PhronesisThe floor
Introspection has a real jurisdiction. Ericsson and Simon mapped it: verbal reports of heeded contents — the thoughts passing through attention while a task runs — are valid data ("Verbal Reports as Data," Psychological Review, 1980; Protocol Analysis, 1984, revised 1993). A person doing mental arithmetic aloud says the intermediate steps; the steps match one predicted solution strategy and no other; the trace converges with reaction times and eye fixations. Their review found that thinking aloud slows the work somewhat and leaves its course unchanged. Reports of what appeared are documentation. Retrospective reports keep this standing too, if given immediately after short tasks.
Even the ceiling's own cartographers grant the floor. Nisbett and Wilson: "we do indeed have direct access to a great storehouse of private knowledge." Personal history. Current focus of attention. Sensations, emotions, evaluations, plans. These are contents, and about contents you are the standing authority — the one observer with access to the inside of the instrument.
The ceiling
Causes are a different jurisdiction. Nisbett and Wilson, "Telling more than we can know" (Psychological Review, 1977): "there may be little or no direct introspective access to higher order cognitive processes." The plainest demonstration: under the guise of a consumer survey, 52 passersby judged which of four identical pairs of nylon stockings was best quality. Position drove the judgment — the rightmost pair was preferred to the leftmost by nearly four to one. "No subject ever mentioned spontaneously the position of the article in the array." Asked directly, virtually all denied it — some, the paper reports, with a worried glance suggesting they had misunderstood the question or were dealing with a madman. The authors kept their own ceiling in view: why position mattered was never obvious, even to them. The finding is narrower and stranger than the pop version — an unnoticed situational cause moved the judgment, and every subject reported attribute-based reasons instead.
Where access ends, something still answers. Nisbett and Wilson's account: a priori causal theories — judgments of what would plausibly cause such a response. The theory fills the gap and arrives wearing the feel of memory. So causal self-reports are accurate when the real cause is salient and plausible, and wrong with the same fluency when it is neither. Confidence carries no signal about which case you are in. Their line for the whole terrain: "The only mystery is why people are so poor at telling the difference between private facts that can be known with near certainty and mental processes to which there may be no access at all."
The same ceiling, twice more
Johansson, Hall, Sikström and Olsson reached the ceiling at the level of a single choice (Science, 2005). 120 participants chose the more attractive of two faces; on three trials each, a sleight of hand delivered the rejected face instead. Concurrent detection: 13 percent of manipulated trials. Every criterion counted — immediate, retrospective, suspected — no more than 26 percent. Roughly three of four swaps passed with no detection at all. Then the question: why did you choose her? The answers came, and on the swapped trials they matched ordinary reports on emotionality, specificity and certainty — same detail, same confidence. Some cited features only the never-chosen face had: "I like earrings!" — for a choice whose original face wore none. About a tenth quietly described the face actually chosen, the original intention surfacing inside the wrong report. The authors named the effect choice blindness, cited Nisbett and Wilson, and called the earrings case what it is: an indisputable instance of telling more than we can know, isolated in one trial.
The third way through is anatomical. Gazzaniga and LeDoux, The Integrated Mind (1978), patient P.S., split-brain: a chicken claw shown to the speaking left hemisphere, a snow scene to the mute right. The right hand picks a chicken card; the left hand picks a shovel. Asked why, the left hemisphere — which never saw the snow — answers at once: the claw goes with the chicken, and the shovel is for cleaning the chicken shed. (That wording is the standard retelling; Gazzaniga's own tellings vary slightly. The structure — cause out of reach, reason produced anyway — is documented.) Gazzaniga named the module the interpreter. The part that produces reasons is separate from the parts that produce acts, and it reports either way.
The hinge
Polanyi, The Tacit Dimension (1966), page four: "we can know more than we can tell." Skill outruns articulation — the bicycle balanced, the face recognized, the grammar never said. Eleven years later, Nisbett and Wilson titled their paper with the words run the other way: telling more than we can know. No Polanyi citation is on record and the paper never glosses its own title; the inversion is a fact about the words, standardly read as an echo, unconfirmed as intent. It maps the terrain regardless. The two clauses are the two sides of one line.
Below the ceiling you know more than you can tell. Above it you tell more than you can know.
Ericsson and Simon close the frame from their side. Their model marks which reports fail — among them, delayed why-questions about factors never attended — and Nisbett and Wilson's demonstrations are built of those. Read together, the two literatures draw the same line from both ends: contents report; causes, when the real cause is out of view, confabulate. (One method note belongs here: Nisbett and Wilson's evidence is between-subject — no individual subject could have seen the manipulated factor — a limit pressed by Smith and Miller, White, and Ericsson and Simon themselves.)
Below the ceiling
Autopsychoanalysis works below the ceiling, or it works on invented ground. The practice splits cleanly along the line.
Its observations sit on the floor. What you noticed, what you felt, what you avoided, what you returned to, what the dream contained — contents. The waking observations carry the authority Ericsson and Simon defended; the dream report carries the weaker authority of any delayed recall — which is why: gather them close to the event, dated and concrete. They are data.
Its causal stories sit above the ceiling. Why the pattern, what the avoidance protects, what the dream means — process claims, made of the same material as the earrings and the chicken shed. Fluent, confident, indistinguishable from real reasons from the inside. The correction — and this is the proposal — keeps the practice and moves the certainty. Hold every causal story as a hypothesis. State it, derive what it predicts, test it against behavior you have yet to observe. "I avoid the call because it threatens the plan" predicts something about the next call; watch the next call. A story that predicts earns weight. A story that only explains backward stays a story — held, useful, uncertified. The observations keep their authority; the explanations earn theirs.
A harder skeptical line exists: Schwitzgebel (2008) argues we are prone to gross error even about our own current conscious experience. This note marks that line and stays with the narrower, better-documented one.
A live checker sits at /field/ceiling: paste a report and it marks the causal language — it marks language, not truth.
Kin to Autopsychoanalysis — the practice this corrects — with Dangers for the failure modes, Tacit Knowledge for the floor's far edge, and When Thought Becomes Artificial for the same Polanyi limit met from the machine side.
Rests on: Nisbett & Wilson, "Telling more than we can know: Verbal reports on mental processes," Psychological Review 84(3), 231–259 (1977) — position-effect studies at pp. 243–244; Johansson, Hall, Sikström & Olsson, "Failure to detect mismatches between intention and outcome in a simple decision task," Science 310, 116–119 (2005); Gazzaniga & LeDoux, The Integrated Mind (Plenum, 1978) — first statement of the interpreter and the chicken-claw case; Ericsson & Simon, "Verbal reports as data," Psychological Review 87(3), 215–251 (1980), and Protocol Analysis (MIT Press, 1984; revised 1993); Polanyi, The Tacit Dimension (Doubleday, 1966), p. 4; Schwitzgebel, "The Unreliability of Naive Introspection," The Philosophical Review 117(2), 245–273 (2008). The findings beneath the floor and the ceiling are established, cited as such; the floor/ceiling framing and the corollary — observations as data, causal stories as hypotheses tested against future behavior — are the proposal, offered to be argued with.