Modelling the mind: Nietzsche’s epistemic ends in his account of drive interaction

ABSTRACT

Nietzsche offers us an account of how different drives interact with one another; it is rich but also appears to risk the homunculus fallacy. Competing attempts to deflect this charge on his behalf share an implicit consensus about the ‘epistemic ends’ of the account: they assume Nietzsche is trying to provide true explanations of psychological phenomena. I argue against this consensus. I claim that Nietzsche's characterisations of drive interaction are to be taken as fictive and are not intended to have explanatory value. They nevertheless facilitate genuine epistemic achievement. Drawing on Catherine Elgin's account of the epistemic role of idealised models in science, I argue that Nietzsche's account of drive interaction is a ‘model of the mind’ that, despite relying on falsehoods, can exemplify features of our psychology that aid us in making novel predictions. We then see that Nietzsche neatly sidesteps the homunculus fallacy; we can further understand more fully what Nietzsche hopes his drive psychology will teach us. We can now resolve, for example, outstanding interpretative puzzles about the relationship between psychic integration and Nietzsche’s distinctive notion of spiritual health.

KEYWORDS:

• Nietzsche
• drives
• health
• homunculus
• explanation

1. Introduction

Nietzsche considered himself a psychologist – indeed, one ‘without equal’ (EH Books 5).¹ His philosophical psychology centres on the drive, loosely a kind of behavioural disposition (Riccardi 2021); Nietzsche’s account of the drives gets especially interesting, but also obscure, when he attempts to characterise how they interact with one another, and it’s on this that I will focus. Interpreters have recently emphasised how Nietzsche uses his account of drive interaction to help us understand a variety of phenomena, including spiritual health, value judgements and agency (as examples see Clark and Dudrick 2012; Riccardi 2021; Richardson 2009). The rich socio-political language that Nietzsche tends to use to characterise how the drives interact, however, has led to worries that Nietzsche commits himself to a fallacious homuncularism. This has received much attention, with Clark and Dudrick (2012) and Riccardi (2021) offering two notable competing readings which aim to solve the problem. While there’s much disagreement between them, I’m interested in an implicit background consensus about what I’ll term the ‘epistemic ends’ of Nietzsche’s account that emerges in this discussion:

CONSENSUS VIEW: In providing his account of drive interaction, Nietzsche aims at a true representation of how the drives work, and in so doing to provide explanations of various psychological phenomena.

In this paper, I argue against this view. I will claim that Nietzsche’s account of drive interaction is not aimed at representing them truly, but instead should be taken as fictive. Following Suárez (2008), I say ‘fictive’ – rather than ‘fictional’ – to distinguish myself from a view on which Nietzsche would be read as denying that drives as entities exist. Nietzsche fairly clearly thinks drives are real features of human psychology: I merely argue that many of the claims he makes about how these entities interact aren’t intended to be taken as true. We might suppose that a fictive account won’t be able to teach us much, but that isn’t so: despite being false, these fictive representations can still facilitate genuine epistemic achievement. To understand how, however, it helps to acknowledge that Nietzsche is not aiming at providing genuine explanations of psychological phenomena; Nietzsche instead seeks to help us in the descriptive project of seeing patterns in the phenomena that we otherwise might’ve missed, thereby facilitating novel predictions. To illustrate how, I’ll draw on Catherine Elgin’s (2017) work on the epistemic role of idealised models in science. I’ll argue Nietzsche’s socio-political account of drive interaction should be taken as an idealised model of the mind which aims at exemplifying features of its target to facilitate prediction, rather than representing it accurately. I take all this to be important, because getting clear on Nietzsche’s epistemic ends will help us then clarify the concrete things Nietzsche hopes his psychological work will teach us, and how it does so without resorting to fallacy.

In section 2, I will start by drawing out two key assumptions in the secondary literature that underpin the consensus view. In section 3, I present textual evidence to suggest Nietzsche rejects these assumptions. In section 4, drawing on Elgin, I flesh out how Nietzsche’s account of drive interaction might still facilitate epistemic achievement despite being false. In section 5, I illustrate how rejecting the consensus view allows Nietzsche to sidestep any worries about homuncularism, while also enabling us to appreciate more fully the subtlety and power of his account of human psychology. Rejecting the consensus view can help us clarify, for instance, Nietzsche’s distinctive notion of spiritual health.

2. Two key assumptions underpinning the consensus view

I’ll start by clarifying Nietzsche’s notion of a drive. I’ll follow ‘most scholars [who] agree that at bottom [drives] are behavioural dispositions towards specific patterns of goal-directed behaviour’ (Riccardi 2018, 22–23; see also Riccardi 2021, 22): a simple example would be a sex drive which disposes agents towards various forms of sexual activity. Per Nietzsche, drives shape our actions (D 129; GS 335) and certain introspectively accessible mental phenomena like thoughts and feelings (D 119; GS 333).

I’m interested here in Nietzsche’s account of how drives interact. While this account features prominently in his work, it risks appearing particularly philosophically confused. The concern is that Nietzsche appears to treat the drives as human-like entities or ‘homunculi’. He frequently characterises how drives interact in ways that make them seem like agents in socio-political structures: he talks, for instance, of how a drive craves to be ‘master of all the other drives’ (BGE 6), of the soul as a ‘society constructed out of drives and affects’ (BGE 12), and the structure of Caesar’s drives being ‘five paces away from tyranny’ in a manner analogous to the political organisation of Ancient Rome (TI Skirmishes 38). What is often termed the homunculus fallacy can be broken out into two distinct problems that have been emphasised in the literature (Clark and Dudrick 2012, 197; Riccardi 2021, 49).² First, there is a risk of explanatory regress: in attempting to explain person-level capacities, subpersonal entities are posited which already possess and employ the capacity to be explained, rendering the explanation idle. Second, positing such homunculi can lead to a form of mereological fallacy where we ascribe capacities to parts that could only really belong to the whole, leading to a kind of category mistake: parts of persons just aren’t the kinds of things that can, say, ‘crave mastery’.

The homunculus fallacy is a familiar one in the history of philosophy and psychology. Dennett (1978) offers a notable solution to it in psychology for certain kinds of functional analysis: it is to this solution that Riccardi and Clark and Dudrick turn to deal with the charge of fallacious homuncularism levied against Nietzsche’s account of drive interaction, although they deploy it in different ways. Dennett’s solution to the mereological fallacy hinges on the very method of functional analysis: by treating psychological capacities as ‘complex systems’, he attempts 'to explain how a system works by understanding the functional contributions of its components’, thereby facilitating the ascription of psychological predicates to parts of persons (Drayson 2012, 4). Dennett then critically shows how functional analysis can avoid explanatory regress too: he notes that homunculi ‘are bogeymen only if they duplicate entire the talents they are rung in to explain’ (1978, 123; cited by Clark and Dudrick 2012, 198); explanatory regress can be avoided if the homunculi are ‘less clever’ (Dennett 1978, 80; cited by Riccardi 2021, 59) than the capacity to be explained.

Turning to Nietzsche, Clark and Dudrick claim that the drives are just such ‘less clever’ homunculi: while drives are characterised as occupying a political hierarchy, Clark and Dudrick argue that this need not imply that they have full person-level capacities, noting that ‘certain lower animals’ are taken to ‘act politically’ too (Clark and Dudrick 2012, 199). Having thus defended a sense in which drives could have genuine authority over one another, Clark and Dudrick go on to offer a normative reading of what it is for drives to have an ‘order of rank … [in which they] stand with respect to each other’ (BGE 6). Riccardi, in contrast, aims to explain how the capacity to engage in political relations can be seen as a relatively simple one by pointing to a Humean deflationary account of authority, arguing that understanding such relations as purely causal dispositions allows Nietzsche to avoid the homunculus fallacy.

My aim here is not to re-litigate the debate between these readings, however. Rather, I’m interested in the shared background assumptions about Nietzsche’s epistemic ends that they both take for granted – what I’ve termed the ‘consensus view’ of these ends. I’ll focus here on breaking out and clarifying the two assumptions that underpin this view. The first concerns explanation: both interpreters take it to be clear that Nietzsche’s account of drive interaction is involved in a project of explanation, and in particular what Dennett terms ‘subpersonal’ or ‘top-down’ explanation: that is, analysing ‘the highest levels of psychological organisation … into more and more detailed smaller systems or processes’ (1978, 110; cited in Clark and Dudrick 2012, 196).³ This gives us the first assumption underpinning the consensus view:

EXPLANATION ASSUMPTION: Nietzsche’s account of drive interaction is principally aimed at explaining various psychological phenomena.

Defenders of the consensus view highlight a range of psychological capacities which we might take Nietzsche’s account of drive interaction to explain: they suggest, for instance, that Nietzsche intends to explain our capacity to ‘act on values’ (Clark and Dudrick 2012, 199) or possess evaluative attitudes (Riccardi 2021, 176–177).

The second assumption underpinning the consensus view is then as follows:

VERACITY ASSUMPTION: Nietzsche’s characterisations of how the drives interact are intended to be taken as true (or at least approximately true, with perhaps modest paraphrase).

Clark and Dudrick make this particularly explicit: they emphasise that Nietzsche shouldn’t just be taken to be saying that drives ‘can be interpreted as political agents’ but that they ‘really are’ so (2012, 199). Riccardi follows Clark and Dudrick here, affirming that his dispositional construal of drives genuinely exercising authority should be taken to be ‘true … of the way in which [the drives] interact and are hierarchically arranged’ (2021, 60). I further suspect the veracity assumption is bound up for these commentators with the explanation assumption: it is commonly held that explanations can only succeed if they are true (De Regt 2015).

We’ve thus delineated the two key assumptions of the consensus view. I will now argue, however, that there are good textual reasons to suppose that Nietzsche would oppose both of the above assumptions, and thus the consensus view. Nietzsche, I suggest, intends much of his drive talk to be taken as fictive. He nevertheless hopes it can facilitate genuine epistemic achievement: to see how, it helps to understand that Nietzsche isn’t trying to facilitate explanation in the first place, but instead aims to help us identify patterns and regularities which might facilitate prediction.

3. Nietzsche against the consensus view

3.1. Nietzsche against the veracity assumption

Let’s start with the veracity assumption. It might sound odd to suggest that Nietzsche rejects this: defenders of the veracity assumption will note that he certainly tends to speak as if his claims about the drives are true. This isn’t decisive, however: fictions are typically presented as if they are true too. Nietzsche does, furthermore, appear several times to imply that much of his psychological analysis relies on a distinctive form of falsification and idealisation. In BGE 24, for instance, Nietzsche claims that we now realise ‘how it is precisely the best science [die beste Wissenschaft] that will best know how to keep us in this simplified … well-invented, well-falsified world’. Now Wissenschaft has a notably broader sense than the English ‘science’: metaphysics, for example, would also plausibly fall within its remit. But I think when Nietzsche talks of ‘the best science’ here he most likely has his own psychological work in mind as at least one example. Nietzsche concludes the immediately preceding passage, BGE 23, by insisting that ‘psychology again be recognised as queen of the sciences [Herrin der Wissenschaften], and that the rest of the sciences exist to serve and prepare for it’. Indeed, Nietzsche here appears to be directly challenging Kant’s claim that metaphysics be considered queen of the sciences, seemingly relegating metaphysics to the level of the ‘rest of the sciences’ that simply exist to ‘serve and prepare’ for psychology. This all suggests that Nietzsche specifically has psychology in mind in BGE 24 when he talks of the best science; given, furthermore, that Nietzsche takes himself to be a ‘psychologist without equal’ (EH Books 5), his psychological work is presumably an exemplar of this kind of the very best science, and that implies that it ‘best knows how’ to keep us in a ‘well-invented, well-falsified world’.

BGE 24 is, however, sometimes read as a more straightforward statement of the ‘falsification thesis’ often attributed to Nietzsche – the view that all beliefs are false, at least in some sense (Clark 1990, 103). To avoid the various philosophical issues that such a thesis presents, we might presume that Nietzsche still takes many of his core claims about the world to be true in a distinct, thinner sense,⁴ and so BGE 24 might not be taken as an issue for the veracity assumption after all: his psychological work might still aim at this thinner notion of truth. BGE 24 certainly starts by appearing to suggest that all human cognition involves pervasive falsification of some sort, exclaiming: ‘[w]hat a strange simplification and falsification people live in!’. Science, Nietzsche then says, ‘could arise only on this solidified, granite foundation of ignorance’: Nietzsche appears to accept that the pervasive falsification found in cognition extends to science. Importantly, however, that isn’t where the story ends: Nietzsche explicitly denies that science is engaged in a struggle against this falsification, claiming that science is not the ‘opposite’ of the foundational ‘will to not know’ but instead its ‘refinement’. Talk of refinement is particularly interesting here: Nietzsche implies science deploys a distinctive, ‘refined’ form of falsification to achieve its epistemic ends. This is hammered home when Nietzsche suggests that ‘precisely the best science … will best know [my emphasis] how to keep us in this simplified … well-invented, well-falsified world’. Nietzsche’s claim that we should focus on how ‘precisely’ the best science ‘best keeps us [my emphasis]’ in this falsified world again points to a distinctive form of falsification in the best science which goes beyond the thoroughgoing falsification already found in cognition. My suggestion is that Nietzsche here points to the familiar role of idealisation in science: that is, the role that idealised representations, which are transparently false even on a thinner notion of truth which Nietzsche would otherwise countenance, can play in epistemic achievement.

TI Skirmishes 7–8 reinforce such a reading of BGE 24, offering an explicit discussion of methodology in psychology. Starting with TI Skirmishes 7, this aphorism in part consists of a polemic against the ‘anti-artistic, factual’ type of psychologist which Nietzsche sees exemplified in the Goncourts. But in offering this attack, Nietzsche also sets up a positive vision for what he thinks psychology should be; the passage is titled ‘Moral for Psychologists’, and Nietzsche here details a second type of psychologist which he contrasts to that anti-artistic, factual type: this is the type of the ‘born psychologist’ – ‘an artist who is whole and complete’. Given Nietzsche’s own frequent boasts about his psychological acumen (EH Books 5; EH Destiny 6), he must consider himself a ‘born psychologist’ if anyone is one. Investigating the sense in which a born psychologist is an ‘artist’ should thus help us better understand Nietzsche’s psychological method.

The immediately subsequent aphorism, TI Skirmishes 8, provides helpful context here.⁵ This passage emphasises the importance of ‘idealising’ for ‘art or any sort of aesthetic action or vision’: this process ‘does not consist in removing or weeding out things that are small and incidental. Much more decisive is an enormous drive to force out the main features so that everything else disappears in the process’. This points, then, to the fact that an artistic psychology at its best does not just rely on a ‘light’ kind of falsification which abstracts and simplifies the phenomena, but a much deeper kind which even drives out ‘the main features so that everything else disappears in the process’, fitting into the idea of a more thoroughgoing form of idealisation that I outlined above. This, then, offers us a chance to make sense of how, in BGE 24, science, particularly at its best, is a ‘refinement’ of the pre-existing ‘will to not know’ that Nietzsche sees as the basis for any kind of cognition.

Notably, furthermore, Nietzsche is not in TI Skirmishes 7 denying that there are facts altogether, and thus a thinner sense in which science can get to the truth: he does not deny the existence of the ‘petit faits’ of the ‘anti-artistic, factual psychologist’. Rather, he is making the point that a scientific method which limits itself only to these is unnecessarily encumbered, and a more thoroughgoing form of falsification than that which might already pervade cognition is central to some of our most valuable scientific practices.

Now this all suggests that Nietzsche makes substantial use of idealisation in his psychological work; given that his account of the drives is central to his psychological analyses, we should expect to see idealisation deployed in this account.⁶ It is a further challenge pinning down which parts of his drive psychology he takes to be idealised; the claim I want to make here is that, at the very least, the problematic agential characterisations of drive interaction are likely cases of the idealisation Nietzsche refers to – if anything in his drive psychology is idealised and not aimed at accurate representation, these detailed portrayals of competing agents engaged in complex social interactions within a single mind seem like the most plausible candidates. While Nietzsche isn’t fully explicit, he further appears to hint that some of his most famous characterisations of drive interaction aren’t aimed at true representation: the extensively discussed characterisation of the soul as a ‘society constructed out of drives and affects’ in BGE 12 is in fact offered tentatively only as one of several ‘versions’ of a revised ‘soul hypothesis’; the ‘new psychologist’ positing such an idea, Nietzsche then further emphasises, is ‘condemned to invention’. Talk of ‘invention’ implies that the characterisation is not aimed primarily at representing the world accurately: when we invent a claim (as opposed to, say, speculating that a claim is true) we are typically not aiming at truth. Nietzsche indeed talks of ‘invention’ [erfinden/Erfindung] elsewhere early in BGE to explicitly highlight the falsity of claims: he talks, for instance, of ‘Plato’s invention of pure spirit and the Good in itself’ as a ‘dogmatist’s error’ (BGE P; see also BGE 11).⁷

I do not intend, though, to suggest that idealisation goes all the way down: in particular, I do not think that Nietzsche’s claims that humans have drives, that is various ‘behavioural dispositions towards specific patterns of goal-directed behaviour’ (Riccardi 2018, 22), are to be taken as false. Even if Nietzsche thought that we might struggle to identify the drives with precision, I do not think he would deny that drives are genuine features of human psychology: I expect he would find it uncontroversial that humans are disposed to at least some patterns of goal-directed behaviour, such as seeking out food and water. That means that if we attribute to Nietzsche a view on which the self is, say, identical with a drive or set of drives (Reginster 2003), or at least exists in a relationship of mutual dependence with them (Anderson 2012), Nietzsche can still hold selves to exist.

Importantly, furthermore, it appears Nietzsche still takes the fictive talk in his drive psychology to facilitate a kind of epistemic achievement: as we’ve seen, he is not modest about his success as a psychologist. Again, turning to BGE 12 is helpful here: Nietzsche emphasises that the ‘new psychologist’, in finding themselves condemned to ‘invention’ (i.e. fiction), might nevertheless in so doing find themselves condemned to ‘discovery’ (i.e. epistemic achievement). I’ll now turn to spelling out the kind of epistemic achievement I take Nietzsche to have been aiming at.

3.2. Nietzsche against the explanation assumption

Explanation is certainly one potential epistemic end of science; it is not the only one, however.⁸ Another, distinct epistemic goal that philosophers of science often emphasise is ‘knowledge that’ as opposed to ‘knowledge why’: science is interested in the project of identifying and describing the regularities found in nature, particularly for purposes of prediction and control, alongside the project of understanding why things happen the way they do.⁹

The two aims are not entirely unrelated: identifying regularities in nature also sometimes helps us explain the phenomena we observe, and good explanations may also help us make predictions. But it is important to distinguish them because the goals can come apart: a barometric reading may reliably predict the advent of a storm, but identifying this regularity doesn’t allow us to explain the storm (Salmon 1989, 47).¹⁰ The two goals can even be in tension: accurate predictions, particularly of complex systems, are often facilitated by tools poorly suited to explanation. In climate modelling, for instance, scientists might use a range of complex computational models in concert to reach a final prediction, but ‘none of those models are expected to accurately represent the causal influences on climate, nor to explain climate change … [indeed] accurate representation of specific causal influence is impossible, for the models represent influences in multiple, incompatible ways’ (Potochnik 2015, 76).

As we’ve noted, defenders of the consensus view take Nietzsche’s account of drive interaction to be involved in a project of explanation.¹¹ I think they are mistaken here: in particular, I take Nietzsche’s aims for his account of drive interaction to be more straightforwardly ‘descriptive’ rather than explanatory, concerned principally with identifying regularities in nature to thereby facilitate prediction and control. Nietzsche’s comments on the proper aims of science certainly seem to point in this direction: we see him express outright scepticism as to whether science can satisfactorily explain things at all. Consider the following:

We call it ‘explanation’ [Erkärung], but ‘description’ [Beschreibung] is what distinguishes us from older stages of knowledge and science. We are better at describing – we explain just as little as all our predecessors … The series of ‘causes’ faces us much more completely in each case; we reason, ‘this and that must precede for that to follow’ – but we haven’t thereby understood anything … (GS 112)

This theme is then developed further in BGE 21: here Nietzsche claims that ‘we should use “cause” and “effect” only as pure concepts, which is to say as conventional fictions for the purpose of description [Bezeichnung] and communication, not explanation [Erklärung]’. In these passages Nietzsche appears to be distinguishing between two aims of science, description and explanation, and asserting that even when we appear to be using causal concepts or offering explanations, we are really using them fictionally for the purposes of ‘description and communication’. To clarify what is going on here, Stoll provides helpful context on Schopenhauer’s influence:

[Schopenhauer] divides scientific investigation into two main branches: ‘it is either’, he writes, ‘description of forms [Beschreibung von Gestalten], which I call morphology, or explanation [Erklärung] of alterations, which I call etiology’ ([]WWR 1:96). Morphology is particularly characteristic of disciplines like botany, which deal with the classification and conceptual ordering of particulars into types, and those types into hierarchical genus-species relations. Etiology, by contrast, is the part of science that tries to determine causal laws and is the main business of modern science. But this mode of explanation, according to Schopenhauer, turns out to be only explanation in an impoverished sense. For, what we get at the end of our investigations is merely another list of forces whose nature we still fail to comprehend. Thus, Schopenhauer says, ‘it may be argued that all natural science basically achieves nothing more than what botany does, namely, the collection and classification of the homogenous’ ([]WWR 2:174); ‘even the most perfect etiological explanation of the whole of nature would really be no more than a catalogue of inexplicable forces and a certain specification of the rule’ according to which they operate ([]WWR 1:98). In other words, science, in the end, turns back into a merely descriptive rather than properly explanatory enterprise. (Stoll 2018, 526)

The crux of Schopenhauer’s rejection of explanation as a viable aim of science is a demanding conception of what it is to genuinely explain something. As Stoll highlights, it ‘was widely shared by German philosophers in the nineteenth century’ that ‘in order to have full insight into some explanandum, its ground must already be, or be made, familiar or “intuitive” to us in some sense’ (2018, 526). Science, by itself, can’t make the panoply of regularities it uncovers genuinely intuitive on Schopenhauer’s view; science can only uncover the brute facts of nature, and explaining these is an extra-empirical task.¹² And ‘description’ or ‘morphology’ for Schopenhauer is centrally bound up with facilitating prediction and control, as we might expect: ‘it involves the determination of empirical laws, which allow us to predict phenomena precisely and control our environment’ (Stoll 2018, 536).¹³

We see this concern with identifying regularities to aid in the prediction and control of nature frequently echoed by Nietzsche: in GS 112, he glosses description as that which helps us to reason ‘this and that must precede for that to follow’. In BGE 14, furthermore, having just rejected the view that physics can be genuinely explanatory in comments reminiscent of GS 112 and BGE 21, he goes on to insist that it can nevertheless be valuable for ‘a sturdy, industrious race of machinists and bridge-builders of the future, people with tough work to do’: he here appears to point to the way science’s descriptive capacity can aid our ability to manipulate nature to achieve our ends.¹⁴ In a 1884 note produced between the publications of GS and BGE, Nietzsche further characterises a core part of scientific method as the ‘transformation of nature into concepts for the purpose of controlling nature’ (NF 1884 26[170]); this suggests that when Nietzsche claims in BGE that ‘we should use “cause” and “effect” only as ‘pure concepts, which is to say conventional fictions for the purpose of description and communication, not explanation’, he sees this project of description and communication as tightly bound up with a project of predicting and controlling nature, as others did before and after him.

Intriguingly, we further see Nietzsche directly characterise psychology in particular as a discipline interested in ‘morphology’, Schopenhauer’s term for disciplines engaged in the pursuit of descriptive knowledge: he talks of ‘grasp[ing] psychology as morphology and the doctrine of the development of the will to power, which is what I have done’ (BGE 23). We see Schopenhauer’s influence on the notion of morphology, and the firm distancing of morphology from any attempt at explanation, in a notebook entry written the year before BGE was published: ‘In fact nothing is explained through morphological representation, supposing it were completed, but a tremendous fact is described’ (NF 1885: 36[28]).

These passages all look to me to pose a serious issue for the explanation assumption: to the extent that Nietzsche was sceptical about the possibility of genuine explanation in science in general, and in psychology in particular, it seems like a mistake to assume explanation is the principle epistemic end of his account of drive interaction. Defenders of the consensus view might respond that Nietzsche’s view of what it is to explain successfully is a very demanding one, and indeed one that we would likely reject now. But I am interested here in how Nietzsche understood the epistemic ends of his account of drive interaction, and his demanding notion of explanation looks to have pushed him to locate the epistemic value of such an account not in successful explanation, but instead in description: identifying the regularities of nature for prediction and control.

The defender of the explanation assumption might at this point take a different tack: while acknowledging the problematic passages above in GS and BGE, they might, following Leiter (2015, 18) and Clark (1990, 103–105), argue that Nietzsche’s scepticism about causality and the possibility of genuine explanation hinges on his problematic falsification thesis, which, on Clark’s influential developmental reading, is abandoned in the works after BGE. Nietzsche’s final works certainly appear to endorse the ‘healthy concepts of cause and effect’ (A 49) and seem to explicitly offer causal explanations (indeed, in TI Errors 5 Nietzsche appears to offer what he terms a ‘psychological explanation’).

The view that Nietzsche abandons in his final works the position we see him take in GS 112 and BGE 21 is controversial, however. Stoll (2018) denies that Nietzsche’s rejection of the viability of explanation hinges on any ‘falsification thesis’ that he may have at one point endorsed; as we’ve seen, Stoll suggests Nietzsche’s rejection of the value of explanation instead hinges on a particularly demanding conception of what genuine explanation consists in. Hussain (2004), on the other hand, rejects Clark’s view that Nietzsche abandons the ‘falsification thesis’ attributed to him: he highlights sceptical claims about causality as late as TI, and suggests that when Nietzsche appears to talk of cause and effect or explanation in his final works, he still does so only in the sense pointed to in BGE 21 – he uses these terms as ‘conventional fictions’ for the purpose of ‘description and communication’, not to make genuine explanatory claims.

If we side with Stoll and Hussain in this debate, then the explanation assumption remains in trouble. Even if we do not, however, the plausibility of the explanation assumption still looks strained. Many passages where Nietzsche details his account of drive interaction occur in or before BGE, where we still see Nietzsche rejecting explanation as a viable epistemic end of science: Nietzsche’s famous claim that we should see the ‘soul as a society constructed out of drives and affects’, for instance, which features prominently in both Riccardi and Clark and Dudrick’s account, is made in BGE 12. And, as we have already seen, Nietzsche emphasises the importance of distortion and idealisation to good psychology as late as TI in Skirmishes 7-8: while Nietzsche may have moved beyond the claim that all beliefs are false, and thus that causal explanations are always fictional and fail to genuinely explain, to the extent that his fictive account of drive interaction still here hinges on false, heavily idealised characterisations of how the mind works, it seems most plausible to assume that he would have retained for it the more circumscribed aim of description, rather than explanation. This also makes the most philosophical sense: as we’ve noted, it is tough to see how false characterisations of how the mind works could genuinely explain.

We might finally wonder, here, how a fictive account of drive interaction might help us with the task of description and prediction. As Stoll (2018) highlights, however, Nietzsche would have been able to see a path forward towards solving this problem. Du Bois-Reymond, a noticeable influence on Nietzsche’s philosophy of science, prominently argued that fictional posits (such as atoms) could nevertheless aid us in prediction and control, as Stoll summarises:

[atoms on Du Bois-Reymond’s view] serve a merely heuristic function as logical subjects for the predication of certain kinematic and dynamical states … [t]hey do not allow us to comprehend anything about the nature of matter, but serve as shorthand for communicating, predicting, and modelling regularities in a readily intelligible way. The mention of such atoms in physical theories, however, is ontologically non-committal. (2018, 530)

We see, then, from a thinker that Nietzsche knew well, the idea that fictions can facilitate achieving something akin to his notion of ‘description’. The idea is nevertheless fairly inchoate. I’ll now turn to a prominent account of how falsehoods in science can facilitate epistemic achievement; this will help us clarify and reconstruct how all this might work.

4. Elgin’s account of the role of felicitous falsehoods in science and art

4.1. An overview of Elgin’s account

We can start here by noting that models in science are often heavily idealised and misrepresent various features of their targets (Elgin 2017). The Lotka-Volterra model frequently deployed to help understand predator-prey population dynamics, for instance, assumes that population sizes change in a continuous way (so you could have, say, 32.3 wolves in a population) and that all individuals in a species are homogenous, with age and size having no impact on the likelihood of death (Knuuttila and Loettgers 2016). It is clear these features cannot be true of any real biological systems, but models such as this continue to be taken to have clear epistemic value in the scientific community.

Elgin (2017) provides a compelling account of how this works in True Enough, where she argues that epistemology needs to acknowledge the role that can be played by ‘felicitous falsehoods’: these are inaccurate representations whose inaccuracy does not undermine (or even facilitates) their epistemic function (3). Her claim is that felicitous falsehoods are not mere heuristics, but ‘ineliminable and epistemically valuable components of the understanding science supplies’ (1).

Central to her argument is the way falsehoods (in science but also art) can provide a special kind of ‘epistemic access’ to their targets (2017, 2). Reality is messy, and it is easy to miss patterns that are really there in the ‘blooming, buzzing confusion’ that the world confronts us with (250). Falsehoods, through simplifying, distorting and exaggerating, can exemplify features of their target and thereby make salient patterns and regularities which otherwise might have been overlooked, facilitating genuine epistemic achievement.

The Lotka-Volterra model, thanks to its transparently false assumption that populations can be continuous, makes mathematically tractable a set of differential equations about how population numbers develop. When we compare the predictions of the model to how various predator-prey populations fluctuate in reality, we often see that patterns very close to those predicted by the model were there all along: it thereby ‘discloses a regularity’ while remaining indifferent on the ‘precise reproductive mechanisms’ underpinning it (Elgin 2022, 14). Models can thus ‘equip[s] us to recognise and marshal information we already have’ (Elgin 2017, 243); on Elgin’s account, they can grant us the know how to wield the rest of our knowledge base more effectively.

Importantly, Elgin highlights that falsehoods in art can similarly facilitate epistemic achievement through exemplification. Lolita, for instance, in being ‘written from the perspective of an utterly unreliable narrator’, exemplifies the limits on our introspective capacities (2017, 240). This is interesting because, as we have noted, Nietzsche sees strong continuities between good psychological work and art (TI Skirmishes 7); he further considers novelists like Stendhal and Dostoevsky exemplar psychologists (TI Skirmishes 45; BGE 39). Elgin’s account helps us make sense of this; it is also thus well placed to help us see the continuities between Nietzsche’s rich, colourful, almost literary accounts of drive interaction and more rigorous, often quantitative standard examples of scientific models, such as the Lotka-Volterra model mentioned above.

On Elgin’s view, while we should not believe felicitous falsehoods, we can still accept them, that is take them as ‘bas[e]s for action … when one’s ends are cognitive’ (Elgin 2017, 19). The standard by which we assess the epistemic value of a felicitous falsehood, and which governs whether we ought to accept it or not, is its capacity to further our cognitive ends; while Elgin opens up some room for variation in cognitive ends across different practitioners and disciplines, a prominent potential cognitive end she highlights is accurate prediction (2017, 112–113).¹⁵ In this way, literally false models still need a ‘tether to the facts’ if we are to accept them; we might accept the Lotka-Volterra model but not accept certain tenets of astrology because we find the predictions astrology would allow us to draw are not tethered to the facts, while those of the Lotka-Volterra model are.

On Nietzsche’s ‘descriptive’ or ‘morphological’ conception of the epistemic value of psychology, we might then accept his ‘felicitously false’ account of drive interaction if it exemplifies or make salient patterns and regularities in observable person-level psychological phenomena that we otherwise might have missed, and if these regularities then help us predict and control the development of these psychological phenomena. I will give a concrete example of how I take this to play out in section 5.2, where I suggest that Nietzsche’s account of drive interaction allows us to identify and predict those who will exhibit signs of ‘Nietzschean health’.

4.2. Revisiting the veracity assumption

Talk here of how Nietzsche’s account of drive interaction still requires a ‘tether to the facts’ might prompt us to revisit the veracity assumption: we might wonder, if Nietzsche’s account is only accepted if it ‘gets things rights’ in a certain way and facilitates correct predictions, whether it isn’t true in some sense. Even if not literally true, Nietzsche’s characterisations of drive interaction might still be true with some kind of paraphrase in much the same way that we might talk of literally false metaphors being ‘true’.¹⁶

I do not think conceding that Nietzsche’s account is ultimately tethered to the facts forces us to accept the veracity assumption in light of the evidence already presented against it, however. As Elgin highlights, felicitous falsehoods can facilitate epistemic achievement by making salient features we otherwise would have missed, thereby equipping us to wield the rest of our knowledge base to better achieve our cognitive ends. The inferences that felicitous falsehoods help us to draw are tethered to the facts, but that doesn’t transform those falsehoods themselves into true representations.

An example helps make clear the way a model’s capacity to facilitate a cognitive end like predictive success can come apart from true representation.¹⁷ Weisberg highlights how students are often taught G.N. Lewis’ electron-pair model of chemical bonding, which treats ‘chemical bonds … as electron pairs shared between two atoms’ (2013, 107). We now know this model provides a distorted and inaccurate picture, and that models incorporating insights from quantum mechanics provide a more accurate representation of how chemical bonding works. These more complex models are, however, difficult for students to employ towards making accurate predictions without a lot of background knowledge; the electron pair model of chemical bonding better allows certain practitioners (students) to make predictions which are reasonably accurate when they don’t yet grasp the principles of quantum mechanics.¹⁸ This greater predictive capacity is ultimately tethered to the facts, but relies on the model’s distortions and simplifications of the underlying causal structure; this does not make it, to my mind, ‘truer’ than the more complex and representationally accurate competing models in any meaningful sense.

5. Upshots of rejecting the consensus view

5.1. Nietzsche does not commit himself to a fallacious form of homuncularism in the way originally feared

With the above in hand, we can now see how Nietzsche sidesteps entirely the two fallacies he has typically been charged with. The mereological fallacy can be dealt with once the veracity assumption is dropped: category mistakes of this kind are less problematic if we take his talk to be fictive. In fiction, it is frequently the case that we can make sense of representations involving conceptual impossibilities: we can make sense of Lewis Carroll’s Cheshire Cat disappearing except for its grin, say, despite it plausibly being a conceptual impossibility to have a grin present without a face (Nolan 2021; see also Gendler 2000).

Similarly, worries about homuncular explanatory regress recede once we stop taking Nietzsche to be involved in the project of explaining our psychological capacities. We might worry, however, about being too quick here: we might worry that Nietzsche’s attempt to exemplify features of human psychology through felicitously false representations of drives as people leads nevertheless to some problem of regress. We seemingly already have to understand human behaviour to understand what talk of the drives as agents is meant to exemplify.

Here, however, we should return to Elgin’s account of how exemplification works: she notes that some features of a representation made to exemplify will always be inert, because an exemplification is only ever a likeness (2017, 265). When Dalton used a set of real billiard balls to model atoms in the 1800s, the material they were made out of, as well as the fact that they were themselves divisible and admittedly composed of atoms, were inert in the representation (while the fact that they were all of the same shape and mass was salient). This is why this model, while now outdated, still facilitated genuine understanding, even though it sought to model atoms with things that themselves were composed just of atoms. Similarly, when Nietzsche treats the drives as conscious agents in his modelling of the human psyche, I suggest that the fact that these agents would themselves hypothetically have a complex psychic drive economy is broadly inert in the cases Nietzsche focuses on. When Nietzsche suggests we treat our psyche like a society, the features Nietzsche is trying to exemplify, perhaps that there are issues when you give any one element too much control,¹⁹ could be understood even by someone who had never heard of Nietzsche’s drive psychology. The issue of explanatory regress can thus be avoided in the same way it was for Dalton’s model of the atom.

5.2. Opposing the consensus view allows us to more fully appreciate the richness of Nietzsche’s drive psychology

It’s worth highlighting here a key drawback for those who subscribe to the consensus view: they typically only attempt to offer a reading of a certain subset of the rich agential portrayals Nietzsche provides of drive interaction, focusing on the most easily assimilable agential characterisations he deploys.²⁰ Such accounts implicitly have to excise many of the more colourful characterisations Nietzsche offers in their attempts to show how he can avoid the homunculus fallacy. This is because they are constrained by trying to find a way to interpret Nietzsche’s characterisations as true without committing the category mistakes that characterise the mereological fallacy; with our review of Elgin in hand, however, we can instead focus on what they might exemplify despite being obviously false, opening ourselves up to the full nuance of Nietzsche’s characterisations. I think doing so might help us start to resolve outstanding interpretative puzzles in his account of the drives.

As an example, take the puzzle of how we are to understand the relationship between Nietzsche’s notion of ‘health’ and a certain ‘structure’ of the drives. On a prominent reading (Gemes 2009; Richardson 2009), one enters a state of health (characterised, for instance, by the tendency to affirm life and the possession of substantial psychic and physical energy; see Huddleston 2017) as a result of a relationship of integration between the drives, in which ‘there is a dominant master drive … that reins the other drives in’ (152) so that they find expression in a manner that is ‘consonant’ (Gemes 2009) with that master drive’s purpose. Gemes (2009) argues that this characterisation of drive interaction draws our attention to cases typified by a well-known Freudian analysis of Da Vinci’s sexual expression: rather than actually engaging in same-sex sexual activity, according to Freud, Leonardo ‘reined in’ his homoerotic drive, redirecting it in the service of a master artistic drive towards the drawing of idealised nude male bodies such as the Vitruvian Man. We can thus take talk of a master drive reining others in in this sense to exemplify something about health: Nietzsche’s socio-political drive model helps us to notice how individuals like Da Vinci who enjoy the great reserves of energy that enable substantial achievement typically give expression to a wide range of typical human tendencies in the pursuit of their life’s main goal. On my reading, it thus attempts to make salient a regularity about certain healthy individuals; if it gets things right, it should thus help us predict who will go on to demonstrate features of Nietzschean health (such as a tendency to affirm life or demonstrate substantial reserves of psychic and physical energy). It might even prompt us to think of ways we can change our own behaviour that might increase the likelihood of demonstrating such features ourselves.²¹

While much textual evidence has been adduced to support such an account, it nevertheless seems in subtle tension with some of the examples of ‘healthy’ and ‘sick’ individuals that Nietzsche gives. As Huddleston (2017, 154–156) notes, Nietzsche explicitly highlights the way that Socrates co-opted baser sexual and aggressive drives towards a master drive oriented towards reason (TI Socrates 8), which seems like a textbook case of successful integration akin to the Da Vinci case. The issue here is that it is clear that Socrates is not healthy by Nietzsche’s lights: he is portrayed in this very same segment of TI as ‘sick’ (TI Socrates 12).

It is interesting to contrast Caesar here, an exemplary psychological case study in Nietzsche’s view (BGE 200; TI Skirmishes 38); while Caesar might’ve had, say, tightly integrated aggressive, political and intellectual drives, some of his behaviour seems integrated to a much lesser extent. Contrast Caesar’s sex drive to Socrates’: while Socrates rejects the handsome Alcibiades’ advances in the Symposium, preferring to give erotic drives expression through dialectic, Caesar is notorious for having had several wives and many more mistresses. While some of these relationships ended up tightly bound up with his political and military achievements (his affair with Cleopatra, perhaps), many others seem only tangentially related to his life’s central project. The degree of integration of this drive certainly seems much weaker than in the case of Socrates.

I suggest looking here at one often overlooked but particularly rich version of the socio-political drive ‘model’ that Nietzsche offers us: that of the Roman empire. As Huddleston (2019, 92) highlights, when talking about health and decadence, Nietzsche sees a ‘macrocosm-microcosm relation between culture and the individual’, with cultures similarly more healthy and less decadent when they exhibit a certain unity by integrating their different components. Nietzsche is particularly effusive about the Roman empire, healthy and unified as it was in contrast to the ‘decadent’, ‘anarchist’ regime of Christianity that would succeed it (A 58).²² He notably directly analogises it to the supremely healthy Caesar, comparing its political organisation to his structure of drives in being similarly ‘five paces away from tyranny’ (TI Skirmishes 38).

I think we can fruitfully start to resolve the puzzle about the counterexample of Socrates that we started with if we take a serious look at what Nietzsche might be trying to exemplify with the Roman empire when he says it, like Caesar’s drive organisation, is ‘five paces away from tyranny’ (Socrates’ integrating master drive of reason is in contrast a ‘tyrant’ in TI Socrates 10). Consider a stylised account of how Rome works: it has a strong imperial core closely aligned with aiding the flourishing and expansion of the empire, drawing on the talents of military, intellectual and administrative elites. As you radiate further away from the core, however, government becomes increasingly decentralised, with distant provinces given substantial autonomy in how they govern their affairs, with particularly powerful or culturally distinct groups given greater autonomy.

This highlights an important point that the Roman empire exemplifies: the Roman system was distinctive in its wariness towards integrating too far; while greater integration is broadly desirable for an imperial regime, all else equal, as it grants greater control to the metropole, the Romans realised that all else isn’t in fact always equal. Integration comes with trade-offs: if you push too hard, too fast with peoples who are too different to the Romans, you risk doing more harm than good to the regime through provoking unrest and rebellion. This is at various points explicitly highlighted as a key feature of Rome’s strength by Nietzsche: in A 58 he highlights ‘the history of the Roman provinces’ as of particular interest if we are to understand the Roman empire, with ‘its design … calculated to prove itself over millennia’; in BGE 46, Nietzsche repeatedly emphasises Rome’s ‘noble and frivolous tolerance’.

Sprawling empires, and the Roman empire in particular, then neatly exemplify the delicate trade-offs at play in their quest to expand their control and integrate far-flung peoples. We can then see the sense in which they are always ‘five paces from tyranny’: they are integrated just up to the point where, if they went any further, they’d tip into being felt to be tyrants, rather than masters, and ultimately trigger uprising and anarchy. So, I suggest this model helps clarify the puzzle about Socrates and Caesar: Nietzsche certainly thinks a kind of integration predicts markers of psychic health, but he seeks also to highlight that there is a different, delicate balance to be struck for each drive, and if a drive’s expression becomes too indirect you risk provoking sickness in yourself through ‘tyrannising’ your instincts. Socrates, in Nietzsche’s view, is just such a case of overshooting the mark; Caesar, in contrast, successfully hits the Goldilocks zone.²³ This model thus helps draw out and make salient that monomaniacs are unlikely to end up exhibiting signs of Nietzschean heath.

Stepping back, this serves as an example of the avenues that might be opened up to us if, instead of trying to paraphrase away Nietzsche’s agential characterisations of drive interaction and focus on those that can plausibly be rendered true, we instead ask exactly what he might’ve been trying to exemplify by choosing the rich models that he did.

6. Conclusion

I’ve argued here against the consensus view of the epistemic ends of Nietzsche’s account of drive interaction, which takes Nietzsche to be involved in a project of providing true explanations of a range of psychological phenomena. Instead, I offer a reading on which Nietzsche’s rich agential characterisations of drive interaction are fictive but can nevertheless have epistemic value through exemplifying features of our psychology that we otherwise might’ve missed, thereby helping us to identify psychological regularities and make novel predictions. I take such a reading to be supported by Nietzsche’s explicit comments on scientific methodology. It draws further support from the way it allows Nietzsche to smoothly sidestep worries about the homunculus fallacy. Finally, whereas those who hold the consensus view largely discard them, my opposing account gives us the scope to more fully appreciate the significance of the most rich, intricate socio-political characterisations of the psyche that Nietzsche offers us.²⁴

Disclosure statement

No potential conflict of interest was reported by the author.

Notes

1 I use the following abbreviations for Nietzsche’s texts: Daybreak – D; The Gay Science – GS; Beyond Good and Evil – BGE; Twilight of the Idols – TI; The Antichrist – A; Ecce Home – EH; Unpublished Fragments – NF [Year]. All translations of the published works are from the ‘Cambridge Texts in the History of Philosophy’ editions; the translations of the unpublished fragments are from the Stanford University Press editions of Nietzsche’s complete works. I use the following abbreviations for Schopenhauer’s texts: The World as Will and Representation, Volume I – WWR I; The World as Will and Representation, Volume II – WWR II. I use the E.F.J. Payne translation for these two texts.

2 See also Drayson (2012) for further discussion of these two components of the homunculus fallacy.

3 See also Riccardi (2021, 49). For further elaboration on what makes an explanation distinctively subpersonal, see Drayson (2012).

4 Anderson (2005) offers one notable development of this strategy.

5 I thank Christopher Fowles for the suggestion that I use this passage to clarify TI Skirmishes 7.

6 Nietzsche, to be clear, considers his account of the drives to be ‘psychology’ as he understands it: BGE 12 and TI Skirmishes 38 are examples of cases where he indicates this.

7 I suspect that Nietzsche’s use of idealisation extends beyond his claims about drive interaction to some of his more speculative claims about individual drives. Consider an interpretative puzzle raised by Stern (2015): Nietzsche’s claim that our ability to identify and name the drives is inevitably extremely limited (D 119) seems in tension with his various highly specific characterisations of specific drives in his psychological analyses (consider, for instance, his discussion of the prominence of the ‘histrionic instinct’ in diplomats in GS 361). This puzzle could be overcome if we took these latter claims to be fictive; I suspect they may well be, but I do not have the space to fully develop this case here.

8 I am indebted to Brian Leiter for various comments and suggestions which helped me clarify my views on the explanation assumption, and in particular for pushing me to draw out the importance of prediction as an alternative epistemic end.

9 See Salmon (1989, 3–4) and Potochnik (2015) for further discussion. Note that, while prediction may have pragmatic value, it is nevertheless a distinctively epistemic aim: a good prediction is one that is right.

10 For a detailed overview of the issues facing the thesis that predictions and explanations are symmetrical, see Salmon (1989, Ch. 2).

11 Little is said about whether Nietzsche values prediction too, to the extent the goals come apart.

12 Interestingly, this backdrop is also highlighted by Salmon (1989, 4–5) in his classic history of the development of views of explanation in the 20th century: he notes that in a ‘philosophical context dominated by post-Kantian and post-Hegelian German idealism’ in the late 19^th and early 20th centuries, many philosophers ‘argued there could be no genuine explanation of any fact of nature that did not involve an extra-empirical appeal … Many scientific philosophers (as well as philosophical scientists) reacted to this attitude by denying that science is in any way concerned with explanation’.

13 Stoll highlights a second strand to morphological or descriptive knowledge for Schopenhauer: it doesn’t just seek to predict but has a further ontological upshot in helping us ‘determine[] which forces make up the material world and the priority relations between them’ (2018, 536). It is not clear to me that Nietzsche took up this second strand; while Stoll suggests he does, his primary evidence looks to come from unpublished notes and letters (534). But even if Nietzsche does accept this second more robust aim of description, it would not be sufficient for genuine explanation on the demanding conception of explanation in play.

14 Compare also GS 335, where Nietzsche suggests knowledge of the regularities of nature is an important aid in the project of self-creation: ‘we must become the best students and discoverers of everything lawful and necessary in the world: we must become physicists in order to be creators’.

15 Elgin often refers to this broad group of cognitive ends as contributing to what she terms, following Kvanvig (2003), ‘objectual understanding’: objectual understanding is the understanding of a topic or subject matter. She is at pains to distinguish objectual understanding from ‘understanding why’, however; she thus accommodates a view on which understanding might be measured by predictive success rather than the ability to offer true explanations. Someone might demonstrate an (objectual) understanding of the New York subway system by showing an ability to correctly predict the best route to take at different times of day and in different weather conditions; they may, despite this predictive ability, be unable to explain their know how – they might not be able to explain why their proffered routes are the best ones to take, even if they in fact are so (Elgin, 2017, Ch. 3).

16 This claim is certainly still itself philosophically controversial; for one account on which metaphors can in some sense be true, see Hills (1997). I will, for the sake of argument, concede that such a notion is viable here.

17 See Weisberg’s account of the fidelity criteria by which we evaluate models (2013, Ch. 3) for further discussion on why we should separate our evaluation of a model’s predictive power from its ability to accurately represent its target’s causal structure.

18 We see here, recalling Elgin’s account, how the value of felicitously false models depends centrally on how they enable a given practitioner to make the best use of the rest of their knowledge base.

19 See section 5.2 for further discussion.

20 Riccardi (2021), for instance, focuses heavily on how a causal dispositional account can be provided for talk of drives ‘commanding and obeying’.

21 Whether Nietzsche’s model is in fact a good one, and actually has predictive power, is not a question that I will attempt to answer here.

22 Note the near identical terms used to characterise Socrates (TI Socrates 4).

23 What this balance is will vary from person to person, based on the strengths and natures of their various drives: ‘health … of course could look in one person like the opposite of health in another’ (GS 120). The suggestion is not that everyone needs to behave exactly as Caesar did if they want to be healthy.

24 This paper benefited greatly from the comments of audience members at the London Nietzsche Seminar and the 2023 International Society for Nietzsche Studies workshop in Atlanta. I am grateful for detailed written comments from Ian Dunkle, Joseph Alfon, Daniel Davis and Brian Leiter which prompted substantial revisions of the piece. I owe a particular debt of gratitude to Andrew Huddleston for encouraging me to pursue this project, and for having the patience to offer extensive guidance on its numerous iterations.