The target article contains an excellent prescriptive exposition for scientific sanity checks. But it also argues there is an a priori impenetrability that challenges us to consider anew the relationship between perceptual experience and cognitive attribution.
When a person is asked to put on a heavy backpack in an experiment, experimenters can probe their perceptual experience of space by multiple methods. For example, in the original report of their backpack effect Bhalla & Proffitt (Reference Bhalla and Proffitt1999), argued that verbal judgments reflected (distortions of) conscious perception, whereas haptic matching tasks represented (undistorted) unconscious perception. When we (Durgin et al. Reference Durgin, Baird, Greenburg, Russell, Shaughnessy and Waymouth2009) simply provided an explanation for the backpack (“it contains EMG [electromyography] equipment so we can monitor your ankle muscles”), the verbal effect went away. Moreover, later investigations of haptic matching tasks (Durgin et al. Reference Durgin, Hajnal, Li, Tonge and Stigliani2010; Reference Durgin, Hajnal, Li, Tonge and Stigliani2011b; Li & Durgin Reference Li and Durgin2011; Shaffer et al. Reference Shaffer, McManama, Swank, Williams and Durgin2014) suggest that they are controlled by conscious perception (often the comparison of misperceived haptic orientation with misperceived hill orientation). In the end, the dissociations between the biasing of verbal and the non-biasing of haptic matching appeared to reflect a dissociation between biases in judgment (attributional effects) and perceptual stability.
Some have summarized our work on experimental demand as showing that the effects of effort (say) can be eliminated by beliefs, but our data go much further than that. We realized that in studies in which blood sugar was manipulated quite carefully (Schnall et al. Reference Schnall, Zadra and Proffitt2010), the experimental manipulations purporting to show that lower blood sugar led to higher slant estimates confounded several uncontrolled contaminants. The blood sugar manipulation (supplying a drink that did or did not contain sugar to participants in a state of low blood sugar from fasting) was always followed by a cognitive task, such as a Stroop task simply to pass the time (might this juxtaposition matter?). When replicating this procedure, we (Durgin et al. Reference Durgin, Klein, Spiegel, Strawser and Williams2012) found that this sequence led participants to look for a relationship between the drink and the cognitive task, rather than the drink and the hill. Moreover, nearly all subjects who were not confident about being able to tell sugar from sweetener assumed the drink we gave them contained sugar (might this belief matter?). Finally, the studies of blood sugar all employed a heavy backpack manipulation prior to the exposure to the hill. The backpack manipulation, in particular, seemed like a puzzling choice in an otherwise relatively clean experiment. What if low blood sugar just makes people more cooperative with the demand of the backpack?
Shaffer et al. (Reference Shaffer, McManama, Swank and Durgin2013) removed the intermediate cognitive task and used an EMG deception in conjunction with a cover story about the sugar/no-sugar drink itself (that it improves the EMG signal). Therefore, we could remove the experimental demand of the heavy backpack, leaving clever participants to notice that the experiment involved a drink manipulation followed (after time sitting to absorb the fluid) by a hill judgment. (To test for any residual backpack effect, for half of the subjects the experimenter carried the backpack; there was no effect of wearing the backpack on estimates of the hill slant.) Although we had also provided a cover story for the sweet drink, a quarter of our participants reported in a written survey that they thought the drink was intended to affect their perception of the hill.
Consider that these suspicious/insightful participants were evenly divided between those who had been given sugar in their drink and those who had not, whereas all had arrived at the lab in a state of low blood sugar due to fasting. But consider also that nearly everyone assumed that the drink contained sugar (even though in this experiment we told them that it contained only electrolytes). Given the belief that the drink contained sugar, participants who cooperated with experimental demand should have given lower estimates of the hill slant than those who did not cooperate with experimental demand. But if the effect of low blood sugar (previously fasting participants not actually given sugar) is to increase cooperation, then we should see the ironic effect that among those who deduced that the experiment concerned sugar, those who had received no sugar in their drinks should actually give lower estimates than everyone else – and that is exactly what happened.
Therefore, in the case where effort theory made a clear and direct prediction (low blood sugar should increase slant estimates), ironically, demand prediction won out: Low blood sugar increased cooperation with experimental demand, which, in this experimental context, led to lower estimates of hill slant. Note that the magnitude of these ironic effects was just as large as the effects observed in backpack studies, but in the opposite direction.
In addition to the methodological motivations emphasized in the target article (control experiments to test alternative hypotheses are important), in our pursuit of these questions a fundamental theoretical commitment has developed: In the perception of large-scale space, at least, we do seem to see large perceptual distortions that are fairly stable across individual and contexts (e.g., Durgin Reference Durgin2014). The best explanation of these distortions appeals to information-theoretic coding advantages (comparable to Huffman coding) that tend to perceptually magnify certain angular variables that are quite useful for the control of locomotor action. The crucial theoretical point here is that systematic and stable perceptual bias (causing the exaggeration of apparent slant and the underestimation of apparent distance) can be advantageous for action control, but that momentary destabilization of space perception by desire, fatigue, and so forth would tend to undermine the whole point of perception as a guide for action.
I have no a priori commitment to the impenetrability of perception. I thought the hill experiments were cool when I first learned of them. But the more one looks into them, the less likely they seem to have been correctly interpreted. Cognitive systems, in general, must walk a tightrope between stability and flexibility: We need to be able to take in new information without losing the old. It may be that perceptual systems tend actually to be self-stabilizing but that attribution systems are free to reflect more of the totality of our knowledge (boy, that hill seems steep!).
The target article contains an excellent prescriptive exposition for scientific sanity checks. But it also argues there is an a priori impenetrability that challenges us to consider anew the relationship between perceptual experience and cognitive attribution.
When a person is asked to put on a heavy backpack in an experiment, experimenters can probe their perceptual experience of space by multiple methods. For example, in the original report of their backpack effect Bhalla & Proffitt (Reference Bhalla and Proffitt1999), argued that verbal judgments reflected (distortions of) conscious perception, whereas haptic matching tasks represented (undistorted) unconscious perception. When we (Durgin et al. Reference Durgin, Baird, Greenburg, Russell, Shaughnessy and Waymouth2009) simply provided an explanation for the backpack (“it contains EMG [electromyography] equipment so we can monitor your ankle muscles”), the verbal effect went away. Moreover, later investigations of haptic matching tasks (Durgin et al. Reference Durgin, Hajnal, Li, Tonge and Stigliani2010; Reference Durgin, Hajnal, Li, Tonge and Stigliani2011b; Li & Durgin Reference Li and Durgin2011; Shaffer et al. Reference Shaffer, McManama, Swank, Williams and Durgin2014) suggest that they are controlled by conscious perception (often the comparison of misperceived haptic orientation with misperceived hill orientation). In the end, the dissociations between the biasing of verbal and the non-biasing of haptic matching appeared to reflect a dissociation between biases in judgment (attributional effects) and perceptual stability.
Some have summarized our work on experimental demand as showing that the effects of effort (say) can be eliminated by beliefs, but our data go much further than that. We realized that in studies in which blood sugar was manipulated quite carefully (Schnall et al. Reference Schnall, Zadra and Proffitt2010), the experimental manipulations purporting to show that lower blood sugar led to higher slant estimates confounded several uncontrolled contaminants. The blood sugar manipulation (supplying a drink that did or did not contain sugar to participants in a state of low blood sugar from fasting) was always followed by a cognitive task, such as a Stroop task simply to pass the time (might this juxtaposition matter?). When replicating this procedure, we (Durgin et al. Reference Durgin, Klein, Spiegel, Strawser and Williams2012) found that this sequence led participants to look for a relationship between the drink and the cognitive task, rather than the drink and the hill. Moreover, nearly all subjects who were not confident about being able to tell sugar from sweetener assumed the drink we gave them contained sugar (might this belief matter?). Finally, the studies of blood sugar all employed a heavy backpack manipulation prior to the exposure to the hill. The backpack manipulation, in particular, seemed like a puzzling choice in an otherwise relatively clean experiment. What if low blood sugar just makes people more cooperative with the demand of the backpack?
Shaffer et al. (Reference Shaffer, McManama, Swank and Durgin2013) removed the intermediate cognitive task and used an EMG deception in conjunction with a cover story about the sugar/no-sugar drink itself (that it improves the EMG signal). Therefore, we could remove the experimental demand of the heavy backpack, leaving clever participants to notice that the experiment involved a drink manipulation followed (after time sitting to absorb the fluid) by a hill judgment. (To test for any residual backpack effect, for half of the subjects the experimenter carried the backpack; there was no effect of wearing the backpack on estimates of the hill slant.) Although we had also provided a cover story for the sweet drink, a quarter of our participants reported in a written survey that they thought the drink was intended to affect their perception of the hill.
Consider that these suspicious/insightful participants were evenly divided between those who had been given sugar in their drink and those who had not, whereas all had arrived at the lab in a state of low blood sugar due to fasting. But consider also that nearly everyone assumed that the drink contained sugar (even though in this experiment we told them that it contained only electrolytes). Given the belief that the drink contained sugar, participants who cooperated with experimental demand should have given lower estimates of the hill slant than those who did not cooperate with experimental demand. But if the effect of low blood sugar (previously fasting participants not actually given sugar) is to increase cooperation, then we should see the ironic effect that among those who deduced that the experiment concerned sugar, those who had received no sugar in their drinks should actually give lower estimates than everyone else – and that is exactly what happened.
Therefore, in the case where effort theory made a clear and direct prediction (low blood sugar should increase slant estimates), ironically, demand prediction won out: Low blood sugar increased cooperation with experimental demand, which, in this experimental context, led to lower estimates of hill slant. Note that the magnitude of these ironic effects was just as large as the effects observed in backpack studies, but in the opposite direction.
In addition to the methodological motivations emphasized in the target article (control experiments to test alternative hypotheses are important), in our pursuit of these questions a fundamental theoretical commitment has developed: In the perception of large-scale space, at least, we do seem to see large perceptual distortions that are fairly stable across individual and contexts (e.g., Durgin Reference Durgin2014). The best explanation of these distortions appeals to information-theoretic coding advantages (comparable to Huffman coding) that tend to perceptually magnify certain angular variables that are quite useful for the control of locomotor action. The crucial theoretical point here is that systematic and stable perceptual bias (causing the exaggeration of apparent slant and the underestimation of apparent distance) can be advantageous for action control, but that momentary destabilization of space perception by desire, fatigue, and so forth would tend to undermine the whole point of perception as a guide for action.
I have no a priori commitment to the impenetrability of perception. I thought the hill experiments were cool when I first learned of them. But the more one looks into them, the less likely they seem to have been correctly interpreted. Cognitive systems, in general, must walk a tightrope between stability and flexibility: We need to be able to take in new information without losing the old. It may be that perceptual systems tend actually to be self-stabilizing but that attribution systems are free to reflect more of the totality of our knowledge (boy, that hill seems steep!).