One score hides six control problems

A person can become less distractible by changing the environment, tightening selection, noticing capture sooner, reducing elaboration, widening the field or entering a state in which ordinary sensory competition has altered. These are not interchangeable achievements.

The everyday word attention encourages a spotlight picture. A beam points at the breath, wanders, and is dragged back. This picture is useful for one problem: whether a selected object remains privileged. It says little about how the object was selected, what entered the surrounding field, how quickly wandering became known, whether the return was harsh or easy, and how much effort the system spent maintaining the beam.

Contemplative instructions often change several of those properties together. Research then labels the whole package “meditation”, measures time on target, and attributes any difference to a generic training effect. That is like evaluating two vehicles only by whether they reached the same address. One may have followed a stable road, another may have corrected repeated skids, and a third may have been carried on a train. Arrival does not identify the mechanism.

A minimal state description can begin with six dimensions. Sensory coupling asks how readily a sight, sound or bodily signal recruits processing. Target stability asks how continuously a selected object remains privileged. Meta-awareness asks how quickly a change in the attentional state becomes known. Aperture runs from narrow selection to broad monitoring. Reactivity measures how strongly an event triggers grasping, resistance or elaboration. Effort records the control cost required to maintain the current organisation.

These dimensions are a practitioner synthesis, not a hidden ontology of meditation. They are useful only if each becomes connected to a discriminating observation. Sensory coupling can be probed by unexpected but safe events. Stability can be estimated through target reports and behavioural variability. Meta-awareness can be approached through the latency between wandering and noticing. Aperture can be tested by peripheral detection. Reactivity needs measures of carry-over, rather than only detection. Effort should be separated from success because a session can be accurate and exhausting.

A state is not a trait

Write the current configuration as x_t = [C,S,M,A,R,E]. The letters denote sensory coupling, stability, meta-awareness, aperture, reactivity and effort at time t. A practice instruction is an operator P acting in a context k, so the next state is represented as x_t+1 = P(x_t,k) + ε. The term ε collects unmodelled variation such as fatigue, expectation and ordinary life events.

The notation prevents three common mistakes. First, a temporary state change is not automatically a durable trait. Second, the same operator can move different starting states in different directions. A narrowing instruction may stabilise a scattered field but intensify an already rigid one. Third, context is causal. Ten quiet minutes after adequate sleep are not the same input as ten minutes during acute stress, on retreat or after several nights of reduced sleep.

A useful study therefore measures at least three times: before the session, during the intended transition and after ordinary demands resume. The last point matters most for transfer. If target stability rises only while the audio is playing, the configured scaffold produced the result. If faster noticing survives a task switch or a later notification, the training may have changed a reusable transition. The unit of evidence is a state change plus its retention boundary.

flowchart LR
  A["The bell at minute twelve"] --> B["One score hides six control<br/>problems"]
  B --> C["Two traditions, no shared ladder"]
  C --> D["Training changes transition rates"]
  D -. "re-test" .-> A
  classDef first fill:#0a1424,stroke:#007D68,stroke-width:2px,color:#e6f3ff
  classDef step fill:#0a1424,stroke:#3f7da0,stroke-width:1px,color:#e6f3ff
  class A first
  class B,C,D step

Two traditions, no shared ladder

Yoga and Buddhist sources both describe disciplined changes in attention, yet they disagree about the structure of mind, the role of self, the interpretation of absorption and the end for which training is undertaken. Similar exercises do not erase those differences.

In the Yoga Sūtra, pratyāhāra appears after posture and breath regulation. Sūtra 2.54 describes the senses as no longer joined to their own objects and as following, as it were, the form of citta; 2.55 associates this with mastery of the senses.¹ The claim is subtler than closing the eyes. A closed eye may still be governed by remembered images, anticipated sounds or bodily discomfort. The text concerns the relation between sense capacities, their objects and the organisation of mind.

The next three terms are also relational. Dhāraṇā binds citta to a place. Dhyāna is the continuity of the same cognition there. Samādhi is described as the object alone shining forth, as though the practitioner's own form were absent. The three jointly form saṃyama in the following sūtra. A staircase image misses the changing structure: location, continuity and object-dominance are more than longer durations of the same act.

Early Buddhist texts organise attention within a different path. Sense restraint in DN 2 does not require that sights and sounds vanish. It asks the practitioner not to seize the themes and details through which unwholesome states gain purchase.⁵ The Satipaṭṭhāna discourse directs sustained contemplation of body, feeling, mind and dhammas with ardency, clear comprehension and mindfulness.³ The Ānāpānasati discourse uses breathing within a sixteen-step cultivation that includes bodily, affective, mental and insight-related dimensions.⁴

Modern research often pulls two training families from this larger field. Focused-attention practice selects an object, detects wandering, disengages and returns. Open-monitoring practice reduces explicit selection and monitors experience from moment to moment without elaborating each event. Lutz and colleagues proposed this contrast as a useful research taxonomy, not as a claim that all Buddhist systems divide practice exactly that way.⁶ Later taxonomies have shown why a single pair is insufficient: meditation methods differ in body involvement, imagery, affective cultivation, perspective and metaphysical framing as well as attentional mode.¹⁰

Mindfulness is not passive openness

The modern phrase “open monitoring” can suggest a camera that records whatever passes. The early Buddhist instructions are more active and selective than that image. The practitioner remembers the relevant frame, recognises qualities of body, feeling and mind, and cultivates or abandons states within a normative path. Even when no single sensory object is privileged, the practice has a purpose, a set of distinctions and a correction policy.

This matters experimentally. A participant instructed merely to “notice everything” may scan, wait, daydream or monitor for success. Another participant may recognise grasping as grasping and stop feeding it. Both produce broad reports, but only the second instruction specifies what should happen when an event becomes reactive. The difference can appear after the stimulus has gone, through reduced carry-over rather than improved detection.

Yoga's pratyāhāra also resists a passive reading. The senses are not praised for becoming weak. The associated claim is mastery, which implies that contact can be governed rather than merely lost. A useful perturbation therefore asks for reversible deployment: can the practitioner reduce recruitment and then detect a relevant signal when instructed? Trained non-capture should preserve the capacity to engage. Without that test, fatigue, dissociation and skilled restraint can look deceptively similar.

Practice instructions are operators, not labels

Consider the instruction “stay with the breath”. In one setting it means narrow selection at the nostrils, count departures and return. In another it means know the whole body while breathing, then become sensitive to feeling, mind and impermanence. In a third it may prepare entry into absorptive unification. The noun breath meditation therefore underdetermines the operation.

A practice record needs at least five fields: object, attentional operation, correction rule, intended transition and context. “Breath” names an object. “Return whenever wandering is noticed” names a correction rule. “Let all events be known without choosing among them” names a different operation. “Reduce external stimulation and remain available to interruption” names a context and boundary. Without these fields, comparing studies by technique name is often comparing packages with unknown active ingredients.

Training changes transition rates

A practice need not create a permanently different state to matter. It may change how often attention leaves, how quickly departure becomes known, how costly return feels, or how much of the next event inherits the previous reaction.

Focused-attention practice makes this visible because wandering is not an exception to the exercise. Hasenkamp and colleagues modelled a recurring sequence of mind wandering, awareness of wandering, shifting back and sustained focus.¹¹ The value of the model is not that four brain networks map neatly onto four inner acts. Its value is temporal. “Better attention” can result from fewer departures, shorter unnoticed departures, faster shifts, longer stable periods or some mixture.

Now change the instruction. In open monitoring, a sound or thought is not automatically a failure because it entered the field. The relevant transition is often from event to elaboration. Did the event become a story, a judgement, an impulse to suppress, or merely another transient occurrence? A probe designed for focused attention may therefore mis-score open monitoring. It treats breadth as wandering and can reward rigid narrowing.

Published studies support the modest claim that practice families can recruit different patterns and produce different task effects. Focused and open monitoring have been associated with differing attentional and connectivity profiles, while results vary with design, expertise and comparison condition.¹⁴ Reviews of focused-attention neurophysiology continue to find substantial heterogeneity in samples, tasks and signals.¹⁶ This is evidence against a generic meditation variable, not evidence that every traditional distinction has a unique neural signature.

Four ledgers prevent a theory victory

A convincing result needs four ledgers. The behavioural ledger records what changed in a task. The experiential ledger records what the participant noticed and how the transition unfolded. The physiological ledger records measured bodily or neural covariation. The instruction ledger records what was actually practised. None can silently substitute for another.

Suppose experienced practitioners show a distinctive connectivity pattern during open monitoring. The physiological finding is real within its method. It does not by itself show that participants were non-reactive, that the pattern caused the reported state, or that a traditional insight occurred. Demand characteristics, breathing, posture, expertise, expectation and selection into the sample remain rival generators. Conversely, a precise report without a perturbation may distinguish experiences while leaving mechanism open.

The strongest design makes the ledgers constrain one another. A stated reduction in elaboration should predict shorter behavioural carry-over after a distractor. If that relation disappears when instructions are swapped or when the same report is induced by suggestion, the original explanation weakens. If a physiological marker tracks the report but not the behavioural consequence, it may be a correlate of posture, arousal or reporting. Convergence earns force when each ledger rules out a different rival.

Order can become part of the intervention

If focused attention improves the reliability of noticing and returning, it may create a more stable platform for broad monitoring. If open monitoring reduces over-control or increases present-moment sensitivity, it may make later focus less rigid. Both predictions can be true for different starting states. Training order is therefore not a ceremonial inheritance. It is a causal hypothesis about prerequisites.

A small non-randomised study of thirty-three Japanese university students compared four weeks of focused attention followed by four weeks of open monitoring with the reverse order and a wait-list group. The order appeared to influence the process of skill acquisition more clearly than final questionnaire levels; several skills improved earlier in the focused-first sequence, while present-moment awareness improved earlier in the open-first sequence.¹⁵ The sample, measures and design do not justify a universal curriculum. They do make path dependence empirically discussable.

Absorption is a regime change, not a trophy

Absorption presents a harder measurement problem. Reports may include strong unification, altered effort, reduced discursive thought, changes in pleasure or equanimity, and diminished salience of ordinary sensory events. Yet traditions disagree about factors and stages, practitioners use terms differently, and advanced samples are small. A highly intensive 7T fMRI study of one adept across repeated sessions is scientifically valuable because it links dense phenomenology with repeated measurement. It remains a single-case, exploratory design.¹⁷

The state-space therefore treats absorption as a candidate regime, not the top value of concentration. A regime changes the relations among variables. High stability may become less effortful. Sensory coupling may fall without active suppression. The field may feel unified rather than narrow. These claims require post-state perturbations because an unresponsive or quiet channel can be caused by absorption, fatigue, demand compliance, drowsiness or simple non-report.

stateDiagram-v2
  [*] --> Claim
  state "A state is not a trait" as Claim
  state "Figure 1. A practice<br/>landscape, not a staircase" as Test
  state "Figure 2. One observed score,<br/>three causal generators" as Revision
  Claim --> Test: expose mechanism
  Test --> Decision: evidence survives
  Test --> Revision: rival survives
  Revision --> Claim: narrow or rebuild
  Decision --> [*]

Choose a trajectory, not a prestige practice

The right practice is not the one with the highest cultural status or the deepest reported state. It is the smallest, safest intervention that can test the current bottleneck without sacrificing capabilities the situation still requires.

A useful practice decision begins with a contrast. What happens now that should happen differently? “I want better attention” is too broad. “I notice email capture only after five minutes” points to meta-awareness. “I notice immediately but keep rehearsing the message” points to reactivity. “I can focus narrowly but miss relevant peripheral signals” points to aperture switching. “Every sound recruits me despite low emotional interest” suggests sensory coupling or environmental load. Each description implies a different measure and a different failure mode.

The next question is dosage. A ten-minute exercise, a daily hour and a silent retreat are not larger versions of the same causal input. Dose changes fatigue, sleep, social context, expectation and the chance that unusual experiences become self-reinforcing. A safe system therefore records minutes, frequency, recent escalation, setting, guidance and delayed effects. It treats adverse-effect monitoring as part of the practice, not an embarrassment after it.

That requirement is empirical. A systematic review found a pooled adverse-event prevalence of 8.3 per cent, with estimates varying sharply by study design and likely under-reporting in trials that did not ask systematically.²⁰ In three mindfulness-based programme variants examined with active harm monitoring, lasting negative effects occurred in 6 to 14 per cent of the sample and were associated with dysregulated arousal, including hyperarousal and dissociation.²¹ These figures should not be universalised. They show that “no complaint” is a weak safety measure when no one asks the relevant questions.

Every trajectory needs a recovery contract

A stop rule answers when to interrupt. A recovery contract answers what must be true before continuing. It should name the last known stable dose, the ordinary activities that indicate orientation, the person authorised to review the event, and the observations required before any restart. “Try again more gently” is not a contract because it leaves dose, evidence and authority undefined.

The recovery boundary should extend beyond the cushion. Sleep returning to baseline, ordinary conversation remaining coherent, work or care responsibilities being manageable and the person retaining genuine choice about practice are stronger release signals than a calm report immediately after a session. In research, the same logic requires follow-up windows and a record of withdrawals. Participants who stop because of harm cannot be erased from the denominator without changing the safety claim.

Recovery also protects interpretation. Reduced sensory engagement can be a trained and reversible state, or it can be unwanted detachment. Lower emotional response can be equanimity, exhaustion or blunting. The difference appears through voluntariness, context sensitivity, function and reversibility. A beneficial trajectory must retain a route back to ordinary agency.

The Attention Trajectory Workbench

The workbench below is a decision instrument, not a diagnostic or a teacher. It turns the state-space into explicit assumptions. Move the six sliders, choose a low-intensity practice family and inspect the illustrative direction of change. Then add safety observations. The model does not predict an individual outcome. It exposes what a recommendation is assuming and which measure should falsify it.

The browser tool uses a transparent directional model. It does not estimate clinical risk, attainment or individual benefit.

A smallest useful experiment

The lowest-risk research implementation compares two brief, ordinary-breathing practices while preserving normal life. It is not treatment, retreat training or an attempt to induce absorption. Exclude breath retention, forced breathing, sensory deprivation, sleep restriction, fasting and prolonged sitting. A person with current severe destabilisation, recent meditation-related harm or a condition for which inward attention has been clinically discouraged should not use a self-guided protocol.

A positive result permits a narrow statement: under this instruction and dose, this person showed a repeatable change in the predicted transition without observed short-term harm. A negative result means the chosen operator, measure, dose or state model failed to distinguish the conditions. It does not prove incapacity, invalidate a tradition or establish that all practices are equivalent.

from __future__ import annotations

from dataclasses import dataclass, asdict
from typing import Literal

Practice = Literal["focus", "wholebody", "open", "withdrawal", "absorption"]


@dataclass(frozen=True)
class State:
    sensory: float
    stability: float
    meta: float
    aperture: float
    reactivity: float
    effort: float

    def checked(self) -> "State":
        values = asdict(self)
        if any(not 0 <= value <= 10 for value in values.values()):
            raise ValueError("Every state value must be between 0 and 10")
        return self


DELTAS: dict[Practice, State] = {
    "focus": State(-0.4, 1.8, 0.8, -1.2, -0.4, 0.5),
    "wholebody": State(-0.2, 0.7, 1.2, 0.8, -0.8, -0.2),
    "open": State(0.2, -0.2, 1.5, 1.8, -1.1, -0.5),
    "withdrawal": State(-1.6, 0.4, 0.3, -0.4, -0.2, 0.2),
    "absorption": State(-2.2, 2.2, 0.2, -1.8, -1.0, -0.8),
}


def clamp(value: float) -> float:
    return max(0.0, min(10.0, value))


def apply_operator(state: State, practice: Practice) -> State:
    state.checked()
    delta = DELTAS[practice]
    return State(**{
        key: clamp(getattr(state, key) + getattr(delta, key))
        for key in asdict(state)
    })


def safety_verdict(*, sleep: bool, anxiety: bool, dissociation: bool,
                   trauma: bool, function: bool, compulsion: bool,
                   minutes: int) -> tuple[str, str]:
    if dissociation or trauma or function:
        return "red", "Stop and seek qualified support; use urgent local help for immediate danger."
    if sleep or anxiety or compulsion or minutes > 30:
        return "amber", "Pause or reduce dose, restore ordinary routines, and consult qualified support."
    return "green", "Continue only at the matched low dose with next-day readback."


if __name__ == "__main__":
    start = State(6, 4, 4, 5, 6, 7)
    print("before", asdict(start))
    print("after", asdict(apply_operator(start, "focus")))
    print("safety", safety_verdict(
        sleep=False, anxiety=False, dissociation=False,
        trauma=False, function=False, compulsion=False, minutes=10,
    ))

sequenceDiagram
  participant Q as Question
  participant M as Mechanism
  participant E as Evidence
  participant D as Decision
  Q->>M: The bell at minute twelve
  M->>E: A state is not a trait
  E-->>M: compare rival explanations
  E->>D: Govern transitions instead of ranking states
  D-->>Q: revise scope or act

Govern transitions instead of ranking states

The old decision asks how many minutes a person practised and whether concentration improved. The better decision asks which control variable needed to change, which operator was actually delivered, what transition occurred, what capability was traded away and whether the result remained safe after the session ended.

For practitioners, this means choosing the smallest practice that tests the current bottleneck. Someone who notices late may need a clear object and repeated return. Someone who notices immediately but elaborates may need a practice that changes reactivity. Someone with rigid narrow control may need safe aperture flexibility. Someone pursuing absorption needs competent lineage guidance, interruptibility and recovery measures, not a leaderboard.

For researchers, the independent variable must become a configured instruction rather than a meditation label. The outcome must become a vector and a trajectory rather than one attention score. The design must include rival generators, perturbations, learning order, delayed readback and active harm monitoring. A finding then licences a claim about a particular operator under stated conditions.

For teachers and programme designers, the adverse-effect register must be able to halt progression. Ordinary functioning, sleep, orientation and voluntary choice outrank completion of a curriculum. The decisive shift is from ranking states to governing transitions. Once that shift is made, Yoga and Buddhist practice can inform attention science without being flattened into one ladder, and science can test bounded relations without pretending to validate a tradition's final aims.