Formal instruction and "doing useful stuff"

It is useful at a very early stage to start thinking about the role of formal instruction in the typical SLA journey of most learners. If one accepts the Meaningful Input hypothesis, this role becomes clear. The key is understanding that the “primary goal” of formal L2 instruction for learners is most often not what some highfalutin educational theorist might suggest - L2 competence - but rather the imminently practical - “getting an A”, or “keeping as far away from Spanish class as possible”. The naive assumption, while shared by most lay-people and an unfortunately large number of SLA scholars, that success in formal L2 instruction equates to L2 competence, would appear to be the cause here. ??? more here ???

Without getting into a full-blown critique of the role of mandatory education in modern society, it is certainly worth dwelling on a few of the “unspoken goals” or “unfortunate side-effects” of government-mandated schooling in most of the world today. One such unfortunate side-effect is that, by and large, it is a zero-sum game - there are winners and losers - and this results in a significant stratification of observed outcomes*. The negative reinforcement that ensues from consistently low marks turns many learners into consistent failures, whether or not these have any relationship with purported “aptitude” or “intelligence”. Another side-effect of standard syllabi and classroom environments is the presentation of a necessarily small range of content and development of a small range of skills. These almost invariably favour the dominant ethnic and social groups, and largely perpetuate the narratives of these groups around what is important to learn, and what successful learning looks like. Some commentators even view these as the “true” goal of standardised education, with actual “learning of things about the world” a mere side-effect (??? find a citation ???). Whether you are cynical or simply accepting of some negative side-effects of a “necessary evil”, these side-effects are well-known, well-defined and clearly measurable (??? cite something ???). Some of the skills and practices developed in today’s formal instructional environments are clearly useful in the modern world though, so it is an over-reaction to view them entirely in a negative light. For example, while it is obviously unnatural and unhealthy for humans to sit at a desk doing something they are not interested in for several hours without a break for a large proportion of most days of the year, this is clearly a very useful behaviour in many formal working environments (desk jobs), and what many view as “part of becoming educated/an adult”.

These side-effects are also critical in understanding the role and place of L2 instruction in most learners’ L2 journeys. The family life, support structures and content are almost certainly tailored to the dominant group in society (or at least those in power in the educational domain). This means that a good deal of formal L2 instruction is optimised for creating meaningfulness only for a subset of learners in any typical group of learners (particularly in the K12 environment, though universities and commercial environments also apply here as “that is where the money is”). While this may indeed be a “necessary evil”, it is not an “evil” that we should ignore in designing effective L2 instruction environments and curricula.

* Footnote ??? this obviously interplays delicately with social (class?) issues, but full discussion of this point will need to wait till later.

representationalism and

While Krashen appears to subscribe to some version of representationalism (??? get quote ???), at least insofar as that means there are “things” called “languages” that are “governed by rules” and these “rules” are “acquired in a certain order”, this is by no means a requirement of the Meaningful IO hypothsis (or his theories either). In fact, “doing useful stuff” is arguably a much simpler and more powerful explanatory tool. Learners typically acquire usage patterns in a particular order because they are typically exposed to them in Meaningful Input earlier or later. Due to learners’ differing proclivities (different types of formal/informal instructional activities, autonomous learning, native speaker interactions, etc.), these may vary somewhat but due to the way humans interact (certain forms are almost everywhere) and how most formal instruction is structured, there will likely be solid similarities between individuals. One doesn’t need to appeal to a Language Acquisition Device or “then Nature of langue X” for this - “that’s just usually how it happens” is as explanatory as one needs to be, at least for any practical SLA theory, instructional programme or technology.

What does Meaningful Input look like, and how do we define and operationalise "i+1"?

Because Meaningful Input necessarily needs to be emotionally/psychologically meaningful and useful for the individual learners, and individuals paths through the L2 competence development journey will vary greatly, there is necessarily no standard order. That is not to say there will be no communality between learners, that formal instruction can’t play a useful role, nor that certain “meaningless” or “learning” activities can’t be used effectively for speeding up actual acquisition. In fact, one useful way to think of word lists (SRS, etc.), grammar drills and other such activities is that they provide very useful extra-linguistic skills and knowledge in order for actual acquisition to take place. Just what activities are useful depends on a number of factors, mainly to do with the available sources of emotionally meaningful input and how comprehensible they are to a learner at a given point in their acquisition journey.

As Meaningful Input is necessarily personal, and individuals change over time, what is meaningful at a certain age is almost certain to be different at another. Krashen points out that L2 learning is essentially the same as L1 learning, and the ideal situation for comprehensible input is having a “language mother”. Basically a person who will tailor language “instruction” to the individual personally and to whom the individual is deeply emotionally and psychologically dependent. Typical human care-givers are constantly accompanying their care-giving (feeding, putting to bed, kissing grazed knees, etc.) with comprehensible input. Typical care-givers do this for multiple hours per day, and it takes the typical child multiple years to develop any sort of “linguistic competence” - we are often willing to coo and smile very proudly at a 2 year old who can barely scrape together a thousand words but a 13 (or 30) year old who hasn’t managed a thousand words with only a few hours per week after the same two years will likely be considered “bad at learning language” or “not very bright”. We are simply comparing apples and oranges. While it is theoretically possible to provide such an environment for adults who might find themselves in such an emotional relationship, such relationships are rare and it would generally be prohibitively expensive to try and emulate. We will see that it is also not needed.

As humans age and they are able to become more independent for basic survival activities, their “emotional centres” change and develop. The annoyed, sweaty serving lady dishing out a dollop of slop at the school canteen will hardly elicit the same emotional response to an adolescent as a primary care-giver spoon-feeding a hungry infant. Thankfully, it turns out that language is pretty much everywhere where humans are, so however an individual’s interests and passions develop, there will (at least for the major languages) be some kind of language-rich context or content that they would be interested/passionate about. Now it may well be that there is a quite large comprehension gap between what is readily available and the learner’s current competence level. While a dedicated human tutor with similar interests would be ideal for bridging this gap, that again is prohibitive in terms of cost for the vast majority of learners. This is where “meaningless” learning activities have traditionally added value, and technology is now on the verge of changing the situation markedly.

When humans derive meaning from any linguistically-rich context, they employ a variety of “non-linguistic” skills. Whether it be face-to-face buying a baguette in a Parisian boulangerie or reading a journal article on quantum mechanics, many “not strictly linguistic” skills are employed to derive sufficient meaning for the task at hand, particularly if it is to be done quickly and efficiently. These skills are used by natives and L2 learners alike, and it is precisely using these where Krashen claims that we find the i+1 - the level is slightly above what the learner would be able to manage without them but can get the overall meaning with them, and doing so in this context is where acquisition of new words, expressions and grammatical constructions actually happens. Not accidentally, the claim is that this is also where native speakers develop L1 competence, and it is here that we can find the direct parallels of Krashen’s i+1 to Vygotsky’s Zone of Proximal Development.

Let’s say that we have an early intermediate Chinese learner who is interested in science fiction (much like the author). If the learner doesn’t have any available reading texts that are simplified and have interesting content, then there are three cost-effective means for making most content comprehensible and consumable in a “reasonable” amount of time - simplifying the text, memorising translations of enough of the (right) unknown words to be able to get the meaning (and possibly doing some grammar drills of relevant constructions) before reading, enriching the text with comprehension aids, or some combination of the three. Because most language-rich content can be accompanied with textual representations, these principles can be extended to other content types, such as video, music, etc., allowing to cater for a very wide range of content preferences.

While text simplification is an interesting area, there are several reasons for it to be rather limited in practical scope. For one, it only works well with primarily textual forms (though does include comics and the like). If you start simplifying song lyrics or subtitles then it no longer becomes possible to follow the text and audio at the same time, and following text and audio at the same time is known to provide for robust acquisition over time (??? find citation ???). The text is also no longer “authentic”, and unless done by a human, the “translations” (to a simpler language) can suffer from inaccuracies. Simplified versions for L2 learners is definitely widely recognised to be an effective means for providing comprehensible input, and if there is content that a learner is interested in, it can be an excellent way of getting Meaningful Input.

Both text simplification and the remaining two means can benefit very deeply from something that is only tractable for large numbers of learners using digital means - personalisation. If a learner has already “acquired the word” “snout”, for example, there is no point simplifying it to “nose” in a text, and not doing so when not needed reduces the unwanted side-effects of simplification. What a learner really needs is simplification of, or help for, only the things they haven’t already acquired. But how do we know what they have acquired? Well part of the issue with language acquisition studies is precisely measuring when something has been “properly acquired” - it’s very hard to pinpoint. What is much easier to determine is whether the language element has been learned. If a particular dictionary definition/translation has been learned, then while the word might not have been “acquired”, it is definitely available to the learner as an “extra-linguistic” resource for helping comprehension, at least until forgotten. The same is true of expressions and grammatical forms. It can also be said for proactive glossing - there is no benefit adding a gloss for a word that the learner can remember an L1 definition for. If the learner gets help for all the “directly linguistic” aspects of a text they they haven’t learnt yet and for none of the aspects they have, then as long as the system can be updated in near realtime, the system can effectively be used to provide Meaningful Input. In addition, if the reading environment also allows for the learner to signal to the reading system that he no longer needs a word to be glossed (or simplified, etc.), then the system can also assume that a word has either been learnt or acquired (or both).

So i+1 is neither language nor learner independent - it is directly a consequence of the individual developmental history of the learner.

Sidebar - So i+1 is neither language nor learner independent - it is directly a consequence of the individual developmental history of the learner. With sufficiently well-designed, connected, “learning” tools and content consumption application environments, the system, with sufficient use, can get a very accurate picture of the current state of the learner, how they have developed, and can predict what the immediate future needs of the learner will be.

With such a learner model available, it is also possible to do many other useful things related to both language learning and language acquisition. While formal exams such as the HSK for Chinese, the DALF for French, or IELTS for English may be poor proxies for measuring L2 competence, they are a reality for a very large number of learners. Many learners have strong motivation for passing these exams and with an accurate learner model, a system could generate personalised help via drills, word lists or practice reading. Any (practice) testing for these exams within the system could also feed back in to inform the system about comprehension aids needed for Meaningful Input.

Learning vs Acquisition

Sidebar ??? - At the end of the day, it is somewhat irrelevant where you stand on the division of learning/acquired and its usefulness in actual language acquisition. Few would argue that those things that are “learnt” are remembered for very long unless they are practiced regularly (preferably in comprehensible input or output…), so whether you make a distinction between “learnt” and “acquired”, “learnt” and “durably learnt” or “learnt” and “fully internalised”, etc. doesn’t really matter. The timescales and investments that are necessary for developing advanced L2 competence measure in the thousands of hours and for most learners span multiple years (or even decades). If targeted, personalised “learning” provides “extra-linguistic” scaffolding for facilitating acquisition through expanded Meaningful Input availability or whether you think (extensive) content consumption helps “cement” language elements that have been learnt (and don’t distinguish learning/acquisition), both sides can agree on the practical steps to take.

However, some sorts of language learning or acquisition activity will likely be more or less efficient, and some may have little value.

Meaningful Input and the classroom

So if engagement or meaning is required for acquisition to take place and what is meaningful for a given group of individuals is different, how is it possible to allow for or provide Meaningful Input in the classroom with a diverse group?

Krashen points out that even traditional grammar classes can provide Meaningful Input (??? find citation ???), the reason being that if the instructor and textual material provide comprehensible input, and the learners find it meaningful, then it qualifies. As he further points out, this is only the case for certain learners, learners who are actually interested in formal grammar study. One observation that we might make is that possibly a significant majority of SLA researchers and L2 curricula designers fall into that category. Many (most?) are also likely to be members of the dominant social group, whose narrative and definition of “successful learning” happens to coincide with their own individual and group successes. They may have indeed received significant amounts of Meaningful Input during grammar classes and mistakenly conclude that the key ingrediants were a property of the material and instructional context alone, rather than a necessary combination of that and their own attitudes/motivations towards the classes.

Another difficult issue to tackle when attempting to pin down meaningfulness is that something that may not be meaningful in one context may become so in another. Situations are meaningful for inviduals for a vast number of reasons. Many (though alas never enough) individuals have had the experience where they have performed mediocrely in a particular subject and then one year, often due to a rapport (and sometimes sentimental crush) with a teacher, a transformation will occur. Suddenly the learner will find a passion for the subject and their marks will often, over time, increase significantly. This suggests that meaningfulness can appear under certain conditions. And if it can appear, it can also be created/fostered, and even designed for or engineered. “Good teachers” do this, sometimes unknowingly but often as a result of good ol’ fashioned hard work. Something may have meaning for an individual for cultural reasons, family reasons, sexual reasons, or any number of others. While some particularly empathic, hard-working teachers may be able to motivate most learners in a given instructional environment, it will rarely be all. Some teachers may not be considered particularly empathic or hard-working and still generate a significant motivational response in some learners - general predictions might be able to be made but for a given individual or group of individuals it will likely be impossible to predict what the outcome will be.

Input may also become meaningful for some learners due to external circumstances. Take, for example, an English class with content related to health. Under normal circumstances, there will be a certain percentage of learners that are interested in health matters (dieting, exercise/sports, budding doctors, health professional parents, etc.) and a certain percentage who, at least at a given time T, are not. What if an epidemic sweeps the globe and is on the front page of most newspapers every day (as it is as this is being written)? What if you personally have to take (extreme) precautionary measures in order not to be infected yourself? While it is likely impossible to predict with certainty whether this will translate to meaningfulness for a given individual, it is likely to for most. A year after the epidemic, when it is no longer front-page news, the same content will likely fade into the background and no longer have the meaningfulness it does at time T.

So there are essentially two ways for input to become Meaningful when considering a typical instructional environment. Comprehensible Input can be made emotionally meaningful, or emotionally meaningful input can be made comprehensible. The first can occur through blind luck (the lesson dealing with health can happen during an epidemic) or skill/charisma on the part of the instructor in making content meaningful. For the second, the instructor can train the students to be more skilled in extracting comprehension from material that is already psychologically meaningful to them. It is in this second area where new technologies can provide a massive boost to the learning tools that are currently available. A successful instructional environment will very likely do both of these, as they are complementary rather than mutually exclusive.

Meaningful Input in society

Educational researchers have long talked about the importance of creating instructional environments that are low in stress, and this has resulted in significant evolutions over the past decades in how instructional programmes are designed and tested (??? get citation ???). A consequential part of this attempts to reduce the anxiety level, both during learning and assessment activities, and much research suggests that this can provide for not only improved learning, but more accurate assessment. Krashen’s Affective/Motivation hypothesis can be seen as a clear expression of this.

While there are large sections of the public (lay-people, government and even academics) that will often rail against such measures, claiming that it has gone too far and is “not preparing learners for the real world”, there is no necessary explanatory gap. Whether it be L2 or other forms of acquisition, a reduction of performance on tests focussed on learnt, conscious knowledge is indeed possible, even if usable competence in the subject domain has increased. It may also be an implicit acceptance of the stratifying, standardising role of traditional instructional environments - if learners are not skilled in sitting for long periods of time performing emotionally meaningless tasks, then many of our current forms of work will struggle to find suitable applicants. These are real considerations but fall well outside the domain of influence of L2 theorising and practical application, even though they can strongly control the observed outcomes.

Thankfully for L2 instruction and acquisition more generally, a lot of acquisition will usually need to take place outside the typical instruction environment - L2 acquisition takes far more time than is available within typical classroom environments. This means that the (social, cultural, economic, etc.) constraints that may inhibit Meaningful Input for some learners in a given instructional context need not mean they are unable get any at all. The internet and the availability of virtually all content in a digital form at the tap of a fingertip on a touchscreen means that it is now far easier than ever before for most learners to access L2 content that is emotionally meaningful to them.

Meaningful Input and autonomy

As devices, and communication technologies more generally, have made available virtually all content at a reasonable price*, the importance of institution/government-run/financed media centres providing L2 material has also diminished. This is clearly of net benefit for many learners, whose content preferences were never catered for there, just as they never will be in formal instructional environments. What this allows for is a restructuring of formal L2 instructional environments to focus on developing the skills for to learners to be able to access Meaningful Input themselves, particularly at the post-beginner level. This also suggests that focus of technologies developed specifically for L2 acquisition should also be directed at this goal.

  • While some may argue that many learners can not access all their preferred content legally due to their financial situation, it is also a fact that many means exist to obtain content “for free” illegally or in murky legal territory. Whatever your opinion on the morality of this issue (whether it should cost so much or whether consuming it illegally is immoral), it is a fact of life, just like the way most nations have organised their formal instructional environments.

While the technologies talked about here could theoretically be combined with other digital mass education (online courses, digital tutors, etc.) to provide for a complete programme of learning without physical presence in a classroom, there is also clear value in traditional, instructor-led environments, particularly if the technologies and in-person, formal environments interact and rely on each others’ strengths. Because it can function effectively independently, the technologies can be adopted or integrated whenever the formal instructional context is ready (teacher-training, curricula design, government approval, etc.).

The role of Meaningful Output.

Output is little more than an effective means for soliciting comprehensible input within Krashen’s framework. If we ignore long-form, “projection” writing/speech for the moment (but see “the 4 skills” section below), and restrict ourselves to verbal and written interaction, then that also holds for the Meaningful IO hypothesis. However, we should note that in many situations, particulary spoken, we aren’t just talking about any old kind of Meaningful Input, we are talking about deeply Meaningful Input, provided it is for genuine communicative purposes, and not simply reading out an unrealistic, canned conversation from a textbook.

Humans naturally accomodate (see the discussion of primary care-givers above), and speakers with any level of experience interacting with non-natives will quickly adapt their language to an appropriate level. What can be a little frustrating for learners is that it may take quite a bit of interaction before a real interlocutor settles on a reasonably optimal level. Here accent and grammatical production accuracy can be crucial in making sure the interlocutor doesn’t over-compensate. The converse is also true (as the author experiences whenever he tries to speak German) - a stock expression (so with accurate grammar) spoken with an excellent accent can result in a deluge of fast and incomprehensible language that results in total communication failure.

One potentially useful area for some learners who find producing some speech sounds in a classroom context is precisely by letting them know consciously where they have difficulties, and helping them to be able pronounce certain sounds. There a few more frustrating (and embarassing) moments for a language learner when they have finally built up the courage to speak, they think they’ve “got it right” and a pronunciation issue means their interlocutor can’t understand. A useful activity for a given individual might well be simply practicing such sounds, to try and develop their ability to produce comprehensible output.

Linguistic accommodation

Sidebar - Linguistic accommodation - we can also note that the language-level negotiation phenomenon looks extremely like accommodation from the socio-linguistics field. Members of different varieties of a single L1 usually negotiate on a particular accent and (sub-)dialect - if they can’t - or refuse to - then communication usually breaks down very quickly.

Real-time communication (to which we can include text messanging) is necessarily deeply Meaningful - if a learner is engaged in communication then they are doing it for a very real reason (even if that is just “verbal grooming”) and live human interaction has very clear psychological (and physiological) correlates. When the input is also adapted to be comprehensible, this is clearly going to provide significant language acquisition value.

Output without Meaning, however, particularly in the context of a stressful classroom performance, is also predicted to provide little or even negative value, just as it does for Krashen.

The 4 skills - or what is language really?

The “4 skills” are regularly taught separately, often as separate skills or even in explicit classes. It is worth considering whether the Meaningful Input hypothesis has anything to say here.

To begin with, however, we must take a closer look at what we really mean by “language” - are there really 4 skills to it? What about 200 years ago, were there 4 skills then? What about 10000 years ago, before writing was invented? While the precise dates are argued about when “painting/drawing/marking” became “writing” (or rather we added writing to these), no one seriously argues that “human language” didn’t exist before writing. While some (many?) might argue that someone who can speak and hear fine but can’t read or write is “language deficient” or “language defective” in a modern society, no one would have done so before writing was invented. Now if it was just a matter of magical moment in history where some happy thinker had a eureka moment and with a “simple 15 minute trick that scribes don’t want you to know” any (normal) human could simply pick up a pen and the words would flow a river, then there would be little further to say on the matter. It turns out, however, that reading and writing are neither automatic, nor easy, even in your native language. Most learners take several years to be able to read and write basic texts and take decades to develop “full competence”. As you might expect for something that isn’t natural and takes many years to learn, individuals attain markedly different levels of “competence”. Recently, studies of the brain have started to show that in fact the brain is significantly reorganised by high levels of literacy. The brain’s actual physical structure changes (as it does with other similar long-practiced activities, like “The Knowledge” for London’s taxi drivers, etc.). Just how “natural” that is is a matter for debate. Because writing is not simply “speaking with your hands” (as deaf children who can sign perfectly well but still struggle to learn to read and write like their hearing fellows do), the relationship between speech and writing is far from clear.

What is clear about the relationship between reading and writing is that unless you are competent in doing them in at least one of your highly competent LXes, doing them in your L2 is not going to be easy. As others have pointed out elsewhere (??? find quote, maybe from Nation somewhere???), in some contexts learners may not have highly developed reading/writing skills in any language and so finding Meaningful Input for reading is going to be a challenge, and the most successful strategy for helping learners with L2 reading starts by helping them with reading in L1. “That’s not my job” some teachers will inevitably complain but the fact remains that if they want their learners to achieve better outcomes, that is what they must do. With problem readers, there may also be a negative affective filter associated with the reading process/activity that further increases the bar for finding Meaningful Input for learners. These challenges must not be overlooked.

Following on from the physical and developmental differences that are introduced by the different modalities of reading/writing and speaking/listening, we find some quite distinct lexical and grammatical differences. Many corpora studies (??? find some???) show marked differences in the frequencies of both lexical and grammatical items, and also the kinds of things that are typically written about. While there are a great number of different kinds of writing (text messaging, academic essays, advertising, etc.) and indeed kinds of speaking (speeches, baby talk, two-way radio, gossiping in person, etc.), not all people will participate in all kinds of either. Attempting to teach or force a learner to participate in a kind of language activity that they are not familiar or comfortable with in the L1 may well lack meaningfulness and raise a negative affective filter.

One of the interesting corollaries of the emergence of writing and the different affordances the “memoryless information retention” allows is that, at least in a largely literate society, it will change the nature of speech of the members, even for those who are illiterate. In a predominantly literate society, literacy can come to pervade almost all aspect of life, and with it all aspects of language. People’s reflexes and even intuitions about language (grammaticality, wordiness, etc.) are changed at the physical (brain organisation), personal (“speaking like a book”) and cultural (recounting of long stories, songs). In very real ways one can claim that literacy, or at least its consequences, has now become a fundamental part of the basic linguistic activities of all members of such societies. This is one of the ways in which the fundamental linguistic principle of the “primacy of speech” is simply wrong for many societies - reading and writing have become so intertwined with the normal verbal cultural practices that someone learning an L2 will only be able to participate in a very narrow range of activities if they are not able to at least read and probably write in the L2.

"L2" competence - what, exactly, is that

Whatever definition we give to “the X language”, it is clear that there can never be a single individual who “knows it all”, whether we restrict it to the verbal or not. In reality, even a monoglot individual will be more or less proficient in a number of registers or varieties of their L1. An individual’s accent, choice of words and grammatical constructions and even topics of conversation change markedly depending on the participants (and this in addition to the differences between spoken/written mentioned above) - sometimes only in frequency but often also in what words are produced and the associated order and morphology. This is normal L1 behaviour. In reality, those who attain “native-like” fluency in an L2 will also do this. It is a simple fact that if one “talks like a native speaker” then one doesn’t speak the same way when watching a football match with a group of males as they do when they are reciting a poem on stage, or whispering sweet nothings into their beloved’s ear. While many may not be fully conscious of how they change their language between these different situations, they often have at least some appreciation of it. What’s more, everybody knows it is meaningful - “speaking like a professor” at the pub in front of a football match will typically be interpreted by other participants as wanting to create and maintain significant social distance, and doing so is quite likely to be one of the overt manifestations of this when an individual is trying to do this, whether they are aware they are doing it or not. Furthermore, the various registers will often differ, sometimes greatly and even in writting, between regions of a “single language”. Take English for example - can you teach “English”? If so, who speaks it? Alas, there are no “English” speakers, only “British English” or “American English” speakers, etc. Further scrutiny uncovers that there are no more “British English” speakers than there are “English” speakers - a Londoner will regularly speak differently to a Scouse. Even if they are capable, at least if they are instructed to do so specifically, of writing prose or in an academic style that differs very little, much of what they actually produce will be noticeably different.

Identity and Meaning

In addition to things such as hair styles, clothes, mannerisms and other means, a significant channel for expression of identity and group membership is done through linguistic “choices” (conscious or not). In addition to a Londoner sounding different from a Scouse, a working class East Londoner will often sound very different to one of the landed gentry one might come across in Kensington. A Londoner whose ancestors came from West Indies (originally of African origin) might sound very different from a caucasian a couple of street over. The differences are measurable and usually noticeable to all. These linguistic choices are part of our identities, and are necessarily Meaningful. As an L2 learner, the input they receive will be imbued with this meaning and, if they are successful, register variation will become part of their repertoire. Part of how successful they will be at acquiring an L2 will necessarily involve navigating this, and whether they are able to get sufficient Meaningful Input. If we accept that affective/psychological meaning is an integral part of Meaningful Input, that means learners must be able to get enough input from “their people”. Obviously that doesn’t necessarily mean “people that look like them” (I will never look Chinese, for example), though that can play a powerfully motivating role for many. One of the wonderful things about learning another language is the fact that one can indeed use the opportunity to develop and experiment with another identity. As with all such matters, however, teachers trying to steamroll learners to adopt an identity similar to that of the teacher will usually fail miserably, unless the teacher is particularly charismatic. Gender, appearance and many other factors come into play in identity construction and development, and all of these necessarily play a part in the construction of an L2 identity and so consequently what will qualify as Meaningful Input for any given individual.

So what L2 registers are learners supposed to acquire? Just

If we accept that the language patterns of the dominant social group will naturally take the lion’s share of attention in any formal L2 instruction, and that this might make developing L2 competence considerably harder for some learners who do not identify with such groups, should we allow learners to chose their own registers and modalities? While that is one solution for fully autonomous learning, it is likely not only impossible but also undesirable for much government-mandated/sanctioned, formal instruction. The reason is that, as alluded to above, much formal instruction has an associated goal of formatting learners to a certain set of behaviours and setting out what is “acceptable” or “proper”. This is largely inescapable, at least within the confines of how we currently construct society in most of the world.

If we think about this in more abstract terms, the problem of what to learn/acquire and test may become clearer. If our goal is to teach L2 competence, and by L2 competence we mean “those elements of linguistic behaviour that are demonstrated by native speakers of the L2 of normal neurological development”, then we are faced with an interesting issue - any native speaker should necessarily score 100% on a L2 competence test. If even a single one doesn’t, it is not the native speaker who is defective, it is the test.

The Meaningful IO hypothesis and ultimate attainment

One common topic talked about in SLA literature, particularly in relation to assessment, is whether there is any ultimate ceiling that a learner will be able get over. This is often broached in discussions of what age L2 learning/acquisition/instruction should begin but is also discussed in relation to the notions of “aptitude”, “intelligence” and “quality of instruction” (??? cite ???). Krashen claims that many learners may be capable of producing more native-like speech (and writing?) than they are typically observed producing (??? cite ???) and suggests that the affective filter may come into play here (??? cite). When we look at findings and discussions from a wide range of sources (L2 motivation theory, ???, find some other source, maybe psychology, etc.) in the context of the Meaningful IO hypothesis, the reasons become clear, particularly in relation to Meaningful Output.

As we have seen, meaning (utility) and identity are everywhere in language. As we have also seen, accent and register are used by everyone to assert identity. Output needs to comprehensible and efficient - but it doesn’t need to be indistinguishable from a native speaker in order to be both of those. The Meaningful IO hypothesis therefore predicts that there will indeed likely be a ceiling at the point where the learner - or rather the actual individual human being who needs to interact in a real language environment - arrives at a point where they are able to achieve their interactive (comprehension) goals and their emotional goals (express their identity). The vast majority of native speakers “have an accent”, at least for some of their registers, and they use this to assert membership of (a) particular social group(s). L2 speakers are in exactly the same position - having an accent (that doesn’t happen to correspond to any “native” accent) will identify them as a member of a particular social group - “foreigners” or “foreigners from X (country)”. In fact, not having an accent may ostracise them from other members of their community. As such, if the individual has no (possibly deep) psychological imperative to be identified as a member of a particular native-speaker group, then their accuracy will likely not improve above a certain level, so they can continue to be reliably identified as “a foreigner”.

While only anecdotal, the author has discussed this issue with a number of “perfect” L2 speakers and almost all are conscious of these issues, and the relationship to the expression of their identities. The author, a native speaker of NZ English, has been confused as a native speaker of both Russian (started learning at 16 years of age) and French (started at 19 years of age) by educated, native speaker adults of both speech communities. Like the author, most post-adolescent learners who attain native-confusable levels of an L2 appear to be deeply motivated to “be considered a native X”. It is something which is pervasive in their interactions with native speakers (“at the back of their minds”), and something that they are definitely aware of. In spite of any putative “aptitude” they may be considered to have, almost all the author has discussed this with were convinced that their near-native level was not the result of any “God-given talent” but rather the result of copious amounts of good ol’ hard work. They consciously monitored their ability to be confused as a native speaker and worked at sounding like one until they achieved it (at least for some people some of the time). It is also relatively easy to perform a lay-psychoanalysis of the reasons - attaining near-native fluency is regarded as a significant intellectual endeavour and will result in large amounts of envy, praise and positive comments about intellectual capabilities.

Furthermore, the relative importance of age also becomes perfectly clear. Many commentators ascribe the relative ease with which younger learners attain native-equivalence and how this “becomes harder with age” (??? cite ???) or is even “impossible after adolescence” (??? cite ???). This is often attributed to changes in the brain, with some proposing hard limits at puberty (??? cite ???). Immigrant learners are often studied and used as “proof” of these theories. Alas, what these theories often completely ignore is the importance of age at different stages of identity construction, and the differing pressures that immigrant learners can come under at these different ages. As most people reading are probably aware, many traditional schooling environments can be particularly brutal in terms of social pressures. Young immigrants will regularly be ridiculed if they “can’t speak properly” and as such, have a deeply meaningful reason to sound like their peers. If the expectation is that the learner will stay in their adopted community, then there are deep psychological reasons to be considered a “proper” member of the group, “one of them” and not a “foreigner”. Many aspects of identity may have crystalised by puberty, and particularly if a given individual has a series of difficult experiences around their integration, they may settle on membership of their native community (as an “X” or as an “immigrant from X”) as an important part of their identity. Note, however, that this is not to say that every equally motivated (whatever that might actually correspond to in a “scientific” sense) will achieve the same level of native-ness at the same speed, even taking into account equal exposure, support structures, etc. It is also not to say that certain behavioural skills (imitation, memory, etc.) won’t make things easier for some than others. What is does predict is that, given enough time, any sufficiently normal individual can eventually attain native-like fluency if doing so is an important part of their identity.

All these phenomena are predicted by the Meaningful IO hypothesis. Learners will cease to “improve” when they have reached their comprehensibility and “nativeness” (including for writing, as needs dictate) goals because the particular form of their output contributes to its meaningfulness.

Implicit bias in theories and in technologies - allowing for Meaning for those not like us.

A common complaint levelled at some technologies, and recently particularly those involving “AI” algorithms and technologies (machine/deep learning, etc.), is the presence of implicit biases. The same can clearly be levelled a educational theorists and practicioners. As already mentioned above, these biases have clear effects often in favour of a particular social and/or ethnic group. One advantage of the Meaningful Input hypothesis is that, while unable to change society or remove implicit and explicit biases from outside the SLA domain, it clearly lays out a set of parameters that enable bias to be identified within it, and so they can be addressed by the stakeholders, if they decide to. One of the realities of “real life” in complex societies is that completely removing bias is often impossible, and too much of a focus on reducing it can handicap a system to a point where it is no longer useful to anyone. Not identifying the biases however, can lead to results that aren’t obvious and a search for improvements where none can be found.

Via personalisation, some of these biases can be mitigated, at least to a certain extent. By allowing learners to have a great degree of control over the input they receive, particularly for content, they can develop their competence in directions that are most meaningful to them. For many learners, this will not be the typical “polite and obediant member of the dominant social group” types of behaviour that fill most language learning textbooks. However, the tools available for enabling personalisation need to be able to personalise over a wide variety of registers (and dialects), and as a casual look at the auto-generated subtitles of any Youtube video with a non-standard-American accent (for English) can attest, sometimes these tools fail miserably. While more work is certainly needed on the base technologies (electronic dictionaries, NLP platforms, etc.) in this area, part of the solution is simply acknowledging that there is a problem to all the stakeholders and accepting that when outcomes are poor, this may be a contributing factor. If learners know that the tools may not be optimised for the kinds of language they are attempting to get Meaningful Input from and why this is so, this information can be added to their repertoire of extra-linguistic and affective information.

How do we operationalise "meaningfulness"? Can it be measured? How much is "enough"? How do we measure its benefit?

One of the tenets of Krashen’s work is that you can’t force L2 output. In the right environment, it will emerge naturally to allow learners to participate in meaningful activity when they are ready to do so.

One of the consequences of Krashen’s work, and particularly the distinction between learning and acquiring, is that measuring acquisition is necessarily going to be difficult. A significant proportion of language testing tests “learnt” language, or skills that don’t rely exclusively on “L2 competence” per se (see ??? above). While the theory predicts that learnt material can be used to facilitate acquisition, measuring learnt material is not the same as measuring acquired material.

Meaningfulness and memory

Memory is a overarching term that is applied to a number of aspects of the learning process (short-term vs long-term, “muscle memory”, episodic memory, etc.). Exactly how these various types of memory relate to the learning vs acquisition dichotomy will require further elucidation but we can make a few observations that suggest a significant role for meaningfulness strongly correlating to memory effectivenss.

Scholars and teachers have been thinking about the importance of memory and how to “develop a good one” since ancient times. Several ancient Greeks talked about the “method of loci” in various texts (??? get citations from wikipedia and several other similar techniques are talked about in the literature. While it might be charged that such memorisation of lists or facts is aligned with learning rather than acquisition, the power of association with deeply familiar contexts indicates that at least some types of information are more easily retained in this way.

Meaningfulness and aptitude (and how real is "aptitude")

A common lay-conception about L2 competence is that there are fundamental differences in humans, and however competent a given individual may be in their L1, some people just “aren’t built for second language learning”. Many scholars also include a notion of aptitude in their theories (??? get citation???) though some, such as Krashen, question how predictive aptitude is of the ultimate level of L2 competence a given learner will eventually acquire (??? get Krashen quote).

TODO: - Look at what aptitude really is - can it be developed or is it innate? - Is it really separate from meaningfulness (or at least an ability to find lots of meaningful input)?

Predictions of the theory.
  • It is possible to “learn” language elements, and get usually high marks on language tests, without acquiring them.
  • Learning language elements will make getting Meaningful Input easier but will not be sufficient for acquisition.
  • Acquiring language elements makes learning them much easier, though learning still needs to be done explicitly. (a native speaker will not necessarily get 100% on a language exam but should be able to perform very well if they have the requisite reading/writing skills and receive a minimun of training).
A programme for bringing Meaningful L2 input to the masses - bringing it all together.

So what would a realistic, successful L2 acquisition context look like? First of all, it can’t require massive, fundamental changes to our education systems. Reform in government-mandated programmes can be painfully slow at the best of times but none of the arguments presented above require a “revolution” in terms of how we structure our L2 classrooms. It should also not require any fundamentally new, specific, complicated skills on the side of either teachers or learners. It should be able to piggy-back on the devices, network access and subscriptions that “normal people” already have. Ideally, it also won’t require any substantial investment outside of the existing instructional context, and not require any new or unproven technologies.

Fortunately, there is good reason to believe that all these conditions are met, at least for the L2 languages that are the focus of most learning today. In fact, there is reason to believe that a significant problem to be solved has already been solved - how learners discover new content that is (or will likely be) meaningful to them - recommendation. Almost all successful content platforms have some form of content recommendation system, at least for native speakers. These recommendation engines are often black boxes but are also often very successful at identifying content that will appeal to a given individual. While there may be beneficial to integrate linguistic considerations into the existing recommendation engines, it is definitely not a necessity, nor even necessarily desirable. They are successful precisely because they are usually successful in finding the content that people are most passionate about, not because they attempt to optimise on some other dimension.

Now a given learner may not be able be able to comprehend a given piece of content that a recommendation engine suggests in its “raw” form but we have already indicated several means (simplification, pre-training and personalised input enhancement) for turning authentic content into comprehensible input. Take a typical content search in a search engine. A learner will search for certain keywords or with a question or phrase. The results will be presented in a search results interface belonging to the application or platform in question (think a web browser or your netflix application). Modern, well-designed applications are almost all engineered modularly, and either allow for plugins/modules to be added or already have all major funcationality implemented as a plugin/module. If a plugin application has access to these results, the application can directly enrich the interface with information relevant to learner-user. If the plugin also has access to an accurate model of the learner’s competence (or at least what they have “learned”) then it can also make accurate predictions about how comprehensible a given piece of content will be, what language elements could usefully be learnt before tackling a particular piece of content and what comprehension aids are likely necessary when actually attempting to consume it.

Because the primary focus is on finding emotionally/psychologically meaningful content and we have means for making this comprehensible, there is no need to pre-analyse large corpora or modify existing recommendation engines or interfaces. When the results come in analysis of the target content can start (in parallel) and then update the interface with information such as comprehensibility scores. When a given content item is chosen, the plugin will be able to enhance most input with comprehension aids in place, leveraging the investments already made by the content provider in user interface design and technologies.

Enhancing Meaningful Input with aids for learning

As mentioned previously, when language elements have been acquired they are almost certainly more easily consciously learnt. Given that the same mechanisms for helping in situ comprehension can be tweaked to also facilitate and promote conscious learning, the system can also help out here. Let’s say a learner has a language exam in the coming weeks and the exam requires the conscious knowledge (memorising, etc.) of certain lexical items and grammar patterns. Because the system has already analysed the content and compared it to the existing learner model, it is trivial to add “expected benefit” scores associated with a given piece of new content. Various “benefits” are easy to imagine, such as the presence/quantity of elements included in a given course of instruction (a set of language learning materials from a given publisher used in a particular formal instructional context) at the current level of advancement through the course a learner is currently taking.

The Meaningful IO hypothesis and Natural Order

Because the Meaningful Input hypothesis rejects the utility of the abstract idea of a single, definable object of study - a fixed code that exists outside of the real languaging activities of real human beings, e.g., “English” or “French” or “Chinese” - in favour of a more realistic mosaic of (sub-)dialects, registers, and language mashups that are at the disposal of a language learner (or native speaker for that matter) for interaction and self-expression, the idea of a Natural Order to the acquisition of a fixed set of Rules that belong to The Language loses its meaning. Given a certain social, political and instructional context, there will often be a rough ordering that emerges for most learners in “the real world”. There is no necessity to this however, and it is a function only of the frequency with which the learner encounters particular language elements in comprehensible input and the importance they bring the expression of psychological/emotional meaning - how Meaningful they are. Because the Meaningful IO hypothesis agrees with Krashen in that the explicit teaching of (particularly) grammar has relatively little interest for most and no interest for others for developing L2 competence, this doesn’t represent a major point of divergence in the practical implementation of instructional programmes or technologies between the two theories.

Engineering Meaningful IO

When we step back and consider the real purpose of research into SLA, it behoves us to make explicit what the true goals are. When we then turn to designing instructional programmes and software to help us in achieving these goals, we need to take into account the current state of affairs and how what we are proposing will change this current state. A programme or software that may be optimal given a certain (abstract) set of circumstances is totally useless if, in reality, that certain set of circumstances only exists in the minds of academics that concoct them and not in the real world. It is also necessary to determine (or at least analyse in depth) whether the programmes we devise will be acceptable in terms of their associated “cost” - both monetary and in terms of any side-effects that the changes might bring about.

The “true goal” of the Meaningful IO programme is to provide a (hopefully significant) increase in L2 competence in terms of both efficiency and ultimate attainment level for the majority of L2 learners.

There are a number of requirements that must also be met for it to be implementable in the real world: - it should not reduce the efficiency or ultimate attainment level for most learners, and if it does for a particular group or groups then those groups can not be the most vulnerable/underprivileged (for fairness reasons) or the most privileged (politically impossible). Ideally it will raise the level significantly for those who are currently best catered to and reduce the import of the advantages that group current has, increasing the levels for the others further still. - it should not need significant investment by either education providers or individuals - it should not require qualities of the instructors or learners that today’s instructors or learners don’t possess

Very useful properties for the programme to have (increasing likelihood of adoption): - increase the scores using today’s current assessment techniques (that focus to a greater or lesser extent on conscous learning).

Meaningful IO, the human body and formal instruction/autonomous learning

In addition to the points raised elsewhere in our discussion, education research has turned up a number of very important factors in predicting educational outcomes more generally. Let’s look at a few of the typical factors that are important for education/learning regarding learners as human beings, and how we might engineer for better outcomes given these constraints.

When we talk about Affective Filters or emotional/psychological meaningfulness, we have already noted that learners’ filters can be modified through external events (an epidemic, etc.) or through a combination of external and (psychological) internal (a quiet person in a stressful classroom/assessment environment, etc.). However, physical factors can also play a very significant role. For example, a learner who has trouble seeing what is written on the blackboard may have to concentrate much harder on actually seeing what is written and have less mental energy to devote to conscious learning or understanding/acquisition. Some learners (for many different reasons) may be more receptive at the beginning of a 40-minute instruction session than at the en of a 2-hour one.

Specifically in terms of the Affective filter, what might be both comprehensible and psychologically meaningful for a given learner L in a given instructional context IC at a specific time T were they to have had a nutritious breakfast may be neither if they have not. If the learner is unable to concentrate properly because they have low blood sugar and feel hungry then it doesn’t matter how good a teacher is or how passionate they are about the content, learning outcomes will suffer. While this issue has been widely studied and some governments attempt to address it, it is unrealistic to think it can ever be solved completely (or with an “acceptable” cost to everyone in a given context). The fact remains that if a learner who never gets a nutritious breakfast is getting the majority of their opportunities for Meaningful IO at 10am on Monday and Thursday mornings, their outcomes are very likely to be poor*. A similar point might be made about exercise (blood flow to the brain), either at a given time T (straight after a break and a long walk between classrooms) or as a more general condition (physical fitness better

* Here is one area where the dominant social group typically has an advantage in most cultures - care-givers are typically far more concerned with formal educational achievement and often ensure that learners under their care are properly fed.

So how do we engineer around this?


TODO: add sections dealing with how the theory differs or includes each of Krashen’s hypotheses

Krashen’s theory of second language acquisition consists of five main hypotheses:

the Acquisition-Learning hypothesis;

??? meta-knowledge vs interactional skill - knowing how to describe vs knowing how to do. Different kinds of knowledge? Different kinds of intelligence? Comparison with sports? Best theoretician on kicking a rugby ball vs best goal scorer. Comparison with L1 writing? Best literary critic vs best author.

The nature of rules and “meta-knowledge” - language acquisition as internalisation of a set of “rules” inherent to “a language” or as a interactional skill in “doing real stuff” that relies on context-specific, negotiated patterns? Integrational (non-representationalist, constructivist enactivism) vs representational views of language.

Meaningful IO and the "science" of language acquisition

Models in science - but Linguistics is a SCIENCE!!!

Linguistists, long maligned as merely members of the humanities, have been very proactive in proclaiming the “scientific” nature of their endeavours. Case studies are frowned upon and many books have been written about how to “do proper linguistic science”, with the appropriate schooling of budding linguists in proper experiment design and how results need to be analysed and presented. While one can only salute this general tendency, it appears that many linguists have not bothered to think through the consequences of this very deeply. It is certainly true that many a theoretical physicist can also be guilty of confusing a particular theory or model with The Truth About The Universe, any moderate reflection on the history of science suggests that is premature to say the least. As George Fox puts it - All models are wrong but some are useful. If, as most linguists would claim, we are “doing science”, then what we are doing is proposing models of the world that correspond to a particular theory. The models can be tested in a wide variety of real situations (experiments), and if we find significant amounts of counter-evidence, then a given model is typically thrown out because it isn’t fit-for-purpose - it doesn’t help us make useful predictions (“do useful stuff”).

So models are “convenient fictions” that help us think about a particular phenomenon in a principled and useful way - useful for making predictions, useful for doing real things. A “good” model will also likely be very fertile, and will let us do useful things in places we didn’t first expect it.

So are scientific models really The Truth About The Universe? Or are “All models are wrong but some are useful”. - So there are not “really, truly” context-independent “languages” with “fixed elements” (“phonemes”, “words”, “rules”) that are “acquired” for “transmitting ideas” but this way of looking at language can be useful for some things. This way of looking at the world is particularly effective for creating a relatively homogeneous community of practice over a large distances, large time-scales and with large numbers of participants. Such communities are critical for successful nation-building and cross-community interactions, and many other organisational needs for modern society. - There are other ways of looking at it (metaphors, parallels, etc.) that are useful in other ways. So what do we really want to test for? Conscious learning or acquisition?

If conscious learning is not necessary for L1, then why necessary for L2?

Sidebar - If conscious learning is not necessary for L1, then why for L2? Many L1 learners in modern English language environments no longer have formal grammar instruction in their L1 yet are still able to produce “beautiful” and “grammatical” speech. This is noted by Krashen, and learners appeal to “not feeling right” rather than a strict “it’s not grammatical because of rule X”. If conscious learning is not necessary for L1, then by what magic must it be necessary for L2? ??? what do I want to say here?

the Monitor hypothesis;

Mainly agree, for production the same and largely the same for situations where the learner has enough time

Differences: - consciously learned rules (and definitions) are useful for more than just production, they can also be used as “extra-linguistic” information for analysis when the learner has issues in comprehension

Predictions: Can be a useful tool for situations where there is enough time - i.e., (low-pressure?) reading but is neither necessary nor sufficient (“just another tool in the learner’s toolbox”)

the Input hypothesis;


Affective Filter hypothesis;

The Affective filter tends to apply mainly to formal instructional environments, and would predict that high-stress or traumatic environments are not going to be propicious for language development.

the Natural Order hypothesis.

Disagree (??? do I, re-read his stuff!!!) - there will be little more than general tendencies, there are no such things as languages

ZPD and i+1

Sidebar - As some authors have pointed out (??? find ref for ZPD and i+1 dissimilarity), it appears to share much common ground with Vygotsky’s notion of ZPD and is sometimes even considered equivalent. The term is not rigorously defined by Krashen however, and as such has been appropriated by many different scholars from markedly different theoretical viewpoints.

TODO: Sidebar - make more explicit just what "i+1" is

Sidebar - One consequence is to make more explicit just what “i+1” is, how to provide for it in real learning environments and how this fits in to formal instruction. Another consequence is a clear roadmap for the kinds of technologies that will help learners develop L2 competence, and what kinds of technologies will not, or only do so indirectly and ineffeciently.


Goals of instruction for post-adolescents - explain the following: - Describe the true nature of standard language and prestige varieties - the politics of standard varieties and group dominance - explain that virtually all native-speakers habitually use more than one variety, potentially many - there are no “bad” or “incorrect” varieties, just appropriate or inappropriate - certain varieties are appropriate for certain situations (“prestige” in a job interview, “local” or “social-group” dialects at a football match - formal L2 examinations are generally written by the dominant group (or some outside power), so that is what is taught and needs to be (re-)produced

  • Goal should be to be able to navigate real life, not pass exams

    • learners need to be “realistic” - what will you really need? when will you need it?
  • Because everyone is different, and there is no such thing as a “bad” human being, every variety is valid as an entry point

    • Find a group or groups in the target L2 that you identify with
      • Find authentic language sources that enable you to connect with this group (literature/texts, music, video, immigrant groups/clubs, online discussion boards, etc.)
      • the only “bad” groups in a target L2 community of practice for an individual are groups that the individual doesn’t identify with
    • Knowing a variety doesn’t mean knowing it will get you full marks on an exam - you need to know how it relates to the standard variety
  • People’s identities are multiple and change

    • Someone may like performance cars at 13, gangsta rap at 16, and climate change at 19 - or all of these at the same time.
    • from high-school on (and usually from when they start school), normal native speakers are exposed to standard varieties, so native speakers will necessarily have that as one of their varieties, at least for understanding
  • Vocab lists and grammar can be useful for understanding

  • Real language will thus be a mix of standard and non-standard for most people - you should know both for the things that are meaningful to you as an individual. Not important to your teachers or parents, to you.

Tech: ??? Think about gamification

??? Threshold effects - example of smartphone - nothing new with iPhone all tech already existed and many people realised it would be a transformative tech (it just makes sense!) - but Jobs understood the threshold effect of usability - over a certain level of usability the learning curve and cognitive load while using it became acceptable for large numbers of users. Language tech needs the same - it’s all there, it just needs to be put together in a usable package.

Identity: ??? Identity and L2 - was (is?) trend to look at case studies and individual biographies rather than rely only (predominantly?) on “normal experiments”. Krashen likes “normal experiments” but also does quite a bit of case study stuff!

??? Krashen SLA and SLL book (198?) p77 “On the other hand, formal operations may be at least partly responsible for a fossilization of progress in subconscious language acquisition. One effect of formal operations on acquisition may be a result of what can be termed the indirect effects of formal operations, namely the affective changes that occur in adolescence that are catalyzed by formal operations. These changes result in the self consciousness and feelings of vulnerability often observed in this age group. Elkind (1970) has argued that formal operations allows one to conceptualize the thoughts of others-this leads the adolescent to the false conclusions that others are thinking about him and are focusing on just what he considers to be his inadequacies. Such feeling may generate at least in part attitudes unfavorable for the successful acquisition of a second language, which may actoto discourage the acquiirer from interacting with primary linguistic data, and/or may act to strenegthen an “affective filter” (Dulay and Burt, 1977) that prevents the acquirer from utilizing all the input he hears for further language acquisition.”

Meaningful IO, language and LX

The definition of Meaningful IO makes no mention of an L1 or even an L2. This is intentional - the theory predicts that they will indeed be learnt in the same way. Indeed, the theory predicts that a speaker will undergo a similar process when interacting in the various communities of practice that they participate in over their lifetimes, whether that be with members of “a different language family” or a new social or regional variety of their L1. In fact, further still, Meaningful IO has no need for a notion of “language” (as separated from other “extra-linguistic” means of communication) or “languages” (English, French, etc.).

As the famous saying goes - “a language is a dialect with an army and a navy”.

Meaningful IO and "extra-linguistic" communication

So called “extra-linguistic communication” is treated no differently to “linguistic” communication for Meaningful IO. Gestures, gaits, postures, mannerisms, dress, fashion accessories and other conscious and non-conscious actions are very often an integral part of group identity. People of a particular race who grew up in a community very different from their heritage community (e.g, a Chinese-American in Taiwan) will often be able to be identified merely by their gait, or by their clothing and mannerisms (e.g., a Somali-American in Somalia).

??? more here ???

Meaningful IO, formal instruction, assessment and learning strategies

Meaningful IO, by separating out clearly what “L2 knowledge competence” is (the gap between the current and the idealised L2 self, smaller means higher competence) and what it is not (a given mark on a given language exam, etc.), offers guidance on what activities might be most productive and efficient in taking an individual closer to that goal.

As is broached at length in other parts of the discussion, one needs to make a clear distinction between the goal of L2 competence, and the goal of achieving a certain score on an L2 assessment (mark for a formal instruction course or internationally recognised assessment). Strategies which optimise for one may be very far from optimal for the other.

Meaningful IO and formal assessment

Every form of assessment is necessarily a proxy. It attempts to measure general abilities or knowledge based on a (usually very short) set of artifacts (quiz/exam results, projects, pieces of writing, oral evaluation, etc.). This is uncontroversial.

An unfortunate reality, however, is that many people (both lay-people and SLA researchers) confuse the results of L2 assessments with actual L2 competence. The problem is that behaviours optimised for performance on an assessment are not necessarily optimised for an individual acting in real language contexts.

Good assessments will typically acknowledge these facts and attempt to create techniques that are very hard to optimise for - the goal is to be the best proxy possible, given the resources available (expense, assessor time, learner time, etc.). However, if the particular set of assessment techniques present in a programme themselves are a fantastic proxy for achieving their stated goal (knowledge of/ability to do X), then we also need to ascertain whether the X is actually L2 competence, or something else. In reality, language assessments will often themselves not be testing for L2 competence but rather for a proxy of L2 competence, so they are a proxy of a proxy (e.g, “what is meant by phrase X?” = proxy of ability to identify paraphrases = proxy of a certain aspect of a certain conception of L2 competence)

Stress, topic familiarity, particular day (physical state, illness), etc.

Teacher/Societal goals of L2 for individual X vs learner goals - ideal L2 selves and ideal L2 others

When talking about the construction of an idealised L2 self, one must consider that this idealised self is essentially personal. However, people don’t only have idealised L2 selves, people also have “idealised L2 others”. This becomes important in many contexts, and can be a critical determinant for a given learner’s likelihood in realising their idealised L2 self. This is because learners are not atoms in a vacuum, they are embedded in a social context and if their idealised L2 self is different from the idealised version of the individual that powerful others have, significant tensions may appear and may make L2 identity realisation very difficult for an individual.

A few examples: - an individual may be in a relationship with a person from a given L2 community, and may have an idealised L2 self where they can interact not only with the person (which they might habitually do in either their L1, or with varying degrees of success the L2), but also with the family and friends of the partner. The idealised L2 self might be working in the L2 community in a job conducted predominantly in the L2, ordering food from a restaurant on a date with the partner, watching football at the pub with L2 locals, writing to the partner’s parents, etc. The partner may, for various reasons, not see this at all - they may not want to return to the L2 community, and they may see their L2 identity as something they don’t want to share. They may want the individual to not be able to interact freely with parents, or go to an L2-speaking pub, etc. - a learner in a formal instructional context may have an idealised L2 self that is markedly different from the idealised L2 other of their teacher. A teacher may desire that a given student can write an essay on a particular topic, using certain vocabulary and grammatical constructions. The teacher may also want a given student to adopt a certain set of (usually positive) beliefs and attitudes about the L2 culture, or desire that learners become interested in particular topics. For example, the teacher may have learnt significant amounts of L2 via their love of L2 pop music, and desire that their students also become interested. That might enable the teacher to use increasing amounts of L2 pop music in instructional activities and assessments, something they believe will increase their own enjoyment teaching. (This is not to say this is a bad thing - there is plenty of evidence to suggest that teaching a topic (or via a topic) you are passionate about yourself may be highly effective. The teacher may have also developed highly effective techniques for maximising language learning value around the given topic, and teaching those techniques may be allow many students to gain useful generalised learning skills and strategies). - Governments may have macro-economic or macro-social goals related to L2 competence. Social cohesion might be a goal, such as the development of a lingua franca in a multi-lingual nation (India, Canada, China, etc.) or region (Europe, etc.). Economic goals might be to facilitate economic exchange or economic integration (L2 English so Chinese businesses can do business internationally, etc.). In some cases, the idealised L2 others might conceivably even have specific competencies (ability to negotiate an oil contract over email) and not have others (ability to chat about feminist literature in a coffee shop). Sometimes the goals will be multi-layered, e.g, there may be a belief that strong economic integration is the most effective means for avoiding inter-group conflict (e.g, no more wars in Europe).

So other stakeholders (parents, teachers, partners, children, governments, community leaders, etc.) have an influence on how easy it will be for an L2 learner to pursue their idealised L2 self. In reality, idealised L2 selves are not fixed, a particular individual’s idealised L2 self will be constantly renegotiated as they interact with others, and those others’ idealised L2 other, in their daily lives. Being dynamic, others’ idealised L2 others may also evolve as contexts change (the L2 partner’s parents may become incapacitated and require them to return home to take care of them, requiring our learner to get a job, which may require significant L2 competence; the partner might realise that they are respected by their friends more if the individual can order at an L2 speaking restaurant, etc.).

Ideal L2 technologies will be able to provide value to all the influencers, and allow for effective negotiation of the L2 identities (self and other) for all these parties. The technologies must recognise the potential tensions, and allow for maximised outcomes for all. To give a specific example, an ideal technology might be able to both allow a learner to participate in an online community around fly-fishing and also develop some of the skills necessary to negotiate contracts over email at the same time. And ideal technology will also give visibility on how it achieves these goals to all the stakeholders, allowing them to better understand the complex and dynamic interrelations.

Meaningful IO and access to certain L2 communities or materials

Meaningful IO rejects the utility of abstracting away the wider communal and macro-social influences and tensions the affect L2 acquisition. The realisation of the idealised L2 self may at times be in significant tension with the idealised L2 other of other important actors in a given learners orbit. A learner may

Meaningful IO, PUMLs and learning strategies

While the ultimate goal of SLA technology to optimise for Meaningful IO is to provide learners with a way to meaningfully interact with other members of a linguistic community of practice, there are undoubtedly facts about the world that enable technology to further strenghten acquisition via conscious learning techniques, memory-focused features, composition aids, gamification, etc. Though the presence of an accurate PUML will significantly change the affordances available to technologists, teachers and learners, a great deal of research into learning strategies has already been done, and many of the technologies proposed on this site are simply adaptations of well-known strategies in technology (PUML, SRS, concordancing, extensive reading/media, auto-generated questions, etc.). Many of these strategies are known to be effective, and it would be foolish to ignore them - if learners are happy with a given technique, then they should have tools not only to continue using it but also to see how that technique relates and compares to other techniques they use. Certain techniques will curry favour more or less with the other stakeholders in the learners orbit (teachers, government curricula, international examiners, university entrance boards, etc.) as they will develop certain skills that might be valued. As already mentioned, successful technologies need to provide value to all actors, not just the learners and not just the teachers.

Glossing, extensive reading and comprehensible input

The basic question: Is “glossing” good for getting the “comprehensible input” we need for “extensive reading”?

The problem: Glossing results in the literature are mixed, some good, some show no effect.

The hypothesis: There are different kinds of glosses, some are good for some things, but not other things, others may not be good for anything.

Accurate, dynamic, personalised, in-stream single words L1 glosses, supported by in situ dictionary definitions, sentence translations and grammar explanations are effective in facilitating comprehension of authentic L2 texts and do not significantly negate the majority of the language learning benefits associated with extensive reading. Automated, personalised L1 glossing of ditigal texts, while often inaccurate, is sufficiently accurate for achieving the main purpose of allowing a significant majority of digital L2 texts to be L1-glossed for the major language pairs, and as such gives very cost-effective and free choice of extensive reading material to most L2 learners around the world.

What problem are we trying to solve? Why?

Before actually undertaking any expensive/time-consuming problem-solving activity it is salutory to get a very good picture of the problem that needs to be solved, and then decide whether the problem genuinely does need to be solved. The “5-whys” is a well-known root problem identification technique and we should do something similar to understand what the purpose of glossing is. Glossing L2 texts can be a very complicated and time-consuming activity, and if it doesn’t have the desired benefits we hope it will, we certainly shouldn’t waste our time doing it or studying it.

So what is this desired benefit? Let’s start with two words: L2 competence. One issue with a lot of the both practical and experimental work around language acquisition is that even defining what L2 competence is is a significant feat in itself, and actually measuring it is very hard.

While any such definition is obviously fraught with many problems of its own, let’s start with a definition of “complete L2 competence”.

Complete L2 competence is the ability to participate fully and efficiently in the general and specific daily activities of a given individual in a group of practice most participants would regard as "the X-speaking community".

Let us take a moment to tease out some of what the above definition entails - this is directly relevant to many of the choices and arguments to come later surrounding glossing, so please bear with me! To start with, we can note that it makes no explicit appeal to knowledge of a certain “thing” or “things” (words, grammar, expressions, etc.) or a particular skill (pronunciation accuracy, oratory skills, reading speed, etc.). What this means in practice is that the ability to recognise/comprehend or produce particular forms is not important, except insofar as they are important in carrying out the general and specific activities of a given individual. It is fundamentally practical and situated. As an example, the spoken hortation “Please make sure to prune the magnolias today” is likely something that a gardener working in a stately home in Kensington needs to be able to understand quickly and accurately, while it is not certain that it is required for a theoretical physicist living in a flat in downtown Hong Kong. Likewise, “Please make sure to make sure the latest proton sim stats are updated before the draft goes to the professor” written in an email is something that a theoretical physicist might well encounter in his daily life and need to act upon but far from certain for a gardener.

This directly brings us to the relationship between L2 competence and “L1 competence” and is a crucial point to make and accept. I am a “native speaker of English”, with first class honours in Linguistics, a bachelor’s in computing from a second university and currently doing a PhD in Computer-Assisted Language Learning in a third, all of which were English-medium (and of which two were in an “English-native” country). It is unlikely I would be considered language-deficient in English by anyone, including an English literature professor. After I wrote the sentence “Please make sure to prune the magnolias today”, I realised that I didn’t know whether that made sense, so I looked up “prune magnolia” on Google. I didn’t know whether a magnolia was a tree or a flower like a daffodil with no branches (I also had to look up the spelling of the word “daffodil”). Does one “prune” a flower like a daffodil? Google immediately gave results for “pruning the magnolia tree”, and I was able to continue with my writing. However, were I in overalls before a certain “lord of the manor” awaiting instructions for the day’s work (and without a cellphone), I would not be able to perform the task required of me - I still have no idea what a magnolia tree looks like (and have resisted the urge to look it up)! Worse still, I actually spent three months working as a gardener (and general handiman) during one summer of my first university degree! What this highlights is that both L1 and L2 knowledge is both “knowledge of a language” and “knowledge of the world”.

A common question type in many modern language examination systems (HSK, ???what about others???, etc.) is to present pictures and ask for an appropriate reaction to them in combination with some spoken or written language cue. Before looking it up, had you presented me with a picture of a magnolia tree without flowers and a picture of some daffodil-like flowers planted in the ground, I would almost certainly have chosen the daffodils in a language exam, getting it wrong.

Accepting this (and the true value of free extensive reading I will argue) is

(comprehension? extensive reading? intensive reading? vocab learning? grammar? topic familiarity? general cultural familiarity? attitude towards L2 culture?)

Comprehensible input: What is comprehensible input? What is comprehension? What is input? Comprehensible input - what is it? what is it for? How is it measured?

Glossing: What is glossing in general? What are the different kinds of gloss? (pictures, sound, single words,

Pre-ubiquotous digital age Dumb glossing:

Everything connected age Smart glossing:

Meaningful IO and ultimate attainment

The Meaningful IO theory of SLA takes a socio-cultural, post-structuralist stance on language and learning but recognises the practical value of conscious learning and formal instruction for a majority of learners. The realities of today’s societies and technologies are such that literacy, and literate practices, form the bedrock of how the majority of how today’s formal communicating and learning happens. Learners do not live in socio-historical vacuums, isolated from the wider social narratives and practices, and as such are equipped with a set of skills that enable them to participate in literate communities of practice. The vast majority of L2 learners are indeed not embarking on journeys to participate in communities of practice of hunter-gatherers or feudal peasants, rather it is hoped that they will be able to navigate and find new identities in the literate and digitally mediated communities of today and tomorrow.

The written word is everywhere, and while it is but one of the many ways in which humans interact, and build and renegotiate identities, it can serve as a key to unlocking the many other digital and non-digital stages where humans’ communicative communities of practice evolve. Importantly, for most literate learners, it can enable them to participate in the vast sea of communities of L2 content and the written-word-rich social communities available in their pockets and elsewhere. For a significant proportion of L2 learners, the physical L2 others are not within earshot and even when they are, many learners struggle to find their voice in realtime face-to-face contexts. Non realtime activities, or realtime activities where the learner can engage as an (semi-)anonymous observer, can provide learners with a platform where they can develop their identity at their own pace, in a way that let’s them engage more actively when they feel confident enough to do so on their own terms. As one of the goals of L2 will almost always be to participate actively in L2 communities of practice in a variety of forms, both digitally mediated and “face-to-face”, literate practices can never provide learners with the full spectrum of competencies. A majority of learners will benefit, however, by being able to engage in these communities with an L2 identity that they have already been able to start developing, giving them a firmer footing from which to express themselves.

Language, and as such L2 acquisition, is fundamentally a doing - languaging. L2 acquisition is about fashioning an identity that enables the learner to participate in a greater range of linguistic communities of practice, in a way that enables them to express and evolve their identities as they desire in negotiation with their significant others. It is fundamentally dynamic, situated, and contextual. But another key aspect of today’s realities for learners is a certain conception of learning, both of L2 and other subject and skill areas, that of the learner-as-cup. Much of today’s formal instruction is based on the idea of the learner-as-cup, a cup that needs to be filled with knowledge of facts about the world. While there are many reasons to resist this view of learning and knowledge, the simple fact that it has become the dominant narrative around education in the lay person’s mind means it has taken on a psychological reality for a large majority of cultures and learners. It is a key feature of the “modern world”, and as participants in the “modern world”, learners must navigate this reality. Because for a majority of learners the cup-oriented model of learning a set of “facts” about “the L2” will not directly lead to successful participation in L2 communities of practice, filling such a cup should not be seen as the basis or goal of any learning journey. It can, however, be used by the learner and many other interested parties as a tool for better understanding where a learner is in relation to their idealised literate L2 self (for the learner) and the idealised literate L2 other (for the other parties).

Thankfully, with today’s technologies, much of this can finally fade into the background and we can collect large troves of information simply by observing how real humans interact in digital L2 communities of practice. Furthermore, the literacy-focused learner models that technology can develop can also provide the data for powerful tools to help learners get meaning and negotiate identities as they attempt to do so in the real world. Interaction with these tools will provide a further rich vein of data for ever more precise and useful models for allowing learners to actually do languaging and negotiate their L2 identities.

Through this learners can build and inspect a digital L2 representation of certain (literacy-focused) aspects of their L2 selves. This digital L2 self, quantified on the basis of a learner’s own personal search for meaning in the communities of practice they themselves seek out and inhabit, will for many have a much deeper relevance than a model designed by, controlled by and optimised for the idealised L2 other of powerful significant others. Such other-focused models are not a picture of the learner’s own journey but are pictures of how successful they are in conforming to someone else’s idealised L2 other, that of teachers, parents, religious groups, companies, governments, etc. Because learners live in the real world, and in the real world the idealised L2 others of those in significant power relationships with the learner can shape and seek to control their journeys, the learner can use their quantified L2 self to help negotiate with these powerful others in a way that empowers them and satisfies the others.

Because this representation belongs to and serves the learner-as-person-acting-in-the-world, and not some external force in a specific formal context, the learner takes it with them as they navigate and negotiate their identities through the multitude of contexts and roles they inhabit over the course of their lives. The representation will evolve in realtime with the learner, adapting to new contexts and giving them powerful insights as to how their literate L2 selves develop over time and throughout the important phases of their lives.

Such a tool is fundamentally a tool for giving learners visibility and control over their ultimate L2 attainment. It is not a tool that has an end, it is a lifelong companion that enables them to abstract from, visualise and analyse their literate L2 selves, and to fashion and renegotiate goals they may have with respect to them.