Show someone a packaging mock-up and ask "what do you think?" and you will get an answer. Usually a polite one, often an evaluative one, and rarely the one you needed. The object on the table has done half the work of the interview already: it has told the participant what the conversation is about, roughly what kind of opinion is expected, and that somebody in the room probably made the thing. What happens next depends almost entirely on the question you attach to it.
A stimulus-based interview is one where participants respond to a specific prompt (an image, a piece of text, a video, a prototype, a scenario) so the researcher can see how they perceive it, what they feel about it, or how they would reason through a decision. Market researchers use it to test brand concepts and advertising. Education researchers hand teachers a lesson task or a classroom vignette and ask how they would respond, as a way to get at pedagogical orientation without asking teachers to describe their beliefs in the abstract. Researchers working from semiotics use cultural products such as ads, films and packaging to explore how people decode meaning, and how differently two people can read the same sign.
Across all three, the stimulus is there because people are often better at reacting than at reporting. Ask a teacher what they believe about student error and you get the professional answer. Show them a worksheet with a wrong answer on it and ask what they would say to the child, and you see something closer to practice. The stimulus makes the abstract concrete. It also brings its own risks, and each question type deals with them differently.
First-reaction questions
The simplest stimulus question asks what the participant noticed or felt in the first few seconds. "What's the first thing you see?" "What goes through your head looking at this?" These questions work because they come before the participant has built a considered position. They are good at showing salience, meaning what draws the eye and what goes unread, and at catching emotional valence before it gets rationalised.
Their weakness is that first reactions are thin, and people know they are being watched while they have them. A participant who senses the interviewer is invested will soften a negative reaction into "it's interesting". It helps to keep the first question as neutral as you can (describe, don't evaluate), and to ask for description before judgement: "Tell me what this is" comes before "Do you like it?" A participant who has to describe a concept in their own words also shows you whether they understood it, and misreadings are some of the most useful data a concept test produces.
Um, so it's... a meal kit? Or no, it's like a subscription for the spices. I think. Is it?
That hesitation tells you more about the concept's clarity than any rating would.
Interpretive questions
Interpretive questions ask what the stimulus means to the participant: "What is this trying to say?" "Who do you think this is for?" "What kind of company would make this?" This is where semiotic work lives, and it is where the emic and etic distinction stops being theory and becomes practical. The team has an intended reading. The participant has their own reading, built from their own references, and the gap between the two is the finding.
Good interpretive questions leave room for readings the team never intended. "Who is this for?" gets you much further than "Does this feel premium?", because the second one gives the participant your category to agree with. Agreeing is easy, and acquiescence is strongest when the answer is already in the question, a pattern we looked at in how interviewers end up co-authoring participants' answers. When someone gives an interpretation, follow it: "What makes you say it's for older people?" The reason they give, whether the font, the colour or a memory of their grandmother's kitchen, is the thick description you are after.
Comparative questions
Putting two or more stimuli side by side and asking which fits better, which is clearer or how they differ is one of the most common designs in concept and message research. Comparison is useful because it gives participants a reference point. Many people find it hard to judge a single option and much easier to say why one feels warmer than another.
It brings its own artefacts. Whatever is shown first becomes the anchor, and the contrast between options can inflate differences that wouldn't matter to someone who only ever saw one of them out in the world. If you are comparing, rotate the order and look at whether preferences hold across orders, as discussed in how stimulus order shapes concept test results. And ask about the losing option as carefully as the winner. What someone rejects, and why, often tells you more about their criteria than what they pick.
Projective questions
Projective questions ask participants to answer through someone else: "What would a friend say if they saw this on your counter?" "If this brand were a person at a party, what would they be like?" "Finish the sentence: people who buy this are..." The reasoning behind them, which goes back a long way in consumer psychology, is that displacing the answer onto a third party lowers social desirability pressure. People will say what "other people" think more readily than what they think themselves.
This works some of the time, and it is worth being honest about when it doesn't. Projective answers can turn into performance, with participants competing to be witty about the brand-as-person, and the analyst then has to decide how much of it is a real association and how much is improvisation. Treat projective material as a source of hypotheses and vocabulary. It is weak evidence of individual belief, and it reads best next to more direct questions in the same interview.
Scenario and vignette questions
Vignettes are short constructed situations: "A student hands in this answer. What do you do next?" "Your manager forwards you this email on a Friday afternoon. Walk me through what happens." The education research tradition of giving teachers lesson tasks belongs here. So does a lot of B2B and healthcare work, where you can't watch the real decision but can reconstruct its shape.
Vignettes are strong because they turn values into choices. They are weak because the choice is hypothetical, and stated intentions drift from behaviour in well-documented ways, which we covered in the gap between stated and revealed behaviour. The fix is to tie the vignette back to experience: once someone has answered, ask "When did something like this last happen to you?" and let the real story correct the idealised one. A vignette that feels familiar is also more useful than a clever one. If participants say "that would never happen", take that seriously, because it is data about their world.
Task-based questions
Sometimes the stimulus is something to use rather than look at: a prototype, a form, a lesson plan to adapt, a set of cards to sort. The questions here are mostly about process: "What are you looking for right now?" "What did you expect to happen there?" Task-based stimuli show decision-making in motion better than any other type, but they also turn the interview into a test, and participants respond to tests by trying to pass. Make it clear the thing is being evaluated and they aren't, and notice when a participant is performing competence rather than showing you how they would really go about it. Lucy Suchman's work on plans and situated actions is still the best reminder that what people do with an artefact depends on the situation they are in, not just on the artefact itself.
Choosing among them
These types are not interchangeable, and a common failure is mixing them without meaning to, asking interpretive questions in what was meant to be a usability study, or evaluative ones in what was meant to be exploratory. The question type should follow from what the study is meant to decide, the same logic as picking between concept, product and message testing. As a rough guide:
- First-reaction questions for salience and clarity.
- Interpretive questions for meaning and the gap between intended and received readings.
- Comparative questions when a real choice between options is coming.
- Projective questions for associations and vocabulary that direct questions don't reach.
- Vignettes for values and reasoning in situations you can't observe.
- Task-based questions for process and expectation.
A typical stimulus guide moves from the least leading of these to the most leading: describe, then interpret, then evaluate, then compare. Once you've asked someone whether they like something, you can't get back the reaction they had before they thought about liking it.
It is also worth being reflexive about the stimulus itself. Its fidelity, how polished it is, who made it and how it is introduced all shape the answers. A rough sketch invites people to build on it, while a finished render invites them to critique it. A teacher shown a vignette written in the language of a particular pedagogy may recognise the framing and answer in its vocabulary. None of this makes stimulus work invalid. It means the stimulus belongs in your analysis as one of the conditions that produced the data, not as a neutral window.
How Qualz.ai supports stimulus-based interviews
In Qualz.ai you can attach images, text, video and links to specific points in a discussion guide, so every participant sees the same stimulus at the same moment, and comparative designs can rotate order across sessions. AI-moderated interviews follow the describe-before-evaluate sequence you set and probe on the participant's own words ("what makes you say it's for older people?") rather than moving on after the first answer. In analysis, each response stays linked to the stimulus that prompted it, so you can compare readings of one concept across participants, or see whether a preference holds up when order changes.
What the platform doesn't do is choose your stimulus or tell you which question type fits your decision. Those are still the researcher's calls. Nor does AI moderation fully replace a skilled human in a sensitive vignette or a messy task session, a trade-off we set out in our comparison of AI and human moderation. If you have stimuli and a guide ready and want to see how they would run, book an information session and we can set up a pilot together with your materials.
