‘In Itself’: A New Investigation of Kant’s Adverbial Wording of Transcendental Idealism

Abstract

This article offers the first systematic investigation of the linguistic forms in which Kant expresses his transcendental idealism since Gerold Prauss’ seminal book Kant und das Problem der Dinge an sich. It is argued that Prauss’ own argument for the claim that ‘in itself’ is an adverbial expression that standardly modifies verbs of philosophical reflection is flawed and that there is hence very poor exegetical evidence for so-called ‘methodological two-aspect’ interpretations of Kant’s transcendental idealism. A comprehensive investigation of Kant’s adverbial uses of ‘in itself’ rather reveals that there are various groups of verbs that ‘in itself’ modifies, which can more or less all be reduced to a standard use in which ‘in itself’ modifies verbs of predicative being such as the copula. It is also discussed how these findings can be accommodated by the two main alternative kinds of interpretations, that is, two-object and ontological two-aspect interpretations.

1. Introduction

One of the main exegetical debates about Kant’s transcendental idealism concerns the question of how to understand his distinction between appearances and things in themselves. According to so-called ‘two-world’, or ‘two-object’, interpretations, Kant’s doctrine implies a distinction between two different, non-overlapping kinds of objects, which are denoted by the sortal terms ‘appearances’ and ‘things in themselves’. The first kind is one of intentional entities whose being can be reduced to their being represented through sensible intuitions that are conceptualized by the understanding, while the second exist independently of whether and how we represent them. Kant’s doctrine of the ideality of space and time is then interpreted as implying that only appearances exist in space and time, whereas things in themselves do not. ¹

So-called ‘two-aspect’ interpretations, on the other hand, take Kant’s appearance/in-itself distinction not to be one between two different kinds of objects but rather between two different kinds of aspects of the very same objects. It is a distinction between the way things appear to us when we represent them in sensible intuition and the way these very things are in themselves, that is, independently of whether and how we represent them. Consequently, Kant’s doctrine is interpreted as saying that the objects that we encounter in experience are not by themselves in space and time but only insofar as they appear to us. ²

One of the criteria that are used in order to decide the question of which of these two interpretations gets Kant right is, of course, how well they can deal with the overall textual evidence. Since Gerold Prauss’ seminal book Kant und das Problem der Dinge an sich, one part of this textual evidence is considered to be the syntactic form of Kant’s actual wording of the ‘in-itself’/‘appearance’ distinction (Prauss 1974). Does Kant use ‘in itself’ as an adverbial expression, that is, as a modifier of a verbal clause, or as an adnominal expression, that is, as a modifier of a noun phrase? This question seems to have a bearing on the debate between two-object and two-aspect interpreters. For when the two-object interpreter says, ‘No thing in itself is spatial’, she uses the expression ‘in itself’ adnominally, that is, as a modifier of the noun ‘thing’, whereas when the two-aspect interpreter says, ‘No thing is spatial in itself’, she uses ‘in itself’ adverbially, that is, as modifier of the copula ‘is’ or of the verbal phrase ‘is spatial’. There are undoubtedly passages in which Kant uses ‘in itself’ adnominally and passages in which he uses it adverbially. ³ The question is whether one of the two uses is somehow more fundamental and, if yes, which one that is.

In his book, Prauss famously argued that a careful and systematic look at Kant’s overall use of the term ‘thing in itself’ (and related terms) shows that it is the adverbial use that the interpreter should trust. Prauss also claimed that his findings support a specific variant of the two-aspect view. He argued that Kant’s standard verbal complement for the adverbial use of ‘in itself’ is verbs such as ‘consider’ or ‘regard’ which express modes of philosophical inquiry. Hence, the two aspects that Kant wants to distinguish correspond to two different ways in which we as philosophers can consider empirical objects in our philosophical theories, rather than to any ontological difference on the side of the considered objects. This general idea is known under the label ‘methodological two-aspect interpretation’ and has become the most influential through the version of it that Henry Allison presented (Allison 1983, 2004). It contrasts not only with two-object interpretations but also with so-called ‘ontological two-aspect interpretations’, which interpret Kant’s appearance/in-itself distinction as capturing some ontological difference, namely, one between two different kinds of properties of the objects we encounter in experience. ⁴

It is a curious feature about the role of Prauss’ argument for Kant scholarship that, although it is cited in almost every comprehensive treatment of Kant’s transcendental idealism, it has never been scrutinized with any systematic rigour. Some interpreters have claimed that Prauss’ linguistic findings support their own reading (see, e.g. Allison 1983: 8; Allison 1987: 160n7; Allison 2004: 51f; Collins 1999: 58n16., 150); others have remained unimpressed by it (cf. Guyer 1987: 334n4; Van Cleve 1999: 49n; Allais 2015: ch. 4; Jauernig 2021, 14n42 ⁵ ) and have noted that there are passages that do not seem to fit Prauss’ analysis (cf. Strube 1976; Aquila 1979: 296; Aquila 1983: 89–91; Ameriks 1982: 6–7). However, to my knowledge, no one has ever answered Prauss’ analysis by offering their own systematic investigation of Kant’s overall use of the expression ‘in itself’. I think that this is unfortunate. On the one hand, I agree with Prauss that a detailed description of the linguistic forms in which Kant expresses his transcendental idealism should be an important part of the evidence for any interpretation of the latter, and I also agree with him that such a description reveals that Kant’s fundamental use of ‘in itself’ is adverbial. On the other hand, I think that Prauss’ own argument for the priority of the adverbial use is rather awkward and – more importantly – that his claim that the standard complement of ‘in itself’ is verbs of philosophical consideration is very poorly supported by the evidence he presents for it. Or so I will argue in the following.

In the next section, I will give a short summary of Prauss’ argument and explain why I think it is not conclusive. In section 3, I will present the results of my own systematic investigation of all occurrences of the phrase an sich (‘in itself’) in all relevant Kantian works. In section 4, I will conclude by sketching what does and what does not follow from my results for an interpretation of Kant’s transcendental idealism.

2. Prauss’ argument

Prauss’ argument consists of two steps. In the first step, Prauss undertakes a systematic count of all occurrences of the phrase ‘Ding/Gegenstand/Sache/Objekt an sich’ on the one hand and the longer phrase ‘Ding/Gegenstand/Sache/Objekt an sich selbst’ on the other in volumes 3, 4, and 5 of the Academy Edition of Kant’s writings (Prauss 1974: 13–15). The reason why he thinks this is relevant is that, in his opinion, whereas the German an sich can be used adnominally, the longer phrase an sich selbst cannot be so used and sounds incomplete because the adverb an sich selbst needs to be completed by a verbal phrase and not by a noun. He notes that only 13% of all occurrences altogether are of the shorter form, about half of which occur in contexts in which Kant also uses the longer form. That means that only 6% of all occurrences unambiguously display the linguistic form of the expression Ding an sich, which has standardly been used by German interpreters of Kant’s transcendental idealism from very early on as a term for a special class of entities (Prauss 1974: 16). Prauss concludes that this understanding is mistaken. He takes the result of his word count to prove that Ding an sich is a deviant abbreviation for the longer phrase Ding an sich selbst, which itself is only an abbreviation because it is in need of a verbal complement for an sich selbst.

Prauss’ second step consists in the claim that Kant’s standard complements of the adverb an sich selbst are the verbs betrachten and also ansehen or erwägen, which correspond to the English verbs ‘consider’ or ‘regard’. Hence, Kant’s use of the expressions Ding an sich and Ding an sich selbst should be understood as an abbreviation of the longer ‘Ding an sich selbst betrachtet’ (‘thing considered in itself’) (Prauss 1974: 20). Prauss notes that this form corresponds to the Latin ‘res in se (or: per se) spectata’, which Kant would have found in Baumgarten’s Metaphysics, for example (Prauss 1974: 20n11). Prauss concludes from this that Kant’s talk about things in themselves does not carry any ontological commitment with respect to a realm of reality beyond that of empirical objects. To consider objects in themselves and to consider them as appearances are just two different kinds of attitudes that we can take towards those empirical objects when we theorize about them in philosophical reflection (Prauss 1974: 30).

I think that both steps of Prauss’ argument are unconvincing. One notable aspect about the first step is that it does not translate into English, for both Ding an sich and Ding an sich selbst are translated by the same ‘thing in itself’ in English. This is not due to a lack of expressive power of the English language, for also in German, there is no substantial syntactical difference between Ding an sich and Ding an sich selbst. It is correct that an sich selbst on its own cannot serve as a modifier of a noun, but the same is true for an sich. Just like the expressions ‘for itself’ or ‘independently of us’, an sich needs a verbal complement and hence, unless taken in a technical sense as an abbreviation for a longer phrase, cannot be used as a modifier of a noun. It might be that in the case of Ding an sich selbst, the gap for a verbal complement is more obvious to the contemporary reader’s ear than in the case of Ding an sich. But that might just be due to the fact that Ding an sich has been standardly used in a technical philosophical sense as a noun phrase in the reception of Kant’s philosophy for more than 200 years. It does not show that an sich is not per se as little an adnominal expression as an sich selbst. So, although Prauss is perfectly right that the expression ‘thing in itself’ is to be regarded as an abbreviation for a longer form where the adverbial phrase ‘in itself’ is completed by a verbal complement, his whole impressive word count seems superfluous for showing this. ⁶

On the other hand, Prauss is much too quick when he jumps from the insight that ‘in itself’ per se needs a verbal complement to the assumption that it cannot be understood in an adnominal way when being used as an abbreviation for a longer expression where this verbal complement is added. It is true that given Prauss’ own choice for the verbal complement of ‘in itself’, the resulting longer expression ‘considered in itself’ is not an adnominal expression that serves to specify a certain subclass of objects. However, this does not follow from the fact alone that ‘in itself’ itself is adverbial. To see this, consider the following analogous case. The expression ‘on Monday’ is an adverbial expression and is so used in a sentence like ‘My students were late on Monday’. However, it can also be used as an abbreviation of an adnominal expression, for example, in, ‘My students on Monday are better than my students on Wednesday’. Here, ‘my students on Monday’ is an abbreviation for ‘the students that I teach on Monday’, and hence, ‘my students on Monday’ serves to restrict the class of my students to a certain subclass of them, namely, those that I teach on Monday. Analogously, if ‘things in themselves’ were an abbreviation not for ‘things considered in themselves’ but rather for ‘things that exist in themselves’ or for ‘things that have their properties in themselves’, the phrase could be understood as a noun phrase that denotes a certain class of objects. This shows that Prauss’ argument against the two-object interpretation cannot solely be based on the insight that ‘in itself’ has the syntactical form of an adverbial expression, which is the main idea of the first step of his argument, but that it also needs the support of the second step of the argument, that is, the claim that the standard complement for ‘in itself’ is ‘consider’.

Now, in stark contrast to the impressive word count that substantiates the priority of the use of an sich selbst over an sich in the first step of his argument, Prauss’ evidence for the claim in the second step is rather thin. It basically consists in a footnote that lists thirteen passages in which Kant says that we can consider things in different ways, namely (as things) in themselves and as appearances (Prauss 1974: 21n9). Only one of these passages, taken from the Metaphysical Foundations of Natural Science, entails the noun phrase ‘thing considered in itself’ which Prauss claims is the phrase standardly abbreviated by ‘thing in itself’ (MFNS, 4: 507). In any case, in order to show that the thirteen passages really represent Kant’s standard complement for the adverb ‘in itself’, it is certainly not enough to show that they exist. What would rather be required is a comparison of those passages with Kant’s overall adverbial use of ‘in itself’. ⁷ In fact, it would be plausible to apply a procedure similar to the one that Prauss himself used in the first step of his argument: one would have to investigate what other verbs Kant in fact uses as complements of the adverb ‘in itself’, divide them into groups of verbs with a similar meaning, and then count how many occurrences there are in each group. Prauss does nothing of this sort, and the reader simply has to take his word for the claim that ‘consider’ is Kant’s standard complement for the adverb ‘in itself’. If we do take the effort of a systematic investigation of Kant’s overall adverbial uses of ‘in itself’, it becomes apparent that Prauss’ claim is untenable. This is what I will show in the next section.

3. Kant’s actual verbal complements of ‘in itself’

For the following investigation, I have examined all explicitly adverbial uses of the phrase ‘in itself’ in the Kantian writings that are included in volumes 3, 4, and 5 of the Academy Edition, ⁸ as well as in On a Discovery in volume 8 and in the Progress essay in volume 20. To be a bit more precise, I have searched for occurrences of the expression an sich in the online version of these writings in the Bonner Kant-Korpus. ⁹ I have then identified those occurrences of an sich that are explicitly adverbial, that is, in which an sich is preceded or followed by a verbal phrase that is modified by it. That means that I have sorted out adnominal uses where no verbal complement fills the syntactic role of the complement of an sich or an sich selbst. For example, I would have sorted out the occurrence in the sentence ‘Dinge an sich sind unerkennbar’ (‘Things in themselves are incognizable’), where an sich modifies Dinge, whereas I would have counted in ‘Dinge sind an sich unerkennbar’ (‘Things are incognizable in themselves’), where an sich modifies sind unerkennbar, or ‘Wie die Dinge an sich selbst sind können wir nicht erkennen’ (‘We cannot cognize how things are in themselves’), where an sich selbst modifies sind. There are several constructions where both an adnominal and an adverbial reading of ‘in itself’ are syntactically possible. For example, the German sentence ‘Wir wissen nicht, wie Dinge an sich selbst beschaffen sind’ allows for two readings, which are translated into English as ‘We do not know how things in themselves are constituted’ (where ‘in themselves’ modifies ‘things’) and ‘We do not know how things are constituted in themselves’ (where ‘in themselves’ modifies ‘are constituted’). I have sorted out these syntactically ambiguous cases unless I had a clear feeling that the adverbial reading is the intended one. (Obviously, there is room for disagreement in these cases.) Last but not least, I have neglected adverbial uses of ‘in itself’ that are not clearly related to a formulation of transcendental idealism (e.g. when Kant says that something is necessary in itself, good in itself, or an end in itself) ¹⁰ and occurrences of an sich where it precedes a verb but is a lexical part of the verbal phrase rather than an adverbial modifier, as in the German ‘etwas Merkwürdiges an sich haben’ (‘to have something strange about them’). The intention behind all of this was to have as complete a collection of explicit verbal complements of uncontroversial adverbial uses of ‘in itself’ as possible and hence to be in a position to judge which verbal complements Kant actually uses and whether there is something like a standard use.

As a matter of fact, Kant uses a great variety of verbs as complements for ‘in itself’, and it would have made no sense to count occurrences of any one of them in particular. Instead, I have sorted the verbs into the following five groups:

1. 1. Verbs of predicative being or instantiation

2. Typical constructions are as follows: ‘x ist an sich F’ (‘x is F in itself’), ‘x ist an sich selbst so-und-so beschaffen’ (‘x is constituted in such-and-such way in itself’), ‘F-heit kommt x an sich selbst zu’ (‘x pertains to F in itself’), and ‘das-und-das Merkmal ist x an sich selbst eigen’ (‘such-and-such feature adheres to/is proper to x in itself’). ¹¹ Here, ‘in itself’ is used as a modifier of the copula or of other verbs that express the instantiation of a property or of verbs by which we characterize what a certain object is like. ¹² A passage that is paradigmatic for this kind of use is the following: ‘We have therefore wanted to say … that the things that we intuit are not in themselves what we intuit them to be, nor are their relations so constituted in themselves as they appear to us’ (A42/B59; emphasis mine).

3. 2. Verbs of epistemic access

4. Typical constructions are as follows: ‘x ist an sich selbst (un)bekannt/gegeben/etc’. (‘x is (un)known/given/etc. in itself’) and ‘A erkennt/kennt/ist bekannt mit/nimmt an/bestimmt/etc. x an sich selbst’ (‘A cognizes/knows/is acquainted with/assumes/determines/etc. x in itself’). A paradigmatic passage of this kind is as follows: ‘…I am not only warranted but compelled to realize this idea, i.e., to posit an actual object, but only as a something in general with which I am not acquainted in itself at all…’ (A677/B705; emphasis mine).

5. 3. Verbs of predication or semantic reference

6. Typical constructions are as follows: ‘x stellt y an sich selbst vor’ (‘x represents y in itself’), ‘x wird y an sich selbst zugeschrieben/beigezählt/etc’. (‘x is attributed to y in itself’), and ‘Vorstellungen (Urteile) beziehen sich auf/sind gültig (wahr) von/betreffen/etc. x in itself’ (‘representations (judgements) refer to/are valid (true) of/concern/etc. x in itself’). A paradigmatic passage for this use is as follows: ‘Time … cannot be attributed to objects in themselves (without their relation to our intuition) neither as subsisting nor as inhering…’ (A36/B52; emphasis mine).

7. 4. Verbs of existence

8. Typical constructions are as follows: ‘x existiert/subsistiert an sich selbst’ (‘x exists/subsists in itself’). A paradigmatic passage here is as follows: ‘Thus it is also false that the world (the sum total of all appearances) is a whole existing in itself’ (A506–7/B534–5; emphasis mine).

9. 5. Verbs of reflective consideration

10. Typical constructions are as follows: ‘x an sich selbst betrachten/ansehen/erwägen’ (‘to consider/regard x in itself’). A paradigmatic passage for this use is as follows: ‘…appearance always has two sides, one where the object is considered in itself … the other where the form of the intuition of this object is considered…’ (A38/B55; emphasis mine).

Before I present the result of my word count, let me stress that the procedure that I have applied when identifying the relevant occurrences and when sorting them in these five groups is obviously much more prone to error and to controversy, in particular cases, than Prauss’ more or less mechanical count. First, I may well have overlooked occurrences of an sich or have prematurely sorted them out as irrelevant. (In fact, given the enormous number of occurrences of an sich in the relevant volumes, it is rather likely that this has happened.) Second, I may have miscategorized occurrences, or at least I may have categorized them in a way with which other interpreters disagree. Again, given that we are talking about texts by Kant and given the syntactical and semantical intricacy of many of the passages in which an sich occurs, it is probably unavoidable that this is the case. However, I am not really worried about the fact that the boundaries of the following five groups of occurrences are consequently a bit blurred. The general picture that my count shows is so unambiguous that a small number of false positives or false negatives would actually make no difference.

In the following scheme, the occurrences of an sich are listed and counted for each of the five groups. ¹³ Following Prauss’ convention, each occurrence is represented by a five-digit number whose first three digits represent the page number of the volume in question and whose last two digits stand for the number of the line in which an sich occurs. I have also indicated in which Kantian work the passages occur and added the A and B pagination for passages from the First Critique. ¹⁴

Cases where an sich modifies a verb of predicative being or instantiation

(vol. 3) CpR ²: 05211(A23/B37) / 05533(A27/B43) / 05708(A30/B45) / 06031(A34/B51) / 06102(A35/B51) / 06120(A36/B52) / 06236(A37/B54n) / 06322(A39/B56) / 06511(A42/B59) / 06512(A42/B59) / 06517(A42/B59) / 06601(A43/B60) / 06606(A43/B60) / 06620(A44/B61) / 06621(A44/B61) / 06713(A45/B62) / 06729(A46/B63) / 06803(A46/B64) / 06903(A48B65) / 06906(A48/B65) / 06913(A48/B66) / 07003(B67) / 07116(B69) / 07123(B70) / 07130(B70n) / 12231(B156) / 12707(B164) / 16823(A190/B235) / 16902(A191/B236) / 19104(B274) / 22418(A277/B233) / 22436(A277/B233) / 23228(A291/B347) / 26831(B409) / 29911(A431/B459) / 33123(A478/B506) / 33922(A492/B520) / 34009(A493/B521) / 34013(A493/B521) / 34024(A493-4/B522) / 34619(A504/B532) / 34622(A504/B532) / 34704(A505/B533) / 34704(A505/B533) / 34713(A505/B533) / 34716(A505/B533) / 34722(A506/B534) / 35001(A510/B538) / 35208(A514/B542) / 35209(A514/B542) / 35313(A516/B544) / 35613(A521/B549) / 35926(A526/B554) / 35928(A526/B554) / 35932(A527/B555) / 36518(A536/B564) / 36610(A538/B566) / 36622(A538/B566) / 36720(A540/B568) / 37630(A556/B584) / 41808(A628/B656) / 44410(A672/B700) / 44515(A674/B702) / 44735(A678/B706) / 44814(A679/B707) / 44921(A681/B709) / 45034(A684/B712) / 45124(A685/B713) / 45810(A697/B725) / 45820(A698/B726) / 48516(A740/B768) / 48518(A740/B768) / 50815(A780/B808) // (vol. 4) CpR ¹: 08910(A120) / 16313(A249) / 16421(A251) / 16425(A252) / 16511(A253) / 22621(A359) / 22624(A359) / 22633(A360) / 22703(A360) / 23234(A370) / 23404(A372) / 23808(A379) / 23823(A380) / 24106(A385) Prol.: 28214 / 28216 / 28219 / 28305 / 28409 / 28514 / 28519 / 28621 / 28717 / 28904 / 28908 / 28927 / 28934 / 29137 / 29327 / 29411 / 29833 / 30825 / 31436 / 34123 / 34228 / 34235 / 35026 / 35128 / 35425 / 35505 / 35723 / 35919 / 36416 // GMM: 45104 / 45117 / 45123 / 45131 // MFNS: 48108 / 48419 / 50704 / 50720 / 50721 // (vol. 5) CprR: 10118 / 10126 / 13328 // CJ: 34509 / 35305 / 36508 / 45630 / 46236 / 46536 / 48223 // (vol. 8) Discovery: 21918 // (vol. 20) Progress: 26633 / 26810 / 26815 / 28825 / 29013 / 29630 / 30013 / 33516 / 34002.

Total number of occurrences: 145 (≈ 60% of total adverbial usage)

Cases where an sich modifies a verb of epistemic access

(vol. 3) CpR ²: 05628(B44) / 05709(A30/B45) / 06306(A38/B55) / 06933(B67) / 15205(B207) / 33937(A492/B521) / 34105(A494/B523) / 34106(A494/B523) / 34426(A500/B529) / 34923(A509/B537) / 34925(A509/B537) / 35205(A514/B542) / 35632(A521/B549n) / 38220(A566/B594) / 39222(A584/B612) / 41310(A619/B647) / 44507(A674/B702) / 44510(A674/B702) / 44526(A674/B702) / 44630(A676/B704) / 44705(A677/B705) / 44708(A677/B705) / 44713(A677/B705) / 44803(A679/B707) / 44809(A679/B707) / 44909(A681/B709) / 45016(A683/B711) / 45202(A686/B714) / 48308(A737/B765) / 48514(A740/B768) / 51534(A793/B821) / 51537(A793/B821) // (vol. 4) CpR ¹: 08603(A114) / 23215(A369) / 24118(A385) // Prol.: 28624 / 28828 / 29829 / 29905 / 31812 / 35935 // MFNS: 50632 / 50708 // (vol. 8) Discovery: 21921 / 22135.

Total number of occurrences: 45 (≈ 18.6% of total adverbial usage)

Cases where an sich modifies a verb of predication or semantic reference

(vol. 3) CpR ²: 06124(A36/B52) / 06709(A45/B62) / 07133(B70n) / 07134(B70n) / 13509(A139/B178) / 13513(A139/B178) / 20412(A238/B298) / 21316(A257/B313) / 26415(A344/B402) / 26910(B410) / 32602(A468/B496) / 33209(A479/B507) / 34118(A495/B523) / 35332(A517/B545) / 35517(A519/B547) // (vol. 4) CpR ¹: 25221(A405) // Prol.: 28727 / 28917 / 29426 / 29705 / 32824 / 33515 / 35712 / 35907 / 36136 // (vol. 8) Discovery: 21822.

Total number of occurrences: 26 (≈ 10.7% of total adverbial usage)

Cases where an sich modifies a verb of existence

(vol. 3) CpR ²: 06516(A42/B59) / 12703(B164) / 23136(A289/B346) / 29721(A429/B457n) / 33826(A491/B519) / 33902(A491/B519) / 33927(A492/B520) / 33930(A492/B520) / 34019(A493/B521) / 34702(A505/B533) / 34716(A505/B533) / 34725(A506/B534) / 34802(A506/B534) / 34806(A506/B535) / 44930(A682/B710) // (vol. 4) CpR ¹: 23223(A369) / 23829(A380) // Prol: 34125 // (vol. 5) CprR: 10210.

Total number of occurrences: 19 (≈ 7.9% of total adverbial usage)

Cases where an sich modifies a verb of reflective consideration

(vol. 3) CpR ²: 05613(A28/B44) / 06302(A38/B55) / 21224(A256/B312) / 25219(A324/B381) // (vol. 4) CpR ¹: 22610(A358) // Prol.: 31123 // MFNS: 50736

Total number of occurrences: 7 (≈ 2.9% of total adverbial usage)

This overview gives us a clear picture of how Kant in fact uses the adverbial modifier ‘in itself’. The first thing to note is that Prauss’ claim that the adverbial modifier an sich selbst (‘in itself’) takes verbs of philosophical reflection such as ‘consider’ as its standard complement has no basis in Kant’s actual use whatsoever. The number of cases where Kant explicitly does use such a complement is marginal. With seven occurrences altogether, this use is by far the least representative one. ¹⁵

On the other hand, the overview also provides us with a clear answer to the question of what kinds of verbal phrases ‘in itself’ do standardly modify. With 60% of the overall usage, it is the first group that contains verbal phrases by which we describe what things are like and what qualities or constitution they have, that is, the copula or other verbs that express the instantiation relation between things and their properties. The idea behind Kant’s use seems to be that things have certain of their qualities or features only with respect to, or in relation to, another thing, while they have others in themselves, that is, without any further qualification or relativization. This is a common way of speaking. For example, we sometimes say that things are not interesting in themselves but only interesting for people with certain aims or preferences, such that one and the same object could be interesting for me but not interesting for you. Hence, being interesting would count as a feature that does not pertain to objects in themselves. Or you might think that colours are secondary qualities and that an object is not red in itself but that it has this colour only for beings with a visual sensory system like ours. In fact, Kant himself uses this analogy in order to point out what he means when he says that objects are not spatio-temporal in themselves but ‘only as appearances’ or ‘only for us’, that is, for beings with space and time as their a priori forms of intuition (P, 4: 289 and CpR B: 69–70n). The use also corresponds to those cases in which Kant uses the phrase ‘in itself’ outside of his characterizations of transcendental idealism, for example, when he speaks about a good will that is not good for attaining something else but good ‘in itself’. ¹⁶

Taking ‘in itself’ to standardly modify verbal phrases that describe what things are like has the further advantage that it allows us to assume that there is a systematic connection between all its adverbial uses and hence to avoid the assumption of brute ambiguities. For there are plausible ways to translate the uses of the other groups into those of the first. Take the cases of the second group, where ‘in itself’ modifies a verb of cognitive access as in ‘x is unknown in itself’ or ‘we cannot cognize x in itself’. Instead of assuming that these constructions specify a particular mode of knowing or cognizing, we can understand ‘x is unknown in itself’ as an abbreviation for ‘it is unknown how x is constituted in itself’, and we can interpret ‘we cannot cognize x in itself’ in the sense of ‘we cannot cognize x as it is in itself’. In both cases, ‘in itself’ would modify a verb of predicative being in the more explicit formulation.

A similar approach seems plausible in the case of occurrences of the third group where ‘in itself’ modifies verbs of semantic reference. For ‘x represents y in itself’ can easily be understood as ‘x represents how y is in itself’ or as ‘x represents y as y is in itself’. Analogously, to say that a certain judgement or principle is true or valid of x in itself can be understood as saying that this judgement correctly describes how x is in itself. ¹⁷ Moreover, if there are certain features, such as being red or being extended in space, that do not pertain to objects in themselves, then, analogously, there should be a mode of attributing the respective predicates ‘red’ and ‘extended in space’ not to objects in themselves but to objects in relation to something else. Attributing colour concepts, or the concepts of space and time, to objects in themselves would then mean to apply the respective predicates without noticing that their correct attribution would need to mention a further relativization. ¹⁸

As for cases of the fourth group, in which ‘in itself’ modifies a verb of existence, it should be noted that a considerable number of them are also disguised cases of modifications of predicative being. In nine of the nineteen passages of the fourth group, Kant either says that certain things do not exist in themselves as Fs or with certain properties, ¹⁹ or he speaks about how things exist in themselves. ²⁰ Although these passages contain the word ‘exist’, they just seem variants of saying that things are not F in themselves, or have the property in question in themselves, and of speaking about how things are in themselves. Two further passages concern the question of whether the world is ‘a whole existing in itself’, ²¹ which again seems but a stylistic variant of the question of whether the world composes a whole in itself. And one passage contains the formulation an sich gegründete Existenz (‘existence grounded in itself’), in which an sich strictly speaking modifies gegründet (‘grounded’) rather than existieren (‘exist’). So, we are left with only seven occurrences where ‘in itself’ unambiguously modifies a verb of existence. We should recognize these cases as deviations from Kant’s standard use of ‘in itself’ as a modifier of verbs of predicative being. However, since those deviations constitute only such a minute part of Kant’s overall usage and since none of them appears in a context in which Kant introduces or explains the basic claims of transcendental idealism, we should not base our interpretation of what Kant means by ‘things in themselves’ on them.

Finally, there is also an obvious systematic connection between the cases in group 1 and the ‘Prauss cases’ in group 5. To consider an object in itself just means to consider it as having its properties in itself, and to say that certain features do not pertain to it when considered in itself means that these features are not among those that the object has independently of us.

The above investigation of Kant’s explicit adverbial use of ‘in itself’ is also helpful for those cases that are not explicitly adverbial. The first thing to note is that a substantial number of these cases are but stylistic variants of explicitly adverbial cases of group 1. Constructions of the form ‘A is a property/determination/character/condition/form/connection of x in itself/xs in themselves’ are systematically equivalent to phrases of the form ‘A is a property/determination/character/etc. that attaches/pertains/belongs to x in itself/xs in themselves’. To see this, think of the temporal adverb ‘tomorrow’ in the two sentences ‘I am worried that my children might be tired tomorrow’ and ‘I am worried about the condition of my children tomorrow’. It is clear that the phrase ‘my children tomorrow’ in the second sentence is not meant to signify a particular subclass of my children but that ‘tomorrow’ modifies the relation between my children and their condition that is expressed in the genitive construction. What I am worried about is the condition that my children might be in tomorrow.

There are a considerable number of cases that are implicitly adverbial in the sense just explained, namely:

(vol. 3) CpR ²: 05502(A26/B42) / 06530(A43/B60) / 06633(A44/B62) / 06804(A46/B64) / 07217(B71) / 09807(B114) / 20912(B306) / 21901(A267/B323) / 22404(A276/B332) / 22417(A276/B332) / 23629(A297/B353) / 33932(A492/B520) / 34022(A493/B522) / 36428(A535/B563) / 36705(A539/B567) / 36708(A539/B567) / 47223(A719/B747) // (vol. 4) Prol.: 28809 / 29429 / 31004 / 31108 / 33604 GMM: 46106 // MFNS: 48416 / 50619 / 50626 // (vol. 8) Discovery: 22001 / 24017 // (vol. 20) Progress: 26707/28035.

Total number of occurrences: 30 ²²

Now, there are also a large number of cases that cannot be transformed into explicitly adverbial ones in this manner. Here are three typical examples:

Appearances are not things in themselves. (A165/B206)

If one regards the two propositions, ‘The world is infinite in magnitude’, ‘The world is finite in magnitude’, as contradictory opposites, then one assumes that the world (the whole series of appearances) is a thing in itself. (A504/B532)

…if appearances are things in themselves, then freedom cannot be saved. (A536/B 564)

In these and other analogous cases, the phrases ‘in itself’ and ‘in themselves’ are used as adnominal modifiers of the noun ‘thing(s)’, and the phrase ‘things in themselves’ signifies a certain kind of objects. We have already seen that since ‘in itself’ and ‘in themselves’ per se belong to the category of adverbial expressions, we should understand these cases as a technical philosophical use and take them to be abbreviations of phrases where a verbal phrase is explicitly used. Given the results of our investigation, it is most plausible to assume that the verbs that are left out in the abbreviation are verbs that expresses predicative being. Hence, I propose to understand the terms ‘thing in itself’ and ‘things in themselves’ as standard abbreviations of the following form:

‘thing in itself’ ≈ ‘thing that has its properties in itself’

‘things in themselves’ ≈ ‘things that have their properties in themselves’

The term ‘property’ in these formulations is not meant in any technical philosophical sense or to carry any ontological burden. Kant often uses the term Beschaffenheit(en) when he speaks about what things are like in a nominalized form. ²³

According to this proposal, the first quote above says that appearances are not things that have their properties in themselves. The second quote says that when we take the world to be necessarily either finite or infinite, then we presuppose that the world is a thing that has its properties, for example, its magnitude, in itself. And the third quote says that if appearances are things that have their properties in themselves, then none of those appearances – also not human beings – would be free.

Let us next see what consequences the above investigation of the linguistic form in which Kant expresses transcendental idealism has for the three main variants of its interpretation that were mentioned at the beginning of this article.

4. Consequences for the debate about Kant’s transcendental idealism

The methodological two-aspect interpretation has often been criticized for not being able to explicate what it means to ‘consider’ an empirical object as an appearance but not as a thing in itself in a way that allows us to let the appearance/in-itself distinction do all the theoretical work that Kant assigns to it. For example, it seems unclear how one and the same thing could be causally determined and spatial when considered in the first sense but transcendentally free and non-spatial when considered in the second (cf. Guyer 1987: 336–42; Van Cleve 1999: 6–8; Allais 2015: ch. 4). We have now seen that the methodological interpretation can also not be substantiated by what has always seemed to be one of its main selling points, namely, Kant’s own adverbial use of the phrase ‘in itself’. In fact, if we read through all those many passages in which Kant does not use ‘in itself’ as a modifier of a verb of philosophical consideration, it is really astonishing how methodological two-aspect interpretations could ever have become so influential and popular as they have been for such a long time. The claim that ‘thing in itself’ is Kant’s standard abbreviation for ‘thing considered in itself’ is nothing but an interpretational myth. This is not to say that it is impossible to assign to the idea of considering an object in different ways a role in an explication of Kant’s transcendental idealism. However, whoever wants to give this idea a new shot would not only have to do so in a philosophically more convincing way than has been done by previous adherents of this idea, but they would also have to take into account that Kant’s standard use is one in which ‘in itself’ modifies a verb of predicative being.

How do the two alternatives to the methodological two-aspect interpretation, the ontological two-object and two-aspect interpretations, fare in light of our textual investigation? Let us begin with what we have found out about cases in which ‘in itself’ or ‘in themselves’ is used adnominally. We have seen that these expressions are per se adverbial and need a complement, which is hidden in the adnominal use. Two-object interpreters who acknowledge this usually interpret ‘things in themselves’ as short for ‘things that exist in themselves’ (cf. Aquila 1979: 297; Aquila 1983: 89; Jauernig 2021: 29). Given this explication, they can interpret the claim ‘Appearances are not things in themselves’ (A165/B206) above as clear evidence for their two-object view: if appearances are not things that exist in themselves, they seem to have no being beyond their being perceived by us. However, we have seen above that it is textually implausible to take ‘things in themselves’ to be an abbreviation for ‘things that exist in themselves’ and that we should rather read it in the sense of ‘things that have their properties in themselves’. Now, given this explication, the quote no longer indisputably supports the two-object interpretation. It only does so if we interpret the phrase in the sense of ‘things that have some of their properties in themselves’, for then the claim says that appearances do not have any of their properties independently of us. However, if we read the phrase in the sense of ‘things that have all of their properties in themselves’ (which in my ears is linguistically more natural), then the two-aspect interpreter can easily agree that appearances are not things in themselves, for it is precisely her view that appearances have their spatio-temporal properties only in relation to us and hence not all of their properties in themselves.

In fact, it is very natural to read a claim such as ‘appearances are not things that have their properties in themselves’ as being implicitly restricted to those properties that appearances have qua being appearances. We see this if we look to analogous cases. The sentence ‘Politicians are entities whose properties are socially constituted’ is most naturally understood as saying that the properties that politicians have qua politicians are socially constituted. It is not meant to exclude that those beings who are in fact politicians, that is, human being, also have other properties (e.g. biological ones) that are independent of any social relations. So, it is not contradictory to say that, as politicians, members of parliaments are beings that owe their properties to social facts, whereas, as biological organisms, they are beings whose properties are independent of such social facts. Analogously, the claim that appearances are things that do not have their properties in themselves is naturally understood as meaning that appearances do not have those properties in themselves that they have qua being appearances, for example, their spatio-temporal extension. In fact, Kant himself often uses ‘qua’ or ‘as’-constructions to make clear that one and the same objects can have some of their properties only for us while they have others in themselves. ²⁴ But the claim that appearances do not have their appearance-properties in themselves is not controversial between the two-object and the two-aspect interpreter. In order to settle the question of whether appearances also have further properties in themselves or not, both parties will need additional textual evidence beyond claims of the form ‘appearances are not things in themselves’.

Let us now look at the explicitly adverbial use of the phrases ‘in itself’ and ‘in themselves’, which has been the main focus of our inquiry here. Ontological two-aspect interpretations can take the exegetical evidence above as grist to their mills. If we take Kant’s transcendental idealism to be based on a distinction between the way things are in themselves and the way they appear to us or between properties that empirical objects have independently of us and properties that they have for us, then we should expect Kant to use ‘in itself’ standardly as an adverbial modifier of verbs of predicative being or instantiation, just as the above investigation has concluded. In fact, if one takes the effort to click through all the passages of group 1 above, one will find it hard to believe that Kant could ever have wanted to say anything in these passages but what the ontological two-aspect interpretation ascribes to him, because the idea that there are two distinct classes of objects – things in themselves and appearances – is entirely absent from most of them. ²⁵

How can two-object interpretations deal with all the numerous explicitly adverbial uses of ‘in itself’ that we have found? Traditionally, two-object interpreters acknowledge passages as potential problems for their interpretation in which Kant explicitly says that one and the same objects appear to us and are also somehow constituted in themselves (cf. Bxxvi–xxviii, B 69, A38/B55, B156, A360). However, in light of many other things that Kant says, and which they take to be compatible only with their own interpretation, they either try to explain away those passages as a kind of loose talk (cf. Van Cleve 1999: 144–5 and Jauernig 2021: 18), or they point out that some of these passages concern the case of human thinkers and agents for whom there might be special philosophical reasons to accept that there is one common entity that has both appearance- and in-itself properties (cf. Jauernig 2021: 18n54). Given this response, it is noteworthy that the above textual investigation poses a new potential threat to two-object interpretations. For if Kant’s standard use of the phrase ‘in itself’ is one in which it is a modifier of verbs that express what things are like and not one that identifies a certain class of entities, then any interpretation must give a convincing reading of this use.

So, let us look at the passage that we have taken to be paradigmatic of the use of ‘in itself’ as a modifier of verbs of predicative being:

We have therefore wanted to say … that the things that we intuit are not in themselves what we intuit them to be, nor are their relations so constituted in themselves as they appear to us. (A42/B59)

The two-object interpreter has to decide which of the two kinds of objects her interpretation assumes Kant is talking about here. Her things in themselves seem not to be possible candidates, for they are not objects that we intuit. Her merely intentional objects, on the other hand, have no being in themselves distinct from the way we represent them, whereas the passage at least strongly suggests that there is a way that the objects are, which differs from the way they appear to us.

There are many passages in group 1 above that are similar in this respect: they speak about how certain things are in themselves, but they do not leave us any exegetical room for taking these things to be entities that are distinct from the empirical objects that we encounter in experience. Here are a few examples (the underlining is mine):

This reality of space and time leaves the certainty of experiential cognition untouched: for we are just as certain of that whether these forms necessarily adhere to the things in themselves or only to our intuition of these things. (A39/B56)

Through mere relations no thing is cognized in itself: it is therefore right to judge that since nothing is given to us through outer sense except mere representations of relations, outer sense can also contain in its representation only the relation of an object to the subject, and not the inner, which pertains to the object in itself. (B67)

It is only by means of the form of sensory intuition that we can intuit things a priori, though by this means we can cognize objects only as they appear to us (to our senses), not as they may be in themselves… (P, 4: 283)

I grant by all means that there are bodies without us, that is, things which, though quite unknown as to what they are in themselves, we yet know through the representations which their influence on our sensibility procures in us… (P, 4: 289)

All these passages speak about how certain things are in themselves. These things cannot be taken to form any separate class of ‘things in themselves’, because the syntactical structure of the sentences implies that the underlined noun phrases in each quote refer to the same kind of entities. The class of objects that are somehow in themselves includes objects of which we have an intuition (quote 1), objects whose relation to human subjects is represented in the outer sense (quote 2), objects that appear to us (quote 3), and bodies (quote 4). Now, it also does not seem to be an option for the two-object interpreter to take the things Kant is talking about here to be what she calls purely intentional objects, for purely intentional objects do not have a way they are in themselves which could be unknown to us, and they also certainly have no ‘influence on our sensibility’ (quote 4). ²⁶

I think it is important to point to these cases of an adverbial use of ‘in itself’/‘in themselves’ and to the potential problems they can raise for a two-object interpretation. Although they are not among the passages in which Kant explicitly identifies the things that appear to us and the things that are somehow in themselves (cf. Bxxvi–xxviii, B69), they carry the identity assumption as an implicit prerequisite of the use of anaphoric pronouns. Whereas cases of the former kind are accepted by two-object interpreters as potentially problematic for their view (cf. Van Cleve 1999: 144–5 and Jauernig 2021: 18), cases of the latter kind are given much less attention although they do not seem less threatening.

Let me stress that I do not want to suggest that considerations about the linguistic form of the mentioned passages and similar ones will decide the fate of two-object interpretations all by themselves. It is just one piece of textual evidence that needs to be weighed against other criteria of exegetical adequacy, for example, the extent to which different interpretations can explain how the appearance/in-itself distinction can do the kinds of philosophical work that Kant wants it to do (e.g. provide a resolution to the antinomies). And it also needs to be evaluated in light of further textual evidence. This evidence includes, among other things, Kant’s ubiquitous claims that appearances are mere representations, which are arguably much easier to integrate into two-object interpretations than into two-aspect interpretations and whose understanding certainly needs other interpretational devices beyond the method of word counting. Ideally, an interpretation of Kant’s transcendental idealism should be able to accommodate all the textual evidence and leave room for both two-aspect and two-object talk. ²⁷

However, I still think that the results of the above investigation are important and that the linguistic form of the sentences by which Kant expresses his doctrine of transcendental idealism deserves more attention than it often gets in the debate about the doctrine itself. Some of the central questions in this debate appear in a new light once we remind ourselves that the adverbial formulations of the appearance/in-itself distinction are the more fundamental ones. For example, some interpreters have found it impossible to believe that Kant could have accepted the consequence of the two-aspect interpretation ‘that every empirical object corresponds to exactly one thing in itself’ and that we hence have ‘knowledge of the number of things in themselves’ (cf. Van Cleve 1999: 160–1; Walker 2010: 824; Marshall 2013: 524–5; Stang 2014: 124–31). Formulated in this way, these consequences indeed sound odd. The formulations suggest that there are two classes of objects – the class of empirical objects and the class of things in themselves, of which we can identify the elements in each of these classes separately and of which we can then recognize a one-to-one correspondence between them (just as we can identify the natural numbers and the even numbers and then see that there are as many of the first as there are of the second). It seems indeed absurd to assume that such a procedure could be compatible with the Kantian restriction of human cognition to the realm of appearances. However, once we reformulate the implications of the two-aspect interpretation in adverbial form, they seem far less outrageous, for then, all that the two-aspect interpretation forces us to accept is that every empirical object is also somehow constituted in itself – and hence, trivially, is identical to an object that is somehow constituted in itself – and that it must be so constituted in order to appear to us by affecting us. ²⁸ This claim might be problematic for various reasons. But it certainly does not justify any incredulous stares, ²⁹ especially since it corresponds more or less exactly to things Kant himself says, for example, when he claims that ‘the word “appearance” must already indicate a relation to something the immediate representation of which is, to be sure sensible, but which in itself, without this constitution of our sensibility … must be something, i.e., an object independent of sensibility’ (A251 f.). ³⁰

I think that it is a warning of distorting effects of nominalizations in philosophical theories and their interpretations like the one just mentioned that we should take to be the deeper lesson of Prauss’ seminal study. Prauss was absolutely right to point out that since the earliest days of the reception of Kant’s transcendental idealism, interpreters have used the term ‘things in themselves’ as denoting a class of entities without feeling the slightest need to add a further explanation how this adnominal use can be reconciled with the adverbial character of ‘in itself’ (Prauss 1974: 24–30). And he deserves eternal glory for reminding us that this attitude is unacceptable, especially since it is still predominant in so many publications on Kant’s transcendental idealism today.