Individual diet specialization (IDS) is widespread and can affect the ecological and evolutionary dynamics of populations in significant ways. Extrinsic factors (e.g., food abundance) and individual variation in energetic needs, morphology, or physiology, have been suggested as drivers of IDS. Behavioral traits like exploration and boldness can also impact foraging decisions, although their effects on IDS have not yet been investigated. Specifically, variation among individuals in exploratory behavior and their position along the exploration/exploitation trade-off may affect their foraging behavior, acquisition of food items and home-range size, which may in turn influence the diversity of their diet. Here we analyzed stable carbon and nitrogen isotopes in hair of wild eastern chipmunks, Tamias striatus, to investigate the influence of individual differences in exploration on IDS. We found that exploration profile, sex, and yearly fluctuations in food availability explained differences in the degree of dietary specialization and in plasticity in stable carbon and stable nitrogen over time. Thus, consistent individual differences in exploration can be an important driver of within-population niche specialization and could therefore affect within-species competition. Our results highlight the need for a more thorough investigation of the mechanisms underlying the link between individual behavioral differences and diet specialization in wild animal populations.

From 2012 to 2016, we live-trapped wild eastern chipmunks, Tamias striatus, within three sites near Mansonville, southern Québec, Canada (45°05’N; 72°25’W) on rectangular grids of 7.84 ha (sites 1 and 2) and 4.00 ha (site 3), in mixed northern hardwood forests. We ran linear trapping transects composed of Longworth traps placed every 40 m and alternately placed along parallel adjacent transects. We trapped daily during the activity period of chipmunks from early May until early September every year. At each capture, we recorded identity, sex, body mass, and age.

Exploration test

We quantified among-individual differences in exploration using a novel environment test in an open-field arena on 396 individuals from 2012 to 2017. The experimental arena consists of an empty white box (80 × 80 × 40 cm), with no refuge, and a grid marked on the bottom. This arena was placed at the center of the trapping grid. Before the test, we placed the captured individual for one minute in an opaque tubular entrance for acclimation. Exploration in the novel environment was video recorded for 90 s. Using The Observer XT11 software (Noldus), an exploration score was calculated for each test by counting the number of lines crossed by an individual during 90 s (Réale et al. 2007). We ran two tests per individual at 1-year intervals (generally during the first and the second year of life) to obtain a phenotypic value representative of the individual’s life. We then obtained the exploration profile of each individual by calculating best linear unbiased predictors (BLUPs) for exploration using an LMM controlling for date, hour of the test, waiting time before the test, site, and number of trials, all specified as fixed effects, and year, chipmunk and experimenter IDs as random effects. We used a method proposed by Dingemanse et al. (2020) to account for the variance around the BLUPs in the statistical analyses.

Hair samples and stable isotope analysis

We collected hair samples from 107 adult individuals over four successive years from 2013 to 2016. Each hair-sampled individual had been tested for exploration during the study period. We cut the recent darker hair before and after molting in late June, close to the skin, at the same place on the right thigh. Samples of the main food items available (except fungi and insects) were collected between June and October 2013 and dried until constant weight before measuring isotope values (δ¹⁵N/δ¹³C).

The protocol for isotopic analysis followed Dammhahn and Goodman (2014). Isotope data are presented in ‰ δ¹⁵N relative to nitrogen in air or δ¹³C relative to Pee Dee Belemnite calculated as follows: δX = [(Rsample/Rstandard) − 1] × 103, where δX is either δ¹⁵N or δ¹³C, and R is the respective ¹⁵N/¹⁴N or ¹³C/¹²C ratio. For quantification of δ¹⁵N and δ¹³C, approximately 1 g of total dry hair was enclosed into tin capsules and processed through an isotope ratio mass spectrometer. We carried out mass spectrometry analyses on the 2013-14 hair and food source samples at the Center for Stable Isotope Research & Analysis (KOSI) in Göttingen (Germany) using an isotope ratio mass spectrophotometer (Delta Plus, Finnigan MAT, Bremen, Germany). We carried out the analyses on the 2015-16 samples at the GEOTOP laboratories of the Université du Québec à Montréal (Montréal, Canada), using an isotope ratio mass spectrophotometer (Isoprime 100, Elementar, Vario MicroCube). Internal reference materials are normalized on NBS 19-LSVEC for δ¹³C and on IAEA-N1, N-2 and N-3 for δ¹⁵N. To obtain non-biased niche width metrics (i.e., ellipse area, see below), we corrected the raw data for laboratory effects. We used two conservative linear models with the laboratory and year of analysis as fixed effects on δ¹⁵N and δ¹³C values as response variables. On GEOTOP data, we subtracted 0.57‰ in δ¹⁵N and added 0.38‰ in δ¹³C (i.e., estimates of laboratory effect).

Red maple and American beech masting

Chipmunks feed mostly on seeds from the dominant American beech and red maple and also the less common sugar maple (Acer saccharum) These tree species produce seeds in mast events every 2–3 years in southern Québec. To quantify temporal variation in the availability of key food sources for chipmunks, we estimated red maple and American beech seed productions. We used 26 plastic buckets (with 0.06 m² opening) installed on two metallic poles at 50 cm above the ground and 1 m away from the trunk of a tree. On each site, we chose 8 to 13 mature red maple and 13 mature American beech trees uniformly distributed on the trapping grid and with a minimum circumference at breast height of 31 cm. We collected bucket contents twice yearly, in July and October. For each tree species, we calculated the average number of seeds collected per m². We use the term mast year for 2013 and 2015 and non-mast year for 2014 and 2016 to indicate temporal variability in food availability.