The fundamental relationships of traffic flow describe traffic behaviour in
terms of three variables: flow, density and speed. Among other things, the
y are useful for predicting the effect of changing demand conditions on the
level of service. Their basic form was postulated In the 1930s, and many e
xperiments have been carried out since then to determine their precise shap
e. But they have not yielded consistent results, partly because of confusio
n about the causal links among the variables, which operate differently dep
ending on whether the speeds and flows are sampled over short periods of ti
me, or over long periods of time. The mechanism of cause and effect is diff
erent and the roles of the explanatory and response variables (flow and spe
ed respectively) are reversed. In this paper, the authors use simple models
to show that the shape of the relationship between speed and flow varies a
ccording to the unit of observation, and that in the steady-state case, the
form differs from the one that is conventionally pictured. The paper concl
udes by suggesting a modified representation, and briefly discusses the pra
ctical consequences.