In his 1991 book Origins of the Modern Mind, Merlin Donald presents the evolution of human cognition as a sequence of three transitions in dominant representational systems. Humans progressed from other primates by developing gestural, linguistic, and written storage and thought structures, thereby developing what Donald calls "mimetic," "mythic," and "theoretic" cultures. His approach is interdisciplinary and somewhat speculative -- he combines data from anthropology, archaeology, cognitive psychology, linguistics, and neurobiology to reconstruct the stages of evolutionary human cognitive development. In this review I aim primarily to explore some of Donald's ideas beyond the lurch where he occasionally leaves them.
Donald spends the first quarter of the book presenting an impressive collection of neurological, anthropological, and psychological information about the human brain and cognition and how it differs from those of other primates. This portion of the book culminates in a picture of primate memory, and hence cognition, as episodic. The only representational form of memory available to non-human primates, Donald shows, is the memory of specific events. Primate cognition thus consists in recalling specific past events and categorical judgments and applying them straightforwardly to present circumstances. They are able to use symbols by recalling their use in the past, but they are incapable of representational generation. In primate culture, information like tool use is spread by direct and literal imitation and new ideas are hard to come by.
Donald then proposes that Homo erectus underwent a fundamental cognitive shift that set them apart from all other primates, including Australiopithecus and Homo habilis. Erectus developed sophisticated tools, spread across vast territory and varied climates, and developed "society where cooperation and social coordination of action were central to the species' survival strategy." (p. 163) These advances were made possible not because of a simple increase in total brain mass, but through a radical new cognitive capability -- mimesis. Donald argues that while erectus couldn't speak, they had a generative intentional representational form of communication available in gestures and mime. Mimetic culture allowed our ancestors to model group structure, moving from relationships between individuals to relationships between social roles; establish group norms; voluntarily display emotion; coordinate hunting, allowing some hunters to drive the animal to an appropriate spot and others to attack it; and easily teach skills like tool making. Knowledge was no longer passed purely through genes and memory of life events; erectus could pass hard-won knowledge like "don't eat these berries" directly to friends and family. Learning and reinforcement took the form of direct instruction, reciprocal games, and group ritual, "a collective act in which individuals play different roles" (p. 175).
Donald illustrates several of the ways in which mimesis still permeates modern human cognition. Pre-verbal children, deaf-mutes without sign language, and some people with brain damage are all able to communicate fairly effectively using only mime, gestures, facial expressions, and prosodic vocalizations. Group ritual (and its theatrical descendants) remains a crucial part of present-day hunting-gathering culture and even plays a role in modern western culture. Furthermore, mimetic action is more effective at conveying emotion, maintaining crowd control, and teaching certain tasks than language. Human mimetic skill has thus been retained vestigally, just as the mimetic adaptation retained and encapsulated our episodic memory capacity.
The second transition of human cognition that Donald presents is from mimetic culture to mythic culture. Mimesis could only take hominid culture so far, and language developed to allow for more precise, better developed, and better conveyed communication between individuals. Continuing his careful and critical examination of archaeological, neurological, psychological, and linguistic evidence, Donald states that intraspecies competition provided the selection pressure for the language adaptation. But social communication wasn't the only benefit of language. Just as mimesis offered people a memory and cognitive structure vastly better than that possessed in purely episodic culture, semantic symbolism offered a similar boon to the members of "oral-semiotic" culture. Lexical capacity, the root of semantics, doesn't depend on speech, and we therefore possess "a general-purpose capacity that extends beyond the vocal-auditory pathway" (p. 254).
Donald proposes that myth is the primary function of language in a culture dominated by linguistic cognition, which he therefore labels "mythic culture." Donald suggests that mythic cultures passed (and continue to pass) collective knowledge about survival through a vast mythic heritage, complete with oral lore, totemic art, and mimetic song, dance, and ritual. The fundamental (in terms of importance and function) aspects of linguistics thus reside at the narrative level, rather than at the level of the sentence and proposition, the primary focus of linguists. Although Dunbar's commentary proposes that basic social exchange, not myth, was the primary function of language, the primacy of narrative rather than syntax remains.
Just as the mimetic mental system can access the memories, conclusions, and abilities of the episodic, the linguistic center has access to the other two cognitive modes. We can thus linguistically present (though often not very clearly) information and skills we have learned by mimesis or direct experience. However, the mimetic system, Donald thinks, cannot access the internals of the linguistic system, nor can the episodic system access either of the others. Donald's presentation of Brother John, an epileptic who would lose access to internal and external linguistic skill, throws doubt on this flow -- during his episodic- and mimesis-only spells, he was able to act on information obtained during his lifetime of linguistic learning and interact within linguistically-based culture.
Also present in mythic culture, though appearing much later than Donald places speech, are symbolic pictures like those found in southern European caves. Donald argues that these hunting and fertility images were used "to explore and develop the mythic ideas that were already the governing cognitive constructs of human society" (p. 282). Symbolic art, in Donald's view, is handled by the same cognitive system that handles symbolic language, although he clearly shows that language and visual processing take place in separate places in the brain. Humans would later synthesize symbolic art and symbolic language, the third transition proposed by Donald.
The key new feature in the fourth stage of human cognition is writing. While words and semantic symbolic communication are tens if not hundreds of thousands of years old, only within the last 8,000 years or so have people systematically recorded them in external media. Beginning with cuneiform, hieroglyphs, and ideograms, human memory was no longer restricted to the bounds of the body, but could now be held in "external storage systems." The cognitive changes accompanying this memory change include the emergence of information retrieval knowledge as more important than rote memorization and the ability to overcome working memory limitations in thought processes using an "external memory field." The ability to critically examine, piece by piece, exact writings led to the rise of "theoretic culture" -- science, philosophy, and other deep investigations into the nature of the world.
Donald presents the ancient Greeks as the inventors of the critical piece of theoretic culture. By "recording the verbalizations and speculations," rather than just recording facts, the Greeks used writing as more than an extended data store, but as a method of developing and refining ideas. (p. 342) I don't see that this cognitive skill was invented by the Greeks; that people present incomplete and unrefined ideas orally is clear to anyone who has speculated around a campfire or practiced inquisitive gossip. Storing these ideas in a written form did, of course, allow scientific and philosophical inquiry to take on the rapid pace and careful methods that it did, but I am unconvinced that the third transition brought humans a qualitatively different cognitive process.
While we now possess vastly superior abilities in developing and refining ideas, due in large part to the benefits provided by external media, the cognitive operations are still carried out in the human brain. The specifics of data retrieval have changed and in some domains (such as math and logic) the symbols have become significantly non-oral, but the cognitive paradigm is still one of semantic and linguistic processing of each datum. I believe that we did (and continue to) undergo a cognitive change, but it came not from writing but from computers. With the aid of computers, people can solve problems without developing a useful internal representation or performing any significant internal cognition. In a sense this is not qualitatively new, since the results of any computer program can be generated, in theory, by a human with pencil and paper. But the change resides in the locus of thought -- both the human and the computer do part of the processing, so the "thinker" is both in tandem. I am unwilling to propose, however, that consciousness resides outside the human body, if only because such consciousness would violate our intuitive definition, which is the best one we have. Donald does state "the locus of attentional control can reside, at least temporarily, in external memory" (p. 372). By this and his examples, I take it he means that media like television can determine the object and flow of our awareness, a fact clear to anyone who watches a well-made film. This does not mean, as I understand it, that that thing which is attentive, and hence conscious, is anywhere outside of the person.
Commentaries in Behavioral and Brain Sciences by Bickerton, Thompson, and Mitchell & Miles criticized Donald's presentation of Brother John as an example of mimetic cognition without linguistic cognition. Brother John had the benefit of a lifetime of linguistic thought in his favor and he was also acting within a linguistic culture. In his response, Dunbar indicates that his purpose was to show that Brother John was able to be conscious and function without direct access to his linguistic skills. This resolves the first problem, but does little to dismiss the environmental point posed by Mitchell and Miles. How much of cognition is a function of a person's physical and social environment? Donald sketches an answer to this question on p. 309 when he says "The individual picks and chooses, acquires skills and knowledge from society, but nevertheless exists as an easily identifiable unit within that society... memory and thought occur only in the individual mind or brain and therefore are to be regarded as attributes of the individual." He then presents computer networks as a metaphor for the relationship between a person and external media. His network metaphor focuses on storage, even though he skirts with the idea of offloading computation to other society members. But the cognitive advantages of a social network go beyond data storage; Brother John was able to check into a hotel and perform other basic travel tasks because he could communicate his need for action by other parties. All coordinated social action, such as hunting and ritual performance, depends on such performance delegation. Responding to Costall, who Donald interprets as "saying that psychology should abandon the internalist model altogether" (BBS p. 784), Donald indicates that the internalist approach remains valid on at least one level of investigation. This is certainly true; Donald just doesn't adequately explore the important approach of group cognition, though he suggests that "social coordination was one of the prime movers of both mimetic and linguistic evolution" (BBS p. 784). His reason for hesitancy to pursue this lead is revealed in his response to Feldman -- at his time of writing, the field of distributed computing was not well developed, and Donald was hesitant to call upon it for empirical and data and helpful models. Distributed computing has since blossomed, providing fertile modeling possibilities to explore the cognitive community as more than a storage device. The picture of group cognition that I have just presented differs from the earlier discussion of computer qua delegated thinker in that there the computer is used in a generative capacity, producing new knowledge. In social coordination, a shared representation of the situation is developed, the end of which is some group action.
Throughout the book, Donald takes an evolutionary approach to the development of human cognitive capacities. After dismissing quantitative increase in brain size and encephalization as sources of uniquely human cognition, Donald presents the three cognitive transitions by arguing for them in evolutionary terms. He proposes that each of the adaptations is vestigial -- modern humans possess episodic and mimetic cognitive capabilities in addition to visuo-symbolic and linguistic ones. Donald's picture of cognitive evolution is thus one where each new skill could develop from the existing cognitive landscape. He also presents each transition as selected for under pressure, though not as rigorously as one might desire.
Donald's arguments are not as heavily based in evolutionary theory as the present day school of evolutionary psychology. He uses evolution as the underlying mechanism of human cognitive development, but looks to neurological, anthropological, and archaeological data to support his historical speculations. Furthermore, Donald grants a significant role to culture and training in human cognitive capacities, especially (and understandably) for the visuo-symbolic stage. Opposed to Donald's approach, thinkers like Cosmides and Tooby and their camp approach human cognition by investigating what specific mental adaptations would have provided for genetic fitness. Donald is more focused on cultural transmission and individual cognition than on reproduction driven capacities.
The picture of human mental makeup is quite different in Donald than in the evolutionary psychology school. For instance, Donald isn't as module-happy as the evolutionary psychologists. Rather than a Swiss Army Knife with a different module designed for everything from cheater detection to causal reasoning, Donald paints a picture of more general modules to match the four modes of human cognition. These domain general modules are described at the functional level and are far from a set of tabula rasa -- Donald spends a fair amount of space discussing the effects of brain lesions and modular cognitive capacity loss. To counter the Swiss Army Knife metaphor, one might think of Donald's picture of the human mind as a workshop with various types of tools all available to work on the task at hand. While this proposal avoids the problem of the naive homunculus while leaving room for consciousness and cross-modal cognition, it does a poor job explaining why, for instance, people are better at solving the Wason selection task when cast in terms of cheater detection than in its abstract setting. Partly this is because Donald's goal is to explain how our minds reached their current state, not the intimate details of how they work. It would be interesting, though, to see Donald's mental architecture used to explain the myriad of intriguing data from experimental psychology.
Donald thus differs on most of the core principles of evolutionary psychology outlined by Cosmides and Tooby 1997 Rather than "neural circuits ... designed by natural selection to solve problems that our ancestors faced during our species' evolutionary history ... specialized for solving different adaptive problems," Donald proposes that our neural circuitry developed to be able to understand and solve the wide variety of challenges presented to our ancestors. Donald is also ambivalent on the claim that "our modern skulls house a stone age mind." He clearly thinks that modern non-literate aboriginal cultures retain the mythic culture developed in the Upper Paleolithic period, but he believes that visuo-symbolic thought is qualitatively different from the Stone Age mind. Thus, the Stone Age skulls of literate people house a modern mind.
Despite Donald's evolutionary approach to human cognition, he presents precious little evidence or speculation along the axes of the nature/nurture debate. The reader is left wondering how much of human mimesis is genetically provided and how much is picked up from culture. Donald states (on p. 189) that "the mimetic layer of representation survives under the surface, in forms that remain universal, not necessarily because they are genetically programmed but because mimesis forms the core of an ancient root culture that is distinctly human." He does not, however, pursue empirically the interplay of genes and this "ancient root culture," which leaves us to ponder. How much of Jung's mythic "collective unconscious" (adopted by Joseph Campbell) is present in each child and how much of it is a product of recurrent themes in long-standing myths? How might mythic narrative understanding, including its implications about the local environment, be modeled? And while it's fairly clear why a social group with, say, mythic culture would outcompete a social group without it, Donald gives no indication what selection pressures and reproductive strategies would operate within such a culture. While Donald's book is not necessarily made worse for not pursuing these questions, they must be answered before we can hope to arrive at a complete picture of the mind and its evolution.
Donald's focus on origin rather than function presents several other questions worthy of exploration. What problems is mimetic cognition adept at solving and what types of problem require symbolic thought with linguistic capabilities? How does the answer tie into computability classes? Does Newell's 1980 problem space hypothesis, that "the fundamental organizational unit of all human goal-oriented symbolic activity is the problem space" apply to mimetic activity, which is human, goal-oriented, and representational, but not symbolic? Are the visuo-somatic components of mimesis (gesture, expression) processed differently than the aural components (rudimentary song and, by possible extension, percussion)? (Donald thinks that these are handled by the same general mimetic system, but they could be handled in different ways within that system, just as symbolic language and symbolic imagery are processed by different parts of the brain.)
Donald's challenge to the artificial intelligence community is something of a call to refocus -- "The first formal model of truly human representation, far in the future, will have to cope with the problem of mimesis before it considers the problem of linguistic invention." While it's true that mimesis must be addressed in any full simulation of human cognitive skill, it may be more challenging than linguistic skill. Mimetic skill arises from biological evolution and it is intimately tied to the mime's bodily self-image and depends on the assumption that the interlocutor possesses similar mimetic cognition. Even if mimetic culture was implemented among robots, they might not be able to use mimetic skill to communicate much with humans, unless the robot was fairly human in form (another set of profound challenges). Analogously, hundreds of years spent imitating birds were fruitless in producing human flight. This is not to discount such a work of robotics -- they would doubtless produce fascinating cultural adaptations to their environment, they just wouldn't be the same adaptations that humans developed. Donald acknowledges Lutz's comment that mimetic robots must interact with other robots of the same type, and defers judgments until the necessary research has taken place.
On the other hand, producing a computer program capable of linguistic, and perhaps even mythic, thought seems less daunting, in part because we have linguistic computational models, but in larger part because we already have the tools needed to communicate linguistically without calling on biological properties of humans. An understanding of human mythic subtlety may require human experience, or at least simulated human cognition, but there is no reason to think that uniquely computer mythologies couldn't develop, given a linguistic (and even non-mimetic) base. More simply, rudimentary computer narrative understanding may be within our grasp within a few years. However, the lofty goal of artificial intelligence able to reproduce or model all human cognitive abilities is, I think, misguided. Distant seems the need for a computer able to react to the biological challenges faced by humans, from determining which recently consumed food the stomach rejected to mimetically conveying how to care for a twisted ankle.
Structurally, Origins of the Modern Mind is slightly jarring. Donald has a tendency to jump from domain to domain and oscillate between detail and architecture. In many places, the reader is left to construct for herself the argument from data to theory. Furthermore, as has been noted by commentators, Donald is unclear about the source of many of his claims, especially archaeological and anthropological, many of which his reviewers dispute. Taken as a whole, however, his theory was not hard to discern, and it is not surprising that a work as broadly interdisciplinary as this one should encounter difficulties with material in dispute within a given field. I found the book highly illuminating and I look forward to applying its ideas widely.
Cosmides, L. & Tooby, J. (1997). Evolutionary Psychology: A Primer. http://www.psych.ucsb.edu/research/cep/index.html
Donald, M. (1991). Origins of the Modern Mind: Three Stages in the
Evolution of Culture and Cognition. Cambridge, MA: Harvard University
Press.
Donald, M. (1993). "On the evolution of representational capacities"
(author's response) from Behavioral and Brain Sciences (1993) 16:4
737-791.
Commentaries: