Bernardo Alvarez Rodríguez
Marcelino Cuesta
Raquel Díaz Ramila
Juan Manuel Jiménez García
Dolores Paz Caballero
Facultad de Psicología
Universidad de Oviedo
Plaza Feijóo s/n
33003 Oviedo - España
e-mail:pazcabal@correo.uniovi.es

ABSTRACT.

The aim of this work was to compare "Thurstone" and "Likert" techniques for scale construction. To achieve this goal, two scales were ellaborated from both the same item set and sample of subjects and then scored and analysed according to either Likert' or Thurstone' procedure. Thus, thirty items were chosen for each scale, eightteen of them were shared. Reliability coefficients (a) were higher for Likert-scale, even when responses were dichotomized. Regarding validity, correlation with an internal criterion was better for Likert than for Thurstone-scale, although factorial validity was not satisfactory in neither of them. Results confirm the adventages of Likert-scale and justify the masive use of it. These adventages could be, probably, due to the inclusion in the Likert-scale of items posited in the higher and lower end of the attitudinal continuum.

Key words: scale construction, Thurstone scales, Likert scales, reliability, validity.

1.- Introducción
2.- Método
3.- Análisis de datos
4.- Resultados
5.- Conclusiones
6.- Referencias