Visually estimating age of white-tailed deer (Odocoileus virginianus) using physical characteristics, commonly referred to as “aging on the hoof” (AOTH), has gained in popularity as part of management programs. However, AOTH has not been evaluated in reference to its accuracy or the accuracy of its users; and most importantly, AOTH is an unstandardized method for estimating age. To assess accuracy of AOTH as it currently is applied, we developed an accuracy examination consisting of a series of photographs of 70 wild, known-aged, antlered, male deer from south-central Oklahoma, USA, ranging in age from approximately 1.5 to ≥7.5 years. We distributed a pre-assessment questionnaire and the accuracy examination to registrants from the 2009 annual meeting of the Southeast Deer Study Group and to select individuals known to use AOTH. One hundred six wildlife professionals that commonly used the technique completed the assessment and examination. Overall accuracy averaged 36% when placing deer into year classes (min. = 16%; max. = 56%). Accuracy tended to be greater for younger year-classes (1.5–2.5 yr); accuracy generally declined as age increased. On average, 62%, 43%, 25%, 30%, 25%,15%, and 31% of deer were placed into the correct year-class for the 1.5, 2.5, 3.5, 4.5, 5.5, 6.5, and ≥7.5 year-classes, respectively. We provide the first known accuracy assessment by users of the AOTH technique. It appears that the AOTH technique lacks accuracy for placing deer into specific year-classes, which has implications for selective harvest decisions and research use. Accuracy may be improved by developing standardized protocols and criteria for using the technique and training observers. © 2013 The Authors Wildlife Society Bulletin published by Wiley Periodicals, Inc. on behalf of The Wildlife Society
Estimating age of harvested, white-tailed deer (Odocoileus virginianus) has a long-standing history for its use in management and research. Common methods to age deer include tooth replacement and wear (Severinghaus 1949), cementum annuli (Gilbert 1966, Ransom 1966), and more recently, aging deer using physical characteristics of body development. Severinghaus (1949) described an aging method for white-tailed deer based on tooth eruption and wear. Many consider this the “Holy Grail” of deer aging methods; however, contemporary evaluations of the technique (Gee et al. 2002, Meares 2005, Lewis 2010) concluded that accuracy was acceptable for fawns and yearlings, but declined dramatically for year-classes ≥2.5. Estimating age using tooth replacement and wear is the most widespread technique because animals can be aged at time of harvest or during live capture in the field. The cementum annuli technique involves sending samples of removed incisors to a laboratory for analysis, and therefore is performed best on dead animals to avoid the invasive removal of teeth from a live animal.
An alternative method of estimating age, primarily for male white-tailed deer, uses physical characteristics under the assumption that the magnitude of these physical characteristics are related predictably to ontogeny. This method is referred commonly to as “aging on the hoof” (AOTH) and has become popular in the past 10–15 years among deer enthusiasts and managers, as well as some deer researchers, despite lack of scientific development and testing (Demarais et al. 1999, Richards and Brothers 2003, Hellickson et al. 2009). Aging on the hoof is based on loosely defined physical characteristics that often include body and antler size, chest depth, spinal alignment, abdominal tautness, eye squint, and many others (DeYoung et al. 1989, Kroll 1996, Richards and Brothers 2003, Quality Deer Management Association 2012). If this method has acceptable accuracy and repeatability, it could be very useful in various management scenarios and age-specific research.
Accurately estimating age of wildlife is important for research and management. This is particularly true in species as popular and intensively managed as white-tailed deer. For example, deer researchers often incorporate estimated ages into analyses involving parameters such as body mass, antler size, reproductive success, population reconstruction, or survival (Strickland and Demarais 2000, Gove et al. 2002, DelGiudice et al. 2007, Webb et al. 2010). If age is a variable of interest in an analysis (i.e., used for making age-specific inferences) then the accuracy of the technique used to estimate ages can impact the usefulness of the data and the level of inference that can be made. Therefore, researchers and managers should be aware of the limitations of aging techniques (Dapson 1980) and potential ramifications of incorrect age assignments and plan analyses accordingly (Foley et al. 2012).
The popularity of AOTH appears to be growing based on its use and press received; however, despite the popularity and application of AOTH, standardization and validation of the technique using known-aged deer is lacking. For these reasons, we initiated this assessment and examination to determine levels of accuracy of wildlife professionals applying the AOTH method using physical characteristics of antlered, male deer of known-age that were captured in photographs during annual camera surveys. Our goal was not to develop a standardized protocol, but rather to assess the accuracy of AOTH as it currently is being applied and to examine the ensuing implications. Additionally, we sought to gather information on the participants' use of AOTH techniques and their perceptions about its accuracy. Combining the information gathered from the accuracy examination, and the participants' responses about the application of the technique, will help managers and researchers in interpreting results that may later be used to make management recommendations and inferences.
Although this study was an online questionnaire and examination, we obtained photographs of known-aged, male white-tailed deer from a study area located in south-central Oklahoma, USA. We described to the participants the study area, deer population, and management practices that may be considered important for the physical development of white-tailed deer. We captured and photographed deer on the Samuel Roberts Noble Foundation Wildlife Unit (NFWU) located in Coal, Hughes, and Pontotoc counties, Oklahoma. The NFWU was approximately 1,214 ha in size and was located 8.0 km south of Allen, Oklahoma, in the Cross Timbers and Prairies ecoregion (Gee et al. 1994). The study area was approximately 60% wooded and 40% open, with a high degree of interspersion (Gee et al. 1994). Annual precipitation averaged 105.5 cm during a 30-year period between 1971 and 2000 (Ada, Pontotoc County, OK; Oklahoma Mesonet, https://climate.mesonet.org). A 15-strand high-tensile electric fence surrounded the area beginning in 1993, which provided varying degrees of deer population control (Webb et al. 2009a, 2010). Habitat management included livestock grazing, rest from grazing, prescribed burning, and selective control of woody plants with single-stem herbicide treatments. Artificial feeding (e.g., shelled corn, food plots, etc.) was not an implemented management practice, although we provided shelled corn during capture and camera surveys. The deer population consisted of descendants from stockings of white-tailed deer (O. v. texanus) by the Oklahoma Department of Wildlife Conservation (Heffelfinger 2011).
Capture, Handling, and Camera Surveys
First, we obtained a sample of known-aged, antlered, male white-tailed deer using drop-net systems (Ramsey 1968; Gee et al. 1999, 2003) baited with shelled corn from 1997 to 2005. We sedated deer using Xylazine (3–6 mg/kg; Phoenix Scientific, St. Joseph, MO) or a mixture of Telazol® and Xylazine (4.4 mg/kg Telazol®; Fort Dodge Animal Health, Fort Dodge, IA, plus 2.2 mg/kg Xylazine). We used yohimbine (Abbott Laboratories, North Chicago, IL) at 0.125 mg/kg or tolazine at 0.4 mg/kg as an antagonist to the xylazine. We estimated deer age based on tooth replacement and wear (Severinghaus 1949, Gee et al. 2002). We placed deer into 2 groups: 1) known-aged deer consisting of animals first captured as fawns (approx. 0.5 yr) or yearlings (approx. 1.5 yr), and 2) deer with estimated ages consisting of animals first captured when ≥2.5 years of age. Because fawns and yearlings can be definitively aged (Severinghaus 1949, Gee et al. 2002), we only considered photographic records of deer first captured as a fawn or yearling (see below). To identify these deer, we marked them with 2 plastic livestock ear tags (i.e., one in each ear) containing unique color and number combinations. We used double-tagging to increase the probability that an individual retained at least one tag for future field or photographic identification.
Next, we obtained photographic recaptures of deer during infrared-triggered camera surveys (Jacobson et al. 1997, McKinley et al. 2006) that were conducted annually during January from 1998 to 2006. We used TrailMaster (TM-35; TrailMaster, Lenexa, KS) camera systems and Canon Sure-Shot A-1 cameras (Canon U.S.A., Inc., Lake Success, NY). We affixed camera systems to metal T-posts approximately 1.5 m above ground level. We baited camera sites with shelled corn (2–3 m from the camera) to facilitate ear-tag identification. We identified and catalogued photographs of identifiable known-aged, male deer in digital database management systems.
We developed an online questionnaire to characterize the participants' education and career-field, extent of use of AOTH, and their perceptions about AOTH. To obtain an adequate sample of AOTH users, we contacted attendants from the 2009 Annual Meeting of the Southeast Deer Study Group (through an e-mail registry obtained at time of registration), as well as other users of the AOTH technique. Participants were required to complete and submit the questionnaire prior to viewing or working on the accuracy examination (maintained via a secure log-in through The Samuel Roberts Noble Foundation website; www.noble.org). For the purposes of this accuracy examination, it was important to classify users based on their education and use of AOTH techniques. We categorized participants into 3 groups: 1) professional users, 2) professional non-users, and 3) non-professional users. Professional users (n = 106) were characterized by 1) a college degree (at minimum, a bachelors), 2) employment in a wildlife-related field, and 3) use of AOTH techniques. We report only the results for the professional user group because this group was the largest sample and we were interested in obtaining the most accurate assessment of AOTH.
Although AOTH has not been evaluated critically (e.g., accuracy and quantitative description and documentation of the variation of pertinent physical characteristics), numerous sources (books, videos, magazines, etc.) provide information to “teach” users the technique. We found that participants often learned the AOTH method from multiple sources. Most users (82%) were self-taught to some degree, while many also used books (59%), magazines (42%), friends or relatives (33%), videos (30%), seminars (30%), and television programs (17%). From the aforementioned sources, users were taught to look for certain physical characteristics to help distinguish among year-classes, with certain angles of viewing providing more information than others. Participants ranked chest depth, body size, belly tightness and girth, chest girth, and neck thickness as most important followed by lankiness, hindquarter appearance, antler size, facial profile, and spinal alignment. The least important variables were presence or absence of eye squint, antler symmetry, and coat color. Users also ranked the side view as most important followed by front quartering, frontal, rear quartering, and rear views.
AOTH Accuracy Examination
We designed the accuracy examination to include a wide-range of ages, as well as multiple photos and angles for assessing physical characteristics of individual deer (Fig. 1). The assessment consisted of 583 photographs of 70 wild, known-aged, male white-tailed deer with antlers from south-central Oklahoma, ranging in age from 1.5 to 12.5 years. We photographed deer in January, but use the conventional aging scheme of half-year increments. Additionally, these photographs were taken after most of the breeding had been done on the study area (Webb et al. 2009b), so these accuracy assessments reflect the post-breeding phase, which Flinn (2010) found to have higher accuracy in aging deer than before or during the breeding season. To maintain anonymity of individual deer in the sample, we obscured ear-tag numbers using Adobe Photoshop Elements 7.0 (Adobe Systems, Inc., San Jose, CA). In an effort to give participants as much information as possible upon which to base their estimates, a series of 2–9 ( = 8.3) photographs of each deer were provided from various angles (side, frontal, and quartering views; Fig. 2). The initial screenshot shown to viewers consisted of all available pictures of an individual deer. The participant had the capability to select individual photographs for closer scrutiny (i.e., enlargement). After viewing all photographs of an individual deer, participants were required to select an estimated age from a drop-down box with all possible ages (1.5–12.5 yr), designed to reduce data-entry errors. For analysis, deer ≥7.5 years old were grouped together to increase sample size. After selecting an age, and moving to the next photograph, participants could no longer change their estimate. However, participants could exit the examination at any time and log back in at the point of departure. For presentation of results, we report descriptive statistics (means, percentages); unless otherwise noted, we report percentages to the nearest whole number.
Twenty-eight percent, 58%, and 14% of the professional users (n = 106) had B.S., M.S., and Ph.D. degrees in a wildlife-related major, respectively. Forty-three percent worked for a federal or state agency, while 31% and 25% were employed in the private sector or by a university, respectively. Greater than half of respondents were employed in the field for ≥5 years (24% for 5–9.9 yr; 31% for ≥10 yr), while the remaining 6% and 40% were employed for 0–1.9 years and 2.0–4.9 years, respectively. Most participants (57%) used AOTH to estimate ages of ≥25 deer annually. Of the remaining respondents, 6%, 16%, and 22% annually used the technique to estimate ages on 1–5, 6–10, and 11–25 deer, respectively.
AOTH Accuracy Examination
For the 106 professional users, the average overall accuracy was 36%. User scores ranged from 16% to 56%. Average accuracy rates tended to decline with increasing age (Table 1). For example, average accuracy was greatest for the 1.5 year-class (62%) and lowest for the 6.5 year-class (15%; Table 1). Average accuracy for the ≥7.5 year-class does not represent a single year-class; therefore, accuracy of 31% is misleading when comparing with single year-classes because this group contained several year classes (i.e., 7.5–12.5 yr). The number and magnitude of incorrect age assignments varied greatly (Table 1). For year-classes 2.5–4.5, the greatest percentage of age estimates tended to overestimate age by 1–3 years. For example, 57% of the deer in the 3.5 year-class were estimated to be ≥4.5 years old. For year-classes ≥5.5, the greatest percentage of age estimates tended to underestimate age by 1–4 years. For instance, 67% of the estimates for 6.5 year-olds were underestimated, placing them into ≤5.5 year-old age classes. For all year-classes ≥2.5, estimates spanned 6–7 years, including the year of correct year-class assignment (Table 1).
The accuracy levels presented in each column of Table 1 represent the error within actual year-class. For instance, 3.5-year-old deer were assigned to all other estimated year classes (Table 1). Often overlooked, and of equal importance, is the error within estimated year-class, which can be determined by reading across the rows. For example, 25% of the 3.5 year-class was aged correctly; however, 3%, 32%, 24%,18%, 19%, and 16% of the 1.5, 2.5, 4.5, 5.5, 6.5, and ≥7.5 year-classes, respectively, were incorrectly assigned to the 3.5 year-class (Table 1). This one example, using 3.5-year-old deer shows how the variation in physical development can lead to inaccuracy of AOTH. Similar to the 3.5-year-olds being classified to all other estimated ages, deer 4.5 to ≥7.5 years were incorrectly assigned to all other ages except the yearling age class (Table 1).
To our knowledge, this was the first attempt to assess the accuracy of using physical characteristics of deer to visually estimate age using AOTH techniques. We sought to optimize accuracy and maximize participant knowledge about AOTH by targeting and analyzing the responses of participants that are wildlife professionals and users of the technique, thus reducing the potential negative bias associated with novice participants. Over 90% of the participating wildlife professionals had used AOTH for >2 years, and >50% had used the technique for ≥5 years. Therefore, the levels of accuracy obtained by the professional users may offer a best-case scenario on the levels of accuracy achievable using current unstandardized AOTH techniques. Despite our efforts to maximize accuracy by targeting knowledgeable wildlife professionals, accuracy levels reported herein may be much lower than what can be achieved if a standardized protocol is developed and subjected to scientific scrutiny. In addition to assessing only experienced users, we further maximized inference of the technique by using deer that were of known age—the only real way to assess accuracy. Standardization of AOTH techniques also would require the use of known-aged deer to determine year-class–specific physical characteristics. Most (80%) users of AOTH were self-taught with books and magazines (i.e., those sources that make up the “popular literature” and that do not undergo rigorous peer-review or testing); therefore, development of a more standardized protocol for using AOTH may be presented best through scientific testing and publication.
Currently, there are no clearly defined criteria for AOTH, although Hellickson et al. (2008) measured several quantitative traits (e.g., antler and body traits) to assess the predictive ability of the measured traits at estimating age; their sample, however, did not use known-aged deer. Additionally, these techniques are more applicable to estimating age from harvested deer because more precise estimates of trait size can result in accurate statistical separation of age-classes. The authors acknowledged that providing this information could lead to criteria for AOTH, but that testing in the field was needed.
Through our pre-assessment survey, we also identified several consistencies with respect to using AOTH techniques. Viewing deer from the side, or front quarter, was considered important for most participants. The consistent rankings of certain views and body characteristics indicate that participants had some common sources of information regarding AOTH. One interesting finding is that antler size was not deemed important for AOTH, although Hellickson et al. (2008) found that predictive models containing antler trait measurements had the greatest predictive power for estimating age. In a more recent study, antler size was one of the most important physical traits for developing a morphometric computer-based model for estimating age of male deer using photographs (Flinn 2010); the application of computer software and quantitative assessment of morphometric traits may prove to be the most useful tool for AOTH and teaching users a defined set of traits.
Despite familiarity with the technique, participants' overall accuracy was less than expected by their own standards ( = 36%). Participant expectations of AOTH, relative to accuracy of the technique, generally were not supported. Most users in this study (85%) thought that accuracy would exceed 80% for the yearling age-class, and although accuracy expectations declined as age-class increased, 47% of the participants believed accuracy of AOTH would exceed 50% for the older age classification (i.e., ≥6.5 yr). Most participants overestimated the accuracy of AOTH for all specific year-classes. There was significant disparity between expectations and reality for 1.5–3.5 year-classes, which often are the year-classes of interest in selective harvest programs (Demarais et al. 2005). Several types of deer management are being practiced that use selective harvest as a management tool, but if there is inaccuracy of aging deer, then users of AOTH techniques may need to consider the level of accuracy acceptable for each type of management or research program. For example, most professional users believed ≥70% accuracy is required for making management-related selective harvest decisions and ≥80% accuracy is required for research purposes. The average overall accuracy of 36% in our study did not meet either of those requirements. Furthermore, the best-case scenario of 62% accuracy for the 1.5 year-class failed to meet those requirements as well. These findings reveal an overall view that accurate aging is important, but AOTH techniques fall short of expectations.
One generality we observed was a decrease in the level of accuracy as age increased (Table 1). There are 2 error-related scenarios that may occur when using AOTH: 1) overestimating age, or 2) underestimating age. Which source of error is most serious depends on the use of information and the level of accuracy and precision needed for its application. Here we highlight ways a management program may be influenced by these sources of error. First, there was a tendency to overestimate age from 1.5 to 4.5 years of age, especially for 2.5- and 3.5-year-old deer. Deer ≥5.5 years of age tended to have ages underestimated. Last, incorrect estimates of age spanned several year classes for each age (i.e., an over- or underestimation by 1 or 2 yr, indicated as ±1 or ±2 yr error) and became more severe as deer were aged further from their true age. For example, inaccuracy of ±2 years means that a deer that is 4.5 years of age could be misidentified as a deer anywhere from 2.5 to 6.5 years of age; that 5-year span is often greater than the life expectancy of the deer.
Relative to underestimating age, management programs that use selective harvest may mistakenly protect a deer with less than average antler characteristics (typically termed an “inferior” deer), when in fact, the management program targets “inferior” deer for selective removal (Webb et al. 2012). Overestimating age can remove a deer with greater phenotypic potential (i.e., antler size) at an early age, which can lead to high-grading or reduced cohort-specific antler size at maturity (Strickland et al. 2001, Demarais et al. 2005). Based on our findings, a management strategy seeking to protect 1.5-year-old males from harvest would put 38% of the 1.5-year-olds at risk of being harvested (i.e., through overestimation of age). Similarly, a management program seeking to protect males ≤2.5 years old from harvest would be placing 50% of the 2.5-year-olds and 4% of the 1.5-year-olds at risk of harvest because age was overestimated. In such instances, protecting an additional estimated year-class would greatly reduce the percentage of the protected year-class(es) placed at risk for harvest.
Accuracy levels of the AOTH technique in our study never approached those considered necessary for management and research; emphasizing the point that managers and researchers should be cognizant of the inaccuracies associated with this technique. It is unknown whether accuracy can be improved until a standardized protocol is developed through scientific testing, and users of the AOTH technique are trained accordingly. Development of explicit quantitative criteria for use in a qualitative setting (i.e., AOTH), and site-specific validation using known-aged animals, will be requisite for testing accuracy of the technique in hopes of maximizing success. Managers should use caution when making selective harvest decisions based on AOTH because the inaccuracies of the technique, and resultant overlap of physical characteristics in year-classes (cf Fig. 3), could make selective harvest by year-class less effective. Therefore, protection of an additional age-class may reduce unwanted harvest of younger male deer.
The Samuel Roberts Noble Foundation provided funding for this research. We thank all volunteers during deer captures; respondents participating in the accuracy assessment; J. Beard, M. Howry, and K. Staggs for technical assistance; and R. Stevens, C. N. Jacques, and several anonymous reviewers for improving early drafts of this manuscript.