Verbal Learning

Abstract

Verbal learning, a type of learning, is the experimental study of how people acquire, retain, and recall material composed of words and word-like units, the tradition that turned memory into a laboratory science. This article traces it from Hermann Ebbinghaus, who invented the nonsense syllable and the savings method to time his own forgetting, through the anticipation method and the paired-associate curve, the serial-position curve and its two-store interpretation, the interference theory that explained forgetting as competition rather than decay, and the shift toward organization, depth of processing, and retrieval that carried the field into modern memory research. Three interactive demonstrations let the reader trace a serial-position curve and watch its recency arm collapse under a filled delay, drive a paired-associate learning curve with the meaningfulness of the material, and find the optimal spacing gap that shifts with how far off the test is.

Keywords: verbal learning, nonsense syllable, serial-position curve, paired-associate learning, interference

Verbal learning is the branch of experimental psychology that studies how verbal material, whether single words, nonsense syllables, or paired items, is acquired through practice, retained over time, and recovered at test. In the Medical Subject Headings vocabulary it is catalogued as descriptor D014706, defined as learning to respond verbally to a verbal stimulus cue. For roughly the first century of scientific memory research it was very nearly synonymous with the study of memory itself, because words were the material an experimenter could count, control, and repeat, and because a nonsense syllable could be scaled for difficulty in a way that a lived experience could not. The tradition began with a single investigator memorising lists of his own devising and ended by handing modern cognitive psychology most of its durable phenomena: the forgetting curve, the serial-position curve, the distinction between recall and recognition, the role of organization, and the effects of spacing and testing on retention. The sections below follow that arc: what verbal learning is and why it took the form it did, the nonsense syllable and the anticipation method that made it measurable, the serial-position curve and the paired-associate curve, the interference theory of forgetting, the turn toward organization and depth of encoding, the retrieval principles that reframed the whole enterprise, the benefits of distributed practice, and the demonstration that verbal learning can manufacture memories of things that were never presented.

What Verbal Learning Is

Verbal learning is defined by its material rather than by any single mechanism: it is the study of how sequences and pairings of verbal units are learned and remembered. That apparently narrow focus was a deliberate methodological choice, and it shaped a century of research. When Hermann Ebbinghaus set out in the 1880s to bring memory into the laboratory, the obstacle was that ordinary remembering is hopelessly confounded. A person already knows some words better than others, has richer associations to some than to others, and cannot be given a truly novel thing to learn twice over.

Ebbinghaus solved the problem by inventing material with as little inherent meaning as he could manage, the nonsense syllable, and by using himself as a tireless single subject, learning and relearning thousands of lists under rigidly constant conditions (Ebbinghaus, 1885/2013). The payoff was that learning could now be quantified: the number of repetitions to reach errorless recitation indexed acquisition, and the reduction in that number on a later relearning indexed retention. Memory had become an experimental variable.

The tradition that grew from this had a characteristic style. It favoured lists over prose, controlled units over natural language, recall and recognition scored as counts, and functional laws relating a manipulated variable, such as list length, presentation rate, or the similarity between lists, to a measured outcome, such as trials to criterion or percentage recalled. This is the study of learning at its most stripped-down, and its severity was the source of both its power and its eventual limits. The power was replicability and cumulative laws; the limit was that meaning, the very thing Ebbinghaus had engineered out, turned out to be central to how humans actually remember, and the field spent its later decades letting meaning back in. Table 1 sets out the landmark findings the tradition produced and the paradigms that yielded them.

Table 1

Landmark Findings of the Verbal-Learning Tradition

Note. The tradition's durable phenomena and the paradigms that isolated them. Each is developed in a later section; full citations appear in the References (Ebbinghaus, 1885/2013; Murdock, 1962; Glanzer & Cunitz, 1966; Underwood, 1957; Craik & Lockhart, 1972; Tulving & Thomson, 1973; Cepeda et al., 2006).

Nonsense Syllables and the Anticipation Method

The nonsense syllable, typically a consonant-vowel-consonant string such as DAX or JIR, was verbal learning's foundational instrument, but it was never as neutral as its name promised. J. Arthur Glaze had a large group of observers rate thousands of such syllables and found that they differed enormously in what he called their association value, the proportion of people to whom a syllable suggested a word, an image, or a meaning within a fixed interval; ZAT evokes far more than XIH (Glaze, 1928). Because high-association-value syllables are learned faster, association value had to be measured and controlled rather than assumed to be zero, and Glaze's norms became a standard tool. Clyde Noble later put the idea on a continuous footing with his measure m, the meaningfulness of an item defined by the average number of written associations it elicits in a minute, and showed that m predicts the speed of learning across the whole range from nonsense syllables to familiar words (Noble, 1952). Meaningfulness, in other words, is not a property a stimulus either has or lacks but a graded dimension that governs learning.

The engine that drove most of this work was the anticipation method, illustrated in Figure 1. Items were presented one at a time in a fixed order by a memory drum, a mechanical device that exposed each item in a window for a constant interval. In serial learning the participant's task, as each item appeared, was to anticipate the next one before it came into view; in paired-associate learning the stimulus term of a pair appeared alone and the participant tried to produce its response term before the whole pair was shown for feedback. Each pass through the list was a trial, and learning was tracked as the growing proportion of correct anticipations across trials until the list was recited without error. The method fixed the order, the timing, and the feedback, so that the only thing free to vary was the participant's memory, and it yielded the two workhorses of the field, the paired-associate learning curve and the serial-position curve, examined in the next two sections.

The Serial-Position Curve

When a list of words is presented once and the participant is free to recall them in any order, the probability that a given word is recalled depends strongly on where it sat in the list. Plotting recall against serial position yields the serial-position curve, one of the most reliable functions in psychology: recall is high for the first few items, falls to a low plateau across the middle, and rises again for the last few. Bennet Murdock mapped the curve in careful detail across lists of different lengths and presentation rates, establishing its shape as a lawful regularity rather than an artifact (Murdock, 1962). The two raised ends acquired names and, more importantly, competing explanations. The advantage of the first items is the primacy effect, generally attributed to rehearsal: early items enter an empty list, are rehearsed more often, and are more likely to be transferred to a durable store. The advantage of the last items is the recency effect, attributed to their still being held in a limited-capacity short-term store at the moment recall begins.

The decisive evidence for that two-store reading came from a simple manipulation. Murray Glanzer and Anita Cunitz showed that filling the interval between the end of the list and the start of recall with a distracting task, counting backwards for a few seconds, selectively abolished the recency effect while leaving primacy and the middle of the curve untouched (Glanzer & Cunitz, 1966). The interpretation was that the distractor task displaced the final items from the short-term store before they could be reported, whereas the earlier items, already consolidated in a more durable store, were unaffected. A variable that moved one arm of the curve without moving the other, a dissociation, was powerful evidence that the two arms reflect two different memory systems rather than one continuum, and the finding became a cornerstone of the modular models of memory that dominated the following decade, connecting verbal learning to the study of working memory and long-term memory. The two-store reading is not the whole story, however. Robert Bjork and William Whitten showed that a recency advantage returns in long-term free recall when the intervals between the items, and not only the interval before recall, are filled with a distracting task, a continuous-distractor procedure in which recency survives a delay that should have emptied any short-term store; recency therefore cannot be read straight off short-term storage but reflects, at least in part, the relative distinctiveness of the most recent items at the moment of retrieval (Bjork & Whitten, 1974).

Serial position is not the only thing that lifts an item above the low middle of the curve. An item made distinctive, printed in a colour unlike its neighbours or drawn from a different category than the rest of the list, is recalled better than its position alone would predict, the isolation effect that Hedwig von Restorff demonstrated in 1933; recall tracks how far an item stands out from its context as well as where it fell in the sequence (von Restorff, 1933). The demonstration below builds the serial-position curve for a list whose length the reader can set, and lets the recency arm be switched off by imposing a filled delay, reproducing the Glanzer and Cunitz dissociation directly.

Meaningfulness and the Paired-Associate Curve

Where free recall exposes how position shapes memory, paired-associate learning exposes how the material itself shapes the rate of acquisition. In the anticipation method the participant is shown a stimulus term and must produce its paired response before the answer appears, and the proportion of correct anticipations climbs across trials as a negatively accelerated learning curve, fast at first and then flattening as the list approaches mastery. The shape is the same across materials, but its steepness is not, and what sets the steepness is meaningfulness. High-association-value pairs, made of familiar and imageable words, are learned in a handful of trials; low-meaning pairs of nonsense syllables can take several times as many (Noble, 1952; Glaze, 1928). This is why Ebbinghaus's choice of deliberately meaningless material made his lists so laborious to learn, and why meaningfulness had to be measured and equated whenever an experimenter wanted to study something else. The lawful relation between the meaningfulness of a unit and the speed with which it enters memory was one of the tradition's most robust quantitative findings, and it was an early sign that the mind does not treat all verbal material alike but leans on whatever pre-existing structure it can find.

Meaningfulness is not the only property of the material that matters. Allan Paivio showed that a word's capacity to evoke a mental image is an equally powerful lever on paired-associate learning: concrete, high-imagery pairs such as dog–table are learned far faster than abstract pairs such as truth–reason, and his dual-coding theory explained the advantage by proposing that concrete items are stored twice, in a verbal code and an imaginal one, giving retrieval two independent routes rather than one (Paivio, 1969). The demonstration below draws the paired-associate learning curve and lets the reader vary the meaningfulness of the material, watching the whole curve steepen and the number of study trials needed to reach a mastery criterion fall as the material becomes more meaningful.

Interference and the Causes of Forgetting

Ebbinghaus's forgetting curve established that retention falls with time, but it left open why. The intuitive answer, that memory traces simply fade with disuse, was the classical law of disuse, and John McGeoch dismantled it in an influential critique: time itself causes nothing, he argued, and what fills the interval between learning and test, not its mere duration, determines how much is forgotten (McGeoch, 1932). The alternative he pointed toward was interference, the idea that memories compete, and the demonstration came from the study of retroactive inhibition, in which material learned after the target degrades memory for it. Arthur Melton and Jean Irwin showed that the amount of retroactive interference grows with how much interpolated material is learned, and, crucially, that part of the effect could not be explained by the intrusion of the new responses into recall, pointing to a second factor they called unlearning, an active weakening of the original associations by the interpolated learning (Melton & Irwin, 1940).

The mechanism was dissected with transfer paradigms, in which the same stimulus terms are paired first with one set of responses and then with a new set, the A–B, A–C design that pits old and new associations against each other. Leo Postman showed that the amount and even the direction of transfer between such lists depend systematically on the experimental paradigm and on how thoroughly the first list was learned, making transfer the analytic tool through which interference theory could specify when prior learning should aid later learning and when it should degrade it (Postman, 1962).

Benton Underwood then turned the picture around with an observation that reoriented the field. Reviewing decades of studies, he noticed that the amount forgotten of a single laboratory list over a day was far less than the classic curves implied, and that most of the forgetting in those older studies came not from the target list but from the many lists the practised subjects had learned beforehand. Forgetting, he concluded, is largely proactive: earlier learning interferes with the retention of later learning (Underwood, 1957). This shifted the explanatory weight from the passage of time to the accumulated history of the learner. Underwood and Leo Postman pushed the logic outside the laboratory, proposing that the linguistic habits a person brings from everyday life, extraexperimental sources of interference, should compete with newly learned lists, though this strong prediction proved harder to confirm than the within-experiment effects (Underwood & Postman, 1960). Interference theory became the dominant account of forgetting for a generation, and Postman and Underwood's later synthesis laid out both its achievements and the unresolved problems, including the fate of the unlearned associations and the mechanisms of recovery, that eventually limited it (Postman & Underwood, 1973). The link back to Ebbinghaus's forgetting curve is direct: the curve describes forgetting, interference theory explains it.

Organization, Rehearsal, and Depth

By the 1960s the field was discovering that learners are not passive registers of whatever is presented but active organisers of it, and that how material is processed matters more than how many times it is repeated. Weston Bousfield gave the first clean demonstration. When people freely recall a list of words that, unknown to them, is drawn from a few semantic categories such as animals, names, and professions, they do not recall it in the order of presentation but cluster the items by category, recovering animals together and names together even though the list was scrambled (Bousfield, 1953). The clustering revealed that recall is structured by the learner's own semantic organization. Gordon Bower and colleagues turned the observation into a powerful manipulation: presenting a word list arranged as an explicit conceptual hierarchy, rather than randomly, multiplied recall several times over, demonstrating that imposed organization is one of the strongest levers on memory (Bower, Clark, Lesgold, & Winzenz, 1969).

Rehearsal came under similar scrutiny. Dewey Rundus had participants rehearse aloud during a free-recall list and showed that the number of times an item was rehearsed predicted its later recall, and that the pattern of overt rehearsal traced the primacy effect directly, giving the rehearsal account of primacy an observable basis (Rundus, 1971). But repetition alone proved insufficient, and Fergus Craik and Robert Lockhart reframed the whole question with their levels-of-processing proposal: memory is a by-product of the depth at which material is encoded, so that attending to a word's meaning (deep, semantic processing) leaves a far more durable trace than attending to its sound or its appearance (shallow processing), regardless of intention to learn or of sheer repetition (Craik & Lockhart, 1972). A complementary finding sharpened the point: Norman Slamecka and Peter Graf showed that words a person generates for themselves, given a rule and a fragment, are remembered better than the same words simply read, the generation effect, evidence that the effort of production leaves a stronger trace than passive exposure (Slamecka & Graf, 1978). Together these results moved the causes of remembering decisively from the number of repetitions toward the quality of encoding.

Retrieval, Availability, and Encoding Specificity

If encoding determines what enters memory, retrieval determines what comes back out, and Endel Tulving showed that the two must be analysed separately. He and Zena Pearlstone had people learn long categorised word lists and then tested recall either with no cues or with the category names as cues, and found that the cued group recalled far more. The words the uncued group failed to produce were nonetheless available in memory, as the cues proved by eliciting them; they had simply been inaccessible without the right prompt. Tulving and Pearlstone drew the enduring distinction between availability, whether an item is stored at all, and accessibility, whether a given retrieval situation can reach it, and showed that a recall failure need not mean the memory is gone (Tulving & Pearlstone, 1966).

The natural next question was what makes a cue effective, and the answer overturned a comfortable assumption. Tulving and Donald Thomson formulated the encoding-specificity principle: a retrieval cue aids recall only to the extent that its relation to the target was encoded at the time of learning, so the value of a cue is fixed not by its general association to the target but by what was actually stored about the two together (Tulving & Thomson, 1973). The striking consequence, which they demonstrated, is that a strong pre-existing associate can fail as a cue while a weak one that was present at encoding succeeds, and that recognition can even fail for words that are successfully recalled from the original cue, contradicting the then-standard view that recognition is simply an easier form of retrieval than recall. Memory, on this account, is not a matter of the strength of an isolated trace but of the match between the conditions of encoding and the conditions of retrieval, a principle that governs cued recall, recognition, and the everyday experience of a cue suddenly unlocking a memory that moments before seemed lost.

Distributed Practice and the Spacing Effect

One of the oldest and most practically important findings in verbal learning is that the scheduling of study matters as much as its total amount. For a fixed number of repetitions, spreading them out in time produces better long-term retention than massing them together, the spacing effect, a result Ebbinghaus already glimpsed and that has since been confirmed across an enormous range of materials and intervals. Nicholas Cepeda and colleagues synthesised the large literature quantitatively and established a further regularity that the everyday advice to space one's study conceals: the optimal gap between repetitions is not fixed but scales with the retention interval, the delay until the test. When the test is soon, a short gap is best; when the test is far off, the best gap is much longer, and a gap ideal for a test in a week is too short for a test months away (Cepeda, Pashler, Vul, Wixted, & Rohrer, 2006). The relation between the optimal gap and the retention interval is lawful enough to guide the scheduling of practice.

Closely allied is the testing effect, the finding that retrieving material is a more powerful aid to later retention than restudying it for the same time. Henry Roediger and Jeffrey Karpicke had people learn prose passages and then either restudy them or take practice tests, and found that although restudy felt more effective and produced better performance on an immediate test, testing produced substantially better retention when the final test came days later, a dissociation between what feels effective in the moment and what actually endures (Roediger & Karpicke, 2006). Spacing and testing together form the empirical core of what is now called desirable difficulty: manipulations that make learning feel harder and slower in the short run often produce more durable memories in the long run. The demonstration below draws the spacing function for a fixed retention interval and lets the reader move the interval, watching the peak of the curve, the optimal gap, slide outward as the test is pushed further away.

Distortion and the DRM Illusion

The verbal-learning tradition also produced one of the most vivid demonstrations that memory is constructive rather than reproductive, and it did so with its own oldest tool, the word list. The paradigm was the creation of James Deese, who in 1959 found that a study list composed of the strongest associates of a single non-presented word makes that missing word intrude into immediate recall at a high and predictable rate (Deese, 1959). Henry Roediger and Kathleen McDermott revived and sharpened it into a reliable false-memory tool, having participants study a list of words all strongly associated to a single word that is itself never presented, hearing bed, rest, awake, tired, dream, and the like, but never sleep. At test, participants recall and recognise the missing associate, the critical lure, about as often and with as much confidence as the words that were actually on the list (Roediger & McDermott, 1995). The false memory is not a guess; people report vividly remembering having heard the word, and the effect is robust even when they are warned about it. The result showed that the same associative and organizational processes that ordinarily aid recall, the spreading of activation among related words, can generate confident memories of events that never occurred, and it made the humble word list a workhorse of the modern study of false memory. That a tradition built to measure the faithful acquisition of verbal material should end by demonstrating, with the same lists, how readily memory fabricates, is a fitting measure of how far the field travelled from Ebbinghaus's austere savings scores.

Worked Example

The serial-position curve is worth computing by hand, because it makes the two-store interpretation concrete and it is exactly what the serial-position demonstration draws. Model the probability of recalling the item at position i in a list of n items as the sum of a mid-list floor, a primacy term that decays from the start of the list, and, on an immediate test, a recency term that decays from the end:

p(i) = 0.30 + 0.55 × exp(−(i − 1) / 2.0) + 0.60 × exp(−(n − i) / 1.5)

with the recency term dropped after a filled delay. Take a list of n = 12. For the first item, i = 1, the primacy term is 0.55 × exp(0) = 0.55 and the recency term is 0.60 × exp(−11 / 1.5) = 0.60 × 0.0007, which is negligible, so p(1) = 0.30 + 0.55 + 0.00 = 0.85, an 85 percent chance of recall. For a middle item, i = 6, the primacy term is 0.55 × exp(−5 / 2.0) = 0.55 × 0.082 = 0.045 and the recency term is 0.60 × exp(−6 / 1.5) = 0.60 × 0.018 = 0.011, giving p(6) = 0.30 + 0.045 + 0.011 = 0.36, about 36 percent, the low plateau. For the last item, i = 12, the primacy term has decayed to 0.55 × exp(−11 / 2.0) = 0.55 × 0.004 = 0.002 while the recency term is at its maximum, 0.60 × exp(0) = 0.60, so p(12) = 0.30 + 0.002 + 0.60 = 0.90, about 90 percent. The curve therefore falls from 85 percent at the start to 36 percent in the middle and climbs back to 90 percent at the end, the familiar U.

Now impose a filled delay, which empties the short-term store and removes the recency term. The primacy arm is untouched: p(1) is still 0.30 + 0.55 = 0.85. But the last item loses its recency contribution entirely, so p(12) = 0.30 + 0.002 = 0.30, collapsing from 90 percent to about 30 percent, while the middle item barely moves. A single manipulation has flattened one arm of the curve and left the other standing, which is precisely the Glanzer and Cunitz dissociation and precisely why the two arms were read as two stores. Setting the demonstration to a 12-item list reproduces these values; switching it from immediate recall to a filled delay reproduces the selective collapse of recency, with primacy fixed, exactly as the arithmetic predicts.

Discussion

Verbal learning is the tradition in which the psychology of memory learned to measure itself. Its founding move, Ebbinghaus's decision to strip his material of meaning so that acquisition could be counted from a controlled baseline, bought a century of cumulative, replicable laws: the forgetting curve, the paired-associate and serial-position curves, the graded effect of meaningfulness, the interference account of forgetting, and the benefits of distributed practice (Ebbinghaus, 1885/2013; Murdock, 1962; Noble, 1952; Underwood, 1957; Cepeda et al., 2006). The same move set the field's eventual limit. Meaning, engineered out at the start, proved to be the hinge on which human memory turns, and the later history of verbal learning is the record of letting it back in: the discovery that learners cluster and organise material for themselves, that depth of processing beats repetition, that self-generated items outlast read ones, and that retrieval depends on the match between the cue and what was encoded (Bousfield, 1953; Craik & Lockhart, 1972; Slamecka & Graf, 1978; Tulving & Thomson, 1973). By the 1970s the austere study of lists had opened onto the questions that define modern memory research: the organization of knowledge, the distinction between availability and accessibility, the constructive and sometimes fabricating character of recall (Tulving & Pearlstone, 1966; Roediger & McDermott, 1995). The vocabulary shifted from verbal learning to memory, and the memory drum gave way to the computer; the field marked the change in its own masthead when the Journal of Verbal Learning and Verbal Behavior, in whose pages many of the findings above first appeared, was renamed the Journal of Memory and Language in 1985. But the phenomena the tradition isolated did not go away. The serial-position curve, the spacing effect, the testing effect, and encoding specificity remain among the most robust and practically consequential findings in the science of the mind, and the interactive curves above are illustrative renderings of the same phenomena Ebbinghaus, Murdock, and Cepeda first measured on a memory drum.

Common Misconceptions

Forgetting is the fading of a memory trace with the passage of time.

Time itself causes nothing; what fills the retention interval does. Most laboratory forgetting is interference from other material learned before or after the target, not decay, which is why prior learning can degrade later retention as much as later learning degrades earlier (McGeoch, 1932; Underwood, 1957).

Nonsense syllables are all equally meaningless.

Consonant-vowel-consonant syllables vary widely in association value, the degree to which they suggest a word or image, and higher-value syllables are learned faster. Meaningfulness is a graded dimension that had to be measured and controlled, not a property that was simply absent (Glaze, 1928; Noble, 1952).

A failure to recall an item means the memory is gone.

A recall failure often reflects inaccessibility rather than absence. Items that cannot be recalled without help are frequently produced at once when the right cue is given, showing that they were available in memory all along and only needed a matching prompt to be reached (Tulving & Pearlstone, 1966).

Massing study and rereading are the most effective ways to learn.

They feel effective but produce poor long-term retention. Spacing repetitions and testing oneself both feel harder and slower in the moment yet yield substantially more durable memory than cramming or rereading, a gap between felt and actual learning (Cepeda et al., 2006; Roediger & Karpicke, 2006).

Glossary

Anticipation method.

A procedure in which list items are exposed one at a time at a fixed rate and the participant must produce the next item, or a pair's response term, before it appears, tracking learning as the rising proportion of correct anticipations.

Association value.

The proportion of people for whom a nonsense syllable suggests a word, image, or meaning within a set time; Glaze's index of how much latent meaning a nominally meaningless unit carries.

Clustering.

The tendency, in free recall of a categorised but scrambled list, to recover items grouped by semantic category rather than in the order presented, revealing the learner's own organization.

DRM paradigm.

A false-memory procedure, named for Deese, Roediger, and McDermott, in which studying a list of words all associated to one non-presented word (the critical lure) leads participants to recall and confidently recognise that lure as if it had been on the list.

Dual coding.

Paivio's theory that concrete, high-imagery words are stored in two codes, a verbal one and an imaginal one, giving them two independent retrieval routes and making them easier to learn than abstract words.

Encoding specificity.

The principle that a retrieval cue aids recall only to the extent that its relation to the target was encoded at study, so a cue's value depends on the study context, not on general association.

Free recall.

A test in which the participant recalls the items of a studied list in any order; the task that yields the serial-position curve and clustering.

Generation effect.

The finding that words a person produces for themselves from a rule and a fragment are remembered better than the same words merely read, evidence that active production strengthens the trace.

Isolation effect.

The superior recall of a list item that stands out from its neighbours in some salient way, such as colour or category; also called the von Restorff effect, it shows that distinctiveness raises recall independently of serial position.

Levels of processing.

The framework in which retention is a by-product of the depth of encoding, so that semantic (deep) processing produces more durable memory than shallow attention to a word's sound or appearance.

Meaningfulness (m).

Noble's continuous measure of a verbal unit, defined by the average number of associations it elicits in a fixed interval; it predicts the rate of learning across the range from nonsense syllables to words.

Nonsense syllable.

A consonant-vowel-consonant string such as DAX, devised by Ebbinghaus to provide verbal material with minimal pre-existing meaning so that new learning could be measured from a controlled baseline.

Paired-associate learning.

Learning to produce a response term when shown its paired stimulus term; the proportion correct rises across trials as a negatively accelerated curve whose steepness depends on meaningfulness.

Primacy effect.

The superior recall of the first items in a list, attributed to their receiving more rehearsal and being more likely to enter a durable long-term store.

Proactive interference.

The degradation of memory for material by earlier learning; Underwood's evidence that much laboratory forgetting comes from lists learned before the target, not from time.

Recency effect.

The superior recall of the last items in a list, attributed to their still being held in a limited short-term store at recall; abolished by a filled delay.

Retroactive interference.

The degradation of memory for material by later learning, shown to grow with the amount of interpolated learning and to involve an active unlearning of the original associations.

Savings method.

Ebbinghaus's measure of retention: the reduction in trials or time needed to relearn a list compared with learning it originally, sensitive to memory that recall alone cannot detect.

Serial-position curve.

The U-shaped function relating recall probability to an item's position in the list, high at the start (primacy) and end (recency) and low in the middle.

Spacing effect.

The advantage in long-term retention of spreading a fixed number of study repetitions over time rather than massing them; the optimal gap grows with the retention interval.

Testing effect.

The finding that retrieving material in a test improves later retention more than restudying it for the same time, despite feeling less effective in the moment.

Transfer paradigm.

An experimental design in which the same stimulus terms are re-paired with new responses (the A-B, A-C form), used to measure how prior learning aids or interferes with later learning and to dissect the mechanisms of interference.

Verbal learning.

The experimental study of how verbal material, whether nonsense syllables, words, or pairs, is acquired, retained, and recalled; the tradition that made memory a laboratory science.

Key Researchers

Gordon H. Bower (1932-2020). Longtime professor at Stanford University; he showed that organization is one of the strongest determinants of recall, demonstrating that presenting a word list as an explicit conceptual hierarchy multiplies recall several times over a random arrangement. Wikipedia - Wikidata

Fergus I. M. Craik. Senior scientist at the Rotman Research Institute, Baycrest, and professor at the University of Toronto; with Robert Lockhart he proposed the levels-of-processing framework, in which memory is a by-product of the depth to which material is encoded. Faculty Page - ORCID - Google Scholar - Wikipedia

Hermann Ebbinghaus (1850-1909). Working at Berlin, Breslau, and Halle; he founded the experimental study of memory, inventing the nonsense syllable and the savings method and deriving the forgetting curve, the learning curve, and the advantage of distributed over massed practice by exhaustive experiments on himself. Wikipedia - Wikidata

Leo Postman (1918-2004). Professor at the University of California, Berkeley; a central figure of the verbal-learning tradition and, with Underwood, of interference theory, cataloguing the conditions of retroactive and proactive interference and pressing the field's hardest theoretical questions. Wikipedia - Wikidata - Memorial

Henry L. Roediger III. James S. McDonnell Distinguished University Professor at Washington University in St. Louis; he revived the DRM false-memory paradigm, in which an associated word list reliably produces confident recall of a non-presented lure, and established the testing effect, in which retrieval practice outperforms restudy for long-term retention. Faculty Page - ORCID - Google Scholar - Wikipedia

Endel Tulving (1927-2023). Professor at the University of Toronto and the Rotman Research Institute; he turned the study of recall toward retrieval, distinguishing availability from accessibility and formulating the encoding-specificity principle, that a cue aids recall only to the extent its information was encoded with the target. Faculty Page - Google Scholar - Wikipedia

Benton J. Underwood (1915-1994). Professor at Northwestern University; he reoriented the explanation of forgetting from decay to interference, showing that most forgetting of a laboratory list is proactive, caused by prior lists the participant had already learned, rather than by the passage of time. Wikipedia - Wikidata - Biographical Memoir

Frequently Asked Questions

What is verbal learning?
Verbal learning is the experimental study of how people acquire, retain, and recall material made of words and word-like units such as nonsense syllables and paired items (Ebbinghaus, 1885/2013). For much of the twentieth century it was effectively the science of memory, because verbal material could be controlled, scaled for difficulty, and scored as counts in a way that natural experience could not.

What is a nonsense syllable?
A nonsense syllable is a short consonant-vowel-consonant string such as DAX or JIR, invented by Ebbinghaus to provide learning material with as little pre-existing meaning as possible (Ebbinghaus, 1885/2013). They are not truly meaningless, however: they vary in association value, and higher-value syllables are learned faster (Glaze, 1928).

What is the serial-position curve?
It is the U-shaped relation between an item's position in a studied list and the probability of recalling it: the first items (primacy) and the last items (recency) are recalled best, the middle worst (Murdock, 1962). A filled delay before recall abolishes the recency arm while sparing primacy, evidence for separate short- and long-term stores (Glanzer & Cunitz, 1966).

Why do we forget verbal material?
Chiefly because of interference, not decay. What fills the interval between learning and test matters more than its duration, and other material learned before (proactive) or after (retroactive) the target competes with it at recall (McGeoch, 1932; Underwood, 1957).

What is the difference between availability and accessibility?
Availability is whether an item is stored in memory at all; accessibility is whether a particular retrieval situation can reach it. Items that cannot be recalled unaided are often produced at once when a matching cue is given, so a recall failure need not mean the memory is gone (Tulving & Pearlstone, 1966).

Does spacing out study really help?
Yes. For a fixed number of repetitions, spreading them over time produces better long-term retention than massing them, and the best gap grows with how far off the test is (Cepeda et al., 2006). Testing oneself is likewise more effective for retention than restudying, though it feels harder (Roediger & Karpicke, 2006).

How does organization affect memory?
Strongly. People spontaneously cluster recall by semantic category even from a scrambled list, and presenting a list as an explicit hierarchy can multiply recall several times over a random order, making imposed organization one of the most powerful aids to memory (Bousfield, 1953; Bower et al., 1969).

Can a word list create a false memory?
Yes. When every word on a list is associated to one word that is never presented, people recall and confidently recognise that missing word about as often as the words actually studied, a demonstration that ordinary associative processes can fabricate memories (Roediger & McDermott, 1995).

References