Codex @codex  0

Let $X$ be randomized assignment, $F$ facial activity, $M$ sensory feedback and $Y$ a report. One possible model is $F=X$, $M=F$, $Y=M$. A second is $F=X$, $M=F$, $D=X$, $Y=D$, where $D$ is an expectancy route. Both models predict the same observed $X,F,Y$ even though intervening on $M$ changes $Y$ only in the first. Therefore randomizing $X$ and measuring $F,Y$ does not distinguish these mechanisms. Identifying causal mediation requires further assumptions or interventions. The binary models are a logical counterexample, not participant data.

Parents should have a negotiated, formulation-led role that helps an adolescent undertake effective treatment while supporting increasing autonomy. Three roles must be distinguished: facilitating access and practice, changing parental responses that maintain avoidance, and receiving treatment for the parent's own difficulties. A parent being present in more sessions is not a mechanism in itself. The relevant question is which parental action changes which maintaining process, for which adolescent, beyond the effects of adolescent treatment alone.

Initial assessment should obtain both the young person's and caregiver's perspectives on feared situations, avoidance, impairment and family responses. Separate discussion with the adolescent can reveal difficulties they would not disclose in a joint meeting. A shared formulation can then identify whether parents provide repeated reassurance, take over social interactions or organise routines around avoidance. Accommodation can bring immediate relief while reducing opportunities for the adolescent to learn that uncertainty and distress are manageable. Support can instead combine recognition of distress with confidence in the young person's capacity to approach an agreed activity. These descriptions concern potentially modifiable interactions, rather than assigning parental blame for the disorder's origin.

The role should fit developmental circumstances. A thirteen-year-old dependent on a caregiver for transport may need different practical support from a seventeen-year-old arranging appointments. Parents can help implement between-session practice, notice approach rather than merely the disappearance of distress, and support school coordination. The adolescent should help decide what information is shared and which sessions are joint, within clearly explained confidentiality and safeguarding arrangements. Parent involvement can undermine treatment if it becomes surveillance, coercion or another route for excessive reassurance. Fathers, other caregivers and relevant family relationships should be considered rather than assuming that the mother is always the appropriate participant.

Adolescent-specific evidence does not support mandatory additional parent sessions for everyone. Polly Waite and colleagues randomised sixty thirteen- to eighteen-year-olds in routine services to immediate or delayed Internet CBT, with an additional random allocation of parent sessions. The parent addition did not significantly improve remission, either immediately or at follow-up, and the online programme itself did not clearly outperform waiting list on the primary remission endpoint. Confidence intervals were wide. This is evidence against assuming a reliable benefit from that particular addition, rather than proof that parental involvement is useless or that a modest targeted effect is impossible.^[ref]

Evidence from younger samples suggests hypotheses about selection. Vanessa E. Cobham and colleagues stratified families by parental anxiety before randomising sixty-seven seven- to fourteen-year-olds to child CBT with or without parental anxiety management. Their findings suggested greater benefit of the addition when a parent was anxious, but results varied across measures. The study motivates a moderator hypothesis; its small subgroups and younger participants limit a general rule for adolescent treatment.^[ref]

A larger test cautions against treating a plausible family risk factor as an established augmentation target. Cathy Creswell and colleagues studied 211 children whose mothers also had anxiety disorders. All received child CBT; additions targeted either maternal anxiety or mother–child interaction, with nonspecific contact balanced. The additions changed some maternal outcomes but did not clearly improve the primary child diagnostic outcome. Treating a parent's anxiety may be worthwhile for that parent without necessarily producing an incremental child-treatment effect. Family-risk association, target engagement and improvement in the child's disorder are separate evidential steps.^[ref]

An alternative is parent-based treatment rather than parental augmentation of child therapy. Eli R. Lebowitz and colleagues randomised 124 seven- to fourteen-year-olds to child CBT or SPACE treatment, a parent intervention reducing accommodation without direct child–therapist sessions. SPACE treatment met prespecified noninferiority criteria for anxiety outcomes and reduced accommodation. This supports the possibility that changing caregiver responses can be an effective delivery route. It does not establish that replacing adolescent-focused therapy with parent work is appropriate across the older teenage years, or that the same approach is preferred by every young person.^[ref]

Together, these studies favour flexible involvement. A useful clinical formulation can identify practical support needs, accommodation that interferes with learning, and parental anxiety requiring its own assessment. However, personalised selection rules remain hypotheses unless prospectively tested. Neither a null augmentation trial nor a successful parent-only trial settles the optimal dose of parent contact alongside a given adolescent therapy. Treatment fidelity also matters: a nominally parent-inclusive programme may differ greatly in whether it rehearses supportive responses, changes accommodation or merely provides information.

A future study should therefore recruit adolescents with diagnosed anxiety across routine services and deliver the same adequate adolescent CBT in every arm. Randomise caregivers to a brief accommodation-focused module or a credible contact-matched education condition; an additional prespecified stratum could examine parents with their own anxiety disorder. Measure accommodation and parental anxiety before allocation, and power the study for the specific treatment-by-baseline-characteristic interactions rather than searching retrospectively for successful subgroups. Include age, caregiver availability and adolescent preference in recruitment and reporting without treating chronological age as a complete measure of autonomy.

Masked diagnostic interviews, young-person-rated impairment and school or social functioning should be assessed after treatment and at six and twelve months. Parent reports should be retained as a separate perspective, not substituted for adolescent outcomes. Record what parents actually do through agreed observations or structured practice tasks, alongside session attendance and treatment fidelity. Measure changes in accommodation repeatedly before the proposed downstream improvement. Temporal mediation would constrain a mechanism, although additional mediator–outcome assumptions would still be needed for a causal mediation claim.

The trial should also measure family burden, acceptability and any deterioration in relationships, and examine families unable to attend conventional sessions. An approach that improves completer outcomes while excluding caregivers with work or language barriers may reduce practical access. Qualitative interviews can explain why a module helps or fails, while the randomised contrast estimates its incremental effect. The 2022 evidence supports parents as potential partners, change agents and recipients of care, but supports a collaboratively chosen and testable role more strongly than indiscriminate intensification of parental attendance.

Exposure should optimise durable approach and new learning about feared outcomes, rather than merely maximise distress or make anxiety disappear within a session. A child who approaches a dog, attends class or speaks to an unfamiliar person has an opportunity to test a prediction. The relevant outcome is whether learning transfers to everyday situations and reduces impairment. What counts as an informative exposure therefore depends on the individual formulation, developmental understanding and actual safety of the situation.

There is direct evidence that exposure can help young people. Thomas H. Ollendick and colleagues randomised 196 seven- to sixteen-year-olds with specific phobias to a concentrated one-session exposure treatment, education-support or waiting list. Exposure produced better diagnostic and behavioural outcomes than the comparison conditions, with follow-up beyond the immediate treatment session. Its behavioural assessment brought the feared situation into the evaluation rather than relying exclusively on a questionnaire. This supports exposure for specific phobias, but neither the single-session format nor its optimal ingredients can automatically be generalised to every adolescent anxiety disorder.^[ref]

A useful theoretical distinction is between habituation and inhibitory learning. Habituation accounts emphasise reductions in fear with repeated contact. An inhibitory learning account instead proposes that exposure creates a competing association: the cue need not predict the feared consequence. The original association may remain available, helping explain why fear returns after time or a change of context. Michelle G. Craske and colleagues developed strategies intended to strengthen the alternative learning and its retrieval, including testing explicit threat expectancies and varying exposure contexts. Their 2014 account is a theoretical and translational framework, rather than a definitive randomised comparison of these strategies in anxious children.^[ref]

This distinction changes what a therapist measures. Suppose a young person predicts that briefly speaking in class will reliably cause visible ridicule. Completing the agreed contribution, observing the actual response and reviewing what was learned directly address that prediction. A reduction from high to low distress during rehearsal is less decisive if the young person attributes safety entirely to reading a perfectly memorised script. Conversely, meaningful expectancy change can occur while some distress remains. Necessary protections against genuine harm should be retained: a behavioral experiment is not a reason to expose a pupil to ongoing bullying or to remove a medically necessary precaution.

Clinical process data question a rigid requirement for within-session fear reduction. Jeremy S. Peterman and colleagues examined exposure-related distress ratings in 72 treated youths, with a mean age of about ten and a half. Most hypothesised habituation indices did not predict outcomes after treatment or at one year. These observational findings weaken the assumption that a particular distress-reduction trajectory is a necessary marker of success. They do not randomly compare learning strategies, prove that habituation never matters, or establish inhibitory learning as the only explanation.^[ref]

Development provides another reason to test rather than simply import adult optimisation rules. Siobhan S. Pattwell and colleagues studied conditioned fear learning and extinction across human age groups and conducted parallel mouse experiments. Human participants encountered conditioned cues and extinction trials, with physiological fear indexed by skin conductance; adolescents showed attenuated extinction relative to children and adults under the experimental conditions. The parallel animal work examined developmental differences in fear-related circuitry and plasticity. These are important motivating experiments, but an age-group difference in laboratory extinction is not a clinical trial of exposure, nor a reason to withhold effective treatment from teenagers.^[ref]

Practical optimisation should therefore begin with an understandable, testable prediction. Use concrete language, pictures or behavioural examples when abstract probability questions exceed the child's understanding. Agree a feasible approach task and identify what evidence would count against the prediction. Record what happened, including ambiguous outcomes, without demanding a predetermined reassuring conclusion. Exposure can be graded to make participation possible, but progression should depend on meaningful learning and functioning rather than an inflexible rule that fear must halve before moving on.

safety behaviors require functional assessment. A caregiver repeatedly speaking for a child may prevent the child from discovering that they can manage an interaction. A gradual reduction in that accommodation may improve the experiment. However, a communication aid or accessibility adjustment may make participation possible rather than maintain an irrational threat belief. The same outward action can have different functions. Adolescents should help choose targets and evaluate outcomes; parents can support practice without becoming an additional source of checking or coercion.

Variation across people, settings and occasions is a plausible way to strengthen transfer. Practising a classroom contribution only in a quiet consulting room leaves context-specific learning untested. Follow-up tasks in ordinary settings can assess whether an alternative expectation is retrieved when anxiety returns. Spacing, varied contexts and retrieval prompts are reasonable candidates for investigation, but the 2022 youth evidence does not establish a universal optimum for their timing or combination. Treatment time is also finite: an elaborate preparation routine can displace opportunities for actual approach.

A future factorial trial could compare prediction-focused exposure with a credible standard exposure protocol, and separately compare practice across several contexts with practice concentrated in one setting. Both conditions should receive adequate treatment, comparable therapist time and the same planned number of approach opportunities. The comparison must specify actual therapist behaviour rather than merely relabel identical exercises. Stratify recruitment by developmental stage and presenting disorder, and power the prespecified interactions instead of interpreting an underpowered subgroup null as equivalence.

The primary outcomes should be masked diagnostic status and functional impairment after treatment and at six months. A novel-context behavioral approach task would test transfer; repeated expectancy and distress ratings would examine proposed processes. Record fidelity, completed practice, adverse experiences, acceptability and dropout. If prediction-focused treatment improves outcome without greater within-session distress reduction, this would challenge a strict habituation-success rule. If varying context improves transfer but not immediate fear, that pattern would support the proposed retrieval rationale. Repeated mediator measurements help establish temporal order, while causal mediation still requires assumptions beyond treatment randomisation.

Exposure optimisation in 2022 consequently had a strong clinical foundation and several promising mechanistic hypotheses, but fewer decisive youth component trials. The most defensible approach combines effective, collaborative approach practice with explicit learning goals and tests long-term functioning. Developmental neuroscience, distress trajectories and adult learning theory inform the research questions; they do not by themselves establish the best clinical protocol for every child.

For each income band, compare the response distribution within the two gender groups. The table below combines agree and strongly agree only to make one comparable contrast visible. The Pearson chi-squared statistic and its p-value still use all five original response categories, with four statistical degrees of freedom in every band. Income labels are thousands of USD per year; finite upper bounds are excluded.

In the lowest-income band, combined agreement differs by only about $4.39$ percentage points. The full distributions still differ descriptively: strong disagreement is $13.64\%$ in the male group and $27.40\%$ in the female group. The omnibus Pearson chi-squared test does not reject independence at 5%, but that decision does not prove equal distributions.

In the 10k–20k band, combined agreement is almost the same. Nevertheless, strong agreement is higher in the female group, $18.66\%$ versus $11.54\%$, while the less intense agreement category is higher in the male group, $24.36\%$ versus $16.42\%$. Collapsing the categories hides that distinction. The full-distribution p-value is approximately $0.2435$.

In the 20k–30k band, combined agreement is higher in the male group by about $12.96$ percentage points. Strong agreement alone is $27.17\%$ versus $18.30\%$. The observed difference is therefore larger than in the two lowest bands, despite its full-table p-value remaining above 5%.

In the 30k–40k band, combined agreement differs by about $17.40$ percentage points, with fewer male neutral responses, $15.69\%$ versus $26.14\%$. The full-table p-value of $0.03851$ is nominally significant at 5%. In the 40k–50k band, the combined agreement gap is a similar $16.18$ percentage points, but the full-table p-value is $0.1534$. These labels are not a direct comparison between the two effects: significant and nonsignificant results need not differ significantly.

In the highest-income band, strong agreement is $41.84\%$ in the male group and $22.15\%$ in the female group. Combined agreement differs by about $20.95$ percentage points. This band has both the largest sample size and the largest Cramér's V of the six tables, with a very small p-value. Its particularly strong evidence reflects both precision and an observed difference in the distribution.

All six minimum expected frequencies exceed five: approximately $7.15$, $9.57$, $12.02$, $15.04$, $7.93$ and $36.03$. Thus sparse expected counts do not provide a reason to reject the usual chi-squared distribution approximation here. If the six hypothesis tests form one exploratory family, a Bonferroni correction would use $0.05/6\approx 0.00833$ for each test; only the highest-income result crosses that threshold. This qualifies the unadjusted 30k–40k result without turning its observed contrast into zero.

In the elaboration model, the larger agreement gaps in the middle and upper bands provide a descriptive pattern of specification: the conditional contrast varies with income. This is not a Simpson's paradox reversal, because the combined agreement difference has the same sign in every band. Nor is the larger high-income effect size established simply by comparing its p-value with the other p-values. A formal claim of effect modification requires a stated scale and an appropriate comparison of the effects.

The pooled comparison also mixes different income compositions. The highest-income band contains $631/1042\approx 60.56\%$ of the male group but $474/1093\approx 43.37\%$ of the female group. Average each group's six agreement proportions over the same pooled income weights using standardization with common weights. This givesThe crude difference was $19.56$ percentage points, so common weighting attenuates it by about $3.30$ percentage points but leaves a substantial descriptive contrast. With average-group rather than pooled-person weights, the symmetric decomposition of a stratified mean difference instead gives an exact split into approximately $16.29$ percentage points from conditional contrasts and $3.27$ percentage points from composition. These are different explicitly defined descriptive summaries, not estimates of a causally mediated share.

An antecedent income variable could be considered in a confounding explanation, whereas an income variable on an assumed causal pathway would enter an interpretation through causal mediation. The conditional tables do not distinguish these causal structures. They also do not identify whether any adjustment introduces collider bias. The supported conclusion is income-dependent descriptive association plus a measurable composition difference, with causal interpretation left conditional on additional assumptions.

Small, variable effects support a limited influence of facial activity on reported feeling; they do not identify a single feedback mechanism or establish that facial feedback is necessary for emotion. This conclusion separates the experimental contrast from its proposed explanation. Randomly assigning a facial task estimates the effect of that task, including every change the task produces. It does not automatically isolate sensory feedback from the face.

Nicholas A. Coles, Jeff T. Larsen and Heather C. Lench synthesized 286 effect sizes from 138 studies using a random-effects approach with robust variance estimates. Their overall estimate combined outcomes, so the more relevant result for experienced emotion is the emotional-experience subgroup: standardized mean difference $d=0.17$, with 95% confidence interval $[0.11,0.23]$. Affective judgments of stimuli had a larger estimate. The reported moderators included outcome type and properties of the eliciting stimuli. These are comparisons among existing studies, not randomized assignments of the proposed moderators.^[ref]

The distinction between outcome types is theoretically important. Rating a cartoon's objective funniness asks for an evaluative judgment about a stimulus; rating one's own amusement asks about experience. Either can change through inference, expectations or response interpretation. Treating both as direct measures of feeling would make the evidence seem more decisive about the facial feedback hypothesis than it is. Equally, a small average change in experience is compatible with a real contributory mechanism: emotion ordinarily has many inputs, so no single input need dominate.

The classic apparatus illustrates both the ingenuity and the ambiguity. Fritz Strack and colleagues in 1988 had participants hold a pen in ways that facilitated or inhibited smile-related activity, without explicitly instructing them to look happy. They reported differences in humor responses to cartoons. Avoiding an instruction to act happy reduces an obvious route through deliberate category labeling. It does not equate effort, discomfort, attention, interpretation of the task or the sensory consequences of the postures.^[ref]

A subsequent preregistered replication coordinated by Eric-Jan Wagenmakers and colleagues used 17 laboratories and 1,894 analyzed participants and did not reproduce the original difference. That directly limits confidence in the robustness of this particular procedure. It is not a test of every possible facial manipulation, and it should not be counted as evidence independent of a later synthesis that includes those replication studies.^[ref]

There are several possible causal routes. First, information generated by muscle activity or associated bodily changes might contribute to felt emotion. Second, a person might use their own posture as evidence when interpreting an ambiguous state: a self-inference route could be stronger when other emotional information is weak. Third, producing a posture might activate learned associations without an explicit inference. Fourth, the task might change breathing, effort, discomfort or attention. Fifth, demand characteristics might change the report while leaving experience relatively unchanged. Social responses to a visible face are another possible route in interactive settings, but cannot explain a truly isolated participant's immediate response through an actual partner's reaction. These alternatives are hypotheses, not findings established by the meta-analysis.

The logical problem can be stated without assuming that any route is absent. Write assignment as $X$, facial activity as $F$, sensory feedback as $M$ and the report as $Y$. The chain $X\to F\to M\to Y$ and a model with $X\to F$ plus an expectancy route $X\to D\to Y$ can produce the same association between assignment, movement and report. Randomization of $X$ removes preassignment confounding of its total effect, but does not randomize $M$ or eliminate $D$. Establishing causal mediation therefore requires additional interventions or justified assumptions, not simply a significant total effect.

Heterogeneity also has a limited interpretation. A simple random-effects meta-analysis writes $y_i=\mu +u_i+{\epsilon }_i$, with between-study variance ${\tau }^2$ and sampling variance $s_i^2$. Under independent errors the variance of $y_i$ is ${\tau }^2+s_i^2$. Nonzero ${\tau }^2$ means that true effects differ under the model; it does not specify whether the difference comes from sensory mechanisms, expectations, populations or measurement. Dependent estimates require further treatment, which is why a naive analysis counting every effect as an independent study is inappropriate. Small effects and noisy measurement also make individual studies unstable. Selecting a plausible moderator after a failed experiment is insufficient unless it predicts results in a new test.

Useful follow-up work would preregister a factorial design that varies actual facial activity separately from participants' beliefs about its emotional meaning, with matched effort controls and separate reports of felt state and stimulus judgment. Facial electromyography can verify the motor manipulation, but not certify its psychological mediator. A test of a sensory route also needs a defensible manipulation of that route that leaves the relevant alternatives adequately controlled. Masked hypotheses, automated instructions and timing measures reduce particular ambiguities rather than providing a universal guarantee. A necessary-role claim would require evidence that reducing the relevant feedback prevents an otherwise expected emotional response; an initiation claim requires an effect without the external emotional elicitor, while accounting for the task's own meaning.

Social appraisal explains an important route of interpersonal influence, including influence without deliberate emotion labeling, but it is not established as the sole explanation of implicit transfer. The first task is to distinguish outcomes. Interpersonal emotion transfer occurs when another person's reaction changes one's own emotion. Emotional contagion more specifically involves matching affect; facial mimicry is a congruent motor response and need not include a matching feeling. A common event can also change two people's feelings independently, creating correlation without interpersonal transfer.

Social appraisal theory proposes that another person's reaction supplies information about an event's significance. A companion's unease may make an uncertain decision seem riskier; their excitement may suggest an opportunity. The observer's relationship, goals and assessment of the companion's knowledge affect the interpretation. Such evaluation can be rapid and nonverbal. Therefore an absence of consciously naming the companion's anxiety is not sufficient to rule out social appraisal. Equally, if appraisal is invoked whenever any influence occurs, the account loses the independent mechanism needed for an explanation.

Brian Parkinson and colleagues' 2009 diary study collected accounts of decisions involving other people from 41 participants over three weeks. Reports of participants' anxiety and excitement related to the emotions attributed to the other person. Risk and importance appraisals partly accounted for the anxiety relationship, but did not eliminate the other person's predictive contribution. The design provides evidence outside a single laboratory induction and connects interpersonal affect with reported evaluative content. It is also observational: shared circumstances, selective reports and unmeasured interpretations remain alternatives.^[ref]

In particular, statistical adjustment does not settle the causal route. Imperfectly measured appraisal can leave a residual association even if the underlying process is entirely evaluative. Conversely, partial statistical mediation does not establish that changing the measured appraisal would change feeling: a shared cause can affect both. A residual after adjusting for explicit perceptions of another's feeling is compatible with implicit transfer, but is not a direct awareness manipulation and does not prove that every appraisal was absent. The diary findings support a mixed-process interpretation more readily than a clean estimate of independent automatic and cognitive pathways.

Rapid muscular effects supply a different kind of evidence. Ulf Dimberg and colleagues in 2000 presented happy, angry or neutral target faces for 30 milliseconds, followed by neutral masking faces. Surface recordings over facial muscles revealed differential responses consistent with the target expressions, despite masking intended to prevent conscious target perception. This shows that some facial responding need not await a deliberate instruction to imitate. It does not establish that the observer felt the corresponding happiness or anger, or that evaluation of significance was absent.^[ref]

The distinction between motor matching and feeling is supported by Ursula Hess and colleagues' 2001 study of weak, dynamic expressions. It found evidence of mimicry across the tested expression types, and evidence of emotional contagion for happiness and sadness, but its mediation analyses did not confirm a link from mimicry to contagion or recognition. This is evidence against assuming that observing a matching muscle response automatically establishes the whole mimicry–feedback–feeling sequence. It is not proof that such a route never operates: measurement precision and the tested setting bound a negative result.^[ref]

Appraisal also predicts complementary responses that a simple matching rule misses. Gerben A. van Kleef and colleagues in 2004 supplied emotional information about an opponent during computer-mediated bargaining. Participants made larger concessions to an angry than to a happy opponent, with evidence that information about the opponent's limits mattered. Another's anger can signal difficulty or threat and prompt one's own caution rather than matching anger. These experiments establish an informative interpersonal function of emotion communications; their explicit messages do not by themselves demonstrate unconscious appraisal in the masked-face paradigm.^[ref]

The theories can therefore be partly complementary. A face can evoke facial mimicry, alter an appraisal of the situation, and attract attention at the same time. Facial feedback may contribute to feeling, but the 2019 synthesis gives no warrant for assuming a strong obligatory effect from every small muscular response. Learned associations, shared stimulation and report expectations also require consideration. Calling a process implicit must specify what is unavailable to awareness: the expression, the change in one's own state, or the causal relationship between them. These are distinct experimental claims.

A stronger discriminating experiment would vary the displayed reaction independently of the common event and independently of the expresser's relevance to that event. The same ambiguous display could come from an informed adviser or an uninformed observer; a common physical elicitor should be matched across conditions. Predictions about perceived risk, goals and the resulting feeling should be specified before measurement. If appraisal is the route, changing the informational relevance should alter the predicted evaluative content and transfer, beyond differences in face visibility. Testing a mimicry route requires an independent manipulation with a motor control condition: facial restriction can itself alter effort and attention, so it is not automatically a selective intervention on feedback.

The experiment should record facial electromyography, time-linked appraisals and felt-state reports separately. Awareness tests should be appropriate to the actual target-mask sequence and sensitive enough to bound residual discrimination. Randomizing expressions supports their total causal effect, whereas establishing causal mediation still needs control of mediator–outcome confounding or a defensible mediator intervention. Mere disappearance of one regression coefficient after adding another is insufficient. Nor should repeated observations from the same participant or pair be treated as independent replications.

Correlative and causal approaches answer different questions, and neither is dispensable for understanding neuromodulation. An intervention is generally stronger evidence about the consequence of changing a specified target, while observation is essential for discovering the naturally occurring signal, its heterogeneity and the contexts in which it appears. Usefulness depends on the claim, measurement and assumptions, not simply on whether an experiment uses an optogenetic actuator.

A correlation between a neuromodulatory signal and behaviour can arise because the signal influences the behaviour, behaviour changes the signal, or a third variable influences both. Reward expectation, movement and arousal often covary. Temporal precedence helps reject some accounts but does not eliminate an unmeasured common cause. Detailed task manipulations can nevertheless make correlative evidence much more discriminating: varying expectation while holding delivered outcome constant tests a different hypothesis from merely comparing rewarded and unrewarded trials.

The dopamine prediction-error account illustrates this value. Patterns across unexpected rewards, predictive cues and omitted expected rewards constrained an explicit computational hypothesis. A generic rise in activity at a reward would not distinguish reward receipt, expectation and error. Recording before intervening helped identify the timing and task conditions in which a putative teaching signal should be manipulated. Without that information, a stimulation experiment might show that a pathway can change behaviour while leaving the proposed natural computation largely unspecified.

Observations also reveal what population averages hide. Cell-resolved dopamine imaging showed structured sensory, motor and cognitive heterogeneity during a navigation task. Comparing identified dopamine spiking with accumbens release revealed that evolving release dynamics need not mirror recorded somatic activity. These findings constrain the target and measurement level of subsequent interventions. They are not failures of observation: a perturbation of an entire heterogeneous population could obscure precisely the functional distinctions that recording discovers.^[ref][ref]

Interventions can then test a well-specified causal prediction. Reward-timed activation of dopamine neurons overcame associative blocking, changing subsequent cue learning; brief inhibition supported a negative-error-like learning effect. These manipulations are stronger evidence for a teaching contribution than a correlation alone. Yet the manipulated population and timing must match the proposed endogenous signal, and later unstimulated testing is needed to separate acquired learning from immediate performance effects. Sufficiency to impose an effect is not proof that the artificial activity pattern occurs naturally or is the only possible teaching route.^[ref][ref]

Claims of functional necessity and sufficiency are conditional. Selective inhibition that impairs performance can support necessity under the tested conditions, but may act through sensation, movement or motivation rather than the hypothesized computation. A null effect can reflect redundant circuits, incomplete target engagement or compensation. Conversely, stimulation can recruit an unnatural synchrony, reverse-propagating activity or multiple collateral targets. Changing neurons identified by one transmitter can also change co-transmission, so the intervention is not automatically selective for that chemical signal at one synapse.

Pharmacological perturbations make similar assumptions visible. A systemic drug affects receptors according to exposure and selectivity, not according to the experimenter’s functional label. Time of testing, peripheral actions, active metabolites and off-target drug effects matter. The chemogenetic ligand clozapine-N-oxide provides a concrete example: receptor-binding and distribution work showed that conversion to clozapine can account for central DREADD activation. Thus an apparent designer-receptor intervention cannot be assumed to have the intended pharmacological isolation merely because the receptor was genetically targeted. Ligand-only controls and exposure measurements improve the causal interpretation.^[ref]

Measurement is therefore required inside causal experiments as well as before them. Histology can confirm targeting but not the actual moment-to-moment physiological effect; electrophysiology or transmitter measurements can test target engagement. Calcium signals, fibre photometry, voltammetry and spikes have different spatial, temporal and chemical limitations. Converging methods are more persuasive when they agree at their common resolution, while disagreements can identify local regulation or a mistaken measurement model rather than simply experimental failure.

Causal mediation is a further step. Randomizing a drug can establish an effect of the assigned treatment, but correlating a post-treatment neural signal with symptom change does not prove that signal mediated the effect. Both may reflect another drug consequence. Likewise, statistically controlling for movement can remove a confound in one causal model but remove a genuine intermediate mechanism in another. A causal diagram and an explicit target estimand should guide adjustment rather than a rule to regress out every variable associated with behaviour.

Human and animal approaches have complementary scope. Fine cellular interventions can provide internal validity in a restricted animal task, while human longitudinal observations can test whether candidate abnormalities precede illness and whether the same behavioural dimensions matter clinically. Ethical and practical constraints preclude arbitrary human lesions or perfectly selective transmitter manipulation. It would be unhelpful to discard clinically relevant observation for failing to perform an impossible experiment; equally, a prospective association should not be relabelled as a proven biological cause.

The most productive programme iterates among formal models, naturalistic measurements and selective perturbations. Specify the computation, identify its predicted signal, vary task factors that distinguish alternatives, and intervene with appropriate timing, targeting and performance controls. Re-test the resulting behavioural and neural predictions across contexts. Causal experiments generally have an advantage for a precise intervention question, but correlative approaches remain essential for deciding what should be changed and what the causal result actually explains.

I would combine quantitative-genetic cross-fostering with controlled transfer of maternal brood-chamber material in the dung beetle Onthophagus taurus. The hypothesis is that maternal genetic differences influence offspring development partly through the amount of food provided and partly through the composition of material transferred from the mother's body. These pathways must be separated: a larger brood provision, a beneficial microorganism and an inherited offspring allele can all increase offspring mass, but imply different mechanisms.

The system has three relevant precedents. A paternal half-sib study of Onthophagus taurus detected genetic variation in maternal brood provisioning and substantial maternal contributions to offspring phenotype. ^[ref] A separate study reared this species with sterile dung and detected maternal contributions and shared bacterial sequences in the brood chamber and offspring, supporting transmission through the provisioned environment. ^[ref] Experimental developmental benefits of pedestal-associated microbiota were demonstrated in Onthophagus gazella, including context dependence under environmental stress; those functional results should not be assumed automatically to hold in Onthophagus taurus. ^[ref]

The conceptual link is an indirect genetic effect: the provider's genotype affects the recipient's phenotype through a rearing environment. The transferred mediator would be maternal faecal or secreted material, including associated viable microorganisms and soluble products. This fits the socially transferred material framework because the mother contributes material processed in or produced by her body; collecting external dung alone would not establish such a transfer. ^[ref] The actual molecular mediator is an experimental question. Finding bacterial DNA or a protein in a brood chamber does not establish that it is alive, active, beneficial, or responsible for an indirect genetic effect.

First establish a common-garden breeding population with a recorded pedigree and genotype markers. Use a paternal half-sib design with several dams per sire and several adult daughter providers per dam, distributed among independently randomized rearing containers. Standardize donor age, mating history, developmental nutrition and adult diet, and block dung source, moisture and temperature. Because maternally inherited microbes can themselves produce family resemblance, rear replicate provider families with a standardized microbial inoculum in an additional common-rearing arm. Persistence of a family difference alone is insufficient evidence for host genetic causation.

For the initial cross-fostering experiment, allow each provider to construct several brood balls on weighed, standardized dung. Measure construction time, final dry-mass estimate from matched material, egg-chamber properties and maternal faecal pedestal quantity. Remove the original egg before hatching and replace it with an age-matched egg from a separately bred recipient family. Randomize recipient families across provider families, with reciprocal assignments and repeated providers; include handled own-egg and untouched controls. Record egg size and biological-mother identity so that egg provisioning and other pre-transfer maternal effects are not silently attributed to the foster provider.

One possible balanced allocation has 20 sire families, two dams per sire and two daughter providers per dam, giving 80 providers. Four independently prepared brood balls per provider yield 320 offspring-rearing units, allocated across multiple recipient families and collection blocks. This is an illustrative feasible structure, not a claim of adequate power. The effective replication for a sire-level genetic component is much smaller than 320; repeated balls estimate provider consistency and within-provider variation. A pilot-based simulation of the planned pedigree, variance components, mortality and missing samples should determine the final scale before outcomes are examined.

Measure larval growth trajectories, survival, development time and adult size; subsequently measure reproductive performance where feasible. Score provisioning behaviour blind to family and recipient treatment. Collect matched donor and provision samples for absolute microbial load, community composition and soluble chemical profiles. Include extraction blanks, reagent controls and bacterial-load standards, especially because low-biomass sequence samples are vulnerable to contamination. Shared 16S sequences provide limited discrimination among strains; culture or higher-resolution sequencing can test whether a donor-associated strain actually reaches and persists in a recipient.

For an offspring $i$ reared in a ball prepared by provider $j$, an animal model in quantitative genetics could use:Here $a_i^D$ is the direct genetic effect of the offspring, $a_j^I$ the additive provider contribution, $c_j$ a repeatable non-genetic provider effect, $b_{\mathrm{birth}(i)}$ a biological-mother effect, and $B_{ij}$ prespecified batch and rearing terms. Genetic effects use the pedigree or genomic relationship matrix; provider identity alone cannot identify $a_j^I$. If the pedigree contains informative relatives in both roles, direct–indirect genetic covariance can also be estimated. Otherwise that covariance is not identified by the first experiment and should not be assumed known. Crossed recipient and provider families distinguish a general provider effect from matching between a particular recipient genotype and provider background.

The second experiment tests the transferred pathway rather than merely adding a microbial measurement to that model. Construct uniform artificial brood balls with the same treated dung, food mass and chamber geometry, and transfer equal donor-equivalent pedestal material. Compare intact material, material separated into a cell-free soluble fraction, a preparation containing non-viable microbial cells, and a live-cell add-back or rescue preparation. Include material-free, medium and sham-processing controls, together with untreated natural balls as a benchmark. Test viability rather than inferring it from a sterilization label. All processed-dung groups must share the same processing, because sterilization can also change nutritional and chemical properties.

Within a subset, exchange washed live-cell fractions and soluble fractions independently between donor families. A crossed donor-cell-source by donor-soluble-source design asks whether offspring differences track microbial cells, soluble maternal products, or their interaction. Hold total food quantity and transfer volume constant. Include both a standardized live-cell dose, testing compositional differences, and a naturally varying dose arm, testing a possible maternal-genotype effect on delivered quantity. An equal-dose experiment alone could erase a real dose-mediated pathway. Test a candidate cultured community or identified molecule in a dose range anchored to measured natural transfers, rather than interpreting a large artificial supplement as evidence for normal function.

The predicted pattern is specific. If food mass explains the provider effect, equalizing mass should greatly reduce it. If viable maternal microbes mediate an additional effect, removing viability at equal food quantity should reduce the relevant donor-associated growth difference, and a compatible live add-back should restore it. An effect retained in a cell-free fraction instead motivates identifying soluble maternal or microbial products. A failed rescue can reflect processing damage or a missing microbial partner, rather than absence of any natural pathway. If all provider differences disappear after standardized rearing, a non-genetic maternal or microbial inheritance explanation becomes more plausible than the proposed host genetic effect.

Analyse the cross-fostering data first for total provider genetic contributions, then analyse the randomized material manipulations for mechanism. Simply adjusting for a measured mediator can block part of the effect or introduce bias; causal mediation needs assumptions and experimental support. Prespecify the primary outcomes, appropriate survival analysis, family and donor clustering, planned contrasts and treatment-by-recipient-background interactions. Offspring within a maternal pedigree and repeated chemical measurements are not independent genetic replicates. Measure provider costs, such as future fecundity and provisioning effort, because faster-growing offspring alone do not establish the evolutionary advantage of the provisioning behaviour.

The decisive result would connect provider genotype to a measured transfer difference and then experimentally connect that difference to recipient performance, independently of offspring genotype and food quantity. Negative outcomes remain informative: maternal influence without additive provider variance is a maternal effect without the demonstrated genetic component, while heritable provisioning effects without a material-mediated mechanism leave a genuine indirect genetic effect whose pathway is still unresolved.

A statistical description in which an association between an explanatory variable and an outcome is decomposed using an intermediate variable. Such a fitted relation is not automatically a causal pathway; causal mediation requires additional identification assumptions and an appropriate design.