The self construct in cognitive science

The concept of the self remains one of the more enigmatic problems in cognitive science. Many theories of the self and the construction of identity have been explored by psychologists and philosophers throughout history, but a cognitive model of mechanisms that give rise to conceptions of the self has never emerged. It may seem like a daunting task, but recent research has begun to investigate the cognitive construct of the self and things look promising. Interesting data from developmental, pathological, functional imaging, and neuropsychological research has provided some insight regarding the development and disorders of elements that comprise the self. Research in areas such as autobiographical memory, narrative, inference of causality, and mental state representation are intrinsically linked to the study of self in a number of ways. What follows is an overview of historical and recent ideas of note in the investigation of a cognitive scientific model of mechanisms that give rise to different aspects of self; an overview of the developmental course of self; and some exploration of how autobiographical memory, narrative, causal inference, and representations of mental states are significant in pursuing such an investigation.

The self is a complex concept that has been approached from many perspectives throughout history. Many models of 'self' include a variety of sub-selves with separate functions and mechanisms, and variable relationships with mental life and cognition. In 1998, Neisser proposed a model that included five distinct kinds of self knowledge. In this model, a self termed the 'ecological self' was described as the immediate sense of self based on a physical perspective of the environment and self awareness. The 'interpersonal self' mediated behavior when relating and communicating with others. The 'extended self' went beyond the scale of the ecological self to be able to remember the past and intend future actions, and the 'private self' consisted of our ability to engage in a private mental life inaccessible to others. The 'conceptual self' was described as the mental representation of the self, the self that we evaluate, defend, and seek to improve. These multiple and functionally diverse proposed selves are already representative of the obscure nature of the question of self in cognitive science.

Recent philosophical discourses on the self have identified two aspects that must be accounted for in a contemporary discussion of the subject: the immediate sense of the self as an agent, and the self that endures through changes in mental states and over time. The immediate sense of self-awareness and agency roughly approximates Neisser's ecological self, while the coherent and continuous self must contain attributes of Neisser's interpersonal, extended, private, and conceptual self. Many different views have emerged, but this general framework is a useful place to begin when considering this issue.

The self as outlined above can be considered to be composed of the sensations of self-ownership and agency, body centered spatial perspectivity, and perception of a long term unity of beliefs and attitudes(Gallagher 2002). There is some evidence that both of these selves arise from processes in the same neurobiological system, a system loosely distributed in the frontal-lobe and implicated in a wide variety of cognitive tasks. The neuroanatomical data supports a view of the self as a complex phenomenon arising from the interactions of related but distinct cognitive systems.

Philosophy has been instrumental in thinking about the nature of the self, and in cognitive science self is still largely a philosophical issue. Heidegger's work in phenomenology gave rise to a very useful way of thinking about the self that is consistent with some cognitive models. The phenomenologists held that humans perceive a pragmatically structured world of human opportunities and liabilities, in which the self is implicit. This philosophical model is still relevant today, and is basically consistent with many cognitive processes proposed and empirically tried since its conception. This description of the organization of perceptual information as a function of the self is still central to many contemporary accounts of the self. Cognitive scientists have recently postulated models of some of the central mechanisms that give rise to both the immediate and the temporally extended self, as well as their interactions with other concepts and conceptual frameworks.

Research on ownership and agency in both motor action and in cognition has correlated perception of this basic 'self' with mental activity in the prefrontal cortex, SMA, and cerebellum (Fourneret and Jeanerod 1998). Agency is experienced when a person perceives themself as the author of an action; but even during an uninitiated action a sense or perception of ownership over the body in motion arises. Cognitive scientists gained some insight into the mechanisms of experiencing ownership and agency when it appeared that these sensations were selectively deficient in some schizophrenics.

Certain schizophrenic patients experience delusions that their body is under the control of another person or thing, or experience thoughts that they do not perceive as their own in an experience called 'thought insertion'. It is hypothesized that these symptoms are the result in a deficit in the mechanism(s) that give rise to perception of awareness and agency. The self-monitoring mechanisms that underly sensations of self ownership and agency could be the mechanisms deficient in the schizophrenic patients. The researchers hypothesize that a sense of agency arises from anticipatory pre-movement motor commands relating to relevant efferent nerves. An unconscious mechanism compares an efference copy of motor demands with motor intentions and allows for rapid error correction. This mechanism is thought to anticipate the sensory feedback from the motor action and give rise to a sensation of self-agency. If the efference copy is not properly generated, a sense of ownership will persist but the sensation of agency will not occur (Gallagher 2000). Some evidence supports a deficit in schizophrenics' pre-action monitoring, and indicates that their motor-control and ability to perform intended actions remains intact(Frith and Done 1988). Abnormal pre-movement brain potentials have also been observed(Gallagher 2000). This data suggests that the activity of pre-motor areas is critical in experiencing the sense of agency that is the most immediate component of the perceived self.

This model is thought to extend to the phenomena of thought and inner speech, and describes thinking as an action that must be matched to intention in much the same way as motor actions in order to feel self-generated. This would explain thought insertion experiences in schizophrenics, the experience of perceiving thoughts that do not feel self-generated. This self-monitoring and reporting mechanism roughly describes a way in which perception of the self as an agent relative to the immediate environment might arise.

The involvement of pre-motor areas in a cognitive event that is largely perceptual brings to mind recent popular work on the involvement of motor areas in abstract cognitive events, most notably studies on the activation of motor neurons in Chimpanzees. This research describes patterns of activation in motor-areas arising as a result of observing the intentional actions of others. Significantly, these patterns of activation are the same when the animals execute the same or similar actions themselves (Williams et al 2005). This discovery has been applied to the research of the differences in representation of the 'self' and the 'other' in human subjects. This elemental distinction between experience of self relative to other can be considered a part of the immediate awareness of being a distinct entity. A more complex self is implicated in attribution of mental states to others, and the physical implementation of these two kinds of mental representation is an interesting topic with ties to popular theories about motor-neurons.

It has been observed that people are able to infer meaning, predict future actions, and interpret behavior as a result of attributing mental states to others. Two popular views on the subject have emerged, "theory theory" and "simulation theory". "Theory theory" describes the system that attributes mental states as the result of experiential learning that gives rise to a theory of human behavior. "Simulation theory" proposes that people attribute mental states by simulating the experience of the other in order to experience their corresponding mental state (Tirassa et al 2005). The mirror-neuron findings and subsequent functional imaging studies in humans provide support for the simulation theory. So what does this research on representation of the other have to do with the self? The developmental and functional imaging literature indicates that the self/other distinction is central to the concept of self. Through development and into adult life, evaluations of our selves in relation to similarities and differences of others plays a formative role in the construction of self.

Limited self-awareness appears to be an innate mechanism present from birth that develops primarily through interacting with others. A concept of the nature of self begins as immediate spatial and perceptual awareness and a recognition of sameness between self and human others. Infants less than one hour old can imitate conspecific facial gesture in a manner that rules out reflexivity (Meltzoff and Moore 1984). This behavior demonstrates an ability to match gesture that requires the infant to locate and use body parts proprioceptively and recognize the seen face as being of the same kind as it's own; human babies will not imitate non-human actions and gestures (Legerstee 1991). It has been demonstrated that infants can use information from their own action capabilities to understand the actions of others, and throughout development and on into adult life the self continues to be largely a result of interpersonal interaction and self-evaluation in relation to others(Ochsner et al 2005). Functional imaging studies have shown overlapping but distinct regions of neural activity thought to underly representations of self-originating and simulated mental states, and it has been hypothesized that the perception of agency and ownership of self-generated actions and thoughts is essential for distinguishing perceptions of the two. If the same mental events represent a self-generated action and the experience of observing the action of a separate agent, then the qualitative difference between inferring agency and perceiving it internally might be the most salient thing that separates our perception of our own mental states from perceptions of the mental states of others.

The perception of ourselves as the author and subject of our actions may be very important in monitoring our interactions with the physical and social environments we have to navigate. The conceptually complex extended self may also play a self-monitoring role; that of monitoring goal oriented behaviors over time and despite transitions in mental states. This self is extended in time and mediated by narrative, and is postulated to play a role in constructing a continuity between our past and our future. A vague sense of this cohesive mechanism has been postulated since Hume, in 1739, described "a bundle of momentary impressions strung together by imagination." Narrative is thought to aid in the construction of a coherent self-schema, a useful if not perfectly accurate sense of continuous identity. The explicit role of the extended self as outlined above is to represent a continuous self that extends in time, a combination of Neisser's extended and conceived selves.

One of the most commonly investigated components of the temporally extended self is autobiographical memory(AM). AM is a kind of memory not characterized by its mechanisms but by its content; AM contains memories of personally experienced events and abstract personal knowledge such as knowledge of places one has lived and people one has known. AM is thought to be a very central and intricate part of the self that contains information central to constructing the cohesive self. A model of the physical implementation of AM describes the structure of AM and the constructive nature of the interaction between past selves represented in memory and the present, working-self schema.

Memory is thought to represent past experience on multiple levels of specificity, from lifetime periods (e.g. 'when Rover was alive') to general events (e.g. 'my vacation to Florida') to event specific knowledge, or ESK. Recall of ESK preserves something of the phenomenal nature of an experience, such as the warm sensation of Rover in your lap and the sounds of traffic as you drove to Florida. The experience of ESK results in a perceived sense of yourself in the past (Gardiner, Richardson-Klavehn 2000). The intentional recall of an autobiographical memory is put in motion by the intention to remember a specific event. The knowledge structure of memory causes a pattern of activation across levels of informational specificity channeled through the current self's goal structures, resulting in the reconstruction of personal memories consistent with current self-views and goals. Evidence indicates that consciousness has preferential access to memories relevant to personality traits and long term goals (Bakermans-Kranenberg 1993). AM's are typically episodic and reflect singular and important novel experiences, significant emotional experiences, and generalized shemas of frequently occurring events.

Autobiographical memory is thought to be central to identity construction. The coherent narrative of significant personal experiences and self-knowledge represented in AM is an accessible source of the distinctive features and events that comprise the complex representation of the self in time. The self is also thought to play an organizational role in memory, and self-relevant information has been demonstrated to result in more vivid and easily accessible memories. It has also been observed that self-relevant information is preferentially allocated attentional resources and singled out for higher-order processing (Grey et al 2004). Similar attentional and higher-order processing effects have been observed for strongly emotional information, which may be primarily self-relevant. It is hypothesized that self-relevant information is charged with affect in order to organize world and life information relative to goals, obligations, and desires of the coherent self.

This view of AM formation is consistent with reports that vivid autobiographical memories predominantly represent episodes from formative or transitional periods of life, such as moving away to college or choosing a career (Oschner et al 2005). Many representations of the self in AM relate to the self's enduring relationship to specific social groups, for example 'male,' 'Catholic' ect.. Studies indicate that this kind of group affiliation information is sufficiently self-relevant to contribute to the content of AM, and in this way personal and social group histories are aligned. An example of this phenomenon is described in a classic study in which White Americans tended to remember where they were when JFK died and not where they were when MLK died, while the opposite held true for African Americans of the same generation. In this way, AM aids in a construction of the self over time relative to the public history of a significant social group.

AM is central to formation and maintenance of perceptions of self that persist over time, but the self and belief structures shaped by past experiences and AM also shape perception of experiences and memory reconstruction. It is thought that self-schema and self-perception cause memories of past events or attitudes to reflect self consistency rather than veridicality. Neisser's classic study of the reconstructive nature of autobiographical memory is still an excellent example of the overwhelming tendency to consider the past self in terms of the present one. During the Watergate scandal, key witness John Dean's memory of his conversations with the President and role in the scandal were demonstrated to be at odds with evidence collected from audio recordings of the conversations Dean testified about. While his recall of specific episodes was inaccurate, he demonstrated excellent recall of the general meaning of repeated conversations and the chronology of the event. His perception of his past self was determined in large part by his self-image at the time of the trial, and his self reported prediction of the coming disaster was not supported by the content of the tapes. This finding has been frequently replicated, and subjects are often surprised when their memory of past beliefs or attitudes is more positively correlated with their current beliefs and attitudes than to the remembered instance.

Perception of consistency over time is dependent on a number of mechanisms and hypothesized to be a result of the organization of structures that mediate between mind and world. Just as perception of qualia like color and speech is constructed by neural mechanisms that organize external phenomena into useful information, construction of self-percept is guided by lower level mechanisms that help us make sense of the world at the level of actions of intentional agents. Dennett has proposed the idea that narrative is the mechanism through which disorganized sensory information is organized by chronology and causality in the stories we tell ourselves (Dennett 1990). Narrative acts as an organizational tool linked to innate theories of human agency and causality, and we understand our world by telling ourselves causal stories about current and past events.

The left hemisphere is thought to be the primary seat of narrative, and has specifically been implicated as the source of an interpreter of inferred causality. The human tendency to perceive and infer causality is consistent and automatic. An understanding of relations between causes and effects is critical to making sense of our constantly changing physical world, and even greater sensitivity to causal relations is required by the demands of our complex social organization.

Research on mirror neurons in primates has demonstrated that the goal of an intentional action can be inferred from incomplete information. The left hemisphere may be involved in elaborating on available information to generate hypotheses about likely causal explanations for events, including self generated actions. Such causal attributions come naturally and automatically to normal adults, even when they do not appear to contribute to performance. A recent study demonstrated that the left hemisphere in split-brain patients generated a causal explanation for the behavior of the dissociated right hemisphere regardless of the utility of doing so or accessibility to right hemisphere causal mental state information. This finding, and further evidence of the automatic generation of causal explanations by the left hemisphere, is evidence of the inferential nature of causal stories about even self-originating action causation. The left hemisphere of split-brain patients also performed poorly on a visual memory task, and tended to falsely recognize novel objects consistent with common scenes presented for encoding (Gazzaniga 2000). This illustrates the role of narrative in detecting common structures and generalizing. These observations support the constructive nature of causality attribution in personal narrative, and provide evidence that this may be a left-lateralized activity.

The tendency to perceive causality appears to be a foundational principle in the organization of event information, and narrative is the mature form of causal-temporal representation of event information. Narrative comprehension and production rely on the same widely distributed frontal lobe areas that support theory-of-mind and mental state processes generally, as well as working memory and areas thought to be necessary to causal-temporal ordering of information. In addition, AM and most human communication assumes the narrative form in order to represent and communicate a coherent and organized self within experience. Memories of events originally encoded as causal narratives are more easily recalled than individual events outside of a causal context, and the underlying narrative structures of repeated events are stored and then elaborated on in reconstructive memory formation. Construction of a solid causal explanation for events can determine our perception of the significance and tone of personal episodes.

In addition to structuring autobiographical memory, the inclination to attribute events to underlying causes provides the framework for a folk-psychological interpretation of behavior. Folk psychology serves to order a stream of behavioral data by identifying causes such as motive, obligation, sentiment, and personality traits. Both external(situational) and internal(agent-originating) causes are recognized, and trends in behavioral data indicate a systematic difference between self and other causal attribution. People tend to explain their own behavior in terms of external causes acting on the self, and the behavior of others in terms of the personality, and mental states of the agent. The constructive nature of causal inference on narrative allows for this kind of flexibility, and for perceived consistency of self-image (Mar 2004).

Data from the developmental and pathological literature provides evidence that the ability to perceive and attribute causal relations is central to engaging in appropriate social interaction. Behavioral explanations and their flexible interpretations can impact the perception of the self by others and the distribution of responsibility and blame. The ability to detect deceptive or manipulative behaviors is dependent on reliability of inferential strategy. An appropriate perception of internally and externally causal factors could have been highly adaptive in an environment of aggressive social competition. In development, children's causal stories tend to be more situational than those told by adults. Though capable and more than willing to report causation of their actions and experiences by the age of 3, causal stories are typically bereft of self-originating causal mental entities such as needs, goals, and wants until the age of 5. This is attributed to the more complex and temporally extended self that emerges in early AM around the same time, along with the theory-of-mind ability to adopt multiple perspectives (and pass the false-belief task) (Atance and Meltzoff 2005). Conversely, paranoid schizophrenics tend to disproportionately perceive the internal mental states of others as causal, and as a result suffer illusions of persecution(Langdon et al 2005). Somewhere in between is the normal adult behavior of generating causal narratives which allow for natural and rapid interpretations of personal events and interpersonal interactions.

Early development of the self-concept is thought to be strongly related to interpersonal interaction. Infants are extremely sensitive to self/other similarities and differences, and may learn a lot about themselves by understanding their similarity to other human agents. At birth, neonates have an awareness of being relative to physical objects and bodily perception, and an interpersonal self that is shaped through interaction with others, typically adult care-givers. A tendency to imitate and share affect observed on conspecific faces develops quickly and 'social embodiment' information such as postures and facial expression continue to play a central role in social information processing. Within weeks, consciousness of self-originating goals and actions as distinct from the goals and actions of others emerges. At 12 months, there is evidence of a perceived distinction between the emotions of others and the self, and at 18 months infants demonstrate discrimination between their own desires and preferences and those of others, and can ascribe agency. It is very telling that this development is articulated through steps in establishing the self-other boundary.

The developing self is increasingly capable of representing the world relative to personal goals, and interpersonal interaction continues to be a force in self formation and perception. Receptive language skills emerge before language generation, and are related to recall and cohesion. The development of these skills may be a small step in the development of the extended self, and is thought to be related to the development of the ability to interpret the contributions of others in extended discourse, illustrating the natural and automatic capacity for generating cohesive interpretations of incomplete information.

At 4 or 5, children exhibit the ability to represent the self extended in time, and begin to use more qualitative and self-referential language in causal interpretations of events. At the same time the ability to represent mental content develops in complexity to include the possibility of false representations of reality and disparate perspectives of the same event. There is some speculation about the coincident emergence of these two abilities (perception of extended self and mature performance on false belief tasks) being attributable to maturation of the theory-of-mind ability to adopt multiple perspectives. AM emerges around this same time, and a unified sense of self over time and defined by personal narrative begins to develop. Again, the self concept increases in complexity as a function of it's elaborated relationship with the other. The neural pathways that allowed the self to develop through imitating and assimilating to the other now allow an articulated self to simulate internal states and represent their predictive significance relative to the self.

The path to the cognitive scientific self is long and convoluted, and is not made easier by the composite, interactional nature of its component parts. Even the basic distinction of the self from the other turns out to be a vague one. An important step in the functional imaging literature was made by an appropriately indirect and backwards inference about the mechanisms of self-referential thinking. Certain tasks of attribution of intention and mental state reasoning have been correlated with activation patterns closer to baseline than other cognitive tasks. The state of the brain during rest, or the default state of the brain when not faced with a specific task, is used as a baseline in many fMRI and PET studies. The similarity of patterns of activation in mental-state related tasks to patterns of activation in the resting, baseline state caused researchers to speculate as to the content and nature of the resting state and it's relationship to thinking about thinking. It was surmised that some level of self/other/mental state representation or reasoning takes place during the resting state. Further investigation revealed that mental-state reasoning that required more internally directed attention, for example judgments of subjective experience, caused an even lower decrease in activity associated with the resting state. The direction of this correlation indicates that the resting-state is an introspective state, with attention directed towards internal experiences and processes.

Degree of activation appeared to be a function of the relative proportion of internal to externally directed attention to mental-state information, which supports a common neural basis of self/other representation. In fact, self/other relevant cognitive processes generally have been correlated with the same set of neural structures. This set of neural structures has been implicated in face perception, autobiographical memory, self-evaluation, causal-temporal ordering of information, representing the mental states of others, and monitoring, execution and perception of goal oriented actions. Individual observed instances of localized function are as variable as the tasks associated with them, but generally the medio-dorsal frontal cortex, orbito-frontal cortex, and STS bilaterally; the anterior cingulate cortex, temporo parietal junction, pre-motor, and motor cortex of the right hemisphere; and the temporopolar cortex of the left hemisphere are consistently correlated with processing of mental state information. Additionally, the posterior cingulate cortex has been correlated with self-awareness and thinking directed towards the intentions of the self specifically (Harris et al 2005, Decety and Sommerville 2003, Metzinger and Gallese 2003).

Overlapping cognitive architecture, dependent conceptual development, and data from primate research on motor neurons all point to the relationship of motor-pathways and mental state representation as a promising and interesting area for future research. Many intriguing claims have been made about the relationship of motor-pathways to understanding the other, including reports of patients with R hemisphere lesions resulting in denial of hemiplegia that extends to denial of the motor-deficits of other patients(Tirassa 2005 reporting on Ramachandran). The connectedness of the self and the other invites investigators of the self to look to deficits traditionally thought of as social, such as autistic spectrum disorders. Other interesting topics include the assimilation of AM, belief structures, perception and narrative processes into a more comprehensive model of the extended self. Rather than the unwieldy philosophical issue we began with, closer inspection of the nature of the problem of self in cognitive science reveals a difficult but manageable problem in social cognition.