Sketch for a General Theory of Human Psychology
Is it possible to come up with high-level organizational principles of human psychology? The task has been attempted before–as in associationist psychology, classical conditioning theory, and computationalism. The hope is to discover general principles that cover a wide variety of psychological phenomena, thus unifying what appears disparate. In this paper I make another attempt at this project, by integrating what we have learned since Chomsky introduced his theoretical framework, also adding some emphases that may be less familiar. I will be operating at a high level of abstraction.
I shall consider four types of competence: linguistic, logical, geometrical, and social. By linguistic competence I mean the ability to produce and comprehend grammatical sentences of one’s native language (or the cognitive structure that underlies this ability). By logical competence I mean the ability to reason logically, i.e. according to valid rules of inference—to follow trains of reasoning produced by others and to produce such trains oneself. By geometrical competence I mean the ability to classify and manipulate geometrical forms—to tell triangles from squares, to grasp geometrical relations, to master school geometry upon suitable instruction. By social competence I mean the ability to grasp social relationships and dynamics, to read the minds of others, to understand social groupings (family, village, nation). No doubt these competences break down into sub-competences, with a good deal of inner complexity. Thus we have phonetic, syntactic, and semantic competence; we have competence in deductive and inductive logic, in modal or deontic logic, and so on; geometric competence can include basic sensori-motor tasks such as sorting objects according to shape, as well as theoretical grasp of abstract geometry; social competence will involve a whole host of abilities employed in social interactions, predicting the behavior of others, and moral evaluation (e.g. knowing what a promise is and that breaking promises is blameworthy). In these four competences we see huge areas of the human mind at work, with enormous sophistication and complexity; so if we can discern general features of the operative systems we will have discovered something very general about the mind. My question is whether we can unify these four competences by articulating general organizational principles, thus providing a synoptic picture of what is distinctive to the human mind. More broadly, I want to know what the mind must be such that it is capable of the competences in question—its most general properties.
A number of questions can be asked about these four competences, as follows. What are the universal features of each competence, i.e. what features do all humans share that possess such competences? What are the linguistic universals, the logical universals, the geometrical universals, and the social universals? What are the specific principles involved in each competence? What do they have in common, if anything? How are the competences expressed in action (performance)? To what extent are the underlying principles innate? What does the schedule of acquisition look like? How did the competences evolve? Are any of them more basic than the others? Might any derive from the antecedent presence of others? How do they interact with each other? How are they realized in the brain? What are their characteristic pathologies? To what extent are the operative principles conscious? We think of each competence as psychologically real and we ask questions about their internal structure, their origins, their interactions, their physical realization, their overt expression in behavior, how basic each is, and so on. We try to do justice to their distinctive properties, so that our model of each competence is not impoverished or distorted. We take them for what they are, instead of trying to shoehorn them into a preconceived theoretical box.
What I now propose to do is list the most general features the four competences have in common. Many of these will be quite familiar, though my interpretation of them may not be. I am looking for universals acrossthese domains, not within them. The mind will then be characterized as the system that has these inter-competence universals: that is, there are abstract principles that are given specialized form in each specific competence. There is an abstract operational schematism that gets exemplified in linguistic, logical, geometrical, and social competence. Let me emphasize that I am offering only a sketch here, with little detail or empirical confirmation.
(i) The first feature goes by several names: generative, recursive, combinatorial, compositional, creative, infinite. The point is usually applied to linguistic competence—in language we can produce a potential infinity of sentences based on combinations of primitive parts. Sentences have structure and our mastery of language reflects that structure. But the same basic point applies to the other three competences. Logical arguments also have structure and our grasp of them is a projection from mastery of primitive modes of inference. A complex chain of reasoning is a composite of smaller bits of reasoning. Also, we grasp abstract rules of inference that apply to infinitely many potential cases. Grammatical rules generate well-formed sentences; logical rules generate valid arguments. In geometry we have figures composed of primitive parts—lines, planes, solids. Infinitely many figures can be produced by iteration of basic geometric components. We demonstrate a grasp of these principles of combination in our ability to build complex objects using simpler objects of certain shapes—as with basic building-block operations (architecture is a more sophisticated expression of our geometrical competence). In social cognition we grasp social units as combinations of simpler elements (people and other animals): thus we grasp the concepts of family, friend, village, pack, herd, marriage, nation, and so on. We understand how individuals combine with others to form certain kinds of social unit. We also understand social relations, such as promising, contracting, befriending, lying, cheating, and so on. Our grasp of morality is part of this competence, which is about right and wrong in social relations. We function as we do socially only because we have this kind of social cognition.
It is customary to express the point by saying that we can analyzecomplex structures into parts—the wholes are not taken as primitive. I would add that we are also capable of synthesis, as we fuse the elements into wholes. In reception we analyze; in production we synthesize. We grasp the basic units as elements of a potential synthesis—words, propositions, shapes, and individuals. So we see the parts in relation to each other and to constructed wholes. We see the elements according to their roles—what they do in relation to other elements (“if xwere combined with y, we would get z”). Words combine into sentences, propositions combine into arguments, shapes combine into geometrical structures, and people combine into social formations. In each case the competence involves grasp of part-whole structures, where wholes can in turn become parts, and so on indefinitely (“recursion”). This abstract principle is therefore universal to the four competences: it is the general idea of a generative system.
(ii) The second feature I shall call segmentation. By this I mean that the mind conceives the elements of a combination as discrete entities sharply distinguished from other entities. Thus we conceive of words as clearly individuated, as genuine units with their own identity; and similarly for propositions, shapes, and individuals. We do not regard these elements as intrinsically fuzzy or continuous with other elements. It is a well known fact that the acoustic signals of speech are physically far less well defined than what we hear, far more continuous than heard speech (as revealed by a speech spectrograph); we experience these signals as discrete units (“phonetic segmentation”). We actively segment the stimulus. Much the same is true of the visual stimulus: we segment the ambient array into sharply defined objects. Thus we impose segmental structure onto the world—we insist on sharply demarcated units. No doubt this aids the mind-brain’s combinatorial proclivities, for now we have nicely defined units with which to work. It isn’t the world that foists the segments on us; rather, we foist segments onto the world, in order to facilitate our psychological operations. In any case, in each of our four domains the mind works with a basic “vocabulary” of discrete elements—things than can function as manageable segments of a larger whole.
(iii) Thirdly, we have the notion of rule-governed principles of combination. Not just anything goes; you have to play by the rules. Words must be combined according to grammatical rules if the output is to be successful. Here we encounter modal notions: you mustcombine words thus and so and not just higgledy-piggledy. Similarly, you mustinfer conclusions from premises according to valid logical rules, and not just anyway you feel like. And there are geometrical rules too: you can only construct a triangle by combining lines in a certain way; you can only build a house by setting bricks of certain shapes one upon another. Breaking such rules produces monsters like Escher drawings or round squares. Thus we have the notion of geometrical necessity. It is much the same for social arrangements: there are rules about what social formations are permissible, as with marriage or employment arrangements. Thus we employ the idea of social obligations and social freedoms—what is required by social rules and what is not. Deontological ethics is precisely a theory of social rules. To be sure, the rules are of different kinds in our four cases, but in each case we have the idea of rule-governed combinations—those that obey the rules and those that do not. Putting this together with the first two features, we can assert the following: the abstract schematism involves combining discretely segmented units into synthetic wholes according to precise rules of combination. We operate with rules in each of the four areas and we recognize what constitutes obedience to a rule and what does not.
(iv) The next feature is a corollary of the previous one: each of our four domains incorporates a prescriptive or normative dimension. That is, notions of right and wrong can be correctly applied to the domain. There is a right way and a wrong way to combine words, determined by the grammatical rules; nonsensical combinations are deemed undesirable; and you can be criticized for flouting the rules of grammar. I don’t mean what is called “prescriptivism” about usage; I just mean basic rules of sentence formation. Split infinitives and dangling participles are fine, but it is bad to produce a string like “Barking it’s sing John car very”. You are expected to meansomething by what you say. It is goodto speak meaningfully. In the case of logic, prescriptivism is clearly right: you oughtto reason logically, and illogical reasoning opens you up to warranted criticism. We use logic precisely in order to evaluatearguments. In geometry too there is a right and wrong way to draw an equilateral triangle and, as Plato observed, we have the idea of the perfect triangle, which no drawn triangle ever quite attains. Indeed, Plato’s entire conception of geometry sees it as a repository of value—those perfect unchanging Forms that elevate us in their very contemplation. In the case of social competence we need look no further than ordinary morality, with its many prescriptions about conduct in relation to others (“stealing is wrong”). Obviously morality is the domain of right and wrong. In each area there is heavy infusion of value judgment—of a sense of rightness and wrongness, perfection and imperfection. It is not all value-neutral description but is shot through with approval and disapproval, praise and blame. And we act as we do becauseof these normative judgments—clearly in the case of morality but no less so in the other cases. We try to draw the perfect triangle, we make an effort to reason logically, and we are ashamed to make grammatical blunders (not that we often do, save in pathological cases such as aphasia). We are guided by the governing norms of the competence, respectful of their demands. We see things under normative conceptions. Here the human mind is saturated with notions of value and it proceeds accordingly. Thus the rules are not experienced as arbitrary but as conducive to genuine values: it is a good thing to speak grammatically, commendable to reason logically, admirable to draw triangles as close to the ideal as possible, and right to act morally. Human psychology is steeped in evaluations of many kinds (though this is not something you would guess from typical behaviorist psychology or even computational psychology: I will come back this point).
(v) The four competences, as so far characterized, are quite abstract in their general mode of operation: they must be described in highly abstract language in order to bring out their commonalities. We are accustomed to the abstractness of grammatical rules (a point often made by Chomsky), and the abstract nature of logic is also well attested, as are the abstractness of geometry and moral rules. But now we perceive a higher level of abstractness, as we discern what these competences have in common: the idea of a process that is generative, segmental, rule-governed, and norm-guided. In principle, this very abstract structure could be implemented in many ways, as it is in the four competences considered here; it is neutral with respect to more specific expressions. Maybe in Martians the creation of art is subsumed by a system with this abstract character, which appears not to be the case for humans; maybe in other terrestrial species so-called “language” does not fall under the general schematism I am sketching (dolphins, bees). We might think of the schematism as a kind of “super-competence”—an abstract structure that lies behind and makes possible the specific competences we have discussed. Where this super-competence came from, and how it was specialized into the four specific competences, we don’t know; but it is conceivable that it pre-dates them and has some entirely alien origin (as it might be, our ability to negotiate trees in our dim arboreal past). In any case, the deep architectural principles of the human mind are extremely abstract—multiply adaptable schemas, not specific interpreted contents. Specific contents get slotted into the abstract schema, but it has a nature and psychological reality that transcends its particular exemplifications. Just as universal human grammar is abstract relative to particular human languages, so the general schematism is abstract relative to universal grammar. Thus the schematism can show up as the basis of various types of competence: that is the picture that is emerging. The four competences are no doubt quite modular, but it may be that they stem from something universal—something with a higher level of abstractness. We can try to investigate the nature of this abstract schematism as such, formulating as best we can its general properties.
(vi) The competences are all cognitive. That may seem like a triviality, but it is not. The word “cognition” refers specifically to knowledge, not mere belief or other mental representations. In each area we knowthings to be so: we know that a given sentence is grammatical because we know the rules of grammar; we know that a certain inference is logically valid (we don’t just conjecture that this is so); we know what a triangle is and that no perfect triangle has ever been drawn; we know that stealing is wrong (we don’t merely have a tentative opinion about it). So in our sketch for a general human psychology we need to make it explicit that we are dealing with states of knowledge—the concept of knowledge becomes a central concept for psychology. We are characterizing systems of knowledge, properly so-called—not just “internal representations”. The study of our mere conjectures about remote history or deep space may not be a study of systems of knowledge, given our ignorance in these areas; but we are not similarly ignorant about what is grammatical or logical or triangular or morally right. The output of the abstract rule-governed generative schematism is knowledge in the most straightforward sense in these cases.
(vii) It will be useful to have a short label for the schematism I am describing, so let us call it the “forms and norms” schematism. Then we can express the next feature by saying that the forms and norms schematism is doubly universal: first, it is universal across human beings—everyone is equipped with it, short of devastating brain pathology; second, it is universal across a variety of human competences, being shared by (at least) the four competences I am describing. It is doubtful that it is possessed by other species, except perhaps in a very rudimentary form; and it may not be shared by all human psychological capacities, especially those inherited during evolution from earlier types mind (such as the ancestral fish that led ultimately to us). Basic sensori-motor skills and innate reflexes don’t have this kind of abstract structure. It is an interesting question whether our musical ability is a forms and norms system (music theory makes it seem so, but mere receptivity to beat and melody seems too primitive). It does seem that what is most distinctive of the human mind centrally involves a full-blown forms and norms structure: generative, segmental, rule-governed, evaluative, abstract, cognitive.
(viii) Our language faculty appears to incorporate both a conscious and an unconscious component: we are conscious of sentences as grammatical and we can articulate a good deal about the rules of grammar, but it is also true that the competence includes an unconscious level—which is why we find it hard to formulate universal human grammar. Much the same seems to hold of the other three competences: we reason logically not by consciously formulating the laws of logic but by having an implicit grasp of them (it took Aristotle and Frege to bring these implicitly grasped laws to explicit awareness); our understanding of geometry is largely implicit until we start studying the subject in school (recall Socrates and the slave boy in the Meno); and much of morality is not consciously formulated but instinctively acted upon. So we can say that the forms and norms schematism has both a conscious and an unconscious representation in the mind. Perhaps the underlying abstract structure once had a purely unconscious representation, but once it became exploited by more specific competences its character became more conscious to us—though it still remains largely unconscious. It is certainly true that we do not, in the ordinary course of life, experience ourselves asengaging in abstract operations with the character I have tried to describe; instead the schematism just whirs away inside us, quietly going about its work.
(ix) Chomsky has long urged that the structures of universal human grammar are innate. What about the other three? Without going into the matter in detail, it seems safe to assume that much the same is true of them: our logical faculty is an innate component of the human mind, as is our geometrical faculty, and evidence is accumulating that moral psychology has an innate basis. If the underlying forms and norms structure is itself innate, which seems overwhelmingly likely, then it will not be surprising if the faculties it grounds are also innate. These areas of knowledge are not like our knowledge of history or geography or what is fashionable this season—all these being clearly acquired. But the four competences have a strong claim to innateness, for reasons that are now well appreciated. This dovetails with the previous point, since what is innate is likely to be unconscious: the schematism is specified in our DNA and grows in the brain during the course of maturation, only becoming conscious along the edges, so to speak. Again, we see a commonality that confirms the idea that we are here dealing with a psychologically real internal structure, hard-wired and universal.
I have now enumerated, briefly and dogmatically, the common features that I see as holding over the four competences I am considering. I now want to articulate further what the internal character of the forms and norms schematism is, as well as point to how adopting this perspective alters the way one sees human psychology. The general character of the schematism will be familiar from work done by philosophers, psychologists, and linguists over the last several decades, variously formulated and with varying emphases. I have merely brought these ideas together, while imparting my own spin. A useful metaphor is that of a network: the elements of a network exist in relation to other elements of the network, united by linking relations. Thus we have the conception of language as consisting of a vast network of signs that link with each other in various way, coming into proximity with each other to form phrases and sentences, according to fixed rules. In logic we think of propositions as laid out in logical space, linked by logical relations such as entailment or inconsistency, with rules about what propositions can be inferred from what. Our psychological structure as logicians has to mirror the objective logical structure in some way, so that we can move around it cognitively. In geometry the metaphor of a space become literal, since geometrical forms are conceived as regions of space, carved out in a particular way. Figures can be conjoined with other figures, or laid over them, fitting or not fitting. The spatial world looks like a huge mosaic of geometrical figures, regular and irregular (hence Plato’s doctrine that the essence of the material world is geometry). And social groupings are another kind of network: patterns of connection between people, linkages, aggregations, hierarchies, and collectivities. Each person has a place in this “social mosaic”, and what we are partly depends on our social role (cf. “semantic role” for words). It is all a matter of systems of discrete elements that combine and recombine according to rules, generating endless new wholes, with a heavy dose of the normative (this one good, that one bad). Accordingly, we need in the mind representations for the basic units, representations for rules of permissible combination, and a device to evaluate the outcomes. The mind needs to be able to segment and amalgamate, and it needs a grasp of the point of this mental work. The basic form of a mental operation is thus: segment-amalgamate-evaluate (SAE). The human mind is (among other things) an SAE device.
The first two parts of SAE have been well recognized: the mind must be able to analyze and synthesize, to break down and build up. It cannot build up unless it has first broken down—for it needs segmented elements as the building blocks of constructive operations. If there were no words in sentences, we would have to invent them. Given that we want to have sentences, and given that we are finite creatures, we had better find a way to analyze sentences into finitely many constituent and re-combinable parts. Similarly with the visible world: we need a finite stock of visual primitives if we are to make sense of the huge variety of visual scenes the world can present. We also need something like fixed persons to make sense of social life: we need the idea of the same person being a member of many groups or moving from one group to another. That is, we need the idea of an atom if we are to have the idea of a molecule. And where would we be without the notion of determinate shapes and sizes and combinations thereof? But once we have the elements, neatly segmented, we also need rules to combine them—we must be able to synthesize according to rules. Thus we arrive at the idea of the mind as a machine for analysis and synthesis that incorporates rules. This is all pretty orthodox today, even if it sounded revolutionary fifty years ago.
But where is evaluation in all this? It tends not to get mentioned. So I want to carve out the rightful place of evaluation in the SAE model; I want to give it its due. And my first point is simply that the mind is also a normative machine: it evaluates things. Sometimes this is acknowledged but then scanted: the normative dimension is regarded as essentially epiphenomenal. Yes, we engage in evaluations—of sentences, arguments, shapes, and social actions—but none of that makes any difference to anything. For how can values influence facts? How can the grammatical rightnessof a sentence play any real role in what we do with it? This is no more possible than moral values playing a causal role in the world. And here we reach the nub: mental causation cannot be influenced by values. So if the mind is indeed steeped in values, as I suggested, then these must be epiphenomenal, and hence hardly worth mentioning. The causation must be ordinary mechanical causation, of the same kind that we find in the purely physical world; but then there cannot be any such thing as evaluative causation.
What should we say about this line of thought? First, there is confusion in it. The claim is not that values themselves figure in mental causation but rather that judgmentsof value do. It isn’t that we produce a grammatical sentence because of its havingthe objective value of being grammatically correct; we do so because we takeit to be grammatically correct. Compare: I refrain from stealing something not because it iswrong to steal but because I deemit wrong to steal. But these normative attitudes are not themselves values—they are psychological facts. So why can’t our attitudes towards values causally influence our actions and our mental operations? Why did I go into a particular restaurant? Because I believed they serve good food there and I wanted good food (not as a result of the goodness of the food considered independently of what I believe and desire). This is no more problematic than acting on any other kind of belief and desire. So there is nothing metaphysically to prevent us from crediting the mind with a host of evaluative attitudes that influence the way it works. We could even postulate an unconscious Grammar Evaluator that issues verdicts on strings of words put together by our grammar module, determining which strings will actually get uttered. It says things like “This one good” or “That one bad”. The judgments it makes could have causal powers in respect of what sentences get uttered. And the same could be said for our logical faculty: it issues normative verdicts on arguments in process and can facilitate or halt that process. At any rate, there is no argument derived from the metaphysics of causation to prevent such a hypothesis. Psychological causation by attitudes with evaluative contents seems no more problematic than other sorts of psychological causation. It is true that some theorists are allergic to the use evaluative notions in scientific theories, but their objection cannot stem from considerations about causation. And it is surely obvious that human beings are deeply evaluative creatures—they are always going on about right and wrong, perfection and imperfection, praise and blame.
This rejoinder is fine so far as it goes, but I don’t think it goes quite far enough. For I think that the objective rightness involved ispart of the overall psychological story: we can truly say that certain outcomes occurred becauseof an objective rightness in things. For example, it is perfectly true to say that I don’t produce verbal strings like “Barking it’s sing John car very” precisely because that sentence is grammatically defective (bad, wrong). It is also true to say that I don’t steal things precisely because it is wrong to steal: that is, the fact that it is wrong to steal is whyI don’t steal. There are true “because” statements linking values with psychological facts, as there are true “because” statements linking physical facts with psychological facts; and I think this is an important point about human psychology (it is doubtful that animals can be made subject to such value-mind explanation). So I want to bring values into psychology proper, as part of the SAE package. In order to explore this question fully I would need to go into the entire metaphysics of causation and explanation, which I do not propose to do here. I will say simply that the mechanical model of causation has long been obsolete even in physics (gravity is not a kind of mechanical contact causation). I would favor a more Aristotelian approach to causation, in which causation is made correlative with why-questions and is linked closely to explanation. So while it is true that values cannot literally make physical contact with minds (the two cannot touch) they may yet figure in answers to why-questions. If we ask why I don’t produce nonsense sentences, then the answer is that they are patently ungrammatical and nonsensical: it is becauseof that fact that I don’t utter them. It is an entirely verbal question whether we should speak of this as “causation”. What matters is that it tells us why things happen. We can have perfectly true and informative “because” statements of the kind in question. Indeed, there are true law-like general statements of the type, as in: “Normal speakers don’t utter ungrammatical sentences simply because they areungrammatical”. In the same way we can say “People don’t steal simply because stealing is wrong”. The wrongness of stealing explains why people believe it is wrong to steal, and that belief explains their non-stealing actions. Truth explains belief (which is not to deny that other factors can come into play). This strikes me as simple common sense; and it is important to acknowledge that what happens in people’s minds can have this kind of explanation. Thus a comprehensive psychology will include values as part of its explanatory framework. To put it differently, the human mind is sensitive to values, unlike other animal minds: we think in terms of values and values are part of the explanation of our actions and mental processes. To bleach value out of the study of mind is to miss this important fact, producing a misleading model of how things work (orthodox computationalism is guilty of this). In a slogan: the mind crunches values as well as symbols.
Let me emphasize how modest this claim really is. It says no more than that we are aware of values because of their existence and that this awareness affects what we do. Thus we are aware that it is good for sentences to be grammatical or for arguments to be valid, and that awareness affects our actions. This is whywe put together only grammatical strings and respect valid inferences and keep our promises. Psychology therefore needs to build values into its conceptual framework, simply because the human mind is a value-sensitive device (unlike the merely physical world or the botanical world or most of the animal world). It is also a generative, segmental, and rule-governed device: these are all just facts about the kind of thing it is. Each aspect of SAE must be fully and robustly acknowledged.
Further questions arise. If this is the essential nature of human cognition, how did it arise in evolution? What pre-adaptations made it possible? How does it develop in the child’s mind? Are there other mental faculties with the same general structure? If so, do they derive from any of the four we have considered, singly or in combination (physics, arithmetic, chess, etc)? How is the SAE schematism implemented in the brain’s neural hardware? The last question is especially difficult when it comes to value: for how do brains and values connect? But none of these questions is easy, once we take on board the full reality and abstractness of the forms and norms schematism and its place in the mind’s overall landscape. This is why what I have offered here is little more than a sketch, an aspiration. Perhaps we can be comforted by the reflection that these are at least (and at last) the right questions.
This view is not as silly as it sounds, given the actual conditions under which the intelligence of our ancestors evolved. If the brains of our ancestors evolved to cope with life in the trees, they would need to develop mental representations of the branching structure of trees, which would be necessary to both sensory and motor competence. That would be the most important part of the environment to gain competence in negotiating. Once the geometric structure of trees was mastered it could be generalized and applied elsewhere, so that the tree schema might underlie other forms of competence: for example, social and family relations might be modeled on the structure of a tree. And of course we do speak of “branches” of a family and indeed of “family trees”. Could the tree-like structure of grammar itself be a transformed application of the early mental representation of trees? How could human cognition notbe shaped by the arboreal environment in which our ancestors evolved and lived for millions of years? The brain of the gibbon must above all be a tree-adapted brain with a finely tuned understanding of the properties of trees; and gibbons have evolved a sophisticated form of language. Intelligence is apt to be niche-specific. The genes are geared to the particular environment in which they exist, with respect to both body and mind. Thus tree genes must be part of our genetic inheritance.