Yogācāra: storehouse, seeds and transformation

The owner who never appears

The puzzle has three pieces. Buddhist analysis denies an unchanging self behind the five aggregates. Conditioned events arise and cease. Yet actions have delayed effects, habits become easier to enact, and a mind can return from sleep or profound meditative cessation. If every moment is new, what carries anything forward?

Modern readers often hear “no-self” as a claim that persons are unreal in every sense. That is too crude. A person can remain a useful conventional designation for an organised stream, just as “storm” names a temporally extended pattern without naming an extra object beyond air, water and motion. The philosophical pressure concerns what must exist ultimately to make continuity possible. Vasubandhu’s arguments reject a separate person over and above causally active processes; the person is not an additional controller that perceptions and actions secretly obey.[7][11]

The likely intuition is simple: if learning occurs, there must be one thing that had the earlier experience and now possesses its result. The hidden assumption is stronger: causal continuity is being treated as numerical identity. Something can influence its successor without remaining self-identical through the succession. A flame lights another flame. A river’s earlier current deposits sediment that bends later current. A language changes through speakers who never share one permanent mind.

Thought experiment: the replaced violinist

Imagine a violinist whose relevant biological structures are replaced one functional unit at a time. Each replacement preserves the causal dispositions required for timing, pressure and correction. After years, no original unit remains, but the player still adjusts a note before consciously naming the error. Did the skill disappear at the first replacement, at the last, or never?

If skill requires an unchanged bearer, every answer looks arbitrary. If skill is a historically produced organisation of dispositions, gradual replacement need not interrupt it. This does not prove Buddhist no-self. It isolates the assumption that sameness of material or metaphysical substance is required for learned continuity.

Figure 1. The riverbed model distinguishes persistence of a substance from persistence of causal influence. Earlier flow changes the conditions of later flow, even though no water drop acts as the river’s owner. Conceptual synthesis. The river comparison is prominent in Yogācāra sources, including the Triṃśikā and discussions of the Saṃdhinirmocana Sūtra.[2][8]

Why the problem became acute

Early Yogācāra materials did not invent the problem from nothing. Earlier Buddhist schools already used notions such as underlying tendencies and seeds to explain how a momentary act could alter a later stream. The Sautrāntika seed theory is an important predecessor.[9] What Yogācāra added was a more explicit stratification of cognition: intermittent manifest consciousness, an afflicted self-making process, and a subliminal continuity called ālayavijñāna.[5]

Scholars disagree about why the new layer first became necessary. One reconstruction treats it as a response to no-self, momentariness, karma and rebirth. Another traces its decisive origin to a technical puzzle about nirodhasamāpatti, an attainment in which ordinary conscious activity ceases. How can a practitioner remain alive and later resume cognition if no preceding cognition is available to cause the next one? Schmithausen emphasised this meditative context; Griffiths analyses the associated mind-body problem; Yamabe argues that meditative and bodily experience may have been more directly formative than abstract theory alone.[4][12][14]

Contested historical interpretation

The paper does not need to choose one origin story. The same conceptual mechanism can answer several pressures: how latent tendencies remain effective, how intermittent cognition resumes, and how a stream bears consequences without a permanent owner. What matters is the type of continuity being proposed.

A useful continuity claim needs more than temporal order. Rain follows a bell without inheriting anything from it. A recording follows a button press, but the recording does not become easier to play because it was played yesterday. Dispositional continuity requires three relations. A later moment must inherit a selective bias from earlier moments, that bias must help explain what later occurs, and later activity must be able to revise the bias. Inheritance without causal relevance is merely a record. Causal relevance without revisability is a fixed mechanism. Revisability without inheritance is a sequence of fresh starts.

This yields a distinction that a snapshot cannot reveal. Two streams can look identical now while carrying different sensitivities to the next event. Conversely, one stream can change almost every observable feature while retaining a causal lineage through successive handovers. Continuity is therefore neither resemblance across time nor the survival of an inner token. It is a relation among histories, present powers and possible transitions.

Worked example: the repaired melody

A pianist and a player piano produce the same eight notes. Stop both after note four, alter the timing of note three during rehearsal, then restart from the beginning. The pianist’s later phrasing may change because the correction altered a disposition governing the whole passage. The player piano repeats the stored sequence unless its programme is edited. The initial performance cannot tell us which kind of continuity is present. The history-sensitive restart can.

The example also shows why “memory” is too broad. The player stores tokens. The pianist carries a modifiable tendency that changes timing under related conditions. Both preserve information, but only one exhibits the reciprocal lineage at issue.

Substance continuity

The same inner entity owns changing states. Change happens to it, but it is not constituted by the change.

Token storage

Past contents survive as retrievable items inside a container. Continuity means keeping the items.

Dispositional continuity

Past events alter causal powers. Continuity means that later states inherit and revise those powers.

Figure 2. The Yogācāra mechanism is easiest to misunderstand when the storehouse metaphor is read as either an enduring substance or a warehouse of unchanged mental tokens. The more useful causal reading is dispositional continuity. Analytical reconstruction, not a claim that every Yogācāra text uses one identical model.

The argument in one pass

Chart · flow

flowchart LR
  A["The owner who never appears"] --> B["The river that rewrites its bed"]
  B --> C["Transformation without a commander"]
  C --> D["What the model earns, and<br/>where it breaks"]
  D -. "re-test" .-> A
  classDef first fill:#0a1424,stroke:#7B4CD8,stroke-width:2px,color:#e6f3ff
  classDef step fill:#0a1424,stroke:#3f7da0,stroke-width:1px,color:#e6f3ff
  class A first
  class B,C,D step

A structural route through Yogācāra: storehouse, seeds and transformation: from The owner who never appears to Make responsibility transform causal…. Reading use: follow the relation, then return to the surrounding argument for evidence and limits.

The river that rewrites its bed

Ālaya is commonly translated as store, storehouse, basis or substratum. None is perfect. “Storehouse” suggests a place containing objects; “substratum” suggests an underlying substance. Yogācāra texts themselves use several etymologies and functions. The Saṃdhinirmocana Sūtra presents this underlying consciousness through terms tied to appropriation and continuity, while later systematisation expands its seed-bearing role.[1] The safer approach is to ask what causal work the term performs.

Vasubandhu’s Triṃśikā calls the first transformation of consciousness the maturation consciousness and says it has all the seeds. It also compares its operation to a river’s current. Later verses describe consciousness and its seeds as changing through reciprocal influence, giving rise to further conceptual constructions and maturation.[2] Asaṅga’s Mahāyānasaṃgraha develops this reciprocal relation in greater detail: manifest forms of cognition arise from dispositional conditions, then “perfume” or condition the continuing stream in return.[3]

Bīja, seed, names a potential for a later result. Vāsanā can mean an impression, habituation or perfuming. The botanical and olfactory metaphors make the same point from different sides. A seed is not the mature plant in miniature. A perfume does not preserve the event that produced it. Both name a changed propensity under suitable conditions.

“Suitable conditions” carries much of the explanatory burden. A seed is not a miniature command waiting to execute. The same inherited tendency may remain quiet, mature weakly or contribute to very different manifest events as bodily state, attention, social setting and competing dispositions change. This relational character prevents the theory from becoming a catalogue in which every possible event is already stored in completed form.

For causal analysis, it helps to treat a disposition as a conditional asymmetry. It can make one feature easier to notice, one interpretation easier to form, one response easier to enact, or one update easier to reinforce. These are different roles. A person may notice criticism quickly yet inhibit retaliation. Another may notice no threat initially but learn defensiveness after repeated humiliation. A single scalar “seed strength” is therefore a teaching simplification, not a faithful inventory of mind.

Worked example: the cup that is not the cause by itself

After becoming ill soon after drinking jasmine tea, Arun feels nausea when he smells it. The smell is not carrying a stored copy of the illness. A previous episode changed how a later cue recruits bodily expectation and avoidance. Serve the same tea during hunger, safety and deliberate re-exposure, and the response may differ. The inherited trace matters, but it matures only with present conditions.

Now block the smell while preserving the visual cup. If nausea falls, olfactory processing mediates part of the effect. Change the label while preserving the aroma. If the response persists, the label is not the main pathway. Such interventions turn the seed metaphor into questions about selective causal routes.

The decisive mechanism

At moment t, inherited dispositions help shape which world becomes salient, how it is divided into grasper and grasped, and which response becomes likely. The enacted cognition then modifies the dispositional stream that conditions moment t + 1. There is no separate owner between the two moments. The causal handover is the continuity.

The classical eight-consciousness scheme makes a further distinction. The six manifest consciousnesses concern ordinary sensory and mental cognition. Kliṣṭamanas, afflicted mentation, takes the subliminal stream as its object and appropriates it as “I” and “mine”. The ālayavijñāna is therefore not the self. It is precisely what self-making mistakes for a self.[8] This is an elegant diagnostic move: the feeling of an enduring owner is itself one conditioned activity within the stream.

Figure 3. The loop is reciprocal, not a one-way retrieval system. The strongest test is to intervene on a disposition or on the enacted response and observe whether later maturation changes while the present cue is held constant. Causal reconstruction grounded in Triṃśikā verses 18–19 and the reciprocal conditioning developed in the Mahāyānasaṃgraha.[2][3]

From metaphor to a minimal causal model

To make the logic explicit, let d_t denote a vector of dispositions at a moment, x_t the present conditions, and y_t the cognition or action that becomes manifest. A minimal model needs only two equations:

y_t = g(d_t, x_t)
d_t+1 = F(d_t, x_t, y_t)

The first equation is maturation: current conditions do not act on a blank system. The second is perfuming: what arises and is enacted changes the stream that follows. There is deliberately no variable for a permanent self. The model does not deny the conventional person. It shows that the causal explanation of learning does not require an additional owner-variable.

The notation also exposes an assumption hidden by the river image. The update function F determines what counts as learning. If F merely appends a verbatim episode, the system has accumulated a record. If it changes which response becomes likely under related conditions, it has acquired a disposition. If practice changes F itself, the system has altered how it learns. Those three levels, content, tendency and update rule, should not be collapsed.

A historically serious model would represent many interacting dispositions, not one number. It would also distinguish the occurrence of an event from its uptake. Two people may undergo the same public incident yet acquire different traces because attention, interpretation, bodily arousal and prior learning differ. The past affects the future through the change it actually produces, rather than only because the event appears in a chronology.

Worked example: one ambiguous message

Let Mira’s defensive disposition be 0.60 on a scale from 0 to 1. The message contributes a threat cue of 0.20. A simple response rule gives a 0.77 probability of defensiveness. If she enacts the defensive response, the update raises the disposition to 0.77. If she notices the impulse and performs a trained counter-response, the same rule lowers it to 0.47. Under the next matched cue, the predicted defensive probabilities are roughly 0.87 and 0.66.

The numbers are illustrative. Their purpose is to expose the causal claim: a present intervention matters because it changes what a later event is able to mature.

Inherited tendency

d = 0.60

Defensive responses are already easier to trigger.

Present cue

x = 0.20

The message is ambiguous, not neutral to this stream.

Two update paths

Defend: d′ = 0.77

Counter-train: d′ = 0.47

Same cue, different enacted trace.

Figure 4. A disposition is not a stored sentence about the past. It is a present causal asymmetry that changes the probability of future responses. Intervention works by updating that asymmetry. Illustrative synthetic values; no empirical measurement is claimed.

The trace-and-disposition test

The model suggests a new experimental distinction between storage and continuity. Call a system historically sensitive across horizon k when changing an earlier event changes a later response under matched current conditions. That alone is insufficient, because the effect may travel through an uncontrolled pathway. A stronger test blocks the proposed update mechanism. If the later difference collapses, the state update mediated the continuity.

Δ_k = E[Y_t+k | do(E_t=e₁), X_t+k=x] − E[Y_t+k | do(E_t=e₀), X_t+k=x]

Practitioner synthesis

The trace-and-disposition test does not verify ālayavijñāna. It operationalises one causal relation that the seed and perfuming framework makes salient. A positive result permits the modest conclusion that prior events alter later behaviour through a mutable dispositional state. A negative result means the proposed state is unnecessary for that task or that the intervention failed. Neither result decides karma, rebirth, phenomenal consciousness or Buddhist metaphysics.

Transformation without a commander

If the stream merely reproduced its inherited tendencies, Yogācāra would offer a theory of fate, not a path. Its practical force lies in the loop’s revisability. Attention, ethical restraint, repeated action, reflection and insight can change what gets reinforced. The mind is conditioned, but conditioning includes the conditions introduced by practice.

This makes learning more than adding content. A new belief can sit on top of an old reaction while the reaction remains dominant. Deeper transformation occurs when the mapping from cue to salience, feeling, intention and action is altered. In classical language, the path culminates in āśrayaparāvṛtti, transformation or reversal of the basis. Texts and commentators differ about its precise scope, but it is not the enthronement of a purified permanent self. The afflicted organisation, especially dual grasping and the seeds that sustain it, is transformed or brought to cessation.[10]

Three depths of change make the claim concrete. A content-level correction changes what one says in a particular case. A dispositional correction changes which response becomes easy across a family of cases. A transformation of the update rule changes what future experience reinforces. Someone may learn the sentence “criticism is useful” while still rehearsing resentment after every meeting. The sentence changes content. Repeatedly attending to specific information rather than social threat may change a disposition. Seeing the self-defensive framing as constructed may alter which reactions receive further reinforcement.

Worked example: an apology at three depths

Leena memorises a courteous apology for customer complaints. In the first week, she says it while internally rehearsing blame. Her output changes, but the old tendency is repeatedly strengthened. Later she learns to identify the customer’s concrete loss before composing a reply. Her attention and response policy change across cases. Eventually, hostile arousal itself becomes a cue to pause and inspect self-protective interpretation. The event that once strengthened blame now initiates a different update.

This scenario is not evidence for a classical attainment. It demonstrates the causal difference between replacing a sentence, reshaping a tendency and changing what subsequent experience teaches.

Figure 5. Moving the mutable state outside a candidate system can preserve input-output behaviour while changing continuity, dependence and interruption. This is why a cloud ledger or model checkpoint should not be casually called an ālayavijñāna. Boundary thought experiment; the right panel is an engineering analogy with explicit break conditions.

Thought experiment: the external storehouse

Suppose every disposition in an adaptive agent is moved to a remote service. The agent queries the service before each action and writes an update afterwards. Latency is negligible, so behaviour remains identical. Has the storehouse moved outside the agent, or has the candidate system expanded to include the service?

The answer depends on the boundary criterion. Yogācāra’s ālayavijñāna is more than any database that makes behaviour history-dependent. It is presented as a continuous basis associated with embodiment, subliminal cognition, karmic maturation and the path. The external service preserves one relation, historical conditioning, while losing others. The experiment therefore yields a design insight, not an equivalence.

An executable trace-and-disposition lab

The lab below starts two streams from the same defensive disposition. Both receive the same ambiguous criticism eight times. One enacts the automatic defensive response; the other performs a counter-trained response. Learning rate controls how strongly an enactment perfumes the next state. Decay controls how much a disposition fades without reinforcement.

Executable lab

Fork one stream, change only the update path

Learning rate 0.18

Decay 0.03

Both branches begin at disposition 0.60. Run the comparison to see how identical cues diverge under different enacted traces.

Figure 6. The live chart is an explicitly illustrative trajectory. A result counts as historical sensitivity only because both branches receive matched cues and differ in the update path. Synthetic simulation. It demonstrates model logic, not human effect size.

JavaScript · core update rule

function step(disposition, cue, policy, learningRate, decay) {
  const enactment = policy === "automatic" ? 0.62 : -0.48;
  const next = (1 - decay) * disposition
             + learningRate * (cue + enactment);
  return Math.max(0, Math.min(1, next));
}

// Same cue, same starting state, different enacted trace.
const automatic = step(0.60, 0.20, "automatic", 0.18, 0.03);
const trained   = step(0.60, 0.20, "trained",   0.18, 0.03);

What the lab tests. It tests whether a mutable state can carry the effect of an earlier enacted response into a later matched situation. It encodes four assumptions: the disposition is scalar, the cue is constant, the update is deterministic, and the two policies differ only in their trace. Change any of those assumptions and the trajectory changes.

A positive result permits one conclusion: under this model, different enactments produce different future propensities. A negative result, such as zero learning rate or overwhelming decay, means no lineage effect survives. The lab cannot establish that real human habits obey this equation, that ethical karma is a scalar, or that a machine implementing it has consciousness.

To use the test outside the toy model, define the candidate disposition before examining results. Specify the situations across which it should generalise, the state expected to mediate it, and the intervention that should weaken it. Then separate acquisition from expression. A person can acquire a tendency that a strong social constraint temporarily suppresses. A system can express a pattern from a prompt without having updated any persistent state. Probe both the latent mediation and the later behaviour.

Failure case: history without a disposition

A service saves the sentence “be more patient” after a complaint and inserts it into the next prompt. Its later reply improves. Remove the sentence and the improvement vanishes immediately, with no transfer to semantically related situations. The service is history-sensitive, but the effect is carried by token retrieval rather than a changed response tendency. The trace-and-disposition test should classify this as stored instruction, not dispositional continuity.

Now fine-tune a policy so that the improvement survives paraphrase and the original sentence is unavailable. That is stronger evidence for a changed tendency, although it still says nothing about a self or a subject of experience.

Optional depth: full intervention protocol and invariants

Run at least three matched histories from the same initial state. Vary only the event at the nominated intervention point. Freeze the update function in one control arm. Re-present an identical cue after a fixed horizon. Record the future response, the latent state and any external pathway that could carry the intervention. The key invariant is that the current cue and policy interface remain matched across arms. The key falsifier is a persistent future difference when the proposed update path is disabled.

Forks, resets and responsibility

Figure 7. A fork makes the difference between shared ancestry and later identity visible. A reset can alter internal disposition while leaving social and material consequences untouched. Illustrative lineage model; it does not purport to settle Buddhist or legal theories of personal identity.

A bounded comparison with adaptive machines

Consider a synthetic tutoring system with four distinct resources: model weights, a learner profile, current lesson state and an evidence log. Only the learner profile is updated by the student’s errors and successes, then used to select later explanations. That profile preserves the structural relation tested above: past interaction changes future response through mutable state.

Configured-system example: two learners, one model

The model weights and algebra question are identical for Nila and Omar. Nila’s profile records a high probability of sign errors after distribution; Omar’s records difficulty translating prose into equations. The system proposes different hints, then updates only the profile dimension supported by the learner’s next action. The evidence log separately records which question, hint, response and update occurred.

Swap the profiles while holding the question fixed. If the hints swap, the profiles causally condition the proposals. Freeze profile updates for one learner. If later adaptation stops, the update path mediates learning. Delete the evidence log. The system may still adapt, but it loses the receipts needed to justify why. This configured example distinguishes dispositional state, current world state and historical evidence without calling all three memory.

The comparison stops there. The Yogācāra source structure includes karmic valuation, embodiment, subliminal cognition, dual grasping, rebirth and liberation. The tutoring profile has none of these by default. Model weights are not seeds, the learner database is not ālayavijñāna, and an audit log is not a mind-stream. The design insight is narrower: separate the state that conditions future proposals from the records that merely describe the past.

Comparison break condition

The analogy fails whenever “history-dependent output” is treated as sufficient for personal continuity, consciousness or moral status. Many systems are history-dependent. Yogācāra’s claim concerns a much richer, tradition-internal account of living cognition and liberation.

How the claim changes under pressure

Chart · state

stateDiagram-v2
  [*] --> Claim
  state "Why the problem became acute" as Claim
  state "From metaphor to a minimal<br/>causal model" as Test
  state "The trace-and-disposition test" as Revision
  Claim --> Test: expose mechanism
  Test --> Decision: evidence survives
  Test --> Revision: rival survives
  Revision --> Claim: narrow or rebuild
  Decision --> [*]

The revision loop begins with Why the problem became acute and keeps counterexamples inside the reasoning path. Reading use: follow the relation, then return to the surrounding argument for evidence and limits.

What the model earns, and where it breaks

The storehouse theory earns a genuine explanatory possibility: delayed influence need not be carried by a permanent person or by unchanged mental tokens. A moment can alter the causal powers of its successor, and those powers can be progressively reshaped. That is enough to explain a form of learning, habit and continuity.

It does not automatically solve every problem. Critics have long worried that the ālayavijñāna performs too many functions traditionally assigned to a self: it maintains continuity, bears karmic seeds and links lives. Yogācāra replies that it is conditioned, momentary in its constituents, continually changing, misappropriated by afflicted mentation and absent or radically transformed at awakening.[8] Whether this avoids a self in substance rather than merely in name remains a live philosophical question.

There are also rival Buddhist explanations. Sautrāntika seed theory sought continuity through a mind-stream without positing a distinct eighth consciousness. Some Abhidharma systems proposed other carriers or relations. Contemporary naturalistic accounts appeal to bodily organisation, neural plasticity, social scaffolding and narrative construction. These rivals can reproduce parts of the same evidence. The relevant question is not which vocabulary sounds deeper, but which intervention differentiates their mechanisms.

The trace-and-disposition schema is compatible with several of those rivals. Neural plasticity could realise the inherited bias. A socially maintained routine could carry part of it outside the body. A narrative could recruit and stabilise it. A dynamical system could describe it as motion among attractors. This underdetermination is deliberate. The schema identifies what a continuity explanation must make testable before it argues about the substrate or ontology that realises the relation.

Its limits are equally important. The two-equation model does not represent karmic valence, the constitution of a lived world, the beginning or ending of a life-stream, rebirth, or liberation. It cannot adjudicate whether Yogācāra is best interpreted as metaphysical idealism, phenomenological analysis or a soteriological strategy. A causal abstraction can preserve a relation without preserving the world in which that relation has meaning.

Figure 8. “Transformation of the basis” is represented here as a change in the landscape of likely responses, not as the purification of a permanent substance. The geometry is explanatory, not a historical Yogācāra diagram. Illustrative synthesis informed by discussions of āśrayaparāvṛtti.[10]

The strongest alternative explanations

A good test must ask what else could produce the same observation. If Mira reacts differently after six months, the cause may be a revised belief, changed workplace safety, medication, fatigue, social support or simple rehearsal. The dispositional model is not confirmed by change alone. It is strengthened when interventions have selective, mediated and durable effects predicted by the proposed state variables.

The same discipline applies to historical interpretation. The theory’s origin may lie in meditative phenomenology, scholastic pressure, scriptural reconciliation or a combination. Its ontology may be read as idealist, representationalist, phenomenological or strategically soteriological. Modern scholarship does not supply one uncontested answer.[6][13] The causal thesis of this paper is deliberately narrower than that debate.

Observation	What it supports	Rival generator	Next intervention	Prohibited inference
Matched cue, different response after controlled history	Historical sensitivity	Uncontrolled environmental change	Randomise history; match current conditions	A permanent self exists or does not exist
Blocking state update removes later difference	Proposed state mediates continuity	Hidden parallel state	Patch or swap candidate state selectively	The state is ālayavijñāna
Same compassionate output from two systems	Behavioural equivalence only	Scripted lookup	Vary prior history and counterfactual cues	Equivalent learning or moral cultivation
Counter-training changes future trajectory	Update rule is revisable	Temporary context priming	Test persistence, transfer and relapse	Transformation of the basis has occurred

Figure 9. The evidence matrix prevents a useful causal test from expanding into a metaphysical verdict. Each observation earns a typed conclusion and a specific next intervention. Decision instrument developed for this paper.

Decision rule

Use the Yogācāra comparison when it helps you identify a causal lineage, a mutable disposition and a reciprocal update. Stop when the comparison is asked to establish consciousness, karma, rebirth, moral status or formal equivalence.

What would strengthen or weaken the proposal

The trace-and-disposition synthesis would be strengthened by experiments in which carefully manipulated histories produce selective later differences, the differences are mediated by identifiable mutable states, and counter-training changes both state and behaviour across contexts. It would be weakened if current input alone predicts later response, if any arbitrary latent variable fits equally well, or if disabling the supposed update path leaves the historical effect intact.

A rival could always say that a different physical or computational mechanism carries the effect. That is not a defect. The synthesis is a causal schema, not a claim that one substrate uniquely realises it. Its smallest useful implementation is the two-equation model above. Its proper domain is the analysis of history-sensitive learning. It should not be applied as a shortcut from behavioural persistence to a claim about a subject of experience.

The evidence conversation

Chart · sequence

sequenceDiagram
  participant Q as Question
  participant M as Mechanism
  participant E as Evidence
  participant D as Decision
  Q->>M: The owner who never appears
  M->>E: Why the problem became acute
  E-->>M: compare rival explanations
  E->>D: Make responsibility transform causal inheritance
  D-->>Q: revise scope or act

Question, mechanism, evidence and decision exchange the burdens developed in Yogācāra: storehouse, seeds and transformation. Reading use: follow the relation, then return to the surrounding argument for evidence and limits.

Make responsibility transform causal inheritance

When a pattern persists, stop asking first, “What inner thing owns it?” Ask four more discriminating questions. Which earlier events changed the stream? What disposition now biases maturation? Which current action will reinforce or weaken it? Where does the relevant causal boundary actually lie?

For personal practice, this changes blame into mechanism without erasing responsibility. A harmful response is neither the expression of an immutable self nor an event with no owner in any practical sense. It belongs to a conditioned lineage, produces consequences, and can be interrupted by conditions that alter the next update. Responsibility becomes the work of transforming a causal inheritance.

For engineered systems, the decision is equally concrete. Do not call every persistent resource “memory”, and do not call model weights a storehouse consciousness. Separate checkpoint, episodic trace, world state, user profile and evidence ledger. Then test which resource actually mediates future behaviour. A system that produces the right words from a script has not thereby learned the disposition those words express.

For philosophy, Yogācāra makes continuity cheaper than a permanent self but more demanding than a sequence of disconnected moments. The price is a disciplined account of causal handover. The gain is a model in which history matters, practice can matter, and transformation does not need a commander standing outside the process.

Glossary

Ālayavijñāna: The store, basis or substratum consciousness in Yogācāra. It is presented as a changing subliminal stream that supports dispositions and continuity, not as a permanent self.
Bīja: “Seed”. A causal potential or disposition capable of maturing under suitable conditions.
Vāsanā: Impression, habituation or “perfuming”. The way an event leaves a conditioning trace in a stream.
Vipāka: Maturation or ripening, especially the arising of a result from prior conditions.
Pravṛttivijñāna: Manifest or occurrent forms of sensory and mental cognition.
Kliṣṭamanas: Afflicted mentation that appropriates the continuing stream as “I” and “mine”.
Āśrayaparāvṛtti: Transformation or reversal of the basis, a soteriological reorganisation associated with overcoming afflicted cognition and dual grasping.
Causal lineage: A sequence linked by inherited and revised causal influence, without requiring one numerically identical substance throughout.

References

Saṃdhinirmocana Sūtra. Unraveling the Intent, translated from Tibetan by the 84000 translation team. 84000: Translating the Words of the Buddha.
Vasubandhu. Triṃśikā Vijñaptimātratā (Thirty Verses), comparative translations including Khenpo David Karma Choephel, Mattia Salvini, Stefan Anacker and others. Verse by Verse.
Asaṅga. A Compendium of the Mahayana: Asaṅga’s Mahāyānasaṃgraha and Its Indian and Tibetan Commentaries. Translated by Karl Brunnhölzl. Snow Lion, 2019. Publisher page.
Schmithausen, Lambert. Ālayavijñāna: On the Origin and the Early Development of a Central Concept of Yogācāra Philosophy. International Institute for Buddhist Studies, 1987; reprint with addenda, 2007. IIBS publication record.
Waldron, William S. The Buddhist Unconscious: The Ālaya-vijñāna in the Context of Indian Buddhist Thought. RoutledgeCurzon, 2003. https://doi.org/10.4324/9780203451175.
Lusthaus, Dan. Buddhist Phenomenology: A Philosophical Investigation of Yogācāra Buddhism and the Ch’eng Wei-shih lun. Routledge, 2002. Publisher page.
Gold, Jonathan C. “Vasubandhu.” Stanford Encyclopedia of Philosophy. Entry.
Szanyi, Szilvia. “Yogācāra.” Stanford Encyclopedia of Philosophy. Entry.
Jaini, Padmanabh S. “The Sautrāntika Theory of Bīja.” Bulletin of the School of Oriental and African Studies 22, no. 2 (1959): 236–249. https://doi.org/10.1017/S0041977X00068683.
Szanyi, Szilvia. “The Changing Meanings of āśraya in Vasubandhu’s Abhidharmakośa(bhāṣya).” Journal of Indian Philosophy 49 (2021): 953–973. https://doi.org/10.1007/s10781-021-09489-8.
Siderits, Mark. Personal Identity and Buddhist Philosophy: Empty Persons. Revised edition. Routledge, 2015. https://doi.org/10.4324/9781315247052.
Yamabe, Nobuyoshi. “Ālayavijñāna from a Practical Point of View.” Journal of Indian Philosophy 46 (2018): 283–319. https://doi.org/10.1007/s10781-018-9347-5.
King, Richard. “Early Yogācāra and Its Relationship with the Madhyamaka School.” Philosophy East and West 44, no. 4 (1994): 659–683. https://doi.org/10.2307/1399757.
Griffiths, Paul J. On Being Mindless: Buddhist Meditation and the Mind-Body Problem. Open Court, 1986. Bibliographic record.