The effect of shape: comparing different presentations of response Botelho, Maria do Carmo (1st author) Instituto Universit´ario de Lisboa (ISCTE-IUL), Unidade de Investiga¸ c˜ao em Desenvolvimento Em- presarial (Unide-IUL) Av For¸ cas Armadas, s/n 1649-026 Lisboa, Portugal E-mail: maria.botelho@iscte.pt Calapez, Teresa (2nd author) Instituto Universit´ario de Lisboa (ISCTE-IUL), Unidade de Investiga¸ c˜ao em Desenvolvimento Em- presarial (Unide-IUL) Av For¸ cas Armadas, s/n 1649-026 Lisboa, Portugal E-mail: teresa.calapez@iscte.pt Ramos, Madalena (3rd author) Instituto Universit´ario de Lisboa (ISCTE-IUL), Centro de Investiga¸ c˜ao e Estudos de Sociologia (CIES- IUL) Av For¸ cas Armadas, s/n 1649-026 Lisboa, Portugal E-mail: madalena.ramos@iscte.pt Attitudes and motivations, intrinsically subjective attributes, need often to be studied and evaluated in order to support decisions in different areas of knowledge. But how to evaluate these subjective, and in many aspects non-measurable, entities? Almost one hundred years ago, rating scales have been proposed, which intend to gather the ”degree of affection” of an individual on a particular object or value. From then on, several studies have evaluated, compared and discussed the behaviour of different rating scales. The earlier discussions are related to the diversity of categories to be included in each variable - essentially on the number, but also on the inclusion of a neutral category (Green and Rao, 1970; Weng, 2004; Moors, 2008). Other studies focus on the adequacy of assumption of equal distances between adjacent categories implicit in the usual quantification (assigning consecutive integers to successive categories) (Green and Rao, 1970; Jamieson, 2004; Carifio, 2007). Different presentations of response have been compared with classical univariate statistics. How- ever, due to the fragility of the classical measure of location (mean) it seems appropriate a robust approach using several types of estimators less sensitive to extreme values and heavy distribution tails. The performance of robust estimators of location - applied to both quantitative and Likert-type variables - is presented in, among others Botelho (2008) and suggests that these type of estimators have a better performance in longer than in shorter items. The aim of this field based work is to study in what measure different presentations of items induce different behaviours in scales distributions using both classical and robust approaches. Participants and Procedures The participants of this study were first year, first-time students, from several degree courses (Social Sciences, Management and Technological Sciences Courses) at a public university institution, whose course plan includes subjects from the scientific area of quantitative methods, defined as the target population. The sample consist in 727 participants, with age range between 16 and 56 years, the average age is 20.9 years (SD=6.7) and the most frequent age is 18 years old. The majority of