Skip to content
Back to search results

Vignette Experiment

Introduced by
Survey and social-psychological research traditions
Type
Randomized experiment using short described scenarios
Family
Experimental survey designs
Font size

What does it do?

A vignette experiment presents respondents with a short, carefully constructed description of a person, situation, or scenario — a vignette — in which one or more features are experimentally manipulated, and then asks for a judgment, attitude, or intended action.

By randomizing which version of the scenario each respondent reads, the researcher isolates the causal effect of each manipulated feature on the elicited judgment, combining the realism of a concrete scenario with the causal leverage of an experiment.

The intuition behind the method

Asking people abstractly what matters in, say, a deserving welfare recipient invites rationalized or socially desirable answers. Instead, show them a concrete vignette — a specific applicant with a named age, work history, and family situation — and ask whether the applicant deserves aid. Now vary just one feature, like whether the applicant lost their job through layoff or quit voluntarily, across randomly assigned respondents. The difference in approval ratings reveals how much that feature actually drives judgment, without respondents ever having to articulate, or even consciously realize, that it mattered.

History

Vignettes have long been used in sociology, social psychology, and medical ethics to study judgment and decision-making. As survey experimentation matured, the experimental vignette became a standard randomized design; Atzmüller and Steiner's 2010 methodological treatment formalized single- and multi-dimensional vignette studies, and Aguinis and Bradley's 2014 best-practice paper codified design recommendations. In political science the approach underpins research on welfare deservingness, immigration attitudes, and candidate evaluation, and it is the conceptual parent of the factorial survey and conjoint experiment, which scale vignette manipulation to many attributes at once.

When to use it

Use a vignette experiment when you want to study how specific, manipulable features of a described situation or person causally shape judgments, attitudes, or intended behavior — deservingness, blame attribution, candidate evaluation, perceived fairness, or moral judgment. They excel when direct questioning would invite social-desirability bias or when the construct is best conveyed concretely. They are weaker when the scenario cannot capture the complexity of real decisions, when behavioral rather than judgmental outcomes are needed, or when many attributes must vary simultaneously, where a factorial-survey or conjoint design is more efficient.

Limitations and considerations

  • A described scenario is still hypothetical; stated judgments may diverge from real decisions under stakes and consequences.
  • Poorly written or implausible vignettes introduce confounds or noise that contaminate the manipulated effect.
  • With many dimensions, the number of distinct vignettes grows quickly, straining sample size unless a fractional design is used.
  • Respondents may attend to unintended surface features of the text rather than the construct the researcher intends to manipulate.

Source conditions do not replace independent scientific review or validation by an analysis engine.

Strengths

  • Concrete scenarios elicit more realistic, less abstract judgments and can reduce social-desirability bias on sensitive topics.
  • Randomized manipulation cleanly identifies the causal effect of each scenario feature on the outcome judgment.
  • Flexible: a single vignette can convey rich context that a battery of abstract questions cannot.
  • Easily embedded in surveys on representative samples, supporting both internal and reasonable external validity.

Applications

  • Studying welfare and immigration deservingness by varying recipients' characteristics in a described case.
  • Measuring blame attribution and punitiveness by manipulating offender and offense features in a scenario.
  • Evaluating perceived fairness of policies or procedures by varying their described attributes.
  • Assessing candidate evaluations by manipulating biographical or position features in a described profile.
  • Examining moral and ethical judgments by varying the circumstances of a described dilemma.

How it works

How it works

1

The researcher writes a realistic, self-contained scenario that fixes a context and embeds the features of theoretical interest. Good vignettes are concrete, plausible, internally consistent, and hold constant everything except the manipulated dimensions so that respondents react to the intended construct rather than incidental cues.

2

Each manipulated dimension of the vignette takes one of a small set of levels — for example, the protagonist's gender, an action's severity, or a policy's cost. The combination of levels defines the version a respondent sees; varying dimensions independently lets their separate effects be identified.

Formula
V = (d_1, d_2, \dots, d_p),\quad d_j \in \text{levels}_j
3

Respondents are randomly assigned to vignette versions with known probabilities. With a single manipulated dimension this is a simple between-subjects comparison; with several dimensions it becomes a factorial allocation in which each respondent receives one combination of levels.

Formula
\Pr(\text{respondent sees vignette } v) = \text{known by design}
4

After reading their assigned vignette, respondents provide the outcome judgment — approval, deservingness, blame, punishment severity, vote intention, or similar — on a rating or choice scale. Each respondent's response is their potential outcome under the version they received.

Formula
Y_i = Y_i(V_i)
5

Judgments are regressed on indicators for the manipulated dimension levels, yielding the average causal effect of each feature. With multiple respondents per condition and randomization, the coefficients are unbiased estimates of how each scenario feature shifts the judgment, and interactions reveal combined effects.

Formula
Y_i = \beta_0 + \sum_{j} \beta_j\, d_{ij} + \varepsilon_i

Common pitfalls

  • Manipulating a dimension in a way that inadvertently changes the scenario's plausibility or tone, confounding the intended feature.
  • Writing vignettes so long or detailed that respondents tire or focus on irrelevant details, adding noise.
  • Failing to pre-test for comprehension and realism, so the manipulation does not land as intended.
  • Treating one manipulated feature's effect as generalizable when it was only tested within a single fixed scenario context.

Frequently asked questions

How does a vignette experiment differ from a factorial survey experiment?

A classic vignette experiment often manipulates one or a few dimensions of a single scenario in a between-subjects design. A factorial survey experiment systematically crosses many dimensions, each with several levels, generating a large universe of vignettes from which respondents rate a sample. The factorial design is essentially a scaled-up, fully crossed vignette study built to estimate the independent effect of many attributes simultaneously and efficiently.

Are stated judgments in vignettes valid measures of real behavior?

Vignettes measure judgments about hypothetical situations, which correlate with but do not equal real behavior under stakes. Validity is strengthened by realistic, well-pretested scenarios and by outcomes framed as concrete decisions rather than abstract opinions. Researchers should treat vignette results as evidence about the determinants of judgment and be cautious extrapolating directly to consequential behavior without corroborating evidence.

How many dimensions can a vignette manipulate at once?

A single between-subjects vignette experiment typically manipulates one to three dimensions cleanly. Beyond that, the number of distinct versions multiplies and within-cell samples shrink. To manipulate many attributes, researchers move to a factorial survey design, presenting each respondent several vignettes drawn from the full factorial universe, or to a conjoint experiment, which handles many attributes through forced choices.

Related methods

Sources

  1. Atzmüller, C., & Steiner, P. M. (2010). Experimental Vignette Studies in Survey Research. Methodology, 6(3), 128–138.

    Open DOI reference
    Copied.
  2. Mutz, D. C. (2011). Population-Based Survey Experiments. Princeton, NJ: Princeton University Press.

    ISBN: 9780691144528
    Copied.
  3. Aguinis, H., & Bradley, K. J. (2014). Best Practice Recommendations for Designing and Implementing Experimental Vignette Methodology Studies. Organizational Research Methods, 17(4), 351–371.

    Open DOI reference
    Copied.