Introduction
Foraging decisions are essential because they directly affect survival through nutrient and energy acquisition (Sih, 2011). In most cases, foragers choose from a variety of food options. Many factors affect these choices including differences in availability and accessibility. Understanding the factors that affect these decisions provides insight into how animals budget their time in addition to the selective pressures which shape feeding preferences (Ydenberg, Brown, & Stephens 2007).
One way researchers document these choices is through cafeteria trials, in which foragers are offered many food items simultaneously (Morrison & Hik, 2008; Woods 2009; Dostaler, Ouellet, Therrien, & Côté 2011). Choices made during these trials presumably reflect choices made during normal foraging in the wild. Cafeteria trials allow for direct tests of the balances between energetic costs and benefits because they allow researchers to control some of the variables that affect foraging decisions such as search time and relative abundance of different food types. It is also important to consider the number of options offered in these experiments, as foraging preferences can vary with the diversity of options (Wang, Wang, Liu, Huang, & Hodgkinson 2011). Moreover, cafeteria trials conducted in the laboratory can control for several environmental variables.
A wide variety of approaches can discriminate foraging preferences from cafeteria trials (Elston, Illius, & Gordon 1996). Analysis of preference using pairwise comparisons can be problematic because not only does this oversimplify foraging decisions, but the probability of observing a spurious preference for any food type increases with the number of comparisons (Thomas & Taylor, 1990). Moreover, ranking foraging preferences based on pairwise comparisons overlooks the importance of the presence of alternative food types on foraging decisions (Woods, 2009). Other problems plague some approaches. For example, if a forager avoids a given resource, this can cause mathematical or computational problems because of zeros in the data (Elston, Illius, & Gordon 1996). In addition, zeros are a problem for statistical approaches that require chi-square tests to determine preferences (Neu, Byers, & Peek 1974), as these tests require most expected frequencies to be greater than one (van Emden, 2008).
Resource consumption curves provide an excellent approach to measuring foraging preferences (Krebs, 1999). Foragers are monitored over time to determine how much resource is consumed in the absence of other resources. The utility of the consumption curve is dependent on the time scale of measurement of the amount consumed. Measuring consumption too often may interrupt or negatively affect foraging behavior, introducing a bias into the analysis. Also, this index assumes preferences are additive and there is no interaction between different resource types.
The preferable approach to analyzing data from cafeteria trials is to develop an index that incorporates preferences across many food types simultaneously. The Shannon Diversity Index accomplishes this task. Considering each food item, the index quantifies not only the number of food types in the diet (analogous to species richness) but also the distribution of food items within each food type (analogous to evenness). Species diversity is useful if foragers avoid certain food types, but evenness is preferable if foragers sample all food types available, and provides the most information concerning feeding preferences. Given a starting amount of resources, a null hypothesis of equal preference can be calculated and compared to the observed value with a simple hypothesis test. One shortcoming of the index is that is only indicates if a preference exists but does not allow for identification of which food type is preferred without further testing. It does, however, allow comparisons between groups of individuals (i.e., different species, males versus females, etc.) and could be applied to any amount of n food types. Its simplicity is also a benefit.
This study assesses the efficacy of using Shannon diversity indices to test hypotheses concerning foraging preferences. When foragers include more resources in their diets, or increase their diet breadth, this should be reflected by higher values of the Shannon Diversity index (Hs ). Nuanced differences in diets can be determined by testing evenness, or the distribution of selected food items into different resource categories. In the case of equal preferences for all food types, values of Shannon Evenness (J’ ) should be unaffected by diet breadth. This interpretation of Evenness provides an excellent null model, as it assumes the complete lack of partial preferences (Stephens & Krebs 1986), and generates a single value to compare to actual diets. Null models demonstrated how the indices reacted to changes in diet breadth, and field preferences were tested using data generated from a series of cafeteria trials on common forest granivores.
Forest granivores provide an excellent model system to address questions about foraging decisions. Seeds are a convenient resource for representing the variety of food items occurring in natural environments because they are ubiquitous, easily measured, and readily manipulated. Moreover, studying seed consumption can lead to insights about how granivores may affect forest composition through their foraging decisions (Meiners & Stiles, 1997; Lobo, Green, & Millar 2013). Foragers encounter many different seed types leading to species-specific preferences. For example, deer mice demonstrate divergent seed preferences: the woodland deer mouse (Peromyscus maniculatus gracilis ), shows a strong preference for red maple (Acer rubrum ) over sugar maple (A. saccharum ) seeds, while the seed preferences of the white-footed deer mouse (P. leucopus ) are less clear (Cramer, 2014). In the Great Lakes region, two common and well-studied seed predators are eastern chipmunks (Tamias striatus ) and least chipmunks (Neotamias minimus ) (Hsia & Francl, 2009; Myers, Lundrigan, Hoffman, Haraminac, & Seto 2009; Penner & Devenport, 2011; Jenkins & Devenport, 2014).
Ecological differences between these species should lead to differential foraging preferences. T. striatus is a habitat generalist (Snyder, 1982) and is more common in the rodent community than N. minimus , a habitat specialist that prefers mesic areas (Verts & Carraway, 2001). There is also a distinctive body size difference:N. minimus is approximately half the mass of T. striatus . In addition, there are behavioral differences between these chipmunks:T. striatus can be active year-round and are larder hoarders, whereas N. minimus undergoes seasonal torpor in the winter months and stores seeds in small dispersed caches.
Regardless of these differences, studies in other systems with multiple species of chipmunk have demonstrated prevalent interspecific competition. This may be ameliorated in Great Lakes forests by limiting similarity, in which foraging choices between similar species is mediated by seed size. Granivores differentially harvest seeds of different sizes, and the size of seed selected for is positively correlated with body size [Brown & Lieberman, 1973; Emmons, 1980). Small seeds may not provide enough energy for larger granivores to survive, but smaller seeds are more accessible to smaller granivores. Thus, it is reasonable to expect that differential seed preferences between these species are based on seed size.
Two sets of hypotheses were generated for the null model analysis and the field-test using actual recorded diets. For the null model, we predicted that the value of the Shannon index would increase as more food types were added to the diet, but that Evenness would remain constant regardless of diet breadth. For the chipmunk model system, we predicted that each species would demonstrate partial preferences based on seed size: T. striatus should select larger seeds and N. minimus select small seeds.