There are many experimental situations in which polymer chains are constrai
ned or localised into a small region of space (e.g. melt chains confined to
a "tube", network chains pinned by crosslinks). We show that detailed cons
ideration of the quenched variables is vital in these experiments. This pap
er provides a crucial link between microscopic models with localising const
raints and scattering: patterns by a generalisation of the Random Phase app
roximation (RPA) which allows for quenched translational variables. A metho
d is developed which deals with correlations between the quenched variables
brought about by incompressiblity (for example, in a polymer melt there ar
e correlations between tubes because of the interaction between chains). As
an example. the generalised RPA is applied to models based on the Warner-E
dwards picture of the tube. Theoretical results for a melt of I-I-shaped co
polymers are compared with experimental scattering. Early results suggest t
hat to fit the scattering: we may be forced to relax one of the central ass
umptions of the tube model, that the tube deforms affinely, that all chains
retract by the same amount or that the tube diameter does not couple to th
e strain.