. 5
( 10)


which was at p1, has moved to p2,™™ we would interpret the subjective
succession of our perceptions differently. For example, I perceive a
tower at point p1, and a moment later I perceive a (qualitatively) iden-
tical tower at point p2. It is impossible for me to suppose something that
I could think of as the antecedent s of a rule, ˜˜if s, then the tower, which
was at p1, has moved to p2.™™ So in this case, I need to order the temporal
relation of the objects of my perceptions differently. I conclude that two
towers that are qualitatively identical exist simultaneously at two distinct
points in space.
The conclusion of the argument, therefore, is: every objective succes-
sion of states ˜˜presupposes something upon which it follows according to
a rule,™™26 that is to say, it has a cause (ratio fiendi or existendi “ both terms
are appropriate here: the reason or ground is a ground of a state™s
coming to be [ratio fiendi], but it is also the only possible version of the
ratio existendi, or ground of existence). The only existence for which one
can seek a ratio existendi or cause is the existence of a state of a substance
that did not exist before. As for the substance itself, the permanent
substratum of every change of state, there is no sense in seeking a ratio
existendi, a ground of existence.

Ratio existendi, ratio fiendi, ratio essendi
Does all of this suffice to explain why the causal principle stated and
proved in the Second Analogy of Experience should take over the role of
the principle of sufficient reason stated in the New Elucidation, in all its
aspects? So far I have only explained how a descendant of the principle

Admittedly, here Kant seems blithely to move from the epistemic point: ˜˜we presuppose
something, upon which the change of states follows, according to a rule,™™ to the ontological
point ˜˜the change of states presupposes something upon which it follows, according to a rule™™
(see for instance A195/B240:
If, therefore, we experience that something happens, then we always presuppose that
something else precedes it, which it follows according to a rule . . . Therefore I always
make my subjective synthesis (of apprehension) objective with respect to a rule in
accordance with which the appearances in their sequence, i.e. as they occur, are determined
through the preceding state . . .
I say more about this move, and try to explain how Kant thinks he can justify it, in ch. 6,
pp. 168“73.

of succession from the New Elucidation managed to take over the role of
the principle of reason of existence, as well as that of the principle of
reason of coming to be. But what happens to the other aspects of the
principle of sufficient reason? And what happens to the objection I
formulated earlier, which was that in the pre-critical period, Kant
jumped too quickly from distinguishing between reason that and reason
why to asserting that there is always a reason why? Well, this is perhaps
where the most interesting aspect of Kant™s critical position comes to the
fore: Kant™s view now provides a response to that objection that his pre-
critical view could not provide. Kant can now assert that for every
determination of a thing there is an antecedently determining reason
(a reason determining by the antecedent), a reason why, whether this
reason is contained in the essence of a thing (ratio essendi) or in its relation
to other things (ratio fiendi vel existendi). But this is because the ˜˜essence™™
of empirical things, or what Kant now calls their ˜˜nature,™™ consists in the
marks under which they can be recognized as appearances, not in the
properties they might have as things in themselves. This restriction is
what makes it possible to assert the universal validity of the principle of
sufficient reason understood as a principle of antecedently determining
reason. The reason for a thing™s determinations may lie in the (relatively
or absolutely) permanent characteristics by which a thing can be recog-
nized as the kind of thing it is (this argument was made in the First
Analogy of Experience, which I have not examined here).27 Or it may lie
in ˜˜something that precedes any change of state, upon which this change
of state follows, according to a rule™™ (this is the argument of the second
Analogy of Experience, which I just briefly recounted). Finally, perman-
ent as well as changing characteristics are determined in the context of
the universal interaction of all things coexisting in space (this is the
argument of the Third Analogy of Experience, the descendant of the
principle of coexistence from the New Elucidation).28
For the essence itself (what I called the relatively or absolutely perman-
ent marks under which a thing is recognized as the kind of thing it is),

The ordinary objects of our perceptual experience, Descartes™ piece of wax, Kant™s planets
in the Third Analogy, and Kant™s ship in the Second Analogy, are only relatively perman-
ent; matter, characterized by extension, figure, and impenetrability, insofar as we take it to
be the ultimate substrate of all spatiotemporal appearances, is absolutely permanent. The
argument that all changes of state of a thing presuppose something permanent was made
in the First Analogy (see A182/B224“A189/B232). On this point, see KCJ, pp. 325“45, and
ch. 2 in this volume, pp. 53“4.
On this principle, see KCJ, pp. 375“93, and ch. 7 in this volume.

there is no reason. It is just a fact about the relation between our
cognitive capacities and the state of things that we recognize bodies in
general under the marks of extension, figure, and impenetrability. It is a
fact about the present use of our recognitional capacities that we recog-
nize beeswax as the kind of thing that is hard, yellowish, and fragrant
under normal conditions of temperature but becomes soft, sticky,
browner, and so on, when heated up. As for the changes of states, for
which the Second Analogy provides a principle of sufficient reason, no
ultimate determining reason, or ground, can be found. For any event,
the search for ˜˜something that precedes, upon which it follows, accord-
ing to a rule,™™ can go on indefinitely. So, Kant™s critical proof of the
principle of sufficient reason is also a severe restriction of its scope and
force. Nevertheless, because he has thus proved a principle of sufficient
reason that is understood as a principle of antecedently determining
reason, reinterpreted in the terms of his critical philosophy and itself
having its ground or reason in the unity of self-consciousness, Kant can
affirm, in the preface to the Critique of Pure Reason and then again in the
introduction to the Transcendental Dialectic, that it is an unavoidable
destiny of reason (this time as a faculty, Vernunft) always to look for a
further reason, or ground (Grund) of the objective determinations of
things, while at the same time it can never claim to have found the
ultimate ground.
Finally, it is clear that we must now distinguish between the principle
of reason of propositions and the principle of reason of things and their
determinations. It is a logical principle that every proposition (assertoric
judgment) must have a reason, without which it would, at best, remain a
merely problematic judgment whose negation could equally be
admitted as problematic (possible). This principle, as Kant points out
in the introduction to the Logic collated by his student Jasche, can be
specified in two ways: an assertoric proposition must (1) have reasons or
grounds (Grunde haben) and (2) not have false consequences (nicht falsche
Folgen haben).29 In the first requisite, we may recognize the mere form of
the modus ponens proper to the antecedently determining reason from
the pre-critical New Elucidation, while in the second, we see that of the
modus tollens proper to the consequently determining reason. But
neither of these two versions of the logical principle of sufficient reason
gives us any access to the reason, or ground, of the determinations of

Cf. Logic, Einl. vii, AAix, p. 51.

things. That there has to be a reason or ground for the determination
of things was proven not from a logical principle of reason for the
truth of propositions but from an elucidation of the conditions under
which we can apprehend a temporal order among the objects of our
This restriction of the principle of reason of things and their determi-
nations to a principle of the determination of an objective temporal
order, and the foundation of reasons, in the plural (whether empirical
or logical), in one transcendental reason or ground, ˜˜transcendental
unity of self-consciousness,™™ allow Kant to present an unprecedented
solution to the problem of the relation between the principle of sufficient
reason and human freedom.

The principle of reason and human freedom: the ground beyond grounds
(the reason beyond reasons)
In 1755 Kant insisted against Crusius that admitting the universal valid-
ity of the principle of sufficient (or determining) reason was compatible
with affirming that human beings are free. For, he said, although it is
true that everything that happens “ and therefore also every human
action “ has an antecedently determining reason, in cases where this
reason (ground) is not external (as in mechanical causality), but internal
(as in divine action, and in those human actions where ˜˜the motives of
understanding applied to the will provoke actions™™),31 the action,
although certain, is not necessitated. But in the Remark on the
Analytic of the Critique of Practical Reason, Kant categorically rejects this
kind of solution. Describing an action as free because its ground is not
external but internal, he now maintains, amounts to attributing to
human beings the ˜˜freedom of a turnspit,™™ which has in itself the source
of its movement, its position, and internal structure at each moment
determining its position at the following moment. The truth is,
Kant now says, that in such a situation each change of state, far from
originating from itself a new series of states, is strictly determined by the

In his dismissive response to Eberhard in 1790, Kant noted that Eberhard entertained a
confusion when he formulated the principle of reason as: ˜˜Everything has its sufficient
reason.™™ ˜˜Everything,™™ Kant remarks, can mean ˜˜every proposition™™ or ˜˜every thing.™™ In
the former case the principle is logical; in the latter it is transcendental (see On a Discovery,
AAviii, pp. 193“4). The confusion he denounces was his own in 1755 “ even if, as we have
seen, he was careful to distinguish reason of truths and reason of existences.
New Elucidation, AAi, p. 401.

change that precedes it.32 In the same way, whatever their mode
of determination (whether according to the rules of skill, the advice of
prudence, or the imperatives of morality), human actions, insofar as
they are events in time, are strictly determined by the events that pre-
cede them in time. The principle of reason, proven in the Second
Analogy, applies to them as it applies to every event. But the distinction
between things as they appear to the senses (phenomena) and things
accessible to the pure intellect (noumena), as well as the discovery of
the equivalence between freely determined action and action deter-
mined under the representation of the moral law, allow Kant at the
same time to adopt a position that is in certain respects very close to
the position of Crusius, which he criticized in the New Elucidation: it
is also true to say that at each instant there is no other antecedently
determining reason of action than the will itself, acting under the repre-
sentation of the moral law “ whether or not the agent makes this law
the supreme principle of the discrimination and ordering of his or
her maxims. The temporal determination of the action is no more
than the expansion over time of a non-temporal relation of the agent
to the moral law for which, at every instant, s/he can and should be held
Significantly, it is again in the vocabulary of 1755 that Kant defines the
relation between the moral law and freedom: freedom is the ratio essendi
of the moral law, and the moral law is the ratio cognoscendi of freedom.34
But this vocabulary really indicates that we have now arrived at the limit
of antecedently determining reasons. For human freedom, there is no
other reason than a ratio cognoscendi, moral law as a Faktum of reason
(Vernunft) (not a given of reason, but rather a production of reason).35 In
the New Elucidation, for the existence of God one could state only a
ground of knowing, and not a ground of being or existing (God, Kant
strikingly stated, was the only being for which existence precedes possi-
bility). With the critical system, for freedom as a property of human
beings we must affirm that we have a ground of knowing but not that we
have a ground of being or existing. Of course, according to Kant the

Critique of Practical Reason, AAv, p. 97, trans. Mary Gregor (Cambridge: Cambridge
University Press, 1997).
On freedom and the moral law, see ch. 9 in this volume.
Critique of Practical Reason, AAv, p. 4n.
Ibid., pp. 31“2.

same ground of knowing “ the moral law “ that leads us to affirm the
existence of human freedom leads us also to postulate the existence of
God as a ground for the synthetic connection between virtue and happi-
ness. But this only serves to widen the gap between this and Leibniz™s
principle of sufficient reason. The existence of God is not affirmed by an
ontological, cosmological, or physico-theological proof (God does not
have in himself his ground of being or existing, nor does the affirmation
of his existence result from the ultimate application to finite things of the
principle of antecedently determining reason). The existence of God is
postulated by virtue of a ratio that is not even a ratio cognoscendi, but
rather a ratio credendi, one which human reason generates from its own
resources as the only possible response to its inescapable demand for the
Highest Good.36
In brief: the thinned-out version of the principle of sufficient reason
defended by Kant in his critical philosophy depends on the unity of self-
consciousness that, he maintains, on the one hand conditions all know-
ledge of objects, and on the other hand conditions the ordered unity of
the maxims of action under the legislation of the moral law. The desti-
nies of the two notions “ unity of self-consciousness, principle of suffi-
cient reason “ are from now on linked, for better or for worse: to debunk
the one is also to debunk the other.37
But there is another way of challenging Kant™s principle of sufficient
reason: in Kant™s argument, as we have seen, the principle in all its
aspects is dependent on an Aristotelian predicative logic (the Wolffian
version of that logic) which provides discursive thought with its forms
and toward which temporal syntheses are guided. To put in question the
relevance of this predicative logic and its role in constituting the struc-
ture of our perceptual world is undermining the principle of sufficient
reason in both of the senses the critical Kant gives it (the logical principle
of reason of propositions, the transcendental principle of reason of the
temporal order of appearances). Of this principle, there seems then only
to remain, at best a modest heuristic principle “ for every thing and every
event, one must seek an explanation,38 for every action one must seek a
reason. And a practical imperative of autonomy: for one™s own actions,

Cf. ibid., pp. 124“32. See also Critique of the Power of Judgment, x84, AAv, pp. 434“6, trans.
Paul Guyer and Eric Matthews (Cambridge: Cambridge University Press, 2000).
As we can see in Schopenhauer: see On the Fourfold Root of the Principle of Sufficient Reason.
David Wiggins ends up with this modest version of the principle of sufficient reason in his
article, ˜˜Sufficient reason: a principle in diverse guises, both ancient and modern,™™ in Acta
Philosophica Fennica, vol. 61 (1996), pp. 117“32.

one should, as much as can be done, be in a position to hold oneself
It is therefore tempting to disconnect Kant™s argument in the Second
Analogy of Experience from Kant™s defense of the old principle of
sufficient reason, namely from any aspect of the principle inherited
from the German rationalists Kant discussed in his pre-critical period.
One may then take the Second Analogy to be part of Kant™s exposition of
the epistemological presuppositions of Newtonian natural science (the
option of neo-Kantianism, taken up today by Michael Friedman).39 Or
one may take Kant™s argument to be an explanation of the necessary
conditions of our ordinary perceptual experience, an argument that can
be reconstructed without any reference at all to Kant™s dubious scholastic
heritage (the option of Strawson and his followers).40 In this chapter,
I have tried to offer a third option. I have tried to show that taking Kant™s
scholastic heritage seriously does not mean reducing his view to this
heritage, but on the contrary enables us to measure the full extent of the
reversal he imposed upon it. Following up and reconstructing Kant™s
argument all the way to its origin in the principle of sufficient reason and
the reversal of its proof, then, echoes more familiar themes in today™s
philosophical concerns: the relation between reasons and causes,
and the determination of reasons from the point of view of a self-
consciousness that has the capacity to generate from itself the norms
of its theoretical and practical activity.41 How and why the modern devel-
opments of these themes differ from Kant™s, and what they nevertheless
owe to him “ it will take more work to try further to clarify these questions.

Readers of KCJ may notice that in my discussion of Kant™s New
Elucidation, I discussed only three versions of Kant™s notion of sufficient
reason (ratio essendi, ratio fiendi, ratio cognoscendi) and ended up taking
Kant™s concept of cause in the Second Analogy, in the critical period, to
be a critical version of the pre-critical ratio fiendi. In this chapter, I take
the Second Analogy to have for its ancestor the Principle of Succession

Cf. Cohen, Kants Theorie der Erfahrung; Michael Friedman, Kant and the Exact Sciences
(Cambridge, Mass.: Harvard University Press, 1992).
Strawson, Bounds of Sense. There are of course other options available beyond these most
influential ones. I say more about contemporary readings of Kant™s Second Analogy below,
ch. 6, esp. pp. 170“2.
McDowell, Mind and World, esp. pp. 114“17.

expounded in Nova Dilucidatio. This principle is itself presented by Kant
as a consequence of the principle of sufficient reason understood as a
principium rationis fiendi. I think this new presentation of the issue is a
more precise way of understanding the relation between Kant™s pre-
critical and critical views. It also has the advantage of making more
perspicuous the striking reversal in Kant™s method of proof.


Incredible as it may seem, scholars continue to disagree about what
exactly Kant was trying to prove in his Second Analogy of Experience “
in that section of the Critique of Pure Reason in which he is supposed to
provide his response to Hume™s skeptical doubt concerning the concept
of cause. Since Kant describes this response as the groundbreaking
initial step into his critical system, disagreement about its interpretation
is not a situation we can easily be satisfied with.
In recent years, a number of new studies have brought valuable insight
into the complexities of Kant™s argument, as well as into the roots of the
persisting disagreements about it. All agree on what constitutes the core of
Kant™s response to Hume: Kant maintains that some representation of
causal relation, rather than resulting “ as Hume claimed “ from the
repeated perception of generically identical successions of events, is pre-
supposed in the very representation of any particular objective succession
of states of a thing.1

Admittedly, succession of events and succession of states of a thing are not the same. Hume™s
standard case of succession, in his explanation of our idea of necessary connection, is that of
a succession of events: the motion of one billiard ball (event A) followed by the motion of
another (event B) (see Enquiry, section 4, part i, p. 29). His question is: how do we acquire
the idea that there is a necessary connection between A and B? Kant™s argument in the
Second Analogy focuses on changes of states of a thing: a ship™s change of position, a
cushion™s change of shape (see A192/B237, A203/B49). He argues, as I shall show in the


Disagreements, however, have recently focused on two main issues:
(1) what is meant by the ˜˜objective succession™™ whose representation,
according to Kant, presupposes some representation of causal relation?
Is it (a) the succession of events or states of affairs as we perceive them in
the objects of our ordinary experience “ the freezing of water, the
moving of a ship, the warming up of a stone? Or is it, rather, (b) the
succession of states of affairs as determined in the context of a scientific
image of the world “ for instance the objective, as opposed to the merely
apparent, succession of positions of heavenly bodies? (2) Just what is
involved in the concept of cause which, according to the Second
Analogy, is presupposed in our representation of objective succession?
Is Kant only asserting (a) that in order to think any particular sequence
of events or states of affairs as an objective sequence, we have to think its
temporal order as in some way constrained, and thus in a loose sense,
causally determined “ without further asserting that this constraint or
˜˜binding down™™ of the temporal order involves any notion of strictly
universal and necessary causal laws? Or does Kant argue, in the Second
Analogy, (b) that every event falls under universal and strictly necessary
causal laws?
Gerd Buchdahl and Henry Allison have argued in favor of answers
(a) to questions (1) and (2) (call this ˜˜the Buchdahl/Allison interpretation™™).
Michael Friedman has argued in favor of answers (b) to both questions
(call this ˜˜the Friedman interpretation™™). The alternative between these
two options has more or less dominated recent discussions of the Second
Analogy.2 It seems, then, that interpreting the objective succession at
stake in the Second Analogy as the temporal order of ordinary objects of
everyday experience commits one to the view that the causal principle
second section, that such changes of states are perceived only under the presupposition that
they are connected to other changes of states according to universal causal laws. A change of
states is of course itself an event, so that in the end Kant™s argument does also concern law-
governed successions of events. But the perception of succession on which the argument is
built is not primarily the perception of a succession of events (this will be clear when we
consider the argument in the second section). In reading the argument as focusing on
successions of states rather than successions of events, I agree with Allison: see Henry
Allison, ˜˜Causality and causal laws in Kant: a critique of Michael Friedman,™™ in P. Parrini
(ed.), Kant and Contemporary Epistemology (Dordrecht: Kluwer, 1994), p. 300. Cf. also Allison,
Kant™s Transcendental Idealism, p. 248; Strawson, The Bounds of Sense, p. 134n; and
Longuenesse, KCJ, pp. 371“2.
Cf. Michael Friedman: ˜˜Kant and the twentieth century,™™ in Parrini, Kant and Contemporary
Epistemology, pp. 27“46, and Allison, ˜˜Causality and causal laws in Kant.™™ Cf. also Friedman,
˜˜Causal laws and the foundations of natural science,™™ in Guyer, The Cambridge Companion to
Kant, pp. 161“99; Gerd Buchdahl, Metaphysics and the Philosophy of Science (Oxford:
Blackwell, 1969), pp. 641“5.

argued for in the Second Analogy is a relatively weak one, involving no
notion of universal causal laws. On the other hand, opting for a notion
of objective temporal order understood as an order determined under
mathematical laws (for which the paradigm would be, for Kant,
Newtonian universal gravitation) commits one to the view that the causal
principle argued for in the Second Analogy is a strong causal principle,
asserting not only that for every event there is a cause, but that this cause is
determined under universal and strictly necessary causal laws.
The position I shall defend in this chapter breaks the terms of the
alternative I have just outlined. For I shall defend answer (a) to question
(1), and answer (b) to question (2). I shall maintain that the objective
succession Kant is concerned with in the Second Analogy of Experience
is the succession of states in the objects of our ordinary perceptual
experience; but I shall also maintain that according to Kant, we can
perceive such objective changes only under the presupposition that
they fall under strictly universal causal laws. I shall, in fact, attribute to
Kant not just this epistemological point, but the more radical ontological
(transcendental) point that in the world of appearances, all changes of
states do fall under strictly universal causal laws.
I am not alone in attributing such an argument to Kant. Strawson has
defended the view that the argument of all three Analogies is concerned
with the ordinary objects of our perceptual experience; and he has also
defended the view that Kant intends the Second Analogy as a proof that
all changes of states in such objects are causally necessitated, namely
determined under strictly universal causal laws. He has famously
endorsed Lovejoy™s charge, however, that Kant™s argument for this
strong conclusion is a ˜˜non-sequitur of numbing grossness.™™ And he
has argued that rather than Kant™s own flawed argument, one could
extract from the Second Analogy a valid argument for a weaker conclu-
sion: we can relate our perceptions to the objects they are the percep-
tions of, only if the changes of states in these objects fall under a unified
pattern of reasonably coherent and stable rules.3 Although I agree that
Kant™s argument in its stronger version does raise problems, I shall
argue that it is definitely not the gross non-sequitur Lovejoy and
Strawson read into the Second Analogy.

Strawson, Bounds of Sense, p. 144. For Strawson™s denunciation of Kant™s ˜˜non-sequitur,™™
see ibid., pp. 137“8; cf. Arthur Lovejoy, ˜˜On Kant™s reply to Hume,™™ Archiv fu Geschichte der
Philosophie, vol. 18c (1906), pp. 380“407; repr. in Moltke S. Gram (ed.), Kant: Disputed
Questions (Chicago: Chicago University Press, 1967), pp. 284“308.

Paul Guyer too offers an interpretation of the Second Analogy along a
pattern similar to the one I defend: he maintains (a) that the Second
Analogy is about ordinary objects of our everyday experience, and
(b) that in the Second Analogy, Kant argues for some version of universal
causal laws. However, Guyer has his own way of weakening Kant™s claim:
he maintains that Kant is concerned only with the conditions for confirm-
ing our beliefs about objective successions. Guyer™s Kant argues that we
can confirm our belief that an objective change has occurred or is occur-
ring only if we can ascertain that this change falls under known causal
laws.4 I maintain that Kant wants to make the stronger claim that we
perceive any objective change at all only under the presupposition that
this change occurs according to universal causal laws.
It is not just with respect to the conclusions I think Kant wants to reach,
that I differ from recent commentators of the Second Analogy. Equally
importantly, I differ from them in the method I adopt. My method
consists in taking Kant at his word when he claims that we can understand
the meaning and role of the categories “ the fundamental concepts
necessarily presupposed, according to Kant, in any representation of
objects “ if we understand their relation to logical forms of judgment.
The logical form corresponding, in Kant™s ˜˜table of logical forms of judg-
ment,™™ to the category of causality, is the form of hypothetical judgment.
I claim that the best way to understand Kant™s argument about causality is
to follow the guideline provided by this form as conceived by Kant, and its
use in empirical knowledge. I follow this guideline in three main steps “
thus the three sections of the chapter.
In the first section of the chapter, I consider Kant™s formulation of the
problem of causality. I argue that Kant™s questioning of the causal
principle and his analysis of the concept of cause are best approached
in the light of his conception of logic, and more particularly in the light of
his conception of hypothetical judgments and hypothetical syllogisms.
In the second part of the chapter, I consider Kant™s proof of the causal
principle in the Second Analogy of Experience. All students of Kant know
that this proof is concerned with the conditions of our perception of
objective succession. But this aspect of Kant™s argument is all too fre-
quently detached from the claimed relation between the causal category
and the logical form of hypothetical judgment. In contrast, I shall argue
that this relation provides an indispensable foundation for understanding

Cf. Paul Guyer, Kant and the Claims of Knowledge (Cambridge: Cambridge University Press,
1987), p. 252.

Kant™s argument on the conditions of time perception. But showing this
will also reveal a fundamental difficulty. The argument Kant provides
does not seem to support the strong causal principle he claims to prove.
I suggest that this apparent discrepancy between Kant™s claim and his
actual argument in the specific context of the Second Analogy is a primary
reason for the persisting disagreements about the meaning of the Second
In the third section of the chapter, I argue that in fact Kant does
provide an answer to the difficulty I raised. This answer, however, relies
not only on the discursive model of thought laid out in the first and
second sections of the chapter, but also on Kant™s conception of space
and time as forms of intuition, as it emerges from the Transcendental
Aesthetic and the Transcendental Deduction of the Categories. Since
Kant™s views on space and time are generally considered to be the most
problematic aspect of the first Critique, it is no great surprise if the
argument of the Analogy seems to come upon its major difficulty at
this point. Clarifying the nature of the difficulty is perhaps the most we
can hope to do. It would be no small feat, however, if this also helped us
resolve some of our disagreements about the nature and import of
Kant˜s proof.

Kant™s problem about causality
Let us start with ˜˜Hume™s problem.™™ Hume distinguished two main
aspects in the problem raised by the concept of cause. The first concerns
the causal principle itself: what is the source, and what is the justification,
of our belief that every event or state of affairs must have a cause? The
second concerns our representation of particular causal connections:
what are the source and justification of our belief, in any particular
case, that one event or state of affairs is the cause of another?5
In the Enquiry, Hume focuses mainly on the second question.6 In the
Treatise of Human Nature, he argues that in answering the second question

Cf. David Hume, A Treatise of Human Nature, ed. L. A. Selby-Bigge and P. H. Nidditch
(Oxford: Clarendon Press, 1993), bk i, part iii, sect. 2, p. 78:
˜˜First, for what reason we pronounce it necessary, that every thing whose existence has a
beginning, should also have a cause? Second, why we conclude that such particular causes
must necessarily have such particular effects; and what is the nature of that particular
inference we draw from the one to the other, and of the belief we repose in it?™™
Hume, Enquiry Concerning Human Understanding, in Enquiries Concerning Human
Understanding and Concerning the Principles of Morals, ed. L. A. Selby-Bigge, 3rd edn, with
text revised and notes by P. H. Nidditch (Oxford: Clarendon Press, 1992), sections 4“7.

one also answers the first. Very briefly, his argument is as follows. No
particular perception of an event or state of affairs provides us with any
idea of its power to produce another event or state of affairs, or with any
idea of a necessary connection between two events or states of affairs. Only
the repetition of similar pairs of events or states of affairs following upon
one another generates in us a customary association of one with the other,
and thus a subjective expectation of perceiving the second upon perceiving
the first. Our idea of a necessary connection between two events or states of
affairs, then, reflects nothing but our own subjective propensity to expect
the second upon perceiving the first, and to form the vivid idea of the first
(which amounts to believing that it exists) upon perceiving the second. But
because of the natural tendency of our mind to ˜˜spread itself upon external
objects,™™7 we tend to attribute to the objects themselves a connection whose
idea really reflects only an expectation in us. This, then, is how we form the
idea of particular cases of causal connections. Because an event has always
been followed by another, we come to believe that every event similar to the
first will always be followed by an event similar to the second. But in truth,
no amount of evidence provided by our memory and senses is sufficient to
justify such a belief.8
According to Hume, our belief in the universal causal principle is just a
generalization of our particular causal beliefs. Associating to every per-
ceived event or state of affairs the vivified idea (and thus the belief in the
existence) of a preceding or succeeding event similar to those that have
always preceded or succeeded it, just is entertaining the general belief
that ˜˜everything that comes into existence must have a cause.™™ Thus
Hume derives our representation of causal connections from the
repeated succession of similar events, and our belief in the universal
causal principle from the generalization of our belief in particular causal

Hume, Treatise, p. 167.
Treatise, bk I, part iii, 2“14; Enquiry, sections 4“7.
See Treatise, bk I, part iii, section 8, pp. 104“5; section 14, p. 172. One may wonder whether
Hume actually accounts for the universal principle as he has first introduced it (i.e. the
principle admitted both in metaphysics and by common understanding: ˜˜every beginning
of existence must have a cause™™), or rather gives an explanation of some broader principle
such as: ˜˜every beginning of existence must have a cause and an effect.™™ Many of Hume™s
formulations, throughout the Enquiry and the Treatise, do not give precedence to cause over
effect in the ideas imagination naturally associates with any given event or state of affairs.
The privilege given to the principle as stated can probably be explained more thoroughly
by looking into Hume™s explanation of what he calls ˜˜the world of judgment™™ (Treatise, p. 74,

In the Introduction to the Critique of Pure Reason, Kant states that
accepting Hume™s account would amount to giving up the very content
of the concept of cause:
[In the proposition: ˜˜every alteration must have a cause™™], the very
concept of a cause so obviously contains the concept of a necessity of
connection with an effect and a strict universality of the rule that it would
be entirely lost if one sought, as Hume did, to derive it from a frequent
association of that which happens with that which precedes and a habit
(thus a merely subjective necessity) of connecting representations arising
from that association. (B5)

The charge may strike us as bizarre: after all, as I just explained, the
whole point of Hume™s psychological derivation of the concept of cause is
to account for the idea of necessary connection ˜˜contained in the concept
of cause,™™ and to explain how we tend to inflate mere observed regular-
ities into ˜˜strictly universal rules.™™ Kant is of course aware of this. What
we must take him to mean, then, is that accepting Hume™s account of
precisely these features would be giving up the concept of cause alto-
gether, because it would mean that as far as objects are concerned, our
idea of causal relation can be reduced to the idea of a non-causal relation:
repeated succession of similar events or states of affairs. The ideas of
˜˜necessity of connection™™ and ˜˜universality of the rule™™ would remain
grounded only in the subjective propensities of our mind. Now, one way
to reject such a reduction is to show that Hume in fact does not give an
accurate account of what we actually think when we think ˜˜the necessity
of connection with an effect™™ and the ˜˜strict universality of the rule™™
contained in the concept of cause. This is indeed what Kant will set out
to show. But what does he himself mean by the ˜˜strict universality of the
rule™™ contained in the concept of cause?
I suggest that we find the beginning of an answer to this question in
the preface to the Prolegomena to Any Future Metaphysics. There Kant
credits Hume with having challenged our reasoning capacity to explain
˜˜by what right she thinks anything could be so constituted that if that
something be posited, something else also must necessarily be posited;
for this is the meaning of the concept of cause.™™10 This question concerns
the second aspect of the problem of causality as defined earlier, namely,
cf. Enquiry, p. 26) namely our belief in the existence of independently existing objects. My
purpose here is not to submit Hume™s account to critical scrutiny, but only to lay out its
overall structure insofar as it should help clarify Kant™s own formulation of ˜˜Hume™s
Prolegomena, AAiv, p. 257.

what is the justification of any particular statement of causal connec-
tion?11 However, the terms in which Kant formulates this question are
quite bizarre, and certainly not Humean: to say that something is the
cause of something else is to say that ˜˜if this something is posited, then
something else must also necessarily be posited.™™ It is the word ˜˜posited™™
that intrigues me here. What Kant credits Hume with, is perhaps
Hume™s problem. But this problem is not stated in Hume™s language.
The language is actually that of the hypothetical syllogism in modus
ponens as defined in the logic textbooks of the time. Indeed, Kant™s
phrasing (˜˜if something is posited, something else also must be posited™™)
reproduces, almost word for word, Christian Wolff™s description of the
inference in modus ponens in a hypothetical syllogism:
x407: If, in a hypothetical syllogism, the antecedent is posited, the con-
sequent must also be posited [si in syllogismo hypothetico antecedens ponitur,
ponendum quoque est consequens].
x408: The antecedent being posited in the minor, the consequent should
also be posited [posito antecedente in minore, ponendum quoque est

In a hypothetical judgment (˜˜If A is B, then C is D™™), the ˜˜if™™ clause is
called the antecedent, the ˜˜then™™ clause is called the consequent. In a
hypothetical syllogism whose major premise is ˜˜If A is B, then C is D,™™
the antecedent of the hypothetical judgment being posited, i.e. asserted,
in the minor premise (˜˜A is B™™), then the consequent should also be
posited, i.e. asserted, in the conclusion (˜˜so, C is D™™).
By presenting the problem of causality in these terms, Kant brings
attention to the fact that the problem of how we can think a particular
causal connection turns out to be the following: how can the relation
between two empirical states of affairs be such that the first can be
thought under the antecedent, the second under the consequent of a
hypothetical judgment that functions as the major premise in a syllogism
in modus ponens, such that ˜˜the antecedent being posited (as the minor
premise), the consequent must be posited (as the conclusion)™™? If this is
correct, the ˜˜strict universality of a rule™™ thought in the concept of cause is
the strict universality of the hypothetical judgment (˜˜If A is B, then C is D™™)

See above, pp. 147“8 and n. 5.
Christian Wolff, Philosophia rationalis sive Logica. I do not mean that Kant™s ˜˜universality of
the rule™™ is the universality of the rule of modus ponens itself. The ˜˜universality of the rule™™ is
the universality of the hypothetical judgment, which is the major premise of the hypothe-
tical syllogism in modus ponens. I shall say more on this in a moment.

that we implicitly presuppose as a premise whenever we represent two
particular states of affairs ˜˜A is B™™ and ˜˜C is D™™ in such a way that ˜˜A is B™™
being posited, ˜˜C is D™™ should also be posited.13
One might wonder what the question thus reformulated still has in
common with ˜˜Hume™s problem.™™ But Hume too moved from the ques-
tion of how we think the necessary connection between two events to the
question of how we make the representation of a mere repetition of
similar sequences of events into the representation of a strictly universal
rule or law, in such a way that all future events similar to the one identified
as a cause should be followed by events similar to the one identified as the
effect. What is interesting about Kant™s formulation is that from the outset
it collapses together the two steps in Hume™s analysis of the problem of
particular causal connection. ˜˜If something is posited, something else
should be posited™™ can mean both that the second ˜˜something™™ necessarily
comes to existence if the first does, and that they are, in effect, respectively
thought under the antecedent and under the consequent of a strictly
universal rule. Hume argued, of course, that not reason, but imagination
(the natural propensity of the mind to form the enlivened idea of the
second upon perceiving the first) is the author of the ˜˜strict universality™™ of
the rule. Kant wants to argue that understanding and reason are at work
in universalizing the connection between what precedes and what follows.
Presenting the problem in the terms borrowed from Wolff™s hypothetical
syllogism helps to bring this out.
But what might Kant mean by the ˜˜strict universality™™ of a hypothetical
judgment? When Kant explains the quantity of judgments, he always gives
examples of categorical judgments “ all, some, one A are/is B. What could
be the universality of a hypothetical judgment “ if A is B, then C is D?
Before considering this problem, we need to say more about the hypothet-
ical form itself. Kant™s hypothetical judgment is quite different from our

In addition to the striking similarity between Wolff™s formulation of the rule of modus
ponens and Kant™s presentation of ˜˜Hume™s problem™™ in the preface to the Prolegomena, we
have other reasons to suppose that Kant had the inference in modus ponens in mind.
Already in his pre-critical Reflections on Metaphysics he characterized the problem of
the causal connection in terms of what he called a synthetic respectus rationis ponentis (see
Reflexion 3753, AAxxvii, p. 283 ). And of course in the first Critique and in the Prolegomena
he relates the category of cause to the form of hypothetical judgment. In the Lectures on
Metaphysics contemporary with the Critique, he gives a more detailed exposition of the
relation between the cause and the antecedent, the effect and the consequent of a
hypothetical judgment (see Metaphysik Volkmann, AAxxviii“1, p. 397). For more on this
point see above, ch. 5, pp. 129“31.

material conditional, in two respects: the nature of the connective, the
nature of the propositions connected. I shall consider each in turn.
First, the connective. In our material conditional, the meaning of the
connective is given by its truth table: the conditional (˜˜if p, then q™™) is false
just in case its antecedent is true and its consequent false; it is true in all
other cases. Not so for the connective ˜˜if . . . then™™ of the hypothetical
judgment, which Kant calls Konsequenz (not to be confused with the
˜˜then . . . ™™ clause, called the consequent in English, die Folge in
German). The truth value of the hypothetical judgment does not
depend on the truth value of its components, but on the truth of the
Konsequenz itself: the hypothetical judgment is true just in case there is
between antecedent and consequent a relation of Konsequenz, i.e. a rela-
tion of ground to consequence. Here is what Kant writes in the Jasche ¨
The matter of hypothetical judgments consists of two judgments that are
connected with one another as ground and consequence. One of these
judgments, which contains the ground, is the antecedent (antecedens, prius),
the other, which is related to it as consequence, is the consequent (conse-
quens, posterius), and the representation of this kind of connection of two judgments
to one another for the unity of consciousness is called the consequentia [my
emphasis] which constitutes the form of hypothetical judgments [ . . . ]
In [a hypothetical judgment] I can . . . connect two false judgments with
one another, for there what matters is only the correctness of the connection “ the
form of the consequentia, on which the logical truth of these judgments rests.14

Note Kant says that antecedent and consequent can both be false: namely,
in asserting the hypothetical, we assert neither the antecedent nor the
consequent. We only assert that there is a relation of consequence
between them. So, they can both be false, or they can both be true. Or
perhaps even (as in our material conditional) the antecedent can be false
and the consequent true, without this putting into question the truth of
the Konsequenz. But the important difference between Kant™s hypothetical
judgment and our material conditional is that in the former, the meaning
of the connective is not fixed by its truth conditions, but on the contrary
the truth conditions are fixed by the meaning of the connective: because
the meaning of ˜˜if . . . then™™ in a hypothetical judgment is that there is a
relation of Konsequenz between antecedent and consequent, the hypothet-
ical judgment is true only if its antecedent is false, or if its antecedent and

Immanuel Kant, Jasche Logic, x25, AAix, pp. 105“6.

consequent are both true. And of course these are necessary, but not
sufficient conditions. For these conditions could be satisfied and there
still be no relation of Konsequenz at all between antecedent and conse-
quent. For instance: ˜˜If my mother is French, New York is in America™™; ˜˜If
the moon is square, I can fly™™; ˜˜If the moon is square, New York is in
America.™™ All three hypothetical judgments are false, because there is no
Konsequenz between antecedent and consequent.
Now, one may wonder, then, what Kant™s form of hypothetical judgment
has at all in common with what we would call a logical form. Answer: what
makes the Konsequenz a logical form is that it grounds the two forms of
inference: modus ponens, modus tollens. For because of the meaning of the
Konsequenz, whoever asserts the antecedent is thereby committed to assert-
ing the consequent (modus ponens); and whoever denies the consequent is
thereby committed to denying the antecedent (modus tollens). Note that
these two forms of inference are just those that the meaning of the material
conditional allows. But we see here the same asymmetry as in the determin-
ation of the truth of the propositions themselves: just as the truth of the
material conditional depends on the truth of antecedent and consequent,
similarly the modus tollens and modus ponens are just rules of separation
stemming from the truth conditions of the conditional: if the conditional
is true and its antecedent is true, then the consequent is true; if the condi-
tional is true and the consequent is false, then the antecedent is false. For
the hypothetical on the other hand, the form of inference is grounded not
in the truth conditions, but in the meaning of the Konsequenz. Nevertheless,
the forms of inference allowed are the same in both cases.
Consider now the propositions connected by the Konsequenz in a
hypothetical judgment. For Kant, the primary model for any judgment is
predication: A is B (subject“copula“predicate). Hypothetical judgments
themselves are to be understood as asserting a relation between predica-
tions. More specifically, according to Kant (who here too takes after Wolff )
what a hypothetical judgment asserts is that the predication expressed by
the consequent can be asserted only under the condition that the predica-
tion expressed in the antecedent be asserted. In the Jasche Logic, Kant writes:
There is an essential difference between the two propositions: all bodies
are divisible, and: if all bodies are composite, then they are divisible. In
the first proposition I assert the state of affairs [die Sache] directly; in the
second, I assert it only under a condition expressed problematically.15

Ibid., p. 106.

In the hypothetical judgment, the predication expressed by the conse-
quent is asserted only under the condition of the predication expressed by
the antecedent. This is why, again, in a hypothetical syllogism, the ante-
cedent being posited, the consequent should also be posited.
Kant says of hypothetical syllogisms that they are not really syllogisms,
because in them there is no middle term: one simply converts the ante-
cedent from a problematic to an assertoric proposition, and thus provides
the ground for asserting the consequent in the conclusion. As Kant explains:
A hypothetical syllogism is one that has a hypothetical proposition for its
major premise. It consists in two propositions: 1- an antecedent and 2- a
consequent, and it is achieved either through modus ponens or through
modus tollens.
Note: Hypothetical syllogisms thus have no middle term, but in them
the Konsequenz of a proposition from another proposition is shown. “
Namely, the major premise expresses the Konsequenz of two proposi-
tions, of which the first is a premise, the second is a conclusion. The
minor premise is a transformation of the problematic condition into a
categorical proposition.16

So, to take up Kant™s example of hypothetical judgment above, a
hypothetical syllogism formed from such a judgment would be: ˜˜If
bodies are composite, then they are divisible; bodies are composite; so,
they are divisible.™™ One might want to make explicit the fact that what is
asserted in the consequent is asserted of all bodies, together with the
Wolffian idea of an added condition, and write: ˜˜All bodies, if composite,
are divisible; all bodies are composite; so, all bodies are divisible.™™
But there can be more complex cases: cases that combine features of
categorical and hypothetical syllogisms. In a categorical syllogism, there
is a middle term by the mediation of which, in the conclusion, a parti-
cular class of objects is subsumed under the predicate of the major
premise. Now, consider the syllogism: ˜˜All stones, if lit by the sun, get
warm; the stones along the river are lit by the sun; therefore, they get
warm.™™ Such a syllogism combines features of a categorical (subsump-
tion of the subject of the minor premise under the subject, and thus
under the predicate, of the major premise) and of the hypothetical
(assertion in the minor premise of the antecedent of the major premise).
I suggest that when Kant talks of the ˜˜strict universality of a rule™™
contained in the concept of cause, what he has in mind is precisely this

Jasche Logic, x75, AAix, p. 129.

kind of mixed premise. And the ˜˜positing™™ of something that results in
the ˜˜positing™™ of something else similarly has the features of both the
categorical subsumption of an instance and the hypothetical assertion of
the antecedent. In other words, to think a causal connection between the
stone™s being lit by the sun and the stone™s becoming hot is to think that
the proposition ˜˜this stone is lit by the sun™™ being posited, the proposi-
tion ˜˜this stone is becoming hot™™ should be posited, which amounts to
thinking the first as an instantiation of the antecedent, the second as an
instantiation of the consequent, in the (implicit) strictly universal rule: all
stones, if lit by the sun, get warm. This is why Kant says that the concept
of cause (the sun™s being by its light the cause of the stone™s getting hot)
contains ˜˜the strict universality of the rule.™™
Three caveats: first of all, what we are talking about in a causal judg-
ment are empirical states of affairs. Kant™s question, like Hume™s, is how
a necessary connection can be thought to exist between two distinct
states of affairs which we know only empirically (matters of fact).
Second, the relation between antecedent and consequent has to be
synthetic: if asserting the predicate of the consequent follows analytically
from asserting the antecedent, then we do not have a causal connection.
˜˜All bodies, if composite, are divisible™™, is such an example. Third, if the
connection is itself an empirical generalization, we do not have a causal
connection. For instance, ˜˜All stones in this garden, if the sun shines on
them, are warm (I™ve checked).™™ This is not a causal connection.
To sum up, for the hypothetical to express a causal connection, it has to
be the case (1) that the states of affairs connected in the judgment are
empirical, (2) that the connection is synthetic, (3) that it has strict univer-
sality. Because only if I can presuppose a premise with strict universality
can I state, given one case (the sun shines of the stone) that this case being
posited, the consequent must be posited: the stone gets warm. In other
words, to think a causal connection between two states of affairs is to think
the one as the posited antecedent (in the minor premise) the other as the
posited consequent (in the conclusion) of a strictly universal rule.
In Kant™s terms, then, the difficulty inherent in the concept of cause
may be reduced to the following: how can a hypothetical judgment be
universally and necessarily true although it is not analytically true (its
consequent is not analytically contained in its antecedent)? When trying
to ground the ˜˜strict universality of the rule™™ contained in the concept of
cause, Kant is on the look-out for hypothetical judgments in which the
connection of antecedent and consequent is as strictly universal and
necessary as is the analytical connection of concepts or propositions; he

is looking for a way to move from the ˜˜positing™™ of the antecedent to the
˜˜positing™™ of its consequent by a modus ponens which is as rigorously
grounded as a modus ponens formed from an analytically true proposi-
tion, even though the connection contained in the supposed premise is
in fact synthetic, and its components are wholly empirical. To say that
there is a causal relation between the stone™s being lit by the sun and its
being warm is to say that the if . . . then connection between these two
states is as necessarily and universally true as the if . . . then connection
between perfect justice and the punishment of the wicked, although in
the case of the stone™s being lit and its getting warm I have only repeated
observation to vouch for my statement of an if . . . then connection.
The transition from a judgment which merely recounts repeated observ-
ation to a causal judgment (which amounts to claiming universal validity
and necessity for the rule: ˜˜If a stone is lit by the sun, then it gets warm™™) is
what Kant calls, in the Prolegomena, a transition from a mere ˜˜judgment of
perception™™ to a ˜˜judgment of experience.™™ A judgment of perception, he
says, holds only ˜˜for me, and in the present state of my perception.™™
A judgment of experience, if true, is true ˜˜for all, and at all times.™™ This
is because what it expresses is not just a repeated combination of per-
ceived events (I have repeatedly experienced that when the stone was lit
up by the sun, it became warm), but a connection in the objects themselves
such that if sun shines on the stone, the stone gets warm.17 But what
makes it possible to assert such a connection? What allows the transition
from the mere statement of a repeatedly observed occurrence (judgment
of perception) to a hypothetical judgment for which we claim the ˜˜strict
universality of a rule™™ (judgment of experience)? Kant™s response is that
we presuppose the necessary truth of another judgment, prior to both the
judgment of perception and the judgment of experience. We presuppose
the truth of a judgment that states that appearances, the objects of our
perception and experience, are ˜˜in themselves determined™™ with respect
to the logical form of our hypothetical judgment. We presuppose, in other
words, that appearances are in themselves, as empirical objects, connected

Cf. the striking manner in which Kant defines the role of the categories in Prolegomena
x21“a, AAiv, p. 304:
˜˜The judgment of experience must add to the sensible intuition and its logical connec-
tion in a judgment (according to which it has been made universal by comparison) some-
thing that determines the synthetic judgment as necessary and thereby as universal; and
this can be nothing other than the concept which represents the intuition as in itself
determined with respect to one form of judgment rather than another [my emphasis]: that is to
say, a concept of that synthetic unity of intuitions which can be represented by a given
logical function of judgment.™™

by a chain of causal connections, or we presuppose the universal validity of
the causal principle. Because we make such a presupposition, we allow
ourselves, upon repeated observation of similar events, to move from
such repeated observation to a causal judgment (a hypothetical for
which we claim the ˜˜strict universality of a rule™™).
What justified such a presupposition? For an answer to this question,
in the Prolegomena Kant merely refers us to the Critique of Pure Reason.18
Before we consider this answer, we can already note that the structure of
Kant™s response to ˜˜Hume™s problem™™ turns out to be the exact reverse
of Hume™s own. Hume derived the universal principle from the parti-
cular cases of causal connections and the particular cases of causal con-
nection from repeated successions of similar events. Kant says that we
derive a causal connection from any given repetition of similar events
because we already have the universal causal principle.
What we need from the Critique, then, are answers to three main
questions. First, is it the case that we presuppose the truth of the causal
principle? Second, supposing we do presuppose its truth, is it indeed
true, i.e. do we have the right to presuppose its truth? (This is, in effect,
the quid juris question of the Critique: by what right do we make use, in
empirical knowledge, of concepts such as that of necessary connection?
[cf.A84/B116].) Third, supposing we do presuppose the truth of the
causal principle, and supposing the principle is indeed true, how does
this general presupposition warrant in any particular case the transition
from observing a mere regularity (what Kant calls a ˜˜judgment of per-
ception™™) to asserting a causal connection, which we claim holds ˜˜for all,
and at all times™™ because it holds of the (empirical) objects themselves,
and is thus what Kant calls a ˜˜judgment of experience™™?19
In order to see what answers, if any, Kant offers to these questions,
I now turn to his argument in the Second Analogy of Experience.

Causality and perception of objective temporal succession
In the first edition of the Critique, Kant gives the following formulation
for the Second Analogy of Experience: ˜˜Everything that happens
(begins to be) presupposes something which it follows in accordance

See the footnote to x22 of the Prolegomena (AAiv, p. 299) where Kant refers his reader to
A137ff, i.e. the beginning of the chapter: On the Schematism of Pure Concepts of the
Cf. Prolegomena, x18, AAiv, pp. 297“8.

with a rule™™ (A189).20 Stressing as it does the notion of a rule, this
formulation is closely related to the logical unpacking of the concept of
cause I have just laid out. Indeed, I intend to show that this logical
unpacking helps clarify the terms and method of Kant™s proof.
As anybody who has battled with the first Critique knows, Kant pro-
vides in the Second Analogy not one, but five different expositions of his
proof of the causal principle, each of them quite tortuous. All of them
arguably share the following steps:
1 Our apprehension is always successive (premise).
2 This by itself does not tell us whether the succession of perceptions in
our apprehension is the perception of an objective succession (a
change of objective states) (premise).
3 We experience the succession in our apprehension as the perception
of an objective succession just in case we consider the subjective
succession as order determinate (i.e. we know that we could not
reverse the order of our perceptions and perceive A after, instead of
before perceiving B; or perhaps better, we are aware, while perceiving
B, that should we at that instant return our gaze to the point where we
perceived A, we would not perceive A there)21 (premise).
4 We consider the subjective succession < A, B > as order determinate
just in case, relating it to an object, we recognize a change of state in
the object, which means that we presuppose that < A, B > follows
from a preceding objective state according to a rule (premise).

In the second edition, the formulation of the Second Analogy is: ˜˜All alterations occur in
accordance with the law of connection of cause and effect™™ (B232). This formulation is
closer to the one I quoted earlier from the introduction (also in B). The B formulation is
interesting in that it stresses Kant™s equation of an event with the alteration of a (relatively)
permanent thing, and thus the connection of the Second Analogy with the First. It also
emphasizes the notion of a law rather than that of a rule, and thus the necessity of the
causal connection. But as I say in the main text, in my view the superiority of the A edition™s
formulation lies in its staying closer to the discursive model from which Kant derives the
meaning of the concept of cause.
For an explanation of the correction I propose after ˜˜perhaps better,™™ see my comment on
premise 3 below, and n. 26.
One may wonder what work premises (1) and (2) do in the argument, since according to
my analysis, (5) follows from (3) and (4), without any reference to (1) and (2). Their only
role, in the pattern I am laying out here, is to prepare the way for (3), although (3) does not
strictly speaking follow logically from them. All one can say is that once one has recognized
that perceptions in our apprehension are always successive, and that one can draw from
this mere succession in apprehension no instruction at all as to the objective order of
things, then one is prepared for the statements of (3) and (4), from which (5) follows.
Thanks to Colin Marshall for pressing me on this point.

5 Therefore, we perceive a succession as an objective succession (a
change of states in an object) just in case we presuppose a preceding
state upon which it follows according to a rule (from [3], [4]).
6 Therefore, every objective succession (every event) presupposes
something upon which it follows according to a rule (from [5]).22
I shall be relatively brief on premises (1), (2), and (3) of my outline,
which do not strike me as extremely problematic. I shall spend more
time on premise (4), for which the formulation I have proposed is both
complex and controversial. And of course there is an obvious difficulty in
the move from (5) to (6) (which respectively answer the first and second
questions we needed the Critique to answer “ see the concluding remarks
of section 1 above: do we presuppose ˜˜something which what happens
follows, according to a rule™™? Yes, if the argument is sound. Does it mean
that what happens ˜˜presupposes something which it follows according to
a rule™™? We need to see if Kant gives an argument for this additional

Premises (1) and (2)
They are, in a peculiar sense, phenomenological descriptions of our
perceptual experience: whatever we perceive, we perceive successively.
Even a permanent, unchanging object or realm of objects, we perceive
by means of a continuous succession of perceptual states. A fortiori,
apprehending or directing our attention to different parts of an unchan-
ging object, or to different objects, or to successively appearing and
disappearing states of one object, is itself always successive. In his
Analogies of Experience, Kant gives examples for each of these three
cases: the first (directing our attention to different parts of one object) is
exemplified by the perception of a house, in which we successively
apprehend the bottom part and then the top part, or the top part and

The obvious difficulty in the transition from (5) to (6) is that Kant seems to move blithely
from an epistemic point (˜˜in experiencing objective succession, we presuppose™™) to an
ontological/transcendental point (˜˜objective succession itself presupposes™™). A possible
ground for the move might be an implicit reliance on the Transcendental Deduction:
˜˜The conditions of possibility of experience are the conditions of possibility of the object of
experience™™ (A111, A158/B197). But I shall also suggest that there is, within the argument
of the Second Analogy itself, a move that is meant to ground the transition from (5) to (6),
and which essentially repeats the argument which, in the Transcendental Deduction,
supported the view that the categories are not just concepts according to which necessarily,
we think about objects, but concepts that universally represent features of the objects
themselves. More on this below, especially in the third section of this chapter.

then the bottom part, the right side and then the left side, or the left side
and then the right side, and so on; the second (directing our attention to
different coexisting objects) is exemplified by perceiving the earth
around us and then the moon, the moon and then the earth; the third
(perceiving different states of one and the same object) is exemplified by
the famous example of perceiving successive positions of a ship sailing
down a river, which I shall analyze in some detail in a moment.23
These examples are meant to show not only that our apprehension is
successive, but also that, notwithstanding this successive character of
our apprehension, we do distinguish, without even having to reason
about it or to reflect upon it, an objective succession (in the case of
the ship) from a succession which is merely subjective and for which the
objective temporal relation is one of simultaneity (in the case of the
parts of the house or in the case of the moon and the earth). In fact,
what requires some reflection in order for us to become aware of it is
the successive character of the subjective apprehension: as I said, we
quite naturally and without any reflection at all perceive the parts of the
house, or the moon and the earth, as simultaneous, without being
aware that our apprehension of them is successive. This makes pre-
mises (1) and (2) rather surprising, but nonetheless, upon reflection,

For Kant™s analysis of these examples, see A190“1/B235“6, A192“3/B237“8, B257.
A standard objection to Kant™s reasoning here is that in fact, contrary to Kant™s claim, the
manifold of representations is not at all always successive. See for instance Lewis White
Beck, Essays on Kant and Hume (New Haven: Yale University Press, 1978), p. 144:
˜˜Kant assumes that the manifold of representations is always successive. This is certainly
wrong. When I open my eyes I do not scan the visual field as if my eyes or my attention
worked like the electron ejector in a television tube, aiming first at one point and then at an
adjacent point. But as a consequence of his sensational atomism, Kant assumes that my
apprehension does work in this way.™™
Beck adds, however, that this error does not destroy Kant™s argument altogether. For
this argument to be relevant, says Beck, it is sufficient that the temporal order of our
perceptions and that of objective states be sometimes different. ˜˜It is the difference in
temporal orders, and not the putatively necessary successivity of representations, which generates
the problem of the Analogies.™™ However, I do not think that Kant makes the error Beck
attributes to him. Kant certainly does admit that an object or a scene can be perceived uno
intuitu. What matters is that we acquire detailed awareness of each of its elements only by
successively apprehending it. And when the latter occurs, our awareness of objective
succession, as opposed to objective simultaneity, is an (implicit) awareness of rule-
governed change, as opposed to rule-governed coexistence (as the Third Analogy will
establish). On Kant™s analysis of our awareness of objective temporal order as an awareness
of the rule-governed character of our perceptions, see below my analysis of the third and
fourth premises.

Premise (3): objective succession and order-determinateness
of the subjective succession
Premise (3) says that we perceive a succession as objective (as a succession of
states in the object) just in case we consider the subjective succession as
order-determinate, i.e. just in case we are aware that we could not reverse
the subjective order of our apprehension and perceive A again after having
perceived B. Of course, as most commentators of Kant have noticed, this
does not mean that the objective succession itself cannot be reversed. When
I perceive that a ship successively occupies positions p1, p2, and p3 (as
opposed to my successively perceiving different, coexisting ships at p1, p2,
p3), I am aware that the ship could reverse its movement and go back from
p3 to p2, and from p2 to p1. But in the present circumstances, my percep-
tion of the scenery could not be reversed. I could not direct my perception
back to p2 or to p1 and expect to see the ship. Whereas if I interpret what I
apprehend successively as a perception of three coexisting ships, I do
expect the succession of my perceptions to be repeatable in the reverse
order. I would see the very same ships if I directed my glance back where I
directed it a moment ago (unless, again, the ships moved; but even then
I would at least expect to see them somewhere).25

Premise (4): order-determinateness of the subjective succession, relation
to an object, supposition of a rule
Premise 4, as I have formulated it, says: ˜˜We consider the subjective
succession as order-determinate just in case, relating it to an object, we
recognize a change of states, which means that we presuppose that it
follows from a preceding state according to a rule.™™

A more disturbing objection might be that raised by the perception of time-reversible
processes, like the movement of a pendulum. How could we possibly think, in this case,
that we could not perceive A (position 1 of the pendulum) after perceiving B (position 2 of
the pendulum)? It is precisely for such a case that I proposed a corrected version of
premise 3 above: we are aware, while perceiving B, that should we at that instant return
our gaze to the point where we perceived A, we would not perceive A there. This alter-
native formulation also has the advantage of stressing the role of imagination: we have to
use our imagination to bring to mind what we would perceive at a point we are not
presently perceiving. Of course, even to this, one might object that we just ˜˜see™™ the
pendulum in its successive positions. But there I would say that Kant™s arguments are
first directed at cases of interrupted perception. Uninterrupted perception itself is inter-
preted in the light of our ability to unify interrupted perceptions. This is how we constitute
for ourselves representations of unified and coherent realms of objects and their temporal

Packed into this one premise are three conditions for interpreting the
subjective succession as order-determinate: (a) we have to relate it to an
object; (b) relating it to an object, we have to recognize a change of states
of the object; (c) recognizing a change of states means presupposing that
it follows from a preceding state according to a rule. I now want to
consider these three conditions one by one.
(a) What makes us consider, or interpret, the subjective succession as
order-determinate is that we relate it to an object. Or more precisely,
what makes us so perceive it is that we implement the mental activity of
relating our subjective apprehension to an object it is the apprehension
of. Now, against the argument so construed one might want to observe
that premises (3) and (4) seem suspiciously circular. Premise (3) says that
to interpret the subjective succession as also objective is to interpret it as
order-determinate. But premise (4) says that what triggers the interpret-
ation of the subjective succession as order-determinate is that we relate
our perceptions to an object. To perceive as objective is to perceive as
order-determinate; but one perceives as order-determinate only if one
relates to an object. Actually, this is not circular because ˜˜perceiving as
objective™™ (in premise [3]) and ˜˜relating to an object™™ (in premise [4]) are
not the same. Relating the subjective succession to an object (premise [4])
might result in considering it as the perception of an objective coexist-
ence, and thus as a merely subjective succession, a succession merely in
apprehension (in which case premise [3] would not be satisfied). This is
so when, relating subjective succession to an object, I recognize that what
I apprehend successively are really three generically identical ships
which exist at the same time, and not three objectively successive states
(different positions in space) of one ship. Similarly, to take Kant™s own
examples, relating the subjective succession of my perceptions to an
object may result in my recognizing that what I am perceiving are the
objectively coexisting parts of one house, or two heavenly bodies existing
at the same time (not one body which has moved from one position in
space to another). So, Kant™s point is that the mental activity of relating
perceptions to objects they are perceptions of, just is what generates
(without our even needing explicitly to reason about it) our representa-
tion of objective correlations in time. And this in turn gives its character
of irreversibility or reversibility to the subjective succession of percep-
tions in our apprehension. In fact, we have quite familiar examples of
what happens when this activity does not take place: in day-dreaming we
may let our eyes wander from one object to the next, let changes in the
objects we are taking in occur, without discerning the occurrence of any

objective event or recognizing any coexistence of objects or, for that
matter, recognizing any persisting object at all.
An interesting aspect of Kant™s argument is that it reverses the depend-
ence relation between the features of our perceptions and our relating
these perceptions to objects we take them to be the perceptions of, found in
seventeenth- and eighteenth-century theory of ideas. For Descartes no less
than for Locke, Berkeley, or Hume, perceiving external objects is a matter
of interpreting features of our ideas and their combinations. The four
authors differ, of course, in the tools they admit we have at our disposal
for interpreting these features. Therefore they also differ in the kinds of
inferences they allow concerning the existence and objective properties of
external objects. But for Kant, the features of our ideas (representations),
and especially the modal characteristics of their temporal relations, depend
themselves upon our mental acts of relating them to objects they are the
perceptions of. We take the succession of our representations to be order-
indeterminate or order-determinate depending on whether we are led to
interpret them as representing an objective simultaneity or an objective
succession. So the order we introduce into the subjective succession of our
representations depends on how we interpret the objective order we take
them to be the representations of. Thus Kant writes:
In the previous example of a house my perceptions could have begun at
its rooftop and ended at the ground, but could also have begun below
and ended above; likewise I could have apprehended the manifold of
intuition from the right or from the left. In the series of perceptions
there was therefore no determinate order that made it necessary when I
had to begin in the apprehension in order to combine the manifold
empirically. But this rule is always to be found in the perception of that
which happens, and it makes the order of perceptions that follow one
another (in the apprehension of this appearance) necessary.
In our case I must therefore derive the subjective succession of apprehension
from the objective succession of appearances [my emphasis], for otherwise the
former would be entirely undetermined and no appearance would be
distinguished from any other. (A192“3/B237“8)26

Or again:
If we investigate what new characteristic is given to our representations
by the relation to an object, and what is the dignity that they thereby

Cf. premise (41) (premise [4] in the first exposition of the proof ) quoted in the appendix

receive, we find that it does nothing beyond making the combination of
representations necessary in a certain way, and subjecting them to a rule,
and conversely, that objective significance is conferred on our represen-
tations only insofar as a certain order in their temporal relation is
necessary. (A197/B242“3)27

Here the seemingly circular aspect of Kant™s argument, mentioned
above, is quite visible. Relating our representations to an object confers
upon our representations a character which they would not otherwise
have (their temporal order-determinateness, or on the contrary their
order-indifference). But this character is what makes them representa-
tions of objective succession or, on the contrary, of objective coexistence.
Strawson, following Lovejoy, famously characterized Kant™s argument in
the Second Analogy as a ˜˜non-sequitur of numbing grossness.™™28 This
damning statement rests, I think, on a misunderstanding of Kant™s mean-
ing when he says that ˜˜in [the perception of an event] we must derive the
subjective succession of apprehension from the objective succession of
appearances.™™ Strawson understands Kant to be saying that we interpret
the subjective succession in our apprehension as irreversible when we think
that it is causally determined by an objective succession. The subjective
succession (say a, b) is then thought to be necessary in that its order
is thought to be constrained by the order of the objective states of affairs
(say A, B). This is a reasonable thing to say, Strawson comments. But from
this Kant then proceeds to his astoundingly gross non-sequitur. He claims
not only that the subjective succession is necessary (being causally deter-
mined by the objective succession); but that the objective succession which
determines it is itself necessary as well. Thus every necessary order of the
subjective succession is the perception of a necessary, i.e. causally deter-
mined, objective succession.29 This certainly is a resounding non-sequitur.
But in fact, if I am right in the analysis I have proposed, this causal/
representational account of perception plays no role in Kant™s argument.
In fact, making use of such an account for the proof of the causal principle
would be an even grosser non-sequitur than the one Strawson denounces:
for the purpose of proving the universal applicability of the concept of
cause it would make use of the very concept whose applicability is in

This passage belongs to the transition between the second and third exposition of the
argument, where Kant explains that to understand his argument we need to reflect on
what we mean by an object: see appendix to this chapter, explanation of ¶11.
Strawson, Bounds of Sense, p. 137.
Ibid., pp. 137“8.

question. But, as I pointed out earlier, Kant™s account of perception here is
not causal, but phenomenological, in an original sense. Kant asks: what is it
about our perception that makes a subjective succession the perception of
an objective succession? He responds in the two steps (3) and (4): what
makes our successive perception in apprehension the perception of an
objective succession is the awareness of its temporal order-determinateness;
what generates the awareness of its temporal order-determinateness is the
fact that we intentionally relate it to an object.
But how does this happen? Why should relating our representations to
an object generate an awareness of its order-determinateness, and why
should this awareness warrant perception of an objective succession? This
is what the second and third conditions stated in premise (4) are meant to
explain. I now consider those two conditions.
(b) and (c) What we want to explain is why relating successively
apprehended perceptions to an object should generate a representation
of order-determinateness (in the case where the subjective succession is
interpreted as the perception of an objective succession) or of order-
indifference (in the case where the subjective succession is interpreted as
the perception of simultaneously existing objects or states of objects).
I shall consider only the first case, which is the one the Second Analogy is
about. Why does relating perceptions to an object generate the order-
determinateness of the subjective succession in the case where it is the
representation of an objective succession?
To relate perceptions to an object of which they are the representa-
tion, is to recognize an object under a concept.30 I perceive patches of
grey out there, in a vaguely rectangular shape. I relate these perceived
patches to an object when I recognize a tower: for instance, the tower of
the science building in the University of Paris. Now, recognizing an
object under a concept is either recognizing it under ˜˜permanent™™
characters, characters it could not cease to have without ceasing to be
the kind of object I identified it as being, or recognizing it under chan-
ging characters: characters it can acquire or lose without ceasing to be
the kind of object it is. The tower could possibly be painted bright red by

Kant argued for this point in the Transcendental Deduction of the Categories; see A103/
B137. This does not mean that all object-related representations presuppose a concept.
Intuitions are also representations we relate to objects (cf. A320/B377). But intuitions
(˜˜representations that relate immediately to the object and are singular™™) (A320/B377, cf.
Jasche Logic, x1), left to themselves, are ˜˜blind™™ (A52/B76). Knowing what we intuit, and
re-identifying objects, i.e. recognizing them as persisting through time, is possible only if
we recognize them under concepts.

angry students, its windows could be broken, or it could be wrapped in
cloth by Christo. None of this would stop me from identifying it as the
tower of the science building. I would say: ˜˜Hey, look what happened!™™
But I still would identify the object. On the contrary, if I saw a similar
tower a hundred miles from Paris, I would either have to doubt my eyes
or have to suppose that this was in fact another tower. A ship moves, but
not a tower. So, in the previous cases I identified the object as the tower
of the science building because there are any number of plausible exter-
nal conditions I could formulate for the change of some of its familiar
characters. In the second case, although it is not entirely impossible that
some external conditions might account for the fact that the science
building has changed position in space, it is highly implausible. In my
identification of the object, then, the supposition of plausible conditions
for the change of state of the object I recognize is just as important as the
recognition of its permanent characters.
But analyzing our perception of objective succession in this way seems
to take us back to something like the syllogistic model I laid out in the
first section of this chapter.31 For Kant is telling us that we interpret a
subjective succession of perceptions as an objective change of states only
if we can suppose the condition of a rule according to which this change
of states occurs. In other words, he is telling us that our judgment about
the object (˜˜the object has altered, it has passed from state 1 to state 2™™)
can be considered as the conclusion of a hypothetical syllogism whose
premises we do not know. For instance, when I perceive the ship as
having changed position, all I actually perceive is a ship in position p1 at
t1 and a ship in position p2 at t2. But I interpret in this way the succession
of my perceptions because it is coherent with my experience of what it is
to be a ship, to suppose that a ship changes position given the right
circumstances. And this means that I can presuppose a rule, without
being able to specify what this rule is, so that what implicitly goes on in
my mind is something like: ˜˜[(1) A ship, if subjected to conditions XYZ,
alters its position; (2) this ship is subjected to conditions XYZ;] (3)
Therefore this ship has altered its position™™ (only this last proposition
expresses what I actually perceive).
The supposition of a rule in this manner is just what Kant asserts, for
instance, in the following passage:

Only ˜˜something like™™ this model because, as we shall see, it is not obvious that here we
have a strictly universal rule, and thus the representation of a necessary connection. I shall
return to this point later.

In the synthesis of the appearances the manifold of representations is
always successive. Now no object at all is hereby represented, since
through this succession, which is common to all apprehensions, nothing
is distinguished from anything else. But as soon as I perceive or presup-
pose [sobald ich aber wahrnehme, oder vorausannehme, my emphasis] that in
this succession there is a relation to the preceding state, from which the
representation follows in conformity with a rule, I represent something
as an event, as something that happens; that is to say, I cognize an object
that I must place in time in a determinate position which, after the
preceding state, cannot be otherwise assigned to it. (A198/B243)32

This gives us the conclusion I stated earlier:
5 Perceiving an objective succession (a change of states in an object) is
presupposing a preceding state upon which it follows according to
a rule.

What is all-important here is that we do not know the rule, but only presuppose one, and
this presupposition is necessary for the perception of an objective succession. On this
point, my analysis of Kant™s argument differs from Paul Guyer™s. According to Guyer, with
the example of the ship Kant intends to show that we can confirm that the ship is sailing
downstream only if this interpretation of our perception is in accordance with known
causal laws:
˜˜Kant™s theory is . . . that it is only if we are in possession of causal laws which dictate that
in the relevant circumstances “ that is, not in general, but in the particular circumstances of
wind, tide, setting of the sails, and so forth, which are assumed to obtain “ the ship could
only sail downstream, that we actually have sufficient evidence to interpret our representa-
tions of it to mean that it is sailing downstream™™ (Guyer, Kant and the Claims of Knowledge,
p. 252). An immediate objection is that this would make Kant™s argument circular: to know
the causal laws compatible with the perceived movement of the boat, we would have to
have already confirmed our perception of this movement on many previous occasions, as
well as confirmed the correlations that can be asserted as causal laws. In response to this,
Guyer insists that Kant™s argument should be understood ˜˜not as a psychological model of
the generation of beliefs, but as an epistemological model of the confirmation of beliefs™™
(p. 258). But Kant claims more than this. He does claim to give an account of the genera-
tion of our belief that a succession is objective. Guyer™s mistake, it seems to me, is to think
that for Kant the rule presupposed in every perception of objective succession is a rule (or a
set of rules) we actually know.
Note that Guyer and Wood translate vorausannehme by ˜˜anticipate.™™ I translate it by
˜˜presuppose.™™ This difference may be related to the difference between my interpretation
and Guyer™s. I suppose Guyer understands Kant to be saying, in this passage, that if we
anticipate that we will be able to find an explanation for the succession (and thus justify our
belief in the existence of an objective succession), then we take it to be objective. I under-
stand Kant as saying that we perceive a succession as objective if and only if we can suppose
an antecedent for a rule according to which it occurs. The two interpretations are not
incompatible: see what I say below in the main text. But my interpretation attributes a
more radical view to Kant: to borrow Guyer™s own terms, I take Kant to intend to provide
an account of the generation of our belief in the existence of an objective succession, rather
than just an account of the justification of that belief.

Now, supposing we accept this line of reasoning, it is still not clear how
it would justify Kant™s move from (5) to (6), namely from the assertion
that ˜˜perceiving something as happening is presupposing something
upon which it follows according to a rule™™ to the assertion of the principle
as Kant formulates it: ˜˜Everything that happens presupposes something
upon which it follows according to a rule.™™
We can say of course that making the presupposition commits us to
strive to confirm our perception by actually determining the rule or set
of rules according to which we can take the perceived happening to have
actually occurred. In this sense, to say that perceiving that something
happens is presupposing something else upon which it follows according
to a rule is also to say that confirming our perception as the perception of
an actual event, or confirming that something has happened, is deter-
mining the rule, or set of rules, which warrants asserting that what
happened is actually the event we perceived.33 And on the contrary,
finding out that such an event is in contradiction with all known rules of
our experience would tend to disconfirm the perception as perception
of the event we think we have identified. If I find a warm stone where
there has been no sun, no spring of hot water, nobody to light a fire, no
other known ˜˜antecedent of a rule,™™ I have to start worrying that per-
haps what I have in front of me is a dangerous, unknown material from
outer space or a particularly weird animal. But the event: ˜˜this stone,
which normally is cold, has become warm™™ is put into question. Still, even
this makes the principle only an epistemic principle, a principle by which
we would confirm or disconfirm our belief that something has hap-
pened, not an ontological principle, universally true of happenings
One could perhaps provide a further answer by pointing out that
Kant has restricted the meaning of ˜˜object™™ to ˜˜object of possible experi-
ence,™™ and that ˜˜everything that happens™™ should be understood as
˜˜everything we can possibly experience as happening,™™ that is to say
˜˜everything we can meaningfully call an event or a change of state of
an object.™™ If we identify an event as event X only by supposing possible
rules according to which it happens, then this is just what it takes to be an
event X. Moreover, it is not just by arbitrary whim that we suppose



. 5
( 10)