Description:
Abstract. The consequences of speeded testing for the structure and validity of a numerical reasoning scale (NRS) were investigated. Confirmatory factor models including an additional factor for representing working speed and models without such a representation were employed for investigating reasoning data collected in speeded paper-and-pencil testing and in only slightly speeded testing. For achieving a complete account of the data, the models also accounted for the item-position effect. The results revealed the factor representing working speed as essential for achieving a good fit in data originating from speeded testing. The reasoning factors based on data due to speeded and slightly speeded testing showed a high correlation among each other. The factor representing working speed was independent of the other factors derived from reasoning data but related to an external score representing processing speed.