Visual object affordances: Object orientation

Abstract

Five experiments systematically investigated whether orientation is a visual object property that affords action. The primary aim was to establish the existence of a pure physical affordance (PPA) of object orientation, independent of any semantic object–action associations or visually salient areas towards which visual attention might be biased. Taken together, the data from these experiments suggest that firstly PPAs of object orientation do exist, and secondly, the behavioural effects that reveal them are larger and more robust when the object appears to be graspable, and is oriented in depth (rather than just frontally) such that its leading edge appears to point outwards in space towards a particular hand of the viewer.

Introduction

Gibson (1979) famously argued that a perceiver’s sensorimotor capabilities highly constrain the kind of visual information that gets accessed. The affordance hypothesis is central to this; the visual system might directly detect visual information about the behavioural possibilities afforded to us by objects and surfaces. Decades later – in one form or another – the notion of affordances continues to attract special interest (e.g. Michaels, 2003) and inspire contemporary theories of perception.

This paper examines one potential source of affordance – an object’s orientation – and asks whether it is an object property that acts as a ‘pure physical affordance’. By pure physical affordance (PPA), we mean an affordance that is solely revealed by the physical structure or arrangement of the object. While the term affordance is convenient and much-used, it is not even clear whether PPAs exist at all. In fact, as we discuss later, visual stimuli used in previous investigations have confounded orientation with semantic sources of affordance and sources of attentional bias. Using carefully selected visual object stimuli, the experiments in this paper go some way towards establishing that PPAs of object orientation do in fact exist.

Several behavioural studies have reported facilitatory effects associated with an object’s orientation, thus providing general support for the affordance hypothesis. However, as we argue in this section, semantic and attentional factors prevent these studies from serving as evidence of PPAs.

Humphreys and Riddoch (2001) reported data from a patient (MP) with spatial neglect. MP’s object detection improved when the object was cued by a description of its associated action (e.g. “find the object you could drink from”) rather than its identity (e.g. “find the cup”). This improvement dissipated, however, when the object’s orientation was away from the viewer (i.e. the handle pointed away from the viewer) such that it was less likely to afford interaction. Such data led Humphreys and Riddoch to argue that there is a direct pragmatic route from vision to action in the brain, realised by action templates that are activated by object affordances.

Riddoch, Edwards, Humphreys, West, and Heafield (1998) reported a case of ‘utilisation’ behaviour whereby patient ES often failed to respond with the task-appropriate hand when reaching for a cup (the rule demanded reaching with the left hand to cups on the left, and with the right hand to cups on the right). ES reached instead with the hand that was afforded by the orientation of the cup’s handle (e.g. even though the cup was on the left, when the cup’s handle pointed to the right, ES would reach with the right hand). Related findings have been discussed by Riddoch et al., 2000, Humphreys and Riddoch, 2000.

Studies using normal participants have also demonstrated that an object’s orientation affords action. Tucker and Ellis (1998) reported a study in which the participants judged whether the household objects were upright or inverted and responded accordingly with a left or right hand key press. Spatial key presses were faster and more accurate when they corresponded with the task irrelevant orientation of the object (e.g. a saucepan with its handle pointing leftwards promoted better left hand response performance). Symes, Ellis, and Tucker (2005) have shown that this orientation-dependent spatial compatibility (OSC) effect is dissociable from other spatial compatibility effects such as the Simon effect (Simon, 1969). In addition, the OSC effect also extends to remembered objects (Derbyshire, Ellis, & Tucker, 2006). In contrast to Tucker and Ellis’ (1998) suggestion that the affordance of object orientation evokes a specific motor response bias (e.g. for the limb most suited to perform the afforded action), Phillips and Ward (2002) argued instead that affordances are coded in a more abstract fashion. In support of this abstract coding view, Phillips and Ward (2002) demonstrated OSC effects using crossed hands and even foot responses (in their study prime symbols that were superimposed over target objects specified which spatial response to make).

We suggest that these kinds of studies do not provide evidence of PPAs, since the stimuli used have tended to be common household objects (such as cups) that have obvious action connotations. We all know, for example, that the handle is the functional part of a cup that is supposed to be grasped. Creem and Proffitt (2001) suggested that picking an object up appropriately by its handle requires a necessary interaction between cognition and action. Such facilitatory effects may therefore owe as much to the object’s functional, semantically derived affordance as they do to any potentially available PPAs.

In cases where the orientation of an abstract object has been manipulated (rather than common household objects), there has also been a potential semantic confound. In the well-known visuomotor priming studies of Craighero and colleagues, for example (Craighero et al., 1998, Craighero et al., 1999), clockwise or counterclockwise grasps of real-world target bars were facilitated when they were compatible with the orientation of a prime object (an outline of a rectangle oriented ±45°). It is possible that the goal-set and nature of the task (i.e. grasping a real-world oriented bar) engendered the visually similar prime objects with some action-related meaning.

As Anderson, Yamagishi, and Karavia (2002) have pointed out, affordance studies have generally investigated visually asymmetrical objects. They argued that visual asymmetry is likely to induce an attentional bias, and it may be this attentional bias (whereby attention is shifted to a specific location on the object) that is responsible for the generation of motor signals. This explanation is well-served by theories of attention that suppose a link between motor programming and attentional control (see Schneider & Deubel, 2002, for a recent review). Schneider and Deubel’s (1995) visual attention model – VAM, for example, assumes that the allocation of visual attention leads to motor programming (e.g. of the oculomotor system).

In a series of reaction time experiments in which line drawings of objects and non-objects were judged to be clockwise or anticlockwise from their ‘normal’ orientation, Anderson et al. (2002) reported facilitation of responses that were spatially compatible with the object’s visually salient feature – even if this contradicted the object’s semantic affordance. It is possible, however, that these results arose because participants used the visually salient feature as a cue for judging orientation. In contrast, a notable feature of Tucker and Ellis’ (1998) and other OSC studies has been the task-irrelevancy of object orientation. OSC effects occur even though orientation is task-irrelevant. Nevertheless, the attention-directing hypothesis is a plausible alternative to the affordance hypothesis and it remains the case that the object stimuli used in affordance studies have tended to be asymmetrical, with visually salient areas that might bias attention.

It was therefore the main aim of the present study to test the hypothesis that an object’s orientation acts as a PPA. We tested this hypothesis by setting up OSC conditions using object stimuli that firstly did not have any semantic affordances (only novel, functionally neutral objects were used), and that secondly were controlled for any visually salient areas. If an object’s orientation activates components of a reach-to-grasp movement, this should be reflected in the speed and/or accuracy of spatial key presses (which to an extent mimic the initiation of a reach-to-grasp movement with a particular hand).

In addition, we systematically varied the visual complexity of the stimuli (see Fig. 1 for examples), in an attempt to vary the strength of our predicted OSC effects. Our prediction here was that the more realistic, three-dimensional and graspable an object appeared, the stronger its PPA would be.

The same basic method applied to each experiment: On each trial a visual stimulus appeared centre screen that was oriented ±45° from the perpendicular. The participants were instructed to respond as fast as possible with a left key press when the object’s surface pattern resembled a ‘wobbly’ wood grain, and a right key press when it resembled a ‘straight’ wood grain (object orientation was therefore completely task-irrelevant). The stimulus orientation and surface pattern were selected at random and with equal probability. As a secondary research question, we introduced different stimulus onset asynchronies (SOAs) in order to examine the time course of any OSC effects found. When the object’s surface pattern was ‘neutral’ (the SOA stimulus), the participants waited until this pattern changed (after 0, 800 or 1200 ms) into a wobbly or straight pattern (see Fig. 1D for an illustration of each pattern). There is some evidence that the size of OSC effects builds in a linear fashion with viewing time (e.g. Phillips & Ward, 2002). Specifically, spatial responses that are compatible with the orientation of the object get faster as object viewing time increases. We were interested in seeing whether this same pattern would occur using visual stimuli that had been controlled for semantic and attentional factors.

Referring back to the objects in Fig. 1, it may not be immediately obvious which orientation would be compatible with a particular response. Why should we expect that the ‘−45° left-down’ orientation of these objects is more compatible with one hand than the other for example? There are at least two reasons for supposing that −45° left-down oriented stimuli are compatible with left responses, and 45° right-down oriented stimuli are compatible with right responses. Firstly, the nearest end of a −45° left-down stimulus (the left end) is closest to the viewer’s left hand (and similarly, the nearest end of a 45° right-down rectangle is closest to the viewer’s right hand). This nearest end might activate the corresponding response side. Secondly, the ease of a would-be-grasp of a given oriented stimulus is not equal for the two hands. A left-hand grasp, for example, would be less biomechanically awkward when made to a −45° left-down object than to a 45° right-down object, and vice versa for the right hand (see Johnson, 2000). As such, objects with −45° left-down orientations should afford left-handed actions, and objects with 45° right-down orientations should afford right-handed actions.¹