View - Kowalewski, M. - Virginia Tech

DIETL AND KELLEY—PREDATOR-PREY ARMS RACESdistributed across all populations of a set ofinteracting species. It is easy to imagine how thesame defense can be adaptive locally even if thesource of the predation pressure changes amongpopulations. The other ecological predictions mayhold in specific cases, but this level of variation islikely lost (or masked or misinterpreted as a highdegree of phenotypic lability) in the fossil record.Thompson’s (1994, 1998) mosaic view of theevolution of species interactions suggests howreciprocal evolutionary change can shape speciesinteractions that are asymmetric at the species level(one species interacting with multiple species—i.e., a diffuse interaction) (Fig. 1) but are specificat the local level. Thus, although a predator mayprey on a number of species, a single prey speciesmay be most important in driving the evolution ofthe predator. There is some theoretical backing forThompson’s argument. Abrams (1991) has shownfrom models that interactions between a predator,a prey species, and a third species (either thepredator’s own predator, or a second prey species)often increase the magnitude of coevolutionaryresponse in the predator or prey (or both) toevolutionary changes in the additional species.Thus multiple prey species may intensify theresponse of a predator to its main prey. If empiricaldata are found to support Abrams’ models, one ofthe tenets of the hypothesis of escalation may bebrought into question (i.e., that predators areunlikely to respond specifically to any one preyspecies because of the catholic diet of mostpredators; Vermeij, 1987, 1994).What are the implications of the mosaichypothesis for the fossil record of speciesinteractions?—In general, the mosaic view of theevolution of species interactions incorporatesvariation that occurs over a larger spatial scale (therange of the species) than we usually consider (i.e.,within a local population) when evaluating armsrace hypotheses. As paleontologists, we are usuallyonly able to trace traits that have spread among allthe populations of a species, but this does not meanthat the coevolutionary meanderings or continualreshaping of interspecific interactions areunimportant. Interactions may come and go withinthe ecological context of a local population throughtime. A prey population with a geographic rangethat extends beyond that of its predator may escapeattack for many generations, lose (or more likelydecrease) its level of defense, and then fall victimagain as the range of the predator expands. Just asa lack of evidence for local adaptation is notconclusive evidence that coevolution (or for thatmatter escalation) has not occurred, the lack ofdirectional trends through successive stratigraphicintervals in traits important to an interaction is notevidence that biotic factors were unimportant.Gould (1990, p. 22) commented that “positivefeedback [should drive the evolution of aninteraction] to its furthest point in a geologicalinstant, while most actual events [trends] span tensof millions of years”; positive feedback is thoughtto require “locking of biotic interactions” over theduration of the trend. In other words, how can longtermbiotically driven trends be accounted for ifreciprocal adaptation is expected to occur onlyearly in the history of the interaction (see alsoFutuyma, 1985)? The mosaic view of selectionsuggests that the degree of adaptation to localconditions may increase or decrease depending onthe amount of contact with other populations. Thespatial distribution of genetic variation “determinesthe degree and scale over which populations havebeen (or are) evolutionarily independent, andconsequently free to evolve in response to localvariation in selection” (Grosberg and Cunningham,2001, p. 61). In general, species with restricted geneflow are expected to exhibit local adaptation tospatially varying selection; as the level of gene flowamong populations with different selective regimesincreases, the selective costs of local adaptationincrease (Grosberg and Cunningham, 2001). Thus,the geographic structure of a species may favor longtermtrends in the evolution of species interactions.Kelley and Hansen (2001) further added that thecondition of “locking” or maintenance of specificpredator-prey interactions is not necessary ifescalation (which does not require a strongcoevolutionary component) is the most importantprocess in the evolution of species interactions.Few interacting species have identical365