Economic theories of choice under uncertainty assume that agents act a
s if they have preferences which govern their choices between risky op
tions. Theories differ as to the exact specification of the preference
structure, but it is common to assume that preferences are complete a
nd satisfy certain consistency requirements such as transitivity and m
onotonicity. In this paper, it is argued that there may be reason to d
oubt whether individuals act as if they have complete and consistent p
references over risky actions. Instead it is suggested that individual
s should be thought of as actively constructing preferences through a
process which I call ''rationalisation''. It is then argued that ratio
nalisation provides a basis for understanding certain experimentally o
bserved ''anomalies'' which appear quite at odds with conventional the
ory.