When Frances Yates set out to recover the classical art of memory, she found herself working from remarkably thin textual evidence. Three Latin sources had survived antiquity with substantial accounts of the mnemonic technique: the anonymous Ad Herennium, long attributed to Cicero; Cicero's own dialogue De oratore; and Quintilian's pedagogical treatise Institutio oratoria. The Greek tradition in which the art almost certainly originated was, for practical purposes, lost. What Yates does in the opening chapter of The Art of Memory (1966) is not merely summarize these three accounts but read them against each other, identifying where they reinforce and where they complicate our picture of what the classical art actually was. The result is a portrait of a technique that was already, in its earliest surviving documentation, both systematically described and quietly disputed.
Simonides and the architecture of memory
The origin myth of the art of memory is vivid enough to have survived two millennia of retelling. Cicero records it in De oratore, and Yates gives it pride of place near the start of her analysis. The poet Simonides of Ceos had been hired to perform a lyric at a banquet in Thessaly. He was called outside briefly, and in his absence the roof collapsed, killing all the guests and mangling the bodies beyond recognition. Simonides was able to identify the dead by recalling where each person had been sitting: the position of each body in the hall restored the name of its occupant. From this grim episode, the tradition held, Simonides drew a general principle. Orderly arrangement is the key to memory. Those who wish to train their memory should select a series of places, imprint on them vivid images of the things they wish to remember, and then, when recall is needed, mentally walk through those places in sequence, recovering each image in turn.
Cicero does not present this story as personal instruction. It enters De oratore through the character Antonius, who describes the art with evident admiration but without claiming to practise it himself. This is a detail Yates notices and takes seriously. Cicero's dialogue form gives him distance from the technique; the art is described rather than prescribed. Antonius explains that Simonides inferred two things from his experience at the banquet: that sight is what most powerfully supports memory, and that a spatial ordering can substitute for the sequential structure of a text or speech. The image does the work of the word; the locus holds the image; the mental walk recovers the sequence.
The Ad Herennium: a technical manual
The anonymous Ad Herennium, composed sometime around 85 BCE and addressed to an unknown Herennius, is the most technically explicit of the three sources. Where Cicero offers anecdote and general principle, the Ad Herennium offers something closer to a manual. Its author divides the art into two branches: memory for things (memoria rerum) and memory for words (memoria verborum). The distinction is foundational, and it immediately introduces tension into the tradition.
Memory for things requires that the orator create an image standing for the substance of what he wishes to say: a passage about an inheritance dispute might be represented by a scene of a sick man lying in bed, with other figures arranged around him to suggest the legal action. The image is not a transcription of the words but a compression of their content into a single memorable scene. Memory for words, by contrast, requires a separate image for each individual word, so that the entire verbal surface of a speech can be reconstructed exactly. The author of the Ad Herennium acknowledges that this is difficult, perhaps prohibitively so, and seems to regard memory for things as the more practical of the two. The ambition to memorize word-for-word, he implies, may be more trouble than it is worth.
What makes the Ad Herennium account particularly striking is its guidance on image-making. Not just any image will serve. The author insists that memory images must be vivid and emotionally charged: figures that are beautiful or comic or grotesque, images of action rather than stillness, scenes that provoke surprise or discomfort. Ordinary things slide past the mind; remarkable things catch and hold. The art of memory, on this account, is partly an art of controlled strangeness.
The Ad Herennium is also notably precise about places. A locus should be of moderate size, well-lit, and sufficiently distinct from its neighbours that the images placed within it do not blur together. The author recommends real architectural settings: houses, colonnades, archways. If a person does not have ready access to varied real buildings, he can construct imaginary ones, provided they are stable and internally consistent. The sequence matters as much as the places themselves, because the ordered series of loci is what produces the ordered series of memories. Walk the building forwards to deliver a speech in order; walk it backwards to reverse the order; skip to any point by counting the places.
Quintilian's reservations
Quintilian, writing in the Institutio oratoria in the late first century CE, occupies a different position in this tradition. He knows the art and describes it accurately. But he is skeptical in a way the earlier sources are not, and his reservations are more interesting than simple dismissal.
His first concern is practical. The method of loci requires, before any memorization can begin, that the practitioner have a rich stock of places already fixed in mind. Building that stock takes time and effort of its own. For a student or orator who needs to memorize something quickly, the infrastructure of the art may cost more to maintain than it saves in application. This is not a frivolous objection; it points to a genuine asymmetry between the overhead required by the system and the work it is meant to reduce.
His second concern cuts deeper. The method, Quintilian argues, works well enough for discrete items, for lists of things or separate points of argument, but it sits uneasily with continuous prose. A speech is not a list; it is a flow of language with syntactic and rhythmic structure. To break that flow into a series of discrete images, one per locus, is to work against the natural grain of how language is processed and remembered. The orator who has memorized his speech through loci must, at the moment of delivery, translate his images back into words, and Quintilian doubts that this translation can be rapid or reliable enough to sustain the momentum of actual speaking. He suspects that direct rehearsal, the old-fashioned method of simply reading and repeating, may in many cases serve better.
Yates treats Quintilian's skepticism with care. She does not dismiss it, but she notes that it may reflect a particular conception of what the art is for. Quintilian is primarily a teacher of rhetoric, concerned with the practical education of orators, and his doubts are professional doubts about whether this particular tool is right for the specific task of public speaking. He is not questioning the psychological principle that spatial ordering aids memory; he is questioning whether the full apparatus of the classical art is worth the trouble for the orator's purposes. The distinction matters for how we read the tradition as a whole.
Where the three sources agree
Despite their differences in tone and emphasis, the three sources converge on a core description. Each takes it as given that the method depends on an ordered series of places and a set of images located within those places. Each recognizes the distinction between memory for things and memory for words, even if they weight it differently. Each connects the technique to the practice of oratory, treating memory as a professional capacity to be trained rather than a natural gift to be admired. And each, to varying degrees, invokes the Simonides story as the founding moment of the art.
There is also a shared assumption about the primacy of visual imagery that Yates finds significant. The method depends not on verbal repetition but on pictorial representation. The thing to be remembered is translated into a scene, and the scene is what the trained memory stores and retrieves. This visual emphasis distinguishes the classical art from rote memorization through repeated oral recitation, and it gives the technique its peculiar character. To use the art of memory is to think spatially about information that is not inherently spatial, to impose an architectural order on the sequential structure of language.
What was lost and why it matters
Yates is careful to frame what these three texts can and cannot tell us. They are Latin sources, and they almost certainly describe a technique that was already old by the time they wrote about it. The Greek tradition that preceded them had not survived in comparable detail. What Yates is working with, and what the medieval and Renaissance traditions would work with after her, is not the art of memory in its original context but the art as refracted through three Roman accounts, each shaped by its own rhetorical and pedagogical concerns.
This matters because the technique as the three sources describe it is not self-explanatory. The Ad Herennium gives precise instructions, but it does not fully account for why the method works, what theory of mind underwrites it, or what relationship the artificial memory has to natural memory. These questions, which later commentators would press urgently, are left open by the Latin sources. The technique arrives in the historical record as a practice, not a theory: a set of instructions without a complete philosophical foundation.
There is also, Yates suggests, a gap in the surviving evidence about how widely the art was actually used. The Ad Herennium writes as though the technique is established and its value uncontested; Quintilian writes as though it is well-known but overrated. Neither tells us how common it was for practising orators to build memory palaces and stock them with grotesque images before a trial or a public address. The art may have been more admired than practised, more described than deployed. Or it may have been a genuine professional tool whose everyday use simply went unrecorded because it was too ordinary to remark upon.
The foundation of a tradition
What the three Latin sources established, regardless of what classical practice actually looked like, was a textual foundation that later centuries would treat as authoritative. The Ad Herennium in particular, circulating under Cicero's name through most of the medieval period, carried enormous prestige. When medieval scholars wanted to understand what the ancient art of memory had been, they turned to these texts. The Renaissance elaborations that Yates traces in later chapters, the magical memory theatres of Giulio Camillo and Giordano Bruno, the combinatorial architectures of Ramon Llull's followers, all have their root in the three classical sources, not because those sources fully determined what came after, but because they provided the authoritative vocabulary: loci, imagines, the ordered walk through architectural space.
Reading Yates's opening chapter is to watch a scholar perform a kind of textual archaeology, carefully distinguishing what the sources say from what later tradition claimed they said. She is interested in the art of memory not as a curiosity but as a case study in how cognitive technologies are transmitted, transformed, and sometimes lost. The three Latin texts she examines are not a transparent window onto ancient practice; they are themselves artifacts, each shaped by its context, each incomplete. But they are what we have, and from them Yates begins the long reconstruction that her book undertakes: the history of a technique for thinking that shaped Western intellectual culture far more deeply than its three thin classical sources might suggest.