Generalised additive models (GAMs) were applied to survey data to assess the influence of dissolved oxygen, water temperature and year of sampling upon the presence/absence of small (≤15 cm TL), medium (16-34 cm TL) and large (≥35 cm TL) size classes of deepwater Cape hake Merluccius paradoxus captured off the west coast of South Africa. Data were obtained from surveys using the RV Dr Fridtjof Nansen conducted in 2003 and from 2005 to 2013 during summer (January-February). Among the variables investigated, oxygen was the most important for the small size class (juveniles), with both low and high constraints (two-sided, 'just right' option), whereas for the medium and large size classes the oxygen effects were one-sided (avoiding lows). This finding, in combination with other published information, suggests that the Orange Banks is a nursery ground for juvenile M. paradoxus and that the area covered by this nursery ground can vary with the optimal oxygen concentration. The temperature constraint was generally wider and weaker than that for oxygen, being two-sided for the small and medium hake and one-sided (avoiding highs) for the large hake. The medium hake displayed the greatest tolerance to the investigated variables, which resulted in the widest distribution for this size class. Temperature, oxygen and sampling year play an important role in determining the distribution of M. paradoxus, but details of the biology (life cycle) of the species, such as its pelagic-demersal transition and associated movements, are no less important. . The list of models considered and the performance measures of each of the models, considered for the three size classes of Merluccius paradoxus, are presented in Tables S1 and S2. The climatologies of bottom temperature and dissolved oxygen are shown in Figure S1. Additional model diagnostics (e.g. calibration plots and residual variograms) are shown in Figures S2 and S3.