The thought is not the route

Before asking which of the five vṛttis is present, we need to stop treating citta as a container of sentences.

In the Sāṃkhya-inflected ontology of classical Yoga, citta is the mutable cognitive instrument. It belongs to prakṛti, the domain of change. It is not identical with puruṣa, the non-agentive seer. Translating citta as “mind” can be convenient, but it easily imports a modern bundle containing brain, awareness, personality and private narrative. The classical term is narrower in some respects and metaphysically more demanding in others. This paper therefore leaves it mostly untranslated.

An editorial choice matters here. Modern readers often meet the sūtras as isolated aphorisms, while the earliest surviving explanatory layer is printed as a separate commentary. Recent philological work argues that the sūtra and bhāṣya layers should be studied as one Pātañjalayogaśāstra, although not every historical question is closed. This paper therefore marks where an argument comes from. The fivefold division and the kliṣṭa-akliṣṭa distinction are in sūtra 1.5-1.6. The explicit wheel in which vṛttis form same-kind saṃskāras and saṃskāras form later vṛttis appears in the commentary. If a reader brackets that commentary, the classificatory distinction remains. The feedback account becomes less secure. Keeping those evidential layers visible prevents a later interpretation from masquerading as the bare aphorism.

The weather image corrects two common mistakes. First, a storm passing across a field is not the field’s identity. “This event functions as viparyaya” is therefore different from “I am deluded.” Second, weather cannot be classified only by its visible outline. Two clouds may look similar while forming through different pressure systems. Likewise, two sentences may match while one is remembered, another inferred and a third merely generated by words.

Two questions hidden in one list

Sūtra 1.5 gives two classifications at once. The event belongs to one of five families, and it is also related to affliction or its interruption. The commentary makes the relation dynamic. An unafflicted event may occur within an afflicted stream, and the reverse may also happen. That makes a one-dimensional ladder impossible.

For practical work, we can make the distinction explicit without pretending to formalise the whole Yoga system. Let a cognitive event be e. Record its family V(e), its object-and-source relation R(e), its loading L(e), and the tendency update it appears to support, ΔS(e).

The last line only says that repeated events change what becomes easier to repeat. Take a deliberately crude score: mark an episode that strengthens aversion as +1, one that interrupts it through clear discrimination as −1, and an unresolved episode as 0. A sequence of +1, +1, 0 and −1 produces a net +1. The number does not measure a saṃskāra. It simply prevents a retrospective story from hiding the direction of practice.

flowchart TD
  A["The meeting at three"] --> B["The thought is not the route"]
  B --> C["Five forms, two axes"]
  C --> D["What survives translation"]
  D -. "re-test" .-> A
  classDef first fill:#0a1424,stroke:#D95545,stroke-width:2px,color:#e6f3ff
  classDef step fill:#0a1424,stroke:#3f7da0,stroke-width:1px,color:#e6f3ff
  class A first
  class B,C,D step

Five forms, two axes

The five families are easiest to grasp by changing one causal property at a time. Start with an event, alter its relation to evidence or time, and watch its classification move.

Pramāṇa: cognition with a warranted route

You look towards a hill and see a continuous column of smoke. You cannot see fire, but you know the relevant relation between smoke and combustion. You infer that something is burning. The classical list under pramāṇa names perception, inference and reliable testimony. These are not three topics that happen to be true. They are three routes by which a cognition can be warranted.

Pramāṇa is therefore more precise than “a good thought”. It may be unpleasant, strategically inconvenient or morally misused. An accurate account of a colleague’s mistake can become material for humiliation. Its truth does not determine its practice loading. This distinction will matter shortly.

Its warrant is also conditional rather than magical. Perception depends on the sense conditions and the object presented. Inference depends on a relation that travels from observed sign to unobserved conclusion. Testimony depends on the speaker’s access and reliability. These dependencies supply concrete failure tests. Improve the light, attack the inference, inspect the speaker’s source or compare independent testimony. A pramāṇa audit is valuable because it tells us where to intervene when confidence and reality part company.

Viparyaya: a fitted appearance defeated by stronger cognition

At twilight, a coiled rope appears to be a snake. The event is not a free-floating concept. It has an object-shaped presentation and prompts bodily readiness. Bring a lamp closer and the apparent snake resolves into fibres. The original cognition is more than uncertain. It is atadrūpapratiṣṭha, established in a form that the object does not have, and it is displaced by a cognition better fitted to the object. The commentary’s example is seeing two moons, corrected by seeing one.

This gives a strict use rule. A disagreement is not automatically viparyaya. Neither is a doubt, a low-confidence guess or a perspective you dislike. The category earns its name when there is a determinate misfit and a stronger cognition that defeats it. Used without that condition, “viparyaya” becomes a Sanskrit way of dismissing someone.

Time matters as well. At the first glimpse, treating the coil as dangerous may be a sensible protective response under uncertainty. After the lamp reveals fibres, continuing to insist on a snake has a different structure. The event has met its defeater but has not updated. This separates the accuracy of the original perception from the flexibility of the later process. The distinction is useful in both contemplative reflection and system assurance: an error-tolerant process can still be reliable if it exposes and incorporates correction.

Vikalpa: word-led construction without the intended object

Imagine a golden mountain floating above the sea. The phrase is intelligible. It combines known predicates and supports further reasoning: perhaps the mountain reflects sunset, perhaps it cannot bear its own weight. Yet no corresponding object need be present. Sūtra 1.9 characterises vikalpa as following verbal cognition while being empty of an object. The commentary also uses ordinary linguistic constructions, not only fantasies. Language can organise action around absences, negations and abstractions.

Counterfactual plans, mathematical idealisations, legal persons, risk scenarios and interface personas all depend on constructions that need not correspond to one present object. Their usefulness comes from controlled relations, not literal existence. A disciplined vikalpa therefore needs an object contract: what the construction stands in for, which operations produced it, where it may guide action and which observation would expose misuse. That is more demanding than either believing the words or dismissing them.

Nidrā: a special event supported by absence

You wake and say, “I slept deeply; there was nothing there.” The statement presents a puzzle. How could an absence be reported if no relevant event occurred? The classical commentary reasons from post-waking recollection to a distinctive sleep cognition. It mentions waking assessments such as having slept pleasantly, painfully or in a stupor. Whatever one makes of the argument, nidrā is not simply a blank outside the system. It is a special case whose support is an absence-cognition.

Smṛti: an earlier object carried into a later event

Memory is defined as the non-loss of an experienced object. The commentary connects the present memory to a prior cognition and a retained saṃskāra. This does not make memory an unedited photograph. The category concerns temporal dependence: the current event presents something as previously experienced. Whether it preserves, recombines or distorts the earlier event is a further question.

A useful record therefore has two temporal layers. The present event may be smṛti because it depends on a prior trace. The remembered content may originally have arisen through pramāṇa, viparyaya or vikalpa. “I remember seeing the snake” tells us how the current scene arrived, not whether there was a snake. This prevents memory from inheriting truth merely by presenting itself as past experience.

The second axis: what recurrence trains

Now change a different property. Keep the cognition true but alter how it is held. A manager accurately remembers a public insult. On the first review, the memory helps establish what happened and supports a proportionate response. On the twentieth review, the same event is rehearsed to renew anger, recruit allies and protect a threatened self-story. Its family can remain smṛti. Its practical loading changes.

What survives translation

A classical category can sharpen a modern experiment without becoming a disguised modern theory. The comparison succeeds only when the preserved relation and the breaking point are both visible.

From pramāṇa to source monitoring

Modern source-monitoring research studies how people attribute a mental experience to perception, memory, imagination or communication. That offers a limited but useful comparison. Both frameworks resist the idea that content carries its own provenance. Both ask how a present cognition came to have its authority. Both predict errors when source features are weak, similar or reconstructed.

The comparison breaks at the level of purpose and ontology. Source monitoring is a family of psychological models and experiments. The five vṛttis sit inside a soteriological system that distinguishes citta from puruṣa and aims at cessation of vṛttis. No experiment on source attribution establishes that metaphysics, and no Sanskrit category supplies a neural mechanism. The bridge preserves a question, not an identity.

Every comparison in this paper uses the same five-part gate. State the source structure in its own vocabulary. State the target structure without borrowing prestige from the source. Name the relation preserved across the bridge. Mark the first point at which the mapping fails. Then derive a test or design change. Here the preserved relation is provenance-sensitive correction. The break appears at ontology and liberative purpose. The resulting test substitutes source routes while holding content fixed. A resemblance that produces no different intervention has not yet earned explanatory status.

Memory, confidence and reconstruction

Constructive-memory research gives a second bounded comparison. Remembering can recombine elements of past experience, and some of the same capacities support imagining possible futures. This helps prevent a crude reading of saṃskāra as a stored photograph. It does not prove that the classical trace is a neural engram or that the two theories carve memory in the same way.

Metacognition research adds another useful separation: first-order accuracy and confidence about accuracy can dissociate. That matters to all five families. A viparyaya may feel certain. A pramāṇa may be held tentatively. A memory may be vivid yet mis-sourced. Confidence is therefore evidence about the agent’s monitoring, not a substitute for the source and object relation.

Dynamic accounts of spontaneous thought make a related move. They examine how constraints tighten or loosen over a trajectory rather than treating “mind-wandering” as one content class. That supports the loop-level analysis developed here. It does not justify translating mind-wandering as vikalpa. One framework studies changing cognitive dynamics; the other places word-led, objectless construction inside a distinct philosophical project. See Christoff et al.

Why sleep is the hardest translation

Contemporary sleep studies can compare reports of dream experience with physiology and neural activity. They can ask whether reported experience varies across rapid-eye-movement and non-rapid-eye-movement sleep, and whether particular patterns predict later reports. But the classical nidrā category is organised around an absence-supported vṛtti and post-waking recollection within Yoga’s own system. The target constructs do not line up automatically.

Even within modern research, the inference from neural pattern to dream experience depends on report timing, awakening procedures and what counts as “experience”. Published critiques of influential dream-correlate studies show that the measurement bridge remains contestable. The honest conclusion is modest: sleep provides a laboratory for separating interval, retained content, report and interpretation. It does not settle whether awareness is present in dreamless sleep.

A configured system: “Backup completed”

Consider an operations assistant that announces, “Backup completed.” The phrase could be copied from a signed effect receipt produced by the storage system. It could be inferred because a workflow reached its final node. It could be retrieved from yesterday’s log. It could be generated because completion usually follows the preceding message. The output is identical. Its authority is not.

A safe system keeps model proposal, world state, typed action, effect receipt and verified outcome separate. The model may propose the sentence. Only an authoritative readback can support release of the claim as fact. This is not an assertion that software has citta, vṛttis or experience. It is an engineering consequence drawn from the route distinction: when consequences depend on truth, provenance must remain outside the fluency of the statement.

The smallest useful test

To see what a content-only record loses, construct twelve synthetic events in pairs. Give each pair the same sentence but different provenance: current perception versus memory, supported inference versus pattern completion, current receipt versus stale receipt. Ask two schemas to record them. The content schema stores the sentence and confidence. The route-aware schema also stores source, object relation, temporal status, available defeater and practice effect.

The point is not that a longer form always wins. It is that the extra fields become decision relevant precisely when identical content demands different correction or authority. If no decision changes after the route fields are added, the framework is unnecessary for that case.

stateDiagram-v2
  [*] --> Claim
  state "Two questions hidden in one list" as Claim
  state "The sentence that survives<br/>every replacement" as Test
  state "The dual-axis vṛtti audit" as Revision
  Claim --> Test: expose mechanism
  Test --> Decision: evidence survives
  Test --> Revision: rival survives
  Revision --> Claim: narrow or rebuild
  Decision --> [*]

From labels to feedback control

The framework earns its keep only if it changes the next intervention. Classification without a change in observation, evidence or practice is decorative.

The loop is the unit of analysis

The commentary on sūtra 1.5 says that vṛttis produce same-kind saṃskāras and saṃskāras produce vṛttis. Read cautiously, this supplies a feedback model. A cognition is not important only because of its present content. It can alter readiness, salience and the ease with which a similar event recurs. A remembered insult can prime attention for fresh disrespect. The next ambiguous expression is then read through that readiness, adding another trace.

This is why a static label fails. It records a frame while discarding the movie. The useful intervention may occur at the source check, the object test, the rehearsal pattern, the action that follows or the environment that repeatedly triggers the loop.

A six-question practice protocol

Apply the audit to one event, not to “my mind today” or “this person”. Write the exact cognition before interpreting it. Then ask six questions.

1. Route: Did this arise through current perception, inference, testimony, recollection, verbal construction, a post-sleep report or an unclear mixture?
2. Object: What is the cognition about, and is a corresponding object intended?
3. Defeater: What stronger observation, testimony or counterexample would revise it?
4. Time: Is the claim current, remembered, imagined, inherited from an earlier source or reported after sleep?
5. Loading: Does this engagement strengthen grasping, aversion or confusion, support discrimination, or remain mixed?
6. Next action: Observe, verify, reframe, pause, seek help, or decline to act?

The protocol is deliberately asymmetric. It permits “unknown” at every disputed point. A forced answer is worse than a visible gap because false precision itself can become a source of viparyaya.

Return to the meeting case. Charu’s sentence is recorded as a present claim arising from recollection. The object is today’s schedule, but the retained source belongs to last month. The decisive defeater is the current invitation. Repeating the old time after seeing the update strengthens defensiveness, so the loading is likely afflicted in this episode. The next action is not “stop having smṛti”. It is to acknowledge the source mismatch, update the calendar and notice the urge to defend the old claim. The classification earns its place because it selects a correction at the right causal point.

Executable lab: the claim-and-practice concordance

The browser-local instrument below exposes the decision logic. It is not a consciousness detector, truth machine, clinical screener or spiritual score. It asks whether the supplied fields support a provisional vṛtti candidate and a separate practice loading. All data stays in the page and is cleared on reset.

How to interpret the lab

What it tests: whether the information supplied supports one provisional event family and a separate loading. What it assumes: that the user can distinguish the event from the person and can report at least some provenance. A positive result permits: a better-targeted check, such as verifying a receipt, testing a defeater or noticing a rehearsal pattern. A negative result means: the source relation is underdetermined, so more observation is required.

What it cannot establish: that the claim is true beyond the evidence entered, that someone has a disorder, that a practice is safe for a particular person, that citta has been measured, or that a machine instantiates any classical category. For significant distress, impaired functioning, safety concerns or consequential action, stop treating the tool as sufficient.

/** @typedef {'pramāṇa'|'viparyaya'|'vikalpa'|'nidrā'|'smṛti'|'underdetermined'} Family */
/** @typedef {'likely kliṣṭa'|'likely akliṣṭa'|'mixed'|'unknown'} Loading */

/**
 * Invariants:
 * 1. Classify one event, never a person.
 * 2. Keep family and loading separate.
 * 3. Preserve unknown rather than forcing certainty.
 * 4. Consequential action always requires external evidence and authority.
 */
function classifyEvent(record) {
  let family = 'underdetermined';
  let reason = 'The supplied route and object relation do not select one family.';

  if (record.temporalStatus === 'post-sleep' ||
      record.sourceRoute === 'post-sleep-report') {
    family = 'nidrā';
    reason = 'The event is a post-sleep, absence-related report.';
  } else if (record.temporalStatus === 'remembered' ||
             record.sourceRoute === 'recollection') {
    family = 'smṛti';
    reason = 'The present event depends on a prior experience or trace.';
  } else if (record.objectRelation === 'contradicted') {
    family = 'viparyaya';
    reason = 'Stronger evidence defeats the presented form.';
  } else if (record.sourceRoute === 'verbal-construction' &&
             record.objectRelation === 'no-object-intended') {
    family = 'vikalpa';
    reason = 'Verbal construction is operating without an intended corresponding object.';
  } else if (['direct-perception', 'inference', 'testimony']
               .includes(record.sourceRoute) &&
             record.objectRelation === 'supported') {
    family = 'pramāṇa';
    reason = 'A stated valid route currently supports the cognition.';
  }

  /** @type {Loading} */
  const loading = ['grasping', 'aversion', 'confusion']
    .includes(record.practiceEffect) ? 'likely kliṣṭa'
    : record.practiceEffect === 'discernment' ? 'likely akliṣṭa'
    : record.practiceEffect === 'mixed' ? 'mixed' : 'unknown';

  return { family, loading, reason };
}