Facial cues to depressive symptoms and their associated personality attributions
Introduction
Depression is a disabling mental disorder which affects 12% of the adult UK population at any one time (Singleton et al., 2001). The effect of this widespread diagnosis is far-reaching; by 2030, depression is proposed to be the second largest cause of worldwide disability, rising to the leading cause in high-income countries (Mathers and Loncar, 2006). At an individual level, those with depression are more likely to have fewer qualifications, a lower socioeconomic status (Singleton et al., 2001), and form a group that have a higher suicide incidence rate than the general population (Simon and Vonkorff, 1998). Depression also has a negative impact on the economy, with patients requiring more doctors, community care and medication than others (Singleton et al., 2001). A deeper understanding of the ability to discriminate between individuals with and without depressive symptoms, alongside how this affects perceptions of social desirability, may aid knowledge of the social implications of perceived depression.
There are two main types of cues that could be utilised to help discriminate between depressed and non-depressed individuals. The first considered here is that of live cues, where participants are able to use cues from body motions. In a study of “thin slices” of non-verbal behaviours and appearance, Waxer (1976) showed two-minute silent video segments of depressed and non-depressed patients to undergraduate and clinical graduate students. The patients' Minnesota Multiphasic Personality Inventory depression subscale scores and participants' depression ratings showed a correlation of 0.60, suggesting an ability to recognise depressive symptoms from non-verbal cues. These live cues included information from physical movements such as body motion, facial expressions, posture and proximity to others which could help to identify individuals with depression. Use of these live cues has been explored in other studies; for example, Michalak et al. (2009) found that body motion in minimal point-light displays was a reliable cue to depression, with depressed participants walking more slowly and with less amplitude in limb movements. Geerts and Brüne (2011) discuss findings of other live cues of depression, such as the ability to predict depression course from the impaired social behaviour of depression-prone patients.
However, recent research indicates an unexpected richness from a second type of cue — static cues. An example is the static, non-expressive face. Many of the Big Five personality traits, as measured by self-report, can be accurately estimated from “passport”-type photographs of people looking straight-ahead and with a neutral expression, both using individual photos (Penton-Voak et al., 2006) and composite images comparing the highest and lowest trait levels (Little and Perrett, 2007, Kramer and Ward, 2010). Composite images are computer-generated averaged faces of a number of individuals. These composite images reduce individual variation and preserve the common facial features of those within the group. Besides Big Five personality traits, the face can also reliably cue socially relevant traits such as sociosexual orientation (Boothroyd et al., 2008) and trustworthiness (Stirrat and Perrett, 2010).
Limited research has been conducted into the ability to recognise clinically related conditions from facial images. Shevlin et al. (2003) demonstrated that participants were able to accurately judge levels of psychoticism from facial photographs. However, as jewellery, clothing and hairstyles of the stimuli were not kept consistent, it cannot be determined whether the accuracy reported is a product of facial cues, or aided by additional cues such as clothing type. Research by Holtzman (2011) has helped to overcome these criticisms by using images in which individuals have removed all make-up and jewellery, and tied their hair away from their face. Even after removing these confounds, participants were able to accurately identify the ‘dark triad’ of personality disorders (machiavellians, narcissists and psychopaths) from composite facial images.
Results like those above raise the question of whether reliable cues for symptoms of depression (and other mental health conditions) may be present in the non-expressive face. The possibility of a link between facial appearance and symptoms of depression is strengthened by two findings. First, high trait Neuroticism is a risk factor for depression; and second, trait Neuroticism is one of the factors reliably signalled on the face. We discuss these findings in turn.
Personality traits are heritable constructs that serve as risk factors for mental health disorders (Krueger et al., 2000, Sen et al., 2003). Neuroticism has been identified as a personality trait with special importance for depression, as it may increase the likelihood of an individual experiencing distress (Bagby et al., 1996), which can in turn manifest itself as anger, anxiety or depression (Bagby et al., 1997). Indeed, individuals with depression are consistently found to have higher than average levels of Neuroticism (Widiger and Trull, 1992, Trull and Sher, 1994, Bagby et al., 1997). Although there is controversy surrounding the direction of the neuroticism–depression link, the general consensus suggests that Neuroticism forms a diathesis for depression (Trull and Sher, 1994, Krueger et al., 2000). The ability to use Neuroticism levels to predict later depression provides some supporting evidence for this theory (Roberts and Kendler, 1999).
Neuroticism also seems to be one of the traits which is most reliably signalled from the neutral face. Kramer and Ward (2010) found that composite images of women with high and low levels of trait Neuroticism could be identified at levels significantly greater than chance, even from internal facial features alone. In fact, Kramer and Ward (2010) found that removing the outer facial features (top of the head, jaw-line, etc.), leaving only the eyes, nose, and mouth areas within the images, did not impair identification of Neuroticism. More recently, Jones et al. (2012) created high and low Neuroticism composite images on the basis of 3D facial scans, to remove potential postural or alignment cues from subsequent comparisons. Again, identification of Neuroticism composites was robust even without these cues.
We sought to determine whether symptoms of depression might be similarly cued in the static, non-expressive face in two experiments. In Experiment 1, we tested the possibility that, within a non-clinical sample, people who score high for symptoms of depression have a distinct facial appearance from those who score low for depressive symptoms, and if so, whether such differences in appearance can be identified by untrained observers. We also explored how appearance impacted observer impressions of “social desirability”, that is, whether the person depicted in the photo was perceived to be a desirable social partner. In particular, we examined whether the facial appearance correlated with depressive symptoms might leave negative impressions on observers. These observer impressions might be important even if they are not accurate, as they could mean that people at risk of depression receive unfavourable social reactions, even in the absence of overt depressive behaviour. This unfavourable reaction could then increase social isolation and negative social consequences associated with depression. In Experiment 2, we created a new database of photographs and depression inventory scores, to replicate the identification task used in Experiment 1, with a new set of composite images.
Section snippets
Method
This experiment consisted of two phases, the first of which (“stimulus creation”) involved the collection of photographs and a measure of depressive symptoms to create our face stimuli. In the second phase, these stimuli were used to investigate the ability of observers to discriminate between faces representing high and low depression symptom occurrence (Experiment 1a), and to measure observer impressions from the high and low images of important social traits (Experiment 1b). All stages of
Identifying levels of depressive symptoms from the static face in a new sample
In Experiment 1 we demonstrated that observers were able to identify, at above-chance levels, facial images representing high and low symptoms of depression. We thought it was important to replicate and generalise this finding. We sought to replicate these results by creating a new database of face images, and generating new high and low composite and warped images. We also wanted to be certain that our results were not sensitive to the depression inventory used, and so this time levels of
General discussion
We created images which captured statistical regularities in the facial appearance of students who did and did not report high levels of depressive symptoms - less formally, depressed and non-depressed faces. We found in two separate studies that untrained observers were able to discriminate between these images more accurately than expected by chance. These results demonstrate two important conclusions about facial appearance and depression. First, there are differences in the facial
Acknowledgement
This work was supported by the Economic and Social Research Council (Grant no. ES/J500197/1).
References (35)
- et al.
Bipolar disorder, unipolar depression and the five-factor model of personality
Journal of Affective Disorders
(1996) - et al.
Affect and personality as predictors of conflict and closeness in young adults' friendships
Journal of Research in Personality
(2000) - et al.
Facial correlates of sociosexuality
Evolution and Human Behavior
(2008) Facing a psychopath: detecting the dark triad from emotionally-neutral faces, using prototypes from the personality faceaurus
Journal of Research in Personality
(2011)- et al.
Can you judge a book by its cover? Evidence of self-stranger agreement on personality at zero acquaintance
Personality and Individual Differences
(2003) - et al.
Relationship between the five-factor model of personality and unipolar, bipolar and schizophrenic patients
Psychiatry Research
(1997) - et al.
BDI-II Manual
(1996) - et al.
What is beautiful is good
Journal of Personality and Social Psychology
(1972) - et al.
Depression: the complexity of self-report measures
Journal of Applied Biobehavioural Research
(2000) - et al.
Assessment of heritability for personality, based on a short-form of the Eysenck Personality Inventory: a study of 12,898 twin pairs
Behavior Genetics
(1980)