Official URL: https://escholarship.org/uc/item/9kb6k4bh

This paper describes two novel experiments that investigate the cognitive basis of surprise in stories depicting simple, everyday events. Participants were asked to read a number of short scenarios, each of which concluded with some surprise event for the central protagonist. The story versions differed in how strongly they predicted this ‘surprise’ conclusion, by varying the degree of attention drawn to key enabling conditions for that conclusion. The effect of this manipulation on participants’ surprise ratings, and the speed with which they read the final sentence, were used as dependent measures. The results of our experiments show that the specific representations built by people in understanding the earlier part of a story have definite effects on their level of surprise at later events. Furthermore, the pattern of reading times for the target sentences supports the explanation given for these differences. We discuss the consistency of these findings with theories of discourse comprehension and describe how such effects might be modelled computationally.