Stress analysis by thermoelastic techniques has become a growing field since the application of sensitive infra-red measuring devices in the late 1960s. It is commonly asserted that by observing the variation in temperature on the surface of a body as the body undergoes a change in stress, the surface stress changes may be determined via a simple linear relation between temperature and stress. In making this assertion, two fundamental approximations are made: i) the stress changes in such a manner that adiabatic conditions are attained; and, ii) the material properties which relate the change in temperature to the change in stress remain constant throughout the loading process and are not significantly affected by either the stress or temperature change. The aim of this review article is to show the potential applications that can arise when the above two assumptions are not made. It will be shown that in many situations, these effects can significantly bias the experimental data. At first glance, such a bias can pose difficulties in the quantitative assessment of the data; if, however, the nature of these effects are understood sufficiently to be modeled mathematically, then important information can be gained which leads to a more powerful tool for stress analysis. If the assumption of adiabaticity is not applied, then the thermoelastic heat generation and conduction process which occurs when a composite laminate is stressed can be modeled. It will be shown how by observing the manner in which the surface generated temperature is biased by heat conducted from subsurface plies, that the strain components may be determined, even though such stress analysis techniques are typically assumed to measure only bulk stresses. Also, if the material properties are not assumed to remain constant with stress, it can be shown how an understanding of the variation in coefficient of thermal expansion with stress can lead to the potential for measuring residual stresses and plastic zones by thermoelastic techniques. This article contains 53 references.