The debate about the role of attitude in the technology acceptance model (TAM) seems to have re-emerged in two prestigious journals in the field of educational technology. Among the publications on this debate, there are authors in favour of excluding the attitude of TAM, whereas others are in favour of including it. These opinions are derived from the results that both groups have obtained in their researches. We have wondered whether the methodology used in these studies could have influenced the results obtained and whether different statistical tools may lead to different conclusions. Two alternatives models were compared: TAM with attitude (TAM-O) and TAM without attitude (TAM-R). To detect the influence of different methodological tools, two structural equations modelling (SEM) approaches were used for analysis: covariance-based SEM (CB-SEM) and partial least squares SEM (PLS-SEM). We found that if PLS-SEM method is used, TAM-O is a better model than TAM-R, whereas if CB-SEM is used, TAM-R is better than TAM-O. We face a paradox that is explained by the influence of the methodology.