Where stereotypes come from: a short history and psychology

Every day, our Stereotypes theme asks players to guess how Google finishes sentences like “Why are Germans so ___?” or “Why do Italians ___?”. The answers are real searches, and they are often funny. But behind the joke sits one of the oldest and most studied ideas in social psychology. Where does a stereotype actually come from? Why do some stick for centuries? And is there any truth in them?

This article is our attempt at a short, honest answer. It starts, oddly enough, in a print shop.

A word borrowed from the printing press

Before it described people, “stereotype” described metal. In the late eighteenth century, printers were looking for a way to reprint popular books without keeping thousands of loose letters locked up in frames. The solution was to take a mould of a page of set type and cast a single, solid plate from it. That plate could be stored, reused and printed from again and again, identically, for years.

The French printer and type founder Firmin Didot is credited with naming and popularising the technique. He built the word from two Greek roots: stereos, meaning solid or firm, and typos, meaning impression or mark. A stereotype was, literally, a solid impression — a copy that no longer changed. Its sibling word, “cliché”, also comes from French printers, who used it for the same kind of cast plate.

It is hard to invent a better metaphor. A stereotype plate saves effort, reproduces the same image every time, and cannot be corrected without being melted down and recast. Keep that image in mind; the psychologists who came later essentially confirmed it.

Lippmann and the “pictures in our heads”

The word made its jump from printing to people in 1922, when the American journalist Walter Lippmann published Public Opinion. Lippmann’s question was a practical one for a democracy: how can citizens form opinions about a world far too large and complicated for any of them to experience directly?

His answer was that they cannot, so they rely on simplified mental images instead — what he famously called the “pictures in our heads”. Lippmann borrowed “stereotype” for these pictures. His key observation, paraphrased, was that we tend to define first and see afterwards: we meet a situation already carrying an idea of what it contains, and that idea shapes what we notice.

Importantly, Lippmann did not treat stereotypes as a moral failing of a few bad people. He treated them as an economy of attention that everyone uses, because nobody has time to meet the world fresh every morning. The danger, in his view, was forgetting that the picture is a picture.

Katz and Braly: measuring the pictures

Lippmann described stereotypes; the next step was to measure them. In 1933, the psychologists Daniel Katz and Kenneth Braly gave Princeton undergraduates a long checklist of adjectives and asked them to pick the traits they thought were typical of a series of national and ethnic groups.

What made the study famous was not any single answer but the agreement. Students who had, in many cases, never met members of a group still chose the same handful of adjectives for it with striking consistency. The pictures were not private; they were shared, like a plate passed from printer to printer. Researchers repeated the study with later generations of Princeton students, and those follow-ups found that some specific traits faded or shifted over the decades — a first hint that stereotypes are durable but not permanent.

The method itself should sound familiar. A list of traits, ranked by how many people pick them, is not so far from what an autocomplete box shows you today:

The big difference is that Katz and Braly asked people what they believed. Autocomplete records what people wondered — which is not the same thing, as we explain below.

Allport: stereotypes as a side effect of thinking

The most influential book on the subject is probably Gordon Allport’s The Nature of Prejudice, published in 1954. Allport made a move that still shapes the field: he argued that the root of stereotyping is not malice but categorisation, one of the most basic things a human mind does.

We sort the world into categories because we must. We cannot treat every dog, chair or stranger as a brand-new puzzle. Allport described this as a kind of least-effort principle: once a category exists, we fill it with a few vivid features and apply it quickly. Most of the time that is harmless and useful. The trouble begins when the category is a group of people, the features are exaggerated or false, and the shortcut becomes resistant to evidence. Allport noted how easily people explain away an individual who does not fit, treating them as an exception rather than a reason to update the picture.

Allport also left a hopeful idea, now called the contact hypothesis: prejudice between groups tends to shrink when their members meet under the right conditions — roughly equal status, shared goals, cooperation rather than competition, and support from the surrounding institutions. Decades of later research have broadly supported the idea that contact helps, even if it is no cure-all.

The Stereotype Content Model: warmth and competence

For a long time, researchers mostly asked whether a stereotype was positive or negative. In 2002, Susan Fiske, Amy Cuddy, Peter Glick and Jun Xu proposed something more precise, the Stereotype Content Model. Their argument is that people judge groups along two main dimensions:

• Warmth — does this group mean well towards us? Is it friendly, sincere, trustworthy?
• Competence — is this group able to act on its intentions? Is it capable, skilled, efficient?

In the model, warmth judgements track whether a group is seen as competing with us, while competence judgements track its perceived status. The insight that made the model famous is that many stereotypes are mixed: high on one dimension and low on the other. A group seen as competent but cold tends to attract envy and grudging respect. A group seen as warm but incompetent tends to attract a patronising kind of pity. Neither is flattering, even when half of it sounds like a compliment.

You can see both dimensions in our own captured data. The German list above is almost pure competence: engineering, orderly, efficient, punctual. Now compare the list for Canadians:

That is almost pure warmth. Neither list is a finding about Germans or Canadians; they are a record of which dimension English-language searchers in the US reach for when they think of each country. The Stereotype Content Model gives us a vocabulary for noticing that pattern rather than just laughing at it. For a tour of how this plays out across the continent, see our piece on European stereotypes in autocomplete.

Stereotype threat: when the picture changes the person

Stereotypes do not only affect how we see others. In 1995, Claude Steele and Joshua Aronson published a study suggesting they can affect how people perform. Black and white students at Stanford took a difficult verbal test. When the test was described as a measure of intellectual ability, the Black students scored lower than their white peers with similar prior results; when the same test was framed as a non-diagnostic lab exercise, that gap largely disappeared.

Steele and Aronson called this stereotype threat: the extra pressure of knowing that a poor performance might seem to confirm a negative belief about your group. The idea spread widely and was tested with many groups and tasks, including women in mathematics.

The “kernel of truth” debate

The hardest question is also the one people ask most: are stereotypes true? The honest answer is that researchers disagree, and that the question is less simple than it sounds.

One camp, sometimes associated with the “kernel of truth” hypothesis, points out that some stereotypes line up reasonably well with measurable group averages. People are not always wildly wrong about statistical differences between groups, this argument goes, and treating every stereotype as pure fiction is its own kind of inaccuracy.

Another line of research finds the opposite for many beliefs, especially about national character. A large cross-cultural study published in Science in 2005 compared people’s beliefs about the “typical” personality of their own nation with average personality scores actually measured in those countries, and found that the two generally did not match. The pictures in people’s heads were shared and confident, but they were not a good guide to the real populations.

Three points tend to get lost in this argument:

1. An average is not a person. Even a stereotype that matches a group average tells you very little about the individual in front of you, because variation within groups is usually far larger than differences between them.
2. Kernels grow. Where there is a real difference, stereotypes tend to exaggerate it and forget the overlap.
3. Some kernels have mundane causes. Plenty of what gets called character is really climate, history, geography or policy. “Why don’t Germans have air conditioning?” is a question about building traditions and summer temperatures, not temperament.

What a search box adds to the story

Autocomplete is not a survey of beliefs. When someone types “why are Kenyans so good at running”, they may be curious about altitude and training camps, sceptical of a claim they heard, or simply watching a marathon. What autocomplete captures is the question, repeated often enough to be predicted. (We explain the mechanics in how Google autocomplete works.)

That is exactly what makes it useful for thinking about stereotypes. Many of the captured questions are about customs and trivia rather than character:

Every one of those has a real, interesting answer, and none of them is an insult. Others are pure Lippmann — a single adjective standing in for millions of people. The same pattern shows up far beyond nationality: in generations, in star signs, in birth order. Seeing them side by side is a quick lesson in how easily a picture forms.

Why a game about stereotypes can help you notice them

We built the Stereotypes theme on a simple belief: a stereotype you can see is easier to question than one you cannot. When you have to guess whether the top completion for Canadians is “nice” or “polite”, you have to consult the picture in your own head — and then you find out that almost everyone else has the same one. That little jolt of recognition is the Katz and Braly result, experienced from the inside.

We do not treat the answers as facts about anyone, and the game never claims they are true. It asks a narrower question — what do people search? — and leaves the bigger one to you. We also filter heavily. Race, skin colour, religion, party politics, sex, health, self-harm, hygiene put-downs and border disputes never make it onto a button, and every recapture of the data is reviewed by a person. Our editorial standards explain the full approach, and autocomplete and bias covers why search predictions need that care in the first place.

If you want to play with the knowledge rather than just read about it, how to play explains the rules, and our strategy guide shows how thinking like a searcher — not like a psychologist — gets you the points.

▶ Play today’s stereotypes round

GuessTheSearch is an independent game and is not affiliated with Google. The completions quoted here were captured in September 2026 in US English and will drift over time, which is rather the point: the plates in our heads can be recast.