Verses 145 to 148 of the Śiva Svarodaya set out something I did not expect to find in a tantric manual: a formal measurement scheme. Five elemental phases, eight dimensions, and the instruction that every phase gets a value on every dimension.
I spent a fortnight building this part of the tool and most of that time was spent on one question. The scheme asks you to combine seven observations into a single reading. Three of those seven cannot be checked by anyone, including you. So what does the combined reading mean?
The answer, I think, is that it means less than any one of its parts. That is not a complaint about old texts. It is the same structural problem that clinical trials have been criticised for since the 1990s, and the fix is identical in both cases.
The scheme is better built than its reputation
Strip the cosmology and what verses 145 to 147 specify is a description matrix. Eight dimensions are named in order: the enumeration of the phases, the junctures of the breath, the signs of the breath, their location in the body, their colours, the vital force, their associated tastes, and the characteristics of their movement.
The dimensions are deliberately heterogeneous. Some are temporal, some spatial, some sensory, some physical. And the commitment is total — every phase receives a value on every dimension, not only on the dimensions that happen to suit it.
That commitment is the clever part. It enforces completeness, so no category gets characterised only where the evidence is convenient. It enables comparison, because you can set two phases side by side dimension against dimension. And it surfaces its own failures: a category that cannot be assigned a value on a required dimension is a problem the scheme itself makes visible.
The nearest modern thing I know is the way psychology characterises affective states. Russell’s circumplex model, published in the Journal of Personality and Social Psychology in 1980, argues that affect is best described not as a list of independent feelings but as position in a two-dimensional space of pleasure and arousal, with every state receiving a value on both axes. The Svarodaya is doing the same kind of work with more axes and no statistics.
I want to be clear that this is a structural compliment, not an empirical one. A well-built matrix filled with wrong values is still wrong. But the architecture is sound, and it is sound in a way that makes the contents testable — which is more than most of what gets written about the breath.
Four dimensions you can actually check
Of the seven observable dimensions, four can be checked by anyone with a mirror, a hand and a clock.
Which nostril is open is verse 153, and it is the only external instrument named in the whole treatise. Breathe onto a cool mirror held a few inches below the nose, and two patches of condensation form. The physics is ordinary: exhaled air is saturated with water vapour at roughly body temperature and condenses on a cooler surface, and the patch it leaves reflects the geometry of the jet that made it. The larger patch marks the more patent nostril.
The reach of the exhalation is verse 158, and it is the most interesting line in this section. Hold the back of the hand in front of the nostrils and move it away until you can no longer feel the breath. Note the distance. That is the only quantitative protocol in the entire text, and what it produces is a crude index of expiratory flow, obtained with no equipment at all.
The direction of the jet is verse 155 — centre, downward, upward, oblique. The junctures of the breath is verse 146, which amounts to rate and pause length, and is simply countable.
Three dimensions that exist only in the report
Location in the body, taste, and the colours seen under ṣaṇmukhī are different in kind. Nobody else can check any of them, and neither can you, in the sense that there is no second reading to compare yours against.
Verse 156 assigns each phase a position: fire at the shoulders, air at the base of the navel, earth at the knees, water at the feet, space in the head. There is no physiological reading I can defend. The locations do not map onto autonomic territories or any known pattern of referred sensation, and the ordering looks like a simple vertical logic — heaviest lowest, subtlest highest — with fire at the shoulders as an anomaly, probably because heat is associated with the upper body. It is schematic ordering rather than observation.
Verse 157 assigns tastes, and is more interesting than it first looks, because the tastes named are the standard rasa categories of Āyurvedic pharmacology. Their appearance here tells you the Svarodaya is drawing on the medical rather than the ritual side of its inheritance. As a diagnostic it is untested. As a thing to notice it has something going for it: resting oral sensation does shift with autonomic state. Frederiksen and colleagues, writing in Oral Surgery, Oral Medicine, Oral Pathology and Oral Radiology in 2024, measured unstimulated salivary flow in forty healthy adults and found it dropped significantly during a mental arithmetic stressor compared with a non-stressful speaking task. That is one crossover study in forty people, so I would not call dry mouth a reliable index of anything. I would call it a real signal worth writing down.
The colours are verses 151 and 152, seen inside ṣaṇmukhī mudrā, with the fingers closing ears, eyes, nostrils and mouth. Something does appear. Reducing patterned visual input reliably produces visual phenomena in healthy people: Merabet and colleagues blindfolded thirteen sighted volunteers for five days and reported in the Journal of Neuro-Ophthalmology in 2004 that ten of them experienced visual hallucinations, from bright points of light up to faces and landscapes, with full insight that none of it was real.
I should be honest about the gap there. That study is ninety-six hours of total visual deprivation. Ṣaṇmukhī is three minutes with your fingers on your eyelids. The mechanism is plausibly the same family and the doses are nowhere near each other, so the study tells you the phenomenon is real in principle, not that the specific colours of verse 152 are what three minutes produces. Nobody has run that experiment.
The flaw is not the soft three
Here is the part I kept circling back to. Three unverifiable dimensions in a seven-dimension scheme is not, by itself, a fatal problem. Plenty of good instruments include self-report. The problem is what the scheme does next: it asks you to merge all seven into one elemental verdict.
That is a composite outcome, and composite outcomes built from components of unequal quality are a well-documented way to mislead yourself.
Cordoba, Schwartz, Woloshin, Bae and Gøtzsche reviewed this for the BMJ in 2010. They took forty randomised trials published in 2008 that used a binary composite as their primary outcome, and had blinded observers judge whether the components were of comparable clinical importance. The results are worth sitting with.
Components were judged not of similar importance in 28 of the 40 trials. In 20 of those, death was pooled with hospital admission — as though the two were interchangeable events. The composite’s own definition changed between abstract, methods and results in 13 trials. Only one trial gave a good rationale for its choice of components. And in 11 of the 16 trials where the composite reached significance, the abstract’s conclusion implied the benefit applied to the most important component, which the data did not show.
The mechanism is simple enough to state in a sentence. When you pool a strong component with a weak one, a result driven entirely by the weak component reads as a result about the strong one. The good items lend their credibility to the bad.
That is exactly what a merged elemental reading does. If your composite says “fire” and the composite was driven by a taste you reported, the reading borrows authority from the mirror patch and the counted breath rate, neither of which contributed. You cannot tell afterwards which dimension decided it, because the scheme never asked you to record them separately.
What disaggregation gets you
The fix in trials is to report components separately as well as pooled. The fix here is the same, and the Svarodaya does not do it. That is my judgement about the text, not something the text says about itself, and I want to be careful about the difference. I have described this before as the text warning against the move. Reading it again, I do not think that is right. The text sets out the matrix and treats the combined reading as the product. The discipline of keeping verifiable and unverifiable dimensions apart is something I am importing.
Imported or not, it is what the tool does. The four measurable dimensions score. The three reported ones are recorded, displayed, and marked “not scored.”
The reason this is worth the extra interface is that it turns a dead end into a question. If the self-reported dimensions turn out to track the measured ones over a few weeks of records, that is a finding, and an interesting one. If they do not, that is a more important finding — and a merged instrument conceals it permanently, because there is no separate column to compare.
| Merged reading | Disaggregated record |
|---|---|
| One verdict per observation | Seven values per observation |
| Cannot say which dimension decided it | Can say exactly which decided it |
| Soft dimensions borrow hard credibility | Each dimension carries its own weight |
| Agreement between dimensions is assumed | Agreement is measurable |
| Nothing to compare next month | A series per dimension |
Why the classification has to be blind
There is a second discipline the mirror test needs, and it is the one I expect people to skip.
Verse 154 says to read the elemental phase from the shape of the condensation patch — square for earth, crescent for water, triangle for fire, circle for air, scattered points for space. I have left that out of the tool entirely, and the reason is not that the shapes are ancient. It is that a condensation patch is an ambiguous stimulus, and the person classifying it already knows which answer the calculation expects.
The size of that effect is measurable. Hróbjartsson and colleagues, in CMAJ in 2013, found 16 randomised trials in which the same outcomes had been assessed by both blinded and non-blinded assessors — 2,854 patients in total, so the comparison is within-trial and the only thing differing is the assessor’s knowledge. The non-blinded assessors produced more optimistic estimates, exaggerating the blinded effect size by 68%, with a confidence interval running from 14% to 230%.
These were trained clinical assessors rating defined outcomes. A practitioner squinting at a patch of fog on a mirror, hoping for a triangle, is in a considerably weaker position.
So if you want to test the shape assignments, there is one protocol that carries information: produce the patch, photograph it, and classify the photograph later without knowing what the breathing state was at the time. Classification made in ignorance of the expected answer is the only classification worth having. I would read that study with real interest. I am not going to build an interface that invites people to do the unblinded version.
The four measurable ones are noisier than they look
Having argued for the measurable four, I should say how far I would trust them, which is not very far.
Take the mirror. The physics is sound and the patch is a real trace, but as far as I can find, nobody has validated mirror condensation against rhinomanometry or acoustic rhinometry. It is a plausible measurement, not a validated one.
What has been tested is the thing the mirror is meant to replace: your own sense of which side is blocked. Chin, Malek, Pratt, Marcells, Sacks and Harvey measured 145 patients by anterior rhinomanometry and compared the result with what each patient said, reporting in the Journal of Laryngology and Otology in 2014. Sensitivity was 86.9% and specificity 41.1%. The positive predictive value was between 54% and 59% depending on the threshold used.
Read that carefully, because it cuts both ways. People are reasonably good at noticing when a side is genuinely obstructed. They are poor at not reporting obstruction when the sides are objectively symmetrical — and in that study 31% of patients were symmetrical. If you ask someone which nostril is dominant, they will usually give you an answer, and when there is no real difference the answer is close to a coin toss.
That is an argument for the mirror, not against it. It is also an argument against the interoceptive certainty that swara yoga tends to cultivate.
The reach of the exhalation has the same shape of problem one level down. Nobody has validated hand-feel distance against a flow meter either, and the instrumented version of the measurement is itself noisy. Hoang and Ho measured peak nasal inspiratory flow in 300 healthy Vietnamese adults and published normal values in Tạp chí Y Dược Huế in 2025: 158.8 ± 41.0 L/min in men and 113.4 ± 31.9 in women. A standard deviation of about a quarter of the mean, in a healthy population, with age and height both significant predictors. They also found the third of three attempts gave the highest average, significantly higher than the first and second, which is a learning effect in a measurement that takes one breath.
If a calibrated meter behaves like that, a hand in the air will behave worse. The reach is a real variable and I would record it. I would not read much into a single morning’s value, and I would not compare my number with yours.
What I kept
The tool scores the four measurable dimensions, records the three reported ones without scoring them, and does not read shapes off the mirror at all.
What I am after is not a better oracle. It is a record that could disappoint me. Seven columns, filled in the same way each morning, each with a fixed criterion, kept for long enough that a pattern would have to show itself rather than be found. Most self-tracking fails not because the wrong variables were chosen but because the variables shifted — recorded when convenient, on dimensions that drifted, with the criteria for each rating never pinned down.
The Svarodaya already solved that part. It fixed the dimensions, fixed their order, and demanded a value on each. Nine hundred years later, the thing I had to add was the discipline of keeping the checkable columns apart from the ones only I can see.
That is a smaller claim than the text makes. It is the one I can defend.
Sources
- Cordoba G, Schwartz L, Woloshin S, Bae H, Gøtzsche PC. Definition, reporting, and interpretation of composite outcomes in clinical trials: systematic review. BMJ (2010), 341:c3920. doi.org
- Hróbjartsson A, Thomsen ASS, Emanuelsson F, Tendal B, Hilden J, Boutron I, Ravaud P, Brorson S. Observer bias in randomized clinical trials with measurement scale outcomes: a systematic review of trials with both blinded and nonblinded assessors. CMAJ (2013), 185(4):E201–E211. doi.org
- Chin D, Malek J, Pratt E, Marcells G, Sacks R, Harvey RJ. Patient self-assessment in discriminating the more obstructed side in nasal breathing. The Journal of Laryngology and Otology (2014), 128(Suppl S1):S34–S39. doi.org
- Hoang PM, Ho DA. Peak nasal inspiratory flow: normal values in a healthy Vietnamese adult population. Tạp chí Y Dược Huế (2025), 15(6):56–59. doi.org
- Merabet LB, Pascual-Leone A and colleagues. Visual hallucinations during prolonged blindfolding in sighted subjects. Journal of Neuro-Ophthalmology (2004). doi.org
- Frederiksen EH, Rojas Ramirez MV, Moreno-Hay I, Miller CS, Carlson CR. Whole unstimulated salivary flow rate decreases during acute stressful condition. Oral Surgery, Oral Medicine, Oral Pathology and Oral Radiology (2024), 137(6):594–603. doi.org
- Russell JA. A circumplex model of affect. Journal of Personality and Social Psychology (1980), 39(6):1161–1178. PDF