An apparatus has been constructed that permits the measurement of time-dependent changes in volume following a sudden change in pressure near the point of vitrification. The same instrument is used for measuring steady state pressure, volume, and temperature properties that are necessary for a proper analysis of the dynamic measurements. The former experiments are referred to as pressure-jump volume relaxation (PJVR) measurements and are considered a direct probe of the structural relaxation process that occurs in all glasses.' This type of experiment relies on the nonequilibrium nature of the glassy state and the relatively slow approach toward equilibrium near the vitrification temperature and pressure.'Experimentally, the measurements are performed as shown in FIGURE 1. Following equilibration on the liquidus line (point A), the sample is subjected to a rapid change in pressure to enter the nonequilibrium glassy state (point B). The pressure is then held constant during the relaxation of the volume toward a new equilibrium state (point C). The cycle is completed by executing another pressure step of equal magnitude, but opposite in sign, resulting in another nonequilibrium state (point D), followed by recovery toward the initial volume (point A). Experiments are then repeated a t different starting pressures and temperatures.The present experiments have been performed on polystyrene ( Tg at zero pressure is 100° C) at temperatures from 110 to 150° C and at pressures of up to 2 kbar using pressure steps of -500 bars. The qualitative observations are analogous to those obtained at atmospheric pressure by rapid changes in t e m p e r a t~r e ,~ namely, 1) nonlinearity with respect to the magnitude of the departure from equilibrium, 2) asymmetry with respect to the sign of the departure from equilibrium, and 3) memory effects associated with complicated temperature or pressure histories. Each of these effects can be accounted for by a phenomenological order parameter model that has been extended to include the effect of p r e~s u r e .~ This model assumes that a distribution of relaxation times exists and that these relaxation times depend on the external variables (temperature and pressure) as well as on the instantaneous departure of the system from equilibrium. This can be represented as a product of independent shift factors 7 = roa , . a d . a p (1) where ro is a reference relaxation time, a , = exp[-B(T -T,)], a6 = exp[ -( 1x)O6/Aa], and ap = exp(bpP). Note that T, is a reference temperature, 0 is proportional to an activation energy, x is an adjustable parameter that reflects the n~nlinearity,~ 6 = ( V / V,) -I , V, is the equilibrium volume, Aa is the change in the 330