Korzybski, hand-built, 1925
The structural differential, and where E‑Prime cuts in.
In the 1920s Alfred Korzybski, the engineer-philosopher who founded general semantics, built a physical teaching device and patented it in 1925: the structural differential. A wall-hung model of a parabola, a disc and a chain of hanging tags, joined by strings, it trains what he called consciousness of abstraction: a felt awareness of everything your nervous system and your language leave out between reality and the words you speak about it. Korzybski traced much human muddle, from marital rows to ideologies, to losing track of that route.
One worked example runs through this page. Your colleague Anna doesn't appear at the 10:00 team meeting. By lunchtime, someone in the office has said: “Anna is selfish.”
The differential reveals every rung that verdict silently climbed.
Reading the device, top to bottom
The parabola: the event level
Reality itself, as ongoing process: Anna's actual morning in full: the sick child, the traffic, the missed alarm, the email that never arrived, right down to the atoms. Korzybski gave the parabola a broken, jagged edge on purpose: the event level extends indefinitely; no frame ever contains it. The dots (holes in the physical device) stand for its characteristics, effectively infinite.
The strings: abstracting
Only some holes carry a string down to the next level. Each string marks a characteristic that survives; every hole without a string, a detail lost. Korzybski called this abstracting, and the device makes you see its price: abstracting means leaving out. The same thing happens again at every level below, with fewer strings each time. Moving down the ladder you never gain information; you only lose it, while sounding ever more certain.
If you know one sentence from this whole world, it probably reads: “the map is not the territory.” The differential simply builds that sentence in wood and string: every level maps the level above it, and every dropped string marks a place where the map thins out. (Connoisseurs will spot the joke: the most famous sentence in general semantics flunks E-Prime. Bourland, in the essay this page leans on, gives it in a form that survives the ban: “the map does not fit the territory.”)
The disc: the object level
Not yet words. Here sits what your senses actually made of the event: an empty chair at 10:00, a silence where Anna's voice would normally sound. Notice how little of the parabola made it down. You experienced a chair, not her morning. For Korzybski the object level remains non-verbal: something your nervous system built, private to you, and already radically incomplete.
The tags: the verbal levels
The moment you speak, you hang the first tag: description: “Anna didn't come to the meeting.” Still close to the ground; anyone could check it. The next tag hangs off that one, not off reality: inference: “it can't have mattered to her.” A guess presented as a fact. The tag below hardens the guess into a generalization, “Anna is selfish”, and here the little word does something drastic: it welds a level 4 guess onto a level 1 person. The verdict feels like a fact precisely because the ladder stays invisible. The chain runs on from there: etcetera: labels about labels (“people like her ruin teams”), each further from the chair, none nearer the morning.
The loop back: the natural order of evaluation
The finest detail of the original device: Korzybski pinned the last tag back onto the parabola. Healthy thinking, he insisted, runs in a cycle, not a one-way descent. You climb down into words and theory, then return to the world and check (did Anna actually sit in A&E all morning?). He prescribed regular returns to what he called “silence on the objective levels” (going quiet and looking again) before bursting back into speech. People get into trouble when they never close the loop: the etcetera level then feeds on itself.
Where E-Prime cuts in
Korzybski singled out two uses of the verb “to be” as the troublemakers. The “is” of identity (“Anna is a slacker”) flatly equates a person with a label: a level 5 object welded to level 1 territory. The “is” of predication (“Anna is selfish”) does the same more quietly, projecting a quality onto her as though it lived in her rather than in your inference about her. He called both identification: collapsing different orders of abstraction into one. The sentence bypasses the ladder it actually climbed. And the listener can't see the bypass either: “Anna is selfish” arrives with no fingerprints on it, no owner, no date, nothing to check.
“The little word is has its tragedies: it names and identifies different things with the greatest innocence; and yet no two are ever identical, and if therein lies the charm of wedding them and calling them one, therein too lies the danger. Whenever I use the word is, except in sheer tautology, I deeply misuse it; and when I discover my error, the world seems to fall asunder, and the members of my family no longer know one another.”
Korzybski's own remedy took years: training with the wall-chart until you felt the levels. His student D. David Bourland, Jr. found a shortcut of his own: make the collapse grammatically unsayable. E-Prime, English minus every form of “to be”, doesn't argue with the verdict; it removes the verb the verdict needs.
Back up the device
The ban does more than stop a sentence. Ed Kellogg, who spoke E-Prime every day for years, found it also pushes you somewhere: it “tends to bring the user back to the level of first person experience”. Run the Anna verdict back up the device, starting at the bottom, and each version gains something the one before lacked.
- Generalization“Anna is selfish”No owner, no date. It needs “is”, so E-Prime takes it away.
- Generalization“Anna acts selfishly”Gains an action: something she did, which you can point to.
- Generalization“Anna seems selfish to me”Gains an owner.
- Inference“I wonder if it mattered to her”Gains a flag: a guess about one event, not about her.
- Description“Anna didn't come to the meeting”Gains a check, though she may have dialled in.
- Description, as close as words get“At 10:00 I looked at Anna's chair and saw it empty. I felt let down.”Gains an observer, a time, what the senses caught and the feeling.
- Object levelThe chair itselfWords stop here: Kellogg's “silent level”.
- Event levelGo and look: ask AnnaHer morning, which you never saw.
E-Prime gets you climbing and keeps you honest on the way up. When you reach for an “is” and can't use it, you have to look again at what you actually saw, which matches Kellogg's advice for a stuck speaker: go back to the experience and describe what happened. Every sentence above the bottom step comes out in E-Prime, and words like seems and acts show you where you have got to, and how much further you could go. The same climb works anywhere: “he is lazy” becomes “he missed two deadlines this month”, and “it is hot” becomes “I feel hot”.
Nobody reaches the top in words. Kellogg calls his goal “not presently attainable, like the North Star”, and warns against describing experience too closely.
The top step matters as much as the grammar. E-Prime restructures the sentence; the differential asks for more: go back to the parabola. “Ask Anna” completes what Korzybski called the natural order of evaluation: world → experience → words → back to the world. Even a perfectly E-Primed verdict, never checked, remains a guess, however well phrased.
The ban on “to be” also delivers something Bourland documented with relish: the role players return. The passive voice (“the data were not available”, “mistakes were made”) becomes nearly impossible without the verb, so the people re-enter the sentence. Bourland records two engineering reports (a failed satellite sensor; an antenna a contractor never switched on) whose E-Prime rewrites forced the authors to dig out who did what, and the digging, he notes, “brought useful information to light”. For anyone who works in organisations, that alone might justify the practice.
And if the full ban sounds impossibly hard, two reassurances from Bourland himself: E-Prime removes only 20 or so words from an English lexicon that scholars put at one to two million; the language's wealth of linking verbs (become, seem, appear, the verbs of the senses) takes over most of the work. And Russian, Hungarian and Mandarin lack an equivalent of “to be” altogether: roughly a fifth of humanity already speaks more or less “is”-free, every day.
One honest caveat, and the real reason for the hard rule. E-Prime works as a blunt instrument: it bans every form of “to be”, including uses that collapse nothing: “she is running”, “there is a chair”. Why not simply avoid the two harmful uses and keep the rest? Because that approach has failed even the masters, what Bourland calls the “Is Trap”. Korzybski himself, who warned against these constructions more than any thinker alive, still used identity and predication in some 37% of the sentences of Science and Sanity; J. Samuel Bois, for years the Institute's chief lecturer, reached about 42% in his own textbook. Piecemeal avoidance, Bourland concluded, “simply [has] not worked. E-Prime provides a simple discipline that does.” Hence our group practises the strict ban, with grace for slips, and Bourland's own history supplies the grace: when Bourland wrote his first paper without “to be”, it gained what he called “a peculiar clarity”… and one “was” crept in anyway.
Korzybski, typically blunt, dismissed the content of all this as “baby talk”. The discipline of knowing which rung you stand on, and looping back to look, matters far more. E-Prime turns that discipline into grammar.