Methods
Null models were constructed in R (R Core Team, 2019). The resource base comprised ten types of food, and foragers would randomly select up to ten items of each food type. Foragers could select zero items for any of the food types, and these were eliminated prior to calculation of the Shannon Index, and allowed for variation in diet breadth. Models were run for 1000 iterations, and it was noted that diet breadth was always greater than five items; stenophagic diets were excluded from analysis due to stochastic factors. Thus, the model was altered to allow for more variation in diet breadth: the diet breadth (analogous to species richness) was a random number between two and 10 and the food item selection was constrained to fit this number of different types of food items. A further constraint was eliminating complete specialist diets (where the diet breadth was one). This allowed calculation of Shannon Evenness: complete specialist diets have no evenness, as all food items are of the same type. So the final model selected food items (between one and 10) for a random number of food types (between two and 10). Shannon Diversity and Evenness were calculated, and this process was iterated 1000 times. These data were analyzed to determine how diet breadth affects Shannon Diversity and Evenness.
The proportion of each food type consumed (pi ) was calculated by dividing the number of food items of that particular type by the total number of all food items consumed. These proportions allowed calculation of the Shannon Diversity index (Hs ),
\begin{equation} H_{S}=-\sum_{i=1}^{S}{p_{i}\log p_{i}}\nonumber \\ \end{equation}
where S is the total number of food types (Maurer & McGill, 2011). In cases where individuals consume portions of every species offered, analysis of evenness provides additional information. Shannon Evenness Index (J’ ) was calculated by dividingHs by the natural log of S (Maurer & McGill, 2011).
The field study was conducted at the University of Notre Dame Environmental Research Center (UNDERC), located on the southwestern border of the Upper Peninsula of Michigan near Land O’Lakes, Wisconsin. Data for T. striatus were collected from May-July 2012, and data for N. minimus were collected in May-July 2013. We assumed that foraging preferences would not vary significantly between years, and that any observed differences in foraging behavior were species-specific. Because these preferences were tested in a laboratory setting using seeds obtained from a commercial supplier, the availability of seeds was constant and variation in seed quality was presumably reduced. Moreover, the trial methodology used in both years was identical.
Chipmunks were trapped using Sherman and Tomahawk traps baited with apples and peanut butter. Any pregnant or lactating females were released immediately at the site of capture and were not used in the study. Once captured, each chipmunk was placed in a glass aquarium with a plywood cover in the laboratory. The bottom of the aquarium was covered in shaved cedar bedding and a plastic container was present with polyfill bedding for nesting. A cloth covered three sides of the aquarium so the chipmunk was not disturbed by other activities in the room during the trial. The open side of the aquarium was placed near a window to allow natural light to enter throughout the day. Striped sunflower seeds and water were available ad libitum during an acclimation period which lasted from the time of capture until 2200h of the same day. The use of animals in this study followed the guidelines of the Animal Behavior Society (Ethical and Animal Care Committee of the Animal Behavior Society, 2012) and the American Society of Mammologists (Sikes & Gannon, 2011).
Seed preferences were determined using cafeteria-style feeding trials, which have been used in previous studies to accurately represent the actual food habits of small mammals (Page, Swihart, & Kazacos 2001; Lobo, Duong, & Millar 2009). At 2200h on the day of capture, the sunflower seeds were removed followed by an 8 hour overnight starvation period. Because chipmunks are diurnal, the subjects should have been inactive for most of the starvation period (Lobo, Duong, & Millar 2009). The starvation period ended at 0600h the following morning with the introduction of the trial seeds: sugar maple (A .saccharum ), beaked hazelnut (Corylus cornuta ), red maple (A . rubrum ), balsam fir (Abies balsamea ), and pin cherry (Prunus pensylvanica ), representing some of the most common woody species in the area (Curtis, 1959). Due to significant size differences, T. striatus were given twenty grams of each seed species and N. minimus were given ten grams of each for a total of 100g and 50g of food offered, respectively. The seeds were placed in an open container such that all were equally available to the chipmunk. At 1800h, twelve hours after the introduction of the seeds, the seeds were removed. The chipmunk was then released at the site of capture. Eleven N. minimus and fourteen T. striatus were used during this study.
The seeds remaining, both in the food container and among the bedding, were separated by species. The final mass of each seed species was measured and was subtracted from the initial seed mass to determine the total amount of each seed species consumed. During the N. minimustrials, some seed species gained mass during the trial. This was likely due to high humidity, so we conservatively assumed no seeds of that species were consumed during that trial. The probability that this strongly biased the analysis is low, as this only occurred in three of the trials.
The expected values of Hs and J’ for the proportion of seed species consumed were calculated assuming no seed preference (i.e. the chipmunk ate the same amount of each seed,pi = 0.20). These values were compared with a simple hypothesis test to the mean Hs andJ’ calculated across all trials for each chipmunk species. Hypothesis tests between the measured values of the Shannon Diversity and Evenness Index and the expected values would show if the chipmunks display a seed preference.
Specific seed preferences would then be determined with repeated measures analysis of variance (RM-ANOVA) that compared the proportion consumed of each food type by chipmunk species. Repeated measures tests were used because the consumption of one seed species is not independent of the consumption of another seed species since all five food types were introduced at the same time. The results of this test would show if seed preferences were consistent between chipmunk species.
Finally, data were transformed using the logit transformation (Warton & Hui, 2011) for use in t-tests of each seed type between values ofT. striatus and N. minimus . These tests would more specifically show for which seed species the chipmunks have different preferences. Since there were some values in which the amount consumed, and thus, the proportion consumed, were zero, a small value (0.00000001) was used to represent that resource. Results are reported as means ± standard errors.