The Knowledge Architects: Building Wisdom in the Information Age

AI-generated audio: the two voices and their conversation in this episode are created by AI. The hosts are not real people.

How we make it: our team chooses each topic and edits the audio. The research and these notes are drafted with AI and read by us before publishing.

Why AI? It lets us bring peer-reviewed science to you for free, in English and German. Sources are below.


Episode Summary

In 1932, a Cambridge psychologist asked English readers to retell a Native American folk tale from memory. They did not simply lose details. They rebuilt the story into one that made sense to them, and then remembered that version, confidently, as though it were the original. Canoes became boats, seal hunting became fishing, and the ghosts, in every single case, quietly disappeared.

Here is the part almost nobody mentions. Thirty-one years earlier, the anthropologist who collected that tale had already had it told to him twice by the same speaker, three and a half years apart, and printed both versions four pages apart in the same volume. Read them side by side and the same thing is happening. The gist holds. Almost every detail moves, down to the name of the village.

This episode is a biography of an idea. Frederic Bartlett named it schema, said in print that he disliked the word, proposed a better one that nobody adopted, and lived to see his own students write the theory off. In the same years his successor at Cambridge declared it best forgotten, artificial intelligence researchers trying to make a program understand a two sentence story about a restaurant discovered they needed exactly the thing he had described, and cited him by name. We follow the idea from the experiment as it really was, through the revival, into the physics laboratory where experts and novices sorted the same twenty-four index cards into two different worlds, and finally into the last four years, in which the mechanistic story has been taken apart largely by the people who built it. That is not a failure. It is what happens when a construct that was too vague to be wrong finally becomes specific enough to fail.


Key Topics Covered

  • The War of the Ghosts: what Bartlett actually did, and which of his findings still stand
  • The second telling: Franz Boas printed two versions of the same tale by the same speaker, Charles Cultee, three and a half years apart, and the gist held while almost every detail moved
  • Bartlett's revolt against nonsense syllables, and why he thought meaning could never be stripped out of a stimulus
  • Effort after meaning: name an ambiguous drawing a battle axe and people draw a battle axe, name it a picture frame and they draw a frame
  • The method as it really was: twenty subjects, no control condition, no fixed retention interval as stated policy, no statistics by declaration, thirteen years of collection
  • What Bartlett did to the story before using it, cutting Boas's 425 words to 330
  • The word he did not want: Henry Head's postural schema, borrowed under protest, and the rejected alternative "organised setting"
  • The burial: his own students abandoned the theory, and his successor in the Cambridge chair published that it was perhaps best forgotten
  • The revival by machines: Minsky's frames, Schank and Abelson's scripts, and the default assignments that arrive pre-filled with things you were never told
  • Where a schema gets in: at encoding, at retrieval, and at the moment of telling
  • The expert's schema: twenty-four physics problems on index cards, and the null results nobody quotes
  • Where the novice deficit actually is: the same keywords circled, with no procedure attached to the far end
  • Chi's own reversal in 2012: teach surface features harder, not less
  • The honest reckoning: Alba and Hasher in 1983, Bergman and Roediger's replication in 1999, and the Edinburgh rats that gave the word teeth
  • The dismantling of 2022 to 2026, carried out mostly by the theory's own authors
  • The U shape: prior knowledge helps most with what clearly fits and what flatly contradicts, and least with what almost fits
  • Diagnosis before explanation: random errors are gaps, consistent errors are a structure

Researchers Mentioned

  • Frederic Bartlett (1886-1969, Cambridge) : invented the study of memory for meaningful material, named the schema, disliked the name, and printed evidence against his own headline claim
  • Franz Boas (1858-1942, Columbia) : collected The War of the Ghosts from Charles Cultee and printed a second telling by the same speaker. He was checking his informant's dialect, not his memory
  • Charles Cultee (Bay Center, Washington) : one of the last speakers of Kathlamet, and the only person who ever told this story twice in its own language
  • Henry Head (London Hospital) : coined the word schema, with Gordon Holmes, for the body's running model of its own posture
  • Oliver Zangwill (Cambridge) : Bartlett's research student and successor in the chair, who told the third Bartlett Lecture that the theory was perhaps best forgotten
  • Ulric Neisser (Cornell) : the single act of rehabilitation, 1967
  • Marvin Minsky (MIT) : frames and default assignments, 1974
  • Roger Schank and Robert Abelson (Yale) : scripts, and the restaurant story that needed knowledge nobody had supplied
  • David Rumelhart (UCSD) : gave schema theory its general psychological statement
  • Gordon Bower, John Black and Terrence Turner (Stanford) : the first serious empirical test of scripts, including the result that cut the other way
  • Richard Anderson and James Pichert (Illinois) : the perspective studies, and the strongest evidence that schemas operate at retrieval
  • Michelene Chi, Paul Feltovich and Robert Glaser (Pittsburgh) : the physics categorisation study
  • Michelene Chi and Kurt VanLehn : the 2012 revision that inverted the 1981 instructional moral
  • Joseph Alba and Lynn Hasher : asked whether memory is schematic and answered, less than you think
  • Henry L. Roediger III and E.T. Bergman (Washington University) : the definitive modern replication, from the camp most hostile to Bartlett's methods
  • James Ost and Alan Costall (Portsmouth) : "Misremembering Bartlett", and the 2022 experiment it had only argued for
  • Dell Hymes : made the twice told comparison in 1984, for the volume's other double text
  • Dorothy Tse and Richard Morris (Edinburgh) : built a schema in a rat and made the construct experimental
  • Vanessa Ghosh and Asaf Gilboa (Toronto) : supplied the four feature definition that made it possible to be wrong
  • Richard Henson (MRC Cognition and Brain Sciences Unit, Cambridge) : co-authored the SLIMM model in 2012 and, in 2026, the paper reporting slim pickings for it

Key Studies and Sources

  • Bartlett, F.C. (1932). Remembering: A Study in Experimental and Social Psychology. Cambridge University Press.
  • Boas, F. (1901). Kathlamet Texts. Bureau of American Ethnology Bulletin 26. [The 1891 telling, pp. 182-184; the 1894 telling, pp. 185-186; the twice told note, p. 5.]
  • Head, H. (1920). Studies in Neurology, Vol. 2, pp. 605-608. [Reprinting Head and Holmes, Brain, 1911-12.]
  • Zangwill, O.L. (1972). "Remembering Revisited." Quarterly Journal of Experimental Psychology, 24(2), 123-138.
  • Neisser, U. (1967). Cognitive Psychology. Appleton Century Crofts.
  • Minsky, M. (1974/1975). "A Framework for Representing Knowledge." MIT AI Memo 306; in P. Winston (Ed.), The Psychology of Computer Vision.
  • Schank, R.C. and Abelson, R.P. (1975). "Scripts, Plans and Knowledge." Proceedings of the 4th IJCAI, 151-157.
  • Rumelhart, D.E. (1980). "Schemata: The Building Blocks of Cognition." In Spiro, Bruce and Brewer (Eds.), Theoretical Issues in Reading Comprehension, 33-58.
  • Rumelhart, D.E. and Norman, D.A. (1976/1978). "Accretion, Tuning and Restructuring: Three Modes of Learning." CHIP Report 7602.
  • Bower, G.H., Black, J.B. and Turner, T.J. (1979). "Scripts in memory for text." Cognitive Psychology, 11, 177-220.
  • Anderson, R.C., Reynolds, R.E., Schallert, D.L. and Goetz, E.T. (1977). "Frameworks for comprehending discourse." American Educational Research Journal, 14(4), 367-381.
  • Pichert, J.W. and Anderson, R.C. (1977). "Taking different perspectives on a story." Journal of Educational Psychology, 69(4), 309-315.
  • Anderson, R.C. and Pichert, J.W. (1978). "Recall of previously unrecallable information following a shift in perspective." Journal of Verbal Learning and Verbal Behavior, 17(1), 1-12.
  • Brewer, W.F. and Treyens, J.C. (1981). "Role of schemata in memory for places." Cognitive Psychology, 13(2), 207-230.
  • Chi, M.T.H., Feltovich, P.J. and Glaser, R. (1981). "Categorization and representation of physics problems by experts and novices." Cognitive Science, 5(2), 121-152.
  • Chi, M.T.H. and VanLehn, K.A. (2012). "Seeing Deep Structure From the Interactions of Surface Features." Educational Psychologist, 47(3), 177-188.
  • Alba, J.W. and Hasher, L. (1983). "Is memory schematic?" Psychological Bulletin, 93(2), 203-231.
  • Brady, T.F., Konkle, T., Alvarez, G.A. and Oliva, A. (2008). "Visual long-term memory has a massive storage capacity for object details." PNAS, 105(38), 14325-14329.
  • Bergman, E.T. and Roediger, H.L. III (1999). "Can Bartlett's repeated reproduction experiments be replicated?" Memory and Cognition, 27(6), 937-947.
  • Ost, J., Udell, J., Dear, S., Zinken, J., Blank, H. and Costall, A. (2022). "The serial reproduction of an urban myth: revisiting Bartlett's schema theory." Memory, 30(6), 775-783.
  • Tse, D., Langston, R.F., Kakeyama, M., Bethus, I., Spooner, P.A., Wood, E.R., Witter, M.P. and Morris, R.G.M. (2007). "Schemas and memory consolidation." Science, 316(5821), 76-82.
  • Ghosh, V.E. and Gilboa, A. (2014). "What is a memory schema? A historical perspective on current neuroscience literature." Neuropsychologia, 53, 104-114.
  • Quent, J.A., Greve, A. and Henson, R.N. (2022). "Shape of U: The Nonmonotonic Relationship Between Object Location Memory and Expectedness." Psychological Science, 33(12), 2084-2097.
  • Samiei, M., Precup, D. and Richards, B.A. (2026). "The schema spectrum: Emergent structures and levels of abstraction in AI and the brain." Neuron, 114(11), 1898-1907.
  • Simonsmeier, B.A., Flaig, M., Deiglmayr, A., Schalk, L. and Schneider, M. (2022). "Domain-Specific Prior Knowledge and Learning: A Meta-Analysis." Educational Psychologist, 57(1), 31-54.

Key Numbers to Remember

  • 20 subjects : seven women and thirteen men, the entire sample of the famous experiment
  • 0 control conditions and 0 statistics : "In this book there will be no statistics whatever" (Bartlett, p. 9)
  • 13 years : roughly 1914 to 1927, the span over which the data were collected
  • 425 words to 330 : Boas's English translation, cut and tidied by Bartlett before use
  • Every subject turned canoes into boats, and no subject kept the ghosts past a single retelling
  • 3.5 years apart, 4 pages apart : Charles Cultee's two tellings, printed in the same volume in 1901
  • 62% : readers who reported that a second interpretation never occurred to them (Anderson et al. 1977)
  • 0.11 : mean rank correlation between two perspectives on which parts of the same text matter (Pichert and Anderson 1977)
  • 7.1% : extra information recalled when the perspective was switched after recall was complete (Anderson and Pichert 1978)
  • 0.29, 0.39, 0.58 : major distortions as a proportion of everything recalled at fifteen minutes, one week and six months (Bergman and Roediger 1999; the six month figure rests on the eight participants who returned)
  • 8.4 against 8.6 : number of piles made by experts and novices sorting physics problems, with the experts taking 18 minutes against 12
  • 7 of 8 novices against 6 of 8 experts : the same index card filed with inclined planes or with a horizontal spring problem
  • 87.6% : recognition of one object against a different exemplar of the same category, after viewing 2,500 objects for three seconds each (Brady et al. 2008)
  • 1 trial, about 2 days : a new flavour and place learned once inside an established schema, and retained after the hippocampus was removed (Tse et al. 2007)
  • 60% : the ceiling in the control arena, where the layout was reshuffled every two sessions
  • 493 studies, 126,050 learners : prior knowledge correlates about 0.53 with what a learner knows at the end, and about zero with normalized gain (Simonsmeier et al. 2022)

Memorable Quotes

"The psychologist, of all people, must not stand in awe of the stimulus."
Frederic Bartlett, Remembering (1932), p. 3

"I strongly dislike the term 'schema'. It is at once too definite and too sketchy."
Bartlett, pp. 200-201, on the word he made famous

"One might almost as well say that because nobody who is suffering from a raging toothache could calmly recite 'Oh, my love's like a red, red rose', the teeth are a repository of lyric poetry."
Bartlett, p. 200, on the idea that the brain is a storehouse

"Remembering is not the re-excitation of innumerable fixed, lifeless and fragmentary traces. It is an imaginative reconstruction, or construction."
Bartlett, p. 213

"Every subject, at some point in the story, introduced 'boats' for canoes."
Bartlett, p. 82

"They show great similarity."
Franz Boas, Kathlamet Texts (1901), p. 5, on the two tellings. His verdict is about the language, not about the memory

"The theory, in my view never very plausible, is perhaps best forgotten."
Oliver Zangwill (1972), Bartlett's own student and successor

"Out of a few stored bone chips, we remember a dinosaur."
Ulric Neisser, Cognitive Psychology (1967)

"62% of the subjects reported that another interpretation never occurred to them."
Anderson, Reynolds, Schallert and Goetz (1977)

"Evidence seems to indicate that the memory representation is far richer and detailed than schema theory would suggest."
Alba and Hasher (1983)

"After 6 months, most of what people recalled of the story was distorted in some form."
Bergman and Roediger (1999)

"A direct instructional implication of our hypothesis is to teach students explicitly to focus on the interactions among the relevant literal surface features as a way to 'see' and understand the deep structure."
Chi and VanLehn (2012), revising her own 1981 conclusion

The Big Idea

Learning is not addition. It is assimilation. What you already have decides what you can take in, what you quietly rewrite, and what you cannot see at all. Ninety years of evidence support the phenomenon: prior knowledge shapes what gets encoded, what counts as important, what can be retrieved, and what seems worth saying out loud. What has never been stable is the machinery underneath. Bartlett borrowed the word schema from a neurologist he was in the middle of contradicting, disliked it in print, and proposed a better one nobody used. The idea was buried by his own students, rebuilt from scratch by engineers who needed it, stretched until it explained everything and therefore forbade nothing, and then slowly, painfully, made specific enough to be proven wrong in interesting places. The practical version is narrower than the slogan and more useful than it: the question worth asking is not what someone needs to know, but what is already in there, and what it will do to the new material.


Next Episode Preview

Episode 34: Mental Models : A schema tells you what kind of thing you are looking at. It does not tell you what that thing will do. Next time we turn to the 1943 proposal that the mind builds small working models of the world and runs them, so that we can try out the future in our heads before it happens.


What is The Knowledge Architects: Building Wisdom in the Information Age?

AI-generated audio: both hosts and their conversations are created by AI, not by real people.

The Knowledge Architects is a free, science-based podcast exploring how we learn, remember, and organize knowledge. Each episode translates peer-reviewed research from cognitive science, neuroscience, and psychology into practical insights, helping you understand how your mind works and how to work with it more effectively. Our team chooses every topic and edits each episode. The research is drafted with AI and read by us before publishing. Brought to you by ElysFlow.