Facemasks and facial recognition

Ben Gooddy talks about his project on facemasks and facial recognition, which he presented at the British Psychological Society Cognitive Section Annual Conference.

On 02- 03 September Oxford Brookes University hosted this year’s British Psychological Society (BPS) Cognitive Section Virtual Annual Conference 2021. Cognitive psychology – for those who may be wondering what exactly this means – is the scientific study of behaviour, with emphasis on investigating mental processes such as perception, attention, and memory.  

Alongside fellow Psychology graduate Fruzsina Urban, I had the privilege of having my research project on surgical facemasks and identity recognition chosen to be presented as an academic poster with 40 minutes of verbal presentation and Q&A time.  

The format of the conference was that there would be a series of talks given by prominent researchers in the morning and afternoon – with my favourite being a talk by Simon Rjosk (Berlin Police) and Meike Ramon (Applied Face Cognition Lab) on the utilisation of super-recognisers within the Berlin police force following the 2016 terrorist attack. Super-recognisers, as the name suggests, are individuals with a particular aptitude for recognising individuals at much higher accuracy rates than the general population. In between such talks, there was a break at midday for lunch and poster viewing that was then followed by presentations and Q&A’s with the poster researchers via zoom on the event website. I presented on Thursday, and Fruzsina presented on Friday.   

My particular research project was titled ‘How do surgical facemasks impair identity recognition in familiar and unfamiliar faces?’ and was originally undertaken for the undergraduate dissertation project for the BSc Psychology course at Oxford Brookes University, supervised by Dr Michael Pilling and assisted by Mr Wakefield Morys-Carter.   

Given the widespread introduction of facemask wearing throughout the Covid-19 pandemic, the study aimed to provide a wide-ranging insight into how surgical facemasks impair identity recognition in familiar and unfamiliar faces, with manipulations of facial expressions and task-expectancy. The study set out with three main hypotheses: 

1. The presence of surgical facemasks would impair subsequent identity recognition of faces shown in a prior task. 
2. Expectancy by participants to memorise faces in the initial task would increase levels of subsequent identity recognition.
3. The presence of angry facial expressions at the initial task would result in higher facial identity recognition accuracy in the subsequent recognition task compared to neutral expressions. 

In the study, 131 participants were recruited online and took part in a two-phase experiment, built to be accessed on both PCs and phones for the maximum potential of individual participation during a socially distanced world. 48 distinct individuals with both angry and neutral portraits from an online face repository for psychological research were used as the stimuli for the experiment. An identical set of images with a surgical facemask superimposed onto each image was combined with the original set to create 192 distinct images. 

The first phase of the experiment (the ‘familiarisation phase’) consisted of a simple gender-sort task where 24 portraits would appear twice displaying varying combinations of a mask/no-mask and neutral/angry expressions. The task involved simply responding to whether the portraits displayed were male or female. Approximately half of the participants were randomly given the additional instruction to try to remember the faces at this phase as they would be tested on them in the subsequent recognition phase of the experiment. 

In the second phase of the experiment (the ‘recognition phase’), participants saw the 24 faces from the previous task with an additional 24 new and unseen faces. All portraits showed neutral expressions and mask/no-mask combinations. The recognition task consisted of responding to whether or not the portrait presented had been seen in the first phase of the experiment – this is what was meant by ‘familiar’ and ‘unfamiliar’ faces in the experiment. 

The findings of the study were the following: 

1. The presence of surgical facemasks at the first (familiarisation( phase has a detrimental effect on the subsequent recognition of facial identity. 
2. Task-expectancy had little to no impact on identity recognition, irrespective of facemask presence.      
3. The presence of angry facial expressions at familiarisation significantly increased subsequent facial recognition. However, this effect becomes complicated when interacting with the presence of a surgical facemask:

• ‘Angry expression + mask’ was more accurate than ‘Angry expression + no-mask’. 
• ‘Neutral expression + no-mask’ was more accurate than ‘Neutral expression + mask’
• Overall, participants who saw angry expressions in the first phase scored higher in recognition than those who saw neutral expressions, irrespective of facemask. 

To summarise, surgical facemasks impair identity recognition in familiar and unfamiliar faces. However, this can be mediated by changes in facial expression at the familiarisation phase. These results notably represent one of the first direct demonstrations of the interaction between emotional expressions and facemask presence on facial recognition accuracy for familiar and unfamiliar faces.

It was a fantastic opportunity to present a research poster at the BPS Cognitive conference and this wouldn’t have happened without support from Oxford Brookes University to send me along to the conference, and so I am incredibly grateful for being granted this experience.      

Want to find out more about Ben’s project? The full conference paper can be found here.

---

Benjamin Gooddy is a recent graduate of the BSc in Psychology, and part-time EEG/TMS research assistant at Oxford Brookes University, currently  studying an MSc in Neuroscience at King’s College London. Ben can be found on LinkedIn here.