According to Wikipedia, in 2004 the American Ornithologists' Union made the cackling goose a new species, splitting it away from the Canada goose. The British followed suit in 2005.
What was the basis for this split? Why was a new species required here? Were offspring of these two found to be consistently infertile, or was the change based on something else?
The "white-cheeked" geese have long been subject to taxonomic controversy. From the 1920s to the 1950s, authorities have classified them into between one and four species. Although Aldrich in 1946 asserted there was near-unanimous agreement among Arctic biologists that there were two species, from the 1950s Delacour's taxonomy was most often followed, which had one species (Branta canadensis) and twelve subspecies.(1)
What changed the consensus were DNA studies. Mitochondrial DNA (mtDNA), which is passed on without recombination, showed distinct sequences supporting two species. One RFLP analysis of mtDNA suggested the split between these two species happened about one million years ago.(1)
The species divide mainly along size and breeding location lines. The Cackling Goose (Branta hutchinsii) is smaller and breeds on the Arctic tundra. The (present) Canada Goose species (Branta canadensis) mostly takes a larger form and breeds mostly in non-tundra areas.(1)
The accepted answer doesn't point out that you can get the answer straight from the AOU if you have access to a science library: they publish explanations of changes to the AOU Checklist in The Auk! The one involving Branta hutchinsii is from Banks et al. 2004; the exact text they use are:
pp. 59–60. Several genetic studies of geese, including recent work with mitochondrial DNA (van Wagner and Baker 1986, Shields and Wilson 1987, Quinn et al. 1991, Paxinos et al. 2002, Scribner et al. 2003) have verified previous suggestions based on differences in voice, nesting habits, habitat, and timing of migration, as well as in color and size (e.g. Brooks 1914, Aldrich 1946, Hellmayr and Conover 1948), that the forms treated as the single species Branta canadensis by all previous AOU Check-lists and most other works actually constitute at least two species, and further that each of the two species may be more closely related to another member of the genus than to each other. Thus, we divide B. canadensis by recognizing a set of smaller-bodied forms as the species B. hutchinsii, and rearrange our representatives of the genus in the sequence bernicla, leucopsis, hutchinsii, canadensis, sandvicensis. Additional analysis may result in further splitting.
So, yes, this is mainly because of mtDNA analysis and because of several studies between 1987 and 2003, as described in the other answer.
