Extended: The hypothesis was proposed that woodland caribou (Rangifer tarandus caribou) in North America had declined due to wolf predation and over-hunting rather than from a shortage of winter lichens (Bergerud, 1974). In 1974, two study areas were selected for testing: for the lichen hypothesis, we selected the Slate Islands in Lake Superior (36 km2), a closed canopy forest without terrestrial lichens, wolves, bears, or moose; for the predation hypothesis, we selected the nearby Pukaskwa National Park (PNP) where terrestrial lichens, wolves, bears, and moose were present. Both areas were monitored from 1974 to 2003 (30 years). The living and dead caribou on the Slates were estimated by the ‘King census’ strip transect (mean length 108±9.3 km, extremes 22-190, total 3026 km) and the Lincoln Index (mean tagged 45±3.6, extremes 15-78). The mean annual population on the Slate Islands based on the strip transects was 262±22 animals (extremes 104-606), or 7.3/km2 (29 years) and from the Lincoln Index 303±64 (extremes 181-482), or 8.4/km2 (23 years). These are the highest densities in North America and have persisted at least since 1949 (56 years). Mountain maple (Acer spicatum) interacted with caribou density creating a record in its age structure which corroborates persistence at relatively high density from c. 1930. The mean percentage of calves was 14.8±0.34% (20 years) in the fall and 14.1±1.95% (19 years) in late winter. The Slate Islands herd was regulated by the density dependent abundance of summer green foods and fall physical condition rather than density independent arboreal lichen availability and snow depths. Two wolves (1 wolf/150 caribou) crossed to the islands in 1993-94 and reduced two calf cohorts (3 and 4.9 per cent calves) while female adult survival declined from a mean of 82% to 71% and the population declined ≈100 animals. In PNP, caribou/moose/wolf populations were estimated by aerial surveys (in some years assisted by telemetry). The caribou population estimates ranged from 31 in 1979 to 9 in 2003 (Y=1267 - 0.628X, r=-0.783, n=21, P<0.01) and extirpation is forecast in 2018. Animals lived within 3 km of Lake Superior (Bergerud, 1985) with an original density of 0.06/km2, similar to many other woodland herds coexisting with wolves (Bergerud, 1992), and 100 times less than the density found on the Slate Islands. The mean moose population was 0.25±0.016/km2 and the wolf population averaged 8.5±0.65/1000 km2. Late winter calf percentages in PNP averaged 16.2±1.89 (25 years); the population was gradually reduced by winter wolf predation (Bergerud, 1989; 1996). The refuge habitat available is apparently insufficient for persistence in an area where the continuous distribution of woodland caribou is fragmented due to moose exceeding 0.10/km2 and thereby supporting wolf densities ≥6.5/1000 km2. A second experimental study was to introduce Slate Island caribou to areas with and without wolves. A release to Bowman Island, where wolves and moose were present, failed due to predation. Bowman Island is adjacent to St. Ignace Island where caribou had persisted into the late 1940s. A second release in 1989 to the mainland in Lake Superior Provincial Park of 39 animals has persisted (<10 animals) because the animals utilize off-shore islands but numbers are also declining. A third release to Montréal Island in 1984 doubled in numbers (up to 20 animals) until Lake Superior froze in 1994 and wolves reached the island. A fourth release was to Michipicoten Island (188 km2) in 1982 where wolves were absent and few lichens were available. This herd increased at λ= 1.18 (8 to ±200, 160 seen 2001) in 19 years. This was the island envisioned for the crucial test of the lichen/predation hypotheses (Bergerud, 1974: p.769). These studies strongly support the idea that ecosystems without predators are limited bottom–up by food and those with wolves top-down by predation; however the proposed crucial test which has been initiated on Michipicoten Island remains to be completed and there is a limited window of opportunity for unequivocal results.