Someone posts a half-joking complaint in the group chat. If the old hand posts it, a string of laughing replies follows. If the person who joined three months ago posts it, they get pulled aside by their manager the next day for "a quick chat." Same line of text, same chat.
The method on this page starts from that asymmetry. Do not ask first whether the sentence was a problem; ask who, from what position, treated it as one. The reading it invites most easily is "labels are just made up by other people," and what it actually says is far colder: whether a label sticks, and what happens once it does, both follow regularities — and a few of them can be measured.
In 1963 the sociologist Howard Becker published Outsiders. As a young man he played piano in Chicago bars to pay for his studies — sociology by day, dance-band musicians by night — and his fieldwork grew out of that. The line from the book that has been quoted for sixty years says: social groups make the rules first, and breaking them is what deviance means. Deviance is not a quality of what the person did; it is a consequence of other people applying a rule to them.
In more detail: for deviance to exist, two things have to be in place at once — someone did the act, and someone successfully applied the rule. So Becker tells you to count four kinds of people: those who did it and were caught, those who did it and were not, those who did not do it and were accused, and those who did neither. Study only the first group and what you are studying is not the behaviour — it is the machinery that catches people.
The rules themselves do not appear on their own. Becker coined the term moral entrepreneur: a rule usually becomes a rule because some person or agency put work into pushing it. His own example is the 1937 Marihuana Tax Act in the United States — he traced the publicity campaign of the Federal Bureau of Narcotics around the legislation and argued that this newly created agency's push was a key link. Historians later argued he overestimated that agency's role; the dispute has not closed.
For a rule to land on an actual person, someone has to be there doing it. In 1968 Aaron Cicourel published The Social Organization of Juvenile Justice, based on going to work alongside police and probation officers in two California cities: how the same kind of incident ends in two sentences of advice on the spot in one case and in a written file in another. What he recorded is that the staff carried a background picture of which children go bad — whether anyone at home is watching, how the child speaks, which part of town they live in — and that picture decided whether a record got written. So "which city has more delinquents" is a number that partly measures children and partly measures two office routines.
Strength of evidence, stated flatly: Becker's book is ethnography plus a conceptual frame, not a falsifiable hypothesis; Cicourel is field observation, and two cities are not a controlled comparison. The objection deserves its strongest version too: if deviance were purely applied, then hurting someone and littering would be indistinguishable — and nearly every known society prohibits a few categories of act, which "somebody pushed for it" cannot explain away. The defensible settlement is a division of labour: the consequences of behaviour belong to the study of behaviour, and "what gets recorded as what" belongs to this lens.
A child gets told in front of the class, once, "you just refuse to learn." Six months later he really has stopped doing homework — and does so with a clear conscience.
In 1951 Edwin Lemert, in Social Pathology, split this into two layers. Primary deviance happens episodically and leaves the person's view of themselves untouched. Secondary deviance comes after the act has been dealt with and the name applied: the person reorganises life around the new identity — defending it, making friends among others called the same thing. The point is that the causes of the second layer are not the causes of the first; the cause is the social reaction itself. Unpack the chain and there are three pathways, and they differ enormously in how measurable they are.
Pathway one carries the most famous evidence and needs the most care. Rosenthal and Jacobson's Pygmalion in the Classroom (1968) told teachers that certain pupils were about to bloom academically — the list was in fact drawn at random — and those pupils scored higher on an intelligence test by the end of the year. The study drew heavy methodological criticism as soon as it appeared. In 1984 Raudenbush pooled 18 experiments of the same kind: the average effect was small (about 0.14 standard deviations), with a clear regularity — if teachers had already spent roughly two weeks with the pupils before receiving the list, the effect essentially disappeared. What survives is not "expectations can remake a child" but expectations matter most where no first impression has formed yet.
Pathway three is the hardest to measure and the easiest to overstate. In 1973 Rosenhan published "On Being Sane in Insane Places" in Science: eight healthy people feigned hearing a voice, were all admitted, and were never detected. The current status has to be stated plainly: in 2019 the journalist Susannah Cahalan spent years in the archives and could find only a very few of the pseudopatients, and found that Rosenhan's own record did not match the paper's description — the study can no longer be used as evidence. What survives is a weaker but steadier observation: once a person is inside an institution of a certain kind, their subsequent conduct gets read through that classification.
Pathway two is the dullest and the best measured. Chicago's juvenile courts assign cases to judges at random, and judges differ in how readily they detain. In 2015 two economists used that fact to compare similar children who happened to draw different judges, and found the detained group markedly less likely to finish high school and more likely to be incarcerated as adults. That is a natural experiment, and the causal inference is far harder than in the other two. The scope has to be fenced in as well: one city, one period, one class of cases. So the next time you hear that a label can ruin a person, ask first: which pathway?
In one team, someone came last in the quarterly review and everybody knows. Someone else has a gap year that never made it onto their résumé, and nobody knows. Both have a hard year ahead, but the hard part is not the same thing.
Goffman published Stigma in 1963. One distinction of his is usually skipped over and is the most useful cell in the book: the discredited and the discreditable. The first lives a life of managing tension — everyone already knows, and the problem is getting through the encounter. The second lives a life of managing information — who knows, who gets told, when. The skills required are completely different, and outsiders usually only see the first kind.
Two moves follow: passing is keeping it unknown; covering is keeping it from taking up the room once it is known. Stigma also spreads to the people standing nearby, which he calls courtesy stigma. And stigma never lives in the attribute — the same thing counts for nothing in another setting, so what you always have to look at is the attribute-plus-setting combination.
In 2001 Bruce Link and Jo Phelan tightened the concept once, and it is worth memorising: stigma requires four things to happen together — a difference gets labelled, the label is linked to a set of stereotypes, the labelled are sorted into "them", and status loss and discrimination follow — plus one condition: the labeller must have the power to make all of it stick. That last clause moves the discussion off "who has it worse" and onto an observable question: this label — who can make it stick, and in which settings?
The status cell can be measured. In 2003 Devah Pager ran an audit study in Milwaukee: comparable young men applied for the same entry-level openings, with résumés differing only in whether they carried a single drug conviction. Among white applicants, callbacks ran at about 34% without a record and about 17% with one — a record cut the chances in half. The experiment also grouped by ethnicity, with corresponding figures of about 14% and 5% for Black applicants. Note that it measures callbacks rather than hires, and that it is one city in one year.
The criticism, stated plainly: the concept is often stretched too far, until every negative judgement counts as stigma and the term empties out; and putting all the effort into "changing attitudes" tends to route around the harder question of whether doors were actually closed.