Examples Offoraging Behavior In Animals

24 hr

48 hr

24 hr

48 hr

Figure 6.2. Results of a grazing study examining how sheep trade off diet preference against intake rate. In this experiment, replicate flocks of sheep were stocked on replicate paddocks in which one-half of the paddock contained white clover and the other half contained perennial ryegrass. Different paddocks were managed to achieve different contrasts in sward surface height (SSH): 6 cm clovervs. 6 cm grass, 3 cm clovervs. 6 cm grass, or 3 cm clovervs. 9 cm grass. The investigators estimated species-specific intake rates forthese sward surface heights to be 3 cm clover = 3.58 ± 0.4 g dry matter/min; 6 cm clover = 4.66 ± 0.8 g dry matter/min; 6 cm grass = 2.49 ± 0.4 g dry matter/min; 9 cm grass = 3.99 ± 0.4 g dry matter/min. The nature ofthe results is complex. Animals could easily have achieved a monospecific diet. Their expressed diet preference is neither based entirely on intake rate nor on plant species, but on some combination ofthe two. To complicate matters, in addition to changing their diet preference, the animals also altered their grazing time and hence their total daily intake. (After Harvey et al. 2000.)

48 hr constraint intensifies, and animals should concentrate on the digestive process, choosing fewer diet items of higher digestibility. However, when time is short or food is less abundant, animals should concentrate on the ingestion process, choosing more food types that have faster handling rates. This requirement for flexible diet selection nicely illustrates why the prior ranking of food types (as in the diet model) may be irrelevant when digestion constrains foraging.

In arid and semiarid environments, water constrains diet selection. For example, Manser and Brotherton (1995) demonstrate this constraint on the diet selection of dwarf antelopes during the dry season. They show that in order to meet minimum daily water requirements, dik-diks (Madoqua kirkii) fed on plant species they normally avoided during the wet season. Given a choice between foods with differing water contents, grasshoppers' diet choice depends on their state of dehydration (Roessingh et al. 1985); they choose higher water content over energy content when dehydrated. Digestive constraints operate for many animal species, whether it's too much sugar in the phloem sap ingested by the aphid seeking nitrogen or too much lignin in the grass eaten by the goat seeking digestible organic matter. David Rauben-heimer contrasts the nutritional challenges faced by herbivores with those faced by carnivores in more detail in box 6.1.

Foraging theorists often think of digestion as a constraint, but students of herbivory have considered the adaptive design of digestive processes. For example, Mathison et al. (1995) have suggested that ruminants have some control over gut retention time, which they can adjust to optimize assimilation rates. Many disagree, noting that the weight ofevidence suggests that mechanistic factors such as particle size determine passage rate (for more discussion, see, e.g., Illius et al. 2000). Ultimately, the animal controls mastication and rumination, which in turn control particle size, so clearly, ruminants do have some degree ofcontrol over this process.

Many plants produce secondary metabolites that either make the plant less nutritious to some animals (e.g., tannins) or make the plant toxic in sufficient quantities (e.g., alkaloids). Guglielmo et al. (1996) demonstrated that the presence of coniferyl benzoate in aspen leaves strongly influenced ruffed grouse (Bonasa umbellus) diet selection. Dearing (1996) found similar results for the North American pika (Ochotonaprinceps), and Tibbets and Faeth (1999) demonstrated that the presence of alkaloid-producing endophytic fungi altered leaf-cutting ants' choice of grass leaves. Bernays and Chapman (1994, chap. 2) and Launchbaugh (1996) give general introductions to the role of plant secondary metabolites in herbivory.

Plant secondary metabolites may also influence diet selection among parts of the same plant. Boer (1999) showed that pyrrolizidine alkaloid concentrations were higher in the youngest (and most nutritious) leaves of Scenecio jaco-baea plants, so that cotton leafworms (Spodoptera exiguq) and a noctuid moth (Mamestra brassicae) both preferred the older leaves. More generally, Hirakawa (1995) noted that when the classic diet model is modified to consider toxins, partial preference may occur for one diet item while all others follow a zero-one rule (see chap. 5 in this volume). Hirakawa also showed that the prey selection criterion changes with the intensity of the toxin constraint, making it impossible to rank diet items a priori. These results are qualitatively different from those reported by Stephens and Krebs (1986).

An animal's state can strongly influence nutritional, digestive, and some secondary metabolite constraints. One approach to the study of current physiological state has been to alter an animal's state through fasting. Experiments routinely use fasting to motivate animals to feed, but fasting should be used with caution because it can alter both diet preference and diet selection (Newman,

Penning et al. 1994; Edwards et al. 1994). States other than hunger per se can be important as well. My colleagues and I demonstrated that sheep that had previously grazed grass had a stronger preference for clover when given a choice between the two, and that sheep that had previously grazed clover had the reverse preference (Newman et al. 1992). Parsons, Newman et al. (1994) demonstrated that such effects can influence diet preference over a period of several days. While the desire to compensate for some imbalance in the previous diet might explain these results, the missing component has yet to be identified. Bernays et al. (1997) suggest that "novelty" per se is the mechanism for incorporating even unpalatable food items into the diet and provide experimental evidence to support this hypothesis in a grasshopper (Schistocerca americana).

Previous state sometimes appears in experiments in the form of hidden variables. Many large mammalian herbivores are maintained on high-energy, low-bulk pelleted foods when not taking part in experiments. These diets can cause gastrointestinal acidity and even ulcers, and subsequent diet selection may be greatly influenced by these pathologies. For example, acidosis leads cattle to self-select more fiber in their diet (see box 6.2).

Raubenheimer and Simpson (1993) have introduced a useful framework for examining the effects of physiological state on diet choice as well as total intake (or feeding time). I describe their framework in figure 6.3, showing how animals may use complementary plants to reach some target intake. More interestingly, their framework gives some insights into foraging behavior when the animal's diet is nutritionally deficient. This basic framework has proved powerful in a variety ofsituations with a variety ofspecies. Here is but one recent example. Behmer et al. (2001) provided locusts (Locusta migratoria) with pairs ofsynthetic food sources that differed in their protein and digestible carbohydrate content (7% P:35% C and 31% P:11% C). Neither food source alone was optimal (for growth), but together they were complementary. The locusts were able, over the course of4 days, to respond to their physiological state by adjusting their intake ofthe two complementary diet items to satisfy their target intake of 19% P:23% C. However, when fed each of these diet items singly, locusts attempted to defend both their protein and carbohydrate goals, as in figure 6.3E. In addition to levels of specific macronutrients (or even micronutrients), digestion rate itselfmay be a physiological state variable that influences diet selection. For example, degus selected food plants based on plant quality (water content and nitrogen:fiber ratio) and on mean gut retention time (Bozinovic and Torres 1998).

So far, the physiological and morphological constraints we have considered affect the processing of food—in other words, the "postingestive" consequences of diet choice. However, many constraints act before herbivores

Figure 6.3. Graphs of nutrient space, with nutrient A on the y-axis and nutrient Bon the x-axis, both measured in grams. The target intake of each nutrient is shown as a solid circle. Any given food item has a fixed ratio of the two nutrients, and we can represent that food item as a line from the origin. Rauben-heimer and Simpson (1993) call these lines "rails." If two complementary foods are available (one rail on each side of the intake target), then the animal can achieve its target by selecting a mixed diet. This system is particularly powerful for investigating dietary priorities. This can be done by feeding animals on a variety of single food items, one at a time, and examining their intake. In each graph, there are several hypothetical food items, each available one at a time. The open circles represent hypothetical intake of each item. (A) Ifwe saw this intake behavior, itwould tell us thatthe animal is more concerned about its intake of nutrient B than of nutrient A and always seeks to satisfy this requirement (although sometimes gut constraints might prevent this, particularly for food items that are quite different from the target ratio). (B) Similarly, this intake behaviorwould demonstrate a desire to always satisfy the nutrientA requirement. (C) Intake behaviorthat always seeks to satisfy both nutrient requirements, even if this means exceeding the total intake target (sum of the x and y coordinates). (D) Intake behaviorthat seeks to meet one nutrient requirement while maintaining total intake at or below the target. (E) A foragerthat seeks the optimal compromise between its two nutrient requirements. This forager eats until a point on the rail that is geometrically closest to the target intake. Raubenheimer and Simpson demonstrated that locusts tend to behave as in part E with respect to carbohydrate and protein. (After Raubenheimer and Simpson 1993.)

ingest their food. A herbivore's spatial memory for locations of different foods or their qualities may limit its diet selection.

Memory constraints are perhaps less important in large vertebrates than our intuition might suggest. Edwards et al. (1996b), Laca (1998), and Dumont and Petit (1998) have demonstrated that some large grazing mammals possess excellent spatial memory and can use it to improve the quality of their diets. For example, sheep with 6 days' experience were able to visit exclusively four patches containing food among thirty-two patches in an 800 m2 grid, using spatial memory alone (Edwards et al. 1996b). Provenza and others have demonstrated that these same animals have very good temporal memories about toxins (e.g., Provenza 1995 a, 1995b, 1996). Of course, there may be significant fitness costs to forgetting that a plant contains a toxin, but in cases that do not involve toxins, the penalty of forgetting the postingestive consequences may be small. Consider an animal grazing two species of grass that differ in protein and carbohydrate. A herbivore may take a mixed diet simply because it cannot remember the nutritional consequences ofthe less preferred species, and must resample that grass to refresh its memory.

In addition to memory constraints, animals may face pre- or postingestive perceptual constraints. For example, can a grazer recognize the difference between two species of grass without eating them? Edwards et al. (1997) showed that sheep could distinguish grass and clover (a common pasture mixture) without eating them. Howery et al. (2000) showed that cattle aided by visual cues associated with preferred and non-preferred foods were more efficient at achieving their preferred diets than uncued animals. This difference was particularly evident when the food items were not located in fixed positions (and hence the cattle could not use spatial memory). Other researchers have demonstrated that several large herbivore species can tell the difference between preferred and non-preferred pelleted foods without eating them (for a review, see Baumont 1996).

Odor can play an important role in phytophagous insect diet choice. Omura et al. (2000) show that oak sap odor stimulates feeding and influences host choice behavior in two butterflies (Kaniska canace and Vanessa indica). Chapman and Ascoli-Christensen (1999) discuss the physiological mechanisms by which sucrose acts as a phagostimulant and nicotine hydrogen tartrate acts as a feeding deterrent in grasshoppers. Of course, odor is not the only cue used by phytophagous insects. Fereres et al. (1999) show that some aphids use color cues to select host plants. Leaf surface chemicals may also be important. Lin et al. (1998a, 1998b) show that alpha-tocopherylquinone acts as a feeding stimulant and forms the basis for host plant choice in cottonwood leaf beetles (Chrysomela scripta) feeding on poplar trees (Populus deltoides). For a thorough and thoughtful discussion of the role of chemical cues in host plant selection by phytophagous insects, see Bernays and Chapman (1994, chap. 4).

Perceptual constraints and cues are obviously important in phytophagous insects, and entomologists have studied them intensively. We do not know how important such perceptual constraints are for larger vertebrate herbivores. Postingestive perceptual constraints may also include the ability to match postingestive consequences with some external cue (Provenzaet al. 1996). For example, grass with higher nitrogen content (hence more crude protein) may also be greener. Villalba and Provenza (2000) easily conditioned sheep to use strong flavors to distinguish between forages with different postingestive consequences. Whether they use such cues in nature awaits further investigation.

6.5 Intake Rate

The product (encounter rate) x (handling time) x (bite mass) specifies a forager's intake rate. Classic foraging models such as the diet model consider handling time and bite mass to be constants, even though these parameters vary considerably. For discrete food items, this is a reasonable simplification because there is likely to be little variation in item size, implying that variation in bite mass and handling time might be small and conveniently ignored. In some cases, however, researchers have found that what we treat as a constraint should be treated as a decision (e.g., Newman et al. 1988). In the case of herbivores, although there are physical limits to bite mass and handling time, studies have often demonstrated that foragers can voluntarily adjust their handling time. Before we consider this possibility, let's first consider bite mass and handling time as constraints.

Short-Term versus Long-Term Intake Rates

We can roughly divide studies of intake rates into two types: (1) studies of short-term intake rates as a basis for diet selection and (2) studies of long-term intake rates as a consequence of diet selection. Neither provides us with a complete picture. Students ofherbivory have debated the relevance ofshort-term intake rates to large grazing mammals (e.g., see Newman et al. 1992; Illius et al. 1999). Short-term rate studies may not be informative because they look at what happens over only a few hundred bites, while large grazing mammals may take tens of thousands of bites in a day. Looking at behavior over a few minutes in isolation ignores the importance of total grazing time. That said, however, diet selection by grazing mammals sometimes correlates with achievable short-term intake rates from the plant species on offer (e.g., Illius etal. 1999).

Short-Term Intake Rate

As before, I divide constraints into environmental and physiological or morphological. On the environmental side, the major constraint is vegetation structure. To ingest vegetation, animals must first sever (prehend) the herbage and then, perhaps, chew (masticate) it. Allden and Whitaker (1970) point out that intake rate in grazers is prehension bite rate multiplied by bite mass. This simple observation has led to extensive (indeed, obsessive) consideration of the determinants of bite mass. Most bite mass studies focus on the role of vegetation structure. Plant height and density influence the maximum bite mass achievable from vegetation of a given species. Researchers have

Figure 6.4. Laca and colleagues used hand-constructed sward boards to investigate the relationship between sward bulk density (g/m3) and sward height (cm) in determining bite mass (g dry matter/m2; shown as contours) for cattle grazing alfalfa and dallisgrass. (After Laca et al. 1992.)

investigated these relationships repeatedly, often using hand-constructed swards or turves; studies by Black and Kenney (1984; Kenney and Black 1984) and Laca et al. (1992; fig. 6.4) provide classic examples. Other approaches have also been tried; a particularly amusing example is that of Burlison et al. (1991; fig. 6.5). Ungar (1996) provides a nice review of this area of research.

Physiological and morphological constraints on intake rate (as opposed to total daily intake, which we will consider shortly) largely focus on jaw morphology. The time it takes to sever the vegetation may be a physical limitation of the jaw muscles (see Newman, Parsons et al. 1994 for discussion), but it may also depend on the tensile strength of the vegetation (see, e.g., Prache and Peyraud 1997). Illius and Gordon (1987) demonstrate that incisor arcade breadth (the distance between the right and left fourth incisors), more closely than body mass, predicts variation in bite mass.

One difficulty with this area of work is that all measurements of animals' short-term intake rates are measurements of what animals do, not what they can do. We know from numerous studies that animals can voluntarily increase their short-term intake rate with no change at all in vegetation structure (e.g., Greenwood and Demment 1988; Dougherty et al. 1989; Newman, Penning et al. 1994). My colleagues and I used a simple mechanistic model to demonstrate the flexibility that grazers have to use behavior to increase intake rate. We showed that within a forage species, grazers have little latitude to alter their handling times, but some flexibility to alter their bite mass (Newman, Penning et al. 1994). I will consider behavioral decisions regarding intake rate in more detail in the next section.

Researchers in the field now appreciate the mechanistic aspects of bite mass and hence intake (e.g., see review by Baumont et al. 2000), and these mechanisms form the basis ofseveral widely used models ofgrazing behavior (e.g., Spalinger and Hobbs 1992; Newman, Parsons et al. 1994; Parsons, Thornley et al. 1994; Thornley et al. 1994; Pastor et al. 1999; Illius 2006).

Long-Term Intake Rate

Long-term intake rates are not easily measured in the field. It is difficult or impossible to see the size of each bite. Peter Penning has developed an excellent device for recording details of intake behavior in larger herbivores at pasture; figure 6.1 shows the device on a cow. By weighing the animal before it goes on the pasture, collecting its dung and urine, weighing it again after a period of time, and correcting for insensible weight loss, intake rates can be estimated over longer time periods. Penning et al. (1991), for example, used the bite recorder to investigate the relationship between pasture surface height, tiller density (which tends to be negatively correlated with pasture surface height), and bite mass. They found that grazing time and prehension bite rate declined with increased pasture surface height, while rumination time and mastication bite rate increased with pasture surface height.

Clearly, the constraints discussed for short-term intake rates sometimes determine long-term intake rates as well, but there are also times when animals behaviorally adjust their short-term intake rates to manipulate their long-term intake rates. A number of studies that have examined the effects

Figure 6.5. Modified metabolism crates allowed sheep access to a 0.56 * 0.46 m area of pasture. (After Burlison etal. 1991.)

of different physiological states for animals grazing the same pastures clearly demonstrate this. For example, Greenwood andDemment (1988), Dougherty et al. (1989), and Newman, Penning et al. (1994) report that fasting in sheep and cows can cause a voluntary increase in intake rate by 27% to 72% with no changes in sward structure. Similar results have been reported during lactation in these animals (e.g., Penning et al. 1995; Gibb et al. 1997; Prache 1997; Patterson et al. 1998). Moreover, Iason et al. (1999) have shown that when grazing time is limited, sheep may voluntarily increase their intake rates; if food is abundant enough, this behavior can compensate for the time limitation (see also Ydenberg and Hurd 1998).

So if animals can voluntarily raise their intake rates, why don't they always eat this quickly? Increased intake rates may reduce vigilance behavior, which imposes a cost (real or perceived) of increased predation in the longer term (Underwood 1982; Illius and FitzGibbon 1994). The choice of intake rate additionally implies a corresponding digestion rate as well as rumination requirements, both of which may have fitness (opportunity cost) consequences (Greenwood and Demment 1988). While physical, morphological, and environmental constraints on intake rate have received extensive attention, researchers have largely ignored the potentially important contribution of behavioral decisions. This area certainly requires further research.

Intake rate in social animals should represent a balance between the need for vigilance and intraspecific competition (both scramble and interference). Rind and Phillips (1999) nicely demonstrated this with cows. They found that prehension bite rates were lower in both small and large foraging aggregations. However, the effects of social constraints on intake rates have not been extensively studied, largely due to the technical difficulty ofestimating intake rates of animals at pasture. It is more common to consider the effects of social constraints on grazing time rather than intake, something I do in the next section.

6.6 Grazing Time

Intake rate multiplied by grazing time determines total daily intake. Many theoretical studies take grazing time to be a constraint (see, e.g., Verlinden and Wiley 1989). In many cases, this assumption is appropriate, as a number of environmental factors can constrain grazing time. Again, the social context may be important. Penning et al. (1993) showed that grazing time is a negatively accelerating function of aggregation size for sheep grazing a monoculture (hence with no diet selection; fig. 6.6), and Sevi et al. (1999) and Rind and Phillips (1999) found similar results in cows and sheep. In addition, Rook

Figure 6.6. Penning et al. replicated flock sizes and used bite recorders (see figure 6.1) to record the grazing behavior of sheep. The sheep were maintained on a monoculture of perennial ryegrass. The data clearly show that individual animals and small groups (< 5 animals) behave differently from larger grazing groups. The best fit curve has the equation [grazing time (min/24 hr)] = 629 - 311 * exp(-0.46 * group size). (After Penning et al. 1993.)

Figure 6.6. Penning et al. replicated flock sizes and used bite recorders (see figure 6.1) to record the grazing behavior of sheep. The sheep were maintained on a monoculture of perennial ryegrass. The data clearly show that individual animals and small groups (< 5 animals) behave differently from larger grazing groups. The best fit curve has the equation [grazing time (min/24 hr)] = 629 - 311 * exp(-0.46 * group size). (After Penning et al. 1993.)

and Penning (1991) and Rook and Huckle (1995) present strong evidence for synchronization in grazing behavior in sheep and cows. Such synchronization may well be a general phenomenon in large social animals. In addition, day length itself can constrain grazing time. On some pastures, lactating ewes may need to graze nearly all the daylight hours to meet their daily energy requirements. The requirements for other fitness-enhancing behaviors, such as vigilance, may also constrain the time available for foraging (Underwood 1982; Illius and FitzGibbon 1994).

Gut passage time constrains the behavior ofsome herbivores. Forage quality can be so poor that animals can starve to death, even on ad libitum food. Plant quality characteristics such as lignin and cell wall content affect gut passage time (for review, see Iason and Van Wieren 1999). The size of food particles entering the gut, which depends in part on the animal's intake behavior, also affects passage time (Gidenne 1992; Kennedy 1995; Wilson and Kennedy 1996; Schettini et al. 1999). Sheep, for instance, have a relatively constant number of jaw movements per minute when grazing (ca. 150; Penning et al. 1991), and these jaw movements must be partitioned between prehension and chewing bites. To increase the prehension bite rate, a herbivore will have to reduce the amount of chewing it does, and this can lead to slower passage rates through the gut (or increase rumination requirements).

Gut size itself, which scales with body size allometrically (see section 6.2), constrains passage time. Bell (1970) and Demment and Van Soest (1985) describe how food use relates to body size (see also Illius and Gordon 1990; Iason and Van Wieren 1999). Size-based differences in forage use operate both between and within species (e.g., sexual dimorphism), and forage use may change as an individual develops. Bernays and Chapman (1970b, 1970a) showed that changes in the mandible of the grasshopper Chorthippusparallelus from the first to the fourth instar correlate with a shift in diet from thinner to thicker-leaved grasses.

The animal's ability to detoxify or excrete plant secondary compounds also constrains grazing time. Lauriault et al. (1990; Dougherty et al. 1991) have shown that grazing time in cattle declines in the presence of alkaloids from endophyte-infected tall fescue. Pfister et al. (1997) have shown that cattle grazing tall larkspur (Delphinium barbeyi), which contains a potentially toxic alkaloid, can regulate their intake to remain below the toxic threshold.

When energy requirements increase (e.g., due to lactation) or food availability declines, animals often voluntarily increase their grazing time in response (Arnold 1975; Clutton-Brock etal. 1982;Penning 1986; Penning etal. 1991). As with the question of intake rate behavior, we might ask why animals don't spend more time grazing when they can. Ultimately, the likely evolutionary reason is that grazing takes time away from other fitness-enhancing activities, such as vigilance, rumination, and social interactions (for a theoretical consideration of this issue, see Thornley et al. 1994; Newman et al. 1995). However, it may be possible to gain some mechanistic insight into the flexibility of the behavioral repertoire by looking at what happens when grazing time constraints are no longer applicable—a situation that happens in captivity when animals are provided with high-quality concentrated feed. Ungulates evolved some 40 million years ago, but we have only housed them in ways that severely curtail their foraging behavior for a few decades. Such studies have led, for example, to the realization that jaw movements serve a function beyond their mechanical effect on food: they promote salivation, which, in ruminants at least, has a vital buffering effect on fermentation in the rumen. It is entirely possible that such a mechanism could, in some pastures, lead to grazing times longer than necessary to satisfy energy demands. Georgia Mason considers this issue in more detail in box 6.2.

6.7 Return to Question One: Where to Eat?

In many cases, habitat or patch choice reflects both diet selection and intake rate considerations. For example, Wallis de Vries and Daleboudt (1994) found that cattle may select patches based on long-term rate maximization (however, see Distel et al. 1995), though additional considerations such as nutrient content (e.g., phosphorus) may also be important (Wallis de Vries and Schippers 1994). More generally, the presence of preferred forage species (e.g., Crane et al. 1997; Watson and Owen-Smith 2000) and differences in forage quantity and nutritional quality (e.g., Wallis de Vries et al. 1999; Van der Wal et al. 2000) seem to be important in these decisions. In some cases, abiotic factors, such as the time since the last fire event (e.g., Irwin 1975; Coppedge and Shaw 1998), affect plant quality in patches. In others, the behavior of the animals themselves alters patch quality; for example, through dung and urine deposition (e.g., Keogh 1975; Day and Detling 1990; Lutge et al. 1995). There is evidence that animals shift their patch preferences in response to both kinds ofconsiderations.

Stokke (1999) found a sex-based difference among elephants in habitat use. Sex-based differences often reflect body size differences, although Perez-Barberia and Gordon (1999) observed sex-based differences in the patch choices of Soay sheep even when the body size differences where removed. Nevertheless, body size differences can mean that "patch quality" is a relative quality. A model by Illius and Gordon (1987) showed that allometric relationships between bite size, metabolic requirements, and body size can explain differences in habitat choice within species, especially between males and females ofdimorphic species. Body size difference may be the mechanism that determines the outcome of interspecific competition, as in the example of cattle and mule deer (Loft et al. 1991).

Mysterud et al. (1999) point out that factors other than food availability can determine habitat choice, but may reflect trade-offs between, for example, food and protective cover. They demonstrated that roe deer (Capreolus capreolus) habitat use did not correlate with the availability of herbs, but did correlate with the availability of canopy cover. Similarly, Ginnett and Demment (1999) found that when female Masai giraffes (Giraffe camelopardalis tippelskirchi) were caring for offspring, they selected habitats without cover for predators. Parasites, too, can influence patch and habitat choice. Cooper et al. (2000) demonstrated a trade-off between patch quality and the presence of sheep dung (infected with Ostertagia circumcincta larvae), and Duncan and Cowtan (1980) showed that horses may choose foraging habitats based on the densities of blood-sucking flies.

Just as they influence other foraging decisions, social interactions can influence habitat and patch choice. A model by Beecham and Farnsworth (1998) demonstrated that a species-specific spacing preference can constrain patch choice and resource utilization, resulting in a short-term reduction in intake rate and an increase in the variability of resource utilization. Bailey (1995)

showed that initial patch selection by groups of steers was often determined by the behavior of one or two individuals. Social learning in early life accounts for sexual differences in roe deer habitat selection that earlier authors attributed to competition (Conradt 2000).

Patch use leads to intake rate depression. The work of Laca et al. (1994) nicely demonstrates this for cattle. They showed that bite mass decreased more in tall, sparse patches than in short, dense patches. While bite mass declined, time per bite did not change, resulting in intake rate depression as exploitation time increased. Patch depletion naturally leads to marginal value-like considerations for patch-leaving rules. Baharav and Rosenzweig (1985) found behavior consistent with Charnov's model (Charnov 1976b) in Dorcas gazelles (Gazella dorcas). According to the marginal value theorem, patch exploitation depends on travel time between patches, and the spatial distribution of patches determines the travel time. Dumont et al. (1998) showed that sheep exploit patches more intensively when they must travel greater distances between patches. More generally, Wallis de Vries (1996) modeled the interacting effects of group size, inter-patch distance, and resource distribution pattern (degree ofaggregation ofpatches) on the spatial distribution offoraging time for an ungulate. He showed that travel costs can be very important, even when small.

Nevertheless, support for the patch model has not been unanimous. For example, Jiang and Hudson (1993) preferred a mechanistic explanation for patch leaving in wapiti based on their lateral neck angle and biokinetic considerations. Lundberg and Danell (1990) argued that marginal value theorem explanations are less useful than optimization of handling time for each ramet for moose browsing birch stands.

6.8 Decision Making with Multiple Objectives

I have been developing a view that herbivore foraging should be understood as a function of the animal's constraints and objectives. For an animal attempting to answer each of the four big questions (where to eat, what to eat, how fast to eat, how long to eat), there are often multiple objectives. For example, sheep choosing a diet from grass/clover pastures would like to graze with conspecifics; avoid areas of previous defecation to guard against parasite infection; eat a 65% clover, 35% grass diet; take bites of 53 mg clover and 30 mg grass at a rate of 83 bites per minute from each, with 17 chews per gram of clover and 27 chews per gram of grass; and of 660 available minutes, spend 334 grazing clover, 166 grazing grass, and 160 not grazing (all weights measured as dry mass; Dumont and Boissy 1999, 2000; Hutchings et al. 1998; Parsons, Newman et al. 1994; Newman, Penning et al. 1994). To some extent, the animal can control each of these objectives. For example, we know that hungry sheep can increase their grazing time by 80—185 minutes, increase their bite masses 54%—290%, decrease their bite rates 5%—41%, and decrease their mastication by 9%—20% per gram (Newman, Penning et al. 1994). We know that sheep whose previous diets comprised a monoculture of either grass or clover choose more or less clover, respectively, than sheep that were recently grazing a mixture of grass and clover (Parsons, Newman et al. 1994). The constraints are the same for the fasted and non-fasted sheep. The animals were all tested on the same pasture; the changes were entirely voluntary.

The reductionist approach to herbivore foraging behavior will not always work, however. In the previous example, we know a good deal about all of these objectives and others, but very little about their relative or absolute importance. When sheep find it impossible to meet all of these objectives simultaneously, how do they respond? Are there some objectives that sheep defend vigorously and others that they sacrifice for higher-order objectives? Which ofthese multiple objectives takes priority, and in which circumstances?

Foraging location, diet choice, intake rate, and grazing time all have fitness consequences, and how herbivores trade off the multiple objectives within and between these broad categories of behavior differs by species, body size, ecosystem, time of year, age, and many other factors. There is not a one-to-one mapping of any of these four dimensions onto fitness. An animal may compensate for a low long-term intake rate by becoming less selective, thus increasing its encounter rate; or more selective, thus improving its diet quality; or it may increase its grazing time, or some combination ofthese tactics. Studying just one of the big questions in isolation may not yield tremendously clear results.

Marc Mangel and others (e.g., Mangel and Clark 1988; Houston and McNamara 1999; Clark and Mangel 2000) have advocated the use of dynamic state variable models (see chap. 1 in this volume). I used this technique to consider how animals select diets and grazing time in the face of intake and passage rate constraints and predation danger (Newman et al. 1995). That work provided a behavioral explanation for total grazing time, which earlier investigators had always treated as a simple constraint. Useful as we felt that model was, it still considered only two of the big questions simultaneously, and I do not believe that the dynamic programming approach will be useful in addressing the integration of the myriad of objectives herbivores routinely face. The problem is computational. Beyond a few state variables and a few decisions, the computational demands of dynamic programming quickly make it intractable. The technique can provide useful analyses ofsome ofthe questions in isolation or in combination, but not all of them.

Computational limitations aside, to apply dynamic programming to understand how herbivores balance their multiple objectives, we ultimately need to know how each objective and their combinations map onto evolutionary fitness. It is unlikely that we will ever reach that goal. How should we proceed? In my opinion, the problem with our ongoing research programs is that they are too mired in "traditional" experimental design. We rely heavily or exclusively on the univariate analysis of variance or multiple regression approaches, but decisions with multiple objectives are, necessarily, multivari-ate problems. To address the questions posed at the start of this section, I think we need to borrow some techniques from economists. Microeconomic theory embraces the notion that preferences are based not on single attributes, but jointly on several attributes. Economists get people to reveal the utility that the attributes of goods or services have for them by examining the tradeoffs that they make between those attributes in the process of making choice decisions.

Economists have developed extensive theory and techniques for addressing decisions with multiple objectives. I find this approach most intuitive when I think about how people value "the environment." All other things being equal, we would like to have clean air, clean water, high biodiversity, charismatic species, unspoiled natural landscapes, rainforests, and so on. However, we would also like to eat; care for our children, the sick, and elderly; improve our education and social welfare systems; control our agricultural pests; and so on. All other things are rarely equal, and we have to make choices and trade-offs among our multiple objectives. The field of research called "economic valuation" is aimed at understanding the choices we make, finding out how much we value particular states ofnature and how combinations ofthese states map onto our utility. In my opinion, this is exactly what we must do for herbivore foraging decisions. Substitute "eat highly nutritious forage" for "have clean air"; "avoid predators" for "control our agricultural pests"; and so on, and the usefulness of economic valuation becomes apparent.

Economists use two broad categories of methods to understand how humans value their multiple environmental objectives: revealed preference techniques and expressed preference techniques. In the remainder of this section, I will briefly introduce some of these techniques and point out how I think they can be used to study herbivore foraging behavior. It is not my intention to teach the full background theory or discuss all the caveats and problems with each technique. Rather, my goal is to stimulate researchers to find out more about these techniques and then apply them where appropriate.

Revealed Preference Techniques

With free-roaming animals, replicated controlled experiments are difficult or impossible. In these cases, we can make use of revealed preference techniques. Economists sometimes call these "indirect techniques" because the researcher does not directly ask people questions, but rather studies what people do and then deduces their preferences from their observed behavior. In a sense, students offoraging behavior do this already, but in relatively unsophisticated ways. Revealed preference techniques offer us a chance to gain deeper insight into how herbivores sort out their multiple objectives. I will quickly review two such techniques, hedonic pricing and travel cost methods. In my opinion, travel cost methods show more promise.

Hedonic Pricing

The most common use of hedonic pricing deals with housing prices. This technique relies on the assumption that an individual's utility for a house is based on the attributes it possess, such as size, location, school district, air quality, general level of environmental quality, and so on. In certain circumstances, one can use multi-market data to derive society's "willingness to pay" for attributes such as clean air.

It may be possible to use either total energy costs or total time costs as a substitute for housing price and the foraging habitat as a substitute for the house. Then, by comparing the choices ofmany individual animals for nearby foraging habitats that differ in many foraging attributes (e.g., forage species available, intake rates achievable, social context, predator/cover attributes, etc.), it may be possible to derive explicit values and trade-offs (substitutions) for each of the foraging attributes. One hedonic pricing study showed that a particular group of people was willing to pay S5,500 for a marginal improvement in nitrogen oxide levels. By knowing the price they were willing to pay, we can see how they might trade off air quality for, say, water quality, whose value a similar study estimated as S41,000. Applying this method in a herbivore study might tell us, say, that gazelles are willing to pay 3.7 MJ of energy for a marginal improvement in predation danger and 2 MJ for a marginal improvement in grazing time. It might then be possible to estimate the trade-off between predation avoidance and grazing time. I reiterate that many caveats would go along with such a study, and the investigator should fully appreciate these problems before undertaking it.

Travel Cost Methods

Travel cost methods use the price people are willing to pay to travel to a non-priced recreation site as a means of inferring the value of environmental attributes that a person experiences at that site. These methods consistently show that as the price of travel (distance) increases, the rate of visits to the site falls. For example, one travel cost study found that a group of people was willing to pay an extra S7 per use to improve water quality from "boatable" to "fishable" and a further $14 per use to improve the quality from "fishable" to "swimmable."

Much as with the hedonic pricing method, we can examine the travel costs (time or energy) that individual animals are willing to pay for a change (improvement) in the dimensions of foraging quality as they move from one foraging habitat to another. The animals reveal their preferences, and their willingness to make trade-offs among preferences, through their willingness to pay the cost of travel. For example, we know from Parsons, Newman et al.'s (1994) work that sheep prefer about 70% clover in the diet when grass and clover monocultures occupy adjacent sides of a pasture. How important is this particular mixture to them? How willing would they be to trade off this mixture for, say, grazing time? One way to see this is to increase the distance the animal has to travel between grass and clover. As we do so, the animal has to pay increasingly higher travel costs to defend the mixture, and we may see how important the 70% objective is versus whatever grazing time objective it has. The sheep example could obviously be studied experimentally, but similar kinds of studies could be conducted in nonexperimental situations with free-roaming animals. We could use travel cost methodology by examining the choices of individuals that must travel different distances for access to each of the choices.

Two comments are warranted here. First, revealed preference techniques will be most useful if we do our best to measure all of the relevant behaviors and objectives. The point is to see how the animal combines or trades off these multiple objectives. This cannot be done if we ignore one or more of the major behavioral decisions the animal must make. Second, we need to recognize that these approaches identify correlations, not mechanisms (for this we need expressed preference approaches). If we are to ultimately understand, and so predict, herbivore behavior, we must understand the relevant mechanisms.

Expressed Preferences

With invertebrates and captive (often agricultural) vertebrates, replicated controlled experiments are possible. In these cases, we don't need to rely on animals revealing their preferences to us; we can ask them directly. Economists use two main expressed preference methods: contingent valuation and conjoint analysis. Again, I will briefly review these two methods with an eye toward how they might be applied to foraging studies. There are extensive

Table 6.1 Measurements of the value of resources to farmed mink

Consumer surplus (kg)

Table 6.1 Measurements of the value of resources to farmed mink

Consumer surplus (kg)

Was this article helpful?

0 0

Post a comment