Working Memory in Language Learning: Thinking in a New Language
Recently, I lost an old friend, a native Italian and naturalized US citizen, whose stories about WWII and Korea had long fascinated me. But among the ideas that Remo elaborated, one that has influenced me most was his description of how he learned English as a native Italian speaker arriving in the US as a young adult in the early 1950s. He explained how, in addition to reading in English and watching movies, he forced himself to think in English whenever he could.
I never fully appreciated that technique, until I implemented it in reverse. As a native speaker of English, I’ve become somewhat fluent in Italian over the past ten years. Consequently, I’ve come to appreciate the value of thinking in Italian whenever I can — and engaging all working-memory systems during those thought processes.
Working Memory: A Central Executive
Working memory conceptualizes brain systems that are capable of holding information in consciousness for a short time period during which they can briefly process this information. In the popular working memory model proposed by Baddeley and Hitch in 1974 [1], a central executive, determines where our brain places its focus, upon what brain processing it shines the light of attention. Since our brains are parallel processors of diverse information, this attention focusing is a key functionality. For example, even considering only sensory inputs, your brain is currently processing your visual input in reading this text and simultaneously processing any sounds or odors in your immediate environment. And the temperature in the room where you’re reading in addition to any pain data coming up from your body.
Then there’s the position of your body and its extremities in space (proprioception) as revealed by the information pouring into your brain from stretch receptors in your muscles and joints. That proprioceptive data processing is a subconscious necessity for skilled movements like typing or throwing a ball. You can, if you choose, de-focus your attention from reading, shifting it to sounds or aromas around you. You might even focus on a single bone or joint during its movements — your wrist position as you type, for example; not something most folks normally do. We can think of this process of attention-focusing by the central executive as bringing other brain processing or its outcomes into conscious awareness.
Although this model is neuropsychological, with its underlying biology still quite mysterious, it does account for many observations — both of patients with memory impairments and of psychological interference experiments with normal subjects. These experiments test someone’s ability to hold and regurgitate different types of information when distractions are also present. They assess the effect of how different sensory inputs to the brain can interfere with processing information we’re paying attention to and how those distractions can even inhibit forming and retaining memories. Some of these experiments and the issues they raise are discussed in the readings below. Here, my purpose is to overview this useful model in terms of how I think it might help someone learn a non-native language as an adult.
Working Memory Subsystems: Inner-Hearing and Inner-Vision
In addition to the central executive, the original model proposed two essential subsystems within working memory. Like an inner voice/ear, the phonological loop allows the internal repetition of verbal information to extend its retention time within the otherwise labile working memory store. Without such repetition, its contents can degrade within a few seconds. For example, I might use this inner voice to repeatedly recite a new phone number to myself to attempt to recall it long enough to add a new entry to my phone database.
But I can also visualize that number as a string of digits in my visuospatial sketch pad, my moment-to-moment inner eye/vision. Individuals described as more-visual learners might depend more on this method of visualizing the phone number than on phonological looping. I sure do; my first mental operation upon hearing a phone number I want to remember is to recite it in the inner voice of my phonological loop, then immediately picture its digits in my sketch pad. This has always worked better for me personally than the continual repetition strategy. This strategy of using both working memory subsystems —inner voice/ear and inner eye/vision — is relevant to my approach to improving my learning of my second language, Italian.
Neurobiologically, I imagine two systems sending reinforcing information to the central executive in the form of closely spaced excitatory neuro-transmission that is thought to underpin effective changes in the synaptic network receiving the dual input. In this case, loop and sketchpad sending these signals spaced closely in time would cause the central exec and its buffer (below) to relay this reinforced information to long-term memory (LTM). How this coding is accomplished is still a topic of uncertainty and controversy.
Episodic Buffer
Unless it’s actively attended to or rehearsed, information in working memory has a short duration of between 1 and 15 seconds. Imagine that you’re furiously taking notes as you listen to a lecture, and you hear a unique phrasing of an important concept, which momentarily captures your attention. But you rush your attention back to finishing off the last entry in your notes. And 30 seconds later, you can’t recall the slick phrasing; frustratingly, it’s disappeared from your phonological loop.
Since the central executive lacks storage capacity, an additional sub-system, the episodic buffer was added to the model, as necessary to integrate phonological, visual, and long-term memory (LTM) information and to transiently store it. This episodic buffer also accounts for diverse observations, particularly those involving subjects with memory deficiencies. For example, as related in [3], there are examples of patients with severe amnesia who can still capably play complex card games like bridge, where they must be able to hold information in mind about the cards being played, at least for the duration of that hand — thus, longer than merely a few seconds. Such a person, with impaired LTM may later entirely forget the hand, yet is able to function well during its playing.
The episodic buffer idea is also necessitated by the evidence that working memory exchanges information with long-term memory (LTM). For example, as discussed in [4], how would you answer the question, how many windows in the front of your house? For most of us, the central executive would effect a transfer of information from your LTM library to form an image in your sketchpad, allowing you to count those windows in your inner vision. During this process, the central executive is focusing attention on the sketchpad’s reconstruction of the image of your house — using information derived from LTM.
Native Language: Grammatical Automaticity
In any case, you’re probably not relying much on this system in speaking your native language(s), the one(s) you learned as an infant/toddler; at least not most of the time. Exactly how the grammatical-phonological structures of that language are imprinted in your brain is quite a complex question and not understood in detail, but that process might be analogized to other learned skilled processes like throwing a ball. Once learned in early childhood, we don’t need to assemble those native grammatical structures in our working memory, nor do we require any forethought about the skilled movements of our vocal cords, lips, and tongue required to pronounce words. Like throwing the ball, the programs (however they’re stored in the brain) seem to execute quasi-automatically without the need to overly focus Central Executive attention on them before speaking or writing.
Although I’m aware of the structure and articulated sounds of what I say or write after the fact, in speaking English, I usually don’t need to rehearse them beforehand. Sometimes, when we sense or know that we’re broaching a delicate topic or uttering something controversial, we do indeed practice our speech with the inner voice in our phonological loop. There are times in my life when I wish I’d done so better. But during casual speech, this isn’t usually the case when speaking in one’s native tongue. Throwing a ball or conversing in your native language, the underlying details are executed with a somewhat similar subconscious automaticity.
Intrinsic Language Structure: Younger and Older Brains
It seems that there’s a particular biology in a toddler’s brain that drinks in the phonemes and grammatical structure of languages like an cognitive elixir. I see evidence for this in my own brain. I never learned to speak or write Italian as a child — except for profane curses and other epithets necessary for surviving growing-up in Brooklyn. But my bilingual grandmother who lived with us often spoke snippets to me. Plus, I overheard my grandmother and mom occasionally converse in Italian. Around sixty years later, when I really began to study Italian with the aim of becoming reasonably fluent, I found that I was always in the upper tier of my peer learner group in one respect — pronunciation of the language’s phonemes. Apparently, my three-year old brain just scooped them out of the air from the overheard conversations and stored them, ready for use sixty years later.
And not just the sounds, but also the Italian rhythm. The way syllables are accented, and because there are so many pronounced terminal vowels (eh, ah, ee, oh), Italian speech has more of a flowing rhythm. Unlike English, chopped by final consonants, which has more cadence, less flow. These are unconscious aspects of a language that we — my three-year-old brain — store subconsciously for a lifetime unless our brains become damaged. So between the phonemes and the flow, I sound more like a native speaker than most of my peer learner group. Certainly, less like a straniero (foreigner), a bumbling Yankee.
Would that the rest of the language came so easily. It does — to young kids, whose brains seem to act as language sponges, storing all that unconscious info about a language. It’s one reason why waiting until junior high or high school to teach language acquisition seems a far less productive strategy than introducing those much younger brains to new languages. While this educational model is more prevalent in Europe, I do have US friends whose kids or grandkids are functionally bilingual at five. Bravissimi!
Languages have structural levels that linguists analyze, but that native speakers rarely think about (unless forced to in school). So beginning early in childhood, native speakers store this grammatical structure at a brain level from which it’s automatically accessed in speech. Accessing grammar in writing a language may be a greater chore for struggling or inexperienced writers, who must often assess their written grammar against rules (especially if they want good grades from their teachers).
Even practiced writers and speakers may need to think about grammar in some instances. But it’s probably more common for grammatical usage to spring from that automatic brain level, analogous to learned muscle movements like throwing a ball. During those learned motions, motor programs are fetched from the cerebellum and executed by the motor cortex. By analogy, in skilled native speakers, it seems like there are grammar programs that execute automatically without rehearsal. They coordinate the structural combinations of phonemes and words, analogous to the motor programs coordinating muscle movements.
Learning a New Language as an Adult
In contrast, when one tries to assimilate another language as an adult, there’s a necessary awareness and learning of the new language’s grammatical rules, actively drawing them from long-term memory to guide the structuring of vocabulary. It’s hardly automatic beginning with accessing/choosing the form and tense of verbs around which to structure the remainder of one’s sentences.
Hence, the central executive needs to focus attention on the new language. I find that as I’m doing something automatic like walking on a trail in Nature, my attention seems unleashed from other cognitive tasks, so I sequentially focus it on both of my working memory subsystems. When a thought or impulse enters my mind, I immediately try to translate and internally pronounce it in Italian in my phonological loop, then quickly visualize it written or typed on my visuospatial sketch pad in Italian, sometimes one word at a time. Then I continue to pronounce it in my phonological loop. When I’m satisfied that its grammar and vocabulary are acceptable, I’ll often speak it aloud several times. Repeated, this process will hopefully send aspects of that small bit of Italian to long-term memory storage.
The closely spaced sequence of using both working memory subsystems — inner eye/vision and inner ear/voice — seems to be what my old friend Remo Deluca was advertising as his method for thinking in English, his foreign language.
There are no subject-matter restrictions, simply a commitment to do this practice using whatever happens to catch my attention from the various thought streams or sensory data entering my consciousness. For example, I’ll begin every hike by telling myself (in the third person), Inspira questa bella giornata, Vincenzo: “Breathe-in this beautiful day, Vin.” If there’s a song in my mind, I might try translating the lyrics from English to Italian. If I’m observing new plant growth in the spring, I might be stimulated to try to describe the different hues of green I’m seeing. Verde luminoso, bright green, verde d’oliva, olive green, and before I knew the Italian for “emerald-green” (verde smeraldo), I pictured and internally articulated verde Irlandese. Irish green, since Ireland is also called “the emerald isle.” So faking substitute vocabulary is allowed.
If I’m occupied with a specific problem or challenge that’s been on my mind, I’ll attempt to express what that problem is in Italian. Lately, the issue of COVID and what part of the immune system is most important to stimulate with a vaccine has been on my biologist’s mind, so I might imagine having a conversation with a colleague about the issue, beginning with: Che ti pensi? nella malattia di COVID, che tipo d’intervento sara il più meglio? “What do you think? in COVID disease, which type of intervention will be the best?” And if I take a moment during my walk to smell the piney fragrances of an evergreen, I might say, thanking the tree, grazie per permettermi di fiutarti; “thanks for letting me smell you.”
Often, I’ll see something else I’d like to attempt to describe in Italian — the shapes of plant leaves, the way a lizard moves or the particular qualities of bird sounds. I try not to internally say the sentence in English with my inner voice, but rather to say it directly in Italian, as I also visualize it printed in my sketch pad. Just a little imagination can generate all sorts of topics from which to form sentences in your second language.
And they don’t have to be outdoor Nature thoughts. For instance, I think in Italian about many aspects of my life , formulating a sentence about a snippet of my own experience, than checking out my sentences with a translator. Recently, I sat down at my keyboard just to play for release after our insanely inept election. I had a truly enjoyable 20 minutes of old songs connected by improvisation. So as I walked back to my office, I thought how satisfying and uplifting it was. And I said to myself: “Mi ho piacuto molto questa serenata. È stata benissima per la mia salute di mente.” Which the translator told me was “I really enjoyed this serenade. It was very good for my mental health.” Those affirmed translations can be moments of serious encouragement to adult language learners — to keep on truckin’, keep working, you’re making progress.
Reaching for Advanced Fluency
This process really only works once you’ve reached the comfort level of having attained intermediate proficiency (levels B1/B2) of a non-native language. You’re able to speak/write it well enough so that the basic grammatical elements are in place, with serviceable vocabulary and some appreciation of the common forms and idioms of the language. These elements are necessary to avoid the assumption that simply translating word-for-word from one language to another will yield an expression that’s both correct and conversational. It often doesn’t, as any good language teacher will tell you.
For example, while the subjunctive mood of verbs has been largely discarded from common usage in American English (perhaps less so in British), Italian uses this verb mood in many situations, starting with any that border on uncertainty. In Italian, the simple declarative, “it rains” is si piove (indicative tense). But, “I hope it rains” is spero che si piova (subjunctive). I may hope it rains, but implicit in that hope is uncertainty of outcome. We still hear the subjunctive used in US English in rare instances. For example, “I wish it were (not was) cooler” is correct, likewise expressing uncertainty of outcome. It may get cooler. Or it may not. And enough folks correctly use that subjunctive form of “to be” to keep the mood alive in English. Barely! In Italian, by comparison, the subjunctive is used with considerable frequency. A volte, mi sembra che in Italiano, il congiuntivo appare dappertutto.[Sometimes it seems to me that in Italian the subjunctive appears everywhere]. So there are many cases where a direct translation of a sentence where the English verb is in the indicative tense will be incorrect because the Italian version of the sentence requires the verb in the subjunctive to be grammatically correct.
Beyond this are the idiomatic and colloquial usages that differ between languages. For example, in Italian, saying in bocca al lupo means good luck, not the literal translation, “into the mouth of a wolf.” In parallel, when we say “break a leg,” we’re also wishing someone good luck in a performance rather than a physical injury. Same idea, different cultural idioms.
While ultimately important in language mastery, this is an issue that doesn’t detract from the value of the process of internally practicing a language using both working memory subsystems during periods when one’s working memory is relatively unoccupied with other tasks. And if your vocabulary is lacking in some key words that are needed to give full expression to your description, it’s a vehicle for looking up the missing words, either immediately or shortly thereafter and learning the new vocabulary in the context of your own ideation and sentence construction. Closer to what interests you.
What Aspects Might Reach Long Term Memory?
Another way to optimize transfer to LTM is to use your phone to either tap out a memo or record a voice memo (my preference); while you still retain the sentence in your episodic buffer.
But suppose I don’t bother recording my newly constructed sentences as voice memos. Is the process still worth it? That is, when one does this process with a diversity of sentences over time, what aspects might transfer to LTM in addition to a reinforced vocabulary of words one already knows? My experience suggests that, even if you don’t later remember many of the exact sentences themselves, somehow information about the grammatical structures and phonemes makes it into LTM storage, thereby increasing your grammatical-lexical information store. And bringing you an step closer to automaticity in using the language. Like a geriatric toddler. I hope that’s what’s happening, and my personal experiences suggest that it might be.
Happy language learning! Felice esperienze in apprendimento della tua lingua preferita.
This post is dedicated to the memory of Remo Deluca, who was one inspiration on the road to tackling a second language relatively late in life. And to my maternal grandmother, Nonna Teresa. Grazie for the sonorous phonemes and the flowing rhythms.
Additional Reading
1. Baddeley, A. D., & Hitch, G. (1974). “Working memory.” In G.H. Bower (Ed.), The psychology of learning and motivation: Advances in research and theory (Vol. 8, pp. 47–89). New York: Academic Press.
2. Baddeley, A., (1992). “Working Memory” Science 255: 556–779
3. Baddeley, A. (2000). “The episodic buffer: a new component of working memory?” Trends in Cognitive Sciences — Vol. 4, №11.
4. McLeod, S. A. (2012). Working memory. Simply Psychology. https://www.simplypsychology.org/working%20memory.html
