The Relation of Inference 
 to Fact in MilVs Logic 
 
 (By , 
 
 ], FORSYTH CRA WFORD 
 
 PHILOSOPHIC STUDIES 
 
 ISSUED UNDER THE DIRECTION OF THE DEPARTMENT OF 
 PHILOSOPHY OF THE UNIVERSITY OF CHICAGO 
 
 NUMBER 5 
 
 THE UNIVERSITY OF CHICAGO PRESS 
 CHICAGO. ILLINOIS 
 
 Agents 
 
 THE CAMBRIDGE UNIVERSITY PRESS, London and Edinburgh 
 THE MARUZEN-KABUSHIKLKAISHA, Tokyo. Osaka, Kyoto 
 KARL W. HIERSEMANN, Leipzig 
 THE BAKER & TAYLOR COMPANY, New York 
 
i 
 
 V , - 
 
 The Department of Philosophy of the University of Chicago issues 
 a series of monographs in philosophy, including ethics, logic, and meta¬ 
 physics, aesthetics, and the history of philosophy. The successive 
 monographs are numbered consecutively with a view to their subsequent 
 publication in volumes. These studies are similar to the series of 
 Contrihutions to Philosophy, but do not contain psychological papers or 
 reprints of articles previously published. 
 
THE RELATION OF INFERENCE TO FACT 
 
 IN MILL’S LOGIC 
 
Digitized by the Internet Archive 
 in 2019 with funding from 
 Princeton Theological Seminary Library 
 
 https://archive.org/details/relationsofinferOOcraw_0 
 
THE RELATION OF INFERENCE TO 
 FACT IN MILL’S LOGIC 
 
 FORSYTH CRAWFORD 
 
 THE UNIVERSITY OF CHICAGO PRESS 
 CHICAGO, ILLINOIS 
 
Copyright igi6 By 
 The University op Chicago 
 
 All Rights Reserved 
 Published April iqi6 
 
 Composed and Printed By 
 The University of Chicago Press 
 Chicago. Illinois, U.S.A. 
 

 CONTENTS 
 
 
 
 CHAPTER 
 
 
 PAGE 
 
 I. 
 
 Subjective Facts as Data . 
 
 . 
 
 7 
 
 II. 
 
 The World of Objective Facts 
 
 . 
 
 . i6 
 
 III. 
 
 The Universality of Attributes and 
 
 Propositions . 
 
 22 
 
 IV. 
 
 Inference as Anticipation of Facts 
 
 . 
 
 • 27 
 
 V. 
 
 Inference as Proof of Propositions 
 
 . 
 
 • 34 
 
 VI. 
 
 Inference as Discovery 
 
 
 • 43 
 
 5 
 
CHAPTER I 
 
 SUBJECTIVE FACTS AS DATA 
 
 The problems of logic are coming to form the center of interest in 
 philosophical discussion. Most present controversies in the field of 
 philosophy turn upon some diflference which is fundamentally logical. 
 There is need, therefore, to re-examine and clarify the underlying con¬ 
 ceptions of logic. Any one of the great historical systems of logic offers 
 a species of laboratory example of the shaping of logical concepts under 
 some specific point of view, and the difficulties which it leaves unresolved 
 offer valuable experimental material for further logical construction. It 
 is with such a purpose that this paper undertakes an analysis of the 
 relation of inference to fact in Mill’s logical system. 
 
 In constructing a theory of inference Mill’s problem was to reconcile 
 the associationism of the English empirical school with the procedure of 
 modern physical science. He occupies alternately the subjective point 
 of view of the one position and the objective point of view of the other. 
 This oscillation of position comes out strikingly in the varied status 
 which he gives to facts, which are at one time the ultimate constituents 
 of consciousness and at another time the things and events of an inde¬ 
 pendent world. When stating the relation of inference to its data and 
 thus assigning it a locus, he takes the subjective point of view. When 
 analyzing the nature of inference and basing its validity, he takes the 
 objective point of view. It will be well, then, at the outset, to examine 
 his conception of both subjective and objective facts, and the transition 
 which he seeks to effect from the one to the other. Over against both 
 of these, facts in the mind and facts in nature. Mill naively utilizes a 
 system of meanings which he never thinks of as sundered from the facts, 
 yet which are assumed as logically independent in his inferential process. 
 An examination of the system of meanings we shall postpone till after 
 we have examined his assumption of facts, and will consider the latter 
 apart from their meaning so far as that can be done without distortion. 
 
 The locus of inference in relation to fact, as assigned by Mill, rests 
 on the distinction between immediate and mediate knowledge. The 
 first is self-evident, the second reasoned. We obtain the first by intui¬ 
 tion, the second by inference. “Truths are known to us in two ways: 
 some are known directly, and of themselves; some through the medium 
 
 7 
 
8 
 
 RELATION OF INFERENCE TO FACT IN MILL’S LOGIC 
 
 of other truths. The former are the subject of Intuition; .... the 
 latter, of Inference.”^ They are ‘‘ truths known to us by immediate con¬ 
 sciousness ” and conclusions which can be drawn from these.Of these 
 two classes of truths, the first does not require proof; such a truth carries 
 its own evidence—“that is, is without evidence in the proper sense of 
 the word.”^ It must therefore be “known beyond possibility of ques¬ 
 tion.”4 One cannot but be certain of it. But only a small part of our 
 knowledge has intuitive character.^ By far the larger portion hangs 
 suspended from these immediate truths by a series of proofs. 
 
 This apportionment of knowledge appears on the face of it very clear 
 and definite. At any rate it seems to offer a hopeful program: to begin 
 with perfectly certain data and to let each step from those data be one 
 of strict proof. Upon examination, however, the simplicity of the matter 
 disappears. Like the gratuitous advice to “be sure you are right, then 
 go ahead,” this program does not tell us how to know when we are right. 
 This conception of primitive and unquestionable data, indeed, is thor¬ 
 oughly ambiguous. While he does not uncover its source. Mill has to 
 acknowledge this ambiguity in two respects: 
 
 First, Mill admits that there is difference of opinion as to the kind 
 of data from which we start. He sets in sharp contrast two theories on 
 this point: the “ontological” theory of the Scotch school, and the 
 empirical theory of his own English predecessors, which he defends.^ 
 The former holds that, in addition to the data admitted by the latter 
 school, the mind is so constituted as to know intuitively the existence and 
 laws of operation of certain objects external to the mind, while the latter 
 recognizes “no ultimate premises but the facts of our subjective con¬ 
 sciousness; our sensations, emotions, intellectual states of mind, and 
 volitions.” Mill considers all thinkers to be agreed that there are some 
 such primitive data, but to disagree as to whether or not such data are 
 exclusively subjective. But surely the very possibility of dispute as to 
 the character of what are to be considered primitive data is a very 
 serious obstacle to the theory that there are any data at all to be taken 
 as unambiguously primitive. 
 
 There is a second ambiguity of more serious character, not so expli¬ 
 citly recognized, though virtually acknowledged by Mill. While it 
 
 ^ Mill, Logic, 8th ed., Harper, igd. References to the Logic are hereafter by page 
 numbers, without repeating the title. To make references to the Logic more definite, 
 letters are added to designate in order each paragraph or part of paragraph on a page. 
 
 2 ige-ioa; Mill, Examination of Hamilton, 137. 
 
 3 21b. 4 20c. s 21c. ^ 520&. 
 
SUBJECTIVE FACTS AS DATA 
 
 9 
 
 seems clear to him that consciousness immediately reveals the primitive 
 data of our knowledge, which consist of our own subjective states, Mill 
 finds it impossible in practice to determine what those data are. Assum¬ 
 ing agreement as to their subjective character, their actual discovery in 
 detail proves impossible. Supposed to be known “beyond the possi¬ 
 bility of question,” they are in fact unrecognizable, because they cannot 
 be distinguished from the inferences that are drawn from them. “We 
 may fancy that we see or feel what we in reality infer. A truth, or a 
 supposed truth, which is really the result of a very rapid inference, may 
 seem to be apprehended intuitively.”^ Observation, which is another 
 name for intuition, is constantly vitiated by confusion with inference. 
 “Observation and inference are intimately blended.”^ Now if we ever 
 mistake inference for observation, it surely follows that we can never be 
 sure we have not done so. There is no safe criterion by which to dis¬ 
 tinguish the two; for the apparent self-evidence of immediate intuition, 
 our only resource in the case, is admitted to be liable to fail. It is idle 
 to urge “a correct discrimination between that, in a result of observa¬ 
 tion, which has really been perceived, and that which is an inference from 
 the perception,”^ for the supposed purity of the observation may still be 
 only our “fancy.” “From my senses,” Mill claims, “I have only the 
 sensations, and those are genuine.”4 No doubt they would be genuine 
 if I could be sure that they are truly mere sensations. “ Errors of sense,” 
 he says again, “are erroneous inferences from sense; .... the decep¬ 
 tion .... is in my judgment.”s But since such errors are always 
 possible, how can I know that there are no further errors of judgment 
 lurking in what I fancy to be perception ? 
 
 Evidently this distinction between intuition and inference proves 
 unworkable, and breaks down for lack of a criterion of discrimination 
 between the two. If we press the difficulty farther to discover its precise 
 nature, we find it involved in Mill’s theory of consciousness, or of expe¬ 
 rience, as consisting of given mental,states. An experience of that kind 
 would properly leave no room for inference; and in grafting inference 
 upon it Mill introduces an alien element in contradiction with his 
 premises. This point will repay closer examination. 
 
 ^ 2od. Cf. Mill, Berkeley, 278. 
 
 2 4506. So Mill contrasts “experience” with “a mistaken supposition of experi¬ 
 ence” (I9S^/), or with “a superficial semblance of experience” (igbu). 
 
 3 45I(;. “We ought to know what part of the assertion rests on consciousness, 
 and is therefore indisputable, what part on inference, and is therefore questionable” 
 
 (S 45 <^)- 
 
 4451a. 
 
 S45od. 
 
lo RELATION OF INFERENCE TO FACT IN MILL’S LOGIC 
 
 Mill professes himself an adherent of Berkeleyan idealism, which he 
 received from his father in the form of an elaborated associational psy¬ 
 chology. To this school of thinkers it seems axiomatic that only our 
 own mental states are immediately known to us. It inevitably follows 
 that consciousness is shut up to its own subjective contents, and that 
 the knowledge of anything else, if the possibility of such knowledge be 
 granted, must be capable of translation into terms of mental states. 
 
 The primary interest of this school, both historically and construct¬ 
 ively, is directed upon the analysis of the knowledge of an external 
 world into constituents of sensation. So strong is this interest in sen¬ 
 sation, an interest which has rightly earned for the school the designation 
 of sensationalist, that one would almost suppose consciousness to be 
 composed solely of a kaleidoscopic succession of sensations. Sensations, 
 in their temporal and spatial discreteness, their qualitative resemblances, 
 and their associational connections, are at any rate taken as the type 
 of mental facts. And it is this sensationalism on which Mill rests when 
 he lays down the distinction between intuition and inference. 
 
 Yet sensations are not the only contents of the mind acknowledged 
 by Mill. With reference to the other contents of the mind there are two 
 conflicting tendencies: either, on the one hand, to reduce them all to 
 mere given mental states of the same immediacy as sensation, in which 
 case there is no opening for inferences to slip between them; or, on the 
 other hand, to lift them to the level of various operations performed upon 
 and with sensation, in which case inference finds a work to do but 
 immediacy is gone. 
 
 There are several of these other classes of mental states recognized 
 by Mill, and one or two factors that he does not think of classing as 
 mental states, but which, if not so classified, do not fit into his original 
 theory of consciousness. Beside sensations, he classes as mental states 
 images, emotions, volitions, and thoughts. Relations he hesitates about 
 so classing. Meanings he does not class here, and on this side of his 
 system they are simply an outstanding factor. Some of these non- 
 sensational mental facts require examination. 
 
 What we now call mental images constitute the simplest of the non- 
 sensational mental states recognized by Mill. Mill usually calls them 
 “ideas,” as against sensations, which he also calls “impressions,” after 
 Hume.^ “Whenever any state of consciousness has once been excited 
 in us, no matter by what cause, an inferior degree of the same state of 
 consciousness, a state of consciousness resembling the former, but 
 
 ^ But of course he has other uses of the term “idea.” 
 
SUBJECTIVE FACTS AS DATA 
 
 II 
 
 inferior in intensity, is capable of being reproduced in us without the 
 
 presence of any such cause as excited it at first.This law is 
 
 expressed by saying, in the language of Hume, that every mental impres¬ 
 sion has its idea.”^ “These ideas, or secondary mental states, are 
 excited by our impressions, or by other ideas, according to certain laws 
 which are called Laws of Association.”^ Here one might be tempted to 
 see between primary and “secondary mental states” a distinction cor¬ 
 responding to that between intuition and inference, or immediate and 
 mediate. But such a comparison would be entirely misleading. Impres¬ 
 sions and ideas stand for Mill upon exactly the same level of immediacy. 
 The distinction which he makes between them on the ground of the 
 presence or absence of a stimulating object^ is from this point of view 
 illegitimate, for it is not based on a difference in their actual analyzable 
 content, but is itself an inference we draw from their organized relations. 
 Again, the laws of association by which these secondary states are aroused 
 by no means coincide with the inferential laws by which we pass from 
 facts directly known to facts known through proof. The confusion of 
 inference and association which sometimes does arise Mill expressly 
 brands as a serious logical fallacy.^ Sensations and images, in short, are 
 on the same level of immediacy, both directly given in consciousness. 
 They are both mental fact. From this point of view we should have 
 no right to treat memory and expectation as in any sense a means of 
 rising inferentially above the immediate data of consciousness. 
 
 The other mental states which must be included in the immediate 
 content of consciousness Mill classifies as thoughts, emotions, and 
 volitions. These are taken exactly in the same way as sensation, all 
 equally immediate. Emotions and volitions^ we may for our present 
 purpose lay aside. But the mental states which he calls thoughts have 
 grave significance for a theory of inference. 
 
 It is certainly disconcerting to find the class of “thoughts” included 
 as a member of this list. Where is inference to come in if it is not 
 thought? But thoughts too are taken as mere mental facts. “Under 
 the word Thought is here to be included whatever we are internally 
 conscious of when we are said to think.”^ These “intellectual states of 
 
 ^ 59I^^. " 592 ^^. ^ 59^d- 
 
 4 521-23, and elsewhere. 
 
 s A volition indeed is made up of two things: a physical effect, preceded by a 
 “state of mind”; the latter is the “intention to produce the effect,” and that belongs 
 to the content of consciousness (516). 
 
12 
 
 RELATION OF INFERENCE TO FACT IN MILL’S LOGIC 
 
 mind” are just as primitive and immediate facts as sensations, for Mill 
 “recognizes no ultimate premises but the, facts of our subjective conscious- ' 
 ness: our sensations, emotions, intellectual states of mind, and volitions.”^ 
 We are here confronted with a curious and most significant difficulty, 
 which brings into question the whole doctrine of immediacy in Mill’s 
 sense. For certainly inferences, knowledge obtained by reasoning, not 
 only what we see and feel but what we “ fancy that we see or feel,” are 
 all in some sense “intellectual states of mind”; they are part of what 
 “we are internally conscious of when we are said to think.” Yet this 
 thinking, which is surely to be set in contrast with the data of inference 
 as the very process of inference if there is such a thing as inference at all, 
 is in turn swept back into the contents of consciousness, all of which are 
 immediately given as the data of inference. The immediacy of all mental 
 states and the genuineness of inference are thus in flat contradiction. 
 
 In the next place, how does Mill treat relations? Between facts of 
 nature, taken in the objective sense that will come before us below. Mill 
 recognizes the possibility of all sorts of relations; but these are all 
 reducible to the subjective relations of coexistence, succession, and 
 resemblance between mental states, and it is the latter that concern us 
 here. Here we meet with the same ambiguity. Mill wavers between two 
 positions. For these relations between objective facts are reduced sub¬ 
 jectively either to relations between states of consciousness or to states 
 of consciousness themselves. On the one hand, “those relations, when 
 considered as subsisting between other things, exist in reality only 
 between .... states of consciousness”;^ on the other hand, “there is 
 no part of what the names expressive of the objective relation imply 
 that is not resolvable into states of consciousness.”^ The peculiar sig¬ 
 nificance of this ambiguity lies here, that only when these relations are 
 taken as themselves states of consciousness can they be treated con¬ 
 sistently with Mill’s general theory of the subjective immediacy of con¬ 
 sciousness; while only as they are treated as somehow different from 
 states of consciousness, added to them, as it were, or involving reorgani¬ 
 zation of them, is there room for a process of inference that, while resting 
 on immediate intuition, shall build upon it a structure of mediated 
 knowledge. The ambiguity is indispensable to the seeming clearness 
 of the original distinction between intuition and inference. So we find 
 Mill asserting on the one side that “resemblance is nothing but our 
 feeling of resemblance; succession is nothing but our feeling of succes- 
 
 ^ 520& (italics mine). 
 
 2 65^ (italics mine). 
 
 3 Gob (italics mine). 
 
SUBJECTIVE FACTS AS DATA 
 
 13 
 
 sion’'^; and on the other side that “our consciousness of the succession 
 of these sensations is not a third sensation or feeling added to them,”^ 
 Mill simply assumes that, because “to have two feelings at all implies 
 having them either successively or else simultaneously,’’ therefore suc¬ 
 cessive sensations are somehow a sensation of succession; and so with 
 the other relations. How this can be is well acknowledged to be a 
 mystery. “These feelings of resemblance and .... dissimilarity are 
 parts of our nature; .... states of consciousness which are peculiar, 
 unresolvable, and inexplicable.”3 They are “either .... states of 
 feeling, or something inextricably involved therein.’’^ 
 
 The final breakdown of this doctrine of immediacy of consciousness 
 is disclosed when a system of meanings is set up over against the merely 
 passing, given facts. But it will be convenient to postpone a discussion 
 of Mill’s system of meanings till after we have examined his attempt to 
 build objective out of subjective facts. We may then be able to trace 
 more clearly the root of Mill’s difficulty. But it is already evident that 
 there is a confusion in Mill’s conception of given mental facts. Indeed, 
 it appears to involve three distinct errors, perhaps all cardinal errors of 
 the associational school: 
 
 First, all analyzed content of experience is made subjective. The 
 unavowed motive for this is probably as follows: The strain of attention 
 required to analyze any content of experience, and the temporary uncer¬ 
 tainty as to the outcome of the analysis, are used as indications that the 
 content when so analyzed is subjective.^ The effort, for example, to 
 distinguish the baffling shades of green seen upon a meadow, or the evan¬ 
 escent chirps and hums heard in the same meadow, acts as a motive to 
 refer back to subjective consciousness the colors and sounds when they 
 are discriminated. In this way all phenomena when analyzed tend to 
 become psychological, and experience is taken as composed of mental 
 states. 
 
 Secondly, these subjective products of analysis are then regarded as 
 genetically prior to the analysis which brought them to consciousness. 
 Because they are now found there, it is supposed that they must pre¬ 
 viously have been there.^ An experience which analytic attention has 
 succeeded in differentiating into elements is taken as having been pre¬ 
 viously constructed by an association of similar elements. So the course 
 
 1636 . ^6oc. ^6ia. ‘^6sb. 
 
 5 Mead, “The Definition of the Psychical,” University of Chicago Decennial 
 Publications, First Series, III, 23-38. 
 
 6 Royce, Outlines of Psychology, 97-117, has an excellent discussion of this fallacy. 
 
14 
 
 RELATION OF INFERENCE TO FACT IN MILL'S LOGIC 
 
 of experience becomes an agglomeration of original, subjective, internally 
 fixed units. This is sensationalism combined with associationism. 
 
 Thirdly, a succession of logically independent units is expected to 
 constitute a single experience. Mental states as mere facts are wholly 
 evanescent, and are never repeated. How then can there be any accu¬ 
 mulation in the series ? At times Mill is fully aware of the difficulty, and 
 acknowledges as inexplicable the power possessed by the series of mental 
 states of knowing itself as a series.^ But this inexplicability never drives 
 him back to a reconsideration of his first assumption. Leaving the 
 mental facts in their bare givenness and independence, he naively adds 
 to them an ‘‘impulse of the generalizing propensity.”^ Now Mill is of 
 course right in regarding generalization as most significant for the theory 
 of inference. The whole possibility of inference to reach after and before 
 rests on generalization. But the propensity to generalize is a somewhat 
 startling addition to the scheme of simply associated mental states. Is 
 it not in fact a contradiction ? Given a succession of barely presented 
 mental contents, one can suppose an addition of more such contents and 
 more varied contents. But how could generalization get any foothold 
 among them? What could generalization possibly mean in terms of 
 such contents, even including immediate feelings of resemblance ? The 
 truth is that Mill has assumed, side by side with his system of mental 
 states, a system of universal meanings. Were it not for these, inference 
 could be only the piling up of more intuitions. With them, inference 
 finds expression in significant propositions. The mysterious propensity 
 of the mind to generalize is simply the illicit acknowledgment, whenever 
 they are needed, of a system of meanings which connect the bare, passing 
 facts of consciousness and elevate them into an organic experience. 
 
 To recapitulate. Mill starts with a sharp distinction between imme¬ 
 diate intuition and mediate inference. The distinction, however, breaks 
 down, because if the given facts of consciousness are made to include all 
 the contents of experience there is no room for inference, while if we 
 allow an element of inference in some of the facts of consciousness any 
 facts may be equally infected and no data are left whose immediacy can 
 be guaranteed. Mill escapes the pressure of the difficulty only because 
 his treatment of inference is disrupted into two parts: On the subjective 
 level he assigns the locus of inference by asserting its distinction from 
 intuition; but he does not on this level discuss the method of inference, 
 and so escapes the need of a criterion of its distinction from intuition. 
 
 ^ Mill, Examination of Hamilton, I, 260-62. 
 
 2 154^, 227a, h. 
 
SUBJECTIVE FACTS AS DATA 
 
 IS 
 
 On the objective level he elaborates a method of inference, but the imme¬ 
 diate data upon which the method must operate he assumes as already 
 provided on the subjective level. On the one level everything is datal, 
 with no room for inference; on the other, there is a method of inference 
 but no starting-point for its operation. This last assertion can receive 
 its justification, however, only after an examination of Mill’s treatment 
 of objective facts, to which we may now turn. 
 
CHAPTER II 
 
 THE WORLD OF OBJECTIVE FACTS 
 
 The world of facts with which MilFs inferential process actually deals 
 is a thoroughly objective world, conceived in terms of pure naturalism; 
 not at all the world of subjective mental facts which he first assumed. 
 When Mill is expounding its method it is these objective facts upon 
 which inference is made to rest. 
 
 How does Mill reach this world of objective facts? Ostensibly he 
 gets it by a process of complication of the associative connections of 
 mental states. The complications become so involved that the original 
 mental states drop out of sight and the complications alone remain, to 
 form an objective world of things and events. It would be aside from 
 our purpose to criticize this transition in detail, but we may note three 
 fundamental characters of the objective world thus reached by Mill. A 
 statement of these will m^ake clearer the nature of the shift from a sub¬ 
 jective to an objective level. It will then also be observed that they 
 correspond closely to the three ‘‘analogies’’ of Kant. 
 
 In the first place, it is a world of things, existing permanently and 
 independent of individual consciousness. The transition to this concep¬ 
 tion from that of mental states is effected by means of the conception of 
 permanent potentialities of sensation.^ On the Berkeleyan level, where 
 experience of objects is interpreted in psychological terms, sensible things 
 are sensations, and the existence of sensible things means their perception 
 by the mind. Nature is but the aggregate of sensible ideas, and the 
 uniformity of nature is a uniformity in the order of sensations. Now 
 Mill, as we have seen, starts on this level; but by means of complicating 
 associations he rises to an objective level, on which the permanent poten¬ 
 tialities of sensation serve as a core around which phenomena are organ¬ 
 ized.^ Substance is thus conceived as a fixed group or set of potentialities 
 of sensation joined together according to a fixed law. These are not 
 “mere vague possibilities,” but “conditional certainties.’’^ They are 
 
 ^ This transition is briefly described in the LogiCy 53-54. The classical passage 
 in which it is elaborated is in Examination of Hamilton, chap. xi. In both these works 
 Mill uses the term “possibilities of sensation.” In his later Essay on Berkeley he uses 
 the more appropriate term “potentialities.” 
 
 2 Logic, 451&, 456c, 425&-26a. 3 Mill, Hamilton, 238. 
 
 16 
 
THE WORLD OF OBJECTIVE FACTS 
 
 17 
 
 considered objective because of their permanence, their reliability, and 
 their social community. First, they are permanent. “The conception 
 I form of the world .... comprises, along with the sensations I am 
 
 feeling, a countless variety of possibilities of sensation.These 
 
 various possibilities are the important thing to me in the world. My 
 present sensations are generally of little importance, and are, moreover, 
 fugitive; the possibilities, on the contrary, are permanent, which is the 
 character that mainly distinguishes our idea of Substance from our 
 notion of sensation.”^ The object is “invested with the quality of per¬ 
 manence, in contrast .... with the temporary character of each of the 
 sensations composing the groupit “presents itself to the mind . . . . 
 as a kind of permanent substratum, under a set of passing experiences or 
 manifestations.” “The whole set of sensations as possible, form a per¬ 
 manent background to any one or more of them that are, at a given 
 moment, actual.Secondly, they are reliable even when they pass out¬ 
 side the individual’s experience.'^ “The reliance of mankind on the real 
 existence of ... . objects, means reliance on the reality and perma¬ 
 nence of Possibilities of ... . sensations, when no such sensations are 
 actually experienced.“Our sensations we carry with us wherever we 
 go, and they never exist where we are not; but when we change our place 
 we do not carry away with us the Permanent Possibilities of Sensation; 
 they remain until we return, or arise and cease under conditions with 
 which our presence has in general nothing to do.”^ Thirdly, they are 
 socially common. For “more than all ... . they are, and will be after 
 we have ceased to feel. Permanent Possibilities of Sensation to other 
 beings than ourselves.” “The permanent possibilities are common to 
 us and to our fellow-creatures: the actual sensations are not.” “The 
 world of Possible Sensations succeeding one another according to laws is 
 as much in other beings as it is in me,” “it has therefore an existence 
 outside of me; it is an External World.” “This puts the final seal to 
 our conception of the groups of possibilities as the fundamental reality 
 in Nature.”^ For such reasons as these “our actual sensations and the 
 permanent possibilities of sensation, stand out in obtrusive contrast to 
 one another”; and “the possibilities .... come to be looked upon as 
 much more real than the actual sensation; nay, as the very realities of 
 which these are only the ... . appearances. 
 
 * Ihid., 237; Berkeley, 279-81. ^ Logic, 425 <^j d. 
 
 2 Mill, Hamilton, 239. ^ Mill, Hamilton, 244. ^ Mill, Hamilton, 242. 
 
 i Ibid., 241. ^ Ibid., 24g-so. ^ Ibid., 240. 
 
i8 
 
 RELATION OF INFERENCE TO FACT IN MILL’S LOGIC 
 
 These groupings in nature, then, no longer subsist for Mill immedi¬ 
 ately between mental states as such, but are composed of mental states 
 only remotely or indirectly. When we once get them we forget their 
 associational origin and treat them in an entirely naturalistic fashion. 
 
 In the second place, the relation of cause and effect holds, not of sen¬ 
 sations directly, but of these objects. Precisely as substances are lifted 
 from groups of sensations to an objective level, so causal connections 
 between these substances come to replace mere associations between 
 sensations.^ “In addition to fixed groups, we also recognize a fixed 
 
 Order in our sensation.Now, of what nature is this fixed order 
 
 among our sensations ? It is a constancy of antecedence and sequence. 
 But the constant antecedence and sequence do not generally exist between 
 
 one actual sensation and another.In almost all the constant 
 
 sequences which occur in Nature, the antecedence and consequence do 
 not obtain between sensations, but between the groups we have been 
 speaking about, of which a very small portion is actual sensation, the 
 greater part being permanent possibilities of sensation. Hence, our 
 ideas of causation .... become connected, not with .... our sensa¬ 
 tions as actual at all, .... but with groups of possibilities of sensa¬ 
 tion.’’^ 
 
 When this conception of causal laws has been obtained it is supposed 
 to be carried to the farthest limit of generalization, and is treated as an 
 absolute character of the objective world. Every event takes place 
 according to a fixed law of sequence, and the whole universe is law- 
 abiding. Mill attempts to deduce this causal character of the whole 
 course of the universe, somev/hat as he did its independent existence, by 
 a complication of subjective association but here even more obviously 
 than there he must leap to another level before there is any place for the 
 
 ^ Mill, Berkeley, 288. 
 
 2 Mill, Hamilton, 239-40. There is much confusion in Mill as to the place of 
 sensation in the causal scheme. Objective things, as potentialities of sensation, 
 apparently include all actual sensations as that part of their content which happens 
 to be experienced. If so, how can sensations be given a place as distinct links in the 
 causal series ? This is to duplicate them. This is similar to the difficulty in Kant of 
 referring sensations to things in themselves as causes of the sensations, although 
 causation is properly a category only of the phenomenal world, which consists of 
 objectively organized sensations. This confusion in Mill suggests the difficulty found 
 by neo-realists in assigning to consciousness a place in the realistic scheme. The 
 following passages in Mill’s Logic may be consulted in this connection {46a, 496, 51a, 
 $26, 576, 6$b-64a, Sib, 82c, 242b, 451&, 589d-9oa). 
 
 ^ 397-405- 
 
THE WORLD OF OBJECTIVE FACTS 
 
 19 
 
 conception of objective things and laws. Ostensibly a generalization 
 from particular causal connections, it is taken as an absolute assumption 
 by which each of them may be tested. 
 
 That the two characters of the objective world already referred to 
 are really assumptions independent of the subjective basis on which they 
 are ostensibly grounded becomes all the more evident when we consider 
 its third character, which is in a sense the limit to which the other two in 
 combination are pushed. For, in the third place. Nature is taken as a 
 single, unitary whole; its objects, through the interaction of their prop¬ 
 erties, forming a completely interrelated system, whose operations are 
 absolutely determined. “The state of the whole universe at any instant 
 we believe to be the consequence of its state at the previous instant; 
 insomuch that one who knew all the agents which exist at the present 
 moment, their collocation in space, and all their properties, .... could 
 predict the whole subsequent history of the universe.” “The whole 
 series of events in the history of the universe, past and future, is ... . 
 capable, in its own nature, of being constructed a priori by anyone 
 whom we can suppose acquainted with the original distribution of all 
 natural agents, and with the whole of their properties.”^ How remote 
 is such a universe from an accidental agglomeration of given mental 
 states! Mill has certainly passed here completely to an objective and 
 universal level. 
 
 Such is the naturalistic scheme within which Mill, when he presents 
 the method of inferential procedure, looks for the data of inference. 
 While we may find that it evinces internal difficulties which unfit it for 
 this purpose, it certainly offers a sphere for the operation of inference 
 very much more hopeful in appearance than that of immediate subjective 
 states. The objective scheme which Mill has thus obtained bears a 
 striking resemblance to that of Kant.^ The three features to which we 
 have called attention correspond closely to the three analogies of Kant 
 
 ^ 2$oc-$ia. 
 
 * Professor De Laguna, Dogmatism and Evolution, 179-82, calls attention to this 
 resemblance, but in the main rejects it, on the grounds, first, that “the forms of con¬ 
 nection which Mill considers .... are not intuitively known or assured,” and 
 secondly, that “it can never be asserted that a given description of any form of con¬ 
 nection is adequate or final.” But the general causal scheme, which is the factor in 
 Mill corresponding to the Kantian categories, is taken as absolute. Kant, on the 
 one hand, recognizes that specific causal laws must be learned from experience; while, 
 on the other hand, the form of connection, when Mill once gets it, operates in as abso¬ 
 lute a sense in Mill as in Kant. It is “adequate” and “final.” See such passages 
 as 237a, 25oc-sia, 345^^* 547^-48a. 
 
20 RELATION OF INFERENCE TO FACT IN MILL’S LOGIC 
 
 as organizing principles for experience, and the correspondence can in 
 fact be pointed out in considerably fuller detail than the present purpose 
 will permit. Is this resemblance merely superficial, or is it significant ? 
 
 There are of course obvious differences. Mill set out from an 
 empirical standpoint, and intended to remain an empiricist. He repu¬ 
 diated any contribution to experience by the structure of the mind itself, 
 and tried, as we have seen, to trace the growth of even the most fixed 
 forms of nature from the simplest associative connections. Certainly 
 Mill did not mean to be a Kantian. Yet the resemblance is not without 
 significance. Mill has moved a considerable part of the way from Hume 
 to Kant. 
 
 Hume and Kant both assume a congeries of mental states to start 
 with. For both, these mental states are organized into objects. Kant’s 
 principles of organization are independent of the content; they are 
 formal, contributed by the intelligence to the content. Hence the 
 organization of the object has universality and necessity. Hume’s 
 principles of organization reside in the content, and are reducible to the 
 principle of custom, or habit, which operates through contentual laws of 
 association. Such organization of objects cannot have universality and 
 necessity, for it rests on mere expectation, and the actual connections 
 may change at any time. Change of organization, while it may surprise, 
 does not rob the object of what for Hume renders it an object. Now 
 what sort of organization of objects does Mill employ ? He starts with 
 that of Hume, and avows it to the end. But after professing to deduce 
 substance and causality as higher complications of empirical association 
 (or habit based on mere empirical relations of content), he uses these 
 objective principles of substance and causality (together with the prin¬ 
 ciple of an absolutely reciprocal system of nature which he could not even 
 profess to deduce by association) as though they were absolute principles 
 which he can bring to the content of experience. Though he would 
 never have avowed their seat in the constitution of the knower, he 
 assumes their universality and necessity.^ That is to say, he uses them 
 just as Kant did. Kant had shown that universality and necessity of 
 objects must be assumed independently of their content; and in uncon¬ 
 scious accordance with this Mill assumes a universality and necessity 
 that, as we have seen, could not have been properly reached by his asso- 
 ciational construction of substance and empirical proof of causation. 
 To be sure. Mill does not admit the conception of necessity; but he has 
 a double use of the idea of universality, one of which amounts to the 
 
 ^ Compare Joseph, Introduction to Logic, 376 ff. 
 
THE WORLD OF OBJECTIVE FACTS 
 
 21 
 
 same thing. That is universal in one sense which happens to be uni¬ 
 formly experienced, and that is universal in another sense which is uni¬ 
 form in nature whether experienced or not, and probably in no case 
 unambiguously experienced; this latter cannot be distinguished from 
 formal necessity. Indeed, Mill’s unconditional universality,^ with its 
 exclusion of transient force,^ is precisely Kant’s necessity. His scheme 
 of objective nature, then, is Kantian rather than Humian. 
 
 The significance of this conclusion for our purpose lies in this, that 
 the two modes of organizing experience allow a very different place for 
 inference. Hume relies directly on custom, and gets no place at all for 
 inference; mental states simply are or are not; they are all equally im¬ 
 mediate. Kant relies not at all on custom, but believes himself to be in 
 possession of a system of organizing principles independent of experience, 
 which put objects into genuinely objective relations, and seem to make 
 a place for inference regarding them. Now Mill holds substantially each 
 of these positions in turn. He relies indirectly on custom, and so far as 
 he does so has no place for inference, as we saw above; but when he has 
 thus reached his objective principles of organization they then work 
 with a universality unlike that of Hume and like that of Kant. Mill 
 therefore is able to produce a logic which was impossible to pure associa- 
 tionism. And his logic, in practically all of its actual construction, can 
 be regarded as based on a purely naturalistic foundation. After once 
 getting to the level where its procedure is worked out. Mill’s logic is no 
 longer sensational and associational, but realistic, and must be judged 
 on that ground. One motive, it is true, among others in constructing 
 his logic was to vindicate the associational philosophy by showing that 
 a logic could be built upon it,^ but such a vindication would of course 
 have to rest in part upon the cogency of the transition from the subjective 
 to the objective level, as well as upon the adequacy of procedure of a 
 quasi-realistic logic after the objective level has been reached. We have 
 tried to show that the transition is not cogent. The difficulties of the 
 realistic logic to which it leads Mill will appear below. 
 
 * 244C-46C. 
 
 * 236!. 
 
 3 Mill, Autobiography, 226. 
 
CHAPTER III 
 
 THE UNIVERSALITY OF ATTRIBUTES AND PROPOSITIONS 
 
 So far we have considered Mill’s facts, whether subjective or objec¬ 
 tive, as bare particulars. This is entirely fair to Mill’s official and 
 avowed empiricism. But such facts would be of no service to the logi¬ 
 cian; for they would be completely cut off from each other, giving no 
 play for an interpretative process to bind them together. Facts as mere 
 facts are never repeated. Mill accordingly imports into them a universal 
 element by endowing them with attributes. This he does quite naively, 
 not suspecting that in using attributes in a universal sense he has aban¬ 
 doned his original attitude toward facts. 
 
 We have already met with attributes in Mill’s discussion of facts as 
 the data of inference. We must now consider their function more par¬ 
 ticularly. Attributes serve Mill in three ways: first, to reinforce the 
 ultimate identity of subjective and objective facts; secondly, to furnish 
 a bond of unity between facts; thirdly, to give content to propositions. 
 This last function of attributes is consciously avowed by Mill; the 
 service of the two former is somewhat surreptitious. 
 
 In his use of the word attribute Mill glides from one meaning to 
 another. At first the attribute is the quality of the sensation, already 
 giving the sensation a universal aspect.^ The attribute is then trans¬ 
 ferred to the thing, as the power the thing has of exciting a sensation.^ 
 This mode of statement is first suggested very hesitatingly, but later 
 accepted without reserve.^ The exciting of actual sensations at length 
 falls out entirely, and attributes become the mere potentialities of causing 
 them.4 xhe world of nature, with all its objective causal relations, can 
 then be described as a realistic system in terms of attributes, now under 
 the name of properties.^ Thus attributes carry Mill over by an easy 
 transition from subjective to objective facts. 
 
 Again, the attribute is the universal element needed by Mill to bind 
 together his merely particular facts. Attributes give meaning to such 
 facts precisely by their universality, and the particulars could get mean¬ 
 ing in no other way. So Mill’s doctrine of meaning is one of intention, 
 not of extension. “We may frame a class without knowing .... any 
 
 ^ 57^, and many passages. 3 636. 
 
 * SjfsSa, 58c. 4 86a. s 25o6-5ia. 
 
 22 
 
UNIVERSALITY OF ATTRIBUTES AND PROPOSITIONS 23 
 
 of the individuals of which it may be composed; we may do so while 
 believing that no such individuals exist. If by the meaning of a general 
 name are to be understood the things which it is the name of, no general 
 
 name, except by accident, has a fixed meaning at all.The only 
 
 mode in which any general name has a definite meaning, is by being a 
 name of an indefinite variety of things; namely, of all things, known or 
 unknown^ past, present, or future, which possess certain definite attri¬ 
 butes.”^ This seems to give something very like a world of meanings 
 over against a world of facts. The meanings are fixed in themselves, 
 and it is a mere accident whether any corresponding particular exists or 
 not. “General names .... have a meaning, quite independently of 
 their being the names of a class. That circumstance is in truth acci¬ 
 dental, it being wholly immaterial to the signification of the name whether 
 there are many objects, or only one, to which it happens to be applicable, 
 or whether there be any at all.” “ Every name the signification of which 
 is constituted by attributes is potentially a name of an indefinite number 
 of objects; but it need not actually be the name of any.”^ It is quite 
 true that Mill denies to these meanings an objective existence; attributes 
 are not real things.^ Yet the meaning spills over the particular fact, and 
 proves to be an identity above the particulars. Spencer criticizes Mill 
 on this ground, charging him with confounding likeness of attributes and 
 their identity. Mill rejoins that likeness of attributes is identity: “The 
 common something which gives a meaning to the general name, Mr. 
 Spencer can only say, .... is the similarity; .... and I rejoin, the 
 
 attribute is precisely that similarity.The things compared are 
 
 many, but the something common to all of them must be conceived as 
 one.”4 
 
 Mill stands here in a position of unstable equilibrium. For him 
 nothing but particular facts can be actual; his entire empiricism is here 
 at stake. Yet the particular can be expressed in terms of attributes, and 
 each attribute is identical, in however many facts. It is an absolute 
 universal. Even further, as we have seen, an attribute need not actually 
 imply a fact at all. The attributes as such, then, present a timeless 
 system of characters without particularity. 
 
 Mill here betrays the inevitable difficulty of assuming that particular 
 facts are independent and isolated. This is due to a failure to realize 
 
 ^ 78e (italics mine). 
 
 2 946. Cf. 84a. In neo-realistic terminology it may be the name of a subsist- 
 ent, not an existent. 
 
 3 1366. 
 
 4 i37(;; Spencer, Principles of Psychology, 125-27. 
 
24 
 
 RELATION OF INFERENCE TO FACT IN MILLS LOGIC 
 
 that it is impossible to speak of particular facts without implying the 
 presence of a person that has an interest in the facts, and to whom there¬ 
 fore they are not mere particulars. Provisionally and for certain pur¬ 
 poses, to be sure, it is necessary to treat facts as independent; but their 
 independence is still the way they are being treated. The results of the 
 logical process cannot be made valid in absolute independence of the 
 interest in that process. An attribute is attributed. Particulars and uni¬ 
 versal are strictly correlative. Any attempt to get a logic out of mere 
 particulars therefore puts one under compulsion, whether surreptitiously, 
 like Mill, or avowedly, like Bradley, to read into the particulars a sys¬ 
 tem of absolute universals. 
 
 In the next place, it is this universality of attributes that furnishes 
 Mill a content for the proposition. This will appear as we examine his 
 doctrine of propositions. 
 
 Mill uses the term “proposition,” rather than “judgment,” because 
 he regards the latter term as emphasizing too much the mental processes 
 involved. It is the import, not the psychology of the judgment, in which 
 he is interested, and this he thinks is best expressed in the term “propo¬ 
 sition.”^ This choice of terms serves to bring into clearer setting the 
 transition from an associational psychology of particulars to a logic of 
 universals. For the import of a proposition is supposed to be quite inde¬ 
 pendent of belief. It is subject to belief, but is not constituted by belief.^ 
 The import is the universal meaning of what is asserted in the proposition, 
 and is not conditioned by the way in which the assertion arose, or the 
 service rendered by the assertion to the one making it. 
 
 Mill does not fully face the difficulty of reconciling such a theory of 
 propositions with his empiricism. His first analysis of the proposition 
 as an assertion of something about something, upon which he dwells at 
 length,^ takes its cue superficially from the verbal form, and for our 
 purpose it can be passed over. Much more significant is Mill’s account 
 of the proposition as an assertion about facts. The true subject of the 
 proposition is one or more facts.^ Indeed, matter-of-fact and assertion 
 are expressly identified.^ But how can this be ? Mere particular facts 
 as such are not asserted; they merely are. The assertion of a fact surely 
 implies a standpoint outside the fact. How then can facts get into 
 propositions? 
 
 ' 73 - 
 
 ® Contrast 27c with 21&, 73a, 74a, and 75&. 
 3 6^d, and elsewhere. 
 
 4 121&. 
 5 836. 
 
UNIVERSALITY OF ATTRIBUTES AND PROPOSITIONS 
 
 25 
 
 Facts get into propositions by virtue of the universality of their 
 attributes. Matter-of-fact and import of proposition are identical^ 
 because every fact can be expressed in terms of universal attributes. 
 The particular existence of a fact is overflowed by its universal character, 
 and it is the latter that gets recognition in Mill’s proposition. Bradley’s 
 vigorous attack on Mill’s associationism leaves this out of consideration. 
 If Mill had remained a consistent associationist, Bradley’s strictures 
 would apply; and they are no doubt justified in so far as Mill fails 
 explicitly to acknowledge a system of universals. But they ignore the 
 use which Mill makes of the universal character of every fact in his 
 doctrine of import. Mill rides two horses, not always very steadily 
 keeping them abreast; it is not entirely clear, however, that Bradley 
 does better to reduce the two to a double-natured monster, with existence 
 and meaning inside one skin.^ 
 
 The difficulties of the position, if no greater in Mill, are perhaps more 
 obvious, and therefore more easily pointed out. The particular fact, 
 taken as merely given, and as such independent of the knower, has to be 
 matched with a universal meaning, belonging to an absolute system of 
 meanings, with an import also as such independent of the knower. The 
 universal character of attributes has played the trick. Committed from 
 the start to particular facts. Mill finds these facts capable of description 
 in terms of universal attributes, which forthwith give content for univer¬ 
 sal propositions. Such is the transition from givenness to import, from 
 fact to proposition. 
 
 The import of propositions assumes a system of time relations, a 
 system of space relations, and a system of qualitative universal attributes. 
 Different attributes may evidently bear to each other time and space 
 relations, and an identical attribute may appear in different times and 
 spaces. These assumptions were not explicitly systematized by Mill,^ 
 but they lie at the basis of the three fundamental relations of matter-of- 
 fact asserted in propositions: coexistence, succession, and resemblance.-^ 
 To these Mill adds two other relations: existence and causation. Exist¬ 
 ence is in the last analysis apparently reducible to the others. Causa¬ 
 tion is a very special form of succession into whose significance we must 
 inquire later, when we consider the problem it presents to inference. 
 
 " 836, 8s&. 
 
 ^ Bradley, Pvinciples of Logic, in various relevant passages; see especially Book II, 
 Part II, chaps, ii and iii. 
 
 3 As they have been, for example, by the neo-realists. 
 
 4 80-85. 
 
\ 
 
 26 RELATION OF INFERENCE TO FACT IN MILL’S LOGIC 
 
 The whole import of propositions reduces at last to accompaniment of 
 attributes.^ 
 
 In this view of the import of propositions Mill seems to have brushed 
 aside the particulars in nature and to have grasped its universal laws. 
 Nature operates always in accordance with ultimate laws that cannot 
 fail to be universal. A complete statement of them would be a complete 
 account of nature. Our propositions are attempts to approximate such 
 a statement. We are not sure that any of the uniformities with which 
 we are yet acquainted are ultimate laws; but we know that there must 
 be ultimate laws; and that every resolution of a derivative law into more 
 general laws brings us nearer to them.”^ The whole course of nature, 
 it seems, is statable in terms of propositions, and the business of infer¬ 
 ence is to approximate as closely as possible to such a statement. It 
 matters not what prompted the statement, for the psychology of belief 
 is irrelevant; the statement if correct expresses an accompaniment of 
 attributes as it is in nature, entirely objective, wholly determined, and 
 absolutely universal. If nature in one aspect consists solely of par¬ 
 ticular facts, in another aspect it consists of universal laws. In his scheme 
 of inductive inference Mill attempts to bring these together, though the 
 result is rather to oscillate between them than to harmonize them. 
 
 ^ 80c, 2>$d. 
 " 345C- 
 
CHAPTER IV 
 
 INFERENCE AS ANTICIPATION OF FACTS 
 
 There are two ways of looking at inference. To distinguish them 
 will give us a point of view from which to criticize Milks account of infer¬ 
 ence. On the one hand, inference may be viewed as a process, whereby 
 knowledge is genuinely increased and there is an actual reconstruction 
 of content. On the other hand, it may be viewed as a relation within 
 content, by which one part of the content implies another. From the 
 first point of view, in describing any case of inference we are bound to 
 consider the state of knowledge of the person that makes the inference, 
 both before and after the inference is made, and to see what has thereby 
 happened to his knowledge and why it happened. Inference thus con¬ 
 sidered is the reconstruction of a specific situation. If, in the light of 
 the knowledge obtained by the inference, we then try to state by what 
 right the inference was made, we pass over into the other view of infer¬ 
 ence. The inference is then no longer a reconstruction of content, but 
 a relation within content. Now the formal analysis of inference always 
 ignores the process of reconstruction. It is bound to take its stand in 
 the light either of the lesser knowledge preceding the inference or of the 
 greater knowledge succeeding the inference; usually the latter. In 
 either case the inference as a movement of thought disappears. For in 
 the earlier content taken just as it stood there was no implication corre¬ 
 sponding to the inference; that is, in the light only of the knowledge 
 previous to the inference the inference cannot be seen as valid, for if it 
 had then appeared as valid it would already have been made. But in the 
 light of the fuller knowledge after the inference, the inference would not 
 need to be made; so that when the content of the later knowledge is 
 analyzed to find an inferential relation within it the inference appears 
 to be tautologous. Taking inference then as a relation within a fixed 
 content, it must be either invalid or tautologous; the premises either 
 involve the conclusion or do not.^ In the study of inference there is of 
 course a place for formal analysis; it need not mislead if, while we pro¬ 
 visionally take that point of view, we expect the inference as analyzed 
 
 ^ “If the Syllogism yields novelty, it begs the question. If it disclaims novelty, 
 it becomes vain repetition. As a form, therefore, it is either futile or false” (Schiller, 
 Formal Logic^ 208). This applies to all formal proof. 
 
 27 
 
28 
 
 RELATION OF INFERENCE TO FACT IN MILL’S LOGIC 
 
 to appear either tautologous or invalid. We can at once escape this 
 dilemma by returning to the other point of view, from which we con¬ 
 sider the inference as it was actually carried through. To do this with 
 full consciousness of what it means would reconstruct the formal analysis 
 of inference, and would relegate it to a subordinate place. 
 
 It is to the credit of Mill that in his theory of inference and its rela¬ 
 tion to fact he set out to introduce a logical reform of this kind. He 
 proposed a logic of truth instead of a logic of consistency.^ He made 
 inference pass from the known to the unknown, a process of discovery. 
 It was to proceed from particular facts, and to lead to particular facts. 
 Mill certainly was feeling for the concrete process of inferential recon¬ 
 struction. Yet in detail he failed to maintain this attitude. His con¬ 
 ception of facts as definitely given interposed itself, and prevented him 
 from appreciating the concrete and really significant reconstructive 
 function of inference, so that he fell back into the formal attitude toward 
 inference as a relation of content. Endeavoring then to extricate him¬ 
 self from this formal position, and to recover the reconstructive process 
 of inference which he had lost, he comes upon the experience of actual 
 discovery, but regards it for the most part as only subsidiary to inference, 
 and even to the end fails to grasp its real function in inference and its 
 true relation to fact. 
 
 In this way Mill oscillates between three quite different conceptions 
 of inference. The conception with which he sets out is that of antici¬ 
 pation of experience. On the ground of given facts other facts are 
 anticipated. His formula for such inference is “from particulars to 
 particulars.” The establishing of this as the character of inference Mill 
 believes to be a vindication of empiricism.^ A universal need not inter¬ 
 vene between the particular facts on which the inference is grounded and 
 the particular facts to which the inference points; if a universal does 
 intervene it is accidental, merely a convenient safeguard against error. 
 Nevertheless it appears that a universal always could be made to inter¬ 
 vene, and if an inference is justified at all, then by the same token a 
 universal is justified. Thus Mill comes to the second conception of 
 inference as the proof of universal propositions. The main body of his 
 inductive technique is built on this conception. But inference as the 
 proof of propositions cannot finally leave out of account the source of 
 propositions, the way in which they come to be proposed for proof. 
 Inference, in short, is variously considered by Mill as the anticipation 
 
 " 25, 236c. 
 
 2 Liard, Les Logiciens Anglais Contemporains, 5th ed., 24-27. 
 
INFERENCE AS ANTICIPATION OF FACTS 
 
 29 
 
 of particular facts, the proof of universal propositions, and the discovery 
 of universal propositions, though Mill seems unaware that he has shifted 
 his ground in passing from one to the other. Mill is not at fault, indeed, 
 in holding to these three conceptions as genuine aspects of the inferential 
 process; his error is in treating each in isolation from the others and in 
 misconceiving the relation of each to facts. Let us examine each of the 
 three in turn. 
 
 Mill finds inference generally acknowledged to be of two kinds: 
 inference from particulars to generals, and from generals to particulars, 
 or induction and ratiocination, the latter being expressed in the syllo¬ 
 gism.^ Mill upholds a third type of inference, not only as valid, but as 
 the foundation of both the others.” Indeed “all inference is from par¬ 
 ticulars to particulars.”"' A universal may or may not intervene; if it 
 does, it merely breaks up the inference from particulars to particulars 
 into two parts, the first ending with the universal, the second beginning 
 with it. The actual inference is always from particular facts to other 
 particular facts, and whether it passes through the universal on the way 
 or leaps directly from facts to facts without a universal does not alter 
 the nature of inference as such. The general proposition is merely a 
 register of inferences already made from the facts, or a formula for 
 making more. 
 
 The doctrine then is this, that inference is in its nature a passage 
 from particular facts to other particular facts^ without the necessary 
 mediation of a universal. There are several fundamental difficulties in 
 this conception of inference. 
 
 First, there is such a process in experience, but it is unreflective, and 
 cannot properly be called inference. It is habit. When Mill offers 
 examples of inference which shall be purely from particulars to par¬ 
 ticulars, he finds himself reduced precisely to cases of habit."* The 
 avoidance of fire by the child, and by the dog as well, and the skilful use 
 of weapons and tools—these do involve anticipations of experience on 
 the basis of previous experience, but they are not logical processes. In 
 some cases they may have grown out of an earlier inferential process, but 
 they need not have done so. They may rest on purely unreflective asso¬ 
 ciation. The mother who prescribes a remedy to a neighbor’s child 
 because it cured her own Lucy^ can scarcely be supposed not to make 
 some use of a “general proposition”; but if she does not, as Mill assumes, 
 
 ' I2SC. 
 
 * 146c. 
 
 3 Or to one other. 
 
 4 1426?, 143a, c, i44<^> b. 
 s 143&. 
 
30 
 
 RELATION OF INFERENCE TO FACT IN MILVS LOGIC 
 
 then she has made no inference at all, but repeats an operation under 
 repeated circumstances by mere force of habit. 
 
 Secondly, when action rises above mere habit and becomes reflective, 
 a universal is required. A particular, if reflectively taken as a particular, 
 implies a universal. Mill begs the question when he finds himself bound 
 to call the particular facts “ cases. An inference from facts to facts is 
 justified only if both sets of facts are seen to be cases; cases, that is, to 
 which the same rule will apply; cases of the same universal.^ 
 
 The emphasis on the universal, it is true, may be shifted either to the 
 beginning of the inference or to the end The universal may then be 
 alternately located at either end of the process. If the cases from which 
 one infers are considered as particular, then any case to which one infers 
 is only one of an indefinite number of possible cases, or is the application 
 of a law. If the cases to which one infers are considered as particular, 
 then they are anticipated on the basis of the law which was exemplified 
 in the cases from which one infers. That is, one infers either from par¬ 
 ticulars to a law which includes other particulars or to particulars from 
 a law that is exemplified in previous particulars. One may be more 
 concerned under given conditions with the origin or with the application 
 of the law, but cannot dispense with it as essential to the inference. 
 Mill seems to do so only because he oscillates from one end of the process 
 to the other, in each case leaving the universal to operate at the opposite 
 end. In the one case the general proposition is a register of observations, 
 and the inference is still to take place in the other case the inference 
 has already taken place, and the general proposition is a formula for 
 making more inferences like it.^ Between the two the inference plays 
 successful hide-and-seek with the universal. 
 
 That inference cannot dispense with a universal is due to the hypo¬ 
 thetical character of the latter. The universal cannot itself be experi¬ 
 enced, and is never a mere record of past experience. It represents a 
 hypothetical interpretation of any future experience so far as it can be 
 taken in terms of the past experience.^ Past experience points to like 
 future experience if it is really the same. If the unobserved particular 
 is a similar ‘‘case,” the inference is valid. The transition through the 
 “if” signifies the mediation of a universal. 
 
 ^ 1256, i^id, 149a. 
 
 2 “A general truth is but an aggregate of particular truths; a comprehensive 
 
 expression, by which an indefinite number of individual facts are afl&rmed and denied 
 at once” But “aggregate” and “indefinite” are contradictory, and this 
 
 contradiction lurks everywhere in Mill’s treatment of “cases.” 
 
 3 141C, 147&, 1516. 142a, hy 148&. 
 
 s 1506. 
 
INFERENCE AS ANTICIPATION OF FACTS 
 
 31 
 
 What do we mean by the universal in inference, and why is it needed ? 
 It is true that all problems and all solutions and therefore all inferences 
 are specific. Had Mill, by the use of his formula “from particulars to 
 particulars” intended to support this aspect of inference, he might well 
 have pressed it farther than he did. But he omitted the reflective ele¬ 
 ment of inference, the necessity we are under to carry with us a set of 
 possible interpretations of new experience. This interpretative tendency 
 is the universal; it is what renders inference reflective where habit would 
 be unreflective. In actual use the universal has no absolute meaning. 
 Absolute meanings are fictions of the logician. When we interpret what 
 we take to be the new “cases” in “the light of” the old, that light is the 
 universal.^ It is the deliberate readiness to treat them, when under the 
 pressure of uncertainty, as cases. General language has crystallized 
 these interpretations into a highly elaborate system, with its own tech¬ 
 nique; but the relative independence and permanence of the language 
 must not be allowed to obscure the essentially instrumental character 
 of the universal which the language expresses. It is here that Mill failed 
 to free himself from the conceptions of formal logic, and let go the very 
 factor which would have revitalized his system: the play of the universal 
 as the reflective interpretation of new experience in the light of the old. 
 
 Thirdly, Mill’s attack upon the syllogism as a petitio principii is due 
 to this failure to appreciate the interpretative function of the universal 
 in the actual process of inference. All inference, if treated as he treated 
 the syllogism, becomes a petitio; for the syllogism avowedly expresses 
 the content of the inference after it has been made, and is therefore 
 tautologous. All inference, when passed through and looked back upon, 
 must appear tautologous; if its content did not admit of tautologous 
 statement it would be invalid. Mill’s strictures apply equally well to his 
 own inferences from and to particulars. Are the earlier and later facts 
 both “cases” ? If they are not, the inference is not valid. If they are, 
 then no inference is needed; for the later cases merely add themselves 
 to the earlier as more of the same. Just so long as Mill takes the content 
 of his particular facts in the same fixed and unquestioned way in which 
 he takes the premises of the syllogism, all actual inference is inevitably 
 shut out. 
 
 * Through the collision and breakdown of habits their content becomes reflective. 
 This reflective content of what was a habit is now a universal. We continue to gain 
 universals in this way; but after this type of experience is once established the system 
 of universals may be built up in other ways, as by verbal instruction. The materials 
 of the universal must always have arisen out of what was unreflective. 
 
32 
 
 RELATION OF INFERENCE TO FACT IN MILL’S LOGIC 
 
 Fourthly, this suggests the difficulty that in Mill’s scheme of inference 
 the fundamental function of inference in meeting an attitude of inquiry 
 is ignored/ Mill takes the knowledge of certain facts and proposes to 
 infer other facts from them. If this is merely the addition of more facts 
 it is not inference, as we have seen. It is inference only if the facts are 
 not adequately known. In any actual case of inference the primary 
 problem is to ascertain what are the facts. We should have no occasion 
 to think of facts in a situation that—so to speak—was all fact. A 
 ‘‘question of fact” emerges because the situation contains more than 
 mere fact; it contains (i) certain contradictions or difficulties that impel 
 to further inquiry about facts; ( 2 ) various more or less tentative con¬ 
 structions which aim to clear away these difficulties. 
 
 Here we strike the root of the matter. Mill has a favorite formula 
 for inference which describes it as procedure from the known to the 
 unknown.^ This formula may no doubt be given an interpretation which 
 renders it a correct description; but it is quite as correct to say that 
 inference is procedure from the unknown to the known. When facts 
 are adequately known there is no occasion for inference. But when our 
 knowledge of the facts breaks down, and it is not known just what the 
 facts are, then inference comes in as an effort to interpret the facts in 
 such a way that they shall be adequately known. Inference, one might 
 say, aims to transform the unknown into the known. The relation 
 between inference and facts is never so simple or so one-sided as Mill 
 supposes; it is a relation of mutual dependence. 
 
 In view of this attitude of inquiry that underlies all inference it is 
 necessary to reject in principle the sharp distinction Mill maintains 
 between the earlier cases which justify the inference and any later case 
 to which it is applied, though in the practical technique of inference the 
 distinction has a relative value. The facts which prompt the inference 
 are first determined as “cases” by the inference, which is designed to 
 meet one of the two questions: “0/ what are they cases ?” or, ^^Are they 
 really cases ?” They are therefore the very facts to which the inference 
 applies. It is true of course that an inferential situation may be ex¬ 
 tremely complex, some factors in it more fixed, others more fluid; and 
 at successive stages of the process it may be useful to consider some the 
 data and others the conclusion. Yet in principle Mill’s sharp separation 
 of the data and the conclusion must be denied. In inference we do not 
 
 This consideration has been well worked out by Professor A. W. Moore, of 
 Chicago, in a seminar on “Modern Logical Theories,” to which the writer is greatly 
 indebted. 
 
 2 1256, 126a, 139a, 2ioe, 223^-24^. 
 
INFERENCE AS ANTICIPATION OF FACTS 
 
 33 
 
 first get a set of facts, ticketed and fully known, to serve as data of an 
 inference, and after an interval another fact conveniently ticketed to 
 receive the application of the inference. That, as we have seen, would 
 be a fair description of the operation of habit, if under knowledge of any¬ 
 thing we were to include any ability to deal with it. But it is not a 
 correct description of inference. In inference the data and the con¬ 
 clusion must come together in organic relation, and it is precisely at the 
 point where they do come together in a single interpretation that infer¬ 
 ence takes place. 
 
 In short. Milks account of inference from particulars to particulars, 
 which actually applies better to habit, implies two situations, an earlier 
 and a later. He oscillates between these, because he seems vaguely to 
 have perceived that in either of them inference may appear. There is 
 no inference unless some factor in the situation is brought into question, 
 but the question may turn upon the formulation of a new universal or 
 upon the selection of an old universal. In both formulation and selec¬ 
 tion of a universal the “cases” are seen to be such only in relation to the 
 universal, but in the one the question looks toward the universal, in the 
 other it looks from it. This is to say that either the major or the minor 
 premise of a possible syllogism may be placed in question. Mill is 
 right in breaking inference from particulars to particulars into two parts; 
 but wrong in considering the universal an unessential division-point 
 between them. He is right in excluding inference sometimes from one 
 part and sometimes from the other; but wrong in denying that it may 
 be in either. 
 
 Indeed, Milks position is shifted when he comes to consider the tech¬ 
 nique of inference. He virtually recognizes the necessity of the universal 
 when he acknowledges that any data which justify an inference to a 
 particular will justify an inference to a universal as well,^ and that the 
 validity of the inference can be shown only by drawing the universal 
 explicitlyThenceforth in his discussion the process of inference is 
 made to consist in the establishing of general propositions. He might 
 well have explicitly taken the same step with reference to the second half 
 of his original type of inference, in the selection of a universal to explain 
 a given situation. He admits this to be often a most arduous intellectual 
 process, but refuses—though not always consistently—to grant it the 
 status of inference,^ on the ground that it is invention, and “invention 
 cannot be reduced to rule.” These two phases of inference remain to 
 be considered in the next two sections. 
 
 * 148^-49^, 155c, 208a. * iS4^f“55^- ^ 208-9. 
 
CHAPTER V 
 
 INFERENCE AS PROOF OF PROPOSITIONS 
 
 The distinction just noted between the establishment and the selec¬ 
 tion of a universal proposition enables Mill to pass over to a second con¬ 
 ception of inference. Before examining the latter it will be well to note 
 again the exact point of the transition. Mill has established, he believes, 
 the general character of inference as a passage from certain particular 
 facts to certain others. But if these certain facts justify the conclusion 
 about any others at all, they justify the conclusion about any others 
 whatever of the same description. Accordingly from premises which 
 justify any conclusion we can always draw a general conclusion. The 
 problem of application, the problem, namely, whether the later facts are 
 of the same description, or of what description they are, is thrown out 
 as irrelevant to inference, because it cannot be reduced to rule; it is 
 apparently too psychological. This leaves for inference only the prob¬ 
 lem of drawing the general proposition from the data, the problem of 
 induction. 
 
 But this is not all. There is a still further limitation in this concep¬ 
 tion of inference as induction. It is not the discovery of general propo¬ 
 sitions but their proof with which he concerns himself. The type of 
 question which his machinery of induction is designed to solve is not, 
 ‘‘What proposition is true?” but rather, “Is a given proposition true?” 
 Mill set out, it is true, with the intention of providing a logic of discovery,^ 
 of the process of advancing from known truths to unknown.^ Logic is 
 to investigate the “pursuit of truth,”^ of “how we come by that portion 
 of our knowledge (much the greater portion) which is not intuitive. ”4 
 But he makes this synonymous with “estimation of evidence,”^ and falls 
 back at once into a formal logic of proof.^ “Our object,” he says, “will 
 be .... to ... . frame a set of rules or canons for testing the suffi¬ 
 ciency of any given evidence to prove any given proposition. ”7 
 
 Why did Milks conception of inference become so completely that of 
 proof ? There were probably two reasons, both suggested in the words 
 just quoted. The first is extrinsic to the character of his theory, and is 
 due to his desire to provide, in a logic of induction, a set of “rules and 
 
 ^ 208a. 3 iqJ. S 235. 
 
 ^ 23&. 4 27&, 51a. 6 121(1, 2lh, 45(Z, 886, 20M. 7 23c. 
 
 34 
 
INFERENCE AS PROOF OF PROPOSITIONS 
 
 35 
 
 canons” that would emulate those of the formal deductive logic in which 
 his father had given him so rigid a training; Mill certainly took great 
 satisfaction in the formulation of his “inductive methods.”^ The other 
 reason is intrinsic to the theory itself, which drove him back to “given 
 evidence” for “given propositions.” Facts for Mill are always given, 
 whether subjective facts of immediate consciousness or objective facts 
 of the system of nature. The proposition is likewise given, for it is the 
 universal meaning of the facts, to be read off from them as they come 
 together. 
 
 The universal propositions the inductive proof of which constitutes 
 inference are of various kinds. They include the statement, for example, 
 of spatial and numerical relations.^ It would be interesting to trace 
 out the relation of inference to matter-of-fact in these, and to inquire 
 whether it is consistent with Mill’s general treatment of induction. That 
 treatment, however, is based almost entirely on the establishment of 
 causal relations,3 and the present criticism of Mill must be limited to this 
 field. 
 
 The character of induction as proof is rooted in Mill’s realistic scheme 
 of nature, (i) Nature consists, as we saw, of certain things, having 
 certain attributes and in certain collocations. ( 2 ) Every change takes 
 place according to a uniform law of succession, each event following and 
 preceding certain other events according to such a causal law. ( 3 ) Nature 
 is a single system of such things and events. This whole system of nature 
 is conceived as definitely determined and proceeding on its way inde¬ 
 pendently of US.4 To be sure, we are not acquainted with all of it, but 
 we are certain that it is all of this character. When we wish to know 
 what the laws of sequence in nature are we must watch the shuffle of 
 facts and observe which go together. Our experiments may vary the 
 shuffle, but the principle of noting the shuffle is the same. The very 
 possibility of identifying the facts in the shuffle rests upon the universal 
 character of their attributes, and the assertion of the accompaniment of 
 certain attributes is a general proposition. It would seem, then, that 
 discovery and proof are essentially identical. We find accompaniment 
 of attributes so far as, within the field of our observation, there is accom¬ 
 paniment of attributes. The facts that teach us the law prove the law. 
 
 ^ Mill hoped that his Logic might serve as a Novum Organum of real practical 
 value. See Logic, 579-80; Autobiography, 226. Cf. Joseph, Introduction to Logic, 342. 
 
 ^ 234^. 
 
 3 2zsb-d, 236c. 
 
 4 These are the “three analogies” in Mill, as noted above. 
 
36 RELATION OF INFERENCE TO FACT IN MILUS LOGIC 
 
 This would seem to be a fair statement of the general conception 
 of proof which underlies Mill’s theory of induction. But the matter is 
 not so simple as it seems. There are grave difficulties raised by it. 
 Partially recognized by Mill, they compel him to introduce deviations 
 in the procedure of induction. Fully faced, they would lead to a recon¬ 
 struction of its underlying principles. Three such difficulties require 
 our consideration: 
 
 First, the inductive methods are all recognized by Mill to be so many 
 means of elimination.^ But elimination is not proof; it is disproof. 
 Elimination must obviously operate upon alternatives; and the very 
 point of the elimination is that neither of the alternatives is proved till 
 at least one other is disproved. There must be a tentative consideration 
 of various apparent uniformities in order that the methods shall have 
 play in the rejection of those which are actually non-uniformities.^ 
 
 Mill might be supposed to answer that this is just what nature does 
 for us. Out of the chaos of sequences, some accidental and some uni¬ 
 form, the accidental are eliminated by the circumstance that they do 
 not recur, which is the principle of the methods, ignoring for convenience 
 their finer complications. But there is no elimination whatever in nature. 
 The facts are all there, and they stay there. Nature is wholly careless 
 whether she presents uniformities or non-uniformities. It is the reflect¬ 
 ive attack upon nature that uses the methods, and their eliminative 
 procedure is exercised upon the tentative efforts to interpret the facts; 
 that is, to find out what the facts really are. Mill’s proofs are disproofs: 
 disproofs of propositions not given ready-made by the facts that furnish 
 the disproofs. 
 
 Where, then, do the propositions come from? Whence arise the 
 assertions about the facts, the interpretations of fact, which are then 
 proved or disproved? To ask such a question, it must be replied, is 
 very likely to imply just Mill’s misconception. We have no system of 
 facts, complete and ready-made, with which to check up our interpreta¬ 
 tions for elimination. There is a constant interplay between fact and 
 interpretation, each reconstituting the other. The only facts we have 
 are the facts as we have successfully interpreted them, and our interpre¬ 
 tations always arise as tentative statements of the facts. Again, the 
 operation of proof and disproof is strictly correlative with the presenta¬ 
 tion of propositions to be tested out; neither can be supposed without at 
 least a tentative form of the other. Nature no more offers a series of 
 
 ^ 278J, 279e-8oa, 281a, 310&. 
 
 = I.e., which are not unconditionally invariable. 
 
INFERENCE AS PROOF OF PROPOSITIONS 
 
 37 
 
 prepared ‘‘evidences” ready to apply to any given proposition than she 
 provides the propositions ready-made. Facts are not all in nature, 
 inferences all in us. Facts likewise grow in the soil of inquiry, and are 
 the fruits quite as much as the roots of our inferential control of nature. 
 
 The second difficulty is due to the impossibility of determining the 
 limits of a fact out of relation to the character of the inquiry about the 
 fact and the proposed interpretation of the fact. It is the difficulty, in 
 any inquiry regarding cause and effect, of setting the limits of an 
 event.^ Mill is already feeling this difficulty when he draws his distinc¬ 
 tion between causation as practically and philosophically understood.^ 
 Causation, according to his own program, is to be understood in the latter 
 way; that, he holds, is the conception underlying genuinely scientific 
 inference, and is based on the facts of nature as they really are. For 
 practical purposes, indeed, we are content to note a selection of the 
 conditions of an event and to call them the cause, or to take two events 
 whose succession is more or less mediate and call them cause and effect, 
 without taking account of all the possible counteracting conditions that 
 might have intervened. But in philosophical strictness we should do 
 neither the one thing nor the other. Philosophically, causation must 
 admit no omissions and no loopholes.^ Now, having set this up as a 
 demand, what Mill actually does in various illustrations is to contrast 
 more nearly philosophical with less nearly philosophical statements of 
 concrete cases of causation, but all his illustrations are statements of 
 approximation only; they never exhibit a single case of causation in 
 the full “philosophical” sense. And they do not because they cannot. 
 « 
 
 ^ The best discussion of this difficulty seen by the writer, though without special 
 reference to Mill, is in Sidgwick, The Process of Argument, chaps, x and xi. A few 
 sentences must be quoted. “ The question where does an occurrence end and ‘ another ’ 
 begin is unanswerable except by an artificial distinction drawn to suit our practical pur¬ 
 poses.” “This kind of analysis may be carried much farther. Given any two steps 
 in the series, we know that some occurrence comes between them, however short the 
 interval, and that this intermediate occurrence .... has just as much right to be 
 classed as belonging to one end of the chain as to the other. It is not properly a chain, 
 in fact, but a stream, or a continuous growth like that between bud and flower.” 
 “No one seriously denies that the links are there, but our practical needs do not always 
 compel us to explore them.” “The separation into Cause and Effect is done for a 
 purpose, and its value depends on its serving that purpose. Whenever a hitch occurs 
 we are bound to look more closely into the details of the connection.” “The line we 
 draw between antecedent and consequent is an artificial one; they are more or less 
 ill-defined parts of a whole which it suits us to pull asunder.” 
 
 2 240c, 241a. 
 
 3 Logic, Book III, chapter v as a whole. 
 
38 RELATION OF INFERENCE TO FACT IN MILL’S LOGIC 
 
 For if Mill were to be pressed on the point, he would find himself 
 completely locked in by the Kantian antinomies. The continuity of 
 time and space renders all change continuous, and there is nothing in 
 the facts taken as they merely “are”—assuming they could be so taken 
 —that would ever indicate the limits of an event, either serially or in 
 section. The continuity of all process in space and time, and perhaps 
 also in qualitative degree, renders the singling out of any “event” pos¬ 
 sible only when relevant to a purpose. Mill admits this for “practical” 
 causation, but he has to assume it covertly for all causation.^ 
 
 The point may be restated thus: Mill reduces the uniformity of the 
 world to a network of particular uniformities of sequence between a 
 group of antecedent phenomena and one consequent phenomenon. This 
 is a correction of the common, crude view (in terms of which, however. 
 Mill often speaks) that every consequent is the result of some one ante¬ 
 cedent, each phenomenon being taken as though an obviously definite 
 unit. But the parts of the world and their properties are more closely 
 intertwined than we crudely suppose. So Mill corrects this view in two 
 respects: First, he insists that in the cause we should enumerate all the 
 conditioning antecedents; that is, all the antecedents whose combina¬ 
 tion is required that the effect may follow. All the conditions, positive 
 and negative, are absorbed into the cause. Secondly, he insists upon 
 the immediate succession of antecedents and consequent. There must 
 be no intermediate steps between. The reason of this is that there may 
 be unsuspected conditioning factors, and with the longer interval and 
 the more opportunity we give unsuspected factors to operate the more 
 likely it is that the expected sequence will not hold. A plot which is to 
 culminate in a year is more liable to be frustrated than one which is to 
 culminate in a day. Mill, it is true, does not correct the popular idea 
 of effect as he does that of cause. He still selects some one consequent 
 which he calls the effect, and thus gives room for the plurality of causes.^ 
 Theoretically this is an inconsistency. An effect is just as complex as a 
 cause, and if we enumerated all its elements the plurality of causes would 
 be impossible. Practically, of course. Mill is quite right; there would 
 
 ^ The supposition of “points’’ and “instants” is a mathematical refinement 
 intended to provide a union of continuity and discreteness by bringing space and 
 time into correspondence with the filled number system. But when we try to apply 
 the formulae based on this supposition, the question of what points and instants to 
 start with raises exactly Mill’s “practical” problem over again, and cannot be met 
 out of relation to a purpose which is dealing with the system of points and instants. 
 
 2 See an acute discussion of plurality of causes in Hobhouse, Theory of Knowledge, 
 457-62. 
 
INFERENCE AS PROOF OF PROPOSITIONS 
 
 39 
 
 usually be no gain in looking for all the elements in the effect. We are 
 not usually so interested in the reasons of things as in how to bring about 
 a particular result. But philosophically speaking, as Mill would say, 
 we have a right to hold him to just as wide an extension of the effect as 
 of the cause. Now when we try to make this view of causal sequence 
 theoretically correct, how are we to decide which of the antecedents and 
 consequents are relevant and which are not ? A finite mind can never 
 grasp all the antecedents. Unsuspected factors constantly enter, and 
 may at any time prove important. Those which we commonly reject 
 because too trivial or remote have an influence upon the effect which 
 though at first perhaps infinitesimal may in time become vast. Theo¬ 
 retically, therefore, we should enumerate all the antecedents and conse¬ 
 quents in the universe. Could we do this, we might say with confidence 
 that were this cause ever repeated its effect would be repeated, and this 
 without reference to the immediateness or remoteness of the cause and 
 effect. This is the uniformity of sequence carried in one direction to its 
 theoretic limit. Within this limit there must be uncertainty and imper¬ 
 fection in our grasp of causal laws. But this limit if theoretically perfect 
 is obviously useless. Again, uniformity of sequence might be carried 
 to theoretic perfection in another way. If the antecedents, as we have 
 seen, are perfectly known, the certainty of the sequence will be absolute, 
 no matter how great the interval between antecedents and consequents; 
 there need be no time limit. The reverse is true, that if the antecedents 
 are not completely known the probability of the sequence is less as the 
 interval between the antecedents and consequents is greater. If, how¬ 
 ever, we bring them into relatively close proximity the probability of 
 sequence becomes relatively great. The limit is reached when the time 
 interval has become infinitesimal. We have then a mere tendency in the 
 antecedents which awaits no interval in which it might be interfered 
 with, and is therefore perfectly certain however small we make the group 
 of antecedents, down to the infinitesimal limit. This result, again, is 
 theoretically perfect but practically useless. Mill is locked up within 
 the dilemma of a construction of causation which is either theoretically 
 incorrect or practically of no avail.* 
 
 Indeed, this is an underestimate of the difficulty. We cannot call 
 the limiting forms of the causal scheme even theoretically correct; they 
 are entirely meaningless. It must be remembered that the problem of 
 causation is to find uniform sequences, and it will be no solution simply 
 
 * Venn, Empirical Logic^ 47-7gives an excellent analysis of the antinomy of 
 causation in Mill. 
 
40 
 
 RELATION OF INFERENCE TO FACT IN MILL’S LOGIC 
 
 to eliminate the problem. If now the absolute certainty of a sequence 
 is assured by screwing the antecedent and consequent so close together 
 that no interval is left in which the sequence might be counteracted, the 
 time relation has disappeared, and this destroys the sequence itself whose 
 certainty it was sought to safeguard. If in the other direction the whole 
 universe is included in cause and effect, the time limit spreads out to 
 infinity, the history of the universe becomes a single event, and sequence 
 is once more destroyed. Theoretically considered, then. Mill’s view of 
 nature as a set of given facts in causal relation breaks up into antinomies 
 when pressed in either direction. We are compelled to return, as Mill 
 did in his actual examples, from the philosophical to the practical view 
 of causation. But to do so requires that we have a criterion of what 
 shall be taken as a single fact, and such a criterion is not provided in the 
 facts themselves in their mere given aspect. In causation practically 
 considered such a criterion is provided; it is found in the personal rela¬ 
 tion of the inquirer to the situation. As this criterion is relative to the 
 inquirer. Mill rejected it as practical, and the conception of causation 
 in which it operates as unscientific. But the conception of causation 
 which aims to be free from this defect collapses for lack of such a 
 criterion. Once more it is seen that there is no place for inference in a 
 world of pure fact. 
 
 This suggests the third difficulty, that proof, as proof, turns out to 
 be merely a relation within given content, as was pointed out above. 
 It is no longer a means of passing from one state of knowledge to another, 
 such a function appearing irrelevant to its character as “given evidence” 
 of a “given proposition,” but it is now a relation within the content of 
 a single state of knowledge. We have already seen how any inference, 
 viewed ex post facto, must be as circular as the syllogism. This circularity 
 centers in the very nature of mere proof, its evidential character lying as 
 a fixed relation within its own content.^ 
 
 How tautologous, for example, are Mill’s proofs of causal laws. He 
 takes his instances, already selected as relevant, orders them under the 
 formulae of the methods, and points to them for the proof. It is there, 
 in them, as they stand. The evidential relation lies within a content 
 assumed to be already completely determined. 
 
 ^ “While the process of thought is still active, the logician .... has nothing to 
 say to it; for his vulturine ‘analysis’ never ventures to attack a living thought. He 
 appears upon the scene when the thinking is defunct and over.‘Logical analy¬ 
 
 sis’ first destroys the real connections between thoughts, and then feigns false ones” 
 (Schiller, Formal Logic, 197-98). 
 
INFERENCE AS PROOF OF PROPOSITIONS 
 
 41 
 
 Proof in this static sense resembles a map by which one may and 
 often does represent the journey one has pursued. In the map, as in 
 the proof, the actual movement or passage of the journey is lost, and 
 that which was discovered in the course of the journey is represented 
 simultaneously with the starting-point. Experience cannot be repro¬ 
 duced in a formula. It can be represented, but the adequacy of such 
 representation is always relative to its purpose. Moreover, in the proof 
 as on the map one may indicate a route that has never been actually 
 followed. One may even amuse one’s self by constructing endless maps 
 and proofs without having journeyed or reasoned at all. And just as 
 the route when projected on the map lets slip the real story of the 
 journey, so proof fails to embody the actual process of inference, the 
 account of which in the end must necessarily be given as a part of some 
 one’s personal biography.^ The results of a scientific discovery are 
 ordinarily our sole concern; having set them out in systematic relations 
 with our other knowledge, we call this their evidence. But if we do 
 undertake to give an account of how the discovery was made, we are 
 drawn back to its setting in the life-story of the scientist. Our systema¬ 
 tized “proof” of the discovery does not reproduce either the problem or 
 the solution as these originally appeared to the discoverer. Inference 
 as discovery will be further considered below. 
 
 It may be worth while to illustrate this point in a different field. 
 Consider the case of a debater, whose mind is definitely made up on the 
 main question, but searching for arguments with which to convince his 
 hearers. This man certainly has a problem before his mind, but it is 
 not the problem expressed by the question in debate. It is the problem 
 of selection of materials that shall stand in certain relations to other 
 materials. This relation may, if you please, be called evidential. It is 
 intended to operate in an inferential process in the hearer’s mind. But 
 to the debater’s own mind"" that so-called evidential relation does not 
 operate inferentially, and the selection of evidential materials is no more 
 inference for him, as regards the question in debate, than would be the 
 selection of rhyming words or the parts of a puzzle picture. Proof, then, 
 may mean two different things. It may mean either the reconstruction 
 of content by which an actual inquiry is met, or a relation within the 
 
 1 In the last analysis is not the historical more fundamental than the scientific 
 point of view? This is perhaps the ultimate issue between instrumentalism and 
 intellectualism. 
 
 2 And to the hearer’s mind when the debate is over and he reviews the argument 
 as a whole. 
 
42 
 
 RELATION OF INFERENCE TO FACT IN MILVS LOGIC 
 
 content as it is looked back on by the inquirer after the reconstruction, 
 or as it is supposed to apply to the problem of some other inquirer. 
 Under the first alternative proof is evidentially operative in an actual 
 inference; under the second alternative it is not. The latter is ‘Tormal ” 
 proof. 
 
 Proof in the latter sense may reach any degree of detachment from 
 actual inference. The content, and the so-called inferential relations 
 within it, may become indefinitely abstract and schematic. Such is the 
 case, for example, in the field of mathematics and of mathematical logic. 
 The inquirer indeed does form a succession of inferences about the con¬ 
 tent as he is building it up, but the relations of ‘implication” exhibited 
 in the content after it is built up are not inferences at all in the same sense; 
 or perhaps in any proper sense.^ 
 
 As the difficulty presents itself in Mill, it seems that the facts must 
 already have been determined before proof appears, and the proof is of 
 the nature of an ex post facto reading of the facts after the real inference 
 has been made. This view of proof leaves untouched the question of 
 how the facts were determined. This question receives partial considera¬ 
 tion under Mill’s treatment of hypothesis, to which we now tiirn. 
 
 ^ In criticizing this sentence of Adamson, “No logical method can be developed 
 save from a most definite conception of the essential nature and modus operandi of 
 thinking,” Professor Marvin says, “To which the reply can in the present day be given: 
 ‘But it has been done, such a symbolic logic actually obtains’” {Journal of Phil., etc., 
 May 8, 1913, p. 275). Even in the present day the claim of symbolic logic to be 
 independent of the nature and method of thinking might be challenged. Its whole 
 scheme of implicational relations is only one stage in the development of an instru¬ 
 ment for the use of actual thinking. 
 
CHAPTER VI 
 INFERENCE AS DISCOVERY 
 
 In his formulation of the inductive methods, Mill set out to describe 
 the process by which general propositions are established. As the induc¬ 
 tive methods are merely means of elimination among alternative propo¬ 
 sitions, the outcome is to reduce inference to disproof. But Mill cannot 
 permanently lay aside as merely psychological the question of how we 
 get the alternative propositions upon which disproof is to operate. He 
 is forced to acknowledge a place in inference for the formation of hypoth¬ 
 eses. Inference as discovery inevitably involves a hypothetical approach 
 to the conclusion; and this may fairly be called Mill’s third conception 
 of inference. In answer to the question where we get the propositions 
 to be established, we must say that they come as hypotheses. 
 
 The official place of hypothesis in Mill’s doctrine of inference is two¬ 
 fold.^ First, it appears as a means of extending the “method of deduc¬ 
 tion,” by substituting a supposed proposition where no suitable one is 
 known, and testing it by the consequences drawn from it.^ Secondly, it 
 appears in several of the processes subsidiary to induction, processes that 
 have to be performed before induction is ready to operate. At neither 
 point is the process of forming hypotheses given a central, constitutive 
 part in inference. Mill can still speak of “perfect induction without any 
 mixture of hypothesis,”^ or of a “theory” that has in it “nothing, strictly 
 speaking, hypothetical.’’^ 
 
 At one point or another of his discussion, indeed. Mill is obliged to 
 acknowledge a hypothetical factor at every step of inference. These 
 admissions appear mainly in the sections that deal with “subsidiary” 
 operations, and the most significant of them appear in passages added, 
 under stress of controversy, in the later editions of his work. They do 
 not lead Mill to reconstruct the non-hypothetical character of his view 
 of inference and its relation to fact. But they do allow a tentative 
 factor through the whole process of inference as it is in operation. Con¬ 
 sider, for example, his treatment of analogy.s Analogy is recognized as 
 hypothetical; but all inference is analogical in so far as it is a taking of 
 
 * On Mill’s doctrine of hypothesis see Dewey, Studies in Logical Theory, 160-68. 
 
 2 350&, c. “ 3606. 
 
 3 3596. ® Book III, chap. xx. 
 
 43 
 
44 RELATION OF INFERENCE TO FACT IN MILL’S LOGIC 
 
 one situation to be like another. Again, abstraction or “ the formation 
 of concepts”^ is but another name of hypothesis. Mill shows with 
 admirable clearness that the mere naming of a fact implies that one has 
 already interpreted it in terms of other experiences, though apparently 
 he had not thought of this when he based inductive proof on ‘‘cases.” 
 “ We cannot describe a fact without implying more than the fact.” “An 
 observation cannot be spoken of in language at all without declaring 
 more than that one observation; without assimilating it to other phe¬ 
 nomena already observed and classified.”^ Mill gives some striking 
 descriptions of the method by which one meets actual problems, as in 
 “the search for a lost object,”^ or the unraveling of “the true history of 
 any occurrence from the involved statements of one or of many wit¬ 
 nesses.” “He extemporizes, from a few of the particulars, a first rude 
 theory of the mode in which the facts took place, and then looks at the 
 other statements one by one, to try whether they can be reconciled with 
 that provisional theory, or what alterations or additions it requires to 
 make it square with them. In this way .... we arrive, by means 
 of hypotheses, at conclusions not hypothetical. ”4 
 
 Of special interest are Mill’s admissions that any experiment—and 
 the inductive methods are all essentially formulae of experimentation— 
 receives its very point from the hypothesis it is to test. “All experi¬ 
 mental inquiry assumes provisionally some theory or hypothesis, which 
 is to be finally held true or not, according as the experiments decide.” 
 
 “The theory itself preceded the proof of its truth.It had to be 
 
 conceived before it could be proved.All the true theories in the 
 
 sciences .... began by being assumed.’’^ “Nearly everything which 
 is now theory was once hypothesis. Even in purely experimental science 
 some inducement is necessary for trying one experiment rather than 
 another.”^ This conflicts with the statement assigning the place of 
 hypotheses in Mill’s original scheme, that they “are invented to enable 
 the Deductive Method to be applied earlier to phenomena ”;7 for it is 
 only on the pronipting of hypotheses that the deductive or any other 
 method is ever applied at all. Without hypothesis Mill’s own methods 
 would be open to the strictures he brings against Bacon’s step-by-step 
 method.^ His attempt to work out a set of inductive methods as self- 
 contained and non-tentative in their “proof” as the syllogism had 
 seemed to be has broken down. 
 
 ^ Book IV, chap. ii. • ^ 45 2a, b. 3 4636. 
 
 4 354. The final words, “conclusions not hypothetical,” could be accepted only 
 subject to the interpretation given below, that they differ only in degree and function. 
 
 s i846-85a. ^ 353C. ’ 35ob. ® 603C-04& and note. 
 
INFERENCE AS DISCOVERY 
 
 45 
 
 But however wide a scope Mill is finally compelled to grant to the 
 operation of hypothesis, he does not adequately appreciate the organic 
 interrelation of hypothesis and fact. Indeed, he allows hypothesis too 
 much scope. “An hypothesis being a mere supposition, there are no 
 other limits to hypothesis than those of the human imagination; we 
 may, if we please, imagine, by way of accounting for an effect, some 
 cause of a kind utterly unknown, and acting according to a law altogether 
 fictitious.”^ This extreme statement, it is true, receives some limita¬ 
 tion later on; but without sufficiently clear recognition of the principle 
 that every genuine hypothesis makes a tentative claim to he fact. Mere play 
 of the imagination may have aesthetic or other values; but only as a 
 hypothesis is a genuine and controlled effort to grasp the facts, and 
 because it is that, does it hold any place as a tentative inference. There 
 is an organic interrelation between hypothesis and fact. In actual 
 inference the two are not held over against each other. Hypotheses are 
 simply our best grasp of the facts—tentative, but genuinely objective— 
 at any one stage of the proceeding. Facts are hypotheses that no longer 
 require a tentative attitude, but have obtained acceptance. Between 
 the two there is progressive interaction, and this interaction constitutes 
 inference. 
 
 Hypotheses, according to Mill, are of two kinds,"* in one of which the 
 existence of the cause is acknowledged but its law of operation is sup¬ 
 posed; in the other the law is acknowledged and the cause supposed. 
 Of this second kind there are again two varieties: the hypothetical cause 
 may be a supposed collocation of agents of a known kind^ or a supposed 
 agent of an unknown kind. The last variety, however, must either be 
 the supposition of a law, in the form of certain properties of the agent, 
 and so fall under the first head, or it collapses into a merely verbal form, 
 without even a tentative claim to be a fact.^ The other two correspond 
 to the two great types of problem, in which either the major or the minor 
 premise is in question; for the difficulty in a situation that needs clear¬ 
 ing up may lie primarily in the need either of constructing or of selecting 
 a concept, the concept standing for the mode of interpretation which we 
 carry over from one experience to another. While Mill’s classification 
 of hypotheses and his selection of examples are somewhat confused in 
 detail, it is at least clear that hypothesis has its place in the clearing up 
 of any problematic situation; it is not limited to the discovery of uni¬ 
 versal propositions only. 
 
 * 349c?. Cf. Hobhouse, Theory of Knowledge, 418. 
 
 2 349c?. ^ 2 > S 9 h -(> 0 ‘ 
 
 4 35 S«-S 9 ^^- Cf. 351^^-53^- 
 
46 RELATION OF INFERENCE TO FACT IN MILL’S LOGIC 
 
 However broad their service in leading over from some facts to others, 
 in their inner character they seem for Mill to be made of a different sort 
 of stuff from the facts which they are intended to explain or by which 
 they are tested. But is not this to misunderstand the character of these 
 facts ? On the one hand, if we had the facts at the start there would be 
 no occasion for any hypothesis and none would arise. It is because in 
 some respect the “facts” have failed us that we are compelled to attempt 
 a reconstruction, and our reconstruction at any stage is the nearest 
 available approach to the facts. If not fact, the hypothesis is the most 
 active possible candidate for that title. When a hypothesis becomes 
 fact, it does not change its inner nature; it simply does its work so well 
 that it satisfies.^ It is idle, therefore, to treat the facts as already given, 
 and the hypothesis as a structure of our own somehow additional to the 
 facts; for we get hold of the facts only as some interpretation of them 
 obtains justification by satisfactorily clearing them up. The interpre¬ 
 tation provides the facts quite as truly as the facts suggest the inter¬ 
 pretation; because the facts for us are always the facts as we interpret 
 them, and our interpretation is always the best statement we can offer, 
 under the circumstances, of what the facts really are. So, on the other 
 hand, the proof of a hypothesis is not the external and accidental bring¬ 
 ing of it to the test of facts which are alien to its nature and which would 
 remain untouched by its own fate. Mill employs no little ingenuity in 
 explaining why a legitimate hypothesis must be capable of proof by 
 “independent evidence.” Is it not simply that if the hypothesis has 
 done its work the facts to which it brings us must be different from the 
 “facts” from which it set out, else these would never have needed the 
 reinterpretation ? 
 
 Mill distinguishes hypotheses which can be proved from those which 
 can only be disproved. “We want to be assured that the law we have 
 hypothetically assumed is a true one; and its leading deductively to true 
 results will afford this assurance, provided the case be such that a false 
 law cannot lead to a true result; provided no law, except the very one 
 which we have assumed, can lead deductively to the same conclusions 
 which that leads to. And this proviso is often realized.”^ But it is 
 realized only inside of a larger system which for the purpose of the 
 inquiry need not be questioned, but is of the same essential character 
 as that of the interpretation within the system. Alternatives which at 
 one stage of an investigation or one period of human knowledge seem to 
 
 ^ Sidgwick, The Process of Argument, 12-21, makes an acute analysis of the relation 
 of fact and inference, and shows it to be one of degree. 
 
 2 35od. 
 
INFERENCE AS DISCOVERY 
 
 47 
 
 be absolute mutual exclusives, cease later to be such through a revision 
 of the system within which they fall. In contrast to the hypothesis that 
 can be definitely proved, Mill considers such hypotheses as that of 
 Descartes’ celestial vortices, which “could not lead to any course of 
 investigation capable of converting it from an hypothesis into a proved 
 
 fact. It might chance to be dwproved.Direct evidence of its 
 
 truth there could not be.”^ In principle, however, there is no difference; 
 all proof is disproof. There is every difference in the degree to which 
 the disproof of one interpretation is felt to be a satisfactory proof of 
 another, but there is never another kind of proof than that. The move¬ 
 ment of inference is positive, not negative; and any bona fide interpre¬ 
 tation of the facts has a right to hold the field till it meets with obstacles 
 and is challenged by a better. 
 
 There seems to be no escape for Mill from an ex post facto view of 
 “facts.” After an inference has done its work we can then look back 
 upon the facts as the inference has now interpreted them, and our 
 rejected or outworn hypotheses we set up in contrast to the facts and 
 call them errors. This ex post facto attitude is the temptation always 
 waiting in the path of the realist, as Mill must be classed in this part of 
 his theory. The very word “fact” plays into his hands.^ For common 
 purposes we very properly confine attention to the results of our infer¬ 
 ences, and forget the process of the inference, as we might kick out a 
 ladder when we have climbed to our goal. We then speak of the facts 
 at an earlier stage as we now at a later stage know them to have been, 
 which is legitimate from the point of view of the later stage, but illegiti¬ 
 mate if we shift the point of view to a consideration of the inference as it 
 took place at the earlier stage. Mill exemplifies this shift when, in con¬ 
 nection with the passage just quoted, he says of the vortices of Descartes, 
 that “ the hypothesis would have been false, though no such direct evi¬ 
 dence of its falsity had been procurable.Here the shift of standpoint 
 is surely obvious. But precisely the same shift, though in more subtle 
 guise, is involved in the assumption that definite, accepted facts are at 
 hand both to prompt and to test hypotheses.^ 
 
 " 3S5<^- 
 
 2 “It is impossible to bring one’s beliefs into harmony with facts, except so far as 
 the facts are known to us” (Joseph, Introduction to Logic, 344). 
 
 3 355 «- 
 
 4 Mill commits the same fallacy when he seriously maintains that if one asserts 
 a general proposition he has thereby actually asserted every case, including those 
 then unknown, that may ever come under it (141a and note). This is to occupy an 
 assumed point of view, not merely of the “innocent bystander,” but of an absolute 
 bystander. 
 
48 RELATION OF INFERENCE TO FACT IN MILL’S LOGIC 
 
 This would seem to afford an explanation also of Mill’s repeated con¬ 
 troversy with Whewell regarding the difference between colligation of 
 facts and induction. It is an ex post facto distinction. An induction, 
 Mill holds, must be either true or false, while a mere colligation need not 
 be either.^ But many an alleged induction has proved later to have 
 been a colligation in Mill’s sense. Mill admits that the colligation is 
 “not the sum of the observations merely” but is “ the sum of the observa¬ 
 tions seen under a new point of view”; but claims that “a real induction 
 is ... . the sum of more than the observations.”^ But this is to forget 
 that a new point of view is more, in the sense of a reconstruction, a better 
 grasp of the facts, and that a real induction adds a general law to the 
 facts only in the sense of knowing them better. The general law has no 
 existence whatever outside the facts, if we remember that the only facts 
 we can speak of are the facts as we take them. But as we look back 
 upon our inferences we distinguish the constructions made from a dis¬ 
 carded “point of view” and those which still hold good; the one we then 
 may very well consider a colligation of the facts under a descriptive con¬ 
 ception, the other we consider the addition of real law. 
 
 It is possible in the same way to explain Mill’s contention regarding 
 the origin of our “conceptions,” which is simply the question of whence 
 our hypotheses arise. Mill, in defense of empiricism, set out boldly with 
 the assertion that “the conceptions .... which we employ for the 
 colligation and methodization of facts, do not develop themselves from 
 within, but are impressed upon the mind from without,” and that they 
 are usually obtained by comparison and abstraction “from the very 
 phenomena which it is their office to colligate.”^ Mill is quite right here 
 as against the extreme a priori position of Whewell. But the concep¬ 
 tion we use is after all a function of our previous experience and of the 
 whole situation, the hypothesis being tentatively taken and the situation 
 gradually cleared up. Mill goes on, in very different terms from those 
 just quoted, to describe the process by which “we advance from a less 
 to a more appropriate general conception, in the progress of our investi¬ 
 gations the process in which the conception is alternately “furnished 
 to the mind” and “furnished by the mind.” “In endeavoring to arrange 
 the facts, at whatever point we begin, we never advance three steps with¬ 
 out forming a general conception, more or less distinct and precise; and 
 . . . . this general conception becomes the clue which we instantly 
 endeavor to trace through the rest of the facts.” “If we are not satis¬ 
 fied with the agreements which we discover among the phenomena” by 
 
 ^218-19. 22id. ^457^. '’463c. 
 
INFERENCE AS DISCOVERY 
 
 49 
 
 using this conception, ‘^we change our path, and look out for other 
 agreements.” ‘‘The different conceptions which the mind successively 
 tries, it either already possessed from its previous experience, or they 
 were supplied to it in the first stage of the corresponding act of compari¬ 
 son.”" This tentative process can be defined only in terms of the 
 mutually constitutive relation of hypotheses and facts. 
 
 Still less can Mill maintain that the contribution between hypothesis 
 and fact is all in one direction when he explains what is meant by “appro¬ 
 priate” conceptions. “The question of appropriateness is relative to 
 the particular object we have in view.”^ “ Some modes of classing things 
 are more valuable than others for human uses, whether of speculation or 
 of practise; and our classifications are not well made, unless the things 
 which they bring together .... agree with each other and differ from 
 other things in the very circumstances which are of primary importance 
 for the purpose .... which we have in view, and which constitutes 
 the problem before us.” Our conceptions must “help us toward what 
 we wish to understand.”^ But, now, what is the criterion of such appro¬ 
 priateness? The answer is suggested by Mill, though he by no means 
 appreciates its full significance. “ That the conception we have obtained 
 is the one we want, can only be known when we have done the work for 
 the sake of which we wanted it; when we completely understand . . . . 
 the phenomena .... with which we concern ourselves.” But “pre¬ 
 mature conceptions we must be continually making up, in our progress 
 to something better. They are an impediment to the progress of knowl¬ 
 edge, only when permanently acquiesced in.”4 The “work to be done” 
 by the hypothesis is to give us the very facts which we can then, if there 
 is occasion so to do, assume to have been present before the hypothesis 
 was suggested. We are entirely justified in reading them back in this 
 way from the later point of view after we have reached it, and so long 
 as we maintain it; but we-could not have held this point of view, or 
 grasped the facts which we thus read back, if the hypothesis had not 
 done its work. 
 
 The cardinal error, then, of Mill’s theory of inference is to read back 
 the results of inference so as to make it seem to operate upon completely 
 determined fact. The facts on which inference is supposed to be based, 
 and to which it is expected to apply, whether these are regarded as the 
 ultimate mental constituents of associational psychology or the things 
 and events of objective naturalism, are by Mill all equally taken to be 
 the sort of definitely determined realities such as they can be known to 
 
 * 4596. " 459^. 3 462e-63a. ^ 463^- 
 
50 
 
 RELATION OF INFERENCE TO FACT IN MILL’S LOGIC 
 
 be only after inference has in any case done its work. Facts when thus 
 determined are found to possess attributes whose universal and system¬ 
 atic character betrays their origin in the reflectively interpretative con¬ 
 structions of experience. In passing from one fact of this kind to another 
 there is no place found for a living act of inference, because the work 
 which inference should do has already been assumed, and inference col¬ 
 lapses into a formal relation within simultaneously presented contents. 
 While a subsidiary place must finally be left open for invention and dis¬ 
 covery, these are not assigned by Mill any constitutive function in the 
 determining of the facts which were given in the first place. Fact and 
 inference therefore fall apart, and in trying to find a vital relation between 
 them Mill oscillates from one to the other. His difficulties could be 
 removed only by recognizing the ultimate unity of inference and fact, 
 and developing their distinction only as an instrumental function in the 
 process of knowledge. 
 
I 
 
 i 
 
 ■ 
 
 'i 
 
 H 
 
 ■( 
 
 i 
 
 \ 
 
 
 A' 
 
 A 
 
 'i 
 
V 
 
 \ 
 
DATE DUE 
 
 
 ^1 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 ' 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 GAYLORD 
 
 
 
 PRINTED IN U.S.A.