Research has left teachers dangling for decades on all sorts of topics that, at first glance, seem to have no connection with one another.

Is music distracting or soothing while studying? What about note-taking - is it a help, or a distraction, as students try to grasp key ideas? And when it comes to rereading and underlining, why do students fall so naturally into those approaches if, as research has revealed, they’re so harmful?

Then there’s perhaps the most contentious question in all of education: do students learn better via direct explanation by instructors, the traditionalist approach? Or are the student-centred approaches, favoured by reform educators, better? Some studies draw one conclusion, others another. And most studies, focused as they are on large-scale effects, avoid the nitty-gritty of what’s happening in the brain.

In schools, this all plays out as a confusing welter of teaching approaches, all claiming a basis on scientific research findings. It’s a little like the parable of the blind men and the elephant, in which each man touches a different part of the elephant’s body and disbelieves the others’ account of the animal. How, they wonder, are others able to draw such radically different conclusions?

Recently, though, we have begun to find some clarity in the blur of confusion. And it is coming from a surprising source: research on working memory.

What is working memory?

Working memory is a mental holding tank in which you can fleetingly retain and perhaps process what you’ve just taken in.

For example, if you are introduced to two new colleagues, Klaus and Zelma, you might temporarily repeat their names. “Klaus looks like Santa Klaus. And Zelma is…”

Oops, now you’re shaking hands with Pete, and Zelma’s name has flown out of your working memory before you could store it.

Working memory is commonly thought of as involving the parts of the brain that temporarily hold and also manipulate - that is, do cognitive work on - the information. But if you’re bold (or crazy) enough to dive into the research literature, you’ll see it’s not as easy as all that. Some definitions of working memory include both processing and storage. Others include only storage. Or only attention and control.

Coming up with a rigorous definition of working memory is a little like trying to dig a hole in the sand. Everybody digs differently, and if you’re not careful, the edges crumble and you don’t have a well-defined hole at all [1].

When it comes to long-term memory, that more permanent storage place, researchers are on somewhat firmer ground [2]. But it’s not as though working memory is completely disconnected from long-term memory - at least not according to some highly influential definitions [3]. The neural assemblies of longterm memory can activate and extend working memory. It’s a little like people in deckchairs who can’t help getting up and joining a conga line when they hear their song [4].

Computer scientists seem to be the first to have used the term “working memory”, back in the 1950s [5]. A few years later, it began cropping up in psychology research, applied to related conceptions involving a limit of “seven, plus or minus two” items [6]. The items could be numbers, letters or even a larger chunk of information, at least if that chunk were familiar.

More recently, Nelson Cowan and others have presented powerful evidence that working memory can, on average, hold only four items or perhaps even fewer [7].

One simple way of digging a definitional hole (so to speak) for working memory ignores potential input streams such as touch, taste, smell and other issues. Instead, it focuses on three components, which can be described as [8]:

• The central executive: The part of your brain that allows you to manipulate information. Sometimes this also implicitly includes a focus of attention that directs your thoughts.
• A phonological loop: The “hearing” portion of what your brain is taking in. It’s also associated with inner vocalisations, such as repeating a number to remember it.
• A visuospatial sketchpad: This involves your visual and spatial intake mechanisms.

This “threesome” approach to understanding working memory is useful because it focuses on the essentials.

Research reveals that the phonological loop centers around the left temporoparietal region, while the visuospatial working memory is cradled in the right temporoparietal region [9]. The central executive is towards the front of the brain; it works in tandem with a “focus of attention” that helps direct your thinking.  This focus of attention involves the parietal lobe (and specifically the intraparietal sulcus), possibly as a central hub of a focus-of-attention network.

In simple terms, you can think of working memory as being like an octopus in the front of the brain (the central executive) that can throw balls (thoughts) that bounce as directed by the focus of attention over towards the audio and visual areas, which, in turn, bounce the ball back to the octopus in the front [10]. This “ball bouncing” from front to back and then from back to front keeps the thought alive in working memory.

The quirky thing about working memory is that whenever the attentional octopus gets distracted from its ball-throwing and catching, the ball can vanish. This leads to one of the fascinating aspects of working memory - its cunning ability to fool students into thinking that they’ve put something into long-term memory.

A student can, for example, stare at a list of 10 new Spanish vocabulary words and think, “I’ve got ’em!” The student does have the words in mind - that is, as long as she is staring at the list. When the student looks away and starts chatting to a friend, the ball drops, and it’s gone.

Similar problems arise when a student glances at the solution to a complex maths problem. “No need to waste time working this out on my own,” she might think, “I’ve got it in mind already.”

And she does have it, at least partly, in mind - temporarily. Students only discover the vanishing act when they take a test. (“I suffer from test anxiety,” can, in fact, be code for, “I feel panic when I reach into long-term memory and nothing’s there.”)

This “false friend” nature of working memory is why students naturally tend towards rereading and underlining. What could be easier and more helpful than running your eyes over the information one more time, underlining to add emphasis?

So how do we get something into our long-term memory? To do this, the tedious process of retrieval practice - that is, using flashcards or simply looking away to recall a key idea - is the best bet [11]. Because it’s harder, students often need to be taught to use retrieval practice [12].

It seems that the terms “working memory” and “intelligence” describe similar underlying processes [13]. And sure enough, those with lower-capacity working memory can struggle more with their learning.

But remember how long-term memory can become a part of working memory? This is good news. It means that if the person with a lower capacity working memory creates and strengthens neural links in long-term memory, those links can extend their working memory. At least when it comes to that topic [14]. Given additional time and well-designed practice, a person with a lower capacity working memory can become as good as - or even better than - the higher capacity student in their area of expertise [15].

This background practice is critical in ways that can at first be difficult to grasp. For example, let’s take the sentence: “The green penguin is eating an apple.”

It would be easy for you to write each letter of the sentence down a minute later.

Now let’s take another sentence: “Зеленый пингвин ест яблоко.”

Unless you’re a Russian speaker, it’s going to be pretty difficult for you to hold all the letters in mind and write them down a minute later, even though the green Russian penguin is similarly eating an apple.

Our “working memory” capacity appears to be much greater or lesser depending on whether our long-term memories have had English or Russian implanted. So your background training matters - a lot.

As John Sweller has pointed out, the intricate, wonderful relationship between working memory and long-term memory is easily the most important factor in human cognition - it goes a long way to understanding how our mind works [16]. And this knowledge should be used to influence how we teach and learn, too.

On a side note, we should point out that, despite enthusiasm by researchers and general population alike, there’s no good evidence as yet that general working memory capacity can be increased through training-although, as we’ve seen, something that looks very much like working memory increase seems to occur within a specific area of practice [17]. 

The transformative effect of education is not that it changes students’ basic working memory capacity. Education instead changes the amount of knowledge held in long-term memory-this is what drives the apparent change in working memory. (This is the “expertise reversal effect,” where, the better a student is, the less guidance they need. [18]) With the right kind of information implanted in long-term memory, people can effortlessly process enormous amounts of information-even if their working memories aren’t that capacious.

What are the implications of this for students?

Let’s take music as an example: students are often told to avoid listening to it while studying. The problem is that some successful students happily listen to music when studying. Why should Sam avoid music when he knows darn well that Edwina listens to it and still gets great grades?

The latest research findings help to solve the puzzle. Music’s effect on studies varies with - you guessed it -working memory capacity [19]. Those with a lower capacity are better off avoiding music altogether in their studies. Those with a higher capacity, on the other hand, can often perform well in exams after studying to music - their higher capacities also help them to focus more easily.

The caveat is that everyone should avoid studying maths while listening to music. Perhaps this might relate to the fact that maths and music use overlapping portions of the brain [20, 21].

What about note-taking? Again, working memory seems to play a role [22]. Those with high capacity can blithely jot notes while also taking in complex explanations from a teacher. But those with lower capacity have trouble simultaneously taking notes and making sense of the instructor’s explanation. They can end up spending a lot more time away from class trying to reconstruct the instructor’s meaning.

Research’s surprising conclusion in all of this? The student with lower-capacity working memory can do well by focusing only on the instructor during the lecture, using others’ notes for review [23].

So the lesson from all of this is that, to be successful in their studies, those with lower-capacity working memories must have more information offloaded into long-term memory. Of course, no one wants to create and strengthen links via excessive “drill and kill” approaches that monotonously suck the enthusiasm and creativity out of students. Innovative teaching adds variety and novel components to the practice, which is why gaming programs such as Smartick, used to teach maths, are so popular.

One more thing: this offloading can provide an advantage. It seems to crystallise and simplify concepts. Ultimately, this means that an industrious low-capacity type can make elegant simplifications that those with higher capacities have difficulty seeing [24].

In a related vein, being tired or tipsy, which lowers working memory capacity, seems to heighten people’s ability to solve problems that need creative insight [25]. 

Having low-capacity working memory can make you better off in education, given the right instruction.

And what are the implications for teachers?

It’s probably no surprise to learn that teaching techniques that work well for students with higher capacity working memories can be harmful for students with lower capacity.

Let’s take, for example, mathematics instruction. Students with high working memory capacity can do well no matter what the instruction type - studentcentred or teacher-directed - although they may particularly flourish with student-centred approaches.

But students who struggle with maths - a common problem among those with lower-capacity working memory [26]  - appear to do worse with studentcentred, and better with teacher-directed, approaches [27].

Research reveals that the teacher-directed approach [28] seems to have the biggest positive effect on these struggling students. Emphasis on procedural fluency appears to strengthen low-capacity students’ grasp of a subject by allowing their longterm memory to enhance their working memory [29]. As these students gain familiarity with the basic concepts, the instruction can shift to a student-centred inquiry approach.

Similarly, reading instruction can have different effects based on students’ initial capabilities - which, importantly, depend on their underlying working memory capacity [30]. A phonics approach is more helpful for those entering school with lower reading ability, while those with higher initial performance flourish under whole language (integrated language arts) instruction [31]. Again, as students gain mastery, the instruction can shift to studentcentred approaches.

The challenge for teachers is that a typical classroom contains students with a hodge-podge of working memory capacities. The typical mixed teaching techniques used by many teachers - student-centred and teacher-directed, phonics and whole language - can be very effective for high-capacity students. But students with lower capacities often need more practice and direct instruction to get them on board so that the student-centred approaches can subsequently take hold.

Without the initial scaffolding provided by teacher-centred approaches, student-centred work can leave students with low-capacity working memories frustrated and confused.

As long-time professor of education Beth Rogowsky notes: “Direct instruction requires a great deal of effort by the teacher and constant student interaction. In student-centred approaches, on the other hand, teachers can ‘set up’ and let the kids do it on their own. You can see the problem with this approach and why it doesn’t work with struggling learners or those who are not motivated. The students who are not motivated need the attention that direct instruction provides [32].” 

By strengthening their long-term memory on a topic, a student can still be successful even if their working memory isn’t capacious.

However, there can be students with exceptionally low working memory capacity. In these cases, it’s not as if a student is actively fidgeting, as with attention deficit hyperactivity disorder. Instead, a busy, sometimes overwhelmed teacher can conclude that a student is just something of a bumbler, especially when the other students seem to be able to follow instructions just fine [33]. Testing the working memory of students who have difficulty holding ideas in mind can go a long way towards capturing a potential learning challenge early on.

Looking forward

The way we teach is willy-nilly, affected by the way we ourselves learn best. Perhaps this is why teaching approaches such as whole language and student-centred maths, which work well for those with high-capacity working memories, are favoured by many highly intelligent reform educational theorists. Growing up, these experts could figure things out pretty well on their own - they didn’t need much extra practice. Consequently, they can’t help but believe that this is how all students learn best. This, in turn, can lead to bias in their research [34].

Along similar lines, one might speculate that teaching programmes promoted by Silicon Valley mega moguls often are the types of programmes the hyper-intelligent moguls themselves would have flourished on when they were growing up. This might be why those programmes flounder when they meet real-world students [35].

The truth is: lower-capacity working memories learn very differently, needing extra reinforcement to strengthen mental representations in long-term memory. Research on working memory is providing remarkable new insights - insights that are helping to reconcile traditional versus reform approaches to education, and that answer other important questions about how students should study.

The best of future teaching will integrate both old and new approaches in individualised ways to help all students flourish, despite their differences.

Barbara Oakley is a professor of engineering at Oakland University in Rochester, Michigan, and the author of books including A Mind for Numbers and Learning How to Learn. The author would like to thank Nelson Cowan, Alan Baddeley, John Sweller, Jo Boaler, Beth Rogowsky, Daniel González de Vega and Berta Gonzalez for their suggestions and insights

If you’d like to learn more, Barbara’s Learn How To Learn video series, a tutorial course in different aspects of learning, can be found at bit.ly/LearningtoLearn

This article originally appeared in the 28 June 2019 issue under the headline “The secret to good teaching? Don’t forget about memory”

Endnotes

[1] Cowan, 2017 is well worth reading.
[2]  Fiebig and Lansner, 2014, Mednick, et al., 2011, Mongillo, et al., 2008, Tonegawa, et al., 2018. Of course, the information flow into long-term memory via the hippocampus is also important-see Schapiro, et al., 2017 for a nifty discussion of the differences in information that is processed through the monosynaptic versus the trisynaptic pathways. 
[3]  See Ericsson and Kintsch, 1995, the contextual discussion in Cowan, 2017, and most recently Cowan, In press 2019. Or is Baddeley’s conception of an episodic buffer the linking mechanism?  
[4]  If you must know, “interareal phase synchrony in the α-, β-, and γ-frequency bands among frontoparietal and visual regions could be a systems level mechanism for coordinating and regulating the maintenance of neuronal object representations in [visual working memory].” Palva, et al., 2010.
[5]  Newell and Simon, 1956. 
[6] Miller, 1956 first used the “seven plus or minus two” phrase in relation to what he called “immediate memory.” Later, in Miller, Galanter, and Pribram’s book Plans and the Structure of Behavior, the term “working memory” was brought into play.(Miller, et al., 1960) The book chapter by Baddeley and Hitch, 1974, which contained many experiments and a theoretical outline, was the most important impetus for the modern field of working memory. See Baddeley’s magnum opus for perhaps the best extent overview of memory: Baddeley, et al., 2015. Cowan, 2017 provides a good overview of not only the definitions involved in working memory, but a sense of the historical development of the field. 
[7] Cowan, 2010, Cowan, 2014.
[8] Baddeley, 2003
[9] Baddeley, 2003, Turi, et al., 2018.
[10] Metaphors and analogies such as what we’re using here are powerful teaching techniques. They allow pre-existing neural patterns to serve as a foundation to more rapidly onboard students onto new ideas. Anderson, 2010, Anderson, 2014. The great quant Emanuel Derman notes that mathematical equations themselves are simply metaphors: (Derman, 2011.)
[11] Karpicke and Grimaldi, 2012, Smith, et al., 2016.
[12] Bjork, 2018.
[13] Alloway and Alloway, 2010 explains that working memory is a relatively pure measure of a child’s learning potential and indicates a child’s capacity to learn, while academic achievement and IQ tests measure knowledge that the child has already learned. See also Shipstead, et al., 2016 and Taub, et al., 2008.
[14] Cowan, In press 2019, Ericsson, et al., 2018.
[15] Agarwal, et al., 2017, Ericsson, et al., 2018
[16] Email correspondence with the author and John Sweller, May 18, 2019.
[17] Baddeley, et al., 2015. As Baddeley notes “This is an area that is certainly worth further investigation, but I would not buy shares in it just yet!”
[18] Chen, et al., 2017
[19] Christopher and Shelton, 2017 also provides a review of the conflicting previous research results about the effect of music on studies.
[20] Cranmore and Tunks, 2015
[21] It seems that when it comes to learning, there’s an exception for nearly every rule. Brilliant mathematician John von Neumann, for example, played marching music so loudly while doing his work at Princeton that he annoyed his neighbor down the hall-Albert Einstein. (Macrae, 1992 p. 48.)
[22] Piolat, et al., 2005.
[23] Jansen, et al., 2017, Kiewra, et al., 1991. The notes should still be reviewed the same day: Liles, et al., 2018.
[24] DeCaro, et al., 2015, DeCaro, 2018, Takeuchi, et al., 2011
[25] DeCaro and Van Stockum Jr, 2018, Jarosz, et al., 2012, Wieth and Zacks, 2011
[26] Clark, et al., 2010, Raghubar, et al., 2010
[27] For a broad overview, see Dehn, 2011 p. 303, who cites three metanalyses (Adams and Engelmann, 1996, Przychodzin, et al., 2004, White, 1988) to conclude “Direct instruction is considered one of the most effective instructional methodologies for students with working memory deficiencies.” More recently, see Morgan, et al., 2015-this massive study involved 3,635 teachers and 13,883 first grade students attending 3,635 classrooms in 1,338 schools. For a more general discussion of direct instruction versus discovery and the impact of these approaches on students, see Klahr and Nigam, 2004. 
      As noted in Geary, et al., 2019 “…meta-analyses conducted by the National Mathematics Advisory Panel, 2008-and consistent with the results from Project Follow Through (Stebbins, 1977)-indicated that students with difficulties in mathematics learning benefit from explicit, teacher-directed instruction (Gersten, et al., 2008) that may help compensate for domain-general deficits.” See also Fuchs, et al., 2013 and Gersten, et al., 2009. There is also a large body of literature involving the benefit for novices of the “worked example effect,” whereby initial guidance using worked problems helps novices (which encompasses those with lower capacity working memory) more than providing no guidance. See for example Chen, et al., 2015. 
      As Dorothy Bishop notes in Nature, researchers in the sciences persist in working in a way almost guaranteed not to deliver meaningful results-all a consequence of publication bias, low statistical power, P-value hacking and HARKing (hypothesizing after results are known) (Bishop, 2019). In education, the situation is even worse: (Tyson, 2014; see Ward, 2018 for a description of Larry Hedge’s efforts to initiate change). Unfortunately, this means that finding gold-standard studies in education-studies where the plans are pre-registered so intentions can’t be altered once the data come in, independent research consultants are used for many aspects of the research process, and the study is massive-is a little like finding hens’ teeth. To those who might demand gold standard proof that direct instruction is more beneficial than student-centered instruction for those with lower capacity working memory, one might reasonably respond, where is the gold standard proof that student-centered instruction is always more beneficial than direct instruction, for all populations?
      Academic turf battles are fierce, perhaps particularly in education, where there’s often an almost defensive sense that “only we can know what’s best for children.” Although cross-disciplinary research is encouraged in many fields, in our experience, this is not the case in education, where those from other disciplines can be made to feel unwelcome unless their insights can be easily aligned with pre-existing theories and approaches. The bottom line is, no matter how definitive a finding (remembering there’s little that’s definitive or replicated in the educational literature), those who have careers vested in certain approaches have little to gain and much to lose by looking with an open mind at new research findings growing from another field, say, neuroscience, that might not support their approach. 
      For a worthwhile and relevant description of how research findings can be attacked, no matter how solid, see Nobel Prize winner Stanley Prusiner’s autobiography Madness and Memory, (Prusiner, 2014). And for a discussion of how geniuses can remain wedded to their erroneous conceptions no matter the evidence, using their intellect to justify how really, they were correct after all, see Nobel Prize winner Santiago Ramon y Cajal’s autobiography: Ramón y Cajal, 1937 (reprint 1989) Perhaps this is why, as research evidence attests, science advances funeral by funeral. (Azoulay, et al., Dec 2015).
[28] Dehn observes (Dehn, 2011): “Many educators use the term direct instruction to refer to any form of explicit teaching. However, direct instruction specifically refers to a structured curriculum that incorporates effective teaching techniques in a scripted fashion. Direct instruction involves small-group instruction; explicit teaching; fast-paced instruction; well sequenced and focused lessons; modeling and shaping of correct responses; reinforcement of appropriate responses; systematic procedures for corrective feedback; continuous assessment of performance; lots of repetition and frequent review of material; and an emphasis on mastery at each step in the learning process.”  (p. 302.) Dehn’s latter definition is how we are referring to direct instruction in this article.
[29] Sweller, et al., 2011. Reform mathematics educators observe that student-centered approaches are effective in part because they offer more representations in memory. The challenge is, if those multiple representations are not well-embedded in long-term memory, they can simply become more confusing for students with low capacity working memory.  It’s important to note that there can be effective or ineffective teaching using either student-centered or teacher-directed approaches. 
[30] Reading comprehension ability is directly related to working memory capacity: Carretti, et al., 2009.
[31] Sonnenschein, et al., 2010, Xue and Meisels, 2004
[32] Email correspondence, Beth Rogowsky and the author, May 23, 2019. 
[33] Gathercole, et al., 2006
[34] Oakley, 2013
[35] Bowles, April 21, 2019