View or print this publication - Northern Research Station - USDA ...

We tlhus neglect two potential sources of neighbor effects because experienced herbivores may well be able to evaluate the
expected value of a patch or a plant from a distance, e.g., by appearance or scent. Hjfilten et al. (1993) have discussed in a
greater length how selection between and within patches can affect associational refuges.
In short, our herbivores are naive, but we assume that they can learn to evaluate the defensive status of the plants if
appropriate cues are present. The consumption process is divided into two parts (bites). From our earlier assumptions, it
follows that the risk of the first bite is m. Each bite by a herbivore implies a fitness cost for the plant. Plant defenses can in
various ways reduce the total cost of herbivory: reducing the fitness cost caused by a herbivore per bite, increasing mortality
(d) of herbivores during a bite, and inducing feeding aversion that can be either unconditional (the herbivore does not take the
second bite of any food) or conditional (the herbivore does not take the second bite if a specific cue is present). In our case of
two bites, unconditional feeding aversion is equivalent to increased mortality as in both cases further herbivory is stopped.
We thus consider only conditional feeding aversion and the parameter a indicates the probability that feeding aversion is
induced during a bite. We also assume that the cost of herbivory is h if the herbivore survives and no feeding aversion is
induced, while the cost of herbivory will be h' if the herbivore happens to die or if a feeding aversion is induced during a bite
(O < h' < h). We assume specifically that non-defensive (N) plants cause no mortality (d - 0, a - 0), while there are two kinds
of defensive plants; those (D) that will kill the herbivore during the bite (d = 1, a - 0), and those (S) that cause a conditional
feeding aversion before any toxic effects appear on herbivore survival (a - 1, d = 0).
Both kinds of defenses (D and S) reduce the average costs that herbivory implies to the player. In addition to these
direct fitness effects, various neighbor effects can arise if herbivores are allowed to move from the player to the neighbor and
vice versa. We assume that a surviving herbivore moves after the first bite with a probability e. For simplicity, feeding
aversion is not allowed to affect this probability, and the parameters h, h', d, and a assume the same values during the second
bite as earlier during the first bite. Taken together, these assumptions imply that the average cost (H) of herbivory over two
bites will be
H=[(1-di)(1-ai)h + (1-di)ah' + d'ih](1 +nii+nij)
ni_= (1-d,)(1-a)(1-e)
nij = { (1-di)(1-ai)ei (i,j=S) }
(1-dj)ej (i v j aS)
where n_ implies the probability of a second bite for the player by a herbivore initially invading the player (denoted by i), and
n_j by a herbivore moving from the neighbor (denoted by j) to the player. The latter term will be (1-d)(1-a)ej if feeding
aversion induced by the neighbor protects the player (i.e., both S), and (1-d)e if the neighbor does not induce a feeding
aversion (i.e., the neighbor N or D) or if the player does not possess the cue required for the maintenance of feeding aversion
(i.e., the neighbor S and the player either N or D). In our specific case (ei = e. l = e), H = h[1 + (l-e) + n_j]for a player adopting
N (d = 0, a = 0), and H = h'(1 + nij)for a player adopting D (d = 1, a = 0) or S (d = 0, a = 1). When H is multiplied by m we
will get, following Augner et al. (1991), a measure of the herbivory load, Hm, for the payoff matrix.
This highly simplified model system elucidates clearly that the fitness of the player may well depend on the neighbor.
In our case, this is so if herbivores can move between the plants (0 < e