This study aimed to identify the law of cognitive test scores with the number of repetitions, and to establish a prediction model that quantitatively describes the changes in cognitive test scores. The experiments were conducted in a chamber that simulated an office environment. Participants performed neurobehavioral, Tsai-Partington, typing and d2 tests at 26 degrees C. The speed and accuracy were used to reflect the cognitive test scores. The results demonstrated that a significant learning effect occurred in these cognitive test scores. The speed and accuracy of completing cognitive tests increased as the number of repetitions increased. The majority of cognitive tests could achieve stability after five repetitions and, at most, eight repetitions. The learning effect had different effects on the speed and accuracy. The percentage improvement of the speed was higher compared with accuracy, but it took more attempts to reach stability. The regression model based on the logistic function could accurately quantify the tendency for the variation in cognitive test scores with the number of repetitions. Compared with the power, linear and logarithmic functions, the logistic function had a higher R-2(> 0.99). A method for correcting the learning effect of these cognitive tests was proposed based on the prediction model. (C) 2019 Elsevier B.V. All rights reserved.