Framework for Investigating Animal Consciousness
© Springer-Verlag Berlin Heidelberg 2014Grace Lee, Judy Illes and Frauke Ohl (eds.)Ethical Issues in Behavioral NeuroscienceCurrent Topics in Behavioral Neurosciences1910.1007/7854_2014_278
A Framework for Investigating Animal Consciousness
Philosophy Department, Pennsylvania State University, 244 Sparks Building, University Park, PA 16802, USA
Center for Brain, Behavior and Cognition, Departments of Ecosystem Science and Management, and Biology, Pennsylvania State University, 410 Forest Resources Building, University Park, PA 16802, USA
Paula Droege (Corresponding author)
Victoria A. Braithwaite
An assessment of consciousness in nonverbal animals requires a framework for research that extends testing methods beyond subjective report. This chapter proposes a working definition of consciousness in terms of temporal representation that provides the critical link between internal phenomenology and external behavior and neural structure. Our claim is that consciousness represents the present moment as distinct from the past and the future in order to flexibly respond to stimuli. We discuss behavioral and neural evidence that indicates the capacity for both flexible response and temporal representation, and we illustrate these capacities in fish, a taxonomic group that challenges human intuitions about consciousness.
KeywordsAnimalsBehaviorConsciousnessFishFlexibilityGoal-directedPainTemporal representationTelencephalonTool use
consciousness defined in terms of teleosemantic theory of temporal representation,
temporal representation links phenomenology, behavior, and neural structure,
four forms of evidence indicate temporal representation and flexible response,
three neural structures essential,
capacities of fish considered as test of framework.
The question is not, Can they reason, nor Can they talk, but Can they suffer? Bentham (1789).
This oft-quoted statement succinctly captures the central insight of utilitarian ethics: the capacity to feel pain ought to be the primary marker of moral concern rather than the capacity for abstract reasoning.1 Rationality may be necessary to be a moral agent, but it is not required to be a moral subject. The force of this insight goes well beyond utilitarianism and strikes at the heart of an ethical relationship to the world. A crying infant compels care, an injured pet elicits sympathy. In humans, the evolutionarily adaptive capacity for empathy is so well developed that viewers of animated geometric figures will interpret them as agents feeling fear, love, and anger (Heider and Simmel 1944). Of course, animated circles do not suffer, which raises the problem of gauging appropriate ethical response. On one hand, concern for suffering requires that we minimize pain in all cases, but on the other hand, the tendency to overextend this empathetic concern means we need some measures other than our own moral feelings to determine which animals do in fact suffer.
To begin with, we can distinguish three elements in human suffering : sensory responsiveness to noxious stimuli, conscious hurtfulness, and self-conscious understanding of the pain’s significance.2 While each of these elements raises difficult research issues, this chapter will focus on conscious hurtfulness as the sin qua non of suffering . Conscious pain is the element in human suffering that we attribute to infants and pets and circles; it is the element that calls for an ethical response.
To determine which animals suffer, then, we need to determine which animals are conscious. In the following, we propose a working definition of consciousness in terms of temporal representation that links internal phenomenology with behavioral function and neural structure. It is worth emphasizing that the proposal offers a working definition of consciousness . Temporal representation is a necessary feature of consciousness , we claim, and is sufficient to distinguish conscious from unconscious representations . For a complete account, more must be said about what conditions are necessary for something to be a “representation,” among other things.3 As a working definition, the test of its value is its usefulness in providing an explanation that accounts for the available data. Phenomenological experience forms the foundational data; it is the data that characterizes consciousness from the subjective perspective. Comparative behavioral and neuropsychological evidence characterize consciousness from the objective perspective. A working definition of consciousness is useful to the extent that it brings these two perspectives into alignment and offers testable hypotheses for the future research. We are particularly interested in the research potential of the present proposal and hope the debate about animal consciousness sparks refinement of the theory as well as methodological recommendations for effective tests. If we have more effective means of knowing when animals are suffering , we can more effectively decide what moral response is appropriate. Ideally, the science of consciousness can provide a useful guide for our moral intuitions.
1 Defining Consciousness
The first, and probably most difficult task in this project is to set out a workable operational definition of consciousness . The point here is not to set out necessary and sufficient conditions of the sort proposed in conceptual analysis. Our goal is to identify a tractable target for scientific research, to map the way a creature experiences the world to the world it is experiencing. For as many years as philosophers have been writing about this topic, the best description of the way a creature experiences the environment around it has been the unhelpful locution “what it’s like” to sense red or feel pain (Nagel 1974; Block 1995; Chalmers 1997; Lycan 1996). Though vague, the phrase resonates because it points to the elusive quality that distinguishes an awake state that is responsive to sensory stimuli from fully unconscious states such as dreamless sleep. When unconscious, there is clearly nothing it is like to sense red or feel pain. Similarly, there is nothing it is like to be an object incapable of consciousness , like a table or a laptop computer. For us, by contrast, waking life is full of sights and sounds and tastes and smells which combine and shift to form a vibrant, changing world, and there is something it is like to experience that world. Later we will say more to refine this rough contrast into a sufficient working definition. Even in us, awake responses to stimuli are often unconscious. Consider, for example, your current body position. To maintain balance and prevent injury, you need to shift position periodically in response to proprioceptive signals. Until your attention was drawn to your body, chances are that these shifts in position were accomplished unconsciously . There was “nothing it was like” for you to shift your body position.
But “what it’s like” merely points at the target phenomenon without adequately identifying it. One tempting refinement of “what it’s like” would be to assume it is a case of self-consciousness . Of course, the instant we begin to think about the nature of consciousness , we become self-conscious by means of introspection. Consciousness cannot be caught unawares, as it were, and this fact raises several methodological difficulties in designing experiments to test human consciousness . Whenever a researcher asks a subject to report on her conscious experience, she must introspect in order to do so. Consequently the capacity for self-consciousness and consciousness appear inseparable. In the investigation of non-human animal consciousness , however, the separability of consciousness and self-consciousness is more clearly apparent. To be conscious of one’s self requires an ability to distinguish one’s self from other selves.4 More than the simple recognition that one’s body is unique with regard to sensation and action, self-consciousness involves the recognition that one has a unique mind. The representation of a unique mind involves, at minimum, a theory of mind, including an understanding of deception, perceptual processes, and empathetic imagination (Perner 1991; Gopnik 1993; Tomasello 1999; Goldman 2006). Given the sophisticated representational abilities required for self-consciousness , it seems likely that consciousness evolved prior to self-consciousness , and so we should expect there to be creatures capable of consciousness that are not capable of self-consciousness . There may be reasons to claim that consciousness entails self-consciousness , but this necessary link must be argued and not assumed.
So far then, we can say that we are looking for a phenomenon that occurs during awake states that are responsive to sensory stimuli, and there is something it is like to be in those states. This definition is insufficient for a science of consciousness. The problem is that it is unclear how to supplement it without begging the question about the nature of consciousness. If we propose that consciousness is a global workspace (Baars 1988, 1997; Baars and Newnan 1994; Dehaene and Naccache 2001; Dehaene and Changeux 2011), this assumes a particular integration function is constitutive of consciousness. If we suggest that consciousness involves being conscious of our mental states, the assumption is that some form of higher-order state is necessary for consciousness.5 Nonetheless, some starting assumption about the nature of consciousness is necessary to articulate a clear target for scientific explanation.
We believe that the phenomenologist Husserl (1905) offered a useful suggestion when he noted the essential role of time in the structure of consciousness. According to Husserl, consciousness represents the world from the perspective of a now-point. We take the items presented to us by consciousness as before us now. Yet this moment is not instantaneous, it includes a brief span of time from just past (retention) to just ahead (protention). A span is necessary to account for the experience of duration, such as the hum of cicadas or the waving of leaves in the breeze. In other words, an essential aspect of what it is like to sense the world is to sense it as now enduring (Droege 2003, 2009).
The representation of time in consciousness differs crucially from the way representations may vary in relation to time. A creature that can vary its action in accord with the seasons or other natural temporal patterns does not need to represent time, it simply needs to respond to the appropriate temporal cues. Even a perception–action sequence that includes a temporal element utilizes time without representing it. The marine Palolo worm Eunice viridis, for example, times its reproductive cycle to a 2-h period in late fall. The precise synchronization of thousands of these organisms is attributable to a combination of biologically based oscillations (daily, annually, lunar, and tidal).6 As long as action is strictly coded to some sort of fixed cycle, no representation of time is required. Only when a creature might decide between actions must it be able to assess the environment as it is now in order to determine how to proceed in the light of its goals. As we will argue, the evolutionary development of flexible behavior demands the ability to represent time. This psychological connection between flexibility and temporal representation along with the proposed definitional connection between temporal representation and consciousness form the links between private experience and public action necessary for assessing the capacity for consciousness in non-human animals.
This claim is controversial, and thus forms a substantive commitment about the nature of consciousness, just as the claim that consciousness is a global workspace or a higher-order state (Baars 1997, 2003; Rosenthal 2005; Dehaene and Changeux 2011). Nonetheless, this description of consciousness has several advantages over other proposed claims. First, unlike the global workspace, our sense of presence is phenomenologically apparent. The world appears before us as right here, right now. What it is like to be conscious includes the temporal immediacy of experience. Note that the coordination of sensations and thoughts into a representation of presence is consistent with the integration function proposed by global workspace theorists and other neuroscientific theories of consciousness (Edelman and Tononi 2000; Gulick 2004; Tononi 2004, 2008; Tononi and Koch 2008; Dehaene and Changeux 2011). The present proposal puts the integration function into an evolutionary and developmental context to explain why creatures come to integrate information in the particular form that is conscious representation. A second, related advantage is that the identification in nonhuman animals of the ability to represent the present moment offers an intermediate evolutionary step between simple sensory response and more sophisticated higher-order representation. As noted above, self-consciousness requires additional evolutionary pressure to develop beyond the capacity for consciousness.
In the remainder of this chapter, we will consider the behavioral and physical evidence for temporal representation , and we have chosen to focus on fish as a particularly hard case for demonstrating consciousness (Braithwaite 2010). Fundamental differences between fish and humans stretch the argument from analogy to its breaking point, and this relative weakness places more weight on functional and theoretical argument. If we can make a convincing case that fish are conscious, despite our lack of intuitions about what it might be like to be them, then we can be sure that the framework of explanation is supporting the theory rather than mere anthropocentric inference.
In sum, conscious states are those it is like something to have. When we are awake and responsive to sensory stimuli, there is something it is like to experience (some of) those sensations. Furthermore, our conscious sensory experience presents the world as appearing before us now, which includes a brief temporal span of events indicating an enduring environment. While temporal representation is not (yet) universally accepted as a criterion for consciousness, we aim to show that it forms a vital part of a research program to establish a framework for consciousness in non-human animals.
2 Behavioral Function
Behavior is certainly the primary and most potent indicator of mental activity, even if it cannot provide a complete explanation of the mind and its capacities. Dogs, cats, horses, and rabbits behave in ways similar to the ways we behave when excited, fearful, happy, or in pain (Fraser 2009). As a result, it is fairly easy to believe that these creatures have similar mental states. When they are awake and responsive to sensory stimuli, it seems reasonable to assume there is something it is like for them to experience their sensations, i.e., they have conscious states.
To reinforce the earlier claim that consciousness and self-consciousness are separate capacities, note that there is no compelling reason to attribute self-consciousness to these creatures. When the squirrels look to be enjoying their game of chase in the tree, it is difficult to imagine that there is nothing it is like for them to consciously feel the joy of their game. However, it is not at all difficult to imagine that they lack the reflective ability to think of themselves as squirrels or as joyful. They may very well be joyful without reflecting self-consciously on their joy. Our ability to imagine that a creature is conscious without being self-conscious does not prove these capacities are separable, but it does show that the claim of necessary connection must be argued.
The limits of this strategy of imagining what it is like to be a creature become apparent in the borderline cases. Fish exhibit some similar behavior: their avoidance of electric shocks looks like fear, and their brisk interactions with one another looks like playful fun. But is it? How could we be sure these are not simply automated responses to stimuli in the way a thermostat responds to changes in temperature, or the way our knee reflex responds to the tap of a doctor’s rubber hammer?
As the question suggests, we need a way to distinguish automated behavior from conscious response. The ability to flexibly adapt to novel environmental conditions is often taken to signal consciousness (Griffin and Speck 2004; Edelman et al. 2005; Seth et al. 2005; Merker 2005, 2007; Edelman and Seth 2009), but the question is, why? Why should consciousness be connected to flexibility , or indeed to any other behavioral or physiological indicator (Dawkins 2006, 2012)? Here is where the definition of consciousness in terms of the representation of presence can help connect behavior to phenomenology. In order to flexibly respond to a stimulus, a creature must be able to represent the present stimulus environment in relation to its goals. The environment appears in a certain way now, and this differs in various ways from the desired environment. There is a predator on the horizon, or no food is available, or scratchy bristles are all around. Without an ability to distinguish how things are now from how things might otherwise be, there is no way to consider alternative paths from here to there. An animal might flee or approach as a simple response to a stimulus, but the ability to consider an alternative possibility involves an assessment of the best action in pursuit of a goal, and this more sophisticated alternative depends on a representation of the present moment. In other words, a representation of now is essential to flexible behavior as well as being essential to the structure of consciousness from the first-person perspective.7
The methodological value of flexibility as behavioral evidence for consciousness lies in the many ways to demonstrate it: (1) differential response to the environment, (2) adaptation to novel situations, (3) manipulation of the environment to accomplish goals, and (4) explicit representation of absent objects.8 We will consider each form of evidence in terms of how it shows both flexibility and temporal representation . Because all of this evidence demonstrates the same basic capacity, several examples could be used to illustrate more than one form.
As noted earlier, fish are a particularly useful group for this sort of investigation, because they are at the edge of human intuitions about consciousness (Allen 2011). If we can get a framework to guide our thinking and research with fish, then we can make a compelling argument about consciousness in other vertebrates as well, and possibly even invertebrates.
Another reason in favor of studying fish is the wide variety of species and the diverse evolutionary pressures on differential species development. Very closely related species exhibit strikingly different capabilities depending on factors such as the complexity of their environment, density of predators, and availability of resources (Braithwaite 2005). This diversity presents the possibility of tracing correlations in neurobiology, behavior, and environment in much richer ways than are possible in the study of mammals (Kotrschal et al. 1998; Gonzalez-Voyer and Kolm 2010).
Differential response calibrated to relevant differences in a situation is the minimum requirement for flexible behavior; all three of the remaining ways to demonstrate flexibility also show the capacity for differential response. While simple single-celled organisms have the basic ability to react to variations in stimuli, flexibility depends on using information learned in the past to select among different possible actions in the present situation.
Dennett (1995, pp. 373–378) provides a helpful schematic of various ways an organism might be designed to respond successfully to its environment, which he calls the Tower of Generate–and–Test. At the base of the tower are Darwinian creatures. Gene recombination and mutation provides these creatures a selective advantage, leading to the reproduction of the advantageous genetic traits. The level of Skinnerian creatures introduces the element of phenotypic plasticity. The behavior of these creatures can be modified by positive and negative reinforcement. This design system works effectively when a specific response is appropriate given a specific stimulus. When a creature develops the capacity to respond in different ways to a set of stimuli, it becomes a Popperian creature. At this level, a creature no longer simply acts based on past associations; it generalizes on past learning conditions to anticipate which current action will best advance its goals. Associative learning remains the foundation for the generalizations made by Popperian creatures. The critical development is that the associations are no longer one-to-one stimulus–response pairings but become a complex weighting of many-to-many relations. Learning, memory, and crucially, integration of information are necessary to differential response. A creature must be able to determine appropriate behavior based on the assessment of a number of variable factors present in the situation (Merker 2007). Popperian creatures behave flexibly, and so on the proposed account they are conscious.
Consider, for example, the complex symbiotic relationship between the cleaner wrasse Labroides dimidiatus and the client fish that need parasites removed. Cleaners establish territorial stations where they feed on the parasites of clients, interacting with clients over 2,000 times a day (Bshary and Grutter 2005). While clients with large home ranges have a choice of several cleaners, they tend to return to the same cleaners, as much or more than 100 times a day (Bshary and Würth 2001). It is in the interest of cleaners, therefore, to provide a valuable service, so clients will continue to return to their station. The mutual benefit of this relationship is complicated by the preference of cleaner fish to feed on the mucus of client fish instead of the parasites. This they do by literally biting a chunk out of the client’s flesh, an action that causes the client to visibly shudder. When bitten by a cleaner, clients will respond by chasing the cleaner or by leaving the station (Bshary and Grutter 2005). Another complication is that some of the client fish are predators, so they too have interests that both favor and counter cooperation.
The variety of factors in cleaner–client interactions has resulted in surprisingly sophisticated behavior by the cleaner wrasse, and to a certain extent in cleaner gobies that also perform these services (Bshary and Würth 2001; Bshary and Grutter 2005; Danisman et al. 2010). First, clients may punish cleaners by chasing them or leaving the station. This reaction causes cleaners to limit their cheating in order to avoid these penalties (Bshary and Grutter 2005). Second, cleaners offer incentives to clients in the form of tactile stimulation: a cleaner hovers above the client, touching the client’s dorsal fin with its pectoral and pelvic fins (Bshary and Würth 2001). Importantly, cleaners offer tactile stimulation in three sorts of situation: (1) in order to induce clients to stop at the station for a cleaning, (2) differentially to predators over non-predator client fish in order to prevent conflict, and (3) after the cleaner has bitten the client, to prevent the client from chasing or leaving.
A second way to demonstrate flexible behavior is the ability to respond to novel situations by inferring future conditions based on information gathered from the past. This description is loaded with mentally sophisticated terms: inference, past/future conditions, information. To warrant attribution of such cognitively rich abilities requires distinguishing them from the simple weighting of past associations. While past associations necessarily figure in the evaluation of a novel situation, the solution must depend on a combination of features never before experienced simultaneously.
According to the temporal representation theory of consciousness , a representation of the present moment allows the creature to assess the current situation in order to determine which action is the best means toward its goals. If a situation is novel, no merely backward-looking algorithm is sufficient to calculate appropriate behavior. Instead the features of the new situation must be combined. In other words, the creature needs a representation of both feature A and feature B as components of the world now in order to know that they need to be combined. The result of this combination specifies the best action in that situation.
The ability of the male cichlids to respond appropriately to a novel set of competitors provides just the sort of cognitively rich inference needed to demonstrate flexible response. Because the species is aggressively territorial, it is useful for them to gain as much information about potential opponents as possible. Watching aggressive interactions is one source of information utilized by cichlids to determine which opponents are more threatening. In an experiment by Grosenick et al. (2007), cichlids were allowed to watch neighbor fish fight with one another in order to determine how much information they were able to gather from observation. Subject fish were shown fights in the following order: A beat B, B beat C, C beat D, and D beat E. In the critical trials, observer fish were placed between super-winner A and super-loser E and between marginally stronger B and marginally weaker D to see how the cichlid would react to these novel pairs. In both cases, the observer cichlid moves toward the weaker fish to assert its dominance.
Because the observer fish has never seen A and E or B and D fight one another, the response cannot be simple association based on past experience. Information about relative fighting strength needs to be assessed given the current opponents in order to determine the best action. This result is particularly striking when the fish identifies D as weaker than B. In this case, both B and D have lost one fight and won one fight, so neither is distinguishable simply as “winner” or “loser.” Only by recognizing B as, in some sense, “winner over C” and D as “loser to C” can the observer act effectively. In other words, the observer fish utilizes memory to identify the relative strength of the particular individuals in the specific current situation. Representation of the features in the world now is needed to respond appropriately.
A third behavioral indicator of flexibility is the ability to manipulate the environment in pursuit of a goal. Differential response and assessment of novel situations are both reactive forms of behavior, showing a highly adaptive but not necessarily forward-thinking creature. Proactive, problem-solving skills, in contrast, require the ability to represent how a change in the current environment might bring about the desired goal. In other words, a creature must be able to represent the difference between the current situation and the goal situation in order to determine what action would bring about the appropriate change.
The best example of manipulating the environment is tool manufacture. To fashion a tool requires an understanding of how a goal can be achieved by means of unrelated items. By twisting or bending or combining objects, an animal can produce a tool specifically designed for a particular task. This shows an understanding of the relation between the tool and the task prior to the execution of the task. Tool manufacture depends on the ability to plan.