A participant opens a menu, pauses and closes it again. The recording shows a detour; it does not show whether the label was misunderstood, the expected option was missing or the participant had simply changed their mind. Asking the person to speak about what they are looking for can give the researcher a better account of that moment.
A think-aloud protocol asks participants to verbalise thoughts as they carry out an activity, usually during usability testing. The aim is to hear what they notice, expect and try to do. It produces a verbal account alongside the behaviour, rather than a complete record of everything happening in someone’s mind.
Listen for expectations while they are still relevant
Imagine a fictional photo-storage product in which a participant wants to remove an image from a shared album. “I think this deletes it for everyone” identifies a concern that the click sequence alone would leave ambiguous. The researcher can compare that expectation with the product’s actual behaviour and the wording on the screen.
The value lies in that connection. A remark about what the participant expects, tied to a visible action, can help explain a difficulty. A broad opinion such as “I like simple apps” may be worth noting, but it does not explain the decision in front of the researcher.
Nielsen Norman Group’s introduction to thinking aloud describes the method and its practical use. It is particularly helpful when the team needs to investigate understanding, although speaking during a task is itself an addition to the ordinary experience.
Decide whether to ask during or after the task
Concurrent think-aloud happens while the participant works. It places the account close to the action, but also asks the person to do two things at once. Some activities are difficult to narrate, and the narration or researcher’s prompts can affect the pace and manner of the attempt.
Retrospective think-aloud asks about the experience afterwards, sometimes while replaying a recording. This keeps the original attempt quieter, but relies on memory and later interpretation. Someone may offer a coherent explanation that they did not have at the time, or fail to recall a brief hesitation.
Choose the arrangement according to the evidence you need. If task timing is central, do not casually compare a narrated session with an ordinary silent attempt. Keep the protocol consistent across conditions, and document it when reporting measures. Neither approach gives unfiltered access to thought.
Make speaking possible without directing the task
Explain the activity in plain language and offer a short practice on an unrelated task. Participants should understand that they do not need to provide a polished commentary, justify each click or impress the researcher with design suggestions. A practice also reveals whether the setup or pace needs adjustment.
When narration stops, allow time before prompting. “What are you thinking?” leaves more room than “Did that warning confuse you?”, which introduces both an object of attention and an interpretation. Repeated questioning can turn a usability task into an interview, so use prompts sparingly and record interventions that may have influenced the attempt.
Adapt the procedure where speaking competes with the activity or a participant’s access needs. Someone using a screen reader, for example, may need time to listen before commenting. The purpose is to obtain useful evidence without making narration an unnecessary obstacle to participation.
Analyse the words with the actions
Keep observations and interpretations distinguishable. In the photo example, the participant’s concern about deleting an image is evidence of their expectation. Whether that concern caused abandonment needs to be considered alongside the sequence, any later explanation and the researcher’s own prompts.
Silence should remain silence in the notes; it is not proof that the participant was comfortable or confused. Equally, a confident explanation can be mistaken. Compare the spoken account with the interface state and outcome, then use moderated follow-up questions where an unresolved interpretation would change the design decision.