Stable isotopes of carbon and nitrogen (δ15N and δ13C) in Antarctic toothfish (Dissostichus mawsoni) muscle were measured to study the trophic connections of Antarctic toothfish in the Ross and Amundsen Sea regions (CCAMLR Subareas 88.1 and SSRU 88.2H respectively). More than 1700 toothfish were sampled by scientific observers on New Zealand vessels between 2005/06 and 2013/14. More than 100 samples each of four by-catch species (grenadiers, icefish, moray cods and deep-sea cod) were also analysed, but only from the Ross Sea region. Appropriate corrections were made to account for the effect of lipid in muscle on δ13C in a way that did not compromise the accuracy of δ15N measurements. We found considerable variability in δ15N and δ13C of toothfish that were not explained by fish length, sex, location, depth or year of capture. We found statistically significant but small relationships between δ15N and fish length, region, and year: (1) on average, larger toothfish fed on prey of higher trophic level; (2) δ15N values of toothfish on the Ross Sea slope were lower (~0.8‰ between medians) than in other areas (northern Ross Sea seamounts, Ross Sea shelf and 88.2H); (3) after allowing for fish length, sex, and location, there was a small but significant reduction in δ15N values (0.3‰) between 2010 and 2014. This study found that δ13C values of toothfish in the Amundsen Sea region were substantially higher than in the Ross Sea region (medians differed by 1.4‰). This result implies that most of toothfish sampled from the Amundsen Sea had not mixed with fish from the Ross Sea region for the 6 months – 2 years before capture (the likely turnover time of protein in fish muscle). The result is consistent with different spawning populations in the two regions. In the north and slope areas of the Ross Sea, “prey polygons” were used to investigate whether δ15N and δ13C values for toothfish were consistent with isotope values for the likely main prey species, assuming standard trophic fractionation factors. For the Ross Sea slope region, the predator-prey polygon largely enclosed the toothfish isotope values, indicating that the by-catch species caught the same area could have comprised most of the toothfish diet. However, the predator and prey isotope data were inconsistent for the northern seamount region because δ13C for by-catch species were higher in the north than the slope area of the Ross Sea, but δ13C for toothfish were similar in the two areas. The reason for this difference is as yet unclear but is consistent with toothfish feeding at a low rate in the northern Ross Sea area and staying there for less than a year.