Informational Influence Research: What the Studies Show

Informational influence research includes experiments on judgment, a distinction between two kinds of social influence, and models of how behavior can spread. They do not all test the same claim. Sherif, Asch, Deutsch and Gerard, and the information-cascade models each contribute something different, and each leaves something unresolved.

The distinction matters when a familiar study is used to support a broad statement about how people decide. A demonstration in one setting is not, by itself, evidence for a universal effect or a reliable estimate of its size.

Start by separating the evidence types

The research grouped under informational influence does several different jobs. An experiment records how participants respond in a task. A conceptual distinction gives researchers language for separating explanations. A model describes what can follow from assumptions. These forms of evidence inform one another but are not substitutes.

Sherif's autokinetic studies and Asch's line experiments are demonstrations using different kinds of judgment. Deutsch and Gerard's work distinguishes informational from normative influence. The cascade models associated with Bikhchandani, Hirshleifer and Welch, and Banerjee describe how people can end up following earlier choices when the reasons behind those choices are not visible.

A claim that "people follow others" is therefore too broad to say what a piece of evidence establishes. It could refer to convergence when the answer is ambiguous, conformity when a group disagrees with an individual's perception, or a sequence in which later choices depend on earlier ones. Those are related questions, not identical findings.

Sherif and Asch answer different questions

Muzafer Sherif's autokinetic studies used a point of light in a dark room. A stationary light can appear to move when there is no reference point to judge its movement against. Participants estimated how far it moved, and their estimates converged on a shared range and remained there.

The task offers no reliable answer independent of the participants' judgments, so the demonstration concerns how judgments can converge under ambiguity.

Solomon Asch's line experiments used a different situation. A participant judged which of three lines most closely matched a standard after hearing three experimenter accomplices give a false, unanimous answer. The task was perceptually easy, so the conflict was not the same kind of uncertainty found in Sherif's setup. Most participants went along with the group at least once.

These studies are often placed together because both involve other people's judgments. The contrast between them is just as important: Sherif examined convergence where the evidence is ambiguous; Asch examined responses when a participant's own perception conflicts with the group. Neither setup establishes one fixed amount of influence that applies across situations.

The studies also do not, on their own, resolve why a participant agrees. Treating others' judgments as evidence and wanting not to stand apart are distinct explanations. A result involving a group cannot automatically be assigned to only one of those channels.

Deutsch and Gerard give the channels names

In 1955, Morton Deutsch and Harold Gerard distinguished normative social influence from informational social influence. They used "normative social influence" for the type most associated with groups and distinguished it from informational social influence, which does not require groups.

The distinction gives readers a way to ask what kind of influence a result might involve; it does not make every observed response self-explanatory.

This is why the line experiments should not be treated as a clean, one-word proof of informational influence. The participant hears a unanimous group, and the task makes disagreement noticeable. That situation leaves room for more than one explanation of a response. The terminology keeps the explanations separate, not collapsed into one label.

The distinction is developed further in normative influence, with cases in normative influence examples. For the broader concept, see informational influence. Neither label turns an individual experiment into a measurement of how strongly people will follow others in every setting.

Cascade models are not another lab demonstration

The information-cascade models of Sushil Bikhchandani, David Hirshleifer and Ivo Welch, and Abhijit Banerjee were developed independently in 1992. They address a structural possibility: when people can see what others chose but cannot see why they chose it, later decisions may reflect earlier behavior more than the decision-makers' own information.

Past a certain point, an individual's information may no longer show up in that person's action.

The motivating examples included choosing a restaurant with a queue or deciding which technology a firm adopts. The same models have been applied to financial markets, product reviews and fashions. These examples make the structure understandable, but applying a model is not the same as directly measuring how often a cascade occurs in ordinary life.

A model can show how a pattern follows from its setup. It cannot, without evidence about the particular case, establish that a real crowd is using that process, that its members share one original source, or that their conclusion is correct. Those are additional claims about the people and information involved.

An IMF staff paper on herd behavior in financial markets says that once a cascade starts, an individual's action does not reflect her private information. A real crowd would have to be shown to be in one. Reading social proof honestly starts there.

This difference matters when a source moves from "a cascade is possible under this model" to "this crowd must be following a cascade." The first describes the model's implication. The second diagnoses a real situation, which the model alone cannot settle.

Reproducibility changes the reach of a claim

A classic demonstration remains evidence about the task it used; questions about reproducibility do not erase that result. But a finding that is difficult to reproduce elsewhere gives less reason to treat it as a dependable pattern across settings.

The original research discussion notes that many memorable behavioral findings have proved harder to reproduce than expected, while the meaning of that difficulty for any particular claim remains argued over. The reproducibility of scientific results entry lays out the positions.

For informational influence, the distinction between repeating a result and extending it is especially useful. Repeating the same kind of judgment bears on whether the result recurs under that setup. Moving from a dark-room estimate or a line judgment to choices involving real stakes, different audiences, or time to reconsider asks a further question: whether the finding travels.

So a study can be important without providing a universal estimate. Sherif and Asch make distinct situations visible; they do not supply a single magnitude that can be carried unchanged into any group, market, or decision. Treat a claim about reach or strength as a separate claim from the existence of the original demonstration.

A citation trail is not an independent result

The number of places repeating a study's conclusion is not necessarily the number of independent pieces of evidence behind it. Several summaries may trace back to one original report. Repetition can make a result familiar without adding another demonstration.

The diagram captures a source-tracing problem, not a new finding about how participants behave. When multiple accounts converge on one report, their agreement should not be mistaken for several independent studies. That distinction is useful when reading claims about the classic experiments as well as claims about later applications of cascade models.

State the conclusion at the right scale

A careful summary can name the result and its boundary in the same sentence. Sherif's autokinetic studies show convergence in an ambiguous judgment task; Asch's line experiments show that participants sometimes followed a false unanimous group on an easy perceptual task.

The two demonstrations are related, but they do not test the same situation. Deutsch and Gerard's distinction helps separate informational and normative explanations, while the cascade models show how following earlier behavior can arise from a particular structure of visible choices and hidden reasons. Neither the distinction nor the models turn the classic demonstrations into a universal law.

That is the useful conclusion from this body of research: it offers named demonstrations, a way to distinguish explanations, and models of a specific pattern. Keep those contributions separate when judging what a cited study actually supports. The errors that follow when they are mixed are covered in informational influence mistakes.