The Secret Operations of the Mind
Department of Philosophy / Cognitive Science Program
Los Angeles, CA 90041
The Secret Operations of the Mind(1)
For my part, my only hope is, that I may contribute a little to the advancement of knowledge, by giving in some particulars a different turn to the speculations of philosophers, and pointing out to them more distinctly those subjects, where alone they can expect assurance and conviction.
- David Hume, A Treatise of Human Nature(2)
It is a common practice among philosophers of psychology to trace the origins of functionalism, and cognitive science more generally, to texts deep within the history of philosophy. Plato, for example, is described by Hubert Dreyfus as a "knowledge engineer" for the view he develops in the Euthyphro of expertise as the mastery of explicit rules and for the doctrine of recollection in the Meno. (Dreyfus, 1990, p. 13-15) Others have argued that Aristotle, Duns Scotus and William of Occam were functionalists. (Smith, 1990) The popularity of functionalist theories of mind in recent years has, not surprisingly, encouraged the search for its antecedents. Properly carried out, the re-interpretation of historical texts in contemporary terms can serve both the historian and the philosopher concerned with contemporary issues. The important question is whether, in particular cases, such reinterpretation sheds any light, either on the contemporary issues or historical texts.
Functionalism may extend back to antiquity, but modern philosophy has seemed to many the age in which functionalism was born. In Artificial Intelligence: The Very Idea, John Haugeland suggests that by the seventeenth century an attempt to articulate a functionalist account of mind was already afoot. (Haugeland, 1984) Haugeland traces the development of functionalism from Copernicus to Galileo, Hobbes, Descartes, and finally to Hume. Though Hobbes explicitly theorized about the mind as a computational device, it was David Hume, on Haugeland's reading, who first undertook a "mental mechanics" which does not invoke a homunculus. Hume's is the first non-homuncular philosophy of mind inspired by Newton's mechanical philosophy of (non-human) nature. Hume shed the theological baggage which had encumbered his rationalist predecessors and founded a true science of human nature. Just as astronomers formulate principles of ever greater generality about the physical realm, so Hume urges, can philosophers about the mental realm. (Hume, 1975, pp. 14-6), (Haugeland 1984, p. 44)
Hume figures prominently in other recent histories of functionalism, including Fodor's critique of procedural semantics. Fodor's argument against the idea that meaning can be understood as compilation to machine language has two parts. First, Fodor argues that the sense in which a natural language sentence can be represented in a machine language isn't the sense needed to account for the meaning of the former. The representation of "Whales are mammals" in the internal memory of the machine doesn't show anything about the relation of "whales" to whales. So what is needed is a language semantically midway between a natural language and a machine language to save the insight of procedural semantics. The second part of the argument is to consider how such an enriched machine language could be achieved. That's to be done by attaching sensory transducers to the machine. Now the machine language representation can connect the natural language meaning of "whale" at one end with the set of sensory stimulations on the transducers at the other.
Fodor's historical move is to equate this second move in procedural semantics with what he calls the "Locke-Hume Program." He takes this program to endorse the view "that every nonlogical concept is reducible to sensation concepts." (Fodor, 1978, p. 213) Fodor's Hume is a sense-data theorist and sense data theories are the forerunners of contemporary procedural semanticists. Procedural semantics, according to Fodor, is tainted with the same problem which "several hundred years" of empiricism have shown to be intractable. You can't get meaning out of sense data. Fodor interprets Hume as a reductionist; perceptions are the (inadequate) pre-AI stand-ins for physical symbols.
According to Daniel Dennett, Hume is pivotal in the history of functionalism. Functionalism before AI and since differ primarily by their relation to what Dennett calls "Hume's Problem." Hume attempted to account for mental representation by the dynamic interaction of mental items called impressions and ideas. The problem is how to characterize such interaction as mental without appealing to the fact that impressions and ideas are the perceptions of a mind. The difficulty backed Hume into an impoverished theory of the self. "The result was his theory of the self as a "bundle" of (nothing but) impressions and ideas." (Dennett, 1978, p.122)
Those who credit Hume with anticipating mechanistic accounts of cognition agree that Hume didn't quite pull it off, that a plausible version of functionalism would have to wait for developments in logic and computation. Having explained mental operations in terms of primitive operations on discrete mental items called "perceptions", Hume can't say what makes the perceptions mental, and so he can't say what makes the whole "science" a science of the mind. What Dennett calls "Hume's Problem" is the homunculus problem. Hume, like Hobbes before him, lacked the notion of an automatic computing machine, and thus he failed to articulate a plausible mental mechanism. (Haugeland, 1984, p. 44) The bundle theory couldn't work until AI introduced the idea that properly coordinated and automated "dumb" homunculi can together give rise to mental properties. Dennett credits Hume with formulating the central problem of cognitive science, but he also unmercifully concludes that Hume's "solution" (the bundle theory) couldn't work.(3)
Although the temptation to see Hume as an early functionalist is great, there is a compelling reason for resisting it: There are other contenders for the title, including rationalist philosophers from whom Hume worked hard to distance himself. How can philosophers who hold such divergent views be seen as providing the groundwork for a common philosophical position? I'll argue that they can't, that including Hume among the founders of functionalism rests on a mistaken reading of some of Hume's central claims. The prehistory of cognitive science is more complex than many have appreciated, but it is well worth exploring.
II. Hume as Functionalist
Functionalism of the sort attributed to Hume occupies the logical space between reductionistic views like the type-identity theory, which identifies the mind with the brain, and dualism, which treats the mind as an object immune from scientific investigation. Functionalists identify the mind not with the physical states which they hold cause mental states, but with the functional organization of such states, an organization which could be instantiated in some other physical system from the one in which it is typically instantiated. For our purposes the following claims can be attributed to the typical functionalist:
(1) Psychology is a science. The mind can be investigated and studied as a natural phenomenon. (Newell and Simon, 1976, pp. 113-126)
(2) Psychological laws are couched in the terms of an autonomous level of functional description. They are not reducible to physical or biological laws, for example. (Pylyshyn 1986)
(3) In principle, everything about the mental can be explained by psychological
laws of the kind specified by (2) together with non-psychological laws.
A few points about these theses are in order. First, they do not jointly single out a particular flavor of functionalism. There's room in (2) for different accounts of the autonomous level of functional explanation. (1) and (3) are held by some non-functionalists, and both are subject to various interpretations. (Churchland, 1986, pp. 277 ff.) It's important to see why (3) is a functionalist claim. Consider the dualist. She holds that scientific inquiry is irrelevant to understanding the way the mind works. The functionalist couldn't disagree more. Science can, in principle, account for all mental phenomena, because mental phenomena are functional states of physical stuff, not mysterious spiritual stuff. We may perish before we fully understand the functional structure of the mind, but that's not an in principle limitation.
Does Hume make claims (1)-(3)? There is some evidence that he does. On the title page Hume introduces his Treatise of Human Nature as "An Attempt to introduce the experimental Method of Reasoning into Moral subjects." A great deal could be said about this pronouncement and the project which it introduces. Historians of psychology are fond of citing the Treatise as an abortive beginning to experimental psychology. (Hothersall, 1984, pp. 43-44) Hume announces that he will employ experiments, but few psychologists think much of the experiments which are found in the Treatise.(4) Whether Hume employed the experimental method, or employed it consistently, is not at issue here. For Hume the mind is an object of empirical inquiry. So Hume is committed to the first thesis of functionalism.
Hume begins the Treatise with a theory of impressions and ideas. The scope of the project is ambitious; it includes nothing less than a categorization of "all the perceptions of the human mind". (T1) There are impressions, which are the original and lively perceptions, and ideas, the "faint images of these in thinking and reasoning." (T1) Complex impressions and ideas are concatenations of simple ones. Hume articulates and defends a principle of the formation of simple ideas: All simple ideas derive from antecedent simple impressions which they resemble. This principle quickly attains a normative status. The only legitimate ideas are those with the right kind of causal history. Ideas must be derived from impressions. Notions such as soul and substance violate this principle. We can't trace the idea of substance to simple impressions and so it has no place in the science of human nature.(5)
These considerations suggest that in the early sections of the Treatise Hume promises a scientific psychology, one based on a theory of the atomic constituents of the mind and their associative relations. The attempt to make good on this promise comes when Hume tries to show how complex cognitive phenomena such as our judgments of duration, continued existence, and belief generally can be explained in terms of connections among simple atomic mental items.
It appears, then, that Hume proposes a new science of human nature, a science which can in principle explain everything about the mind. Haugeland cites the introduction to the Treatise in support of the attribution of this functionalist tenet to Hume. (Haugeland 1984, p. 43) There Hume wrote:
There is no question of importance, whose decision is not compriz'd in the science of man; and there is none, which can be decided with any certainty, before we become acquainted with that science. In pretending therefore to explain the principles of human nature, we in effect propose a compleat system of the sciences, built on a foundation almost entirely new, and the only one upon which they can stand with any security. (Txvi)
As Haugeland reads this passage, Hume holds that everything in human nature has an explanation. The mind, like the Newtonian domain, is a mechanism which defies explanation only when we can't fathom its complexity. Our failure to explain certain features of the mind, then, could never show that such features cannot be explained. Rather, explanatory voids are merely indications of our ignorance, and they are to be filled eventually by more acute or more experienced human minds. Hume's Treatise is an attempt to provide the basic principles which explain all mental phenomena. The principles involve perceptions, that is, impressions and ideas, and their relations.
Hume's enthusiasm with the experimental method does suggest a step in the direction of functionalism and away from the introspective psychology of Descartes. But, I shall argue, there are rationalist strains in contemporary functionalism which don't fit this picture of its empiricist origins. The case for the historical link from Hume to artificial intelligence can only be made if we can square the rationalist tendencies of the latter with the empiricism and anti-rationalism of the former.
III. The Secret Operations of the Mind
We've seen that Hume is considered a forerunner of AI, the Newton of the mind who understands all mental forces and properties in terms of perceptions and the basic principles of association -- resemblance, contiguity, and cause and effect. Is this an accurate assessment of Hume's contribution to cognitive science? We can only answer this question by examining Hume's views on the explanation of mental phenomena. I'll argue that an affirmative answer makes Hume out to be a rationalist about such explanation. Although the opening passages of both the Treatise and the first Enquiry suggest this form of rationalism, Hume's account of our inferential lives in key passages is incompatible with the rationalist reading of Humean explanation. But first I need to say something about rationalist forms of explanation.
Rationalism, as a philosophical doctrine, is perhaps most commonly understood as an epistemological view.(6) Rationalists regard reason as the principal source of knowledge. Empiricists, in contrast, hold that knowledge derives from the senses. The battle between these two doctrines is played out on many fronts. One, relevant to cognitive science, is the issue of innate ideas. One of the first conclusions reached in the Treatise is that there are no innate ideas, since all ideas are derived from impressions. (T7, T157) It doesn't escape Hume's attention that if he's right about innate ideas, then Descartes has to be wrong that the existence of an innate idea of God demonstrates God's existence. (T160) Thus any account which makes Hume out to be a rationalist in this sense is implausible.
Rationalism can also be characterized as the view that everything has an explanation.(7) Descartes embraces this approach to explanation at the close of the Discourse on Method, when he writes "I tried to discover in general the principles or first causes of everything that exists or can exist in the world." (Descartes, 1985i, p. 143) This is a daunting task, and even Descartes is not fully confident that any individual human mind can account for the number and variety of phenomena to be met with in scientific investigations. The limitations Descartes refers to, however, are simply the limitations of finite human inquirers. Leibniz's rationalism about explanation is every bit as thoroughgoing as Descartes'. Individual substances are things which have a concept which is so complete, that we can "deduce from it all the predicates of the subject to which the concept is attributed." (Leibniz 1956, p. 307) In deducing those predicates, however, we make reference to other individual substances to which the substance in question is related. That turns out to be all substances. So the concept of an individual substances includes a complete account of the universe from the standpoint of that substance.
Rationalism so understood seems implausible; we are confronted with the inadequacy of our explanations at every turn. Descartes and Leibniz held that there is an explanation for everything, but not that an individual human mind or a set of minds working together have the capacity to provide every explanation. Full explanations are to be expected only from an infinite being. Thus the rationalist happily admits human limitations, particularly cognitive limitations.(8) The contrasting view is not empiricism, but naturalism. The naturalist holds that accounting for human limitations is illuminating, since there is no transcending them.
It's time to connect both rationalist themes, first, that reason, not the senses, is the primary source of knowledge, and second, that complete explanations are possible. Complete explanations are possible, if only from God, because human reason is just like God's reason, except in its scope. Descartes begins the Discourse on Method with the claim that we all have the same faculty of reason. Descartes thought it followed from the existence of this universal faculty of reason that our understanding of that faculty itself could not be informed by empirical science. The rationalist's view of the senses is in sharp contrast to the account of universal reason. For Leibniz, sensory information depends on one's limited point of view, one's relation to other substances. The senses, in contrast to reason, can only supply incomplete or point-of-view based understanding. It's the kind of understanding dogs have. (Leibniz, 1956, p. 638)
Hume, in contrast, will have none of God and the appeal to a point of view outside of our own. He might, however, still share the rationalist's proclivity for complete explanation, as Dennett, Fodor and Haugeland maintain, while rejecting the rationalist's claim that the senses place limitations on explanation. Cognitive science allows us to be rationalists without believing in God; Hume's crude functionalism, crude because it lacked the idea of an automatic formal system, would be a crippled rationalism, but a form of rationalism nonetheless. In this sense it is quite natural, in light of Hume's optimism about the scope of his proposed science of human nature, to see him as a rationalist about explanation.(9)
Much of the textual support for the claim that Hume is a proto-cognitive scientist comes from the early and optimistic pages of the Treatise and the first Enquiry. In the latter work Hume proposes the experimental method as a sensible alternative to what he calls "easy and obvious" philosophy on the one hand and "obscure and abstruse" philosophy on the other. (Hume, 1975, pp. ii ff.) The matter looks a bit different, however, when one probes further in the texts.
After raising the now-famous skeptical doubts about the justification of beliefs based on causal reasoning, Hume provides a theory of belief in which custom or habit has a central place. Beliefs result when we experience the constant conjunction of causes and effects, what Hume calls the mere "number of past impressions and conjunctions". (T101) After several instances of fire followed by heat, for example, we habitually think of heat when we see fire. Beliefs are the products of such habits. Much of Hume's attention is devoted to studying these habits and accounting for various aspects of habit-based belief formation.(10)
Hume's view that custom or habit alone is at work in belief formation leads to a problem of which Hume was acutely aware: Some beliefs are formed without the repetition of past conjunctions. Hume provides the following example:
A person, who stops short in his journey upon meeting a river in his way, foresees the consequences of his proceeding forward; and his knowledge of these consequence is convey'd to him by past experience, which informs him of such certain conjunctions of causes and effects. But can we think, that on this occasion he reflects on any past experience, and calls to remembrance instances, that he has seen or heard of, in order to discover the effects of water on animal bodies? No surely; this is not the method in which he proceeds in his reasoning.(T103-104)
This passage raises an issue which concerns cognitive science, though the passage is not cited by those who see Hume as cognitive science's champion. The problem belongs to the subfield of AI called knowledge representation. The problem, which contemporary cognitive scientists claim as their own, is this: When we come to the river we make precisely the right inference, that we'll suffocate if we keep walking, and we make that inference in real time. The number of potentially relevant facts we might consider as we approach the river is massive; how do we pick out the facts relevant to the inference? This is a question about common sense reasoning, a question which traditional epistemologists didn't raised because they were preoccupied with the question of how beliefs are justified rather than with the difficult and important issue of how everyday knowledge is structured. (Minsky, 1981, pp. 95-128) But the passage just cited provides evidence that Hume was concerned with the mechanisms of common sense knowledge. More importantly, it illustrates Hume's position on knowledge representation, and thus his views on topics central to cognitive science.
A person who meets a river for the first time knows the effect of water. How do we represent this knowledge? Like other examples of common sense knowledge, it appears that we just have it. There are no antecedent constant conjunctions of water and drowning to prepare the belief, and so Hume can't use his standard account of belief formation. Instead he says that experience can operate on our minds in an "insensible manner," and "experience may produce a belief ... by a secret operation." Hume writes:
The idea of sinking is so closely connected with that of water, and the idea of suffocating with that of sinking, that the mind makes the transition without the assistance of the memory. The custom operates before we have time for reflexion. The objects seem so inseparable, that we interpose not a moment's delay in passing from the one to the other. But as this transition proceeds from experience, and not from any primary connection betwixt the ideas, we must necessarily acknowledge, that experience may produce a belief and a judgment of causes and effects by a secret operation, and without being once thought of. (T104)(11)
Hume's solution, that we form such beliefs by a "secret operation," hardly seems an illuminating explanation. There are several possible interpretations of Hume's secret operations. One interpretation, offered by John Biro, is that the mental operations that produce belief in such cases "are though secret, not occult." (Biro, 1985, p. 259) Biro suggests that secret principles are open to investigation in principle. On his view, Hume refers to such operations as secret merely to indicate that we don't yet have an account of them. Biro is certainly right to reject the interpretation of secret principles as being in conflict with scientific laws of human nature. Hume rejects the occurrence of miracles on the grounds that the evidence for them cannot issue from regularity-induced habit, and that's the only way to get beliefs. So Hume isn't about to endorse principles which conflict with laws of nature. Still, secret principles could be compatible with laws of human nature, without being discoverable by humans. The operation of the understanding in such cases are secret in the sense that empirical science won't reveal any underlying mechanisms due to natural limits to our ability to observe and generalize about our own cognitive mechanisms.
The sense in which the cognitive processes in the river example occupies the space between Biro's reading and the occult reading. Our cognitive powers are surely not unlimited. It is possible that there are cognitive processes we cannot explain because of such limitations. Belief-forming operations of the mind are secret in such cases due to the limitations of human understanding. If we simply lack the resources to understand how we form beliefs in such cases, the processes at work are not occult or miraculous, i.e in conflict with the laws of nature. It does not follow, however, that we can or will ever be able to understand the underlying mechanism. If this interpretation is correct, it undercuts the claim that Hume championed a proto-functionalist account of the mind.
To understand Hume's remarks on the mind's secret operations it is important to contrast what Hume says about the secret operation of experience with the possible secret operation of reason. Just a few pages before the river passage, Hume denies that belief can arise from "any new operation of the reason or imagination." (T102) His evidence for this is that he is never aware of any such operation. Yet we've seen that Hume allows that there are "secret operations" of the mind, operations of which we are unconscious. How can Hume deny that there is any secret operation of reason while he finds no corresponding problem with the claim that experience operates unconsciously in the river example? How can he, in the case of reason, require that mental operations be available to consciousness, yet waive this requirement in the case of experience?
The issue is complicated by the fact that when Hume discusses the river example, it appears that reason is at work, rather than experience. He talks about the "close connection" between the ideas of water, sinking, and suffocating. But Hume immediately points out that the inference (transition) we make doesn't involve a "primary connection betwixt the ideas," and thus he concludes that such inferences are not based on reason, and are therefore secret. Why does the fact that reason is not operating suggest that the operation is secret?
Reason, demonstrative reasoning, in contrast to the operation of custom in causal reasoning, requires the explicit comparison of ideas as the basis for inference.(12) Hume argues that our reasoning in the river case cannot be demonstrative. If there were a "primary connection" between the ideas of water, sinking, and suffocating, we would be aware of that connection, since we can only reason about that of which we are aware. So explicit connections, the ones provided by the faculty of reason, cannot be secret. In contrast, the operation of experience in such cases occurs "before we have time for reflexion." Hume says that in such cases there is no reflection at all, and we draw inferences "without forming any principle concerning it." (T104)
This last claim of Hume's is quite remarkable: Hume was supposed to be the philosopher with the discrete atomic mental items and principles in terms of which the entire mental realm is to be explained. But here Hume asserts that there are no principles underlying this quite common cognitive phenomenon. Does this sound like functionalism? This is the denial of (3), and the rejection of a basic functionalist commitment.
One might grant that we can't introspect the cognitive operations underlying the inference that the river might cause harm but still insist that such inferences can be investigated. There's no reason, however, to think that Hume would allow this move. It is not that Hume is taking the results of introspection as the source for his theory here; rather, there's nothing to suggest that the workings of experience in such cases are open to scrutiny. Hume would not deny that our everyday beliefs about the effect of water on animal bodies are natural, and thus part of the causal order. What he denies is that all natural phenomena can be brought under causal laws by us. An essential part of the science of man is discovering the limits of human reflection on human cognitive faculties.(13)
A less sympathetic reading is that Hume is hedging; he doesn't have the principles and so labels them secret. Hume is still committed to their existence. Saying that we don't form such principles in reasoning just means that Hume can't figure out what they are. And Hume can't be blamed for this; he should be credited with treating a knowledge representation problem, and contemporary cognitive science should be credited for discovering the tools for solving it. (Newell and Simon, 1976, pp.113-115)
Remember that Hume distinguishes between reason and the understanding.(14) A skeptical argument has established that beliefs are not the product of an explicit reasoning process, but are rather the natural products of habit and custom. If we used reason to form beliefs about the effect of water, we'd have principles of the operation of reason. But there need not be principles for the understanding. Belief formation is not an operation of reason, and so there's no requirement that there be discoverable principles of the formation of all beliefs.
Reason, as a faculty which compares ideas, plays no role in belief formation. That's Hume's position. Recent computational accounts of the mind suggest a brand of reason which holds out some chance of dealing with the river problem. Computational systems are reasoning devices, and their mechanisms are open to scrutiny. I think it is clear, however, that this notion of reason was not available to Hume, and there is no hint of it in his writings. Rather, the river passage challenges what Hume thought was a central rationalist doctrine, the claim that we can fully explain the cognitive operations which lead to belief.
When Hume's remarks on the secret operations of the mind are placed in the context of his account of causation and causal inference, it makes sense to see them as underlining a limitation on our ability to understand our cognitive processes. Beliefs are not the products of reason; they are the effects of a natural process. For Hume, it is an empirical fact that such processes are not capable of being fully explained by us.
The characterization of Hume as a functionalist is false, when functionalism is understood as committed to the complete explanation of the mental. It doesn't follow from this that Hume's views are at odds with all versions of functionalism or with other accounts of the mental presented in cognitive science. It's beyond the scope of this paper to explore all the Humean strains in cognitive science, but two points should be emphasized. First, there is a methodological commitment in cognitive science that there can be a science of human nature. Hume also proposed a science of human nature, though this may be obscured by the fact that Hume's proposed science was more ambitious in scope than cognitive science, including all of moral theory. Second, cognitive scientists have become increasingly interested in the ways in which agents fall short of perfect rationality. Here, cognitive science and Hume have much in common. The functionalist, like the rationalist, gets the promise of complete explanation at a cost. For the rationalist, the cost is the idealization of human reasoning in an infinite being. For the functionalist, the idealization is a characterization of cognitive processes in abstraction from real world constraints on finite agents.(15) The rejection of such idealizations does not entail the rejection of the scientific study of human nature, either by Hume or by contemporary cognitive science.
The quotation with which I begin this paper, from the conclusion of Book I of the Treatise, suggests that Hume hoped to illuminate some particulars and some subjects where we can expect "assurance and conviction." A central theme of Hume's philosophy, apparently missed by those ready to have him bear the legacy of cognitive science, is that part of the work of locating areas where we can have assurance and conviction is locating areas where we can't have assurance and conviction. This would indeed be a contribution of the first order to the fundamental concerns of cognitive science, if not the contribution many have thought Hume made.
1. An earlier draft of this paper was delivered at a session of the Fourth Computers and Philosophy Conference . I'd like to thank Herbert Simon for helpful comments at that session. Penetrating comments from an anonymous referee for Minds and Machines on an earlier version of this paper prompted what I hope are significant improvements, though I, of course, am responsible for any remaining defects.
2. (Hume 1973) ; hereafter "T".p. 273.
3. I've argued elsewhere that attributing a bundle theory of the self to Hume rests on misunderstanding of the metaphysical status of perceptions. Cf. (Traiger, 1988)
4. In Book I of the Treatise, see "Of skepticism with regard to the senses" (Book I, Part IV, Section ii), where Hume offers experiments to show that perceptions are mind dependent, and"Of personal identity" (Book I, Part IV, Section vi) for Hume's empirical evidence against the claim that there is a simple idea of the self. In Book II, see Hume's discussion of the "two curious experiments" to confirm his double relation of impressions and ideas doctrine of the origin of the passions in "Of Beauty and Deformity" (Book II, Part I, Section viii), and "Experiments to confirm this system" (Book II, Part II, Section ii). The Twentieth century positivists are fond of quoting the concluding paragraph of the Enquiry concerning Human Understanding (Hume, 1975, p. 165) in which Hume entreats us to "commit to the flames" any volume which contains neither mathematical nor experimental reasoning.
5. This picture doesn't do full justice to the complexity of Hume's position. Hume appreciates that many notions, such as substance, self, and duration, to choose just a few, play a central role in the conceptual scheme of the "vulgar" or common person. Hume's naturalistic psychology attempts to account for the origins of such notions, in light of the fact that they are not derived from impressions. Cf. T15 ff. and also (Livingston 1984, p. 60)
6. In contemporary linguistics, rationalism is a doctrine about the way grammars of natural languages are learned. A rationalist in this sense posits innate learning mechanisms to account for the child's language learning behavior. While this linguistic rationalism has ties to the historical forms of rationalism, they shouldn't be confused. See (Ramsey and Stich, 1990)
7. This characterization of rationalism is due to Wilfrid Sellars. Cf. (Sellars, 1965)
8. Cf. Rene Descartes, Meditations on First Philosophy, in (Descartes, 1985ii, pp. 37-49) and G. Leibniz, Discourse on Metaphysics and "Clarification for Bayle" in (Leibniz, 1956, pp. 313 f.f, p. 495)
9. It is admittedly odd to speak of Hume as a rationalist when he is usually associated with the empiricist tradition. In what follows I will exploit the tension in this characterization to argue against such an interpretation. In spite of its oddness, this is an interpretation of Hume which is often articulated, and which must be evaluated.
10. The passages in the Treatise which deal with belief formation occur in some of the least well known, but most rewarding sections of Book I. See T107 ff.
11. Some editions of the Treatise read differently here. Most notably, in the 1911 Everyman's Library edition, and the 1886 Green and Gross edition, the phrase is not "secret operation" but "separate operation." These readings, however, are in error, and the Selby-Bigge/Nidditch edition used here contains the corrected text.
12. In Book III of the Treatise Hume contrasts imagination and reason on the grounds that we may be conscious of relations of ideas while they fail to have any influence on the imagination. Cf. T339 .
13. In this respect, Hume's project is quite Kantian. Compare, for example, the First Paralogism in the A (first) edition of Immanuel Kant's Critique of Pure Reason, (Kant, 1929, pp. 333-334), with Hume's Treatise Book I, Part IV, Section vi. For a discussion of the overlap of Hume and Kant, see (Beck 1978, pp. 101-109).
14. Hume's terminology makes it difficult to state this without qualification. Sometimes Hume calls the process of drawing inferences from causes to effects "reasoning", and this is clearly part of the understanding. Here I do not contrast understanding from causal reasoning, but rather, reasoning in the strict sense, which is just a matter of what Hume calls "relations of ideas," i.e. demonstrative reasoning.
15. Cf. (Cherniak, 1986) for a discussion of constraints on the reasoning processes of finite agents. Unfortunately, Cherniak saddles Hume with occult processes. He notes that Hume characterizes our ability to come up with appropriate instances of abstract ideas as involving "a kind of magical faculty in our soul." (T24) I read this as an admission that Hume doesn't have an account of the mechanism, not that our ability is a feat of magic. Like the case of secret operations, this ability is, on Hume's view, an empirical fact that is "inexplicable by the utmost efforts of human understanding." (T24)
Beck, Lewis White (1978) "A Prussian Hume and a Scottish Kant", in Lewis White Beck, Essays on Hume and Kant, New Haven: Yale University Press
Biro, John (1985) "Hume and Cognitive Science",
History of Philosophy Quarterly 2 no. 3, July
Cherniak, Christopher (1986) Minimal
Rationality Cambridge: MIT Press
Churchland, Patricia (1986) Neurophilosophy:
Toward a Unified Science of the Mind/Brain, Cambridge: MIT Press
Dennett, Daniel C. (1978) "A Cure for the
Common Code?", in Brainstorms, Montgomery: Bradford Books
Descartes, Rene (1985i) Descartes: Philosophical
Writings, Vol. I, John Cottingham, Robert Stoothoff, Dugald Murdoch,
Anthony Kenny trans., Cambridge: Cambridge University Press
Descartes, Rene (1985ii) Descartes:
Philosophical Writings, Vol. II, John Cottingham, Robert Stoothoff,
Dugald Murdoch, Anthony Kenny trans., Cambridge: Cambridge University Press
Dreyfus, Hubert (1990) "Socratic and Platonic
Sources of Cognitivism", in J-C. Smith, ed.
Historical Foundations of
Cognitive Science Dordrecht, The Netherlands: Kluwer Academic Publishers
Fodor, Jerry A. (1987) Psychosemantics,
Cambridge, MA: MIT Press
Fodor, Jerry A. (1981) "Tom Swift and his
procedural grandmother", Cognition 6, Reprinted in Jerry A. Fodor,
Representations, Cambridge, MA: MIT Press
Haugeland, John (1984) Artificial Intelligence:
The Very Idea, Cambridge: MIT Press
Hothersall, David (1984) History of
Psychology, NY: Random House
Hume, David (1973) A Treatise of Human
Nature, 2nd Edition, L.A. Selby-Bigge, P.H. Nidditch, eds. Oxford:
Oxford University Press
Hume, David (1975) Enquiry concerning
Human Understanding, Oxford: Oxford University Press, L.A. Selby-Bigge
and P.H.Nidditch, eds., third edition
Kant, Immanuel (1929) Critique of Pure
Reason, Norman Kemp Smith, trans., NY: St. Martin's
Leibniz, G. (1956) Philosophical Papers
and Letters, Leroy E. Loemker, ed. Dordrecht, The Netherlands: D. Reidel
Livingston, Donald (1984) Hume's Philosophy
of Common Life, Chicago: University of Chicago Press
Minsky, Marvin (1981) "A Framework for
Representing Knowledge", in John Haugeland, ed.
Mind Design Cambridge:
Minsky, Marvin (1986) The Society of
Mind NY: Simon and Schuster
Newell, Allen and Simon, Herbert, (1976)
"Computer Science as Empirical Inquiry: Symbols and Search", Communications
of the Association for Computing Machinery 19, March
Pylyshyn, Zenon (1986) Computation and
Cognition, Cambridge: MIT Press
Ramsey, William, and Stich, Stephen (1990)
"Connectionism and Three Levels of Nativism", Synthese 82: 177-205
Sellars, Wilfrid (1965) "Meditations Leibniziennes"
American Philosophical Quarterly 2, No. 2: pp. 1-13
Smith, J-C. ed., (1990) Historical Foundations
of Cognitive Science Dordrecht, The Netherlands:Kluwer Academic Publishers
Traiger, Saul (1988) "The Ownership of
Perceptions: A Study of Hume's Metaphysics"; History of Philosophy Quarterly,
V, 1; pp. 41-52