Well, while doing research on Google Scholar for an upcoming post on chimp vs. human evolution, I found the long-lost paper in question, Evolutionary Consequences of Fallback Foods. And it's worth quoting extensively, for clarity's sake (emphasis added):
Although FBFs are not always operationally defined, the term most often refers to items assumed to be of relatively poor nutritional quality and high abundance, eaten particularly during periods when preferred foods are scarce (Hanya 2004; Knott 2005; Laden and Wrangham 2005; Lambert et al. 2004; Ungar 2004, Yamakoshi 2004a).
Accordingly, one can operationally define FBFs as foods whose use is negatively correlated with the availability of preferred foods (cf. Altmann 1998; Conklin-Brittain et al. 1998; Doran et al. 2002). Two points about the concept should be stressed. First, the definition of fallback foods implies a distinction between preference and importance. Whereas preference is a matter of dietary choice, importance is a measure of dietary composition. For example, the importance of a particular item might refer to its use as a proportion of total caloric intake, or the percentage of total feeding time spent eating it. Although preferred foods are in some circumstances important, they need not be, as illustrated in Table I. Similarly, high importance of an item in the diet does not necessarily imply positive selectivity or high preference. Fallback foods tend to be foods of low preference but of high importance seasonally.This is the distinction I was driving at in my "Definitions & Clarifications" post. I.e., the fact that a primate regularly eats a large amount of a given food does not necessarily make that food "preferred," nor does infrequency of consumption make a given food "fallback."
So, if a fallback food is "negatively correlated with the availability of preferred food," what then is a preferred food according to Marshall & Wrangham?
We advocate a classic definition: preferred items are overselected. In other words, preferred food items are selected disproportionately often relative to their abundance within the population’s habitat (Leighton 1993; Manly et al. 2002). Thus, preference is defined as the relationship between 2 parameters: availability and usage. Neither rarity nor frequent utilization is alone sufficient to characterize a food as preferred.Fallback foods influence adaptations for processing (i.e., mastication, digestion, etc.) and preferred foods influence adaptations for harvesting (i.e., locomotion, tool use, etc.).
On the question of whether meat was a "fallback food of last resort," as the Permavegan contends, much will depend on which stage of hominin evolution we choose to focus on. To me, it is clear that by the time Homo arrives on the scene, meat is a preferred food, often over-selected (at the end of the Pleistocene, to great ecological detriment) as our intelligence and technological capacities made us more efficient hunters.
For the australopiths, the picture is a bit murkier. There is some evidence that they ate meat, but whether it served as a preferred food or a fallback food for them is open to debate, which I guess is what this little tete-a-tete between vegans is all about.
But that gives us the additional problem of trying to decide which australopiths or other pre-Homo species we decide to claim is ancestral -- and thus, most relevant -- to Homo. And I don't think we're going to figure that one out any time soon.