How quickly can we recognize a famous face?
Even when pushed to respond as fast as possible, people needed roughly 180 extra milliseconds to tell a famous face from an unknown one compared with simply detecting that a human face was there.
Source
How Fast is Famous Face Recognition?
Study at a glance
- Design
- Human experiment — Within-subject speeded go/no-go categorization of briefly flashed photos in three tasks (human vs animal face, gender, famous vs unknown), plus a separate control group trained on the famous faces over five sessions
- N
- N=18 · 18 participants in the main experiment; 7 different participants in the control (face-learning) experiment
- Population
- Young Caucasian adult volunteers with normal or corrected vision, in Toulouse, France
- Outcome
- Minimum, median and mean go reaction times and accuracy per task; response bias (C) and sensitivity (d′)
Structured fields used in claim comparison tables when every cited study has a complete layer.
What they did
Eighteen young adults saw photographs flashed for 100 ms and lifted their fingers from a pad as fast as possible when a target appeared. In separate blocks the target was a human face among animal faces, a face of their own gender, or a famous face among unknown faces. The gender and famous-face tasks used the same pictures, so any speed difference could not be put down to the images alone. A separate group of seven people then repeatedly studied the famous faces over five sessions to see whether practice would close the speed gap.
What they found
Detecting a human face was fastest and most accurate (97.8% correct), gender judgements were close behind, and famous-face recognition was slowest and least accurate (74.9%). The fastest reliable responses to famous faces came at 467 ms, versus 285 ms for detecting a human face, a gap of around 180-190 ms that appeared in every participant. Recognizing famous faces was also about 160 ms slower than judging gender on the same photos, and participants were more cautious (more conservative bias) in the famous-face task. After learning the faces, the control group reached about 90% accuracy but was still about 120 ms slower than simple face detection.
The limits
What it doesn't show
The main sample was small (18 people, all young and Caucasian) and the control experiment had only seven participants. Famous-face recognition depends on each person's cultural knowledge, so the task was harder and less accurate, and difficulty cannot be ruled out entirely even though several checks argued against it. The study measures behaviour only, so its explanations of the extra time (more ventral-stream processing, feedback, or access to person knowledge) are hypotheses, not tested brain mechanisms. Results come from a bottom-up task with no advance cue about who might appear, so they may not apply when people are looking for a specific person.
Key terms
- Go/no-go task
- A task in which participants respond quickly to one category of stimulus (go) and withhold a response to all others (no-go).
- Minimum reaction time
- The earliest time bin at which correct responses start to reliably outnumber false alarms, taken as the shortest time needed to perform a categorization.
- Superordinate categorization
- Classifying at a broad level, such as deciding whether an image contains a human face rather than an animal face.
- Familiarity
- The feeling that a face has been seen before, which can occur without recalling who the person is.
- Response bias (C)
- A signal-detection measure of how willing someone is to say yes; a conservative bias means they need more evidence before responding.
- Diagnostic information
- The visual features that are necessary and sufficient to solve a particular categorization, which depend on both the task and the image set.
Flashcards
0 of 11 answers reviewed
Research intelligence for this paper
See its role on concept claims, tensions it is part of, placement history, and related discoveries.
Quiz yourself
Which task produced the slowest and least accurate responses?
Common questions
Why include a gender task at all?
The famous-face task used only human faces, which look more alike than human versus animal faces. Running a gender task on the same photos showed that the famous-face slowdown was not just caused by harder-to-distinguish images.
Couldn't people simply be slower because recognizing celebrities is harder?
Partly possible, but faces that were recognized most accurately were not recognized faster, and after learning the faces to about 90% accuracy the control group was still around 120 ms slower than face detection.
Why did earlier studies find identity recognition could be faster than superordinate categorization?
Those studies told participants in advance which person to look for, allowing top-down preparation. Here participants had no clue which face would appear, so recognition had to be driven bottom-up.
More on Face processing