White-tailed deer (Odocoileus virginianus) dispersal and excursion movements impact gene flow, population dynamics, and disease spread. Knowledge of movement characteristics and habitat selection during dispersal could provide the ability to predict how deer may relocate themselves within the landscape while providing managers valuable information regarding corridors for gene flow and disease spread. My objectives were to 1) test the hypothesis that extra-home-range movements occur as a strategy to broaden mating opportunities or as a means of searching for higher quality resources in this fragmented landscape, 2) compare occurrence rates and path movement metrics for dispersal and excursion movements to determine if underlying differences in behavior exist that would allude to mechanisms for accepting the risk of leaving a home range, 3) create and test the performance of expert opinion and step selection function resistance models at predicting deer dispersal movements, and 4) fit single and multiple random walk models to dispersal path data to determine movement states occurring within this behavior. During 2011-2014, I placed GPS collars programmed to take hourly locations on 49 fawn and yearling white-tailed deer in agricultural east-central Illinois to record dispersal and excursion movement paths. Linear mixed effects models were used to test for differences in path characteristics between sexes and ages (e.g., distance, straightness, duration, and speed). I used known-fate models, with demographic, temporal, and home range variables as covariates, to obtain dispersal and excursion occurrence rate estimates. Ten dispersal and 54 excursion movement paths were recorded during the study. Dispersal paths were longer and straighter (P < 0.001), and trended toward being longer in duration (P = 0.080) and faster in speed (P = 0.085), than excursion paths. Dispersal rates differed by sex (annual estimate ± SE with ages pooled: males 0.81 ± 0.12, females 0.16 ± 0.15) and were greatest during the breeding season (14-day estimates for males: winter 0.00 ± 0.01, fawning 0.02 ± 0.1, prebreeding 0.01 ± 0.01, and breeding 0.31 ± 0.15, and females: winter 0.00 ± 0.01, fawning 0.01 ± 0.1, prebreeding 0.01 ± 0.01, and breeding 0.04 ± 0.03). In contrast, I found no evidence that excursion rates were influenced by demographic, temporal, or home range variables (annual: 0.78 ± 0.06). I compared 2 methods of resistance modeling for predicting deer dispersal paths. I created an expert opinion survey and calculated a dispersal step selection function (SSF) to rank habitat variables and create 2 types of resistance maps to dispersal movements. I created least-cost paths with the starting and ending points coinciding with recorded dispersal paths within these 2 resistance maps. I compared the created paths to actual paths and a null straight line path using a path deviation index (PDI), path straightness, and path cost/m as variables of interest. Experts ranked land cover variables differently by season, applying a lower resistance value to agriculture cover during the summer/fall period, so 2 versions of the expert opinion resistance maps were created. For the SSF, I found that both forest cover and streams had significant nonlinear effects on deer dispersal movements. Assuming that all other factors remained constant, deer were more likely (≥ 0.50 probability) to move toward forested habitat when located < 335 m and when > 2795 m away. Deer dispersal movement behavior relating to streams followed a similar trend but with deer always having > 0.56 probability to move toward a stream than away. For least-cost path comparison, I conducted 3 ANOVAs (α = 0.05 throughout) to test for mean differences in calculated path metrics for all paths with path type as a within-subjects effect. I found no difference between the expert opinion survey model, the SSF model, and the null straight line model at predicting dispersal paths. PDI values were similar among all models (F1,9 = 0.004, P = 0.99). The SSF paths (0.91 ± 0.02) were significantly straighter then both the expert opinion (0.57 ± 0.03) and actual deer paths (0.44 ± 0.06; F1, 9 = 32.65, P < 0.001), but the expert opinion path did not differ from the actual path (P = 0.08). Path costs differed within the expert opinion survey resistance map (F1, 9 = 14.21, P < 0.001) with the expert opinion least cost paths (23.64 ± 3.14) having lower resistance/m than both the actual (46.15 ± 3.85) and straight line paths (48.74 ± 3.94; P < 0.001 for both). However, the actual and straight line paths did not differ (P = 0.872). There were no difference in path costs between the actual, SSF least-cost path, and straight line paths within the SSF resistance map (F1, 9 = 0.454, P = 0.64). I constructed and attempted to fit single and multiple random models to collected dispersal locations using WinBUGS v. 1.4.3. I was able to fit a single random walk model to deer dispersal paths but the more complex random walk models did not converge. I used the average parameter values derived from the single model to simulate deer dispersal paths and compared them to observed Net Squared Displacement. My simulated paths underpredicted deer displacement for 0.90 of individuals. Deer in east-central Illinois are very mobile and commonly make excursion movements throughout the year. The fact that I recorded differing dispersal rates within the same study area over a temporally short period from a previous study highlight the need for managers to obtain recent estimates of population parameters when making management decisions. The frequency of excursion movements should not be overlooked by managers as it is a behavior that can influence gene flow and potentially spread disease across the landscape at a localized scale. The preference for forest and stream habitats during dispersal can allow managers to focus surveillance or culling efforts around these types of habitats. The application of the least-cost path modeling technique appears to be ineffective at predicting deer dispersal paths, which emphasizes the importance of validating these types of models with actual data. The results from the random walk analysis highlight the need to collect as many locations as possible during temporally-short movements to understand the mechanisms acting upon them.