An extensive experimental/research program has taken place at the University of Sherbrooke FRP Durability Facility. The program, consisting of four phases, studies the creep performance of FRP bars as well as the overall long-term behaviour of FRP reinforced concrete beams. Phase 1 deals with the creep performance of two types of GFRP bars subjected to different levels of sustained axial load; causing creep rupture at higher levels. In Phase 2, six different types of GFRP bars are tested under two levels of allowable service load, according to the currently available North American standards. The test duration, for the two phases, exceeded 10000 hours (417 days) wherein regular monitoring' of creep strain evolution took place and. the creep coefficient of GFRP bars was calculated. Residual tensile tests and microstructural analysis followed the long-term testing period. It was found that 45 % of the GFRP bars' tensile strength, fu,ave , is a safe limit for GFRP exhibiting sustained load, in standard laboratory conditions. Microstructural analysis shows that the increase in creep strain, after the 10000 hour period, is negligible for GFRP bars under allowable service load. Phase 3 consists of twenty reinforced concrete beams (ten pairs) comprising GFRP, CFRP, and steel reinforcing bars. The dimensions of which are 100 mm x 150 mm x 1800 mm, installed under third-point sustained load, for a period exceeding one year. Exhibiting a maximum applied moment of 25 % of their nominal moment capacity, Mn , all beams were regularly monitored in terms of (i) time-dependent deflection, (ii) strain increase in concrete and reinforcement and (iii) crack widths. Theoretical predictions for immediate deflection were calculated, using three methods (ACI 440.1R-06, CAN/CSA S806-02 and the ISIS Canda Design Manual (2007)), and compared to the obtained experimental results. Results showed that the calculations, regarding immediate deflection, under estimate by 67 %; underestimate by 10 %; overestimate by 11 %, for the aforementioned methods, respectively. The long-term to immediate deflection ratio, .l, was calculated for all beams and compared to ACI 440.1R-06 and CAN/CSA S806-02 predictions. Results showed that the North American standards are conservative as regards long-term deflection prediction. Immediate crack width results were compared to the prediction equations adopted by ACI 440.1R-06 and CAN/CSA S6-06, on the one hand, and by the ISIS Canada Design Manual (2007) on the other hand. Satisfactory results were found when the k b bond-coefficient factor is taken as 1.2 and 1.0, respectively. From the obtained data, the time-dependent kt multiplier, accounting for crack width increase after one year, was deduced as 1.7 and 1.5 for both models, respectively. Phase 4 deals with four full-scale GFRP reinforced concrete beams, of dimensions (215 mm x 400 mm x 4282 mm), subjected to uniform distributed load for a period of six months. Sizeable concrete blocks (of dimensions 610 mm x 762 mm x 1219 mm and weight = 13334 kN) were arranged on top of the beams to simulate sustained uniform distributed load. The main study parameters, of this phase, are (i) bottom reinforcement ratio and (ii) type of upper/compression reinforcement (GFRP and/or steel). The applied moment ranges from 15 to 21 % of the nominal moment capacity for the beams. Numerical modelling took place using a computer program (Fortran-2003) based on the age-adjusted effective modulus method, to predict the long-term deflection of the beams. The creep and shrinkage coefficients were calculated based on the ACI Committee 209 recommendations (1992) and CEB-FIP Model Code (1990). The theoretical curves were in very good agreement with the measured values. Furthermore, the empirical models available in ACI 440.1R-06 and CAN/CSA S806-02 were used for long-term deflection prediction. These predictions showed that both models can serve as upper bound and lower bound limits for the measured long-term deflection curves, respectively. As regards crack width prediction, the equations adopted by ACI 440.1R-06 (same as that of CAN/CSA S6-06) and by the ISIS Canada Design Manual (2007) yield satisfactory results when the kb bond-coefficient factor is 1.2 and 1.0 respectively (similar to phase 3). For both equations the time-dependent kt multiplier is deduced as 1.4, after six months.--Résumé abrégé par UMI.