What changes whether people privately follow cultural practices?

HERETICAL CANDIDATE — Reward can preserve public culture by releasing its private competitors. Repeated public production of A suppresses access to competing B-like action plans; contingent approval then broadens retrieval or releases that suppression, restoring the alternatives rather than merely strengthening A. Thus a reward policy can increase public A expression while reducing subsequent private A adoption, even at a fixed A delivery sequence and unchanged A competence. The feedback connection responsible for private change is expressive-success feedback -> availability of competing actions -> private choice, not engagement -> ranker -> more A. Recommender reset alone cannot erase this recipient state. A local candidate model is c_B(t+1)=c_B(t)-alpha*r_A(t)+gamma*f_A(t)*r_A(t)+eta(t), with c_B bounded competitor accessibility, r_A a standardized A retrieval/production event, f_A contingent positive feedback, alpha ordinary retrieval competition, gamma reward-contingent release and eta nonspecific recovery. A separate q_A reward value can increase public expression. Private log-odds A versus B depend on beta_Q*(q_A-q_B)-beta_C*c_B, with beta_C>0. The nontrivial claim is gamma>0 for social feedback with a causal effect on later private choice after controlling the independently manipulated competitor repertoire; neither positive target valuation nor delivery persistence is sufficient. This stabilizes an outcome-specific SPV_5 contrast and predicts a distinct SPV_6 trajectory. It is a conjecture about competition among remembered cultural options, not an assertion that positive reward universally reduces memory or that adoption does not exist.

Estimate C=[P(A private|contingent)-P(A private|yoked)] with versus without a competing repertoire. The strong version predicts C_competitors<-delta_A while the corresponding public-expression contrast is positive; C_no_competitors is within epsilon_A of zero or positive. Competitor reinstatement abolishes the private contingent-versus-yoked difference by restoring B in the suppressed arm, without reversing the public A reward-value difference or changing independently tested A rule competence. A symmetric A-only refresher does not reproduce the B-specific rescue. Ranker reset has no extra effect under future replay; it changes exposure in the free-running arm. IH_02 instead requires contact-geometry intervention; IH_03 requires tool-cost intervention; IH_04 requires changes in payoff risk. FAMILY-FALSIFYING RESULT: after verified reward contingency, competitor competition and retrieval first stages, private choice is explained within the informative margin by target value, ordinary retrieval-practice/priming and measured B accessibility with no reward-specific release term, or B-specific reinstatement has no adoption effect despite clear restoration of B accessibility. Retain ordinary reward learning/query theory and remove the distinct reward-release cultural family. A private-choice null without successful competitor manipulation is inconclusive.

Compare compact compatible patches with fragmented patches matched on global prevalence and, where feasible, each recipient's immediate A-neighbor count, but differing in boundary length and partner-of-partner structure. In the free-contact arm, the fitted adhesion model must predict the subsequent direction and speed of boundary rearrangement and the resulting confidential A-use trajectory in held-out communities; do not infer adoption from movement of people. After reset/withdrawal, disabling additional sorting versus restoring a saved contact map yields different A abandonment despite identical feed replay. A compatibility adapter that makes A/B collaboration equally effective should remove the topology-dependent adoption contrast while leaving contact identity, content and potential action opportunities present. Neither competitor reinstatement, personal tool replacement nor outcome-variance insurance alone should remove it when deterministic compatibility costs are held fixed. FAMILY-FALSIFYING RESULT: ordinary homophily plus a standard coordination-game model, independently fitted to the same pairwise choices and costs, predicts all held-out reset, remix and adapter trajectories within the prespecified tolerance and the morphogenetic geometric constraint adds no testable separation. Then classify the result as established network coordination/niche construction, not a distinct morphogenetic memetics family. Equivalently precise absence of geometry-dependent private use after a successful topology first stage rejects this candidate in the declared scope.

Jointly useful production aids may sustain private cultural adoption after rewards end predicts instead: In acquisition, independently permit construction of reusable A aids versus equally costly non-reusable preparation, without changing assigned source content, expressive reward totals, total work time or tool visibility. At withdrawal randomize retaining the aid, replacing it with a visually matched inert version, or giving a functionally equivalent neutral aid that makes A and B equally cheap; preregister the private use outcome. IH_03 predicts an adoption-retention contrast following measured execution-cost savings, and a restore-the-original-aid rescue without further social exposure or reward. Transfer a working configuration to a separately trained recipient under neutral ownership labels: savings and adoption should travel with the usable complement rather than with its author's public-feedback history. To test the proposed extension rather than generic switching costs, cross retention of k_1 and k_2 and predict private adoption from externally timed joint versus additive execution costs using held-out task instances. Competitor retrieval, graph remix with compatibility guaranteed, and variance insurance have no independent effect on this pure model once their impacts on tool usefulness and costs are blocked. FAMILY-FALSIFYING RESULT: a standard cost-sensitive choice model plus additive setup savings, independently calibrated before the cultural test, predicts the entire contingent-feedback/asset-reset trajectory within the prespecified tolerance; joint configuration supplies no extra functional dependence. Remove the distinct cultural-complementarity family and report ordinary switching-cost/niche effects. If verified removal of actual savings leaves adoption unchanged within epsilon_A, reject capital as the dominant persistence mechanism even if people continue to like or recognize the aid.

Diversifying practical risks may sustain private cultural adoption after public sharing fades predicts instead: Keep A content, displayed feedback, target outcome sequence, expected payoff, marginal target variance, task effort and contact graph fixed. Randomize the coupling of a participant's already familiar B outcomes with those A outcomes, creating low versus high A/B covariance while preserving each marginal history. Do this as an assigned risk environment, not a post hoc subgroup defined by successful outcomes. With verified risk comprehension and r>0, IH_04 predicts greater private allocation to A when A diversifies B, even if A receives the same or lower expressive approval. Rerandomize a transparent loss-insurance arrangement versus an actuarially matched fixed transfer; the covariance-by-private-adoption contrast should shrink toward zero under insurance while public reproduction can remain reward-sensitive. Ranker reset with replay cannot remove the diversification effect. IH_02 predicts topology/compatibility sensitivity with deterministic returns; IH_03 predicts executable-aid sensitivity after risks are insured; IH_01 predicts B-retrieval specificity after actual outcome risk is held constant. FAMILY-FALSIFYING RESULT: ordinary expected-utility/portfolio choice with independently measured risk sensitivity and costs predicts the entire acquisition/reset/insurance trajectory within informative error bands; there is no residual dependence on the proposed culture-to-risk feedback edge after those actual risks are accounted for. Then retain risk-sensitive adoption as an established economic composition, not a distinct memetic family. Alternatively, a precise absence of a covariance-by-insurance response after confirmed risk first stages rejects this mechanism as the dominant explanation. A positive insurance main effect alone is not evidence for covariance-mediated cultural persistence.

CROSS-DOMAIN TRANSFER — Endogenous contact sorting deposits a persistent compatibility niche. Reproduction of A changes which practical collaborators retain contact; their local contacts then reorganize into compatible patches. The durable substrate is this partner geometry, not a higher A value in each person and not the ranker's historical engagement table. A privately chosen convention remains useful because that geometry changes the actual compatibility cost of future work. Breaking the expression -> partner-retention/sorting edge during acquisition, or remixing contacts after acquisition while preserving feed histories, changes private adoption. Merely deleting the ranker state or expressive likes leaves the already formed niche. The morphogenesis import is differential adhesion: local compatibility costs and constrained contact territories can generate sorting and patch coarsening. The extra, falsifiable restriction is that independently estimated pairwise adhesion and contact-mobility parameters predict adoption differences between geometries with the same global A prevalence and contact budget; a density-only reinforcement model does not. The predicted SPV_5 and SPV_6 effects are limited to conventions whose private use has real partner-dependent functionality. For intrinsically solo practices this mechanism predicts no effect, not a universal hidden social force.

Compare compact compatible patches with fragmented patches matched on global prevalence and, where feasible, each recipient's immediate A-neighbor count, but differing in boundary length and partner-of-partner structure. In the free-contact arm, the fitted adhesion model must predict the subsequent direction and speed of boundary rearrangement and the resulting confidential A-use trajectory in held-out communities; do not infer adoption from movement of people. After reset/withdrawal, disabling additional sorting versus restoring a saved contact map yields different A abandonment despite identical feed replay. A compatibility adapter that makes A/B collaboration equally effective should remove the topology-dependent adoption contrast while leaving contact identity, content and potential action opportunities present. Neither competitor reinstatement, personal tool replacement nor outcome-variance insurance alone should remove it when deterministic compatibility costs are held fixed. FAMILY-FALSIFYING RESULT: ordinary homophily plus a standard coordination-game model, independently fitted to the same pairwise choices and costs, predicts all held-out reset, remix and adapter trajectories within the prespecified tolerance and the morphogenetic geometric constraint adds no testable separation. Then classify the result as established network coordination/niche construction, not a distinct morphogenetic memetics family. Equivalently precise absence of geometry-dependent private use after a successful topology first stage rejects this candidate in the declared scope.

Estimate C=[P(A private|contingent)-P(A private|yoked)] with versus without a competing repertoire. The strong version predicts C_competitors<-delta_A while the corresponding public-expression contrast is positive; C_no_competitors is within epsilon_A of zero or positive. Competitor reinstatement abolishes the private contingent-versus-yoked difference by restoring B in the suppressed arm, without reversing the public A reward-value difference or changing independently tested A rule competence. A symmetric A-only refresher does not reproduce the B-specific rescue. Ranker reset has no extra effect under future replay; it changes exposure in the free-running arm. IH_02 instead requires contact-geometry intervention; IH_03 requires tool-cost intervention; IH_04 requires changes in payoff risk. FAMILY-FALSIFYING RESULT: after verified reward contingency, competitor competition and retrieval first stages, private choice is explained within the informative margin by target value, ordinary retrieval-practice/priming and measured B accessibility with no reward-specific release term, or B-specific reinstatement has no adoption effect despite clear restoration of B accessibility. Retain ordinary reward learning/query theory and remove the distinct reward-release cultural family. A private-choice null without successful competitor manipulation is inconclusive.

Jointly useful production aids may sustain private cultural adoption after rewards end predicts instead: In acquisition, independently permit construction of reusable A aids versus equally costly non-reusable preparation, without changing assigned source content, expressive reward totals, total work time or tool visibility. At withdrawal randomize retaining the aid, replacing it with a visually matched inert version, or giving a functionally equivalent neutral aid that makes A and B equally cheap; preregister the private use outcome. IH_03 predicts an adoption-retention contrast following measured execution-cost savings, and a restore-the-original-aid rescue without further social exposure or reward. Transfer a working configuration to a separately trained recipient under neutral ownership labels: savings and adoption should travel with the usable complement rather than with its author's public-feedback history. To test the proposed extension rather than generic switching costs, cross retention of k_1 and k_2 and predict private adoption from externally timed joint versus additive execution costs using held-out task instances. Competitor retrieval, graph remix with compatibility guaranteed, and variance insurance have no independent effect on this pure model once their impacts on tool usefulness and costs are blocked. FAMILY-FALSIFYING RESULT: a standard cost-sensitive choice model plus additive setup savings, independently calibrated before the cultural test, predicts the entire contingent-feedback/asset-reset trajectory within the prespecified tolerance; joint configuration supplies no extra functional dependence. Remove the distinct cultural-complementarity family and report ordinary switching-cost/niche effects. If verified removal of actual savings leaves adoption unchanged within epsilon_A, reject capital as the dominant persistence mechanism even if people continue to like or recognize the aid.

Diversifying practical risks may sustain private cultural adoption after public sharing fades predicts instead: Keep A content, displayed feedback, target outcome sequence, expected payoff, marginal target variance, task effort and contact graph fixed. Randomize the coupling of a participant's already familiar B outcomes with those A outcomes, creating low versus high A/B covariance while preserving each marginal history. Do this as an assigned risk environment, not a post hoc subgroup defined by successful outcomes. With verified risk comprehension and r>0, IH_04 predicts greater private allocation to A when A diversifies B, even if A receives the same or lower expressive approval. Rerandomize a transparent loss-insurance arrangement versus an actuarially matched fixed transfer; the covariance-by-private-adoption contrast should shrink toward zero under insurance while public reproduction can remain reward-sensitive. Ranker reset with replay cannot remove the diversification effect. IH_02 predicts topology/compatibility sensitivity with deterministic returns; IH_03 predicts executable-aid sensitivity after risks are insured; IH_01 predicts B-retrieval specificity after actual outcome risk is held constant. FAMILY-FALSIFYING RESULT: ordinary expected-utility/portfolio choice with independently measured risk sensitivity and costs predicts the entire acquisition/reset/insurance trajectory within informative error bands; there is no residual dependence on the proposed culture-to-risk feedback edge after those actual risks are accounted for. Then retain risk-sensitive adoption as an established economic composition, not a distinct memetic family. Alternatively, a precise absence of a covariance-by-insurance response after confirmed risk first stages rejects this mechanism as the dominant explanation. A positive insurance main effect alone is not evidence for covariance-mediated cultural persistence.

SCOUT 1 — Human factors and production ergonomics: rewarded reproduction induces durable investments that change the private cost of enactment. Participants prepare reusable templates, shortcuts or physical/digital work arrangements to produce A cheaply; these complements survive the public reward and ranker reset and make later confidential A use cheaper than B. The memory is an executable production configuration, not an intention reminder, a recalled motor plan, an ancestry cue or an ownership right. The operative feedback edge is expression demand -> participant-built complement -> later enactment cost -> adoption. Let k_A be usable A-specific complementary capital, b_A a construction event and delta_k depreciation: k_A(t+1)=(1-delta_k)k_A(t)+b_A(t). Let c_A(k_A) be independently timed private execution cost, with dc_A/dk_A<0. A distinctive candidate extension is that a reward-contingent production history creates cross-step complements whose savings depend on their joint configuration, c_A(k_1,k_2) with a nonzero interaction beyond additive savings. A ranker can induce this configuration without directly reinforcing private preference. SPV_6 remains low while this complement functions even under fixed delivery; resetting the complement rather than the ranker increases abandonment. All options stay executable without the complement, so increased cost is not coded as absence of opportunity.

At withdrawal randomize retaining the aid, replacing it with a visually matched inert version, or giving a functionally equivalent neutral aid that makes A and B equally cheap; preregister the private use outcome. IH_03 predicts an adoption-retention contrast following measured execution-cost savings, and a restore-the-original-aid rescue without further social exposure or reward. Transfer a working configuration to a separately trained recipient under neutral ownership labels: savings and adoption should travel with the usable complement rather than with its author's public-feedback history. To test the proposed extension rather than generic switching costs, cross retention of k_1 and k_2 and predict private adoption from externally timed joint versus additive execution costs using held-out task instances. Competitor retrieval, graph remix with compatibility guaranteed, and variance insurance have no independent effect on this pure model once their impacts on tool usefulness and costs are blocked. FAMILY-FALSIFYING RESULT: a standard cost-sensitive choice model plus additive setup savings, independently calibrated before the cultural test, predicts the entire contingent-feedback/asset-reset trajectory within the prespecified tolerance; joint configuration supplies no extra functional dependence. Remove the distinct cultural-complementarity family and report ordinary switching-cost/niche effects. If verified removal of actual savings leaves adoption unchanged within epsilon_A, reject capital as the dominant persistence mechanism even if people continue to like or recognize the aid.

Estimate C=[P(A private|contingent)-P(A private|yoked)] with versus without a competing repertoire. The strong version predicts C_competitors<-delta_A while the corresponding public-expression contrast is positive; C_no_competitors is within epsilon_A of zero or positive. Competitor reinstatement abolishes the private contingent-versus-yoked difference by restoring B in the suppressed arm, without reversing the public A reward-value difference or changing independently tested A rule competence. A symmetric A-only refresher does not reproduce the B-specific rescue. Ranker reset has no extra effect under future replay; it changes exposure in the free-running arm. IH_02 instead requires contact-geometry intervention; IH_03 requires tool-cost intervention; IH_04 requires changes in payoff risk. FAMILY-FALSIFYING RESULT: after verified reward contingency, competitor competition and retrieval first stages, private choice is explained within the informative margin by target value, ordinary retrieval-practice/priming and measured B accessibility with no reward-specific release term, or B-specific reinstatement has no adoption effect despite clear restoration of B accessibility. Retain ordinary reward learning/query theory and remove the distinct reward-release cultural family. A private-choice null without successful competitor manipulation is inconclusive.

Contact sorting may sustain private cultural adoption by lowering compatibility costs predicts instead: Hold participant composition, assigned feed sequence, expected number of interactions and every person's degree fixed; randomize a small set of allowed degree-preserving partner swaps. Compare compact compatible patches with fragmented patches matched on global prevalence and, where feasible, each recipient's immediate A-neighbor count, but differing in boundary length and partner-of-partner structure. In the free-contact arm, the fitted adhesion model must predict the subsequent direction and speed of boundary rearrangement and the resulting confidential A-use trajectory in held-out communities; do not infer adoption from movement of people. After reset/withdrawal, disabling additional sorting versus restoring a saved contact map yields different A abandonment despite identical feed replay. A compatibility adapter that makes A/B collaboration equally effective should remove the topology-dependent adoption contrast while leaving contact identity, content and potential action opportunities present. Neither competitor reinstatement, personal tool replacement nor outcome-variance insurance alone should remove it when deterministic compatibility costs are held fixed. FAMILY-FALSIFYING RESULT: ordinary homophily plus a standard coordination-game model, independently fitted to the same pairwise choices and costs, predicts all held-out reset, remix and adapter trajectories within the prespecified tolerance and the morphogenetic geometric constraint adds no testable separation. Then classify the result as established network coordination/niche construction, not a distinct morphogenetic memetics family. Equivalently precise absence of geometry-dependent private use after a successful topology first stage rejects this candidate in the declared scope.

Diversifying practical risks may sustain private cultural adoption after public sharing fades predicts instead: Keep A content, displayed feedback, target outcome sequence, expected payoff, marginal target variance, task effort and contact graph fixed. Randomize the coupling of a participant's already familiar B outcomes with those A outcomes, creating low versus high A/B covariance while preserving each marginal history. Do this as an assigned risk environment, not a post hoc subgroup defined by successful outcomes. With verified risk comprehension and r>0, IH_04 predicts greater private allocation to A when A diversifies B, even if A receives the same or lower expressive approval. Rerandomize a transparent loss-insurance arrangement versus an actuarially matched fixed transfer; the covariance-by-private-adoption contrast should shrink toward zero under insurance while public reproduction can remain reward-sensitive. Ranker reset with replay cannot remove the diversification effect. IH_02 predicts topology/compatibility sensitivity with deterministic returns; IH_03 predicts executable-aid sensitivity after risks are insured; IH_01 predicts B-retrieval specificity after actual outcome risk is held constant. FAMILY-FALSIFYING RESULT: ordinary expected-utility/portfolio choice with independently measured risk sensitivity and costs predicts the entire acquisition/reset/insurance trajectory within informative error bands; there is no residual dependence on the proposed culture-to-risk feedback edge after those actual risks are accounted for. Then retain risk-sensitive adoption as an established economic composition, not a distinct memetic family. Alternatively, a precise absence of a covariance-by-insurance response after confirmed risk first stages rejects this mechanism as the dominant explanation. A positive insurance main effect alone is not evidence for covariance-mediated cultural persistence.

SCOUT 2 — Actuarial risk pooling: recipients privately adopt a convention because it hedges the failure of their other practical routines. Adaptive delivery can bring otherwise rare complementary practices into a person's feasible repertoire. Public expressive rewards track popularity, but private practice choice depends on the covariance of future useful outcomes with an already held practice, not only the target's mean reward. The feedback connection is publicly induced participation -> shared practical failure exposure -> value of a diversifying convention -> private adoption. The relevant persistent state is the household/person's portfolio of genuinely usable routines and their joint outcome distribution. This differs from an inherited correlation code: no covariance between cultural representations is copied, and no relational semantic unit is stored. A narrow two-practice model values use fraction w in A and 1-w in B by U(w)=w*mu_A+(1-w)*mu_B-(r/2)[w^2*sigma_A^2+(1-w)^2*sigma_B^2+2*w*(1-w)*sigma_AB]-C(w). Here r is local risk sensitivity and C implementation cost. At matched marginal means/variances, changing sigma_AB changes private A uptake for people already relying on B. Thus resetting the ranker can stop public recurrence while the functional diversification reason for adoption persists. Conversely, insuring the relevant loss removes the reason to maintain A even if target liking, recall, delivery and social approval stay constant. The stabilised outcome is SPV_5 and the subsequent SPV_6 risk-sensitive retention curve.

Randomize the coupling of a participant's already familiar B outcomes with those A outcomes, creating low versus high A/B covariance while preserving each marginal history. Do this as an assigned risk environment, not a post hoc subgroup defined by successful outcomes. With verified risk comprehension and r>0, IH_04 predicts greater private allocation to A when A diversifies B, even if A receives the same or lower expressive approval. Rerandomize a transparent loss-insurance arrangement versus an actuarially matched fixed transfer; the covariance-by-private-adoption contrast should shrink toward zero under insurance while public reproduction can remain reward-sensitive. Ranker reset with replay cannot remove the diversification effect. IH_02 predicts topology/compatibility sensitivity with deterministic returns; IH_03 predicts executable-aid sensitivity after risks are insured; IH_01 predicts B-retrieval specificity after actual outcome risk is held constant. FAMILY-FALSIFYING RESULT: ordinary expected-utility/portfolio choice with independently measured risk sensitivity and costs predicts the entire acquisition/reset/insurance trajectory within informative error bands; there is no residual dependence on the proposed culture-to-risk feedback edge after those actual risks are accounted for. Then retain risk-sensitive adoption as an established economic composition, not a distinct memetic family. Alternatively, a precise absence of a covariance-by-insurance response after confirmed risk first stages rejects this mechanism as the dominant explanation. A positive insurance main effect alone is not evidence for covariance-mediated cultural persistence.

Estimate C=[P(A private|contingent)-P(A private|yoked)] with versus without a competing repertoire. The strong version predicts C_competitors<-delta_A while the corresponding public-expression contrast is positive; C_no_competitors is within epsilon_A of zero or positive. Competitor reinstatement abolishes the private contingent-versus-yoked difference by restoring B in the suppressed arm, without reversing the public A reward-value difference or changing independently tested A rule competence. A symmetric A-only refresher does not reproduce the B-specific rescue. Ranker reset has no extra effect under future replay; it changes exposure in the free-running arm. IH_02 instead requires contact-geometry intervention; IH_03 requires tool-cost intervention; IH_04 requires changes in payoff risk. FAMILY-FALSIFYING RESULT: after verified reward contingency, competitor competition and retrieval first stages, private choice is explained within the informative margin by target value, ordinary retrieval-practice/priming and measured B accessibility with no reward-specific release term, or B-specific reinstatement has no adoption effect despite clear restoration of B accessibility. Retain ordinary reward learning/query theory and remove the distinct reward-release cultural family. A private-choice null without successful competitor manipulation is inconclusive.

Contact sorting may sustain private cultural adoption by lowering compatibility costs predicts instead: Hold participant composition, assigned feed sequence, expected number of interactions and every person's degree fixed; randomize a small set of allowed degree-preserving partner swaps. Compare compact compatible patches with fragmented patches matched on global prevalence and, where feasible, each recipient's immediate A-neighbor count, but differing in boundary length and partner-of-partner structure. In the free-contact arm, the fitted adhesion model must predict the subsequent direction and speed of boundary rearrangement and the resulting confidential A-use trajectory in held-out communities; do not infer adoption from movement of people. After reset/withdrawal, disabling additional sorting versus restoring a saved contact map yields different A abandonment despite identical feed replay. A compatibility adapter that makes A/B collaboration equally effective should remove the topology-dependent adoption contrast while leaving contact identity, content and potential action opportunities present. Neither competitor reinstatement, personal tool replacement nor outcome-variance insurance alone should remove it when deterministic compatibility costs are held fixed. FAMILY-FALSIFYING RESULT: ordinary homophily plus a standard coordination-game model, independently fitted to the same pairwise choices and costs, predicts all held-out reset, remix and adapter trajectories within the prespecified tolerance and the morphogenetic geometric constraint adds no testable separation. Then classify the result as established network coordination/niche construction, not a distinct morphogenetic memetics family. Equivalently precise absence of geometry-dependent private use after a successful topology first stage rejects this candidate in the declared scope.

Jointly useful production aids may sustain private cultural adoption after rewards end predicts instead: In acquisition, independently permit construction of reusable A aids versus equally costly non-reusable preparation, without changing assigned source content, expressive reward totals, total work time or tool visibility. At withdrawal randomize retaining the aid, replacing it with a visually matched inert version, or giving a functionally equivalent neutral aid that makes A and B equally cheap; preregister the private use outcome. IH_03 predicts an adoption-retention contrast following measured execution-cost savings, and a restore-the-original-aid rescue without further social exposure or reward. Transfer a working configuration to a separately trained recipient under neutral ownership labels: savings and adoption should travel with the usable complement rather than with its author's public-feedback history. To test the proposed extension rather than generic switching costs, cross retention of k_1 and k_2 and predict private adoption from externally timed joint versus additive execution costs using held-out task instances. Competitor retrieval, graph remix with compatibility guaranteed, and variance insurance have no independent effect on this pure model once their impacts on tool usefulness and costs are blocked. FAMILY-FALSIFYING RESULT: a standard cost-sensitive choice model plus additive setup savings, independently calibrated before the cultural test, predicts the entire contingent-feedback/asset-reset trajectory within the prespecified tolerance; joint configuration supplies no extra functional dependence. Remove the distinct cultural-complementarity family and report ordinary switching-cost/niche effects. If verified removal of actual savings leaves adoption unchanged within epsilon_A, reject capital as the dominant persistence mechanism even if people continue to like or recognize the aid.