Manual anchoring biases in slant estimation affect matches even for near surfaces

Psychon Bull Rev (2015) 22:1665–1670 DOI 10.3758/s13423-014-0770-7 BRIEF REPORT Manual anchoring biases in slant estimation affect matches even for near surfaces Dennis M. Shaffer & Eric McManama & Frank H. Durgin Published online: 8 September 2015 # Psychonomic Society, Inc. 2014 Abstract People verbally overestimate hill slant by ~15°– 25°, whereas manual estimates (e.g., palm board measures) are thought to be more accurate. The relative accuracy of palm boards has contributed to the widely cited theoretical claim that they tap into an accurate, but unconscious, motor representation of locomotor space. Recently, it was shown that a bias that stems from anchoring the hand at horizontal prior to the estimate can quantitatively account for the difference between manual and verbal estimates of hill slant. The present work extends this observation to manual estimates of nearsurface slant, to test whether the bias derives from manual or visual uncertainty. As with far surfaces, strong manual anchoring effects were obtained for a large range of near-surface slants, including 45°. Moreover, correlations between participants’ manual and verbal estimates further support the conclusion that both measures are based on the same visual representation. Keywords Geographical slant . Action measures . Anchoring . Two systems For 20 years, people’s estimates of slant have frequently been measured both verbally and manually (Bhalla & Proffitt, 1999; Bridgeman & Hoover, 2008; Durgin, Hajnal, Li, Tonge, & Stigliani, 2010; Durgin, Li, & Hajnal, 2010; Hajnal, AbdulMalak, & Durgin, 2011; Li & Durgin, 2011; Proffitt, Bhalla, Gossweiler, & Midgett, 1995; Shaffer, McManama, Swank, & D. M. Shaffer (*) : E. McManama Department of Psychology, Ohio State University–Mansfield, 1760 University Drive, Mansfield, OH 44906, USA e-mail: F. H. Durgin Department of Psychology, Swarthmore College, Swarthmore, PA, USA Durgin, 2013; Stigliani, Li, & Durgin, 2013). Verbal estimates of hill slant have typically been quite exaggerated and are almost always much higher than estimates made by manual matching. It has sometimes been argued that manual measures tap into a more accurate motor representation (e.g., Proffitt et al., 1995) or are simply quite accurate (Feresin & Agostini, 2007; Taylor-Covill & Eves, 2013). An alternative view is that the standard procedures used for manual measures have inadvertently been selected because they produce the theoretically desired accuracy (Durgin, Hajnal, et al., 2010; Durgin & Li, 2011; Shaffer, McManama, Swank, Williams, & Durgin, 2014). For example, egocentric biases in the haptic perception of orientation (Coleman & Durgin, 2014; Kappers, 2004) guarantee that palm boards set at waist level will produce lower estimates than those set higher (e.g., shoulder height). The standard procedure calls for setting the palm board at waist level. Moreover, we have recently shown that when palm boards are adjusted from horizontal, they give much lower hill matches (by 15° to 30°) than when they are adjusted starting from vertical (Shaffer et al., 2014). Again, the standard procedure used in essentially every article on perceived slant has been to have participants start manual adjustment from horizontal. In the present study, we further investigated this anchoring effect. Anchoring effects, including those found both with palm board adjustments and with free-hand matching, are expected under conditions of uncertainty (Tversky & Kahneman, 1974). That is, biases like anchoring are not expected when an exact answer can be produced with certainty. For example, if asked “What is half of 90?,” the answer “45” is not likely to be affected by first mentioning “0.” When one is asked to match one’s hand orientation to the slant of a visible surface, there are two possible sources of uncertainty (or variance): (possibly unconscious) perceptual uncertainty about the slant of the surface to be matched, and (possibly unconscious) uncertainty about the orientation of one’s own hand. Both of 1666 these forms of perceptual uncertainty can be thought of as the basis for making matching tasks susceptible to anchoring. This dual source of variance in perceptual-matching tasks raises the question of whether, in the act of manually matching the orientation of visually perceived hills, the primary source of uncertainty is manual or visual. In the present investigation, we tested for manual anchoring effects when matching near surfaces, because less visual error variance is expected in near space, whereas proprioceptive error variance should remain similar. It has been shown that near surfaces appear to be less exaggerated in slant than do farther surfaces (Bridgeman & Hoover, 2008; Hecht, Shaffer, Keshavarz, & Flint, 2014; Li & Durgin, 2010). Li and Durgin (2010; Li et al., 2013) argued that this effect of viewing distance could be explained by increasing stereoscopic depth compression at farther distances, combined with the systematically exaggerated perceptual coding of slant (Durgin, Li, & Hajnal, 2010). An alternative view is that visual uncertainty is greater at far viewing distances, leading to greater bias. If the latter view were correct, and anchoring in manual matching tasks are due primarily to visual uncertainty, we might expect that manual anchoring effects would be greatly reduced for near surfaces. But if manual anchoring effects are due primarily to perceptual uncertainty in the haptic/proprioceptive system, then large anchoring effects (i.e., of about 20°) would be expected even for manual matches to near surfaces. Manual slant underestimation found for near surfaces (e.g., Durgin, Hajnal, et al., 2010) can be predicted by the shallower verbal estimates that are found for near than for far surfaces (Durgin, 2013; Li & Durgin, 2011). Distance-related changes in perceived slant have been established using both explicit estimates and shape constancy tasks (Li & Durgin, 2010). Moreover, studies that have examined correlations between manual and verbal estimates for a single hill have reported that these correlations (ranging from about .2 to .5) are relatively high, considering the different sources of measurement variance that each type of measure contributes (Shaffer et al., 2014; Stigliani et al., 2013). These observations suggest that anchoring effects on manual estimates concerning near-surface slant would likely continue to be quite large. This is of some importance to establish empirically, however, because it would help to clarify that manual estimates may be exceedingly noisy measures even in near space (Durgin 2013; Durgin, Hajnal, et al., 2010). This is of theoretical importance because palm board measures have often been used to report null effects as one part of a dissociation with verbal measures, whereas these null effects might simply be due to measurement noise. Psychon Bull Rev (2015) 22:1665–1670 manual matches (with either a free hand or a palm board), and the (...truncated)