By the time a word reaches your ears, it has already traveled a great distance — down from somewhere wordless. Think about saying something true. First there is a wordless knowing, a felt sense with no shape yet. Then it begins to take form inwardly, a meaning starting to find its outline. Then the words assemble in your mind. Then, finally, your mouth moves and the sound comes out. The audible word is the last and grossest stage of a long descent. The tradition maps that descent as the four levels of speech.1
From the bottom up: Vaikharī, the gross, audible, spoken word — what comes out of the mouth. Madhyamā, the intermediate, the word in the mind, mental speech before it is voiced. Paśyantī, "the seeing" — a visionary level where the word is still a unified flash, not yet broken into mental language. And Parā, the supreme — the transcendent source, silent, the word before it is a word at all. Speech is a waterfall: it falls from the silent Parā at the top down through vision and mind to the audible splash at the bottom.2
The four-level scheme maps sound as a descent from the transcendent to the audible.3 The grammarian-philosopher Bhartṛhari worked out three levels (Paśyantī, Madhyamā, Vaikharī); the Śaiva tradition added the fourth and highest, Parā — the supreme, transcendent speech, the silent source from which the other three unfold.4
Read top-down, it is a cosmology of how the silent absolute becomes the spoken world:
So every audible word is the bottom of a four-rung ladder that reaches up into silence. And the practice direction reverses it: the contemplative ascends the ladder, tracing speech back up from the gross spoken word, through the mental and the visionary, to the silent Parā at the source — which is the Goddess as sound at her most subtle (the Goddess as sound).5
Why four levels, and why does it matter that the highest is silent?
Because manifestation works by descent — the subtle becoming gross, the unified becoming differentiated, the one becoming many. Speech is the perfect model of this, which is why the tradition uses it. At the top, Parā: one, silent, whole, nothing yet articulated. As it descends, it differentiates: the unified vision of Paśyantī splits into the mental words of Madhyamā, which harden into the physical sounds of Vaikharī. Each step down is a step into more differentiation, more grossness, more separation — the same movement by which the one Goddess becomes the many-thinged world (the alphabet-mothers birthing the world).6
The silence at the top is the crucial part. The source of all speech is not itself speech — it is silence, the Parā, the word before it is a word. This means the deepest level of sound is no-sound, the way the source of all forms is the formless. So when the contemplative ascends speech back to its source, they arrive not at the loudest, most powerful word but at silence — the pregnant silence that holds all words unspoken (the fourth of OM is silence). The fullness is quiet. The source of the waterfall is a still lake.7
This reframes what a word is. The audible word you hear is not the real word — it is the word's last, grossest precipitate, four steps removed from its silent source. Behind every spoken word stands its mental form, behind that its visionary flash, behind that the silence it fell from. To attend only to audible speech is to see only the splash and miss the whole height of the fall.
This page gives the vault the vertical structure of sound — the descent from silent source to audible word — which deepens the Goddess as sound (sound has levels, and the highest is silence) and Mātṛkā (the alphabet manifests at the Vaikharī level, its source higher up). It connects to OM (whose fourth, silent measure is the Parā) and grounds the practice of tracing speech to its source.
Outward, it offers a precise map of the descent from wordless source to spoken word — which illuminates the creative process, the gap between feeling and language, and the layered structure of how meaning becomes form.
The practice direction is where the scheme comes alive, so trace it carefully.8 Most of us live entirely at Vaikharī — the audible, the spoken, the level of words coming out of mouths and the constant inner chatter that is really Madhyamā (mental speech) running on autopilot. We take this surface for the whole of language and the whole of mind.
The contemplative ascent reverses the waterfall. You start at the gross spoken word (Vaikharī) and move up: first to Madhyamā, becoming aware of the mental word before it is voiced — catching speech in the mind. Then up to Paśyantī, the visionary level, where the word is still a single unbroken flash, not yet split into the sequence of mental language — a meaning seen whole before it becomes words. And finally toward Parā, the silent source, where the word dissolves back into the silence it came from. Each step up is subtler, more unified, quieter. The ascent is a journey from noise to silence, from the many to the one, traced along the thread of speech.
The case study reveals why sound is a vehicle in this tradition, not just a topic. Speech is the most available ladder back to the source, because everyone has it and uses it constantly. You do not need an exotic practice; you need to trace the word you are already speaking back up its own descent — through the mental, the visionary, to the silence. Mantra works partly this way: a mantra repeated long enough can carry the attention up the levels, from gross repetition (Vaikharī) toward the silent source (Parā). The four levels turn ordinary speech into a map home.
You are caught, as almost everyone almost always is, in the constant audible-and-mental chatter — Vaikharī and Madhyamā running nonstop. The four-levels practice offers a way to climb out of the noise by going up the ladder rather than trying to silence the bottom by force.
You begin by noticing the mental word before it becomes spoken — catching speech one level up, at Madhyamā, watching thoughts form as words in the mind. Then you reach for the level above that: the moment before a thought has become mental words, when it is still a single unformed flash of meaning (Paśyantī). You are trying to catch the word higher and higher up its own fall, closer to the silent source. And at the edge, you rest toward Parā — the silence from which all the words are falling, the still lake above the waterfall. You are not fighting the chatter at the bottom; you are climbing above it, tracing speech to the silence it comes from. The noise quiets not because you suppressed it but because you went upstream of it.
The main failure is living entirely at Vaikharī — taking the audible and the mental chatter as the whole of language and mind, never suspecting the levels above, mistaking the splash for the fall. The tell is total identification with the constant inner-and-outer talk, no awareness of where it comes from.
A second failure is trying to silence the chatter by force — battling the bottom level head-on, which rarely works, instead of climbing above it by tracing speech to its source. The four levels offer the upstream route the head-on fight misses.
A third failure is mistaking the source for the loudest word — assuming the deepest level of sound must be the most powerful sound, missing that the source is silence, the Parā, the word before it is a word. The fullness is quiet, not loud.
The four-level scheme (Vaikharī, Madhyamā, Paśyantī, Parā), Bhartṛhari's three levels, and the Śaiva addition of the fourth are well-established in the Indian philosophy of language and Kashmir-Śaiva metaphysics; the teacher transmits them accurately.9
A tension worth noting: the scheme is sometimes read metaphysically (the four levels are real cosmic strata of how the absolute becomes the world) and sometimes psychologically/phenomenologically (they describe stages in the actual production of speech in a mind). The teacher's contemplative use leans on the second (you can trace your own speech up the levels), but the cosmological claim is the first (speech descends from the silent absolute). The page holds both — they reinforce each other — but they are distinct claims, and the experiential availability of the lower transitions does not by itself establish the metaphysical reality of Parā.
Open question: is Parā (the silent supreme speech) something a contemplative actually reaches — a real attainable level — or a posited limit, the silence you approach but never arrive at, always one step above wherever your attention has climbed?
Selvalingam draws the four levels from the grammarian tradition (Bhartṛhari) fused with Kashmir Śaivism (the added Parā), converging with the whole Indian śabda-brahman tradition that treats sound/word as the fundamental reality and speech as the ladder between the absolute and the manifest. There is a productive tension between Bhartṛhari's grammarian-philosophical framing (a theory of how language and meaning work) and the Śaiva contemplative-metaphysical framing (the levels as strata of consciousness to be ascended). Bhartṛhari is largely explaining language; the Śaivas are giving you a path. Selvalingam uses both — the philosophy of language grounds the contemplative map — which is the characteristic Tantric move of turning a metaphysics into a practice (as with the Mātṛkā).
The plain bridge: there is a real distance between a wordless knowing and the spoken word, and the journey from one to the other passes through stages — which anyone who has tried to say something true has felt.
creative-practice — Great Writers Master Language: every writer knows the descent — there is a wordless sense of what you mean (something like Paśyantī, a meaning seen whole), and the work is bringing it down through mental drafting (Madhyamā) into the finished words on the page (Vaikharī), losing and finding things at each step. The four levels name what writers do without naming. Read together, the two show that composition is the controlled descent of the waterfall — and that the writer's frustration ("the words never quite catch it") is exactly the gap between the unified flash above and the differentiated words below, which the scheme says is structural, not a personal failing.
psychology — Alexithymia and Speechless Terror: there are inner states that have not descended into words — felt, real, but stuck above Vaikharī, never finding mental or spoken form, and therefore unbearable and unworkable. Alexithymia is the inability to bring feeling down the levels into nameable language. The four-level map gives that clinical picture a structure: health involves the flow from the wordless felt sense down into named, speakable form, and its blockage leaves states trapped at a pre-verbal level. The handshake suggests therapy is partly teaching the descent — helping a feeling fall, safely, into a word.
business / AI — Transformer Architecture & Language-Model Mechanics: a language model has its own layered descent — a high-dimensional internal representation (something like a unified meaning-state) that gets decoded, step by step, into the sequence of surface tokens, the audible word. The internal "meaning" is not the same as the output text; the text is its last, grossest precipitate. The handshake offers a startling structural echo: the ancient four-level descent from silent source to spoken word and the model's descent from internal representation to output tokens share a shape — meaning at the top, differentiated symbols at the bottom — even as it neither proves the metaphysics nor reduces it.
The Sharpest Implication If the audible word is the bottom of a four-rung fall from silence, then almost everyone — you included, almost all the time — lives at the very bottom of language, mistaking the splash for the whole waterfall. The chatter you take for your mind is Vaikharī and Madhyamā, the two grossest levels, running on autopilot, while the visionary and silent sources above go unnoticed your entire life. The unsettling implication is that there is a whole vertical dimension to your own speech and mind that you have never climbed — that "below" your noise is not more noise but, going up, increasing silence, and that the source of all your words is a stillness you have never visited. And the practical edge: you cannot quiet the chatter by fighting it at the bottom. You climb above it. The silence you are looking for is not at the end of the noise; it is upstream of it, at the top of the fall, where the word has not yet become a word.
Generative Questions
Wallis gives the four levels of the Word their Trika metaphysics: Parā-vāk, the Supreme Word, is the deep vibratory matrix from which both words and the manifest world arise, descending through paśyantī (the visionary level where the deepest stories are rooted) and madhyamā to vaikharī (audible speech). His bhāvanā-krama explicitly walks a teaching-passage down these levels — literal/vaikharī, pondered/madhyamā, meditatively-absorbed/paśyantī — as the method of vikalpa-saṃskāra.wallis See Parā-Vāk, Paśyantī, and Vikalpa-Saṃskāra.
A third source has picked up this exact ladder and run it two different directions, neither of them Wallis's meditative bhāvanā-krama and neither of them Selvalingam's cosmology. Content creator Prathamesh Krisang takes the same four rungs — Parā, Paśyantī, Madhyamā, Vaikharī — and treats them as a working production pipeline for on-camera speech rather than a metaphysical map to be contemplated.
The fullest version of this lives on its own page: Parā to Vaikharī: Speaking from Feeling for Creators. Krisang's claim is that the standard advice for talking-head content — write a script, memorize it, perform it — runs the ladder backward, forcing Vaikharī (the spoken word) to be faked without ever passing through Parā (the felt source) at all. His fix is to run the sequence in its real order: feel, then let an image form, then let the image become a mental sentence, then speak — and he claims the whole cascade completes in a literal split second for a fluent speaker, fast enough that voice, face, and hands modulate on their own without separate acting or delivery training.ks-para
A second, narrower application shows up in a different talk: Reading as Madhyamā: Why "Just Listen, Don't Take Notes". Here Krisang runs the ladder in reverse, applying it to reception rather than production. His claim is that reading was never a separate, silent, visual channel outside this scheme — that text on a page is converted into sound at the level of the mind, meaning reading is Madhyamā entered from a different door, not something the four-level map skips over.ks-read The practical instruction built on that claim is blunt: don't take notes while you're being taught something meant to be received this way, because the act of transcribing competes with the same Madhyamā-level processing the instruction is trying to protect.
Set beside this page's own tension — is Parā a level a contemplative genuinely reaches, or a limit only ever approached — Krisang's applied voice-work quietly takes a position the more contemplative framing here never commits to. He treats Parā as something reached constantly, in fractions of a second, by anyone speaking or listening authentically, not as a distant asymptote requiring years of ascent. That's a real divergence in altitude, not just application: where this page holds the descent as cosmology first and practice second, Krisang collapses the two, betting that the structure is genuinely available on demand rather than only after sustained contemplative work. Both readings can be true of the same four-level scheme without either one canceling the other — a structure this old and this well-attested across the tradition can support a fifteen-second video and a lifetime of meditation at the same time.