The Indian epistemological traditions converge on a methodological pair. Anvaya — positive concomitance — is the observation that when X is present, Y is also present. Vyatireka — negative concomitance — is the observation that when X is absent, Y is also absent. The two observations are different in kind, and the Indian frame treats them as a single methodological unit: you have not established concomitance until you have done both arms.
The discipline is what looks obvious in retrospect but is routinely skipped in practice. People notice that the kitchen is fragrant whenever bread is baking — anvaya established — and proceed to attribute the fragrance to bread. The vyatireka arm has not been done. Was the kitchen fragrant on the days bread was not baking? If yes (the windows were open onto the bakery downstairs), the anvaya observation was true and the causal attribution was wrong. Without the vyatireka arm, the cognition that “bread causes the fragrance” rests on half the evidence required.
The Nyaya tradition develops the formal use of the pair in inference. To establish vyāpti (the necessary co-occurrence of the reason with the inferred — see N2), Nyaya requires both anvaya (the reason is present in the sapakṣa, the similar instances where the inferred is known to occur — see N3) and vyatireka (the reason is absent from the vipakṣa, the contrary instances where the inferred is known not to occur). Either alone is insufficient. The pair is the formal device by which the Indian logician separates real concomitance from accidental co-occurrence.
The Vedanta application takes the pair beyond inference and uses it as a general analytic method for figuring out what depends on what. The classical Advaita use is the discrimination of the witness from what is witnessed: the witness is anvaya-present in all cognitions (every cognition has a witness) and would be vyatireka-absent from no cognition (you cannot find a cognition where the witness is gone), establishing the witness as something different in kind from the cognitions it witnesses. The same paired discipline is applied to the kosha layers, the states of consciousness, and the substrate-modification relation.
The general operational form is: identify what you are trying to attribute, identify the candidate cause or condition, observe the cases where it is present (does the effect follow?), and observe the cases where it is absent (does the effect cease?). The frame insists on the second observation as a co-equal requirement; it is no mere refinement of the first.
Where English Falls Short
Western methodology has the same machinery under different names. Mill’s methods include the Method of Agreement (close to anvaya) and the Method of Difference (close to vyatireka) as separate canonical procedures. Causal inference, counterfactual analysis, ablation studies, and A/B testing all implement the paired structure. What the Indian formulation contributes is the framing as a unit. Anvaya-vyatireka is one method with two arms; neither arm alone is the method. The English vocabulary tends to treat the arms as alternatives — Method of Agreement or Method of Difference — and to compose them additively when needed. The Indian frame builds the composition into the concept.
This matters in practice because the absent arm is the one routinely skipped. In product analytics, the team that builds an attribution model has done anvaya: the metric moves when the feature ships. The vyatireka arm — did the metric also move on unrelated changes, would it have moved without the feature — is structurally harder and often omitted. The Indian frame makes the omission visible.
The English “correlation is not causation” warning gestures at the same gap and offers no method. Anvaya-vyatireka gives the method: the paired observation is the corrective. Correlation is anvaya alone; causation requires vyatireka alongside.
The vocabulary of “controlled experiment” assumes the paired structure but localises it in the lab protocol; the cognitive discipline that underlies it goes unnamed. The Indian frame returns it to the cognitive discipline, which makes it available outside the lab — in everyday reasoning, in design decisions, in the analysis of one’s own cognition.
Where it Shows Up
Ablation studies in machine learning. An ablation study is anvaya-vyatireka applied to a model architecture. The component is present (does the metric hold?) and the component is removed (does the metric drop?). The discipline of doing both arms is what makes the conclusion attributable. Skipping the ablation and reporting only the presence-arm is the standard ML failure the frame names structurally.
A/B testing and causal inference. A clean A/B test is anvaya-vyatireka with the randomisation handling confounds. The treatment-arm gives anvaya; the control-arm gives vyatireka. The full causal-inference literature on counterfactuals, propensity-matching, and synthetic controls is elaborated machinery for executing the paired observation where direct randomisation is unavailable.
Debugging and bisection. When a bug appeared after changes A, B, and C, bisection is anvaya-vyatireka applied to the change set. Each change is held out in turn; the bug’s presence or absence in each case localises the cause. The discipline is structural: presence-arm and absence-arm together. Either alone is insufficient.
User research with claim and counter-claim. A user reports that a feature helps with task X. The anvaya arm: users with the feature complete task X faster. The vyatireka arm: users without the feature also complete task X faster, because the population that adopted the feature was self-selected for prior facility. The paired test catches the selection effect; anvaya alone misses it.
Diagnostic reasoning in medicine. A patient presents with symptom set S; the candidate diagnosis is condition C. Anvaya: patients with C have S. Vyatireka: patients without C do not have S, or have S with detectably different presentation. The diagnostic literature on test sensitivity and specificity is the formalisation of the paired structure. Skipping vyatireka is the structural error of taking confirming evidence as decisive.
Personal cultivation and habit attribution. “Meditating in the morning makes my day go better.” Anvaya: yes, on days you meditate the day goes better. Vyatireka: on days you do not meditate, what happens? If the days you skip are also the days you slept badly, the meditation may be a marker for sleep; calling it a cause of the day’s quality outruns the evidence. The paired discipline is what separates correlation from cause in personal observation.
Diagnostic Question
“Have I done both arms here — is the candidate concomitance established by what happens when the property is presentandwhat happens when it is absent, or only by one of these?”
IKS Roots
The Sanskrit terms are anvaya (अन्वय, “following along with,” “concomitance in presence,” from anu-i, “to go along with”) and vyatireka (व्यतिरेक, “exclusion,” “separation,” “concomitance in absence,” from vyati-ric, “to leave out,” “to exclude”). The pair is used across the Indian darshanas with consistent meaning. In Nyaya, it formalises the establishment of vyāpti (necessary concomitance, see N2): the reason must be present in the sapakṣa (positive instances, see N3) and absent from the vipakṣa (negative instances). In Mimamsa, it operates in the analysis of which Vedic injunctions apply to which conditions. In Vedanta, especially Advaita, it is the method of discrimination — viveka — by which the witness is distinguished from what is witnessed, the substrate from what is superimposed (V1). The Brahma-sūtras and Sankara’s bhāṣya use the pair throughout. The Sankhya-Yoga tradition uses it in the analysis of the guṇas. The convergence across schools makes anvaya-vyatireka one of the most reliable methodological transfers from the Indian traditions.
See also N2 (vyāpti — the formal target the paired observation establishes in Nyaya inference) and N3 (sapakṣa / vipakṣa — the instance-sets across which the paired observation is conducted).
Further Reading
Bimal Krishna Matilal, The Character of Logic in India, for the Nyaya-Vedanta methodological apparatus. J.N. Mohanty, Classical Indian Philosophy, for the cross-school use of the pair. Karl Potter, Presuppositions of India’s Philosophies, for the deeper methodological framing. Stephen Phillips, Epistemology in Classical India, for the modern reconstruction of the inferential machinery.