In this paper we apply the laboratory study approach of the new sociol
ogy of scientific practice to a 'thinking science': theoretical physic
s. To specify the work and accomplishments of theoretical physicists w
e choose the notion of 'deconstruction'. Deconstruction involves the e
xpansion of a concrete object, such as an equation, into a series of o
ther objects upon which the 'hardness' of a problem can be shifted and
distributed. In solving an equation, however, the determinate path of
a deconstruction method needs to be supplemented by the exploration o
f clues and guesses, trials and tricks. We trace a series of devices,
and iterations thereof, which physicists mobilize in dealing with hard
problems: formal deconstructions, detours and tricks to identify a wo
rking deconstruction, variation, 'doing examples'', modelling and, fin
ally, thought alliances between subjects.