Perspectiva multifactorial da memóriaconstrução e validação de uma bateria de provas para a população universitária portuguesa

Resumen

The aim of this thesis is to construct and validate a multi-factorial battery to assess memory in university students. The construction of the different tests was based on various theories on memory, namely the Multiple-Component Model of Working Memory (Baddeley & Hitch, 1974). Also, some experimental tasks and standardized memory batteries were considered. From this analysis, six computerized tests were created to assess short-term memory (Digits, Spatial and Colored Digits) and working memory (Letters and Numbers, Visuo-spatial and Arrows and Numbers). Each of the constructs presents three differentiated tests according to different content, namely verbal, non-verbal and mixed. A first study included 15 university students, and was based on commentaries about the format of the tests and the structure of the battery. After making alterations in regards to the suggestions retrieved, the battery was applied to a sample of 89 students attending higher education institutions and results were analyzed according to the Classical Tests Theory and Item Response Theory. The results suggest internal consistency values of Cronbach’s alpha varying between .70 and .87 for the six tasks being studied, as well as a structure based on two factors which explain 62% of the total variance. Through the Item Response Theory, the Partial Credits Model was used for the polytomous data while the Rasch Model was used for the dichotomous items. The data presented indicates items which are adjusted to the Infit and Outfit measures. New modifications were introduced considering some of the items were not adjusted. Also, these alterations occurred because of the intervals of the level of difficulty of the reduced items, as well as because of some Rasch measures for subjects and items. A third study was done with a sample of 547 students from various Portuguese higher education institutions of both the private and public sectors. The Cronbach’s alpha values oscillate between .72 and .89, suggesting a confirmatory factorial analysis of a bifactorial structure including latent variables with verbal and non-verbal content covarying within each other (CFI=.984; TLI=.969; RMSEA=.060). Precise analyses of the derived measures of the Rasch models (Person Separation Reliability - PSR) confirmed similar values to Cronbach’s alpha (PSR between .72 e .88). Also, the convergence validity was secured by the significant correlations between five of the tests of the battery and four sub-tests of the Wechsler Memory Scale III. Differential studies do not mention significant differences according to the areas of study of the participants, but do point out differences in terms of gender, favoring students of the masculine gender. Nonetheless, there is some bias associated to gender for most of the items. From the results obtained, some commentaries and suggestions are made bearing in mind future studies which justify a consolidation of the positive results obtained up to this date.