In order to gain insights into ultrastructural changes taking place during the conversion of hardwoods into mechanical pulps, refiner pulp fibres were studied in detail using several microscopical techniques. Aspen (Populus tremuloides Michx.) and white birch (Betula papyrifera Marsh.) wood chips were used to produce thermomechanical (TMP), chemithermomechanical (CTMP) and chemimechanical (CMP) pulps. Following the hypothesis that there are fundamental differences in the surface and state of the fibres due to species and processing conditions, four pulps for each species and process were analyzed. Trends in fibre characteristic development were obtained within each group, based on the detailed optical analysis of 300 fibre cross-sections for each pulp. Fibre surface quality was the most important aspect of this study. Retention of middle lamella and of the layer, as well as the extent of exposure of the S₂ layer were evaluated. It was found that TMP processing of wood chips produced fibres with more exposure of the S₂ layer. Chemical pretreatment did not improve the extent of S₂ layer exposure nor the extent of fibrillation. However, the TMP fibres remained stiff, producing pulp sheets of low density and strength. Birch fibres showed a marked tendency to produce separation at or near the S₁/S₂ boundary. This resulted in high exposure of S₂ layers in TMP fibres, but produced a sheath of S₁ and ML around fibres from chemically-treated chips. This sheath was sometimes rolled back, exposing the fibre S₂ layer. Aspen TMP pulps showed high proportions of fibres with partially exposed S₂ layer. The application of chemical pretreatments to aspen chips resulted in fibres of similar levels of S₂ exposure than those achieved by TMP processing of this species, but only after reaching freeness levels of about 100 mL CSF. Fibres that showed radial failure were frequent in TMP but not in CTMP nor CMP pulps. The breakdown pattern of tension wood fibres (G-fibres) was also studied. TMP processing showed preferential breakdown of G-fibres, from which the G-layers were freed. This was not the case in the G-fibres from chemically-treated chips, in which the G-layer generally remained inside the fibres. Other categories discussed in the analysis of fibre cross sections included fibres with delamination of the S₂ layer and proportion of fibres distorted due to chemical impregnation. The breakdown of vessel elements (VE) was studied by comparing VE size frequency distributions and the proportion of whole VE that survived refining. TMP reduced VE into small fragments showing virtually no whole VE, while wood softening due to chemical pretreatment was responsible for a high proportion of whole VE in CTMP and CMP pulps. The VE from birch tend to be destroyed more easily than those from aspen, due to the intervessel pitting arrangement of the former. It is concluded that despite superior bonding potential of TMP fibres due to: - large S₂ exposure in fibres on account of separation at or near the S₁/S₂ boundary, - increased fibrillation, - longer fibrils in fines, and - release and exposure of highly cellulosic G-layers from tension wood in the case of aspen, the lack of conformability of TMP fibres, which translates into low sheet density, negates the promising benefits that otherwise would be obtained. / Forestry, Faculty of / Graduate