To test the effects of system response delay and delay variability on users of videotex it was decided to experiment in a context close to that of the service under consideration. Accordingly 165 subjects were drawn from the general public to try videotex in a series of four experiments. An emphasis on performance-related outcome measures was deemed inappropriate. Accordingly, a strategy using stepwise multiple regression followed by factor analysis selected 11 important variables from an original 38. Then these were clustered into four linear combinations or scales labelled session length, passivity, speed of response and difficulty. All four experiments failed to find any significant effects due to mean response delay. Two of the three experiments that tested ranges of randomized delay with rectangular distributions found significant disruptive effects on users. These results replicate earlier studies, and extend their generality to the context of naive users of videotex. The results also support the concept of using response-style scales rather than performance measures in non-work settings, and they contribute to the construct validity of the scales.
展开▼
