1. INTRODUCTION ABSTRACT : We conducted two series of hydraulic fracturing experiments. In one we maintained unequal principal horizontal far-field stresses; in the other the horizontal stresses were isotropic. Tested boreholes were instrumented with a device capable of monitoring diametral changes during pressurization in two perpendicular directions. During borehole pressurization under anisotropic far-field stresses diametral increase at right angles to the future hydrofrac helped isolate two critical pressures, one at which fracture initiated (well before breakdown) and one at which fracture propagation stopped (beyond breakdown). Under isotropic far-field stress fracture initiation appeared to coincide with breakdown. Both results are in accord with a recent fracture mechanics analysis of hydraulic fracture mechanism. The ability to detect the fracture initiation pressure in the field is a prerequisite to obtaining more accurate estimates of the maximum in situ horizontal stress using conventional hydrofrac criteria. The conventional interpretation of the hydraulic fracturing process has been that the breakdown pressure P b , the peak pressure reached during pressurization of the borehole test interval, signifies the instance of hydrofrac initiation. The magnitude of the pressure at which fracture initiates is crucial to the accurate assessment of the maximum horizontal stress H based on the elastic (Hubbert and Willis, 1957) or the poroelastic (Haimson and Fairhurst, 1967) approaches. Both of these criteria assume linear elastic behavior of the rock until tensile failure occurs and fracture initiates. Recently, Detournay and Carbonell (1994) have challenged the assumption that hydraulic fractures necessarily initiate at Pb. They distinguish between three different critical pressures: one responsible for fracture initiation (P1), another for unstable fracture propagation (P11), and the breakdown (peak) pressure (Pb). They speculate that the condition of unstable fracture propagation, defined as the state of continued crack propagation under decreasing test interval pressure, is in effect equal to the breakdown pressure, i.e. P11 = Pb. Furthermore, using a fracture mechanics approach, they define P1 as the pressure at which a critically oriented pre-existing natural short crack around the borehole is in a state of limit equilibrium. They deduce that at slow pressurization rates and uniform far-field stress (sh = sH) fracture propagation at initiation is always unstable. Thus, initiation pressure in this case is also the breakdown pressure (P1 = Pb), as assumed in the conventional elastic and poroelastic hydrofrac criteria. On the other hand, when the two horizontal far-field stresses are non-isotropic (sh<sH), fracture propagation at initiation could be stable or unstable, implying that the fracture initiation pressure in this case is typically smaller than the breakdown pressure (Pi<Pb). This result is of utmost importance since it directly affects the computation of sH from conventional hydrofrac criteria, and questions the accuracy of previous in situ stress estimates based on hydraulic fracturing measurements. In order to verify experimentally Detournay and Carbonell's findings we recently embarked on a laboratory testing project aimed in part at isolating the critical pressure at which hydraulic fractures initiate. The experimental setup and test results are described in the following chapters.