Interobserver error involved in independent attempts to measure cusp base areas of Pan M¹s

Abstract

Cusp base areas measured from digitized images increase the amount of detailed quantitative information one can collect from post-canine crown morphology. Although this method is gaining wide usage for taxonomic analyses of extant and extinct hominoids, the techniques for digitizing images and taking measurements differ between researchers. The aim of this study was to investigate interobserver error in order to help assess the reliability of cusp base area measurement within extant and extinct hominoid taxa. Two of the authors measured individual cusp base areas and total cusp base area of 23 maxillary first molars (M 1 ) of Pan . From these, relative cusp base areas were calculated. No statistically significant interobserver differences were found for either absolute or relative cusp base areas. On average the hypocone and paracone showed the least interobserver error (< 1%) whereas the protocone and metacone showed the most (2.6 -4.5%). We suggest that the larger measurement error in the metacone /protocone is due primarily to either weakly defined fissure patterns and/or the presence of accessory occlusal features. Overall, levels of interobserver error are similar to those found for intraobserver error. The results of our study suggest that if certain prescribed standards are employed then cusp and crown base areas measured by different individuals can be pooled into a single database.

Geometric morphometric analysis of the crown morphology of the lower first premolar of hominins, with special attention to Pleistocene Homo

“…Recently, many dental studies based on image analysis of the occlusal morphology of fossil hominins (e.g., Bailey, 2004;Bailey and Lynch, 2005;Martinan-Torres et al, 2006;Gamez-Robles et al, 2007;Moggi-Cecchi and Boccone, 2007), non human extant primates (e.g., Bailey et al, 2004;Pilbrow, 2006;Hlusko et al, 2007), and recent modern human populations (e.g., Harris and Dinh, 2006;Perez et al, 2006;Bernal, 2007) have been published. Classical morphometric methods applied to image analyses (measurement of diameters and cusp areas) have demonstrated that African robust and non-robust groups differed in their P3 morphology (Wood and Uytterschaut,1987;Suwa et al, 1996).…”

How Reproducibly Can Human Ear Ossicles Be Measured? A Study of Inter‐Observer Error

“…More recently, computed tomography has offered the possibility of directly measuring three-dimensional objects (Hallgrímsson et al, 2008). The other technique is to take measurements on two-dimensional images after projection of the threedimensional objects (Bailey et al, 2004;Quam and Rak, 2008). This is the case when measurements are made on photographic prints or, as in the present study, on the screen of a digital microscope.…”

“…An evaluation of the reproducibility (precision) of measurements should therefore be standard in osteometric analyses to detect problematic measurements and to avoid interpretation based on flawed data affected by severe measurement error (Utermohle and Zegura, 1982). Knowing the contribution of measurement error to total variance is also important when data generated by different researchers are combined into a single data set (Bailey et al, 2004).…”

“…For example, measuring ear ossicles directly using calipers is a technical challenge because of the small size of the bones. On the other hand, measuring on images introduces additional sources of error (e.g., variation in specimen placement, parallax error) into the measurement procedure that are of no concern when direct measurements are taken (Bailey et al, 2004).…”

“…This was done because it was assumed that two measurements of an object taken by different observers tend to differ more than two measurements by the same observer, due to the contribution of betweenobserver bias to total inter-observer error (Healy, 1989;Ulijaszek and Kerr, 1999). It may be worth noting, however, that the assumption that inter-observer error always exceeds intra-observer error has not been substantiated in a number of anthropometric/odontometric studies (Ulijaszek and Kerr, 1999;Bailey et al, 2004). A further reason for using inter-observer error as an indicator of measurement imprecision was that, compared to intra-observer error, it more closely resembles the situation when data obtained by independent researchers applying the same measurement protocol are compared or combined into a larger data set.…”