School of Human Sciences
Face and Voice Recognition Lab
Institute of Lifecourse Development
University of Greenwich
12 April 2021
Katie Read (Student Research Assistant)
Josh Davis (Professor of Applied Psychology)
Thank you to all participants who have recently participated in the Kent Face Matching Test. We have received a large number of e-mails asking about the test and information on what participants' scores mean. Hence this blog.
Fysh, M. C. & Bindemann, M. (2018) The Kent Face Matching Test. British Journal of Psychology, 109, 219-231. https://doi.org/10.1111/bjop.12260
Below is a brief description of the test and summary of scores collected so far. We intend to publish the full data at a later date, so cannot provide too many details now (journals do not normally accept submission if data are previously published). However, the main reason we asked participants to take this test is that we wish to use this test as a measure of ability in our future research. This may replace the Glasgow Face Matching Test (Burton, White, & O’Neill, 2010) that we have previously used in a number of projects.
Kent Face Matching Test
This 40-trial test measures the ability to distinguish between two photos of white-ethnic faces. There are 20 female face pairs and 20 male face pairs. Participants must decide if the images are of the same or different face. There are 20 match trials and 20 mismatch trails. Participants are informed of the equal match-to-mismatch trial proportions in advance and reminded that it does not require memory. Pairs consist of one portrait picture with a neutral face and a smaller ID-style image. The ID image was retrieved from student ID photographs, which were taken at least 3 months prior to the portrait style picture, however the mean interval between the two is 8.8 months. Unlike the portrait image, the ID image has no constraints in terms of facial expression etc., to allow for natural variability found in photos – and this clearly makes the test harder than the Glasgow Face Matching Test.
We sent random invites to a proportion of our 45,000+ volunteer participant database.
The mean scores of this database (on 12 April 2021) on the Glasgow Face Matching Test (M = 36.83, SD = 2.65) are higher than is typically found in research suggesting that members of the database, as a whole, possess better-than-typical ability at face matching.
Therefore, those who contributed to this project may have scored higher than is typical on the Kent Face Matching Test as well. As such the scores are unlikely to be representative of the general population. We will comment on this in the journal article. We collected additional data from participants at the same time that we will report in full.
Please find in Table 1 the percentage of participants achieving each score out of 40. As can be seen, very few participants achieved maximum scores of 40 out of 40 (0.06%). These are draft results.
The second column “Cumulative Percentage” is the percentage of participants achieving that score or above. So, for instance, 65.92% of participants achieved a score of 30 or above. Note: this data set contains the raw data downloaded last week. It has not been ‘cleaned’ to check for duplicate entries or other anomalies (some people reported internet crashes) and the link will remain open for another week or so.
Table 1: Percentage and cumulative percentage of participants achieving each score out of 40
Burton, A. M., White, D., & McNeill, A. (2010). The Glasgow face matching test. Behavior Research Methods, 42(1), 286-291. DOI: 10.3758/BRM.42.1.286