Mill’s Handle

Putting a name on a concept makes it easier to handle.

Giving a stable name to a recurring observation or proposed phenomenon makes it easier to recognise, remember, compare, communicate, investigate, and debate; but the name itself provides no evidence that the proposed category is coherent, the phenomenon is real, or the claim about it is true.

Human beings notice patterns long before they know what to call them. A merchant can notice that people become more willing to take risks when somebody else bears the cost without knowing the term moral hazard. A manager can observe that employees begin manipulating a measurement once their rewards depend upon it without having heard of Goodhart's Law. A historian can notice that institutions gradually acquire interests different from the purposes for which they were founded without possessing an established name for the phenomenon. The absence of terminology does not make these observations impossible.

Something changes, however, when the observation acquires a stable name. What previously required a paragraph to describe can now be summoned in a few words. Earlier examples can be recognised as possible instances of the same thing, people can search for other observations of it, and two people can dispute its meaning without continually reconstructing the original idea. The phenomenon, or at least the claim that there is a phenomenon, has acquired a handle.

That handle does not prove anything. Moral hazard might be invoked where no moral hazard exists, an alleged instance of Goodhart's Law may have another explanation, and a newly coined social or psychological category may turn out to combine several unrelated things. Naming gives the mind something to grasp, but the existence of the handle tells us nothing about the solidity of what we think we are holding. This tension between the enormous usefulness of names and their ability to create unwarranted confidence is what we will call Mill's Handle.

IMill and the Point of Attachment

John Stuart Mill came remarkably close to identifying this problem in A System of Logic in 1843. One chapter bears the unusually appropriate title Of Naming, as Subsidiary to Induction. Mill was interested not merely in language as a means of communicating conclusions but in the role names play in reaching conclusions in the first place. Names, he argued, help fix associations among ideas and give otherwise fleeting thoughts a point of attachment that allows them to be recalled and kept before the mind.

Mill was careful not to make the stronger claim that thought is impossible without language. He considered the proposition that names are indispensable to reasoning an exaggeration, although an exaggeration built around an important truth. People can perceive conjunctions, remember particulars, draw inferences, and recognise similarities without first assigning everything a verbal category. The importance of naming is therefore not that an unnamed thing literally cannot be thought about, but that a named thing becomes much easier to retain, recover, connect with other observations, and submit to sustained reasoning.

This distinction gives us the central metaphor. A heavy object can exist without a handle, and with enough effort we may still move it, turn it around, or carry it somewhere else. Adding a handle does not create the object or tell us what the object is made of, but it dramatically changes what we can conveniently do with it. A good concept-name performs a similar service for thought.

This is why Mill's Handle is not intended as another version of the strong claim that language determines the boundaries of possible thought. The empiriceme is narrower and more defensible. Naming reduces the difficulty of repeatedly handling a concept, and this reduction in cognitive and communicative cost can have consequences far beyond the convenience of having a shorter expression.

IILupyan and the Aliens

Modern experimental psychology provides evidence for something close to Mill's intuition. In a 2007 experiment, Gary Lupyan, David Rakison, and James McClelland taught participants to classify unfamiliar imaginary creatures, described as aliens, into categories. The participants received feedback that was sufficient to learn the classification, but some also heard arbitrary verbal labels attached to the categories. The labels were redundant: they did not contain additional information about which category was correct.

The redundant names nevertheless helped people learn the categories. A follow-up comparison found that merely providing a nonverbal association did not produce the same benefit, leading the researchers to argue that verbal labels can participate in shaping category formation rather than merely being attached to categories after those categories have already been learned. Lupyan subsequently developed this line of research into the label-feedback hypothesis, according to which language can influence perceptual and conceptual processing by activating category-level information.

The experiment should not be inflated into a universal law that anything becomes clear merely because we name it. What it does demonstrate is more modest and more useful: even an informationally redundant label can change how readily people acquire and use a category. The word does not have to supply new evidence to have cognitive value; sometimes it makes existing evidence easier to organise.

That is the first half of Mill's Handle. A name is not simply an abbreviation used after the intellectual work has been completed. Under at least some conditions, attaching a stable verbal handle to a category helps us perform the intellectual work itself.

IIIFrom Repeated Experience to a Common Object

The effect becomes even more important when we move from individual cognition to collective inquiry. Imagine that twelve people independently notice approximately the same pattern, but each describes it differently. One calls it the tendency of a bureaucracy to protect itself, another describes administrators placing institutional survival ahead of mission, a third writes about organisational self-preservation, and a fourth tells several stories without proposing any general principle at all. They may have encountered the same phenomenon, several related phenomena, or nothing more than superficially similar anecdotes.

Without a shared handle, discovering the relationship among these observations is laborious. Each description must be read, remembered, compared, and translated into the terminology of the others before anyone can even ask whether the observations belong together. Give the proposed pattern a stable name, however, and the scattered observations can begin accumulating around a common point. Examples can be added, counterexamples recorded, neighbouring phenomena distinguished, historical antecedents discovered, and evidence from different disciplines compared.

A name therefore does something especially important for collective reasoning: it creates an address for evidence. The evidence may support the claim, weaken it, subdivide it, or eventually destroy it, but researchers at least know where to deliver what they have found. Without such an address, observations that ought to confront one another may pass unnoticed in separate literatures for decades.

This is one reason human knowledge contains so many named effects, laws, paradoxes, principles, biases, rules, syndromes, and fallacies. The names are not merely decorative titles attached to finished discoveries. They allow a proposition to persist through time and move among people who were not present when the original observation was made. A sufficiently useful handle can turn a private recognition into a public problem.

IVA Name Makes Disagreement Easier

We normally think of shared terminology as facilitating agreement, but one of its greatest benefits is that it facilitates precise disagreement. When an observation remains embedded in a story, two people can exchange anecdotes indefinitely without establishing exactly where their disagreement lies. One may be questioning whether the pattern exists at all, another how frequently it occurs, and a third whether the proposed mechanism actually explains it.

Once the disputed phenomenon has a stable identity, the argument can be decomposed. Does it exist? How large is the effect? Under what conditions does it occur? Does the same pattern appear in other societies or periods? Is the category too broad? Are two different mechanisms being confused under one label? Does a rival explanation account for the observations better?

These are much better questions than an exchange of impressions because the debate can accumulate. A new study can answer an earlier study, a historical case can challenge a supposedly universal claim, and a failed replication can weaken a defined proposition instead of merely adding another story to the world. Naming has converted an observation into something capable of receiving organised criticism.

This is especially important to empiricalism because many empiricemes originate outside formal research. They may begin as proverbs, professional maxims, historical observations, jokes, political sayings, management rules, or memorable remarks attributed to a particular person. Before their truth can be assessed, the recurring claim itself has to be identified. Naming is often the operation that makes that identification durable.

VWhen Society Is Missing the Handle

Philosopher Miranda Fricker's discussion of hermeneutical injustice illustrates another dimension of the same process. Her best-known example concerns the emergence of the concept sexual harassment. Before that expression became part of ordinary public vocabulary, the behaviour it described certainly occurred, but people subjected to it could lack an adequate shared concept with which to interpret and communicate what was happening. Fricker describes such cases as gaps in the collective interpretive resources available to a society.

The importance of the new term was therefore not that two words magically caused people to experience something that had never happened before. The term permitted experiences that had previously seemed isolated or difficult to articulate to be grouped, compared, discussed, and eventually disputed within a common conceptual framework. People who discovered that others had encountered the same behaviour could now recognise a class of experiences rather than merely a succession of individual incidents.

Fricker's theory concerns questions of social power and injustice that go well beyond Mill's Handle, and a missing word should not automatically be equated with a hermeneutical injustice. Her example nevertheless demonstrates the collective importance of conceptual handles unusually clearly. Sometimes a society possesses many observations but lacks the category that allows people to see that the observations may belong together.

Once that category exists, the intellectual landscape changes. The new name can enter law, journalism, scholarship, ordinary conversation, criticism, and countercriticism. What had been difficult even to point toward becomes something about which propositions can be made and evidence demanded.

VINames Make Retrieval Possible

There is also a mundane but extraordinarily powerful advantage to naming: retrieval. Knowledge is useful only if it can be found again, whether in an individual memory, a library catalogue, a scientific database, or an internet search. An unnamed observation may be recorded perfectly and nevertheless become almost invisible because every later investigator must guess the words originally used to describe it.

A stable name compresses a complex proposition into a relatively durable search term. Someone encountering survivorship bias for the first time can quickly discover discussions in military analysis, finance, entrepreneurship, medicine, history, and statistics. Without the shared term, each field could describe the same selection problem in its own vocabulary while remaining partially unaware of the others.

The benefit compounds over time because the handle creates a feedback process. A useful name makes a phenomenon easier to retrieve; easier retrieval makes earlier work easier to cite; increased citation makes the term more standardised; greater standardisation makes subsequent work easier to locate. A successful handle can therefore become part of the infrastructure through which a body of knowledge accumulates.

That observation also suggests a testable extension of Mill's Handle. Named claims should, other things being equal, be easier to retrieve and connect across time than equivalent unnamed claims. Memorable names may even attract more discussion and testing than equally important but poorly named observations, which would mean nomenclature can influence not merely how knowledge is organised but what receives attention in the first place.

VIIThe Handle Can Be Attached to the Wrong Thing

Everything said so far makes naming sound beneficial. The danger begins at exactly the point where naming succeeds, because a concept that becomes easier to think about also becomes easier to believe in. Grammar assists the transition: once we turn a proposed pattern into a noun, ordinary sentences begin treating the noun as though it referred to an established object.

Suppose someone coins Executive Empathy Decline for the claim that people lose empathy as they rise through organisational hierarchies. Almost immediately it becomes possible to write that Executive Empathy Decline explains the behaviour of senior management, or that a particular company suffers from Executive Empathy Decline. Yet nothing about capitalisation, repetition, or possession of a memorable name establishes that executives actually lose empathy, that advancement causes the change, that the effect is general, or even that the proposed cases belong to a single coherent phenomenon.

This is the danger of reification: an abstraction, classification, or proposed pattern begins to be treated as though its existence had already been demonstrated. Closely related is the mistake captured by the maxim naming is not explaining. Nilsson and Lövdén used that phrase in discussing theories of cognitive reserve and brain maintenance, warning that a useful theoretical label should not be mistaken for a causal explanation simply because it gives researchers a convenient way of referring to the observations that require explanation.

Mill's Handle therefore contains its own warning label. The better the handle, the easier the concept becomes to carry around, and the greater the temptation to forget that what we are carrying may still be only a hypothesis. Nomenclature lowers the friction of reasoning, but it can also lower the friction of error.

VIIIOne Thing with Two Names, Two Things with One Name

Scientific psychology has a particularly useful vocabulary for another problem created by naming: the jingle-jangle fallacies. The jingle error occurs when different constructs are treated as the same because they have been given the same label. The jangle error runs in the opposite direction: researchers give similar or identical constructs different names and then proceed as though they have discovered different things.

These are not trivial editorial mistakes. If five research traditions study essentially the same phenomenon under five different names, evidence that ought to accumulate may instead remain fragmented. Conversely, if substantially different phenomena are collected beneath one attractive label, inconsistent results may be interpreted as mysterious failures to replicate when the researchers were never studying the same thing in the first place.

An empirical catalogue therefore cannot assume that names reveal the natural boundaries of concepts. Synonyms may need to be merged, broad categories may need to be divided, and inherited names may have to be retained for historical reasons while their ambiguity is explicitly recorded. Naming is indispensable to organisation, but nomenclature itself must remain subject to empirical criticism.

The pair is itself an example of the problem it names. Kelley coined the jangle fallacy in 1927, but the jingle fallacy is older: Thorndike stated it in 1904, and credited the term to Professor Aikins. The phrase jingle-jangle now travels as a single coinage under Kelley's name, which is a handle that has quietly detached itself from half of what it holds.

This point is particularly important when an eponym becomes famous. Parkinson's Law, the Peter Principle, Goodhart's Law, and similar names are exceptionally effective handles, but the memorability of the label is logically independent of the strength of the evidence. A beautifully named proposition may be weak, while a cumbersome and obscure proposition may be extremely well supported.

IXThe Handle Can Move the Thing

When concepts concern human beings, another complication appears. People can learn the classifications applied to them and respond to those classifications. Ian Hacking called this interaction between classifications and the people classified a looping effect: the category can change behaviour, changed behaviour can alter the category, and the object of study becomes a moving target.

A classification of minerals does not normally cause minerals to reconsider their behaviour. A classification of students, patients, employees, voters, criminals, generations, personality types, or social groups can have very different consequences. People may embrace a label, resist it, strategically exploit it, modify their conduct because of it, or change how they treat others to whom the label has been applied. Once the classification becomes socially important, the description itself enters the causal system being described.

This does not mean that social categories are unreal or that empirical investigation of them is impossible. It means that some empiricemes can become reflexive. Publicising an observation about human behaviour may weaken it because people learn to guard against the behaviour, or strengthen it because people begin expecting and reproducing it.

The metaphor of the handle acquires an unexpected second meaning here. Usually a handle merely allows us to manipulate something more easily, but in the human sciences the act of grasping the handle can occasionally move the object. An empiriceme about people may behave differently after it becomes a famous empiriceme.

XWhat Makes a Good Empiriceme Name?

If names have this much influence, choosing them is part of the methodology rather than an afterthought. A good empiriceme name should be memorable enough to function as a handle, distinctive enough to separate the proposition from its neighbours, and short enough to travel easily between contexts. A name that requires the proposition to be re-explained every time has failed at one of the primary jobs for which names exist.

The name should also avoid building more certainty into the claim than the evidence warrants. Calling an observation a law may imply universality where only a tendency has been observed; calling something a fallacy may imply that an argument is necessarily erroneous when its validity actually depends upon circumstances. Historical names should usually be preserved because they are themselves evidence about how the proposition has travelled, but the empirical status of the claim should never be inferred from the grandeur of its title.

Aliases matter for the same reason. Two maxims coined independently may turn out to be versions of one empiriceme, while a single celebrated expression may have accumulated several incompatible meanings. Recording alternative names, earlier formulations, and related concepts makes it possible to examine whether the handles are attached to the same intellectual object.

Provenance matters as well. Mill's Handle is a name coined here; John Stuart Mill did not formulate an empiriceme under that title. His discussion of naming as an aid to induction provides the historical anchor, while Lupyan supplies experimental evidence concerning labels and category formation, Fricker illuminates the consequences of missing shared concepts, Hacking shows how human classifications can become interactive, and the literature on reification and jingle-jangle errors supplies necessary cautions. The proposition is a synthesis, and its name should not conceal that fact.

XIThe Empiriceme About Empiricemes

The idea of an empiriceme is itself an example of Mill's Handle. The world already contained laws, effects, maxims, biases, paradoxes, principles, rules of thumb, professional sayings, historical observations, and folk claims about recurring human and institutional behaviour. What was missing was a neutral collective term for examining these things as members of the same broad class without first deciding whether they were scientific laws, useful heuristics, entertaining sayings, or nonsense.

Giving that class a name makes a second level of inquiry possible. We can now ask whether some kinds of empiricemes are more reliable than others, whether memorable wording causes weak empiricemes to spread more successfully than strong ones, whether similar empiricemes arise independently in different cultures, whether contradictory empiricemes actually describe different boundary conditions, and whether an observation attracts more testing after receiving a memorable name. We can also investigate how empiricemes change as they travel from specialist literature into popular culture.

None of those questions becomes valid merely because the word empiriceme has been coined. Indeed, the category itself must remain open to revision. Perhaps some things now collected together will eventually prove different enough to require separate classes, while other categories currently treated separately will turn out to belong inside it.

That is precisely the discipline demanded by Mill's Handle. Naming should precede judgement, not substitute for it. The purpose of the handle is to make the object easier to examine from every side, including the possibility that closer examination will reveal that we have attached the handle badly.

XIIFrom Anecdote to Testable Claim

Mill's Handle also suggests a useful sequence by which informal observations can mature into knowledge. Someone first notices a pattern, perhaps repeatedly but unsystematically. The pattern is then described, named, compared with other cases, given clearer boundaries, translated into testable propositions, subjected to evidence, and revised in response to failures and exceptions.

Naming sits at a particularly consequential point in that sequence. It is late enough that somebody has already perceived a possible regularity but early enough that the regularity may still be almost entirely untested. This explains why named claims occupy such an ambiguous epistemic position: they are developed enough to travel but not necessarily developed enough to deserve belief.

An empiriceme ledger should make that distinction visible. The catalogue records that a recurring claim exists, where it came from, how it has been formulated, and what phenomena people believe it describes. A separate evidentiary judgement asks whether the claim survives examination.

This distinction is easy to state and difficult to maintain because human beings naturally infer solidity from names. Once a proposition has a title, a history, examples, references, and its own page, it begins to look respectable. Empiricalism therefore needs to resist a bias that its own architecture will inevitably create: inclusion in the catalogue must never be confused with endorsement by the catalogue.

XIIIPredictions from Mill's Handle

A useful empiriceme should expose itself to possible testing, and Mill's Handle generates several predictions. Meaningful labels should under some conditions improve learning or recall of categories compared with equivalent unlabelled material; Lupyan's experiments already provide evidence in this direction. Stable terminology should also make literature concerning a phenomenon easier to retrieve and integrate, while competing terminology should increase the risk that relevant evidence remains divided among separate research traditions.

The principle also predicts less desirable effects. Memorable names may increase the spread of propositions independently of their evidentiary strength, established labels may encourage people to perceive borderline cases as instances of the named phenomenon, and authoritative-sounding terminology may increase confidence that a poorly established construct is real. In human affairs, widespread knowledge of a classification may additionally change the behaviour being classified, producing the sort of looping effects described by Hacking.

Some of these extensions have stronger evidence behind them than others, and they should not be smuggled into the core empiriceme merely because they seem plausible. They are predictions and subsidiary propositions generated by Mill's Handle, which means they can themselves become candidates for investigation. This is how a useful concept should behave: not by ending inquiry, but by producing better questions.

XIVThe Necessary Paradox

Empiricalism therefore rests on a paradox that cannot be eliminated. We need names because unnamed observations are difficult to collect, retrieve, compare, and debate, yet the act of naming encourages us to treat uncertain categories as though they were established things. Avoiding names would leave much human knowledge scattered; trusting names would turn a catalogue of propositions into a catalogue of reified abstractions.

The solution is not to choose between naming and scepticism. It is to institutionalise both. Give the recurring claim a handle strong enough that people can grasp it, carry it into other disciplines, search for it in history, attach evidence to it, argue about its boundaries, and remember where the argument stands. Then refuse to grant the handle any evidentiary weight of its own.

That separation may be one of the foundational rules of empiricalism. A name establishes an object of inquiry, not an object of belief. The empiriceme exists as a proposition because people have formulated and repeated it; whether the phenomenon it proposes exists in the world is a different question.

The name gives the argument somewhere to gather. The evidence decides what remains when it does. That is Mill's Handle.

SourcesSources and further reading

Mill’s Handle is catalogued in the ledger like any other empiriceme, and carries the same caveat it argues for. For what empiricalism is, read What is empiricalism?, or open the ledger.