Perfectionism is a dimension which has been studied very little as a separate entity. It is not even considered as a nosological factor. No classification of the medical sciences underlines its importance other than to speak of a personality trait, of an aspect, or of a parameter. Nevertheless, perfectionism is related to multiple disorders such as depression (18, 20, 36), suicide (8, 16, 55), nutritional problems (11, 28), anxiety (3), obsessive-compulsive personality disorder (53), social phobia (2), as well as insomnia (46). Certain authors stress the possible role of perfectionism in the development or the persistence of a substantial number of these disorders (7, 22, 38). Given these facts, it is all the easier to understand the interest shown by clinicians and researchers in the subject. Better detection and evaluation of its impact on behaviour is important in putting therapies in place (6, 53). Relationships between perfectionism and fear of failure have been approached (21, 51, 54). Correlations between perfectionism and high levels of state and trait anxiety have been demonstrated (23). The evaluation of perfectionism has been dealt with very little. Some questionnaires devote a sub-category to it, such as the Eating Disorder Inventory and the Irrational Beliefs Test. However, recently, it has been recognized that perfectionism is a multidimensional construct. Two Multidimensional Perfectionism Scales have been developed and investigated in relative isolation. Frost, Marten, Lahart and Rosenblate defined perfectionism as the setting of excessively high standards for performance associated with critical self-evaluation. Six dimensions are described: concern over making mistakes, high personal standards, parental expectations, parental criticism, doubt about quality of performance and organization. Internal consistency and validity have been established (25, 26). Hewitt and Flett (30, 31, 33, 35) have developed another approach where three dimensions of perfectionism are described: SOP (Self Oriented Perfectionism) related to high standards and self criticism, SPP (Socially Prescribed Perfectionism) related to the need of approval from others and fear of negative evaluation, OOP (Other-Oriented Perfectionism) reflecting a tendency to set high expectations for others and to evaluate them in a demanding way; this component is related, especially for males, to self-esteem, hostility and authoritarianism. Validity and internal consistency have been established too (30, 31, 35). The Frost and al's Multidimensional Perfectionism Scale and the Hewitt and Flett's scales are closely associated, except concerning the OOP. Because this component could provide new information, we have chosen the second scale, referring to the French translation and validation of Labrecque (45). EMP is the French name of MPS; it is a self-report questionnaire of 45 questions, in fact three subscales of 15 items rated on a 7-point Likert-type scale. MPS was administered to 617 first year students at the university of Liège (table II). Differences are considered according to gender and experience of failure i.e. the fact of repeating an academic year. We realized a component analysis with promax rotation. Among the different possibilities offered by the scree-test the choice of a 4 factor solution stresses the original structure: SOP (14 items), SPP (12 items), OOP (9 items) and anti OOP (10 items); the last one is additional but allows for respecting semantics and saturation of the items. The first aim of confirming validity and internal consistency is satisfactory. In other respects the multidimensional structure of the concept leads to consideration of a positive, adaptive perfectionism and a more negative perfectionism, facilitating psychopathology (59, 60, 61). So it seems interesting to compare the different components of MPS in order to find an eventual sex-failure effect. The evaluation of perfectionism is obvious, considering it as a personality trait, but it can be used also in taking into account stress and its impact, for instance that of academic performance (29, 37, 39, 58). Conferring on MPS more pertinence in gender differentiation and failure evaluation is an other goal of this research. Through the particular choice of statistical results, sex and sex-failure effects can be demonstrated: a MANOVA underlines sex effect (lambda de Wilks = 0.96, p = 0.001) and sex-failure effect (lambda de Wilks = 0.98, p = 0.05). Structure of MPS is different in four groups (FE: women with failure, FnE: women without failure, ME: men with failure, MnE: men without failure). ANOVA show differences of MPS3, MPS1 and MPS2. Far more promising is the use of LISREL method allowing for the construction of a coherent model of relationships between some dimensions of MPS and Test-Anxiety, approached here with THEE (test d'habileté aux études et à leur évaluation) French abbreviated version (49) of TASTE (Test for Ability to Study and Evaluation). In fact according to the literature of fear of failure, girls score higher on anxiety and procrastination but less on self-confidence. The structural model shows different pathways, more especially between SPP (socially prescribed perfectionism), T2 (sense of incompetence) and T1 (anxiety). SOP (self oriented perfectionism) and SPP (socially prescribed perfectionism) by girls are very much correlated; it seems that they are more subjected to society and its exigencies of studying but consequently they are more at risk of anxiety and a sense of incompetence. SOP (self oriented perfectionism) by boys functions more indiscriminately of SPP (socially prescribed perfectionism) and is negatively correlated with self-incompetence; boys are more self-confident but they usually procrastinate more probably because failure expectancies would be particularly harmful for their self-esteem; consequently, failure should be related to something else than their own capacity; this may be an explanation of the high rate of male dropouts and failure in the first year at the university of Liège; also a factor explaining the female domination at the university. In the same way the first choice of studies is moving towards shorter and less difficult orientation (46). In case of failure the model is very similar according to gender: SOP (self oriented perfectionism) and T1 (anxiety) are directly connected; SOP and SPP are in this case better correlated by boys but the path between SPP, sense of incompetence and anxiety is less significant than in girls. In conclusion, providing some modifications according to semantics, the choice of a four factor solution allows for confirmation of the original structure of MPS and for internal consistency. The different components of MPS vary according to gender: SOP and more OOP discriminate men and women; SPP allows for differentiating women with failure. A structural model enhances the role of perfectionism in the cognitive and behavioural contexts; for instance it clarifies its action on fear of failure and success rates according to gender.