Action-Specific Influences on Perception and Post-Perceptual Processes: Present Controversies and Future Directions

Abstract

The action-specific perception account holds that people perceive the environment in terms of their ability to act in it. In this view, for example, decreased ability to climb a hill due to fatigue makes the hill visually appear to be steeper. Though influential, this account has not been universally accepted, and in fact a heated controversy has emerged. The opposing view holds that action capability has little or no influence on perception. Heretofore, the debate has been quite polarized, with efforts largely being focused on supporting one view and dismantling the other. We argue here that polarized debate can impede scientific progress and that the search for similarities between two sides of a debate can sharpen the theoretical focus of both sides and illuminate important avenues for future research. In this paper, we present a synthetic review of this debate, drawing from the literatures of both approaches, to clarify both the surprising similarities and the core differences between them. We critically evaluate existing evidence, discuss possible mechanisms of action-specific effects, and make recommendations for future research. A primary focus of future work will involve not only the development of methods that guard against action-specific post-perceptual effects, but also development of concrete, well-constrained underlying mechanisms. The criteria for what constitutes acceptable control of post-perceptual effects and what constitutes an appropriately specific mechanism vary between approaches, and bridging this gap is a central challenge for future research.

According to the action-specific account of perception, people perceive spatial properties of the environment in terms of their ability to act in it (Proffitt, 2006; Witt, 2011a). For example, people wearing a heavy backpack judge objects as being farther away than people who are not wearing a backpack (Proffitt et al., 2003). According to this account, an object’s geometrical properties (for example, its size, shape, or distance) can be perceived differently depending on the perceiver’s ability to perform an intended action, even when the visual information about the object is exactly the same. Presumably, objects appear farther away because the heavy backpack increases the energetic demands involved in walking to them. Evidence for the action-specific account of perception comes from a variety of experimental paradigms. For example, the ability to reach to targets influences estimated distance to the targets (Kirsch, Herbort, Butz, & Kunde, 2012; Kirsch & Kunde, 2013a; Osiurak, Moragado, & Palluel-Germain, 2012; Witt, Proffitt, & Epstein, 2005). The energetic costs associated with ascending a hill influence the estimated slant of the hill (Bhalla & Proffitt, 1999). The size of the body, and its associated effects on a person’s ability to act, influences estimates of object size (Linkenauger, Leyrer, Buelthoff, & Mohler, 2013; van der Hoort, Guterstam, & Ehrsson, 2011). And performance, or probability of successfully performing an action, influence estimates of the target’s size and speed (Lee, Lee, Carello, & Turvey, 2012; Witt & Proffitt, 2005; Witt & Sugovic, 2010, 2012). These linkages between ability and perceptual judgments have been interpreted as evidence that a person’s ability to act influences the perception of geometrical properties of the environment.

The empirical findings and theoretical interpretations of the action-specific approach have given rise to a heated controversy. Much of the debate has centered on the extent to which past evidence of action-specific effects should be conceived as stemming from differences in how the spatial properties of things appear (that is, differences in the underlying perceptual representation) versus differences in output processing (that is, differences in how people select a behavioral response to communicate about their perception; Durgin et al., 2009; Firestone, 2013; Hutchison & Loomis, 2006; Woods, Philbeck & Danoff, 2009; Shaffer & Flint, 2011). When people don a heavy backpack, for example, do their distance judgments increase because the objects now visually appear to be farther away, or instead because people guess that the experimenters expect them to increase their judgments, and they comply by inflating their responses? This is a fundamental concern that cuts across many domains: if we ask people to describe some aspect of their experience (e.g., their sensory experiences, their cognitions, or their social perceptions), to what degree does their reply actually reflect the experience we want to know about? What factors influence how people communicate about their experiences? What methods may be used to mitigate or control output-related processes to obtain more “pure” measures of the underlying representation of interest? Why does it matter to distinguish between underlying representations and output-related processes, and when might it not matter? These issues are not new to psychological research, but the recent action-specific perception debate has brought them under an intense level of scrutiny. This scrutiny stands to yield new perspectives on perennial issues that are common to so many domains in psychology.

The distinction between the underlying perception and post-perceptual output processes often goes unappreciated. For some, if a person happens to give an accurate verbal judgment of an object’s distance, this is a straightforward indication that perception itself is accurate, and no other psychological processes need be considered. From this perspective, debating the distinction between perceptual vs. output processing has little meaning. For perception researchers, however, discriminating between perceptual and output processing is vitally important, and thus it is critical at the outset to motivate the importance of this distinction. First and foremost, the ability to predict and modify future behavior rests crucially on the accuracy of one’s model of the psychological processes underlying the behavior. Distinguishing perceptual from output-level processes is important for diagnosing and treating neurological disorders, for example. Treatment would proceed quite differently if a patient’s deficit in estimating distances stemmed from difficulty in manipulating numbers (to take one possible output-related process) rather than impaired distance perception. Similarly, in a functional neuroimaging scan, the function of an activated brain region would be interpreted quite differently if a task primarily changes how the participant complies with implicit social demands (to take another output process) rather than changing the visual appearance of an object.

Interventions that seek to improve safety or performance by enhancing visual perception (e.g., in driving, aviation, military, or sports settings) might also be impacted, if the intervention only influencing how people respond in the specific social context of a laboratory experiment, rather than influencing real-world perception when there are no experimenters present. There is agreement among researchers from both sides of the debate that discriminating between perception and output processing is crucial, but they differ strongly in terms of how they feel about the impact of output processing on perceptual judgments: Researchers who have adopted the action-specific perception perspective tend to believe that output processing plays little or no role in explaining past evidence of action-specific effects, while other researchers tend to believe that output processing accounts for most or all of this past evidence. At a more fundamental level, the debate hinges on researchers’ willingness to consider any possible influence of the observer’s transient state of action capability on the perception of geometrical properties of the environment, such as object sizes, object distances, and so forth. We will label the alternative perspectives as “action-specific perception” and “action-resistant perception”, to highlight the degree to which each perspective considers perception to be influenced by action capability.

The literature is currently flooded by one-sided viewpoints championing one perspective over the other (e.g. Durgin et al., 2009; Firestone, 2013; Proffitt, 2006, 2009, 2013; Proffitt & Linkenauger, 2013; Witt, 2011a,b; Witt & Sugovic, 2012, 2013a,b). The thesis of this paper is that a definitive resolution of this debate must include a nuanced evaluation of the possibility that both positions have explanatory power. Indeed, these positions need not be considered mutually exclusive; the relative contribution of action-specific perception and post-perceptual influences might be dynamic and vary dramatically from situation to situation. This being the case, a satisfying account of the action-specific perception issue must characterize the factors that determine the relative contribution of perceptual effects to output-related effects across different situations, as well as the mechanisms underlying these factors.

One drawback of the current literature is that, although there is an extensive corpus of work investigating post-perceptual influences on behavioral responses, many of these studies were conducted long before the notion of action-specific perception was formulated. Thus, there are no consolidated reviews of this wide-ranging literature that are framed in the context of action-specific perception. Readers new to the debate (and even some intimately familiar with it) may find it difficult to appreciate why post-perceptual processing plays such a prominent role in the contentious scientific conversations surrounding these issues. Because the two perspectives differ so strongly in terms of the role they ascribe to post-perceptual processing, there is a pressing need for a synthetic review of this literature. One goal of this paper is to provide a summary of the work underlying post-perceptual processing that captures why these processes are so salient and compelling for some visual space perception researchers as explanations of action-specific effects. We will also propose an extension of the action-resistant perception view that for the first time explicitly characterizes how this approach might account for bona fide action-specific influences on perception (as opposed to characterizing how it might account for action-specific influences on behavioral responses strictly through output processing).

To lay the groundwork for this endeavor, we will first describe a framework that underlies both the action-resistant perception approach and the action-specific perception approach. We then review various types of output processes that might impact judgments of geometrical properties of the environment. From there, we review methods that might be used to discriminate experimentally between genuine perceptual effects and output processes. We then compare and contrast three classes of mechanisms that could generate genuine action-specific influences on perception. Next, we make recommendations for particularly fruitful and theoretically meaningful research directions for future study. This section identifies a set of “best practice” methodologies that show promise for bilateral acceptance. We conclude with thoughts about the relative strengths of the two approaches and a summary of the insights gained through our emphasis on the similarities between them.

A “Modal Model”

For our starting point, we begin with a characterization of the basic model of perceptual processing that both underlies many space perception approaches (for concrete examples, see Foley [1978; 1991], and Gogel, 1990, among a host of others; see Wagner, 2006 for a review) and was the original starting point for the action-specific approach. Many existing theories are designed to account for somewhat different phenomena and thus differ in their specific features. For our purposes, the differences between these theories are relatively unimportant, and as such we will emphasize the global aspects they share in common. The resulting “modal model” is shown in Figure 1. Although the model presented here is framed in terms of visually perceived distance, a similar organization could apply to other perceptual dimensions (e.g., geographical slant or object size) or other perceptual modalities (e.g., audition). While it is admittedly incomplete, the model is intended to characterize the mutually agreed-upon conceptual distinctions that are crucial for discussing the similarities and differences between approaches and their related controversies.

Figure 1.

A modal model of perception. This model captures many of the structural features that the action-resistant perception and action-specific perception approaches share in common. The model shows the relation between visual cues, non-visual factors, perception, and post-perceptual processing. Dashed lines indicate feedback connections. See text for details.

The major components of the model are as follows. All current theories of visual space perception, including the action-specific perception approach, acknowledge the crucial role of visual information for constructing visual perceptual representations. In our backpack example given earlier, there is no controversy surrounding the idea that visual cues are important for specifying the perceived distance of the target, regardless of whether or not the observer is wearing a backpack. In the case of visually-perceived distance, this role is supported by a vast literature (Cutting & Vishton, 1995; Sedgwick, 1986; DaSilva, 1985). Various forms of visual information specifying an object’s distance are available to the brain (e.g., absolute disparity and angular declination). Our perceptual experience of distance is based (at least in part) on these sources of information, or “cues”. One way to describe the linkage between the cues and perceptual experience is by characterizing the relative effectiveness of each cue in determining perceived distance. Often, the cues appear to be combined according to a Bayesian weighted averaging rule, with the effectiveness weights being determined by the cue reliability (Ernst & Banks, 2002; Knill, 2007; see also Massaro & Friedman, 1990). The most reliable cues get weighted most heavily in determining how an object is perceived (Kersten & Yuille, 2003; Landy, Maloney & Johnston, 1995). In principle, attention could play a role in determining the effectiveness weights of specific cues (Gogel & Tietz, 1977; Gogel & Sharkey, 1989). If a particular cue is unreliable or is simply not present, the observer’s perception is largely determined by a combination of the remaining cues. Variations of this framework can account for many aspects of perceived space and layout, including situations involving disruptions or gaps in the ground surface (Sinai, Ooi & He, 1998), well-lit versus darkened viewing contexts (Ooi & He, 2007) and visible contact with the ground versus no visible ground contact (Bian, Braunstein & Andersen, 2005).

In much of the research underlying the modal model, perceived distance is taken to be an internal conscious experience that can be explicitly reported. As an internal experience, perceived distance can only be measured via some kind of behavior. A variety of behaviors might be used, and each behavior is assumed to be subject to a transformation that maps the internal experience of perceived distance onto a behavioral output. This transformation could introduce bias in the behavioral response with respect to the perceived distance. We will refer to this transformation as post-perceptual or output-level processing. In terms of our earlier backpack example, it is possible that donning a backpack has no effect on the perceived target distance, but instead changes some aspect of processing that occurs “downstream” from perception—how people choose a number to describe the distance they perceive, for instance. This possibility has been a central point of contention in the debate, and we will discuss these issues in more detail shortly.

Finally, there is some provision in the modal model for the overt behavioral indications of distance to feed back to earlier stages of processing. Both perspectives agree that information stored from past perceptual experience can serve as a non-visual factor that shapes future perception (e.g., memories that underlie perceptual learning; Epstein, 1967; Kellman & Massey, 2013). Similarly, for action-based responses, moving the body can influence the optic flow field, thereby providing additional visual cues for perceiving and acting (Gibson, 1979). From the action-specific perspective, recurrent connections between action-related signals generated during or in anticipation of a behavioral response and non-visual inputs to perception are particularly crucial. The question remains as to the extent of these recurrent connections.

Figure 1 shows only one box for behavioral response, but in fact this model should generalize well to a variety of response types. That is, many explanations for differences that might arise between response types can be framed in terms of existing components of the model. For example, an observer might verbally estimate a target’s distance as 5 m, but walk 6 m if attempting to walk to the target without vision (using the so-called blindwalking response). This could occur because both response types respond to the same value of perceived distance, but differ in the kind of post-perceptual processing that they tend to engage (Philbeck & Loomis, 1997): observers might tend to stop short when walking if they are concerned about stepping on the target, whereas they may tend to be more influenced by their assumed skill at using numbers if giving a verbal report. Alternatively, systematic differences in distance judgments between response types might happen because there are differences in the perception that controls the response (via differences in visual cues, cue weightings, or non-visual factors).

Given the central importance afforded to action capability in the action-specific perception approach, it may seem surprising that verbal reports and visual matching tasks, rather than action-based measures, have heretofore been the primary response modes used when investigating action-specific perceptual effects (e.g. Bhalla & Proffitt, 1999; Proffitt et al., 2003; Witt et al., 2004). A primary reason for this is that manipulating action capability can alter the action response itself in ways that have nothing to do with the underlying perceptual representation. For example, if action capability were manipulated by asking participants to throw a pencil versus an anvil at a target, the pencil could be thrown farther, but it would be a mistake to attribute this result purely to differences in perceived target distance. In addition, some action-based responses, particularly those that involve on-line visual control of rapid, precise movements, are thought to be guided by visual information processed in a dorsal cortical pathway that is anatomically and functionally distinct from a ventral pathway that presumably subserves conscious visual perception (Milner & Goodale, 1995). The action-specific perception approach seeks to explain conscious visual perception, however, and accordingly has focused on responses that are thought to be controlled by the ventral visual stream. For these reasons, much of the action-specific perception work has used verbal reports and visual matching tasks—response modes that are nearly universal in perception work and are often thought to be sensitive to conscious perception (Da Silva, 1985). Nevertheless, developing new action-based measures is an important ongoing effort that stands to significantly broaden the scope of research questions that can be addressed in this domain (Witt & Sugovic, 2013b).

Output-Level Factors as Alternative Explanations

As we foreshadowed earlier, discriminating between action-specific influences on perception versus output-level processes is a central issue for both the action-specific and action-resistant approaches. Both acknowledge that output-level effects govern the calibration of overt behavioral responses with respect to the underlying perceived distance. A key difference between approaches, however, concerns their assumptions about the relative involvement of perception-level processes versus output-level processes as an explanation of action-specific effects. The action-specific account places special emphasis on the role of the observer’s task and abilities to perform the intended action for visual perception. Some have suggested that this role is so fundamental that perception itself may be implemented in the brain by engaging the motor control pathways that are likely to be required to act on a target (Witt & Proffitt, 2008; Witt, South, & Sugovic, 2014; Witt, Sugovic, & Taylor, 2012). Accordingly, in the action-specific perception account, a person’s ability to act directly influences perception itself, and output-level processing is thought to play a negligible role in explaining action-specific response patterns. The action-resistant perception approach, meanwhile, tends to place more emphasis on the possible impact of output-level factors on responses as the driving factor underlying action-specific effects.

For some, the vigor and tenacity with which adherents of the action-resistant perception approach pursue output-level explanations of action-specific effects can be mystifying. After all, evidence of action-specific effects comes from studies that use commonly-accepted methods for studying perception such as magnitude estimation (e.g., verbal reports and blind walking) and psychophysics. These have been interpreted as valid measures of perception in many studies. Given the use of these common methods, why is there such resistance to the idea that action-specific effects reflect genuine differences in perception? Going further, if one assumes that researchers from the action-resistant perception perspective accept these methods uncritically as measures of perception in other research contexts, but then do not accept them in action-specific perception contexts, this apparent double-standard could be seen as an unfounded bias on the part of the critics.

For researchers from the action-resistant perception perspective, however, acceptance of a behavioral response as an indicator of perception is not so uncritical as it might seem. In this view, the possible influence of post-perceptual biases is not something that can be tested and definitively ruled out for all possible contexts. This approach assumes that any behavioral indication of perception is potentially subject to post-perceptual biases—even common measures such as verbal reports and blind walking. It also assumes that the influence of output biases is sensitive to contextual factors: output biases could play a negligible role in one experimental context, but play a larger role in other contexts. Thus, the possibility of bias by output factors must be considered in each experimental context. These assumptions are supported by decades of research that have been devoted to characterizing the kinds of output-level factors that exist and delineating the circumstances under which they are manifest (e.g., Asch, 1955; Baird, 1963; Carlson, 1977; Da Silva, 1985; Durgin et al., 2009; Epstein, 1963; Gilinsky, 1955; Gogel, 1974; Gogel & Da Silva, 1987; Hastorf, 1950; Mershon, Kennedy & Falacara, 1977; Poulton, 1979; Rogers & Gogel, 1975; Stevens, 1957; Witt & Sugovic, 2013a; Woods et al., 2009). Here, we consider some of the most prominent forms of output-level factors that can account for apparent action-specific perceptual effects. The variety and extent of this evidence provides some motivation for why output factors are at the forefront of action-resistant perception researchers’ minds when evaluating data from action-specific perception experiments.

Demand Characteristics

Certain aspects of the experimental design or experimenter behavior could unintentionally communicate the hypothesis of the experiment (Orne, 1959; Weber & Cook, 1972). Such cues to the experimental hypothesis are called demand characteristics. Behavioral experiments involving humans are typically conducted within a social setting (involving experimenters interacting with participants); if participants believe they know the experimental hypotheses, they may feel compelled to produce responses that support those hypotheses, perhaps in order to fulfill an implied social contract (Durgin et al., 2009; Orne, 1962). In studies investigating action-specific perceptual effects, a particular concern is that when participants arrive at the experimental setting, they may hold preexisting beliefs about the linkage between action capabilities and their experiences when interacting with the world. If this is true, aspects of the experimental methodology might encourage participants to respond in a way that conforms to this belief, without there being any difference in the underlying perceptual variable under study. For example, based on their own experience, participants may believe that carrying something heavy should make one’s destination “seem” farther away. If someone holding this belief is asked to estimate object distances while carrying a heavy box, he or she may interpret the box as a cue that the responses should be inflated to conform with the hypothesized linkage between heavy objects and distances (Durgin et al., 2009; Proffitt, 2006; Woods et al., 2009).

Importantly, researchers operating within the action-specific perception perspective have been mindful of possible demand characteristics in their experiments and have taken steps to reduce task demand (for example, by using cover stories and distractor tasks) in even the earliest work in this domain (Bhalla & Proffitt, 1999; Proffitt et al., 1995), although not in every case (e.g. Proffitt et al., 2003; Witt et al., 2004). Even so, uncontrolled cues to the experimental hypotheses can be very difficult to completely eliminate, and thus some task demand might exist despite researchers’ best intentions. Indeed, experimenters’ motivation to search for such influences could be subtly influenced by whether or not these influences stand to support or disconfirm their hypotheses. In the context of the action-specific perception controversy, these thoughts highlight the need for researchers on both sides of the debate to scrutinize their own motivations so as to strike an appropriate balance in considering the possible role of demand characteristics in individual experiments.

Manipulations of action capability by backpack encumberment

One of the earliest paradigms used for studying action-specific perceptual effects is a valuable example of the issues involved in evaluating experimental demand. Bhalla and Proffitt (1999) noted that when participants donned a heavy backpack, their judgments of hill slope were systematically larger compared to judgments made by unencumbered participants. Having attempted to reduce the role of task demands by using a cover story, the authors interpreted this result as stemming from perceptual differences due to the observers’ capability to act on the hill. In this view, because climbing the hill while wearing the backpack would be more effortful, observers visually perceived the hill to be steeper than when they were unencumbered. This paradigm was designed to capture the personal experiences of the researchers, in which they perceived hills to be steeper when hiking with a heavy backpack. This real-world situation strikes the current authors as one in which bona fide action-specific perceptual effects are especially likely to be manifested, if they are manifested anywhere.

Several aspects of the experimental paradigm, however, make it depart from its real-world analog in ways that reasonably would be expected to increase its sensitivity to demand characteristics and reduce its sensitivity to action-specific perceptual effects. For example, unlike in real world situations, observers know they are participating in an experiment, and the experimenter asks them to put on the heavy backpack as part of the experiment. Frank Durgin et al. have argued that in fact demand characteristics are likely to play the primary explanatory role in this paradigm. In one study (Durgin et al., 2009), one group of participants was asked to wear a backpack, but a cover story was given that the reason was to carry special equipment. The intention of this cover story was to make participants think that wearing the backpack had no bearing on the experimental hypotheses. The researchers predicted that when participants judged the slope of a ramp while wearing a backpack, those who were not given the cover story would be more susceptible to task demands and therefore estimate the ramp to be steeper than those who were given the cover story. The results confirmed this prediction.

This study has been criticized on several grounds, namely its use of a short ramp, the potential impact of demand characteristics in the cover story encouraging observers to ignore possible perceptual effects, and differences between the two groups in the felt weight of the backpack (Proffitt, 2009). Subsequent work involving a real hill has replicated the lack of effect for encumberment when a cover story is provided to disguise the purpose of the backpack (Shaffer, McManama, Swank & Durgin, 2013). An additional complication is that cover story manipulations might themselves influence the perceiver’s intention to act—a factor that has been shown to be crucial for eliciting action-specific effects (Witt et al., 2004, 2005, 2010). For example, if a cover story makes the backpack seem so incidental that perceivers’ intention to act discounts the backpack completely, then neither approach would predict an effect. Indeed, those in the cover story condition of the first experiment (Durgin et al., 2009) reported that the backpack felt significantly lighter than did those in the no-cover story condition, suggesting that participants given the cover story may have been discounting the backpack. Perceived backpack weight was not assessed in a second, similar experiment (Shaffer et al., 2013).

Even if a suitable cover story could be constructed, there are reasons to question whether this paradigm is sufficiently sensitive to bona fide action-specific effects on perception. One issue is that participants are typically asked to stand on a flat field in front of a hill and have little exposure to walking around with the backpack. This gives participants virtually no time to gain experience with the relative energetic costs associated with climbing the hill with versus without the backpack. Participants are also typically not given an explicit goal of walking up the hill. The presence of action-specific effects on perception are thought to be strongly linked with one’s intention to act (Witt et al., 2004, 2005, 2010), and thus if participants do not intend to walk up the hill, one would not expect action-specific perceptual mechanisms to be engaged. Altering the paradigm to correct for these shortcomings runs the risk of increasing the possible involvement of task demands.

In sum, although some contention continues to surround this paradigm, there is clear cause for concern with regards to its susceptibility to demand characteristics. This does not conclusively show that experimental demand (rather than differences in perception) was responsible for past evidence of backpack effects; similarly, this susceptibility to demand in experimental settings does not necessarily rule out bona fide action-specific perceptual differences in the real-world hiking scenario that was the original inspiration of this research domain. Indeed, it is our opinion that if action-specific perceptual effects occur anywhere, they are especially likely to occur under this kind of scenario, in which there are large differences in energetic costs (i.e., due to wearing a heavy backpack and fatigue). It does illustrate, however, that it is particularly difficult in this paradigm to disentangle perceptual effects from post-perceptual influences. To date, the specific paradigms that have been used to assess these effects are not adequate for understanding how people perceive slopes when encumbered.

Compliance with task demand

The impact of demand characteristics in an experiment depends on factors that determine whether or not participants notice the demand, as well as on factors that determine their level of compliance (Asch, 1955; Durgin et al., 2012; Shaffer et al., 2013; Witt & Sugovic, 2013a). These factors likely vary from person to person as well as between specific experimental contexts. Although there is a rich literature on the social psychology of compliance more generally (e.g., Cialdini & Goldstein, 2004), the issue has only recently been taken up in the context of action-specific manipulations. In principle, any intentional behavior might be subject to demand effects. Durgin et al. have reported evidence of compliance with experimental demand in verbal, visual matching, and action-based judgments (e.g., matching one’s hand orientation to a given geographical slant; Durgin, Klein et al., 2012; Durgin, Hajnal et al., 2010; Durgin, 2013; Li & Durgin, 2009).

Presumably, when demand characteristics are driving the results under manipulations of action capability, participants (1) make an inference about the pattern of responses they “should” produce, and (2) modify their responses to match these expectations. With respect to the first assumption, it is important to consider why participants wearing a backpack, for example, would infer that a hill “should” seem steeper. Where would this inference come from other than past experience with hills looking or seeming steeper under loads? Indeed, many action-specific experiments are motivated by real-life experiences in athletes such as baseball players who claim the ball looked as big as a grapefruit after a successful hit, or tennis players who claim that the game seems to move in slow motion when they are “in the zone”.

An alternative view is that an inference that hills “should” seem steeper under loads could come from non-perceptual kinds of past experience. For example, if one is estimating the distance to one’s car when carrying heavy grocery bags, the car’s distance could be encoded and remembered in an abstract, propositional format (“the car is farther than I can comfortably carry these bags”), rather than as a genuine perception that the car visually appeared farther when carrying the bags. One could argue that the affordances provided by the distant car really do change when one is under a load; this being the case, inferences that the car “should” seem farther away under loads could be based on memories of how the affordances of distant objects change under loads. Of course, the action-specific perception account would argue that the bags could indeed make the car visually appear farther way. Nevertheless, the point here is that memory of abstract (though perhaps highly salient) associations could constitute an important source of demand characteristics; the inferences underlying these effects--i.e., inferences about the linkage between effort and geographical slant, egocentric distance, size, and so forth--are certainly based on past experience interacting with the world, but the experience people remember may not necessarily be a perceptual one. Similarly, anecdotal reports of past experiences in the context of athletics may not be founded on perception (as we have defined it here), even though they are expressed in perceptual terms. By contrast, proponents of the action-specific perception approach find such anecdotal reports quite compelling and feel that the simpler interpretation is the one that takes these reports at face value.

With respect to the second assumption mentioned above (that participants modify their responses to match their expectations about the experimental hypotheses), it is important to note that not all participants will modify their responses according to expectations (e.g., Durgin et al. 2009). In the classic Asch conformity experiments, for which there was heightened pressure to select an obviously incorrect answer, only 30% of participants adjusted their response accordingly (Asch, 1955). This tendency for only some participants to be compliant has important implications, because the action-resistant and action-specific accounts make unique predictions for how compliant and non-compliant individuals should respond under manipulations of action capability. According to the action-specific account that these manipulations result in genuine perceptual differences, the effects should be apparent not only in individuals who are generally willing to comply, but also in individuals who tend to resist complying. By contrast, if demand characteristics effects are the sole mechanism underlying apparent action-specific effects, these effects should only be apparent in compliant individuals.

In research leveraging this logic, Witt and Sugovic (2013b) examined the effect of blocking ability (defined in terms of the size of a visible paddle in a Pong-like computer game) on speed judgments. Instructions were designed to bias participants’ responses toward reporting faster (or slower) ball speeds. For example, some participants were instructed to be sure to correctly classify all fast balls as fast and to make no errors in classifying any fast balls as slow, whereas other participants were instructed to be sure to correctly classify all slow balls as slow. Participants then judged the speed of balls while attempting to hit them with paddles of various sizes. Afterwards, the participants were divided into “compliant” and “non-compliant” groups according to whether or not the overall pattern of their speed judgments was shifted in the direction suggested by the instructions. In particular, those who were instructed to be sure to classify all fast balls as fast were considered to be compliant if their mean classifications were faster than the group mean, and non-compliant if their classifications were slower (and vice versa for the other group). The results showed that almost all participants reported faster ball speeds when using a smaller paddle, regardless of their measured level of compliance.

Importantly, this pattern held true even for the “non-compliant” group, who demonstrated an unwillingness to comply with the task demands given in the instructions. If they were unwilling to comply with the demand in the instructions, it is unclear why they would feel any more compelled to comply with a task demand relating to a specific linkage between paddle size and ball speed. It is possible that these participants were navigating within a complex hierarchy of compliance rules, such that they felt compelled to comply with some task demands but were unwilling to comply with others. However, a much simpler and more plausible explanation is that the observed linkage between paddle size and reported ball speed was rooted in genuine perceptual effects and had nothing to do with task demands. This presents strong evidence that experimental demand is unlikely to account for this particular action-specific effect. This technique of determining individual compliancy and examining the relationship between compliancy and action-specific effects is one way to examine, and potentially rule out, explanations that focus on demand characteristics.

Evaluation of past paradigms

Some paradigms are more susceptible to task demands than others. Paradigms in which the key experimental manipulation is obvious (such as putting on a heavy backpack or throwing a heavy ball) are particularly problematic. Ultimately it may be impossible to modify such paradigms to rule out an influence of task demands, and as such, we question whether these paradigms can adequately assess any potential perceptual effect. The paradigms that seem to mitigate task demands best are ones for which experimental manipulations are undetectable by participants. One potential example is the use of sugary drinks versus drinks with artificial sweetener. In one such study, participants were not able to differentiate the drinks based on taste, but the group that drank the sugary drink (which presumably increased their available energy by way of increasing their caloric intake) estimated hills to be less steep than did the group that received no calories (Schnall et al., 2010). However, it is not always possible to conceal the manipulations fully and this can have complex effects on the data. For example, Durgin et al. (2012) have reported that even an artificial sweetener manipulation can be detectable for a sizeable subset of participants, and that participants who had fasted for 3 hours and therefore had lowered blood sugar were more susceptible to experimental demand (i.e., judged a hill as steeper) than participants who had not fasted. Shaffer et al. (2013) argue that participants with higher blood sugar may be more capable of resisting experimental demand. One can imagine improving this paradigm via a combination of food science and psychophysics to develop a truly undetectable caloric intake manipulation, but if blood sugar level is indeed linked with resistance to demand, this would call into question this paradigm as a means of mitigating task demands. As such, this link deserves further exploration. For example, one might predict that if it takes energy to resist demand characteristics, a link between blood sugar and compliance should be observed in other experimental paradigms that have been used to study compliance (for example, the foot-in-the-door procedure or the low-ball procedure; Burger, 1999; Cialdini, Cacioppo, Bassett & Miller, 1978), rather than just action-based settings. If such a link fails to generalize, this paradigm could establish a gold standard for mitigating task demands.

Another approach for concealing the importance of capability to act from participants is to leverage individual differences in their pre-existing baseline level of action capability and analyze their data according to their capability. In this way, participants are not exposed to an obvious action-based manipulation such as donning a heavy backpack; they simply make perceptual judgments. For example, in one study, participants judged the steepness of a staircase, and then were allowed to choose a snack or beverage as compensation (Taylor-Covill & Eves, 2014). The researchers assessed the participants’ action capability indirectly by examining the sugar content of their chosen snack or beverage. The idea was that those who chose a more sugary option were more likely to have depleted energetic resources compared with those who chose less sugary options, a notion supported by several lines of evidence (e.g., Mehta et al., 2012; Page et al., 2011; Piech, Pastorino & Zald, 2010; Wang, Novemsky, Dhar & Baumeister, 2010). It is unlikely in this paradigm that participants would be aware of any task demand encouraging them to change their hill slope judgments according to their energetic resources or to alter their food choice according to their slope judgments.

Paradigms that leverage individual differences are only effective at concealing the hypothesized importance of action capability if participants are not aware that they are being recruited for their level of action capability. Participants recruited at, for example, a pain clinic, or an assisted living facility (Witt et al., 2009; Sugovic & Witt, 2013; respectively) might be likely to infer that their age or chronic pain status is relevant to the hypotheses of the experiment in some way. Another pitfall of the individual differences approach can arise if the participants’ action capability status is inferred on the basis of a variable that covaries with a variety of other factors. For example, several studies have examined the effects of age (Bhalla & Proffitt, 1999; Bian & Andersen, 2013; Sugovic & Witt, 2013) on perceptual judgments, and it is possible that age-related differences in these judgments stems from declines in action capability. However, aging is associated with a host of perceptual and cognitive changes that could impact responses without action capability being involved at all. The success of the individual differences approach for assessing perceptual effects, then, rests crucially on recruiting in a way that does not telegraph the importance of action capability to participants, and on ensuring that participants are well-matched apart from their action capability status.

Response Attitudes

Importantly, not all response biases are the result of participants consciously misreporting their perception when responding so as to comply with an inferred experimental hypothesis. They may make naïve assumptions about what aspect of the experimental context is the focus of the study, and then adopt a “response attitude” (Epstein, 1963; Gilinsky, 1955; Rock, 1986) that gives that assumed aspect more weight as they generate their responses. As an example, consider the following. Researchers typically have precise definitions in mind for the concepts they study, but naïve experimental participants do not necessarily conceive of these concepts so precisely. The instruction to “tell me how far away that mountain is” may seem unambiguous, but from the perspective of a naïve participant, there are several possible interpretations (Carlson, 1977; Mack, 1978; Woods et al., 2009). It may refer to the physical distance of the mountain recalled from memory (e.g., “The map I saw this morning said the mountain is 50 km away”); it might be a distance estimate developed through explicit inferential reasoning when viewing other aspects of the scene (e.g., “I can see 10 streetlights between here and the mountain, and those are at least 2 km apart, so the mountain is at least 20 km away”). It might be the visually-perceived distance (which is sometimes much shorter than physical distances (e.g., the mountain may appear only 10 km away on a clear day). It could even encompass influences from more abstract factors (e.g., being in a hurry may make the mountain “seem” farther away in some sense). Conceivably, each of these connotations could result in a numerically different distance judgment, without there being any difference in the underlying perceived distance. Importantly, some response attitudes could alter judgments by way of focusing the observer’s attention on output-level effects; other response attitudes could have legitimate perceptual effects--for example, by way of eliciting changes in eye movements or attention to specific spatial features of the environment (see below).

Evaluation of past paradigms

One approach to controlling for differences in response attitude is to inform participants of several possible attitudes, and then instruct them to ignore all but one of these when responding. This approach has been used extensively in studies of perceived size and distance, albeit primarily in reduced-cue environments (for reviews, see Carlson, 1977; Da Silva, 1985). Woods et al. (2009) used this kind of instruction manipulation in a well-lit outdoor environment. All participants were informed about three possible factors that might influence distance judgments: (1) the visual appearance of an object’s distance, (2) the physical distance of the object, and (3) non-visual factors. Participants were instructed to respond on the basis of one of these factors (specified in advance by the experimenter) when judging the distance of a target object. Participants threw either a heavy ball or a lighter ball several times without vision toward a target (presumably engaging action-specific perception mechanisms; Witt et al., 2004), and then verbally estimated the target distance. According to the action-specific perception account, throwing the heavy ball should require more effort than throwing the lighter ball, and accordingly the target object should appear farther away. The apparent and physical distance instruction groups showed no differences in distance judgments across the ball weight manipulation. Only the group that was instructed to take “all non-visual factors into account” when responding showed the predicted difference. If the apparent and/or physical (objective) instructions resulted in judgments that were indicative of the apparent target distance, the results suggest that the ball weight manipulation did not influence perception and that the “non-visual factors” effects may have been due to post-perceptual processes. That is, participants in this condition may have taken some salient aspect of the experimental methodology (e.g., the heavy ball), formed a hypothesis about the effect of ball weight on distance judgments, and then produced responses consistent with that hypothesis. More generally, in the absence of explicit instructions specifying which attitude to adopt, subtle aspects of experimenter behavior could influence the response attitude participants assume they should use. Use of specific instructions provides one way to constrain the possible response attitudes that participants may adopt and could help distinguish between perceptual and post-perceptual processing.

This is currently an under-explored area within action-specific paradigms. Nearly all paradigms have asked participants to report on physical or objective properties of the environment (e.g. “What size is the object?”), but, with the exception of Woods et al. (2009), few have gone on to clarify more precisely what response attitude participants should adopt. Thus, in principle, many existing studies from the action-specific perception approach are potentially subject to biases due to response attitudes. One mitigating factor is that if participants are unable to intuit what the hypotheses are, there should not be any systematic differences associated with an action-specific manipulation, even if participants did adopt a response attitude that took non-visual factors into account. For example, if experimenters have been successful in completely concealing an action-specific manipulation (e.g., manipulation of blood sugar; Schnall et al., 2010; Taylor-Covill & Eves, 2014), one would not expect participants to adopt a response attitude that takes that factor into account. Thus, many of the same issues discussed in the Demand Characteristics section above are relevant for evaluating the effect of response attitudes on past paradigms.

One other mitigating factor is that one might expect some action-based measures to be less sensitive to biases stemming from poor control of response attitudes (though perhaps not completely immune). In particular, if a participant is asked to blindwalk to an object or catch a moving object (e.g., Stefanucci & Proffitt, 2009; Witt et al., 2010; Witt & Sugovic, 2013a), the task arguably is framed in a way that emphasizes responding according to the physical location rather than non-visual factors.

To some extent, manipulating action capability within subjects can provide some measure of control for output effects relating to response attitudes. The validity of this approach hinges on the assumption that individuals will tend to adopt the same response attitude across all levels of the action capability manipulation. In principle, this assumption could be verified directly in post-experiment questionnaires, although no experiment investigating action-specific effects has yet done this. A potential drawback of this approach, however, is that by exposing each participant to all levels of action capability, the design may increase a paradigm’s susceptibility to demand characteristics. Thus, the potential benefit of within-subjects designs for mitigating response attitude effects must be weighed against their potential drawbacks in terms of demand characteristics.

Misattribution Effects

Closely related to response attitude effects are misattribution effects. Here, despite instructions to report an object’s size, distance, or speed, participants could be reporting on some other aspect of the stimulus entirely. For instance, when asked about the size of a softball, players might instead report on how easy they think it might be to hit (Lee, Lee, Carello, & Turvey, 2012). According to Gibson (1979), the primary objects of perception are affordances. In this view, it stands to reason that even when asked about size or speed, a participant might (implicitly) report about affordances instead. Misattribution effects are most likely to be of concern in experiments for which an action is frequently performed or at least strongly implied, such as reaching for or grasping objects, trying to catch moving objects, and jumping over gaps.

Experimenter Effects

Even if the hypotheses of an experiment are not communicated to participants, systematic differences in how an experimenter treats the participants during testing could bias the results in a way that happens to coincide with the experimental predictions, thus resulting in apparent action-specific perceptual effects when there are none. For example, if throwing a heavy ball at a target makes one group perform less accurately than a group throwing a light ball, an experimenter may unconsciously provide more encouragement to the heavy ball group or emphasize different aspects of the methodology (Woods et al., 2009). This could result in systematic biases in the group responses, without there being any inferences or assumptions about the experimental hypotheses on the part of the participants. Thus, great care is required to ensure that all individuals and groups are treated identically. Woods et al. (2009) have argued that interpersonal consistency is best maintained when experimenters adopt a neutral affect. In this view, neutral affect guards against unintentional inconsistencies of the kind noted above, stemming from a more positive, supportive affect. Deviations from neutral affect may also be more difficult to characterize when describing the study’s methodology.

Like task demands, experimenter effects can be difficult for a researcher to detect in his or her own experiments. Even if the experimenter treats all participants the same, the participants might be influenced systematically by some personal characteristic that varies from experimenter to experimenter. Within a study, this kind of effect can be minimized by having the same experimenter test all participants; if there are multiple experimenters, no experimenter should test a disproportionate number of participants in each group. These issues are difficult to control across different laboratories, however, as characteristics of the experimenters are rarely reported.

Evaluation of past paradigms

Some paradigms control for experimenter effects more effectively than others. One technique for controlling for experimenter effects is to use double-blind methods, in which the experimenters who interact with participants are themselves unaware of the experimental hypotheses. The Woods, Philbeck and Danoff (2009) study, described in the Response Attitudes section above, is one of the few that has used this technique in the context of action-specific manipulations. Under double-blind conditions, when participants received instructions that emphasized either apparent distance or physical distance, there was no replication of the ball weight effect reported in Witt et al., (2004). It is not clear that experimenter effects were responsible for past evidence of ball weight effects, however, because in another study, Woods et al. attempted to replicate the methods of Witt et al. (2004) as closely as possible (including using experimenters who knew about the hypotheses and interacted with participants from both the heavy and light ball groups), but again there were no ball weight effects. Further research is required to determine the reason for the differing results between Woods et al. (2009) and Witt et al. (2004). Unreported differences in experimenter behavior may be responsible. Nevertheless, the Woods et al. study represents a particularly well-controlled attempt to study action-specific effects. Keeping experimenters blind to the accuracy of participant’s responses (both verbal and throwing responses, in this paradigm) would further guard against the possibility of experimenter effects across treatment groups.

Even if experimenters are not blind to the experimental hypotheses, this may not be a concern to the extent that (1) experimenter interactions with participants are minimized (e.g., by computerized testing) or (2) experimenter interactions with participants are highly stereotyped or follow a written script. Thus, paradigms in which there are extensive unscripted interactions with experimenters who know the hypotheses are potentially more open to experimenter effects. Many studies investigating action-specific effects fall into this category. Examples include the backpack and ball throwing studies (Bhalla & Proffitt, 1999; Witt et al. 2004), studies involving chronic pain or older adults (Bian & Anderson, 2013; Sugovic & Witt, 2013; Witt et al., 2009) and studies involving swimmers and parkour athletes (Witt et al., 2011; Taylor et al., 2011; respectively). By contrast, other studies from the action-specific perception perspective have used methods that would be expected to reduce the likelihood of experimenter effects. For example, in several experiments involving athletes, the experimenters did not know the athletes’ performance levels, thus mitigating potential experimenter effects (e.g. softball: Witt & Proffitt, 2005; golf: Witt et al., 2008). For other experiments, the interaction between experimenter and participant were minimal and restricted mainly to initial instructions. These include tennis and Pong studies (Witt & Sugovic, 2010, 2012), and reaching studies (Kirsch & Kunde, 2013a; Morgado et al., 2013; Witt et al., 2005; Witt, 2011b).

Determining the possible role of experimenter effects in specific studies is not straightforward. If a published study does not mention attempting to minimize experimenter effects, this does not necessarily mean that no efforts were taken in this regard. For many researchers, interacting with participants in a stereotyped way with neutral affect is considered to be a fundamental principle of good methodological practice, and even when these practices are followed during testing, they may not be explicitly mentioned in published work. Nevertheless, given that effective techniques for minimizing experimenter effects are available (e.g., double-blind designs), these techniques should be used to the extent possible in future work in this domain. In addition, any efforts to control experimenter – participant interactions should be explicitly reported.

Memory Effects

There are many examples of action-specific effects for which the target is visible during the response (e.g. Bhalla & Proffitt, 1999; Proffitt et al., 2003; Linkenauger et al., 2009, 2010; Linkenauger, Witt & Proffitt, 2011; Witt & Dorsch, 2009; Witt, Linkenauger, Bakdash, & Proffitt, 2008; van der Hoort et al., 2011, 2014; Witt et al., 2004, 2005; Witt & Sugovic, 2012, 2013b). However, when the target is not visible, processes related to memory are a possible factor in explaining any effect. Presumably, the vision-based perceptual experience of an object’s distance fades away almost instantaneously after vision of the object is occluded, but nevertheless, representations of the object’s location can endure after vision is occluded (e.g., the “spatial image”; Loomis & Philbeck, 2008). Behavioral indications of geometrical object properties are often collected after vision is occluded. In the case of blindfolded walking, for instance, observers view the target object and then attempt to indicate its location by walking to it while blindfolded (e.g., among many others: Philbeck & Loomis, 1997; Rieser, Ashmead, Talor, & Youngquist, 1990; Thomson, 1980; Witt et al., 2004; Wu, Ooi, & He, 2004). Because vision is not available during the walk, the walk must be based on a memory representation rather than directly on visual perception. Such responses could still be informative about perception if the remembered target location was the same as the perceived location (Loomis & Philbeck, 2008). In the transformation from perceptual to memorial representations, however, biases could emerge, such that the remembered location is no longer coincident with the perceived location. Conceivably, an observer’s memory for an object’s location might be influenced by expectations or inferences about where the object “should” be, and these influences may well depend on action capabilities (Cooper, Sterling, Bacon & Bridgeman, 2012). If behavioral responses are based upon the remembered location rather than the perceived location (because, for example, the judgment is made once the object is no longer in view), action-specific effects on memory could be mistaken for bona fide action-specific effects on perception.

As a concrete example, consider the case in which an observer reports on the distances to two equidistant locations, A and B. He might remember that Location A was more difficult to walk to than Location B because he carried an anvil to A and was unencumbered in his walk to B. If memory of the additional effort to walk influences memory of the walked distance, he thus might remember A as being farther away from an initial vantage point than B, even though he did not visually perceive it to be farther way. Put differently, the affordances provided by Locations A and B were different because of differences in the observer’s action capability, and memory of the location might be influenced by the difference in affordances rather than by differences in visually perceived distance.

Evaluation of past paradigms

In the context of judging the size of a hole after throwing a marble successfully or unsuccessfully into the hole, Cooper et al. (2012) found that size judgments did not depend on participants’ throwing success when the hole remained visible (perception condition), but did depend on throwing success when the hole was occluded (memory condition). This suggests that action capability has its effect on memory, rather than perception, in this paradigm. The generalizability of such memory effects to explaining other past action-specific effects remains unclear, however. Memory has not been differentiated from perception in a number of studies of action-specific effects, including those involving softball (Witt & Proffitt, 2005), swimming (Witt et al., 2011), and tennis (Witt & Sugovic, 2010). It is therefore possible that the findings in these studies are attributable to action-specific effects on memory rather than perception, but as yet there is no empirical basis for evaluating this possibility. Furthermore, no studies that have found an effect in perception (i.e. while the target is still visible) have addressed whether the effects get bigger in memory. Thus, the relative roles of perceptual and memory processes in these paradigms remains unknown.

Other Output Effects

All of the foregoing factors may be said to influence the extent to which a behavioral judgment accurately reflects the underlying percept. Other output-related effects have been described, however. Blindfolded walking judgments of egocentric distance, for example, have been found to be subject to response compression biases—a form of stimulus range biases (regression to the mean; e.g., Li, Sun et al., 2013). Estimates of geographical slant obtained by manipulating a palm board device have been shown to be subject to anchoring effects, in which judgments initiated from a vertical palm board position differ systematically from those initiated from a horizontal position (Shaffer et al., 2014). Visual matching tasks are also subject to anchoring effects, such that larger values of a match are often obtained when the adjustable comparison stimuli are initially positioned farther apart than when they are initially close together (e.g. Witt et al., 2004).

Furthermore, many experiments have collected spatial judgments using some variation of the magnitude estimation technique, and this method is known to be subject to a variety of other kinds of biases that presumably occur at the output level (e.g., stimulus range biases and stimulus spacing biases; Poulton, 1979). Much work has been aimed at validating magnitude estimation methods and characterizing the kinds of factors that influence them (e.g., Stevens, 1957). The resounding message of this work is that “avoiding all the biases requires exceedingly rigorous investigations” (Poulton, 1979). To complicate matters, these biases may be applied unconsciously by observers when responding. Even action-based responses are not immune to unconscious biases. Involuntary and unconscious motor activity can occur as a result of suggestion or expectations. This effect, known as ideomotor action, has been implicated via double-blinded studies as the driving force behind phenomena such as facilitated communication and the motion of Ouija board pointers and dowsing rods (e.g., Burgess et al., 1998; for a review, see Spitz, 1997).

Although output-related effects have been acknowledged in the action-specific perception literature, they have not been particularly concerning to proponents of the action-specific approach because it has often been assumed that the effect of many output biases would be consistent across all action-specific conditions. To the extent that this is true, differences between conditions could not be attributable to these kinds of biases (e.g. Proffitt et al., 2003; Witt et al., 2004). Thus, the key consideration is not so much whether a particular response is subject to output-related biases (because one could validly argue that this is true for all behavioral responses), but instead whether the response is subject to biases that operate differently depending on one’s action capability. This remains an active focus of investigation.

Discriminating between perceptual effects and output-level effects

This extensive literature tends to weigh heavily in the minds of researchers from the action-resistant perception perspective when interpreting any response as a possible indication of perception, even if that response has been interpreted as sensitive to perception in past work (e.g., verbal reports; Da Silva, 1985). From the action-resistant perception perspective, output processes take on a devious character akin to Arthur Conan Doyle’s Professor Moriarty—capable of outwitting even a brilliant and determined adversary such as Sherlock Holmes. At least some of the resistance to the notion of action-specific effects stems from the impression that researchers from the action-specific perception approach have not done enough to thwart Moriarty—i.e., have not done due diligence in considering output-level explanations of the observed effects. At worst, failing to aggressively pursue alternative explanations that have such extensive empirical support runs the risk of coming across as confirmation bias (a tendency to only seek out evidence that supports the theory). As a reaction to this line of criticism, proponents of the action-specific perception account have become increasingly mindful of post-perceptual factors when interpreting their results and have begun to apply a variety of techniques to determine the potential role of output-related processes in action-specific effects. We have reviewed some techniques for controlling or mitigating the effects of specific output-level factors, above. Here, we review other strategies that have been employed and discuss their relative strengths and weaknesses.

Generalizability

As we have discussed, a recurring criticism of the action-specific perception approach is that when action-specific effects have been found in past work, they are likely due to post-perceptual processes such as demand characteristics rather than because perception itself has been influenced by the participants’ action capabilities. One argument against this criticism makes reference to the large number of studies in which action-specific perceptual influences have been reported: in this view, if one’s action capability indeed plays a strong role in determining what one perceives, evidence of this role should be observable in a wide variety of situations.

Much research supports this generalizability prediction. Action-specific effects have been reported in athletes, community members, psychology students, adolescents, and special populations such as older adults and those with chronic pain (e.g. Bhalla & Proffitt, 1999; Bian & Andersen, 2013; Cañal-Bruland & van der Kamp, 2009; Sugovic & Witt, 2011, 2013; Taylor, Witt, & Sugovic, 2011; Witt et al., 2009; Witt, Schuck, & Taylor, 2011). Effects have also been reported in a variety of dimensions including estimates of size, distance, slant, height, shape, speed, and weight (e.g. Cañal-Bruland & van der Kamp, 2009; Doerrfeld, Sebanz, & Shiffrar, 2012; Gray, 2013; Linkenauger et al., 2013; Proffitt, Bhalla, Gossweiler, & Midgett, 1995; Proffitt, Stefanucci, Banton, & Epstein, 2003; Stefanucci & Proffitt, 2009; Witt, 2011b; Witt & Sugovic, 2010, 2012). Effects have been reported with a variety of manipulations, including ones in which the manipulation is obvious to participants, such as donning a backpack or wielding a tool (e.g. Osiurak et al., 2012; Witt et al., 2005), as well as with less obvious ones such as drinking juice with sugar versus artificial sweetener or the use of individual differences rather than direct experimental manipulations (e.g. Bhalla & Proffitt, 1999; Schnall, Zadra, & Proffitt, 2010; Taylor-Covill & Eves, 2014; but see also Shaffer et al., 2013).

Generalizability is an important prediction that should be satisfied if one’s action capability indeed plays a strong role in determining what one perceives. By the same token, this kind of evidence is not fully diagnostic with regard to the overall viability of the action-specific perception approach. If output-level biases happen to be very common and occur in a wide variety of situations, the existence of a large number of studies showing action-specific effects might just as likely be due to output-level effects as genuine perceptual influences. Thus, determining the overall likelihood of output-level biases across a wide variety of situations remains an important topic for future research.

Converging Operations

One technique for bolstering evidence that an experimental manipulation influences perception itself rather than post-perceptual processes involves comparing the pattern of responses across multiple response modes. This “converging operations” strategy has been used by researchers from both the action-resistant and action-specific perception perspectives. Importantly, many of these studies from the action-resistant perception perspective that have used this technique compared responses under situations in which the stimulus environments and visual cues are the same across response types, as is often the case in action-specific perception experiments (Foley, 1977; Philbeck & Loomis, 1997; Philbeck, Woods, Kontra, & Zdenkova, 2010; Gogel et al., 1985). Evidence for action-specific perceptual effects has come from experiments using verbal reports, visual matching tasks, and action-based tasks. For example, increased anticipated effort for walking has been linked both with increased verbal estimates of target distances and with increased walked distances when participants attempt to walk to previously-viewed targets without vision (Witt, Proffitt, & Epstein, 2004, 2010; but see also Corlett, Byblow & Taylor, 1990). Another example comes from a task requiring participants to use a computerized fish-catching simulation with large or small nets (Witt & Sugovic, 2013b); virtual fish that are easier to catch as a result of using a larger net lead to verbal reports that the fish is moving slower and also to delayed net release times, which serves as an action-based measure of apparent speed. A third example comes from throwing or pushing a beanbag to land at the target’s location (Linkenauger, Bülthoff, & Mohler, 2015). In these cases, multiple indicators of distance and speed reveal converging evidence.

Particularly strong converging evidence that an effect is perceptual can come from indirect measures. Here, the assumption is that if participants are asked to report on some other dimension than the one of interest, they will be less likely to infer what the experimental hypotheses are and consciously adjust their responses to produce the predicted pattern. For example, given the relationship between retinal size and distance, apparent size can be an indicator of distance (Epstein, Park & Casey, 1961; Sedgwick, 1986)—in such an experiment, participants would judge the size of an object, rather than its distance, and the size judgments would then be converted into distances using a formula such as Emmert’s Law (in which distance equals the object size divided by the tangent of the visual angle that the object subtends). Indirect measures of perceived distance (specifically, the observer’s perceived height above the ground surface) using judgments of apparent size have revealed effects of fear on perceived heights, such that perceived height, as measured indirectly via judgments of object sizes, were larger when participants were fearful of heights (Stefanucci & Proffitt, 2009). In addition, other factors such as apparent shape and apparent parallelism have revealed similar patterns of effects as verbal estimates and visual matching tasks and thus can be considered indicators of apparent distance (Witt, 2011b).

Confirmation of the action-specific perception predictions across a wide variety of contexts and measurement techniques has been taken as the most compelling evidence available that action-specific response patterns can be perceptual in origin. Going further, the similarity of response patterns across so many contexts and measurement techniques has been taken to suggest that a common process underlies this pattern across contexts (e.g. Witt, 2011b; Witt et al., 2010; following Foley, 1977; Philbeck & Loomis, 1997): that is, most or all past evidence of action-specific response patterns is due to bona fide action-specific differences in perception. Thus, from the action-specific perception perspective, this research suggests that post-perceptual processing plays a relatively small or even negligible role in explaining action-specific responses in most contexts.

For researchers operating within the action-resistant perception framework, generalizability and converging operations can be convincing methodologies for addressing the possibility of output-level biases, but even these methods can be subject to systematic output biases and should be interpreted with caution. In the case of indirect measures, one cannot automatically assume that participants draw no inferences about the experimental predictions simply because they are asked to report on size rather than height or distance—ultimately, this assumption might be valid, but it must be approached with the same level of scrutiny as other methods. In short, a nuanced consideration of possible output-level influences in each experimental context is warranted before interpreting action-specific effects as stemming from perception in that context.

Overall Evaluation of Past Paradigms

Earlier, we reviewed existing paradigms in terms of their ability to resist various forms of post-perceptual influence. Here we take a more comprehensive perspective to evaluate the overall likelihood that existing results reflect genuine action-specific influences on perception. In one sense, very few past studies provide effective controls for a broad variety of output-related factors simultaneously, and by this criterion, some researchers might feel that virtually no existing studies provide truly compelling evidence. Nevertheless, some studies provide more compelling evidence than others. Perhaps there is the most cause for doubt in studies that have shown sensitivity to specific kinds of output-related factors. For example, the backpack encumberment and ball throwing paradigms show sensitivity to demand characteristics under at least some circumstances (Durgin et al., 2009; Woods et al., 2009); the link between throwing success and size judgments has been shown to be sensitive to action-specific effects on memory (Cooper et al., 2012). The extent to which past evidence of action-specific effects using these paradigms is indeed due to output effects is unclear, but suspicion that output effects were responsible in the past is certainly heightened by evidence that the paradigm is sensitive to such effects.

There is broad agreement that studies are more compelling to the extent that they provide effective controls for specific kinds of output effects. Double-blind methodology can provide a means of controlling for experimenter effects (Taylor-Colvill & Eves, 2014; Woods et al., 2009). Paradigms involving undetectable action-specific manipulations would be a gold standard for minimizing demand characteristics. One candidate is the use of drinks containing sugar versus artificial sweetener as a manipulation of action capability (Schnall et al., 2010). There has been some contention surrounding this paradigm (Shaffer et al., 2013), but additional development may be able to make it more broadly compelling. Another paradigm that shows promise for minimizing demand characteristics is to collect perceptual judgments before assessing the participant’s action capability (Taylor-Covill & Eves, 2014). In this way, participants do not know that their action capability is relevant to the study at the time they are making their judgments, and experimenters do not know participants’ action capability during the judgments because it has not been measured yet. Thus, this paradigm can be effective for controlling both demand and experimenter effects.

Paradigms using individual differences, such as drink or food preference, can provide a useful way to conceal from participants the hypothesized importance of their action capability and thereby minimize demand characteristics. However, as we have argued, this approach is only effective if the recruitment process does not suggest to participants that their action capability (or some factor related to it) is relevant to the hypotheses. For example, several studies involving athletes collected perceptual judgments before assessing performance, but participants were aware that recruitment took place at softball fields and golf courses (Witt & Proffitt, 2005; Witt et al., 2008). Thus, this may have made the hypothesized importance of their athletic ability salient. As another example, recruiting younger and older adults at an assisted living facility (Sugovic & Witt, 2013) likely made age a salient factor, and recruiting patients and employees at a chronic pain clinic (Witt et al., 2009) likely made pain a salient factor. In contrast, recruiting adults at a public shopping center and measuring their weight and BMI after collecting all perceptual measures is an effective way to conceal an interest in body size (Sugovic & Witt, 2011).

A potential drawback is that attempts to make action capability manipulations undetectable by participants may also minimize the likelihood of finding genuine effects of action on perception. This could happen, for example, if the manipulation is made undetectable by using treatment conditions that contain only very small differences in energetic demands. For instance, if a backpack is worn for only a brief period of time and participants have no reason to anticipate ascending the hill, the conditions are not optimized to find an effect in perception, if one should exist. The presence of the backpack is clearly detectable, but under these conditions, the difference in energetic demands between the backpack and no-backpack conditions is likely quite small. This reduces the sensitivity of the experiment for detecting genuine action-specific perceptual effects and increases the possibility that output-related factors may underlie systematic differences in the results. For this reason, paradigms that involve very small differences in action capability tend to be less compelling as evidence of action-specific perceptual effects.

Another example of potentially insufficient variation in action capability is that of throwing a heavy ball (Witt et al., 2004). Throwing a 2lb (.91kg) ball to targets ranging from 4–10m does not require substantially more effort than throwing a .7lb (.32kg) ball. This could explain why this manipulation sometimes produces significant effects on estimated distance (Witt et al., 2004) and sometimes does not (Woods et al., 2009). A significant challenge for future research is to balance the need to minimize output-related factors, such as task demands, against the need to maximize the amount of variation in action capability; small variations in action capability are not likely to affect perception appreciably.

Yet another paradigm that we question is dart-throwing, at least when conducted with novice dart throwers. Although significant effects of action capability have been reported using this paradigm (Wesp et al., 2004; Cañal-Bruland et al., 2010), novice dart-throwers have poor control over the precision and accuracy of their performance. This would be expected to dramatically reduce the sensitivity of this paradigm to action-specific perceptual effects. Poor sensitivity also increases the likelihood that small variations in methodology across laboratories could lead to inconsistent results. More anecdotally, recent work by one of the current authors (Witt, unpublished data) asked participants to maneuver an airplane icon in a computer display through an aperture or to drop a cargo load from the airplane into an aperture on the ground and then estimate the size of the aperture. Action capability was manipulated by changing factors such as the speed and height of the airplane. None of the action capability manipulations influenced size judgments in this 2D setting. Given that action-specific effects have been found with respect to estimated size of virtual 3D objects (e.g. Linkenauger et al., 2013) and estimated speed of 2D objects (e.g. Witt & Sugovic, 2010, 2012), the reason why these 2D size estimation paradigms failed to elicit effects remains unclear. Although resolving this issue may yield important information about the conditions under which action-specific effects are and are not manifested, the paradigm itself does not appear to provide reliable evidence of these effects.

An example of a paradigm that has made substantial progress towards controlling multiple output-related factors is the so-called Pong task, discussed earlier, in which participants make judgments of ball speed on a computer monitor after they attempt to block the ball with different-sized paddles (Witt & Sugovic, 2010, 2012). Work using this paradigm has leveraged individual differences in compliance to assess the possible effect of experimental demand (Witt & Sugovic, 2013b). This work suggests that demand plays a relatively small role in judgments of ball speed in this paradigm. This paradigm is also effective for minimizing experimenter effects, in that participants perform the task without interacting with the experimenter, apart from the initial instruction phase. Furthermore, paddle size is typically manipulated within-subjects randomly on a trial-by-trial basis; on the plausible assumption that participants would be unlikely to rapidly switch between different response attitudes under these conditions, this paradigm thus also provides a measure of control over response attitudes. The manipulation of paddle size is certainly obvious to participants, so additional strategies to minimize other output-related factors are still warranted. To mitigate this concern, Witt and Sugovic (2013a) used an indirect, action-based measure (Witt & Sugovic, 2013a). Rather than continuously controlling the paddle, participants pressed a trigger to shoot the paddle up, and the challenge was to time the release of the paddle just right in order to catch the target (a “fish” in this case). The center of each paddle was positioned similarly for all 3 paddle sizes so that to maximize catching performance, the different-sized paddles should be released at the same time-point. Participants, however, tended to wait longer to release the large paddle than the small paddle. The experimenters took differences in the time to release the various-sized paddles as an indirect measure of differences in perceived ball speed; in this view, participants waited longer to release the big paddle compared to the small paddle because they saw the fish as moving slower and therefore had to wait longer. Although no paradigm is unassailable, Pong-like tasks such as these arguably come closest to providing evidence of action-related perceptual effects under conditions that control for multiple output-related processes.

Summary

Strategies for minimizing the influence of output-related processes in judgments of distance and other spatial aspects of the environment are summarized in Table 1. Although ultimately it may be impossible to definitively rule out all output-related processing as an explanation for action-specific effects (or any purported perceptual effect), steps may be taken to maximize the discriminability of perceptual and post-perceptual effects and make a convincing case that action-specific effects can be genuinely perceptual. We will outline some of these steps in the “Recommendations for Future Research” section.

Table 1.

Recommended strategies for evaluating or minimizing the role of post-perceptual processes.

Post-perceptual process	Methodology
Task Demand	Cover story for manipulation
Task Demand	Opaque manipulations
Task Demand	Between-subjects manipulation
Task Demand	Indirect or implicit measures
Task Demand	Naturally-occurring differences that are not measured until the end of the experiment
Task Demand	Survey to assess potential awareness of the hypothesis for each participant
Task Demand/Experimenter Effects	Measure of compliance as covariate
Experimenter Effects	Consistent (and neutral) affect/feedback from experimenter
Experimenter Effects	Consistent instructions from experimenter
Experimenter Effects	Experimenters blind to condition
Response Attitude	Within-subject manipulation
Response Attitude	Explicit instructions about which attitude to take
Memory	Assess reports while target is still visible
Misattribution	Clear instructions as to what to estimate

Mechanisms for Genuine Influences on Perception

We have focused thus far on post-perceptual effects that could be mistaken as differences in the underlying perceptual representation, as well as methods for minimizing the influence of such factors in experimentation. To date, these issues have taken center stage of the controversy surrounding the action-specific approach. We now turn to a more neglected area: the mechanisms that could drive bona fide differences in perception. Given that the debate that has dominated the literature has been about the potential role of output-related factors, the consequence is that much less attention has been paid to similarities and differences between the structural features of the action-resistant perception approach and action-specific perception approach. There has been relatively little emphasis on this issue from the action-resistant perception perspective, but nevertheless some possible mechanisms have been discussed, and to some degree accepted, by researchers from this perspective. We will next discuss three possible mechanisms by which one’s potential to act could influence perception.

(1) Modification of Visual Information or Visual Processing

One way that action capability could affect visual perception is by influencing the type or quality of the visual information that gets picked up by the observer. Importantly, this mechanism can be accommodated within the modal model. Even though the distal cue configuration may remain constant, an observer’s goals and abilities may affect where she looks and what she pays attention to (Castelhano, Mack & Henderson, 2009; Henderson, 2003; Rothkopf, Ballard & Hayhoe, 2007); variations in eye movements can accordingly impact task performance (e.g., putting performance in golf; Vickers, 1992). These eye movements and/or shifts in attention might then impact perception. Here, we review evidence suggesting that shifts in eye movements and attention can influence the appearance of geometrical properties of the environment.

Depth and egocentric distance

Perceived depth and distance can be influenced by eye position and eye movements, presumably by incorporating extra-retinal oculomotor signals (Collewijn & Erkelens, 1990; Foley & Richards, 1972; Gogel & Tietz, 1977; Wist & Summons, 1976). If one fixates one object but pays attention to another object at a different distance, this can influence the perceived distance of the fixated object, albeit minimally (Gogel & Tietz, 1977). In outdoor environments, observers localize objects less accurately in egocentric distance when they move their direction of gaze from a far distance inwards toward the object than if they scan from their feet outwards (Wu, Ooi & He, 2004). A similar dependence on scanning direction does not occur in indoor room-sized environments, perhaps because the ground surface plays a less important role when other planar surfaces (walls and a ceiling) are present (Gajewski, Wallin & Philbeck, 2014a). Wu, He and Ooi (2008) have shown that distance judgments are less accurate if observers’ initial attention (and presumably, their direction of gaze) is biased toward locations that lie beyond the peripersonal ground surface than if their initial attention is biased toward locations within this nearby region. This suggests that task-related factors that interfere with processing of visual information from the nearby ground surface can impact distance judgments. Other work has shown that completely occluding the nearby ground plane in outdoor environments has little effect on distance judgments relative to unrestricted viewing conditions (Gajewski et al., 2014b). The explanation for these apparently conflicting reports remains unresolved. The role of eye movements and gaze direction on perceived distance may play a more pronounced role in reduced-cue settings, contexts in which only a very brief glimpse of a novel environment is available, or in more cluttered environments (Gajewski et al., 2014a, b).

Size and speed

In the context of relatively impoverished 2D displays, attended objects have been found to appear larger than unattended objects (Anton-Erxleben, Henrich, & Treue, 2007), and fixated objects appear larger than those in the periphery (Newsome, 1972). Epstein and Broota (1986), furthermore, found that when attention was diverted by asking participants to make numerosity judgments of spots on a stimulus card rather than size judgments of the card, subsequent size judgments were biased toward the card’s projective size. Similar issues have been studied in the domains of obstacle avoidance and steering (e.g., Franchak & Adolph, 2010; Land & Hayhoe, 2001; Matthis & Fajen, 2014; Patla & Vickers, 2003), albeit not in connection with testing for effects of action capability. In an aviation context, novice pilots’ judgments of airport runway size in a flight simulator were correlated with their runway fixation time, as well as with several measures of landing performance (Gray, Navia, & Allsop, 2014). This correlational analysis cannot determine whether these differences in fixation were directly responsible for the differences in judged runway size, however. As such, the role of attention and/or eye position in determining perceived size is uncertain.

Cañal-Bruland, Zhu, van der Kamp, and Masters (2011) examined the role of attention by blocking the view of a golf hole, so as to reduce attention directed to the hole. Putting performance did not influence judged hole size when the hole was occluded, but did influence judged hole size when the hole was visible. The researchers also diverted attention away from the hole by requiring participants to putt the ball between two markers before reaching the hole. The intention was to force attention to the markers rather than the hole. Again, this manipulation eliminated the relationship between putting performance and perceived hole size. These results suggest that attention to the hole was key for eliciting a linkage between action capability and judged size.

Other research, however, does not support the idea that attention is a critical factor. In one set of experiments, a participant’s body was rendered in a virtual environment to be twice as big or half its size. Participants estimated the size of virtual objects placed nearby. The objects were judged to be bigger when placed next to the larger rendered body (van der Hoort et al., 2011). In a follow-up experiment (van der Hoort & Ehrsson, 2014), the researchers constrained participants to look at a stationary fixation point, thereby controlling any differences in eye movements. The effect of rendered body size on perceived object size was just as big when eye movements were constrained as when they were not, suggesting that eye movements and/or attention do not play a role in this particular action-specific effect.

Witt and Sugovic (submitted) found that eye movements do not play a role in explaining the effect of blocking ability on apparent ball speed. In this paradigm, blocking ability is manipulated by varying the size of a paddle in a 2D computerized display and participants make judgments of ball speed after attempting to hit the ball with the paddle (Witt & Sugovic, 2010, 2012, 2013a,b). Witt, Sugovic and Woodman controlled for eye movements by constraining participants to fixate the ball via a secondary task, and also restricted their analyses to trials in which eye tracking showed that participants indeed fixated the ball. The previously-reported influence of paddle size on apparent ball speed persisted even when eye movements were held constant across the paddle size manipulation.

Differences in eye movements/attention due to action capability

For purposes of evaluating this mechanism as a possible explanation of action-specific effects, a more fundamental question is whether there are indeed natural shifts in eye movements or attention associated with differences in action capability. Few studies have investigated this issue. At least in the case of egocentric distance estimation, Gajewski et al. (2014a) found no differences between observers’ natural patterns of eye movements in a distance judgment task as a function of whether a verbal or a blindwalking distance judgment was required. Furthermore, there was no apparent linkage between the natural pattern of eye movements and the accuracy of distance judgments; observers who adopted the strategy of steadily fixating the target performed equally well as those who looked around the room before responding. Additional research is required to determine the generalizability of these findings to other manipulations of intentions to act. Nevertheless, these null results suggest that systematic differences in eye movements and/or attention, if present, may be too small to play a robust role in mediating any bona fide action-related perceptual component in paradigms involving egocentric distance judgments (e.g., Proffitt et al., 2006; Witt et al., 2004).

Summary

Under some circumstances, task manipulations clearly do influence the deployment of eye movements and attention, and differences in eye movements and/or attention clearly can influence perceptual aspects of the environment such as distance and size. However, no research to date provides a satisfactory demonstration of a direct connection between action, attention, and perceived geometrical properties of the environment. Much work remains to be done to evaluate this mechanism, but at present it appears unlikely to provide a robust, across-the-board explanation of action-based effects. It may play a role in specific situations, however, and as such should be evaluated whenever possible.

(2) Modification of Non-visual Processes

Researchers from both the action-specific and action-resistant perception perspectives tend to conceive of perception as a phenomenal experience that excludes influences from conscious cognitive processing, such as explicit reasoning or mental arithmetic. Similarly, there is undeniably some degree of cognitive impenetrability (Fodor, 1983; Pylyshyn, 1999) in layout perception; for example, if a mountain visually appears to be 10 km away, knowing that it is physically 30 km away does not alter the perception that it appears much closer. Both perspectives also agree that the output-level processes discussed above should not be considered to influence perception. Nevertheless, both approaches allow for some kinds of “top-down” influences on perception.

From the action-resistant perception approach, there is an established literature on perceptual learning, in which past experience is thought to influence on-line perception (Fine & Jacobs, 2002). There is evidence that different visual cues to distance require different amounts of time to be extracted during the initial stages of an eye fixation (Gajewski et al., 2010). Given this, and the fact that the region of high acuity is limited to the central 1 deg, some degree of persistence or integration of information is required to maintain perceptual stability of spatial layout. Assumptions or expectations that an object is resting on the ground can influence the object’s perceived location (Wu, He & Ooi, 2014). Similarly, other research suggests that information gained during long glimpses of an environment (e.g., 5 sec) influences subsequent distance judgments when targets are glimpsed very briefly (e.g., 100 ms; Gajewski et al., 2010; Gajewski et al., 2014). Finally, the specific-distance tendency (Gogel, 1990; Gogel & Tietz, 1973; Owens & Leibowitz, 1976) and the intrinsic ground plane (Ooi, Wu & He, 2006) are thought to be internal biases that influence perceived distance to the extent that visual information about an object’s distance is unreliable (e.g., beyond the effective range of distance cues in well-lit natural environments), or unavailable (e.g., in darkened environments). Because these biases do not arise directly from visual stimulation, they might be considered yet another type of non-visual influence on visual perception. Taken together, there are several precedents for the idea that at least some kinds of non-visual factors can influence perception, although the full scope of these influences remains poorly understood.

The action-specific perception approach claims that non-visual action-related factors play a crucial role in perception. However, these processes are thought to be very different than the specific distance tendency and intrinsic ground plane. For example, in this view, action influences perception even when reliable visual cues are available. Also, the information about action is dynamic and detected at the time of perception, rather than being based on a stored representation or internal bias (Witt & Proffitt, 2008). Importantly, this approach holds that the mechanisms responsible for these non-visual influences on perception are limited to unconscious motor-related processes, thus preserving the idea that perception is cognitively impenetrable with respect to conscious knowledge.

Contrasting with the possibility that behavioral potential modifies perception by altering non-visual internal biases or assumptions, the influence of unconscious motor-related processes on visual perception may well be a kind of multimodal interaction (Witt & Riley, 2014). Information from the motor system concerning the perceiver’s ability to act could serve as a non-visual input that gets integrated with the visual cues to determine perceived layout, distance, or size. Much research supports the notion that visual information can influence audition and vice-versa (for a review, see Shams, Kamitani & Shimojo, 2004) even though these effects are typically not included in models of visual perception like the modal model. Often, multimodal interactions are well-described by Bayesian cue combination, in the sense that the influence of each modality is weighted by its relative reliability (Ernst & Banks, 2002; see also Yang & Purves, 2003). Little is known about this in the context of motor-related influences on perception. Indeed, the reliability of this kind of motor information remains unknown. Nevertheless, the notion of multimodal integration provides a conceptual framework, and potentially a toolbox of methodological techniques, for studying action-specific motor-related influences on visual perception (Witt & Riley, 2014). Intention to act might be operationalized using electromyelographic or electroencephalographic data in future work (e.g., van Elk, van Schie, Neggers & Bekkering, 2010).

(3) Scaling of Visual Information

More recently, Proffitt and Linkenauger (2013) have proposed a “perceptual ruler” hypothesis to explain action-specific effects (Proffitt & Linkenauger, 2013). Optical information specifying size, distance, and other spatial properties of objects takes the form of visual angles. For example, binocular disparity is a difference in the visual angle separating the images of two objects between the two eyes. Likewise, texture gradients are informative about the slant of a surface because the angular size of the visible texture elements changes in a systematic way depending on the surface slant. Consequently, in order to perceive dimensions such as distance, size, and slant, optical information needs to be scaled from angles to these dimensions. Because optical information about spatial properties comes in the form of angles, the scaling mechanism must be a non-visual factor. Proffitt and Linkenauger (2013) argue that this scaling mechanism is, and must be, derived from the body.

The perceptual ruler hypothesis is supported by a variety of empirical findings that demonstrate effects of body size on judgments of spatial properties. For instance, manipulations of simulated body size in virtual environments have been shown to influence judgments of the size of other environmental objects (van der Hoort, Guterstam, & Ehrsson, 2011; van der Hoort & Ehrsson, 2014). Furthermore, hand size influences the reported size of graspable objects (Linkenauger et al., 2013; Linkenauger, Mohler, & Proffitt, 2011; Linkenauger, Ramenzoni, & Proffitt, 2010; Linkenauger, Witt, & Proffitt, 2011). Effective arm length, which has been manipulated by giving participants tools with which to reach, influences reported distance to objects presented beyond arm’s reach (Bloesch, Davoli, Roth, Brockmole, & Abrams, 2012; Davoli, Brockmole, & Witt, 2012; Osiurak et al., 2012; Witt, 2011b; Witt & Proffitt, 2008; Witt et al., 2005). According to this perceptual ruler account, it is not just the physical size of the body that provides a scaling mechanism but also the physiological potential and behavioral repertoire of the body. This inclusion allows for the mechanism to explain all action-specific effects to date.

From the action-resistant perception approach, there is general agreement that the information provided by the known egocentric and exocentric distance cues comes in the form of visual angles or visual directions, and that these angular values must be scaled by some aspect of the body in order to be informative (Cutting & Vishton, 1995; Sedgwick, 1986). For each cue, a trigonometric equation can be defined that relates three values: a specific aspect of the observer’s body, an optical (angular) value, and the target’s distance. These equations explicitly show how the body-based units establish a scale that allows visual angles to be transformed into perceived distance. For instance, the distance of a target on the ground is given by the observer’s eye height divided by the tangent of the object’s angular declination below eye level (Cutting & Vishton, 1995; Gajewski, Philbeck, Wirtz & Chichka, 2014; Mon-Williams & Bingham, 2008; Ooi, Wu & He, 2001; Sedgwick, 1986). In a similar way, eye height, along with shoulder width, can be used to scale the visual angles necessary to perceive the width of apertures (Warren & Whang, 1987). The existence of this kind of formal relationship constitutes an “informationally grounded” scaling mechanism, as described at length by Firestone (2013). In cases for which the body- or action-based scaling is informationally grounded, the claim that the body provides a perceptual ruler is generally acceptable to vision scientists from both the action-resistant and action-specific perception approaches. However, more controversially, some body- or action-based scaling factors have been proposed that are not obviously rooted in explicit, informationally-grounded equations.

One example involves using hand size to perceive the size of a small, graspable object. Of course, the size of an unfamiliar object can be determined by its distance from the observer and the visual angle that the object subtends: size is equal to the tangent of the visual angle times distance. Distance can be determined via one or more of the known visual egocentric distance cues, which, as we have argued, are already scaled by body-based factors. The issue here, however, is whether hand size provides a scale for perceiving the object size, over and above the known egocentric distance cues. A relationship between hand size and object size can be informationally grounded, but only if certain criteria are met. These include that the object is already in the hand or next to the hand and that the observer is familiar with the size of her own hand. In this case, the ratio of the visual angles subtended by the hand and object is available. The object’s linear size can be determined with respect to this ratio, because the object size will be this same proportion of the known linear hand size. Several studies demonstrate effects of hand size on perceived object size when these criteria are met (Linkenauger et al., 2010; Linkenauger et al., 2013). However, if any of the specified criteria are not met, the assumptions necessary to scale object size from hand size using these equations are no longer valid. Nevertheless, Linkenauger, Witt and Proffitt (2011) found that hand size influences judged object size even when the hand is not visible and the object and hand are at different distances—conditions that fail to satisfy the assumptions for informational grounding.

Proffitt (2013) has argued that body scaling might be mediated through calibration-like processes, involving learned relationships between visual angles and the outcomes of actions. Proffitt likens this kind of scaling to that of visually-guided actions. When performing a visually-guided action such as catching a fly ball or braking a vehicle to avoid a collision, representations of ball position or distance to obstacle are not necessary to perform the actions successfully. In these cases, the affordance is specified by optical variables, and once an observer has learned the relationship between the optical variable and the action, the observer can successfully perform the action without a representation of size or distance. Proffitt (2013) notes that body scaling in situations like using hand size to perceive the size of a small, graspable object could operate in the same way: observers learn the relationship between hand size, physical object size, and the object’s visual angle, and later are able to perceive object size on the basis of hand size and the object’s visual angle alone. A significant outstanding challenge is for proponents of the perceptual ruler account to specify and demonstrate how aspects of the body can provide a scaling mechanism when these parts of the body are not visible. Finding an informational basis for performance outcomes (such as proportion of softballs hit or golf putts made successfully) seems even more challenging.

For Firestone (2013), the lack of an informational basis for linking one’s body or action capability and the environment is a serious weakness that undermines the logical foundation of the entire action-specific perception approach. In contrast, from the action-specific perception perspective, research investigating these mechanisms is still in its infancy, and development of well-constrained explanations is the ultimate goal even if such explanations are not available at present. Nevertheless, according to this perspective, a particular body part or action capability can still play a functional role in scaling optical information even if there is no formal informational basis for linking the body or action capability with optical information and properties of the environment. To this extent, there is disagreement between the two approaches concerning the level of specificity and constraint in the hypothesized underlying mechanisms that are required to constitute acceptable evidence for action-specific perceptual effects.

Summary and Discussion

To date, relatively little theoretical or empirical work has been devoted to distinguishing this kind of action-specific perceptual influence from those that might operate by changing the proximal information available to the visual system through eye movements or attention. As mentioned above, the action-specific perception approach proposes that action-specific non-visual factors can influence perception even when there are no differences in either the distal or proximal visual information about distance (Proffitt, 2006; Witt, 2011a). Indeed, one goal of this paper is to prompt more research on the mechanisms that could underlie action-specific effects on perception.

There is nothing about these mechanisms, as described here, that would make them mutually-exclusive. Thus, all three could play a role in relative proportions that vary depending on the context. Similarly, these proportions might be different depending on what kind of geometrical property is under consideration (e.g., geographical slant versus egocentric distance versus object size): task-dependent eye movements might play more of a role for judgments of slant while body-based scaling of visual information may play more of a role for distance judgments, for instance. Although the tendency in the literature has been to put forward one or another of these mechanisms as capable of explaining all action-specific effects, few authors have taken the stronger stance that one mechanism is the one and only correct explanation for such effects, to the exclusion of the others. Ruling out other possible mechanisms that are not mutually exclusive is a sizeable challenge, and thus supporting the stronger assertions requires especially compelling theoretical and empirical arguments. At any rate, determining the relative weighting of these three mechanisms, and characterizing them in more detail, are important topics for future research. Techniques are available for recording eye movements and controlling or manipulating the locus of attention, so determining the role of action-specific differences in visual cues (via eye movements or attention) will be relatively straightforward, assuming that appropriate methods are in place to minimize output-related processes. Although some work of this kind is in progress, much remains unknown and this is a particularly fruitful direction for future research. Distinguishing between non-visual influences of action-related information and the perceptual ruler account may be more challenging.

Moving Forward: Recommendations for Future Research

Methodological Concerns

The above discussion shows not only the remarkable complexity and subtlety of post-perceptual processing, but also the great difficulty of controlling for, or otherwise mitigating against, the influence of these factors when studying action-specific effects. While it may be impossible to completely rule out post-perceptual processes as explanations for action-specific effects, there are several methodological features that can help make a more compelling case that action capability manipulations have influenced perception itself (see also Table 1).

Participants in all groups should be treated as identically as possible, to avoid possible experimenter effects. Interpersonal consistency can be enhanced by several means. Experimenters should use neutral affect when interacting with participants, and instructions should either be presented in written form or be highly consistent across participants. Experimenters should not be seen to exert effort themselves, as this could serve to emphasize that action capability is somehow relevant to the experimental hypotheses. Experimenters should also avoid giving feedback or encouragement to participants, because this interaction could have the effect of treating the treatment groups differently. Ideally, the experimenters should be blind to the experimental hypotheses, to further guard against experimenter effects. When feasible, the instructions should specify what response attitude participants should adopt when responding (e.g., report on apparent versus objective distance, ignore abstract connotations of distance). This can guard against variability stemming from individual differences in the assumed response attitude.

Naturally, care should also be taken to conceal the experimental hypotheses from participants. This is especially crucial for building a compelling case for action-specific perceptual effects. Past work has attempted to do this through the use of cover stories (as in Bhalla & Proffitt, 1999; Durgin et al., 2009), non-obvious manipulations (as in Schnall et al., 2010), and between-subjects instead of within-subjects manipulations (Witt et al., 2004; Woods, Philbeck, & Danoff, 2009). Indirect methods (e.g. Stefanucci & Proffitt, 2009; Witt, 2011b) may also be of use—because participants are asked to report on a different aspect of the scene than the one of interest, they may be less likely to infer the correct experimental hypothesis. Another approach is to group participants according to naturally-occurring differences in action capability rather than experimentally manipulating it (e.g. Bhalla & Proffitt, 1999; Sugovic & Witt, 2011, 2013).

Of course, use of these methods for concealing experimental hypotheses does not guarantee that they will be successful, and indeed the extent to which the hypotheses have been concealed in a given experimental paradigm can itself become a source of controversy (Durgin et al., 2009; Proffitt, 2013). Ideally, any attempt to conceal the hypotheses should be accompanied by some measure of its success. Surveys can be useful in assessing the extent to which participants may have been aware of the experimental hypotheses and/or were influenced by them. There is a rich empirical literature on how to construct and administer unbiased surveys (e.g. Schwarz, 1999), and although so far there has been little development of how best to construct surveys specifically in visual space perception contexts, the established principles of survey construction should be consulted as this development moves forward. In this regard, it is important to bear in mind that even if participants are able to deduce an experimental hypothesis when responding to a survey, this does not necessarily mean that they were aware of the hypothesis at the time of making perceptual judgments nor that this awareness influenced their perceptual judgments.

Underlying Mechanisms

In addition to these methodological recommendations, our review has also identified several more substantive issues that suggest important directions for future research. Chief among these is a pressing need to characterize the mechanisms underlying action-specific effects. We have highlighted the importance of controlling or manipulating eye movements and attention for identifying action-specific effects stemming from these factors. Developing methodologies for distinguishing between the other possible mechanisms outlined above remains an important challenge for future research.

Perceptual Stability and Phenomenology

A concern frequently raised about action-specific effects on perception is that they challenge the idea that perception is stable. There are two aspects of this concern. One is phenomenological, pertaining to the extent to which changes in perception ought to be noticeable. Another pertains to how actions might be calibrated to perceptions that change with the observers’ transient state of action capability and intention to act. These issues have important implications for future research. We will address these two aspects in turn.

First, if perception changes when one’s potential to act changes, it is reasonable to expect that the perceiver should notice such changes (Firestone, 2013). In this view, these transient changes in perception would constitute a breakdown in perceptual stability that arguably runs contrary to many people’s subjective experience. Upon donning a backpack, hills do not suddenly appear steeper. Upon grasping a tool, objects do not suddenly appear closer. Firestone (2013) argues that this lack of phenomenological experience of a change is evidence against the idea of action-specific perceptual effects. The central issue at stake concerning the phenomenology of action-specific perception is how much and what kind of change in perception is required for the change to become subjectively noticeable. Perception may tend to resist rapid fluctuations during eye fixations, for example, but be more likely to change rapidly (and unnoticed) across saccades or shifts of attention to match changes in action capability. There is a large literature addressing such issues in the domain of attention—particularly with reference to change blindness (e.g., Rensink, O’Regan & Clark, 1997; Simons, Franconeri & Reimer, 2000) and to contrast appearance (e.g., Anton-Erxleben, Abrams & Carrasco, 2010; Schneider & Komlos, 2008). The full relevance of this literature for perception of spatial features such as size, distance and slant remains poorly understood, however.

Witt (in press) has shown that people are not aware of a change in their perception of size in a dynamic version of the Ebbinghaus illusion. Observers viewed a center circle surrounded by large inducer circles, which switched to become small inducer circles (or vice versa). When probed afterwards about whether they noticed any change in the center circle size at the time of the switch, a large majority of participants responded that they did not. It is possible, even likely, that observers experienced the illusion both before and after the switch, but that there was some hysteresis of perception or continuity of experience during the switch itself such that the perceived size remained unchanged for a short time. That is, observers were not aware of a change in center circle size because the visual system tends to resist or filter out rapid changes in perception. Similarly, one might not be aware of a sudden change in perceived distance when putting on a heavy backpack because the visual system filters out this kind of sudden perceptual change and tends to maintain perceptual stability for some time. Important questions that remain open concern how rapidly perception can change (e.g., in the face of changes in action capability), and to what extent perception exhibits persistence when visual cues or action capability change. Answering these questions could add critical insights into, or constraints upon, the conditions under which action-specific changes in perception might be manifest. Thus, these issues constitute an important potential focus for future research.

Second, perceptual stability may seem to be a crucial precondition for effective control of actions. Perceptual coordinates must be transformed to motor coordinates prior to the execution of an action, and in principle this calibration transformation can correct for biases within perception and support accurate control of actions. If a stationary object appears to be at different distances depending on one’s transient intention to act, however, this might seem to present an insurmountable challenge for these calibration processes and should lead to widespread errors in the control of actions. A transformation function that is effective when a target is physically 10 m away but appears to be at 9 m would not be effective when that same target at 10 m appears to be at 11m. In this view, the fact that our actions continue to be relatively accurate could be taken as evidence that action-specific effects are not perceptual, because otherwise transformation functions would lead to significantly more errors.

However, actions could continue to be well-calibrated to the environment despite action-specific differences in perception if the calibrating transformations are themselves action-specific. For instance, consider an experiment examining heavy versus light ball throwing on estimated distance (Witt et al., 2004). There might be one transformation function for converting perceived distance to throwing coordinates for the light ball and another for the heavy ball. So even if the target looks to be at one distance when intending to throw the light ball and another distance when intending to throw the heavy ball, the transformation functions for each ball could be calibrated to the perceptual bias associated with each ball. Although it may seem implausible to have separate calibration functions for an object depending on the effort associated with throwing to it, the complexity could be reduced if there is a common calibration function for throwing actions that is scaled by one or two effort-related parameters. When the intention to act is established, the effort parameter is set and the appropriate transformation from perceptual coordinates into motor coordinates is applied. This idea has not been tested empirically, and important issues would need to be worked out to characterize this kind of scheme adequately (for example, to what extent is the effort parameter under conscious control). Nevertheless, under this kind of scheme, even if there were action-specific failures of perceptual stability, these failures would not be problematic for maintaining reasonably accurate control of actions. A variety of techniques for investigating motor calibration have been developed (e.g., Durgin et al., 2005; Kurtzer, Dizio & Lackner, 2005; Pan, Coats & Bingham, 2014, among a host of others), and a promising area of future research is to use such techniques to flesh out calibration functions in the domain of human egocentric distance perception.

Replicability and Reproducibility

Earlier we discussed apparent failures to replicate certain aspects of previous action-specific perceptual studies (e.g., Durgin et al., 2009; Durgin et al., 2012; Hutchison & Loomis, 2006; Shaffer et al., 2013; Woods et al., 2009). Some of these apparent failures to replicate have themselves been criticized on methodological grounds (e.g., Proffitt, 2009, 2013; Proffitt, Stefanucci, Banton & Epstein, 2006), and there have been many replications of action-specific findings (Bloesch et al., 2011; Cañal-Bruland et al., 2010; 2012; Cañal-Bruland & Van der Kamp, 2009; Cañal-Bruland et al., 2011; Davoli et al., 2012; Doerrfeld et al., 2012; Eves et al., 2014; Gray, 2013; Gray et al., 2014; Kirsch et al., 2012; Kirsch & Kunde, 2013a,b; Kuylen et al., 2014; Lee et al., 2012; Morgado et al., 2013; Osiurak et al., 2012; Taylor-Covill & Eves, 2013; 2014; Thomas et al., 2014; van der Hoort & Ehrsson, 2014; van der Hoort et al., 2011; Wesp et al., 2004; White et al., 2013). Controversies aside, these dialogs in the literature have illuminated important methodological issues that are important to bear in mind in future research—many of which are summarized in this review. Thus, attempts to replicate past work, not only within laboratories, but across laboratories with a history of past collaboration and, most importantly, across completely unrelated laboratories, is a vital means of characterizing the boundary conditions for action-specific influences and output-related effects. Demonstrating reproducibility across laboratories is crucial for any experimental finding, as this can serve to bolster confidence that systematic effects in the data are not due to specific aspects of a particular experimental context (for example, particular experimenters, laboratory spaces, stimulus materials, procedures, and so on). Such efforts will also be important for establishing norms for future computations of the size of effects to be detected and the sample sizes needed to detect effects of a given size. As methodologies have become more refined for studying action-specific effects, a potentially fruitful avenue of future research would be to attempt to replicate past work in a way that contrasts the original methodology with some of the more refined methods. This would provide an empirical foundation for evaluating the role of the factor that the methodology is designed to control. For example, if using neutral affect when interacting with participants is intended to minimize experimenter effects, experimentally manipulating experimenter affect in a previously-test action-based paradigm (e.g., ball throwing; Witt et al. 2004) would provide an empirical test of this idea.

Conclusions: Moving Towards Reconciliation

The action-specific perception and action-resistant perception approaches assume a very similar underlying model, which we have called the modal model. Both approaches assume that visual information plays a primary role in determining perceptual experience, and both approaches agree that output-related processing is important to control or otherwise factor out when interpreting behavioral responses as indicators of perception. Both also agree that there are some mechanisms that could elicit bona fide influences of action capability on perception, for example, by way of task-dependent differences in eye movements or attention or by the influence of certain non-visual factors. Importantly, this similarity in the assumed model provides a common language for conceptualizing the possible mechanisms underlying action-specific effects on behavioral indications of perception. Research from within the action-resistant perception perspective describes mechanisms that could potentially give rise to action-specific perceptual effects (e.g., eye movements or attention) and that generate non-visual influences on perception (e.g., perceptual biases and perceptual learning).

There are also important differences between the two approaches, however, specifically in terms of the emphasis they place on specific parts of the model. The assumed role of output-level factors is a primary difference between these approaches. It is important to emphasize that although the action-specific perception approach acknowledges that output-level factors can and do influence behavioral indications of perception, this approach proposes that these factors play virtually no role in explaining results of most, if not all, of the experiments targeting action-specific influences on perception. In contrast, the action-resistant perception account assumes that output-related factors account for the large majority of action-related effects.

Another major difference is that the action-specific perception perspective allows for direct influences on perception that do not stem from attention, eye movements, or differential selection or weighting of visual cues. Two alternative mechanisms have been proposed. For one, action-relevant information is integrated with visual information in a similar manner as the multimodal integration of auditory and visual information (Witt & Riley, 2014). For another, action-relevant information provides a mechanism for scaling visual information from angles into units such as distance and size (Proffitt & Linkenauger, 2013). One of the current limitations of the action-specific approach is the relatively sparse amount of empirical data that addresses the underlying mechanism.

Insights Gained by Rejecting All-or-Nothing Conceptualizations of Action-Specific Perception

One insight underlies a recurring theme in this review: namely, that simply debating whether or not action-specific perceptual effects exist is an overly-coarse level of analysis. If one rejects all forms of non-visual influences on perception, one must also reject a subset of research from the action-resistant perception perspective. Alternatively, if one embraces this subset of the action-resistant perception literature, one must acknowledge that legitimate action-specific perceptual effects might occur under at least some circumstances. This perspective highlights the importance of characterizing the boundary conditions under which action-specific perceptual effects do or do not occur (while ruling out output processing as much as possible) and identifying the mechanisms underlying bona fide perceptual effects.

A second insight comes from considering why some examples of non-visual influences on perception (such as those coming from the action-specific perception approach) have met with more intense criticism than other examples rooted more firmly in the action-resistant perception approach (e.g., the Specific Distance Tendency and Intrinsic Ground Plane; Gogel, 1990; Ooi, Wu & He, 2006). One notable difference is that work on non-visual factors has often involved reduced-cue viewing contexts, in which glowing targets are seen in otherwise dark environments. This permits greater control over potentially relevant visual information and accordingly enhances the opportunity to unmask influences of non-visual factors. Reduced-cue contexts thus generally afford a more well-constrained analysis of the factors that play a role in behavioral judgments than do more naturalistic viewing conditions. By contrast, for many action-specific perception researchers, more ecological viewing conditions are favored because the interest is in how perception operates in the context of action rather than isolated from action. This is not to say that reduced-cue settings are necessary for acceptance of evidence of non-visual influences on perception, but just that reduced-cue settings may have contributed to their acceptance. Other potential factors in the acceptance of some non-visual influences on perception by the action-resistant perception approach are that much of the work has been cross-validated using a variety of direct and indirect behavioral measures, that it has generally adopted a cautious and incremental approach, and that it has generally treated post-perceptual influences as complex and insidious.

We acknowledge that there is no “magic formula” that will guarantee bilateral acceptance in a contentious domain. Ultimately, bilateral acceptance of action-specific perceptual effects hinges in part upon the degree to which this research program can (1) adequately guard against post-perceptual effects, and (2) specify concrete and well-constrained underlying mechanisms. In addition, given that definitively ruling out output-level explanations of action-specific effects may be impossible, caution should also be expressed when interpreting results as favoring action-specific influences on perception. At a minimum, there should be some acknowledgement of the complexity of the issue, with some discussion of the relative likelihood of various forms of output-level biases in the studied context. Research founded in the action-specific perception approach has been increasingly mindful of many of these issues. To some extent, the criteria for what constitutes acceptable control of post-perceptual effects and what constitutes an appropriately specific mechanism vary between approaches, and bridging this gap is a central challenge for future research.