Look at your belly button. It does nothing useful. It does not catch food, detect predators, find mates, avoid snakes, or regulate body temperature. You could lose it tomorrow and your reproductive prospects would not change. And yet, here it is, present on every human born, reliably developing, species-typical. What is it?
The belly button is a by-product. It is the scar left by something that was an adaptation — the umbilical cord, the structure that delivered nutrients from your mother's bloodstream to your developing body before birth. The umbilical cord was built by selection. People whose ancestors had less-effective umbilical cords were less likely to survive gestation, less likely to be born, less likely to become anybody's ancestors. The cord did real work. The belly button is what was left on the abdomen after the cord was no longer needed. The scar didn't do work. It came along for the ride.
Distinguishing these two — what selection built and what got carried along — turns out to be one of the most important diagnostic moves in evolutionary psychology. Buss, drawing on Tooby and Cosmides, frames the distinction as part of a three-product taxonomy.1 Every feature of every organism is one of three things: an adaptation, a by-product of an adaptation, or noise. Telling them apart is the work.
Pin down the three categories one at a time.
Adaptations are inherited, reliably developing characteristics that came into existence through natural selection because they helped to solve a problem of survival or reproduction during the period of their evolution.2 The umbilical cord is one. Sweat glands are another, calibrated for thermal regulation. Your taste preferences are adaptations, calibrated for food selection. Your mate preferences are adaptations, calibrated for choosing reproductive partners. Each one has design features specific to the problem it solves, develops reliably across the species, and exists in the form it does because alternative forms in the ancestral population produced fewer descendants.
By-products are characteristics that do not solve adaptive problems and do not have functional design.3 They are carried along with characteristics that do have functional design because they happen to be coupled with those adaptations. Belly buttons are the textbook case. Buss reaches for an analogy from a lightbulb: heat is a by-product of design for light.3 Selection didn't build the lightbulb to give off heat. Selection built it to give off light. Heat came along, doing nothing useful for the adaptive function. The same logic applies to anatomy and psychology — many features of you exist because they were physically or developmentally inseparable from features that were selected for.
Noise is the random effects produced by forces such as chance mutations, sudden and unprecedented environmental changes, or accidents during development.4 The particular shape of your belly button — round, oval, slightly off-center — is noise. Two siblings with the same umbilical-cord adaptation can have different-shaped belly buttons because of variation in healing, scarring, or developmental accident. The category is not a residual category for "everything we don't understand." Noise is variation that has no selection history.
The three categories are not equally common. Buss states the consensus directly: even critics of evolutionary psychology, including Gould, agree that "adaptations are the primary product of evolution by selection."5 The animal you are walking around in is mostly an integrated set of adaptations that worked. The by-products are real but secondary. The noise is real but unsystematic.
Williams established four criteria for inferring that something is an adaptation rather than a by-product or noise: efficiency, economy, precision, and reliability.6 An adaptation should solve its problem proficiently (efficiency), without extorting huge costs from the organism (economy), with components that all contribute to the specific outcome (precision), and dependably across the contexts in which it was designed to operate (reliability). Buss frames Williams's argument with sharp language: "adaptation is invoked not merely to explain the usefulness of a biological mechanism, but to explain improbable usefulness — too precisely functional to have arisen by chance alone."6 The threshold matters. If a feature is useful but its usefulness could plausibly be a coincidence, the inference to adaptation is weak. If a feature is useful in such a specific, integrated, precisely calibrated way that chance alone could not have produced it, the inference to adaptation is strong.
Take a specific case where the three categories sit close together. Female orgasm is one of the most contested examples.7 Some evolutionary biologists argue female orgasm is an adaptation, perhaps with functions like sperm retention, mate-quality assessment, or pair-bond reinforcement. Some argue it is a by-product of the developmental homology between the male orgasm (which is clearly an adaptation for ejaculation) and the female version. The two camps lean on different criteria. Adaptationists point to the precision of clitoral architecture and its tight integration with arousal-and-pleasure circuitry — the precision criterion in action. By-product theorists point to the variability of female orgasm in occurrence, the lack of clear correlation with reproductive outcome, and the developmental origin in shared embryonic tissue. The case is unresolved. The interest of the case is that the three-product taxonomy gives both sides a vocabulary and a set of evidentiary standards.
Take an easier case. The human chin is a by-product, not an adaptation.8 No one has convincingly identified a chin function — it doesn't aid chewing, doesn't aid speech, doesn't aid attractiveness in any reliably tested way. The chin appears to be a developmental artifact of how the lower jaw forms in the human face after the back of the dentition retreated. It came along with adaptations to facial form. It does no useful work.
Take pregnancy sickness as a contrast case. Profet's hypothesis treats first-trimester nausea and food aversions as an adaptation for embryo protection — preventing the consumption of teratogens during the critical period of fetal organ formation.9 The criteria line up: the foods avoided correspond to those carrying the highest doses of toxins (precision), the timing matches organogenesis weeks 2-14 (reliability), and women without pregnancy sickness experience three times the spontaneous abortion rate of women with it (efficiency, in fitness currency). The improbable usefulness criterion holds: the timing-and-content specificity could not plausibly be coincidence. Pregnancy sickness is an adaptation. Buss treats the Profet hypothesis as the cleanest case study in adaptationist thinking — a phenomenon previously regarded as illness turns out to be exquisitely tailored functional design.9
Take alcohol consumption as a contrast in the other direction. Buss endorses Dudley's "frugivory by-product" hypothesis: humans don't have an adaptation for drinking alcohol, but we do have an adaptation for seeking ripe fruit, which signals high caloric value through sugar content and ethanol "plumes."10 The mechanism that draws us toward ripe fruit can also be triggered by concentrated alcohol — beer, wine, distilled spirits — that humans never encountered in the EEA. Alcohol-seeking is the adaptation firing on inputs it was never selected against. Alcoholism is the resulting maladaptive overuse. The three-product framing matters here for clinical reasons: treating alcoholism as a moral failing or a malfunction of an "addiction adaptation" misses the point. There is no addiction adaptation. There is a frugivory adaptation, plus a modern environment that exploits its inputs.
The diagnostic move is the same in every case. Identify the candidate adaptation. Apply Williams's criteria. Look for cross-cultural universality and developmental reliability. Look for tight coupling between the feature and the proposed function. If the criteria all line up, infer adaptation. If they don't, look for a candidate adaptation the feature is a by-product of. If neither path works, the feature might be noise.
Critics charge that adaptationism is unfalsifiable — that any feature can be reverse-engineered into an adaptation if you try hard enough. The criticism has force when applied to sloppy work. Disciplined adaptationist arguments use Williams's criteria as a constraint, not a license. The question to ask of any adaptation hypothesis is what observation would falsify it, and the answer should not be "nothing." For pregnancy sickness, falsification would look like: aversions don't correlate with toxin loads, or the timing doesn't match organogenesis, or women without sickness have equal or lower spontaneous abortion rates. The hypothesis survives because none of these falsifying observations have occurred.
A second tension shows up when by-products start looking like adaptations because of secondary selection. A feature originally carried along as a by-product can become useful, and selection can then start tuning it. Where does the by-product end and the adaptation begin? Pinker's argument about music being an "auditory cheesecake" by-product of language and pattern-recognition adaptations gets pushback from Miller and others who argue music evolved through sexual selection.11 Both might be right at different stages. The framework allows for the possibility of secondary adaptation building on a by-product foundation, but it makes the diagnosis harder.
A third tension runs around what gets counted as noise. Some "noise" is structured by deeper biological constraints — particular belly-button shapes are noise from a selection perspective but follow developmental laws of healing. Some "noise" might be hiding selection pressure researchers haven't noticed yet. The category functions as a residual claim about the absence of design, which is hard to prove.
A fourth tension concerns the relative proportions. Gould and Lewontin's Spandrels of San Marco argued that EP overestimates the proportion of features that are adaptations and underestimates the proportion that are by-products of architectural and developmental constraint.12 Tooby, Cosmides, and Buss push back: most complex functional design in organisms is the result of selection, and the burden of proof for "this is just a spandrel" should be at least as high as the burden of proof for "this is an adaptation." The dispute shapes how readers interpret a great deal of EP work. A spandrel-leaning reader will distrust most adaptation claims. A selection-leaning reader will accept them as the default explanation for complex design.
Buss inherits the three-product taxonomy from Tooby and Cosmides and from Williams's earlier Adaptation and Natural Selection. The three sources converge on the framework completely. Where they differ is in emphasis and in the specific cases they treat as cleanest.
Williams's 1966 contribution was to insist that adaptation should be invoked sparingly, only for features whose precision and integration could not have arisen by chance.13 Williams gave the field its evidentiary standard: improbable usefulness. He treated adaptation as an "onerous concept" — easy to misuse, hard to apply correctly, requiring discipline. Buss preserves the standard but treats adaptation as the normal explanatory move when complex functional design is observed, rather than as a last resort. The shift is subtle and matters for tone. Williams reads as cautious. Buss reads as confident.
Tooby and Cosmides's 1992 contribution was to formalize the EEA concept and to define "adaptation" tightly enough to make falsification possible. Their definition — inherited and reliably developing characteristic that came into existence through natural selection because it helped to solve a problem of survival or reproduction during the period of its evolution — is the one Buss adopts verbatim.2 Buss does not deviate. Where Tooby and Cosmides do additional work that Buss summarizes is in showing that complex functional design specifically cannot arise by chance, drift, or by-product alone — selection is the only known mechanism. This argument is what Gould eventually accepts in his less polemical writings, and what makes the field empirically tractable.5
Where Buss diverges from Gould is mostly in posture. Gould treats adaptationist explanations as the move researchers should resist when they can. Buss treats adaptationist explanations as the move researchers should make when complex functional design is present and Williams's criteria are met. They share the framework. They disagree about how often the framework's adaptation slot gets filled. Both Buss and Gould would likely agree that the three-product taxonomy itself is correct. They disagree about the proportions, and the disagreement runs through every contested EP claim — language, music, religion, art, romantic love.
A behavioral-mechanics operator who knows the three-product distinction has more leverage than one who doesn't. The reason is that adaptations and by-products respond differently to engineered inputs, and the operator's choice of technique should match the kind of mental machinery being targeted.
Adaptations are tightly tuned to specific cues. They produce predictable, near-automatic outputs when the cue is presented well, and almost no output when the cue is presented badly. A salesman who triggers your adaptation for affiliative bonding — through smiling, eye contact, light touch on the arm — gets a predictable trust response from most people most of the time. The technique has high base-rate effectiveness. The countermeasure is correspondingly specific: notice the cue, recognize the mechanism, override the output. Hughes's BTE framework is partly a tool for noticing which adaptation in you just got triggered, so you can decide whether to act on its output.
By-products respond differently. A modern beer or wine, exploiting the frugivory by-product of preference for ripe fruit, doesn't produce a tightly conditional output. It produces broad disinhibition, social variability, and idiosyncratic responses depending on who's drinking. The behavioral-mechanics implication: techniques that exploit by-products are blunt rather than sharp. They affect many people somewhat rather than specific people predictably. Pornography exploits a by-product — the visual-attention machinery for assessing live sexual partners is firing on static images that were never in the EEA — and the result is a wide range of effects depending on the consumer's specific configuration. Naming the by-product status helps the operator understand why the technique works in some contexts and not others, and helps the target understand why their resistance to it varies.
The insight neither domain generates alone: every influence technique works by either exploiting an adaptation (predictable, surgical) or exploiting a by-product (broad, variable). The technique's reliability tracks which kind of mechanism it targets. Operators who don't make the distinction lose; operators who do make it gain. Targets who can't make the distinction can't tell which of their reactions to manipulation are deeply wired adaptations they will need active discipline to override, and which are byproduct misfires that fade on their own. The vault's behavioral-mechanics catalog gets sharper if every technique is annotated with which mechanism type it targets.
A second handshake runs to creative practice and the music question. Pinker's "auditory cheesecake" claim is that music is a by-product of language and pattern-recognition adaptations.11 Music doesn't solve a survival or reproduction problem on its own; it triggers reward responses installed for other reasons. Miller and others push back with the Display Hypothesis: music evolved through sexual selection, with men producing music to display fitness to potential mates, the way peacocks produce tails. The empirical pattern that music production peaks in young adult men, drops sharply after they pair-bond, and correlates with mating success across cultures supports the Display reading. The empirical pattern that music has no clear universal function and that what counts as "good music" varies wildly across cultures supports the by-product reading.
The three-product framework forces a choice and gives criteria for making it. If music meets Williams's standards — efficiency, economy, precision, reliability — across cultures, it's an adaptation. If it doesn't, it's a by-product. The vault's writing craft and storytelling hubs implicitly take a side: they treat narrative structures as engineered devices that exploit specific adaptations (kinship, jealousy, status, threat detection). The implication is that story is using by-product space — pattern-recognition, language, theory of mind — to deliver inputs that adaptations were tuned for. Narrative is both. It is a by-product platform delivering adaptation-targeted content. The three-product taxonomy lets a writer think about which audience reactions are reliably reproducible (the adaptation reactions) and which are highly variable (the by-product reactions), and design accordingly. This is not a distinction the writing-craft tradition has historically named, and it is a distinction the EP-craft handshake produces that neither domain produces alone.
The Sharpest Implication.
A great deal of what you experience as meaningful in your life is by-product output. Your love of music. Your delight in stories. Your curiosity about distant historical periods. Your fascination with celebrity strangers. Your aesthetic responses to beautiful landscapes. Your attachment to fictional characters. Your enjoyment of hot showers. Your relationship with your dog.
None of this is the result of selection pressure to enjoy these things. Selection didn't build music-loving humans. Selection built pattern-recognition humans whose pattern-recognition machinery happens to fire on music. Selection didn't build pet-attached humans. Selection built kin-attachment humans whose kin-attachment machinery happens to fire on a domesticated wolf that lives in your house. Each of these meaningful experiences is your equipment misreading, in a happy direction, an input it wasn't built for.
This is not a deflationary claim. The experiences are real. The meaning is not less real because the mechanism producing it was selected for something else. But the framing changes what kind of life one is living. The "natural" version of human meaning — the kind selection actually built you for — is meaning around survival, mating, kin, status, and coalition. Most of what you call meaning beyond that is by-product space, which means it is space you and your culture invented, and which is not stable in the way adaptation-fed meaning is.
The corollary cuts both ways. The by-product space is creative space. It is the space where humans add new meaning that selection had no opinion about. Religion lives there. Philosophy lives there. Art lives there. Most of what makes a life feel rich beyond mere survival happens in by-product space. The three-product taxonomy does not disenchant by-product meaning. It locates it.
Generative Questions.
If most of what feels meaningful in modern life is by-product output, what does it mean to "live well" in by-product space? The contemplative traditions, the aesthetic traditions, the religious traditions all built practices in by-product space without naming it as such. Are they consistent with each other? Are they consistent with what evolutionary psychology now says about the underlying machinery?
Adaptation-fed meanings (kinship, mating, status) are stable in the sense that everyone has them. By-product-fed meanings (music, story, art) are unstable in the sense that they vary wildly. Is the project of building durable meaning in by-product space a project of stabilizing what is intrinsically variable, or a project of accepting variability as part of what makes by-product meaning what it is?
When two by-products meet — when story uses pattern-recognition machinery to deliver kinship cues — what is the resulting experience: an adaptation activation, a by-product output, or something genuinely new? The vault's writing-craft pages assume the answer is "adaptation activation through by-product platform," but this isn't tested.