This study presents a detailed investigation into size, density and community structure of temperate marine reef fish in the medium-sized Goukamma Marine Protected Area (Goukamma MPA) and adjacent fishing grounds on the south coast of South Africa. The oceanographic conditions, the spatial distribution of the benthic community and the prevailing fishing effort are also described. Life history traits and per-recruit (PR) models for the principle target species, roman (Crysoblephus laticeps) are compared between the protected and exploited area. From the study results, various strategies are proposed for the use of MPAs in the conservation and management of linefish species along South Africa’s south coast. The distribution and topography of reefs in the protected and exploited sections of the study area were found to be comparable. Atmospheric pressure ranged from 992 to 1,032 mb, being significantly lower in summer. Wind speeds ranged from 0.7 to 71.3 km/h. Water temperatures ranged between 9.0 and 22.2 ºC and turbidity between 0.3 and 45.8 NTU. Water temperature and clarity were uniformly low in winter. In summer the water was generally warm, clear and stratified, with a thermocline at around 20 m, although intermittent upwelling events caused water temperature to decrease and clarity to deteriorate. Current speeds ranged between 0.11 and 2.59 km/h and were significantly higher in spring and autumn. Easterly currents prevailed in spring, summer and autumn and westerly and southerly currents in winter. Hake (Merluccius capensis), various resident reef fish and kob (Argyrosomus japonicus) were most frequently targeted by the local linefishery. A significant amount of illegal fishing was found to occur in the protected area. Fishing effort was found to be highest around the border of the MPA (2.7 boats/km²) and lowest in the core of the MPA (0.2 boats/km²). If law enforcement remains poor it may be necessary to adapt the management strategy to extend the reserve, thereby mitigating against illegal fishing and ensuring a core area of no exploitation. Various other alternatives were investigated and it was demonstrated that the amount of fish caught of legal size could be increased by about 23% and post-release mortality of undersized fish reduced by 50% through the introduction of a suit of restrictive measures. Randomly stratified underwater visual census (UVC) and controlled fishing were used to investigate the ichthyofauna and benthic community at protected and exploited sites in the study area. Resulting density and size data from 273 fishing sites and 177 point counts were analyzed using generalized linear models (GLMs). Fish communities were found to vary significantly, depending on the level of exploitation. Roman, the principle reef fish species targeted by the fishery had significantly higher densities within the protected parts of the study area (CPUE: 4.3 fish/anglerhour; UVC: 2.2 fish/point-count) as compared to the exploited part (CPUE: 3.4 fish/anglerhour; UVC: 1.8 fish/point-count), correlating strongly with the observed fishing effort. Also mean sizes were significantly higher in the protected area (299 mm from fishing survey and 233 mm from diving estimates) as compared to the exploited section (283 mm from fishing survey and 198 mm from diving estimates). Although other fish species also had significantly higher mean sizes at protected sites in most cases their densities were significantly lower. This suggests a top-down control of the fish community by the dominant predator (roman). The results of the UVC showed the diversity of the ichthyofauna to be significantly higher inside the protected area. Interestingly this did not apply to the results of the controlled fishing experiment where the diversity of fish in the catch was lower in the protected area - a result that may be explained by the selectivity of fishing for the most aggressive species – and a reminder of the limitations of controlled fishing experiments. Possibly the most important finding of the study revolved around the benthic community. These were significantly different at exploited and protected sites, with algae and crinoids more abundant at exploited sites. Crinoids are the principle food of roman and were low in abundance where roman abundances were high, suggesting that the dominant top predator reduced crinoids. Furthermore, it substantiates the correlation of roman abundance with fishing effort, since habitat preferences can be ruled out by the observed causal predator-prey distribution pattern. Low algae abundances at protected sites correlated with high strepie (Sarpa salpa) frequencies within the fish communities encountered there. Strepie, a shoaling and abundant benthic grazer, does not compete for food with roman, suggesting a high potential for coexistence of the two species. As expected, and found by other studies, life history traits of roman differed between protected and exploited sample-sites. With a significantly lower age-at-maturity and age-atsex- change, the exploited population showed a typical response to fishing effort. The sex ratio of this protogynous hermaphrodite was found to be sustained at healthy levels by phenotypic plasticity. However, one important additional factor was highlighted by the study; the average condition factor of the protected population was significantly lower (0.0283 g/cm³) compared to the exploited population (0.0295 g/cm³). This was probably due to the higher intra-specific competition for lower food abundance in the protected area. Interestingly the diving and fishing survey methods yielded similar mortality results for roman. Total mortality rate estimates derived from length frequency analysis from the diving and fishing survey were not different (0.32 and 0.29 y⁻¹, respectively) as were natural mortality rate estimates (0.24 and 0.19 y⁻¹, respectively). Natural mortality rate (M) estimates indicated by Pauly’s and Hoenig’s relationship were similar (0.25 and 0.23 y⁻¹, respectively). Detailed yield-per-recruit (Y/R) and spawner biomass-per-recruit (SB/R) analyses were presented for different levels of M, varying age-at-recruitment (tR) and fishing mortality (F). Current tR (7.60 y) and F (0.16 and 0.25 y⁻¹, from the diving and fishing dataset, respectively) suggests an optimal exploitation of the population in the exploited part of the study area. However, a separate SB/R analysis of the male part of the population showed their vulnerability to over-exploitation, even at reduced age-at-sex-change from fishing. There therefore remains a high risk of recruitment failure for the roman population. Of course MPAs can be used to measure stock status directly if the influence of factors such as cachability, habitat and sampling method on CPUE assessments can be limited or reduced. The experimental design in this study allowed for contemporary CPUE comparisons across the border of the Goukamma MPA. Results were similar to those obtained by the SB/R analyses. CPUE extrapolations therefore, using small MPAs, can provide reliable and consistent estimates, and offer a practical alternative to conventional assessment strategies. This study has highlighted the importance of ensuring a well structured and comprehensive survey design when undertaking a comparison of protected and exploited marine areas. The results provide a comprehensive framework for future management of the Goukamma MPA and other protected areas along the temperate coastline of South Africa.