Phenomenal Consciousness Disembodied

Phenomenal Consciousness Disembodied * WESLEY BUCKWALTER University of Waterloo [email protected] MARK PHELAN Lawrence University [email protected] Abstract: We evaluate the role of physical embodiment in ordinary mental state ascriptions to disembodied spirits. Presented are five experiments suggesting that embodiment is not a central principle of folk psychology guiding ascriptions of phenomenal consciousness. Results continue to support the important role of functional considerations in theory of mind judgments. 1. Introducing Some Phenomenal Bodies “Suppose we convert the government of China to functionalism, and we convince its officials to realize a human mind for an hour.” This is, of course, the beginning of Ned Block's famed “Nation of China” thought experiment (1978, p. 279). In it, Block asks us to imagine that approximately 1 billion people (roughly the population of China at the time) come together to simulate the inner workings * The authors are grateful to Matthew Kieran, Joshua Knobe, Aaron Meskin, and John Turri for helpful comments on earlier versions of this paper. Consciousness Disembodied | 2 of a normal human brain. Each person is given a two-way radio that enables him or her to communicate with others, much like individual neurons in the brain communicate with other neurons. What's more, this complex communication network is hooked up to a remote, artificial body. As the sensory organs of that body are stimulated, the external state of the body is reported on a system of satellites, visible anywhere within China. According to a set of specified rules, the Chinese respond to these satellite reports by relaying information and commands to one another, and issuing instructions to the body that cause it to execute various behavioral routines. In a sense, the nation of China has now become a “China-Brain” hooked up to an artificial body (just by radio waves rather than electrochemical impulses). Arguably, such a system satisfies a purely functional description of a mind. Broadly speaking, functionalism defines individual mental states in terms of sensory inputs, behavioral outputs, and relations to other mental states. So according to this view, the various states of the nation of China would amount to the thoughts, feelings, desires, and so on, of normal human beings. But could a being of such odd construction be said to have a mind, filled with the same kinds of thoughts, feelings, and desires as us? Part of the original goal of Block's “Nation of China” thought experiment is to demonstrate that it could not. In fact, Block finds it doubtful that we will say that the “China-Brain” has any mental states at all—least of all any “qualitative states, raw feels, or imme- 3 | Buckwalter and Phelan diate phenomenological qualities" (p. 281). It simply is not made of the right stuff for such states to be possible. This intuition has been taken by some as evidence against the claim that certain theories of functionalism adequately capture how the mind works. Intuitions in response to Block's thought experiment continue to be shared and debated. But the basic example has remained influential in both psychology and cognitive science. 1 One prominent position that has emerged from this discussion is the view that only certain sorts of entities are capable of having phenomenally conscious mental states such as emotions or raw feels. In addition to being functionally organized in the right sort of way, entities capable of certain mental states (and in particular phenomenal experiences) it is said, must also have the right sort of biological bodies. They must be made of the right stuff. This traditional view held by certain philosophers of mind about how the mind works has also been explored by others as an important principle of folk psychology guiding ordinary mental state ascription. A number of researchers in philosophy, psychology, and cognitive science have defended versions of what we will call the embodiment hypothesis (Knobe 2008, Knobe & Prinz 2008, Gray et al. 2011). The embodiment hypothesis states that physical embodiment is a major factor that cues ordinary attribution of ex- 1 Much the same could be said for both the original purpose and subsequent legacy of Searle’s “Chinese Room” argument (1980) with respect to the questions of function and embodiment. Consciousness Disembodied | 4 periences, feelings, emotions, and so on, to other entities. On this view, only entities with the right kind of biological body are typically thought capable of having phenomenally conscious experiential states. And conversely, entities without the right kind of biological body, such as robots, groups, ghosts, or perhaps even Block's “Nation of China”, will typically be judged incapable of having these sorts of mental states. To get an idea of how the embodiment hypothesis works, it may be helpful to consider a specific example. One way that advocates of the embodiment hypothesis have defended their view is by looking at people's intuitions about group agents. As Knobe & Prinz (2008) write: It is a striking fact about group agents that we ascribe to them some types of mental states but not others. We might say that Microsoft intends something or wants something or believes something…but there are other kinds of ascriptions that we would never make to Microsoft. For example, we would never say that Microsoft was feeling depressed. (p. 73) Knobe & Prinz conduct several studies that suggest that this is indeed the case.2 They then explain this striking fact by suggesting that there are important differences in how people ascribe intentional states (like intending or wanting) on the one hand, and states requiring phenomenal consciousness (like feeling sad or depressed) 2 For an alternative explanation of Knobe & Prinz's findings, see Phelan et al. (forthcoming) or our discussion in Section 3.5 below. 5 | Buckwalter and Phelan on the other. The embodiment hypothesis claims that attributions of these latter kinds of mental states hinge on an entity’s physical constitution. That is, according to the embodiment hypothesis, we don't attribute phenomenal states or subjective experiences to the Microsoft Corporation because it does not have the right kind of body. This specific hypothesis about the important role of embodiment in folk psychological judgments has not been argued for by Block directly. However this view would be one way of explaining intuitions in Block's original thought experiment, which some philosophers have used as evidence when arguing for certain conclusions in the metaphysics of mind.3 However not everyone has reported having these intuitions, much less agreed that “China-Brain” constitutes an acceptable counterexample to functionalism, or that embodiment plays a crucial role in either possessing or attributing mental states. A number of philosophers and cognitive scientists sympathetic to functional accounts of the mind have suggested that intuitions favouring embodiment in both thought experiments and empirical studies might in fact be trading on subtle cues and distractions related to the functional organization of the target entities. Dennett (1991), for example, argues that Block's thought experiment unfairly relies on a “misdirection of the imagination” because it nonchalantly invites readers to buy into the unlikely idea that the China-brain is complex enough to satisfy the functional roles asso3 After all, the nation of China is one (special kind) of group entity. Consciousness Disembodied | 6 ciated with particular mental states. And, as we have argued elsewhere (Phelan and Buckwalter, forthcoming), it's possible that many of the experimental materials researchers have used when studying the influence of embodiment on the ascription of experiential states also manipulated crucial functional information (e.g., information about inputs, outputs, and other mental states). Before continuing, it may be helpful to pause for a moment to reflect on how this general functionalist picture of the mind might be represented in folk psychology. Suppose, for instance, someone was trying to figure out whether or not an entity named 'Bob' feels happiness or anger. According to this picture, there may be several crucial factors that one might rely on for deciding whether or not Bob is experiencing these states. One factor would involve thinking about how either happiness or anger might stand in the right relation to all the other mental states the entity has (e.g. How does it fit with Bob's goals, desires, or plans?). Relatedly, one might also look to the relationship between the state and the effects of external stimuli (e.g. agents preventing or helping Bob in achieving his goals), or between the state and behavioural responses (e.g. whether Bob is smiling or clenching his fists). Indeed, we have shown in prior work that people often consult these kinds of cues when deciding whether or not to ascribe phenomenal states (Buckwalter and Phelan, 2012). Of course, even if functional information of this sort does cue phenomenal state attributions, it could still be that embodiment 7 | Buckwalter and Phelan (for instance, whether 'Bob' is a normal human being, a group, or an immaterial ghost) constitutes an important ascription cue as well. However, we think there are good reasons to independently reject the embodiment hypothesis. This is what we will attempt to demonstrate. In the remainder of this paper, we examine people’s ascriptions of experiential states to entities lacking a biological body. Our goal is to see if ascriptions of phenomenal states to these sorts of entities differ from ascriptions made about normal human beings, or if they tend to work in the same basic way. For this task, a number of different disembodied entities might have been used. But choices must be made. We chose to begin our examination of disembodied ascription with the phantasmally disembodied—ghosts and spirits. Lacking in any body whatsoever, spirits constitute the ultimate test of the embodiment view. If the embodiment hypothesis is correct, and embodiment is a crucial cue for phenomenal state attribution, then we should expect important differences in ascription between human beings, on the one hand, and disembodied ghosts and spirits, on the other—just as we expect to find important differences in phenomenal state attribution for functional information. If functional information—information about the goals, desires, etc, of an entity—tends to cue mental state ascription independently of whether the entity has a physical body, then it undermines the embodiment hypothesis. This is what we set out to investigate, using spirits as our medium. Consciousness Disembodied | 8 2. Previous Studies on Mental State Attributions to Disembodied Entities We are under no illusion that our investigation into how people attribute mental states to the disembodied is unprecedented. Several influential studies on God and ghosts have already been conducted. Indeed, previous findings seem to offer contradictory evaluations of the embodiment hypothesis. Some have taken the findings of Gray, Gray and Wegner (2007) to support embodiment.4 Gray et al. analyzed comparative attributions of a range of mental states to a cast of “characters” ranging from babies to adults, from robots to animals to the dead, and including the ultimate disembodied entity: God. They found that people were less willing to attribute phenomenal mental states (such as feeling fear, hunger, or pain) to God than to many of the other characters in their study. This finding is consistent with the embodiment view, since, presumably, God is thought to lack a body whereas other of the characters are not. Furthermore, insofar as lack of embodiment is what explains these low attributions to God, the findings are inconsistent with a simple functionalist account of people's phenomenal state attributions. However, as we have argued elsewhere (Phelan and Buckwalter, forthcoming), the findings are readily explicable in functionalist terms, since God is thought to be the ultimate being, who wants for nothing. Surely then he will be thought to suffer fear, hunger, and pain less often 4 Note however that Gray & Wegner (2010) questions whether these prior findings about God are best interpreted as supporting embodiment. 9 | Buckwalter and Phelan than a child or a toad (two other of Gray et al.'s characters). 5 On the other hand, findings from Jesse Bering and colleagues could be interpreted as challenging the disembodiment view. Bering found that adults thought psychological functions—including emotional states—continued after biological death in an agent killed on his daily commute (2002). Bering and Bjorklund (2004) found a similar pattern for children, who continued to attribute emotional and other mental states to a mouse after it was eaten by an alligator. And Bering et al. (2005) found that both secularly and parochially educated children under 10 were proportionally more likely to disagree with statements indicating that psychological functions including emotional states ceased at death. On the assumption that each population thought of the recently deceased as disembodied (a supposition supported by the fact that each population tended to think that biological function ceased at death), these studies present prima facie counter-evidence to the embodiment view. While Bering's work is illustrative, for our present purposes it does not constitute a true test of the embodiment hypothesis. For one thing, it doesn't explicitly compare people's attributions of phenomenal states to the disembodied with their attributions of phenomenal states to normal, embodied humans. Thus it might have missed a tendency to attribute mental states in a way consonant 5 Similar considerations, we think, explain Gray et al.'s findings for other phenomenal states. Consciousness Disembodied | 10 with embodiment. For another thing, it doesn't explicitly manipulate function. Thus it offers no comparison between cues related to embodiment and other salient cues of phenomenal state attribution. Finally, Bering's experimental materials don't explicitly inform participants that the recently deceased agents are disembodied, nor do they ask participants whether they conceive of the dead agents in this way. It thus remains a possibility that experimental participants are not equating death with disembodiment in a way that would shed light on the embodiment hypothesis. Therefore we use this body of research on ordinary beliefs about souls as a point of departure for testing the embodiment view.6 3. Disembodying Ascription We present five experiments investigating people's willingness to ascribe emotional states to disembodied ghosts and spirits. The take home message of this section is that there are certain conditions under which people are perfectly willing to ascribe phenomenal states to these kinds of disembodied entities. Using functional information as a control for comparison, Experiment 1 demonstrates that 6 These are not criticisms of Bering's work, since he wasn't out to investigate the issue of physical realizers at all. In fact, Bering is one of a number of theorists arguing for a particular view about the source of afterlife beliefs, which Bering (2011) calls the simulation constraint hypothesis. Nichols (2007), another proponent of the view, encapsulates the basic idea as follows: “part of the reason we believe in immortality is that we can’t imagine our own nonexistence” (p. 216). Interesting as the connections are between the embodiment hypothesis and afterlife belief, we set them aside. 11 | Buckwalter and Phelan emotional ascriptions of felt emotions to ghosts matches that of ordinary human beings, so long as either of those entities satisfy the relevant functional descriptions. Experiment 2 replicates this effect using experimental probes that lack intentional objects with potentially biasing contextual information (see Section 2.2). Experiments 3-4 display commensurate results for attributions to eternally disembodied spirits. And lastly, using an explicit comparison technique, Experiment 5 provides evidence that in making such ascriptions, participants literally attribute the same phenomenal states to disembodied ghosts and spirits as they attribute to ordinary human beings. 3.1. Experiment 1: Disembodied Ghost 3.1.1. Methods We begin with a between-subjects multifactor experiment designed to test the influence of embodiment and functional cues on ascriptions of phenomenal states.7 Participants in Experiment 1 (N=158, 85 female, median age = 27) were presented with the following story about Bob, his ex-wife Melissa, and their son Henry. Roughly half of the participants received a version where the functional information (in this case, the goal Bob attempts to bring about) is to 7 Participants in this and all subsequent experiments were recruited and tested using commercially available online platforms (Qualtrics and Amazon Mechanical Turk). Participants were located in the United States, and over 85% reported English as a first language. They were paid between $0.30 and $0.45 for their participation. Participants were prohibited from taking more than one study. Consciousness Disembodied | 12 cause his son Henry to hate his mom, while the other half saw a condition in which Bob aimed to make Henry happy: [ANGER/HAPPY] Bob and Melissa have been married for 15 years. After several months of intense bickering and fighting, they decide to get a divorce. Bob moves out of the house, but still tries to spend time with Henry, their ten-year-old son. He also continues to keep close tabs on his ex-wife Melissa. One day, Bob learns that Melissa has started a new romantic relationship. He hires a private investigator to follow the couple, and take photos of them over a romantic dinner. Bob knows that it will [cause Henry to hate his mom, Melissa, if he learns that she/ make Henry incredibly happy if he learns that his mom, Melissa,] has started a new, meaningful relationship. Suddenly, Bob gets an idea. If he leaves the pictures in Henry’s treehouse in the backyard, Henry is sure to find them when he gets home from school that day. So, Bob jumps in his car and drives to Melissa’s house. After reading one of the versions above, half of the participants saw a conclusion to the story where Bob’s physical body is made salient: [EMBODIED] On the drive over however, Bob is in a car accident. Bob emerges from his car and looks over his body. Everything seems to be completely fine—his head, legs and arms. But even though Bob has been in an accident, he won’t let that deter him from his earlier goal. He takes the pictures out of his car and walks them over to Melissa’s house. He carries them 13 | Buckwalter and Phelan over the back fence and into the treehouse, where Henry is sure to see them. The remaining participants saw a conclusion to the story where Bob had no physical body: [DISEMBODIED] On the drive over however, Bob is in a fatal car accident and is killed instantly. Bob emerges from his dead body as a ghost. He now has no form at all—no head, no legs, no arms. Instead, he is something like an invisible force or a spiritual presence. Though he has no limbs with which to touch physical objects, Bob can make objects move without touching them, by floating them through the air. But even though Bob is a ghost, he won’t let that deter him from his earlier goal. He causes the pictures to rise out of his car and to float towards Melissa’s house. He moves them over the back fence and into the treehouse, where Henry is sure to see them. All participants were then asked to rate their level of agreement with the following three statements regarding what Bob both felt and believed at the end of the story: Belief. As Bob moves the pictures into place, he believes Henry will find them in the treehouse after school. Feel Anger. As Bob moves the pictures into place, he feels angry at Melissa for beginning a new relationship. Feel Happiness. As Bob moves the pictures into place, he feels happy for Melissa for beginning a new relationship. Responses were collected on the same seven-item scale anchored Consciousness Disembodied | 14 with positive and negative agreement terms designed to measure people’s willingness to attribute these intentional states (Belief) and experiential states (Feel Anger and Feel Happiness) to Bob. 3.1.2. Results and Discussion We made three main predictions. A large body of prior empirical work has demonstrated that the behavior displayed by an entity is a crucial factor that cues attributions of intentional states to that entity (Heider & Simmel 1944). So our first prediction was that given Bob’s behaviors in the story, participants would signal high levels of agreement with Belief across all conditions in the experiment. Second, we predicted that functional information would have a large impact on phenomenal state attribution, whereby people would signal greater agreement with Feel Anger in the ANGER condition, and greater agreement with Feel Happiness in the HAPPY condition. And lastly, our third prediction was that embodiment would play little to no role in cuing phenomenal state ascription. All three predictions were borne out. We found that people strongly agreed that Bob believes Henry will find the pictures in the treehouse after school.8 Second, there were large effects for function on the way people attributed emotional states to Bob—differences in Bob's non-phenomenal mental states (i.e., his goals) made a big difference in the phenomenal states that were attributed to Bob. 8 Belief Disembodied-Anger (M = 6.39, SD = 1.05), Disembodied-Happy (M = 6.42, SD = 0.81), Embodied-Anger (M = 6.54, SD = 0.70), Embodied-Happy (M = 6.17, SD = 1.23). 15 | Buckwalter and Phelan And third, emotional state attributions appeared completely unaffected by whether or not Bob was embodied.9 These results are represented in the figure below: Figure 1. Mean agreement with mental state attribution in each condition, grouped by mental state probe. All scales ran 1-7. Recall that if the embodiment hypothesis is correct in claiming that 9 A 2 (Embodiment) x 2 (Function) MANOVA was used to compare the influence that body salience and function had on the intentional and experiential states: belief, anger and happiness. The multivariate result was significant only for function, Pillai’s Trace = 0.48, F = 41.36, df = (3,137), p < 0.001. The univariate F tests showed there was a significant difference between attributions of Feel Anger F = 97.79, df = (1,142), p = < 0.001, and Feel Happiness F = 101.16, df = (1,142), p = < 0.001 with respect to function. No differences were detected for Belief. No main or interaction effects were detected for Belief, Feel Anger, or Feel Happiness with respect to embodiment. We also conducted an additional alternative analysis. MannWhitney U test also detected significant differences in Feel Angry within both Embodied U(70) = 178.00, Z = -5.29, p < 0.001, and Disembodied U(71) = 191.00, Z = -5.26, p < 0.001 conditions by function. Consciousness Disembodied | 16 a crucial factor that cues phenomenal state attribution is whether or not an entity has a certain kind of biological body, then participants should be more likely to disagree that Bob feels anger or happiness when he exists only as an “invisible force or a spiritual presence” as compared to when his physical body is made salient. But what we saw was that whether or not Bob had or lacked a physical body in the various conditions of the experiment (Fig.1: solid vs. patterned bars) seemed to play no role in people’s willingness to ascribe these mental states to Bob at the end of the story. On the other hand, functional cues (Fig.1: dark vs. light bars) played a very large role in whether or not people agreed that Bob felt happy or angry. When Bob’s goal was to “cause Henry to hate his mom” people were much more likely to agree that Bob feels angry as he places the pictures rather than happy for Melissa for beginning a new relationship. In fact, they attributed this experiential state at roughly the same level as they attributed the intentional state about belief.10 Lastly, we observe the opposite pattern when Bob wishes to “make Henry incredibly happy.” In such conditions, people were much more likely to agree that Bob feels happy rather than angry with Melissa for beginning a new relationship. 11 10 11 Feel Anger Disembodied-Anger (M = 6.11, SD = 1.19), Disembodied-Happy (M = 3.58, SD = 1.82), Embodied-Anger (M = 6.06, SD = 0.91), EmbodiedHappy (M = 3.75, SD = 1.71). Feel Happiness Disembodied-Anger (M = 1.92, SD = 1.65), Disembodied-Happy (M = 4.31, SD = 1.77), Embodied-Anger (M = 1.43, SD = 0.70), Embodied-Happy (M = 4.14, SD = 1.68). 17 | Buckwalter and Phelan 3.2. Experiment 2: Disembodied Ghost Lacking Intentional Objects The findings from Experiment 1 begin to motivate the following conclusions. First, when it comes to entities like disembodied souls, having or lacking a human biological body is not utilized as an important cue when attributing phenomenal consciousness. In fact, this information seemed to play no role in people’s judgments. Second, participants strongly agreed by comparison that certain entities—embodied or not—can have experiential states when provided with appropriate functional information. Regarding the first point however, one immediate objection surfaces. Prior work in experimental philosophy of mind has suggested that participants’ agreement with phenomenal state attributions are highly sensitive to the amount of contextual information given within experimental probes. Specifically, Arico (2010) found that attributions of phenomenal states to groups that specified an intentional object (e.g., “Canada’s Travel Bureau is experiencing a sudden urge to pursue internet advertising”) were deemed significantly more acceptable than attributions of phenomenal states to groups that lacked an intentional object (e.g. “Canada’s Travel Bureau is experiencing a sudden urge”). In fact, people were much less likely to agree that groups could have a series of phenomenal states when intentional clauses were absent. Thus Arico suggested that that the inclusion of an intentional object in experimental probes provides contextual information that can bias phenomenal ascriptions. Consciousness Disembodied | 18 Perhaps a similar effect could explain the high ascriptions of phenomenal states to disembodied entities in Experiment 1. It could be that participants attributed emotional states to Bob because the probes that were used included intentional clauses, (e.g. “he feels angry at Melissa for beginning a new relationship” vs. “he feels angry”). These clauses might have served to bias disembodied phenomenal state attributions. We conducted our second experiment to rule out this possibility of bias in the probe design. 3.2.1. Methods Participants in Experiment 2 (N=147, 53 female, median age = 32) where presented with the same stimulus material combinations as participants in Experiment 1. However after seeing the vignettes, they were asked to rate their agreement with the following three sentences.12 These sentences were adjusted to account for the worries above by removing the intentional objects from the probe, thereby limiting the potentially biasing contextual information presented: Intention. Bob intends to move the pictures into place. Feel Anger No Object. As he moves the pictures into place, Bob feels angry. Feel Happiness No Object. As he moves the pictures into place, Bob feels happy. Responses were collected on the same seven-item scale anchored 12 For the sake of uniformity in removing intentional objects, we switched the intentional state tested in Experiment 2 from believes to intends. 19 | Buckwalter and Phelan with positive and negative agreement terms. 3.2.2. Results and Discussion We made two predictions. Our first prediction was that we would replicate each of the results uncovered in Experiment 1. Our second prediction was that the absence of the intentional phrases and potentially biasing contextual information in the phenomenal state probes in Experiment 2 (“he feels angry” vs. “he feels angry at Melissa for beginning a new relationship”) would not result in lower rates of phenomenal state ascription to the disembodied entities in the story. Both of these predictions were borne out. First, Experiment 2 replicated each effect found in Experiment 1. 13 Once again, participants overwhelmingly attributed Intention across the board.14 Functional cues continued to play a major role in people’s judgments. Participants were much more likely to agree with Feel Anger No Object in the ANGER condition, and Feel Happiness No Object 13 14 A 2 (Embodiment) x 2 (Function) MANOVA was used to compare the influence that embodiment and function had on the intentional and experiential states: intentionality, anger, and happiness. The multivariate result was significant only for function, Pillai’s Trace = 0.308, F = 20.94, df = (3,141), p < 0.001. The univariate F tests showed there was a significant difference between attributions of Feel Anger F = 60.54, df = (1,146), p = < 0.001, and Feel Happiness F = 14.34, df = (1,146), p = < 0.001 with respect to function. No differences were detected for Intention. And, no main or interaction effects were detected for Intention, Feel Anger No Object, or Feel Happiness No Object with respect to embodiment. Intention Disembodied-Anger (M = 6.50, SD = 0.83), Disembodied-Happy (M = 6.62, SD = 0.49), Embodied-Anger (M = 6.46, SD = 0.56), EmbodiedHappy (M = 6.38, SD = 0.59). Consciousness Disembodied | 20 in the HAPPY condition.15 And lastly, embodiment again seemed to play no role in cuing phenomenal state ascription. Responses in EMBODIMENT and DISEMBODIMENT were nearly indistinguishable. These results can be seen in Figure 2: Figure 2. Mean agreement with mental state attribution in each condition grouped by mental state probe. All scales ran 1-7. Thus Experiment 2 suggests that Arico’s results for contextual information bias for group ascriptions do not extend to phenomenal state ascriptions to disembodied entities such as ghosts. In fact, when the intentional object of the phenomenal state probes are 15 Feel Anger No Object Disembodied-Anger (M = 5.50, SD = 1.33), Disembodied-Happy (M = 3.49, SD = 1.88), Embodied-Anger (M = 5.46, SD = 1.44), Embodied-Happy (M = 3.41, SD = 1.62). Feel Happiness No Object Disembodied-Anger (M = 4.18, SD = 1.72), Disembodied-Happy (M = 5.27, SD = 1.28), Embodied-Anger (M = 3.97, SD = 1.76), Embodied-Happy (M = 4.84, SD = 1.44). 21 | Buckwalter and Phelan dropped in Experiment 2, we again see a clear demonstration of the role that functional information is playing in people’s judgments to these entities. 3.3. Experiment 3: Eternally Disembodied Spirits Taken together, Experiments 1-2 directly challenge the embodiment hypothesis. But one worry about the entities in these experiments is that participants might be conceiving of them as nearly embodied. After all, Bob was not always disembodied; he was until very recently a normal human. So perhaps temporal proximity to embodiment affects people’s judgments about the states they attribute to Bob. It could be that there are specific norms related to the genre of ghost stories such that the ghosts of the recently deceased are attributed phenomenal states.16 To rule out these possibilities we conducted Experiment 3 to see if we could replicate the previous results for spirits that are eternally disembodied. 3.3.1. Methods Experiment 3 mirrored the same between-subjects multifactorial design in Experiments 1-2. Participants (N=118, 41 female, median age = 30) were presented with cases designed to study the effect of embodiment and functional cues on mental state attribution, this time using an entity that was more purely disembodied. For roughly half of the participants, the story started like this: 16 We thank Aaron Meskin for discussion on this point. Consciousness Disembodied | 22 [EMBODIED] Fintan is a very private person. He has little connection with the outside world—no computer, no phone, no car. Instead, he hunts or grows his own food with his bare hands. Though he has no money with which to buy tools, Fintan can make many useful objects with the things he finds around him. The other half saw a story that began as follows: [DISEMBODIED] Fintan is a nature spirit. He has no form at all —no head, no legs, no arms. Instead, he has always existed as a kind of invisible force or a spiritual presence. Though he has no limbs with which to touch physical objects, Fintan can make objects move without touching them, by floating them through the air. Both groups then saw the story continue: For many years, Fintan has lived in Dirk’s Wood beside the Mangahala River. He values the beautiful crystal waters and quiet solitude of the Mangahala above everything else. Recently however, construction has started on the Mangahala Golf Course and Retirement Community. Loggers have begun cutting down segments of Dirk’s Wood to accommodate the project, polluting the entire area. Fintan decides that the only way to stop the destruction of his home is to cause their trucks and chainsaws to break in any way he can. And when the loggers bring in more equipment, he breaks that too. Lastly, to manipulate the functional information specified, parti- 23 | Buckwalter and Phelan cipants saw one of two conclusions to the story: [SAD] But the construction company won’t give in. Realizing that there is nothing he can do to stop the loggers, Fintan leaves Dirk’s Wood. He must now find another place to call home. [HAPPY] Eventually the construction company gives in. Realizing that he has stopped the loggers, Fintan returns to Dirk’s Wood. The place he calls home is now safe. All participants were then asked the following three questions: Comprehension. In the story above, Fintan is: [A human being/A nature spirit with no physical body] Feel Sadness. At the end of the story, Fintan feels sad. Feel Happiness. At the end of the story, Fintan feels happy. Phenomenal state ascription was collected on the same seven-item agreement scale used in Experiments 1-2. 3.3.2. Results and Discussion We made three predictions in Experiment 3. First, we predicted a strong effect for function, whereby people will be much more likely to agree with Feel Sadness rather than Feel Happiness for SAD, and Feel Happiness rather than Feel Sadness for HAPPY. Second, we predicted that embodiment would continue to play no role in people’s judgments, even when the object of attribution, Fintan, is a nature spirit that has never occupied a physical body. And thirdly, we predicted that we would replicate the earlier finding in Experiment 2, that the absence of intentional object clauses does not preclude phenomenal state attribution. Consciousness Disembodied | 24 Again all of these predictions were borne out. We found that function made a very large difference to phenomenal state attribution in these cases.17 People only attributed Feel Happiness or Feel Sadness in HAPPY and SAD, respectively.18 Attribution between EMBODIED and DISEMBODIED conditions was indistinguishable. And lastly, these results again persisted despite using phenomenal state probes lacking intentional objects. These findings are displayed in Figure 3 below: 17 18 A 2 (Embodiment) x 2 (Function) MANOVA was used to compare the influence that embodiment and function had on the experiential states anger and happiness. The multivariate result was significant only for function, Pillai’s Trace = 0.838, F = 284.96, df = (2,110), p < 0.001. The univariate F tests showed there was a significant difference between attributions of Feel Sadness F = 374.93, df = (1,114), p = < 0.001, and Feel Happiness F = 542.54, df = (1,114), p = < 0.001 with respect to function. No main or interaction effects were detected for Feel Sadness or Feel Happiness with respect to embodiment. Three participants were removed for failing Comprehension. Feel Sadness Disembodied-Sad (M = 6.00, SD = 1.29), Disembodied-Happy (M = 2.11, SD = 1.40), Embodied-Sad (M = 6.14, SD = 0.79), Embodied-Happy (M = 1.89, SD = 0.92). Feel Happiness Disembodied-Sad (M = 1.57, SD = 0.97), Disembodied-Happy (M = 5.64, SD = 1.19), Embodied-Sad (M = 1.66, SD = 0.77), Embodied-Happy (M = 5.96, SD = 0.88). 25 | Buckwalter and Phelan Figure 3. Mean agreement with mental state attribution in each condition grouped by mental state probe. All scales ran 1-7. 3.4. Experiment 4: Eternally Disembodied Spirits—An Alternative Measure One worry about Experiments 1-3 is that they all use the same basic technique for collecting phenomenal state attributions, in which participants were asked two questions about states of opposite valence. But perhaps presenting these two questions together created some undue pressure to ascribe phenomenal states. With this worry in mind, we conducted Experiment 4 using a different measure for state attribution based on confidence judgments. 3.4.1. Methods Participants in Experiment 4 (N=120, 37 female, median age = 28) were presented with the same stimulus materials as Experiment 3. However after seeing the materials, they were asked a different set of questions: Attitude Ascription. Which do you think best describes Fintan at the end of the story? [Fintan feels sad/Fintan feels happy] Attitude Confidence. How confident are you with the answer you gave to the previous question? Participants answered Attitude Ascription with dichotomous answer choices above. They answered Attitude Confidence on a sevenitem scale where “1” was anchored with “Not at all Confident” and Consciousness Disembodied | 26 “7” was anchored with “Extremely Confident”. Attitude Ascription was then recoded (“-1” for feels sad, and “+1” for feels happy) and multiplied by Attitude Confidence to create a combined ascription/confidence score (ranging from -7 to +7) for each of the entities in the various combinations of cases. 3.4.2. Results and Discussion We predicted that this alternative measuring technique in Experiment 4 would still result in the same basic findings as seen in Experiment 3. And that is exactly what we found. Participants still relied on functional information as the crucial cue for ascribing phenomenal states—with total indifference to embodiment.19 Mean combined scores (Attribution x Confidence Rating) were significantly lower when Fintan fails to save his home, and significantly higher when Fintan succeeds in defeating the loggers. 20 These results are shown in Figure 4: 19 20 A 2 (Function) x 2 (Embodiment) between-subjects analysis of variance reveals a main effect for the factor of Function, F (1, 119) = 839.08, p < 0.001. No other effects were detected. Combined Measure Disembodied-Sad (M = -6.41, SD = 1.05), Disembodied-Happy (M = 5.34, SD = 2.47), Embodied-Sad (M = -6.60, SD = 0.81), Embodied-Happy (M = 4.81, SD = 3.25). 27 | Buckwalter and Phelan Figure 4. Mean combined score (Attitude Ascription x Attitude Confidence) for each type of entity grouped by function. Scores run from (-7) to (7). 3.5. Experiment 5: Spirits, Groups, and Humans—Explicit Emotional Comparisons The previous experiments appear to demonstrate that people often ascribe emotional states to disembodied entities without hesitation —so long as the appropriate functional cues are present. But then again, how can we be sure that participants are applying the phrase “feels sad” to a disembodied entity as they would to a normal human? Perhaps people merely say that the spirit feels sad, but mean something different than what they mean when they say a human being is sad. In other words, they might not literally attribute the state of sadness to a sprit in the same way they do to a normal human being. In that case, let's say they make an anti-realist ascription. To ensure that participants are literally ascribing phenomenal states in both cases, we need evidence that when people ascribe emotional states to ghosts and spirits, they mean to attribute the same emotional states they attribute to other human beings when they make similar ascriptions. In other words, we need evidence of realist ascriptions. In our fifth study, we set out to provide such evidence. Consciousness Disembodied | 28 As recent experimental work on quantity implicatures (in addition to other work in experimental pragmatics) demonstrates, it is often very difficult to experimentally uncover what people mean by (or how they interpret) particular sentences. 21 However, our task is at least somewhat less daunting since we do not need to uncover what people ultimately mean when they say a spirit is sad. We simply need to demonstrate that people generally mean the same thing by “sad” when they say, for instance, “a spirit is sad” as they do when they say “a person is sad”. There may be numerous ways of examining this question. But one straightforward way is just to ask people to evaluate their mental state ascriptions comparatively. In other words, we could simply ask those who ascribed emotional states to the spirits how similar the emotional states they meant to attribute were to the emotional states they would attribute to a normal person. Of course, we would expect some variance in individual responses to this question, so we would need to compare responses to a similar question asked of those who more or less agreed with emotional state ascriptions to the human character in our stories as well. And since we were predicting no difference between people’s interpretations of emotion words for the spirit or the man, we would also need some other entity to serve as a control, some entity to which people are willing to ascribe emotional states at the verbal level, but to which they do not really mean to attribute exactly the 21 See Noveck and Reboul (2008) for a useful review. 29 | Buckwalter and Phelan same emotional states they attribute to normal persons. For this, we turn again to prior work in the experimental philosophy of mind on group ascriptions. Specifically, Phelan et al. (2012) found that people often offer anti-realist phenomenal state ascriptions to group entities (e.g. the Boeing Corporation).22 Thus group entities seem like the perfect control to use in Experiment 5 when checking for realist ascriptions. Recall the Microsoft Corporation example in Section 1. According to the embodiment hypothesis, people should be thinking about group agents in the same way that they are thinking about disembodied spirits. That is, people should be hesitant to make realist phenomenal ascriptions to both sorts of entities because they lack the right kind of body. So in what follows, we reexamine participants' judgments by asking for explicit comparisons between different emotional states of humans, disembodied spirits, and group entities. 3.5.1. Methods Participants (N=194, 75 female, median age = 26) read vignettes similar to those used in Experiments 3-4. Each vignette began with the introduction of a protagonist that was either a spirit, a human, or a group: [SPIRIT] Fintan is a nature spirit who strives to protect local forests and rivers. He has no form at all—no head, no legs, no 22 Phelan et al. (2012) argue that phenomenal state ascriptions to groups are often distributivist, or that people ascribe states to individual group members rather than to the group as a whole over and above its members. Consciousness Disembodied | 30 arms. Instead, he has always existed as a kind of invisible force or a spiritual presence. Though he has no limbs with which to touch physical objects, Fintan can make objects move without touching them, by floating them through the air. He uses his spiritual abilities to bring an active approach to nature preservation. [HUMAN] Fintan is an individual who strives to protect local forests and rivers. Through hard work and tireless efforts, Fintan works to protect natural areas from development. Though he has little money with which to support his cause, Fintan exploits his own significant technical skills to bring an active approach to nature preservation. [GROUP] FINTAN is an organization set up to protect local forests and rivers. Through charitable donations and the efforts of group members, FINTAN works to protect natural areas from development. Not only does FINTAN support conservation legislation, it also exploits the technical skills of members to bring an active approach to nature preservation. All participants then read a short description of the character’s struggle against a development project, for the spirit it read as follows (with only necessary changes to the character made across other vignettes): For many years, Fintan has worked to protect Dirk’s Wood beside the Mangahala River. The spirit values the beautiful crystal waters and quite solitude of the Mangahala above 31 | Buckwalter and Phelan everything else. Recently however, construction has started on the Mangahala Golf Course and Retirement Community. Loggers have begun cutting down segments of Dirk’s Wood to accommodate the project, polluting the entire area. After an extended struggle, Fintan decides that the only way to stop the destruction of the woods is to cause the loggers' trucks and chainsaws to break in any way possible. When the loggers bring in more equipment, Fintan breaks those too. Lastly, participants were presented with one of two possible endings to the story, where Fintan is either successful or unsuccessful at thwarting the logger's effort: [HAPPY] Eventually the construction company gives in. Realizing that the loggers have been stopped, Fintan celebrates the preservation of Dirk's Wood. [SAD] But the construction company won’t give in. Realizing that nothing can be done to stop the loggers, Fintan gives up on Dirk’s Wood. All participants were then asked the following two questions: Comprehension. In the story above, Fintan is: [A group/A nature spirit with no physical body/A human being] Emotional State Attribution. At the end of the story, Fintan feels [sad/happy]. Finally, those participants who answered Emotional State Attribution with “somewhat agree”, “agree”, or “strongly agree” proceeded to the additional follow-up question designed to measure realist Consciousness Disembodied | 32 ascriptions: Comparison. Consider [Fintan the spirit/Fintan the human being/FINTAN the group (over and above the people that constitute it)]. When you say that this [spirit/man/group] feels [sad/happy], how similar is the feeling of [sadness/happiness] to that of a normal person? Participants responded to this question by rating their level of agreement on a seven-item scale, running from ‘Not at all Similar’ to ‘Exactly the Same.’ 3.5.2. Results and Discussion Given the results of Experiment 1-4, we expected high Emotional state attribution in HAPPY and SAD. And that is exactly what we found. The vast majority of participants (over 80% per condition) signalled at least some agreement with Emotional State Attribution.23 We now move on to an analysis of Comparison for those participants (N=174) agreeing with Emotional State Attribution. If the ascriptions people make to the spirit are realist ascriptions—that is, if people mean the same thing when they attribute emotional states 23 Percent agreement with the emotional state ascription per condition: ‘Sad Spirit’ = 87.9%; ‘Sad Group’ = 81.8%; ‘Sad Man’ = 93.9%; ‘Happy Spirit’ = 93.9%; ‘Happy Group’ = 90.6%; ‘Happy Man’ = 96.7%. Overall, people were somewhat more likely to ascribe happiness (M=6.22) than sadness (M=5.79). A 2 (Emotional state) x 3 (Entity Type) between-subjects analysis of variance reveals a main effect for the factor emotional state, F (1, 193) = 839.08, p < 0.05. No other effects were detected. We set this result aside. 33 | Buckwalter and Phelan to spirits as they do when they attribute emotional states to people —then we would expect (1) high Comparison scores in both HUMAN and SPIRIT (2) with no significant differences between scores in these conditions. And since prior work has suggested that phenomenal state ascriptions to groups are non-realist ascriptions, we would expect (3) significantly lower Comparison scores in GROUP than in both HUMAN and SPIRIT. Again, this is exactly what we found. Despite high scores, there was no significant difference in Comparison between HUMAN and SPIRIT. Participants indicated that they generally mean the same thing by “feeling happy” or “feeling sad” when directed towards a disembodied spirit or a human being. And consistent with prior findings, we also found that Comparison judgments were significantly lower in GROUP than they were for both HUMAN and SPIRIT.24 Results for Comparison are shown in Figure 5 below: 24 A 2 (Emotional State) x 3 (Entity Type) between-subjects analysis of variance reveals a main effect for the factor Entity Type, F (2, 173) = 10.29 p < 0.001. No other effects were detected. A Tukey HSD test revealed significant differences for people’s interpretations of emotional state attributions between GROUP and both SPIRIT (p = 0.001) and HUMAN (p = 0.001). However, no significant difference emerged for HUMAN and SPIRIT (p = 0.897). Six participants were removed for failing Comprehension. Consciousness Disembodied | 34 Figure 5. Mean Comparison judgment for emotional attribution grouped by entity type. All scales ran from 1 to 7. These results suggest that people think of the emotional states they attribute to disembodied entities in the same way as they think of the emotional states they attribute to human beings. In other words, this is evidence that they think these states are similar to the emotional states of normal people. And they think these states are somewhat dissimilar from the emotional states they attribute to groups. Lastly, recall that the embodiment hypothesis predicted that people would be thinking about groups and disembodied spirits in the same way. But it turns out that we see very different results when using our realist measure for comparing ascriptions to these two types of entities. 35 | Buckwalter and Phelan 4. Feeling Beyond Embodiment Our experiments suggest that people are perfectly willing to ascribe emotional states to disembodied entities (ghosts and spirits). Though we think more experiments need to be conducted pursuing the question of realist ascription, we think that these results are a promising first step towards the conclusion that findings across Experiments 1-5 constitute strong evidence against the disembodiment view. It appears that people really do think that under the right conditions, disembodied entities can have the same kinds of emotional states as human beings. What’s more, the data from study 5 suggest that people think of emotional state ascriptions to disembodied entities in the same way as they think of emotional state ascriptions to human beings. These results also corroborate previous research by Buckwalter & Phelan on the important—otherworldly even—role that functional information has on the ascription of phenomenal states to diverse sorts of entities. Completely independently of any influence for body, information about perceptual stimuli, behavioral responses, and other mental states is accompanied by strong attribution of phenomenal states, including, as discussed here, emotional states. The simplest explanation of this fact is to suppose that folk psychology actually identifies phenomenal states with functional roles. Be that as it may, we conclude that when it comes to the psychological factors that cue people's actual attributions of phenomenal states to ghosts, perhaps the only apparition here is the em- Consciousness Disembodied | 36 bodiment hypothesis itself. Returning now to the “Nation of China” thought experiment, it could be that Block's basic insight about what is ultimately required for phenomenal consciousness is still more or less on the right track. After all, the experiments we conducted only speak to the principles of folk psychology that guide ordinary ascriptions of phenomenal states. They don't rule out the metaphysical possibility that cognition requires some sort of embodiment. While this remains a possibility, we would only note that part of the argument for this metaphysical picture of the mind was motivated by the intuition that “China-Brain” does not have mental states in the first place. But if our results for disembodied entities are shown to be sufficiently general, this intuition may not be widely shared. REFERENCES Arico, A. 2010. “Folk Psychology, Consciousness, and Context Effects.” Review of Philosophy and Psychology, 1(3): 371-393. Bering, J. M. 2002. “Intuitive conceptions of dead agents’ minds: The natural foundations of afterlife beliefs as phenomenological boundary,” Journal of Cognition and Culture, 2, 263-308. Bering, J. M. 2011. The God Instinct: The Psychology of Souls, Destiny, and the Meaning of Life. London: Nicholas Brealey Publishing. 37 | Buckwalter and Phelan Bering, J. and Bjorklund, D. 2004. “The Natural Emergence of Afterlife Reasoning as a Developmental Regularity.” Developmental Psychology, 40, 217-33. Bering, J., HernZndez-Blasi, C., and Bjorklund, D. 2005. “The Development of “Afterlife” Beliefs in Religiously and Secularly Schooled Children,” British Journal of Developmental Psychology. Block, N. 1978. “Troubles with Functionalism” in W. Savage, ed., Perception and Cognition: Issues in the Foundations of Psychology, Minnesota Studies in the Philosophy of Science, vol. IX, pp. 261-326. Buckwalter, W., and Phelan, M. 2012. Function and Feeling Machines: A Defense of the Philosophical Conception of Subjective Experience. Philosophical Studies. DOI 10.1007/s11098-012-0039-9. Dennett, D. C. 1991. Consciousness Explained. Boston: Little, Brown and Company Gray, K., & Wegner, D. M. 2010. Blaming God for Our Pain: Human Suffering and the Divine Mind. Pers Soc Psychol Rev, 7-16. Gray, H., Gray, K., and Wegner, D. M. 2007. “Dimensions of mind perception.” Science, 315, 619. Gray, H., Knobe, J., Sheskin, M., Bloom, P., & Barrett, L., B. 2011. “More Than a Body: Mind Perception and the Nature of Objectification” Journal Psychology,101,6, 1207–1220. of Personality and Social Consciousness Disembodied | 38 Heider, F. and Simmel M. 1944. “An experimental study of apparent behavior.” American Journal of Psychology 57, 243– 259. Knobe, J. 2008. “Can a robot, an insect, or God be aware?” Scientific American: Mind. Knobe, J. and Prinz, J. 2008. “Intuitions about consciousness: Experimental studies.” Phenomenology and the Cognitive Sciences, 7, 67–83. Nichols, S. 2007. “Imagination and Immortality: thinking of me.” Synthese, 159(2): 216-233. Noveck, I. A. and Reboul, A. 2008. “Experimental Pragmatics: A Gricean turn in the study of language.” Trends in Cognitive Science, 12(11), 425-31. Phelan, M., Arico, A., and Nichols, S. 2012. “Thinking things and feeling things: On an alleged discontinuity in the folk metaphysics of mind.” Phenomenology and the Cognitive Sciences. DOI: 10.1007/s11097-012-9278-7 Phelan, M. and Buckwalter, W. Forthcoming. “Analytic Functionalism and Mental State Attribution.” Philosophical Topics. Searle, J. 1980. “Minds, Brains, and Programs,” Behavioral and Brain Sciences, (3): 417-458.