Attention has been the focus of some theoretical work in the field. Reference GassGass’ (1988) integrated model of SLA (see also Reference Gass, Behney and PlonskyGass, Behney, & Plonsky, 2020, Ch. 17) gives attention a central place. Thinking on attention in SLA has been strongly influenced by Posner’s theory (see Reference PosnerPosner, 2012; Reference Posner and PetersenPosner & Petersen, 1990; Reference Posner, Rothbart, Milner and RuggPosner & Rothbart, 1992), in which attention is split into three parts: alertness (or vigilance), orientation toward the stimulus, and detection of the target. Reference Tomlin and VillaTomlin and Villa (1994) was essentially an application of the theory to SLA. Posner’s work was also used by Reference RobinsonRobinson (1995), who interpreted noticing as detection (of a formal feature in the input) accompanied by rehearsal of that feature in working memory.

Much of Schmidt’s writing on noticing (see especially Reference Schmidt and RobinsonSchmidt, 2001) focused on attention rather than consciousness, treating the two as essentially equivalent for his purposes. He argued that attention is necessary for all aspects of second language learning. While the claim has appeal, there is a problem, for this and other applications of attention to SLA. As Schmidt noted, the word “attention” has a variety of meanings in psychology. A consensus appears to exist in cognitive theory that there is in fact no single thing to which the term applies. It is a blanket term covering many different processes (e.g., Reference Allport, Meyer and KornblumAllport, 1993; Reference CohenCohen, 2014; Reference Nobre, Mesulam, Kastner and NobreNobre & Mesulam, 2014). For Nobre and Mesulam, the treatment of attention as a domain in itself has probably been a mistake; the features associated with “attention” should instead be studied as inherent parts of a wide variety of processes. There is also a danger of “attention,” like “noticing” being used in a loose, ordinary-language sense and thereby losing what scientific foundation it has. These are general issues for work that invokes attention.

2.6 UG-based Research

Universal grammar (UG) is the innate knowledge of language hypothesized to underlie first language acquisition. The extension to second language acquisition constitutes a rich area of research (see Reference HawkinsHawkins, 2019; Reference Mitchell, Myles and MarsdenMitchell, Myles, & Marsden, 2019; Reference Slabakova, Leal, Dudley and StackSlabakova, Leal, Dudley, & Stack, 2020; Reference White, VanPatten, Keating and WulffWhite, 2020), but input in itself has not been a major concern in this research, a point brought out by Reference Rankin and UnsworthRankin and Unsworth (2016). From a UG perspective, the core of acquisition is establishing the values of various innately given parameters, determining word order for example. This is done through input, of course, but UG researchers have generally been more interested in showing the insufficiency of input for learning.

The acquisition process involves learners analyzing their input, on the basis of the innate principles and the L1, and making deductions from it about the nature of the target language. This is an “input as evidence” conception, the learner seen as a detective using evidence to solve a problem. So the way that input is analyzed or misanalyzed by learners, often reflecting L1 influence, is an important concern. Within this deductive approach, there is now a general recognition that frequency is important.

The bulk of the research that has been done within the UG perspective has assumed Chomskyan linguistic theory, in one form or another. But there is substantial variety among theories. Reference JackendoffJackendoff’s (1997, Reference Jackendoff2002) linguistic theory deviates considerably from Chomsky’s thinking but remains very much within the UG camp. In SLA, the Modular Cognition Framework (e.g., Reference Sharwood Smith and TruscottSharwood Smith & Truscott, 2014b; Reference Truscott and Sharwood SmithTruscott & Sharwood Smith, 2019), along with Reference CarrollCarroll’s (2001) Autonomous Induction Theory, assumes UG but is not committed to any particular linguistic theory. Reference VanPatten, VanPatten, Keating and WulffVanPatten’s (2020) work similarly takes UG as a background assumption but focuses on the processing that produces the input to innate learning processes, leaving the nature of these processes as an open question.

2.7 Input Processing and Processing Instruction

Reference VanPatten, Piske and Young-ScholtenVanPatten’s (2009, Reference VanPatten, Loewen and Sato2017, Reference VanPatten, VanPatten, Keating and Wulff2020) Input Processing and Processing Instruction have spurred a considerable amount of research and are likely to continue to do so. VanPatten’s work combines serious theoretical development, empirical research, and pedagogical application, all focused on input as the key to learning. The work is based on the fundamental point that learning depends on the way that learners process input.

The first issue then is how the processing occurs, and this is the domain of Input Processing (IP). VanPatten accepts the existence of universal grammar as a crucial part of acquisition, but does not give it a direct role in processing. Instead, separate processing mechanisms prepare the input for use by learning mechanisms, including UG. IP is about the processing mechanisms, for which a number of general principles are hypothesized. This understanding of input processing points to possibilities for pedagogical intervention (Processing Instruction, or PI). If the processing principles being used by learners are inappropriate for the language they are acquiring, the learning mechanisms will receive bad input, and learning will suffer as a result. If the processing procedures can be altered to a more appropriate form, learning will benefit. VanPatten stresses that this is not a method or an approach but rather a particular sort of intervention that can be used in a wide assortment of methods.

IP and PI have, again, stimulated a large body of experimental research (see Reference VanPatten, Piske and Young-ScholtenVanPatten, 2009, Reference VanPatten, Loewen and Sato2017, for lengthy lists of references). As the goal of PI is to alter processing strategies, this is the focus of the research, and the main conclusions are that it is successful in this respect, both in absolute terms and relative to other types of intervention. Importantly, it is not claimed to improve learners’ communicative ability.

2.8 Use of the Target Language to Teach Content

Use of the target language to teach content is an increasingly popular approach, appearing in a variety of forms. Considerable research has been carried out in classes of these types (e.g., Reference Collier and ThomasCollier & Thomas, 2017; Reference Feddermann, Möller and BaumertFeddermann, Möller, & Baumert, 2021; Reference Graham, Choi, Davoodi, Razmeh and DixonGraham, Choi, Davoodi, Razmeh, & Dixon, 2018; Reference Martínez AgudoMartínez Agudo, 2020; Reference Watzinger-Tharp, Rubio and TharpWatzinger-Tharp, Rubio, & Tharp, 2018). In the SLA literature, the immersion research in Canada has played an especially important role (e.g., Reference CollierCollier, 1992; Reference Swain and FreedSwain, 1991). The approach tends to do well, though results are variable and it is difficult to draw general conclusions given the variety of pedagogical approaches that fall under this general heading. A common finding is that learners tend to do especially well in comprehension, often attaining native-like ability, but are somewhat less successful with productive skills. Proposed explanations for this limitation have included limited opportunities for production, narrow or impoverished input, and insufficient formal language instruction or insufficient integration of such instruction with content-based teaching (e.g., Reference Snow, Met and GeneseeSnow, Met, & Genesee, 1989; Reference Swain, Gass and MaddenSwain, 1985).

2.9 Interaction, Focus on Form, Output, and Corrective Feedback

The topics of this section are all major research areas in SLA and are intertwined with one another in theory, research, and pedagogy. They are also intertwined with input and so are relevant here, but input is not a central theme and the research does not seek to isolate or focus on input as such and so inferences about its nature or role are not straightforward.

The role of interaction in second language learning constitutes a particularly rich research area, or perhaps a cluster of rich research areas. The target of this work, interaction, is a coherent package of input, output, and feedback, within a communicative context in which meaning is negotiated between the participants. While the role of input is not isolated, most will agree that it is central. The primary motivation for Long’s early, influential proposals was in fact the idea that interaction should contribute greatly to the goal of providing comprehensible input (see Reference LongLong, 1983, Reference Long, Ritchie and Bhatia1996).

There does not seem to be any dispute regarding the value of the overall package for acquisition. This value can probably be accommodated in all major theories, and interaction research can potentially contribute to the development of SLA theory quite broadly. Probably the bulk of the research falls under the heading of focus on form, or more loosely form focus. It should be stressed that, despite the name, form focus as it is commonly recommended now treats meaning as primary, placing the form focus in the context of meaningful activities.

All the topics considered in this section are closely associated with work on consciousness and attention, described earlier, notably work on noticing and the implicit-explicit distinction. The sources cited there are also relevant here. Extensive review and discussion of research in this group of interrelated topics can be found in Reference Gass, Behney and PlonskyGass, Behney, and Plonsky (2020), Reference Gass, Mackey, VanPatten, Keating and WulffGass and Mackey (2020), and Reference Loewen and SatoLoewen and Sato (2017).

2.10 Input and the Development of Phonology

In much SLA work, including work on input, phonology tends to be taken for granted. But phonological processing and learning is in a sense the most fundamental issue, as all spoken input – and any resulting acquisition – begins with perception of the sounds. A substantial body of research exists on the development of second language phonology, and much of it involves input (e.g., Reference Bohn and MunroBohn & Munro, 2007; Reference Flege, Piske and Young-ScholtenFlege, 2009; Reference Kennedy, Trofimovich, Loewen and SatoKennedy & Trofimovich, 2017; Reference Piske and Young-ScholtenPiske & Young-Scholten, 2009; Reference White, Titone, Genesee and SteinhauerWhite, Titone, Genesee, & Steinhauer, 2017; Young-Scholten, 1996).

Largely the same issues arise here as in other areas of input research. First, the importance of input for development is acknowledged, while research is directed to the question of how important it is relative to the other factors, such as the L1 and the age of the learner. The role of consciousness and attention is a common theme in the research, frequently cast in terms of noticing. This concern is naturally accompanied by applications to pedagogy, including issues of output practice, explicit instruction, interaction, and input enhancement. The question of innateness is also pursued – the possible innateness of principles involving hierarchies and markedness as well as stages of development possibly arising from innate constraints (Universal Phonology).

2.11 Conclusion

Identifying distinct branches of research on input is, again, difficult and somewhat arbitrary. We might add research done within given SLA theories, like Reference O’GradyO’Grady’s (2005, Reference O’Grady2015) emergentist model, MacWhinney’s (Unified) Competition Model (e.g., Reference MacWhinney, Gass and MackeyMacWhinney, 2012), and Pienemann’s Processability Theory (see Reference Pienemann, Lenzing, VanPatten, Keating and WulffPienemann & Lenzing, 2020). I have left out these and other theories because, while their treatment of input is important, the research done within them is rarely focused on it. I have also left out most research done within socially or culturally oriented theories because “input,” as a cognitive, information-processing concept, is not widely accepted in these areas and does not play a major role in the research. Likewise for the Complex Dynamic Systems Theory developed for SLA by Reference Larsen-Freeman, VanPatten, Keating and WulffLarsen-Freeman (2020), who rejects the term “input” as dehumanizing (among other objections).

4.1 Input, Perception, and Bird Calls

The first and most important topic is perception. And the first and most important thing to understand about perception, as noted in Section 1, is that it is not a process of taking what is outside and bringing it inside. It is about constructing internal representations of experiences. The representation constructed in this way is connected to already stored objects, and associated with related knowledge and with emotion. The end result of the process is the scene that we perceive, a representation of what is outside but by no means a simple internalization of it. The process is depicted in Figure 1.

Figure 1 Perception

The figure applies to all the senses, individually, so “perceptual” could be read as “visual,” producing images, or as “auditory,” producing sounds, among other possibilities. The perceptual output stage is the border between perceptual systems and higher level systems, particularly the conceptual. It represents the ultimate product of the perceptual processing – the sounds and images, that is – and is thus the immediate input to conceptual processing, which gives meaning to those products.

The existence of distinct perceptual and conceptual systems can be seen in two different types of agnosias (see Reference Behrmann and GoldsteinBehrmann, 2010; Reference Griffiths and GoldsteinGriffiths, 2010). Damage to parts of the brain that deal with conceptual processing can leave you with a clear image or sound but no idea what it is that you are seeing or hearing (associative agnosia). Damage to lower areas, on the other hand, can prevent the formation of a clear image or sound, making it difficult or impossible for the later processes to interpret it (apperceptive agnosia). Note also the bidirectional arrows in the figure, expressing the extensive interaction that occurs among the different systems.

The perceptual process is perhaps best understood through examples. Consider first the case of hearing a sound which you, being an expert on bird vocalization, recognize as the call of a marbled godwit. The air coming from the bird disturbs the air around you, producing vibrations in your ears. The auditory system constructs from these vibrations a representation for the sound “out there” and matches it with already existing sound representations. The candidates for this matching are contained in the store of sounds you experienced in the past – what can be called auditory output representations – one of which is the call of the marbled godwit. When this representation is activated, it activates a connected representation in the conceptual system, MARBLED GODWIT CALL. You have then perceived the call.

What about a bird call that you don’t already know? If you have never heard it before, there is no representation of its sound in the auditory store or its interpretation in the conceptual store. Auditory processing therefore creates a new auditory representation, using sounds that are already present. This new representation will then be connected to a conceptual representation. If you witness the bird making that sound, this new representation will specify that particular bird (the alternative would be a generic “bird sound” representation). You have then learned the sound that a particular bird makes.

We can now ask about the input to this learning. What exactly is it? One possibility is to identify the input as what is out there in the world, since that is what started the whole thing and is what the new representations are representing. But the new auditory representation that you acquire is the product of an elaborate construction process involving a number of intermediate representations (not shown in the figure), so maybe the input has to be placed between the bird and the output representation. The new conceptual representation is based on the auditory representation,Footnote ⁴ so maybe the auditory representation should be considered the real input to learning. Nothing is particularly wrong with any of these options – apart from the fact that each deals with only one part of the process. So maybe we should say the input is the whole perceptual process.

How we choose to apply the term “input” is rather arbitrary. The important point is that it is inseparable from the construction process and the individual representations involved in it. Also worth noting is the contribution of the visual experience, seeing the bird making noise. We could say it is part of the input, or that it provides the information needed to establish the conceptual representation; in other words, that it makes the input comprehensible.

4.2 Input, Perception, and Language

When we go from this simple example to language and language learning, extra complications come in, lots of them. But the basics of perception remain the same. What is distinctive about human language is that it adds a very rich means of connecting sounds (the output of auditory processing, that is) to concepts, allowing us to express an unlimited number of possible ideas and, more immediately relevant, to understand someone else’s expression of them. Figure 2 portrays the perception of language sounds (or written words).

Figure 2 Perception and language

Suppose during the bird noise someone says “That’s a marbled godwit.”Footnote ⁵ This event, out there, causes vibrations in your ears, which trigger auditory processing, culminating in an output representation of the sound of the utterance. To this point perception is essentially the same process as in the previous examples. If you didn’t understand English it could be almost entirely the same process: A conceptual representation would be directly constructed for the auditory output, perhaps consisting of the information that the sound was “something in English,” comparable to the concept in the previous example that you were hearing some sort of bird call.

Things are more interesting, though, if you can understand English or, especially, if you are learning to understand it. In this case processing will take a left turn at the auditory output. Specifically linguistic processes will deal with the sounds, constructing a sentence from them. Conceptual processes then form an interpretation for this sentence, just as they interpreted the bird sound (though with far greater complexity, of course). This linguistic-conceptual processing is carried out by systems having their own specializations, using the output of processing by the perceptual systems.

Note that the linguistic box in Figure 2 is labeled core linguistic knowledge rather than just linguistic knowledge. This is because of the non-monolithic nature of language described in Section 1. The core is the portion that is specifically responsible for the rich sound-meaning connections that make human language special.

At this point we can ask again: Where is “input” in this picture? The term could be used to refer to what is out there, or to any or all of the intermediate steps in the processing, or to the process as a whole. All can be useful ways of talking about the phenomena. Problems arise when we think that input is some real “thing” that needs to be pinned down. The important point is that input is inseparable from perception and therefore shows all the general characteristics of the perceptual process: multi-step construction carried out in terms of already-existing representations (linguistic and other). And it is this construction process that sets up learning and makes it possible.

4.3 Language in the Mind

The previous examples referred to different components in the mind – auditory, visual, conceptual, core linguistic – reflecting the fact that a complex system, such as the human mind, is necessarily composed of interacting parts. These parts represent different types of knowledge, encoded in different ways, associated with different types of processing. In order for the system to function properly, these various aspects of its operations must be segregated in some sense, so that a new face for example is processed as a face and not as an algebraic formula or a string of phonetic features or some mixture of these and other types of knowledge. In the brain this can mean that faces occupy a particular regionFootnote ⁶ and/or that face representations have especially strong connections to one another and to the lower-level visual features that are used in their construction. A central question for research and theory is the nature of the segregation: What exactly are the parts of the system? How do they interact?

The parts, whatever they may be, are often referred to as modules, and when we seek to understand their nature we are studying modularity. Many would avoid these terms, but in a broad sense everyone accepts the idea behind them. The system simply could not work without some sort of segregation of the knowledge types and the processes that construct and work with each type. Important disagreements involve the nature and extent of the segregation and especially the role of our genes – to what extent and in what sense are the modules innately determined? I will return to these questions herein.

When we talk about parts of the system, language is naturally taken to be one of those parts. Linguistic input is necessarily encoded in terms of linguistic features, and stored in a linguistic place. I use the term “place” in a loose sense, simply to mean that some sort of segregation exists. Given the non-monolithic character of language described earlier, it is actually more accurate to speak of places rather than place. Of particular importance here is core language, the knowledge that allows rich connections between sounds and meanings, as pictured in Figure 2.

What happens when there are two languages in one head? A second language has, necessarily, all the components of a first language. We also know that a bilingual’s two languages are intertwined, as seen in the ability of bilinguals to smoothly switch between their languages and merge them in rule-governed ways within a conversation. Research has shown that when a bilingual is using one language, relevant elements of the other are automatically activated, meaning primed for use (for reviews, see Reference Brysbaert and DuyckBrysbaert & Duyck, 2010; Reference de Groot, Starreveld and Schwieterde Groot & Starreveld, 2015; Reference Kroll, Gullifer, McClain, Rossi, Martín and SchwieterKroll, Gullifer, McClain, Rossi, & Martín, 2015; Reference Schwartz and SchwieterSchwartz, 2015). A reasonable interpretation is that the two languages are stored in the same “places.”

The conceptual system has a special status. Unlike visual, auditory, or linguistic systems, it is not dedicated to one type of function but is rather a very general, multi-function system, which is to say we can acquire abstract conceptual knowledge of virtually any type. What all the types have in common is that they are represented in the abstract, domain-general format of concepts. Readers familiar with research on memory might note the resemblance to declarative memory – the kind of knowledge that we can, at least in principle, describe and talk about. I will return to this point later.

The conceptual system’s ability to acquire abstract knowledge on virtually anything naturally extends to language. The knowledge acquired in a linguistics class is an obvious example. More interesting here is the knowledge acquired in a grammar-oriented second language class. Knowing that the past participle verb form is used in English passive sentences is not fundamentally different from knowing that Ouagadougou is the capital of Burkina Faso. Both are instances of abstract conceptual knowledge. What makes the language-related knowledge different is simply that it is about language. This is the metalinguistic knowledge described earlier (1.1)

4.4 Changing the Places in the Mind: Learning

Learning from input means making changes in the places where the processing occurs. Returning to the bird-call examples, if you were one of those benighted individuals who had never heard of a marbled godwit, the linguistic-conceptual processes would form new representations – linguistic and conceptual – to accommodate the new word. This is what sets up learning.

In the account developed in the Modular Cognition Framework, this processing simply is learning (see Reference Sharwood Smith and TruscottSharwood Smith & Truscott, 2014b; Reference Truscott and Sharwood SmithTruscott & Sharwood Smith, 2004). Importantly, there is no conflict here with the idea (see 1.3) that learning means forming internal models of the external world. The claim is that the products of the perceptual (input) process directly, automatically contribute to the developing models, as opposed to the idea that they are first submitted to a separate set of learning processes, which then make the appropriate adjustments in the models. From this point onward, I will assume this parsimonious approach to learning.

Input and the acquisition that results from it should thus be understood as the interaction of processing in the perceptual system itself with the processing in the higher-level systems. The former produces the perceptual output representations which the latter then uses to establish linguistic representations of what is out there. In some cases, the higher-level processing is just a matter of activating (and thereby strengthening) what is already present, as when the learner hears “Good morning” or “How are you?” for the hundredth time. More often, it would mean representing a novel sentence by combining established representations in an already-established manner – when all the words and structures are known but have never been encountered together before.

Most interesting is the case in which the high-level processes are faced with a perceptual output representation that they cannot handle in either of these ways. This might mean a new word (marbled godwit), a new grammatical element (-ed), a new structure (passive), or perhaps something much more subtle, depending on one’s preferred linguistic theory. To form adequate overall representations for this novel perceptual output, the high-level processes will have to establish new representations. These new elements are then part of the interlanguage. If they prove useful in subsequent processing, their continuing use will strengthen them; if not, they will gradually become irrelevant as they get weaker and their rivals get stronger.

A principle that is generally accepted, but not so generally discussed or analyzed, is that understanding the meaning of the input is important for learning from that input. If you don’t understand what “That’s a marbled godwit” means, it is difficult to learn the word “marbled godwit.” If you are going to learn the English past tense marker from input like “Chris talked,” it is very helpful to know that the talking occurred in the past. In terms of Figure 2, understanding the input means having an appropriate conceptual representation of it, which can then be used in linguistic processing – and therefore linguistic learning.

4.5 Input, Perception, and Consciousness

Ideas about consciousness play a central role in SLA, as can be seen in the prominence of the terms noticing, noticing the gap, implicit, and explicit. Even when consciousness is not mentioned, it is often present in the background, in discussions of teaching or correcting language form. To understand input and its role in SLA, we have to look in some depth at consciousness, especially its place in perception.

When we hear (or read) a word or a sentence, what things can be part of the conscious experience and what things cannot? We are certainly aware of the sound; in normal circumstances it is very difficult not to be aware of it. Recall, though, that the sounds we hear are the product of the complex, multi-step construction process that is perception. This process itself is clearly not part of the conscious experience. In terms of the earlier depiction of perception, Figure 2, repeated here with a small addition as Figure 3, the conscious experience is of perceptual output but does not include what comes before that stage. The dashed arrow will be explained shortly.

Figure 3 Consciousness and perception

What about the purely linguistic parts of Figure 3? Are they part of our conscious experience? To anyone who has studied linguistic theory and its complex, unintuitive, and constantly disputed theories of phonology and syntax, it should be clear that the answer is no. Similarly, a language teacher struggling to answer students’ questions about grammar knows that we cannot find the abstract principles underlying language simply by examining our conscious experience. We can learn the principles presented in textbooks, and possibly infer some for ourselves, but they are not there in the conscious experience of hearing language.

This brings us to the trickiest part – the conceptual component: Can we be conscious of the meaning of what we hear? The best answer is “no, but ….” Concepts are abstract, which means they are a step removed from direct experience. Consider, for example, the concept of MUSIC. What would an experience of the abstract concept be like? When we listen to music, or hear the word “music,” the conscious experience is of some particular music, possibly with visual images of musical notes or of a band or a musical instrument. But these things are not MUSIC; they are perceptual elements associated with it. The same considerations apply to more concrete cases, like CUP. An abstract cup has no particular size, shape, color, and so on. But any cup you can experience or imagine does have particular features of this sort. The experiences are perceptual rather than conceptual. This is the perceptual bias of consciousness (Reference BaarsBaars, 1988, Reference Baars1997).

But, when we hear the words “cup” or “music,” there is certainly a sense in which we feel that we are aware of their meaning, even if it is difficult to find anything in the conscious experience that could be called a meaning. There is a feeling that the sounds are making sense – that they are connecting properly to concepts – and that we can express the meaning if we choose to. This feeling is shown perhaps most clearly by its absence when we hear something said in an unfamiliar language. In the two cases we see the contrast between the experience of hearing sounds and the experience of hearing interpreted sounds. The apparent conclusion is that the conceptual, meaning level is not directly part of conscious experience but is always lurking in the background, coloring the perceptual experiences. This is expressed in Figure 3 by the dashed arrow.

Putting all this together, the perceptual output level is where consciousness is found, at the border between perceptual and conceptual processing.Footnote ⁷ The senses present us with a picture of the world around us, which other components of the mind interpret and use for their own purposes. Our conscious experience is of this picture.Footnote ⁸ Awareness of the perceptual output representation is thus a normal part of perception, including the perception that constitutes second language input. When we are aware of “input,” it is specifically a perceptual output representation that we are aware of – a representation that is currently dominating processing at the perceptual output level. To this picture we have to add indirect awareness of meaning, in the form of a conceptual representation of the input.

4.6 Input, Consciousness, and Learning

We can now address an issue that has always been at the heart of SLA, if often implicitly: To what extent and in what ways is learning a conscious process? In particular, what do learners need to be aware of in regard to input? What does it mean to be aware of these things?

First, do learners need to be aware of “input”? There is, again, no “thing” called input, so this question can only be answered in terms of the components of the perceptual process, as this process is what constitutes input. Returning to Figure 3, awareness of input necessarily means awareness of perceptual output representations, because this is where consciousness resides. This is to say learners are aware of the sound of the voices they are hearing, or of the images of what they are reading. A claim that this sort of awareness is unnecessary would amount to a claim that we can learn language while we are asleep or fully absorbed in an unrelated task, and I know of no one in the field who would make such a claim. Thus, awareness of “input” itself is not a topic of research or discussion in SLA.Footnote ⁹ The issue is not learners’ awareness of input but rather their awareness of particular aspects of the input, aspects of its linguistic form in particular. This issue will be addressed shortly.

Perceptual output takes on special significance here because, in addition to being the locus of consciousness, it serves as the material for processing in the conceptual and linguistic systems, including the novel processing that constitutes learning, as shown in Figure 3. This is why awareness of input is so strongly associated with learning in general – why we learn little or nothing from “input” that occurs while we are asleep or distracted. If an input sentence does not come to dominate at the level of perceptual output, then (a) we are not aware of it, and (b) linguistic and conceptual processing can do little or nothing with it; in other words, it does not constitute intake.

Do learners need to be aware of meaning in order to learn from input? This question has not, to my knowledge, generated any debate, and is rarely discussed. It appears to be taken for granted that such awareness is either necessary or extremely valuable, though what exactly it means to be aware of meaning has never been made clear. In terms of the previous discussion, what it means is indirect awareness of conceptual representations – awareness that the perceptual output sounds (or images) are appropriately connecting up to conceptual representations, making them meaningful.

Thus, awareness of perceptual output and its connections to conceptual representations is a normal and probably essential part of learning from input, no matter how that learning occurs. One implication is that the distinction between implicit and explicit learning is not about awareness of input as such, as both involve awareness of perceptual output. The line between them must be drawn elsewhere, a point to which I will return shortly.

In SLA discussions of consciousness, the subject of interest, and the source of controversy, is the remaining question: Do learners need to be aware of linguistic form in their input, things like past tense endings or passive forms or word order? When the conscious experience is of a sentence itself but not of the fact that it contains a subject and a verb or that the two are appearing in that order, what can the higher-level processes do with this implicit linguistic information? If the sentence is in a language that we already know, the answer is clear: They can use it, quickly and efficiently, to construct linguistic representations of the sentence and determine its meaning. To successfully process and understand the sentence you just read, for example, you do not have to be consciously thinking of subjects and verbs or their order in the sentence. You use the implicit information automatically and unconsciously. But what about someone who is just learning the language? Does learning require them to be aware of subjects and verbs and such?

To address the question, we first have to ask what it means to be aware of language form in the input (see Reference Truscott and Sharwood SmithTruscott & Sharwood Smith, 2011, for more detailed discussion of these matters). As described in Section 1, knowledge of language comes in two very different varieties – linguistic and metalinguistic. Linguistic knowledge makes up the core linguistic system of Figure 3, while metalinguistic knowledge is found in the conceptual system. As argued earlier, we cannot be aware of the linguistic representations but can be aware, indirectly, of conceptual representations, including those of metalinguistic knowledge.

What, then, does it mean to be aware of subjects and verbs and such in the input? It means being aware of conceptual, metalinguistic representations, like SUBJECT, VERB, or PAST TENSE. And in order for one of these to be (indirectly) part of conscious experience, it must be connected to a perceptual representation that is currently dominating perceptual output. Thus, awareness of language form is awareness of a perceptual output representation connected to a conceptual, metalinguistic representation of the form. The simplest kind of example is awareness of the [t] sound in “walked” with its connection to a conceptual representation, namely PAST TENSE.

At this point we are no longer talking about the input sentence itself but rather a follow-up representation that isolates one portion or aspect of it. For the case of past tense [t] again, this means a perceptual representation specifically of the [t] portion of the original input, connected to the conceptual representation PAST TENSE. Awareness of these things is awareness that [t] is a past tense form. This awareness might be accompanied, perhaps fleetingly, by the experience of the voice in the head saying “past tense” or of images related to it.

A somewhat different case involves awareness of word order – awareness that the subject is preceding the verb, for example. If the input sentence is “The student walks to school every day,” the follow-up perceptual representation might be “The student walks,” which would be connected to a conceptual representation such as SUBJECT-VERB or SUBJECT PRECEDES VERB. We then have awareness of the information that “The student walks” is an instance of subject preceding verb, possibly with a conscious experience of the voice in the head or of related images. This explicit processing might lead to additional conceptual processing that yields a more abstract metalinguistic representation like SUBJECT PRECEDES VERB IN ENGLISH.

Implicit learning from input is in a sense much simpler than explicit learning. The learner just consciously hears and understands the sentence. Awareness is simply of the interpreted sounds of this sentence – the perceptual output representation.

How then does the awareness of linguistic form, as characterized here, relate to the development of the different types of knowledge of language? Development of the linguistic system, first, is generally recognized to be predominantly if not entirely implicit. Unconscious processes use the interpreted perceptual representation, as is, to make changes in the linguistic system – those changes that are needed to adequately represent the current input. Through many such instances of input processing, the learner gradually acquires the ability to use aspects of form automatically, and is never aware of the linguistic knowledge.

In contrast, metalinguistic knowledge, being simply conceptual knowledge that is about language, is typically the product of explicit learning. The previous descriptions of past tense and word order offer paradigm cases of the acquisition of metalinguistic knowledge (or at least the early stages of the process). But a role for implicit learning is also quite plausible – there is no apparent reason why it should not be possible. Findings of implicit learning under controlled laboratory conditions might well be interpreted in these terms, as the unconscious acquisition of conceptual knowledge about language, distinct from development of the linguistic system itself. The important point, for practical purposes, is that this learning is done without the linguistic specialization found in the linguistic system, and its products are therefore likely to be a pale imitation of genuine linguistic knowledge.

The conception of learning presented here, both implicit and explicit, is depicted in Figure 4. We could perhaps place a small question mark on the vertical “implicit” arrow (see the following section for related discussion).

Figure 4 Implicit and explicit learning from input

4.7 Rethinking the Implicit-Explicit Distinction

To this point I have followed standard practice in SLA by assuming that the implicit-explicit distinction is a fundamental divide in terms of knowledge and learning types. But this assumption is in dispute (in the SLA context, see Reference Ullman, Hickok and SmallUllman, 2016, Reference Ullman, VanPatten, Keating and Wulff2020).Footnote ¹⁰ The original dichotomy of learning and memory types – and the one that best fits with what we know about the brain – was not implicit vs. explicit but rather procedural vs. declarative (see Reference Squire and WixtedSquire & Wixted, 2011), in other words the ability to do things, physical and mental, vs. factual knowledge. For SLA, this distinction offers a good match with the distinction between the linguistic system, which underlies competent use of a language, and the conceptual system with its metalinguistic knowledge. Conceptual and declarative, as they are commonly understood, are not easily distinguished.

Taking procedural (linguistic) vs. declarative (metalinguistic) as the fundamental divide, we can then return to the question of consciousness in language learning, with the same answer offered earlier (see Figure 4). Neurobiological findings (see Reference ChunChun, 2000; Reference HenkeHenke, 2010; Reference Schendan, Searl, Melrose and SternSchendan, Searl, Melrose, & Stern, 2003; Reference Ullman, VanPatten, Keating and WulffUllman, 2020) suggest that procedural knowledge and learning are always implicit while their declarative counterparts tend to be explicit but can also be implicit. This is to say, again, that development of the linguistic system is an entirely implicit process while metalinguistic knowledge is typically, but not necessarily, obtained explicitly.

4.8 “Input” Revisited and Expanded

In thinking about how we interpret things out there, it is natural to focus narrowly on the object or event that we are mainly interested in, like the sound of a bird call or the utterance that is being processed. But perception is more than this, and a treatment of input should recognize this additional complexity. For the case of second language learning, the non-monolithic character of language dictates a broadened view of this sort. Knowledge of language includes a number of components that are not strictly linguistic, each of which is acquired through input and so should be considered a part of the input.

An important part of knowing a language, for example, is knowing how to use it appropriately in given contexts, involving, among other things, a particular speaker, a particular physical setting, and previous utterances and actions. These contexts are part of the perception (the input) and so become connected to the more strictly linguistic parts of the acquired knowledge, creating a composite “language in context” representation. In addition to context, perception also involves the person’s attitude toward what he/she is perceiving – the value attached to it and the emotion that it evokes. To these can be added the goal of the speaker. All of these become connected to the more strictly linguistic knowledge, through input. We thus need a broadened notion of input to include all these elements.

This broadening of “input” is one of the ways in which a social account of second language learning, dealing with such things as contexts and goals, interfaces with a cognitive account. It is naturally accommodated in the Modular Cognition Framework through its concept of internal context (Reference Sharwood SmithSharwood Smith, 2021; Reference Truscott and Sharwood SmithTruscott & Sharwood Smith, 2019).

4.9 How Is Learning Possible? The Issue of Innate Knowledge

Input, in the sense of what is out there, is embedded in the world. And the world is very complex. Our senses are constantly assaulted by an essentially infinite variety of sights and sounds (to say nothing of smells, tastes, and feels), from which we somehow derive our rich knowledge, including knowledge of language. More accurately, our implicit learning mechanisms derive our rich knowledge of language. They somehow manage to pick out, from the vast complexity of the world, exactly those things that are needed for their task and use them to create something remarkable – a human language. How is this possible?

An immediate response is that the implicit mechanisms identify patterns in the information they are receiving from the senses (the input) and in effect reproduce these patterns internally. But while this idea has considerable intuitive appeal, it still leaves us with the question of how the relevant patterns can be extracted from the vast complexity of the world. Patterns are relations among given elements, and the number of possible elements and possible relations among them is astronomical. How do the learning mechanisms know what counts as a relevant element for their current task? How do they know what kinds of relations to look for among those elements?

To get a feel for the enormity of the task, consider some of the information that is contained in a simple sentence like the following.

The man noticed a dog in the yard behind the fence.

The sentence contains exactly three instances of the word the; five other words, with a total of six syllables, appear between the first two instances, and two words, with a total of three syllables, appear between the second and third; the first word in the sentence begins with a voiced consonant while the last word begins and ends with an unvoiced consonant; the past tense marker appears on a verb that has two syllables, the first of which is stressed; an indefinite article immediately follows the past tense marker; the object of each preposition appearing in the sentence is headed by a singular, inanimate, countable noun and so on.

Also relevant is the listener’s background knowledge about the referents, for example, that the agent of the sentence (“the man”) is understood to be an Armenian businessman of medium build with a receding hairline, that the dog might well be a small terrier belonging to the good-natured gentleman down the street, and so on. Also of potential significance is that the person who uttered this sentence was seated at the time, was wearing a light blue shirt with a stain on the right sleeve near the elbow, and had a surprisingly low voice which rose in volume at a point in the middle of the sentence.

A general, unconstrained learning mechanism would have to keep track of all such information in the input it receives, maintaining a record of how often each item co-occurs with each of the others, in combinations of all possible sizes. It would also have to correlate each item and each combination with the current temperature in the room, the scrambled eggs which the learner had for breakfast, and the phase of the moon. And this, of course, is just a tiny sampling of the possibilities that would have to be considered.

This is the problem of computational tractability (see Reference CarruthersCarruthers, 2006; Reference FodorFodor, 1983). A learning mechanism that is not receiving strong guidance on what to look for will have so many possibilities to consider that its task becomes impossible. There has to be something that tells it to look for certain things in the input and disregard everything else and to consider only certain possibilities for the way those things can be associated. This is the essential, foundational element for a theory of learning.

What then is it that is guiding the learning? The approach I adopt here, based on the Modular Cognition Framework, is a UG approach. As such, it hypothesizes that language learning – specifically, learning of core linguistic knowledge – is a narrow, highly constrained process. Patterns in the input are no less real or important in this approach, but the basic elements that make up those patterns are identified in advance, as are the possible ways in which they can be related. The vast majority of possibilities are thus ruled out from the beginning, making the task manageable. In terms of the discussion in 4.7, UG is a system of constraints that evolved to facilitate the procedural learning of language. Such domain-specific adaptations seem quite natural with functions that have proven important for survival, functions like language.

Alternative approaches, commonly called “usage-based” now, hypothesize that learning is carried out by simple, general mechanisms that make no use of specifically linguistic knowledge (see for example Reference Ellis, Wulff, VanPatten, Keating and WulffN. Ellis & Wulff, 2020). But in order for such mechanisms to learn language (or anything else) there still have to be constraints that focus them on particular items and on particular ways those items can be connected. The foundational challenge for such an approach is in establishing what these constraints are and where they come from – without taking any specifically linguistic information as a given. This is a large challenge, and one which has not, to my knowledge, been confronted by SLA theorists. When simple, general learning mechanisms are offered as explanations for language learning, we have to ask how much of the simplicity and generality comes from setting aside the foundational issue faced by all theories of learning.

5.1 The Choices

It is now widely accepted among SLA researchers that natural, implicit processes, operating through input processing, are the heart of second language acquisition (e.g., Reference Lichtman and VanPattenLichtman & VanPatten, 2021; and the various chapters in Reference VanPatten, Keating and WulffVanPatten, Keating, & Wulff, 2020). The apparent implication is that these processes should be the central concern in language teaching. While traditional grammar-based instruction remains prominent in practice, it now receives little respect in the SLA literature. This shift is well illustrated by the fact that the most prominent advocate of such instruction, Robert DeKeyser, is now quite cautious in his statements about it (see DeKeyser, 2020). He accepts the existence and importance of implicit learning and places strong limits on the applicability of his own theory, which represents the underlying logic of traditional grammar-based instruction.

This does not mean that nothing is learned in traditional grammar teaching. As described earlier, the conceptual system has the ability to (effortfully) acquire knowledge (declarative knowledge, that is) on virtually any subject, including language. It can learn the meanings of words or symbols and memorize rules for combining them. In other words, it can treat linguistic symbols like other symbols (mathematical, for example). This is what traditional grammar teaching and learning are about. But the conceptual system’s lack of linguistic specialization greatly limits the extent of its success in any given area, notably in language. If we are satisfied with developing in students the most basic ability to say things in the language, slowly and effortfully (and often wrongly), this kind of learning can be worthwhile. But to go beyond this, we have to go beyond conceptual learning and focus on implicit development of the linguistic system.

So, input and the natural acquisition processes that rely on it are at the heart of second language learning. This recognition leaves us with two general approaches to the use of input in teaching:

1. 1. Trust the natural acquisition processes: facilitate their workings by providing appropriate sources of input, helping learners understand it, and making the experience as pleasurable as possible;

2. 2. Try to assist the natural processes through form-focused interventions: direct those processes to selected aspects of form in the input; select input based on forms it contains; explicitly teach form to support these interventions.

Both approaches take the operation of natural learning processes, using input, as essential for learning. The difference between them regards the potential for successful intervention in the processes. While (2) seeks to make learning more successful by intervening, (1) holds that it is best to simply encourage the conditions in which the natural processes will be most successful. This contrast is about the value of form focus: getting learners to pay attention to and deal with selected aspects of form.

In terms of the discussion in the previous sections, what we are talking about here is perceptual output representations and their role in learning (see Figure 2). In both approaches, when an input sentence “out there” is encountered, learners pay attention to it and become aware of it, which is to say aware of the perceptual output representation of the sentence. On the basis of this representation, phonological and syntactic representations are activated or constructed in the linguistic “place.” The conceptual system seeks to derive a meaning for the sentence, based on this chain of representations and its own contextual and world knowledge. This interpretive processing necessarily interacts with the linguistic processing, helping to activate or establish linguistic representations that are appropriate for the input. The overall process may require changes in the linguistic system, simply to make sense of the perceptual output representation. This is implicit learning.

Where the two pedagogical options differ, again, is in the possibility and desirability of intervening in this process. By “intervening” here, I mean trying to lead the implicit processes to make particular changes in the linguistic system, changes they might not otherwise make.Footnote ¹¹ Option (2) holds that we can and should do so, using what we know about grammar and about learners’ needs. This intervention can take several forms, to be discussed later. Pedagogical option (1), in contrast, says we should trust the natural implicit processes to make the appropriate changes in the linguistic system, simply giving them as many opportunities as possible to do so (providing extensive input), along with any support that the conceptual processing requires (helping learners understand the input).

It is important to avoid simplistic interpretations of the pedagogical options. Authors who favor input-oriented instruction (option 1) are not calling for a ban on all mention of form in the classroom (see for example, Reference Krashen and TerrellKrashen & Terrell, 1988; Reference WinitzWinitz, 2020). Nor are advocates of (2) simply rejecting such a ban; they are arguing that form should have a prominent role in instruction. The contrast is better seen in terms of default assumptions.

1. 1. Form focus is in general not a good thing to do and so should only be used in very limited ways; identifying special cases in which it is appropriate is then a matter for research and teacher reflection.Footnote ¹²

2. 2. Form focus is in general a good thing to do and so should be used extensively.

In the following sections I will consider these two approaches in turn.

5.2 Trusting the Natural Processes

If input, and comprehension of input, is the heart of learning, then a natural goal for instruction is to maximize learners’ exposure to the language, in ways that facilitate comprehension. This approach is thus focused on simply giving the natural processes the best opportunity to do what they do, without trying to alter what they do.

In the extreme case, trusting the natural processes can mean doing nothing more than telling learners “go read something” or “go listen to something” and leaving them on their own to carry out these instructions. While this type of assignment may have value in some contexts (I have used it as one part of class work), it is not the standard approach to input-oriented instruction. Such instruction will typically provide support, of two general types.

One type involves providing appropriate input or guiding learners in selecting their own sources. Crucially, judgments of appropriateness are not based on the particular grammatical forms or vocabulary items contained in the material but rather on comprehensibility and learners’ attitude toward the materials – dealing with the input should be an interesting and pleasurable experience. The other type is help with comprehension, using aids like pictures, graphs, tables, diagrams, and realia, as well as oral or written explanations and background information. Efforts to clarify meaning can involve form in limited ways, where they are deemed useful for understanding the input and do not significantly distract learners from its meaning. Whatever form the clarification takes, its goal is to establish good conceptual representations of the input, which can then contribute to construction of appropriate linguistic (phonological and syntactic) representations – and this is the heart of second language learning.

Earlier, I briefly described the major ways that input is provided in input-oriented instruction – reading, listening, viewing, and content instruction in the target language. Considerable research has sought to test their effects and has obtained favorable overall conclusions, though in some cases these conclusions are complicated by the inclusion of various interventions, including form focus.

5.3 Trying to Assist the Natural Processes: Input-Oriented Interventions

If we accept that second language acquisition results largely from the workings of natural learning mechanisms of some sort, as most of us now do, the question remains of whether we can intervene in those natural processes to make learning more successful. Perhaps teachers can use what we know about language and about teaching techniques to push the natural processes in directions that will be most helpful for the learners, using some type of form-focused instruction. The belief that we can do so is now overwhelmingly favored in the SLA literature.

The term “form focus” is commonly used in this context, to mean any instruction that deliberately focuses learners’ attention on formal features of the language. Such instruction appears in many varieties, only some of which are directly about input, and it is not always easy to separate the two types, in principle or in practice. In this section I will focus on those that are most clearly, explicitly tied to input. The following section will consider the question more broadly, asking if form-focused intervention in general is more effective than the alternative of simply letting the natural processes do their work through meaning-oriented input processing.

The prototypical examples of input-oriented intervention are input enhancement (Reference Sharwood SmithSharwood Smith, 1981) and input flood (Reference Rutherford and Sharwood SmithRutherford & Sharwood Smith, 1985; Reference Trahey and WhiteTrahey & White, 1993). These interventions are manipulations of perceptual output (see Figure 2), seeking to adjust representations at this level in ways that will alter the operations of the implicit learning processes, encouraging them to incorporate past tense endings in the interlanguage, for example. The conclusion from the research described earlier is that we do not have good reason to think that such interventions are effective.

On the theoretical side, Schmidt’s noticing has been offered as a possible basis for form-focused interventions. Such efforts to draw learners’ attention to a form in the input naturally encourage noticing of that form. In genuine applications of his proposal, this is best interpreted as pointing to particular aspects of the input, which can then be used by implicit processes. Schmidt was noncommittal regarding the nature of the implicit learning processes themselves, but they are perhaps most naturally associated with the implicit learning of Reference EllisN. Ellis (2005, Reference Ellis and Rebuschat2015). The idea then is that we can assist learning by encouraging learners to notice appropriate aspects of form in their input.

The (technical) concept of noticing is actually not very helpful here, though. Schmidt’s hypothesis is that learners need to notice appropriate items in the input, but apart from random examples he made no attempt to identify or characterize those items. Nor, to my knowledge, has anyone else taken up the challenge. Teachers who wish to use Schmidt’s ideas are thus left with pedagogical grammars, tradition, and intuition, none of which are useful guides to how natural implicit processes work or to the underlying nature of a functioning interlanguage.

The other problem with applying the concept of noticing is that while we do not know what the targets of noticing should be, we know a lot about what they do not include – all the rules and generalizations that make up pedagogical grammar. Those who try to use noticing as a general guide for teaching are likely to end up abandoning Schmidt’s concept (implicitly, as a general rule) and falling back on the loose notion of “awareness of something,” simply because sticking to the actual meaning of the term would mean dropping most of the topics that they want to include. This describes a number of proposals that have been offered for incorporating form focus in language instruction (e.g., Reference EllisR. Ellis, 1993, Reference Ellis1995; Reference Long, Robinson, Doughty and WilliamsLong & Robinson, 1998; Reference Nassaji and FotosNassaji & Fotos, 2004). These proposals cite Schmidt’s work as their foundation, but they do not in fact use Schmidt’s concept of noticing and are not applications of his Noticing Hypothesis.

Altogether, research on input-oriented interventions offers little reason to think that intervening in the natural processes is beneficial. The theoretical basis for such interventions, in the form of noticing, is less than compelling, to put it mildly. Again, though, research on the effects of form-focused intervention in input processing is not easy to separate from research on form-focused intervention in general. So we now turn to the broader question of whether form-focused intervention is in general superior to simply relying on the natural processes.

5.4 Trying to Assist the Natural Processes: Form-Focused Intervention

There is a near-consensus among SLA researchers that it is in fact superior and that we can and should intervene, quite extensively. The popularity of this view is expressed well by Reference Gass, Behney and PlonskyGass, Behney, and Plonsky (2020): “the debate over whether or not form-focused instruction is superior to input-only language teaching was largely settled [in favor of form focus] in the early-to-mid-1980s” (p. 484).Footnote ¹³ As a description of a social phenomenon, this is more or less accurate. But if it is taken as a claim that the research (then or now) has produced convincing support for the consensus, I suggest that skepticism is appropriate.

5.5 Conclusion

Second language learning, if it is to have any degree of success, largely depends on automatic, unconscious processes. These processes depend on input. Input should therefore be the main focus in language instruction. The implication is that the central issue for teaching is how this focus is to be realized in practice. Two general options are available. First, we can trust the natural processes, trying to create optimal input for them and to avoid interfering in their operations. Alternatively, we can try to enhance the workings of those processes, pointing them to significant aspects of the input and/or cultivating learners’ explicit knowledge of the target language in the hope that it will facilitate the implicit learning process.

The dominant view in SLA now is that the second approach is the right one; nearly everyone who writes on the subject maintains that we can and should intervene in the natural learning process. I have offered here a different perspective: For all the reasons presented in this section, skepticism is in order regarding proposals for form-focused interventions. We do not know what is likely to work, or if anything is likely to work, or if intervention will actually do more harm than good. In this context, the largely undisputed importance of input suggests that we should favor input-oriented approaches. At the very least, we need to recognize that this fundamental issue is still with us, and to adopt a serious attitude toward it.