Numerical-spatial representation affects spatial coding: binding errors across the numerical distance effect

Psychon Bull Rev Numerical-spatial representation affects spatial coding: binding errors across the numerical distance effect Isabel Arend 0 Sharon Naparstek 0 Avishai Henik 0 0 Different brain areas have been identified for carrying out the processing of different features , such as color, form, and motion (e.g., Livingstone & Hubel, 1988). More recently, specializations for complex stimuli including faces, scenes (Epstein, Harris, Stanley , & Kanwisher , 1999), and bodies (Downing, Jiang, Shuman , & Kanwisher, 2001) have also been identified. This distributed processing raises the ques- tion as to how different features are integrated Does numerical-spatial representation affect feature binding? Studies of visual attention show that poor spatial coding leads to illusory conjunctions (ICs). In numerical cognition, it has been shown that numbers and space are not totally dissociated. This association underlies the numerical distance effect (DE): faster responses as the distance between the compared digits becomes larger (2 7 vs. 2 4). We used the DE to test whether numerical-spatial representation is available to visual processes that rely on spatial coding, such as feature binding. Participants reported the larger of two colored numbers. Both numerical distance (distances 2 and 5) and number-space congruity (e.g., congruent pair, 1 3; incongruent pair, 3 1) were analyzed. Results showed a higher proportion of ICs for distance 2 than for distance 5, providing strong evidence that numericalspatial representation (1) entails a strong location code and (2) is available to visual processes that rely on location information. Numerical distance; Feature binding; Illusory conjunctions; Spatial coding - The influential feature integration theory (FIT; Treisman & Gelade, 1980) describes the role of spatial attention in feature binding. According to FIT, miscombinations of visual featuresillusory conjunctions (ICs)appear in the absence of focal attention (Treisman & Schmidt, 1982). A number of studies have examined feature conjunctions. Features that appear close together are more likely to be conjoined than more distant features (e.g., Cohen & Ivry, 1989). With long exposures (up to 1.5 s) and low attentional load, ICs can be obtained in the periphery (Prinzmetal, Henderson, & Ivry, 1995), but not at the fovea, where spatial coding is more precise (Treisman & Schmidt, 1982). Despite several modifications over the years, FIT has maintained the important role of spatial attention as the main mechanism for the integration of independent feature codes (Quinlan, 2003). Recently, on the basis of neuropsychology findings, a multistage model has been developed to account for dissociations between within-dimension and cross-dimension binding (Humphreys, Cinel, Wolfe, Olson, & Klempen, 2000). An example of binding within dimension is the one implying integrating vertical and horizontal segments to make either a T or an L. Cross-dimension binding is the one involving the integration of different feature dimensions, such as color and shape. The multistage account describes cross-dimension and within-dimension bindings as taking place along dorsal and ventral areas, respectively. Similar to FIT, this account still holds that binding is acquired through parietal mechanisms. That is, the parietal attention system stabilizes early conjunctions. There are several representations of space in the primate brain (Andersen, Snyder, Bradley, & Xing, 1997). Such complex representations are known to be used by the visual system for various purposes from object identification to goal-oriented actions. In humans, numerical and spatial information have been shown to be consistently associated. For example, numerical values are represented in a horizontal vector so that small numbers are represented on the left side and large numbers are represented on the right side of space. This numericalspatial association has been documented by a variety of effects, such as the SNARC (spatial numerical association response codes) effect (Dehaene, Bossini, & Giraux, 1993) and the attentional bias effect (Fischer, 2001; for a review, see Hubbard, Piazza, Pinel, & Dehaene, 2005). The existence of such a horizontal vector was first demonstrated by the numerical distance effect reported by Moyer and Landauer (1967). In their study, participants were asked to decide which one of two simultaneously presented digits was numerically larger. Results showed a decrease in reaction time (RT) as the numerical distance between the values increased. The distance effect has been shown to affect RTs even when the numerical value is task irrelevant (Henik & Tzelgov, 1982). The distance effect has been interpreted to reflect an analogical semantic representation of numbers. Numbers are described to be mapped across space in a way that resembles a mental number line. The effect of numerical distance on RTs (distance effect) reflects the distinctiveness between numbers as a function of spatial location across the mental number line. That is, two numbers that are located close together (e.g., 2 and 3) are more difficult to be differentiated than two numbers that are located far apart (e.g., 2 and 7) (see Dehaene & Changeux, 1993; Dehaene, Dupoux, & Mehler, 1990; Moyer & Landauer, 1967). Even though some researchers have suggested that the distance effect may also rely on physical similarity (Cohen, 2009), the sematic component has been consistently shown in number comparison tasks (Goldfarb, Henik, Rubinsten, BlochDavid, & Gertner, 2011). Does spatial coding, derived from long-term representation involving numbers, affect feature binding? If close and distant numbers are indeed cognitively represented in terms of location on a mental number line, we expect numbers that are close together (distance 2) to entail a less precise spatial code than numbers that are far apart (distance 5). Therefore, we expect more ICs for distance 2 than for distance 5. In other words, we should find numerical distance to affect feature binding in a similar way that retinotopic distance has been found to influence binding. Studies looking at the effect of distance on binding have consistently reported that stimuli presented in close proximity produce more binding errors than do stimuli presented farther apart (Ashby, Prinzmetal, Ivry, & Maddox, 1996; Cohen & Ivry, 1989; Prinzmetal, Ivry, Beck, & Shimizu, 2002). We also wanted to test whether numerical spatial arrangement such as left-to-right versus right-to-left would influence binding. That is, if there is a preference for left-to-right mapping (e.g., 13) as opposed to right-to-left mapping (e.g., 31), we should find fewer ICs for the former, since spatial coding would be again more precise, therefore reducing the rate of ICs. However, previous work has found conflicting results regarding the effects of number-line congruity on RTs (Gertner, Henik, & Cohen Kadosh, 2009; Sagiv, Simner, Collins, Butt (...truncated)