Memoire Online - Degree of familiarity, inversion effect and quality of sleep through the type of images used in face recognition

Cindy Schupbach University of Geneva

3 PUBLICATIONS 0 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Face-matching: what results can we obtain between inter-clinical populations and their relatives? View project

Cassandre project : Where is the link between reality and fiction through hallucination and delightful ideas? View project

All content following this page was uploaded by Cindy Schupbach on 22 July 2017. The user has requested enhancement of the downloaded file.

LABORATOIRE EPSYLON, E.A. 4556
Equipe : « Evolution des déterminants psychologiques de la santé et du handicap selon les
âges de la vie -EVOLVE - »
Prs MC GELY-NARGEOT et S. RAFFARD : Responsables

DÉPARTEMENT D E PSYCHOLOGIE
Mention : « Psychologie Clinique, Psychopathologie et Psychologie de la Santé »
Parcours : Neuropsychologie Clinique, Psychopathologie Cognitive, Adulte, Personne Âgée »
Pr. MC GELY-NARGEOT : Directrice de la Mention et des Etudes
Pr. S. RAFFARD : Coordinateur Recherche Parcours
Dr. S. BAYARD (MCF) : Coordinatrice Recherche M1
A. GAYRAL (MC associé) : Coordinatrice Pédagogique et Responsable des Stages

DEGREE OF FAMILIARITY, INVERSION EFFECT AND QUALITY OF SLEEP THROUGH THE TYPE OF IMAGES USED IN FACE RECOGNITION

Through the scientific literature on face recognition, we can only conclude that experimental biases exist and influence the collection of data, which in turn detracts from the credibility of our results. In the face of this, more and more studies seek to make a certain ecological value and to ensure the validity of the generalities that result from it. Finally, this study makes it possible to see whether differences between the two types of images used (natural and standardized) could be observed with regard to the recognition of faces through the inversion effect and the degree of familiarity. It was therefore a face-to-face matching expected participants, recruited within a normal population, aged 18 to 30 years, as well as the Pittsburgh Sleep Quality Index. In this task, the participants had to recognize as quickly and correctly as possible the faces presented to them (their own face, that of the friend or those of the unknown) by pressing the two keys of the indicated keyboard. Finally, it is the faces of unknowns who are most affected by the type of images used, being more quickly or better recognized through standardized images. Compared to the degree of familiarity, his own face and that of the friends are mostly faster and better recognized than those of the unknown, but no significance is found between his face and that of the friends that it is in normal condition or reversed. The inversion effect is rarely present except for the faces of unknowns under certain conditions. Against all expectations, the poor quality of sleep does not significantly affect the recognition of faces except for natural images not match and reversed. Further studies may be considered.

I. INTRODUCTION

The principle purpose of this study is to better understand the recognition and treatment of faces likewise their composants, which can possibly influence them and at what point. But particularly if the treatment or recognition of his own face is different depending on we use ambient images or standardized, an idea of Bortolon Catherine basing on a precedent work (Bortolon, Lorieux, Raffard, 2017). Know and realize if we finally do it the right way in order to conduct our studies on this vast field whose main stimulus is the face. The problematic that we are discussing here is sensitive : How can we give ecological validity to what we are doing in our research practices? The objective will not be completely fulfilled, but will encourage a different vision of what future experiments on this topic could have, which have already begun to appear.

It's accepted that the recorded faces in memory involve assemblies of cells located in both hemispheres (Baird & Burton, 2008), but is there a dominant hemisphere when a face is recognized ? This is what many researchers have asked themselves. Some have concluded that the self-representation, which allows the recognition of one's own face, is not reduced to a particular hemisphere, but is available for the both cerebral hemispheres independently (Uddin, Rayman, Zaidel, 2005). The corpus callosum, affording the transfers between the hemisphere, would be necessary for the different representation of the self (Uddin, 2011). Other authors have found a dissociation between self-perception and the perception of others with the dominant left hemisphere for self-recognition and the dominant right hemisphere for the recognition of other faces (Brady, Campbell, Flaherty, 2004). Reverse thinking also exists: the recognition of the self will be in the right hemisphere (Keenan et al., 1999), as well as the treatment of self-related material (Keenan, Ganis, Freund, Pascual-Leone, 2000). But the main brain regions specifically involved in facial recognition of themselves would be: the right limbic system with the hippocampus; insula and cingulate anterior, middle right temporal lobe, left lower parietal lobe and left prefrontal regions compared to partner recognition that only activates the right part of the insula (Kircher et al., 2000). However, the measures are dependent of the means employed and may involve different neuronal signatures (Butler, Mattingley, Cunnington, Suddendorf, 2012) as evidenced by the divergences within the studies (Kaplan, Aziz-Zadeh, Uddin, Iacobini, 2008; Uddin, Kaplan, Molnar-Szakacs, Zaidel, Iacoboni, 2005).

Recognizing the old self-representation and its current facial appearance involves treatments in distinct neural circuits. The representation of the current self image is maintained and updated through plastic processes inside the cerebral areas, specialized in the treatment of faces, but not specifically of its own face (Apps, Tjadura-Jiménez, Turley, Tsakiris, 2013). It would contain « strong and configural and featural components » (Keyes, 2012). A network of common representations between the self and others seems to be possible and the passing from one to the other can be realized through self-knowledge and the agency (Dcety & Sommerville, 2003) and would influence the structural encoding step (Caharel, Fiori, Rebaï, Bernard, Lalonde, 2006). It is not unthinkable that other components are important, but not yet detected. Nevertheless, the statuts of self-representation as special hasn't yet find a consensus within the scientific literature and needs to be clarified (Gillihan & Farah, 2005).

Generally accepted in the scientific literature, we discern two main types of facial treatment: featural processing and configural processing. The first refers to the possibility for the human being to treat a stimulus by stroke, especially when one perceives an object. The second refers to any phenomenon that involves the perception of the relationships between the different components of a stimulus such as a face. This type of treatment can be divided into three types: the first-order relations ; the hollistic processes and the second-order relations. The first-order relations allows to detect that a stimulus represents a face and not an animal based on the fact that all the faces are composed in the same way, of a basic configuration with certain characteristics : two eyes above a nose, which is above a mouth. This is a crucial step: the other two types of treatment appear only if a face is perceived as a face. In the second type, the person will assemble, interconnect the different facial features in order to form a set. It will now be very difficult to consider the characteristics of the face independently. The latter type reflects the spatial distance between the different internal features of the face, such as the distance between the eyes, which in a way differentiates individuals from one another. For some authors, the inversion effect would destroy the ability to detect the second-order relationships (Sergent, 1984), but would have less effect on the isolated characteristics (Rhodes, Brake, Atkinson, 1993), as well as on first-order relations while others think it interferes with all types of configural processing (Maurer, Le Grand, Mondloch, 2002).

In contrast, the inversion effect (Yin, 1969) shows that faces presented upside down, compared to other non-facial stimuli, are less well recognized and are treated more slowly than faces presented on the right face. A fall in performance is then observed and reflects the inability of

the person to integrate the features of the stimulus into a configurational/hollistic representation in contrast to the treatment of objects, which occurs during traits by traits treatment by traits (analytic processing). The faces presented upside-down would therefore be treated on the basis of individual characteristics only and no longer on the sum of the relationships of the individual characteristics (Thompson, 1980). It is in this sense that he proposes that faces are special stimuli in comparison with other stimuli, which have no inversion effect and are probably treated differently. However, one study shows that it would be possible to develop an inversion effect for one class of stimuli if one is expert. This vulnerability to inversion makes it possible to consider the faces differently: they would not be as unique and the recognition of dogs by an expert would be comparable to the recognition of the faces. The author explain : "With expertise comes the ability to exploit" the second-order relational properties "that individualize members of stimulus classes" (Diamond & Carey, 1986).

An article (Besson et al., 2017) suggests a possible overlap between the types of configural processing discussed above and three types of levels: a level of categorization (detection of human faces), a level of familiarity (recognizing famous people in relation to less familiar ones) and detecting a person in a crowd. More precisely, he established a hierarchical experience of the treatments and established their differences: the categorization of feminine faces would be a first step whereas the individual recognition of faces and the recognition of familiar faces will follow. The categorization level is the fastest in the sense that it takes about 240 ms for a person to decide if a face is a human face and is barely disturbed by the inversion effect. We put about 260 ms to find a designated person within a crowd and is still affected by the inversion effect with a correct level of performance. Finally, it is with 380 ms that one recognizes a face as familiar. This is the level most achieved by the inversion of faces. They deduce that categorization requires little or no treatment whereas individual face processing would need partial hollistic treatments and would correspond to the first-order relations of Maurer, Le Grand, Mondloch (2002). The recognition of familiar faces would be greatly disrupted by the inversion effect and would require the highest level of hollistic treatment.

Through their three experiments, Ellis, Shepherd, Davies (1979) compared the recognition of familiar faces and unknowns, either in whole or with the inner or outer facial features. They conclude that internal and external characteristics are determinant in determining identity, but that for familiar faces, internal characteristics will lead to further recognition. The results for unknown faces show that there is no preferential use between these two categories of facial

features. From their points of view, this can be explained by the fact that famous personalities have been examined on numerous occasions (on television, in magazines, etc.) and that observers have paid more attention to their internal characteristics to understand their emotions,... Attention would thus play a major role and would lead over time to a better representation in memory for internal and external facial features. Such a treatment could be generalized when recognizing familiar persons such as family or friends ...

They have somehow initiated the notion of robust representation of Tong and Nakayama (1999): Proposed as an extreme form of familiarity, robust representations involve enormously over-learned faces that have been encountered under many different conditions and contexts. These are faces that have been seen through dynamic changes (points of view, light, expressions) or gradual changes like the effects of age. They would therefore consist of information that is invariant from the point of view. They allow faster recognition, can facilitate a variety of visual and decision-making processes. It is also acknowledged that robust representations of one's own face require much less attentional resources, but visual experience so that they develop. The results indicate that one's own face can be recognized more quickly than that of strangers under any conditions (either side or back, three-quarters, face, with or without hair, as target or as a distractor). They therefore propose that very familiar faces should be more robustly represented than unfamiliar faces and therefore less affected by the inversion effect.

A study goes against the previous assertion (Caharel, Fiori, Bernard, Lalonde, Rebaï, 2006): the difference of reaction time, clearly higher for familiar faces than for unknown faces during a configurational alteration allows us to propose That these two stimuli are treated qualitatively differently and can solicit various representations in memory: only the faces presented at the place would be anchored in memory. In another study (Keyes, Brady, Reilly, Foxe, 2010), the faces of friends are recognized faster than the faces of unknowns placed at the site, but not during inversion.

We are generally good at matching familiar faces unlike unfamiliar faces, but some authors wonder if photographs are reliable indicators regarding facial appearance. Many questions arise about the methods used in this field and a critical look could allow the improvement of our practices, bringing us a little closer to the reality with regards to the recognition of the faces. In

research, many have concentrated on the recognition of the face of unknown persons, which is not very important in everyday life where the recognition of familiar people is more frequent. The generalization of the results of studies involving different tasks can also be considered as a problem to be taken into consideration. Are people similarly recognized on simple photographs compared to in vivo faces? (Jenkins & Burton, 2011). In part, our study will serve to determine whether our choices regarding the stimuli used in the experimental task influence reaction times and the correct rate of recognition.

Secondly, our study will make it possible to distinguish certain factors influencing the recognition / treatment of the faces and to explain the individual differences in the performances that can only be observed. Only a few studies to my knowledge have been carried out for this purpose (Beattie, Walsh, McLaren, Biello, White, 2016). Using the Glasgow face-matching task (GFMT) with unknown faces, and some measures like the Pittsburgh Sleep Quality Index (PSQI) ; The Epworth Sleepiness Scale (ESS) and the PSI, they showed that participants with sleep disturbances made more errors in the standardized face-to-face test. However, deficits in identifying tasks associated with a restricted amount of sleep are not limited to tasks related to memory recognition of faces (Sheth, Nguyen, Janvelyan, 2009) or emotional recognition (Els Van Der Helm, Gujar, Walker, 2010).

The secondary objectives of this study will be to better understand the familiarity of the faces or the inversion effect, but also to see the impact that the quality of sleep of the participants will have on the treatment and the recognition of the faces within a non-clinical French population. Taking into account the study by Jenkins and Burton (2011), we hypothesize that the style of images used will influence the recognition and treatment of faces according to the degree of familiarity and the inversion effect. Second, on the assumption that a lack of sleep can lead to subtle cognitive changes even in young adults (Benitez & Gunstad, 2012) and that poor sleep quality is associated with a lack of sustained attention (Gobin, Banks, Fins, Tartar, 2015), we hypothesize that the sleep quality of the participants will influence the recognition and treatment of faces according to the degree of familiarity and type of images used following the Bruce and Young model (1986).

- Participants will recognize the photos of the unknowns better and faster through the standardized photos, whether they are placed in the right orientation or upside-down.

- Participants will recognize, whether with natural or standardized images, their face and those of their friend more quickly and with a better rate of correct answers than for unknowns in normal and reverse condition.

- Participants will respond faster and with an average rate of good answers better when they have a good quality of sleep than when they have a poor quality of sleep.

- Participants will respond less quickly and with an average rate of lower correct answers when they have poor sleep quality, than they will see natural images upside down.

- Participants with poor sleep quality will respond less quickly but will have an average rate of good responses comparable with those with good sleep quality when facing their own faces compared to the faces of friends and strangers.