Say "Kodak" and "Lumi" out loud. One snaps; the other glows. You knew that before you considered what either word means — because the meaning is incidental. The sounds did the work.
This is sound symbolism: the documented tendency for speech sounds to carry feeling independently of the words they form. It is not folklore. In the bouba/kiki effect, people across roughly 25 languages overwhelmingly assign the made-up word "bouba" to a rounded shape and "kiki" to a spiky one. Edward Sapir's 1929 experiments showed the same with size — "mil" reads as smaller than "mal" to speakers who have never seen either word. The phenomenon is a form of iconicity: a non-arbitrary link between how a word sounds and what it evokes.
For a brand name, this matters before the first definition is processed. The question is not whether sounds carry feeling — that is settled — but whether we can measure that feeling precisely enough to be useful. We can.
PhonoPair already captures one slice of this signal: a single sharp-versus-soft scalar between two words, derived from the bouba/kiki research. Phonosemantic syllable analysis generalises that one number into six measured dimensions, each grounded in a specific body of sound-symbolic research.
Sharpness — driven by consonant type. Plosives and sibilants read as sharp; sonorants (nasals, liquids, glides) read as soft. This is the bouba/kiki axis, made explicit.
Brightness — driven by vowel position. Front vowels (ee, ih) read as bright and small; back vowels (oh, oo) read as dark and large.
Weight — driven by vowel height and length. Low, long vowels feel heavy; high, short vowels feel light. This is Sapir's "mil/mal" axis.
Energy — driven by voicing. Voiceless stops and sibilants feel energetic; voiced sonorants feel calm. Yorkston and Menon's consumer-judgment work maps directly onto this.
Openness — driven by syllable structure. A syllable ending in a vowel (an open syllable) feels expansive; one closed by a consonant feels contained.
Tension — driven by vowel quality. Tense vowels feel deliberate and sustained; lax vowels feel relaxed and quick.
Each dimension is scored from −1 to +1. None is "good" or "bad" — the point is fit. A meditation brand wants the opposite profile of a power-tool brand, and the score should describe, not judge.
The feeling of a name is not uniform across it. It moves. To see that movement, the analysis breaks each word into syllables and each syllable into its three structural parts: the onset (the consonants before the vowel), the nucleus (the vowel itself), and the coda (any consonants that close it).
English does not mark syllable boundaries explicitly, but they are recoverable from the phoneme sequence using the maximum onset principle — a deterministic rule, not a guess. Working from the arpabet phoneme codes in the CMU Pronouncing Dictionary, the same word always splits the same way.
Take "Prozac". It resolves to two syllables with very different personalities:
PRO — open syllable, back vowel: dark, calm, expansive, low energy. ZAC — closed syllable, sibilant onset, front vowel: bright, sharp, contained, high energy.
The name opens warm and lands hard. That contrast is not a defect — it is arguably the entire reason the name works. A flat, uniform name would not have that arc.
Single-word feeling is only half the story. Most brand names are more than one word, and the more interesting question is how the feeling travels across the whole name — what we call the contour.
For a multi-word name, the analysis computes each word's profile and then the trajectory between them: does sharpness rise, fall, stay flat, or oscillate? Do the words reinforce one coherent feeling, or pull against each other? "Red Bull" builds momentum — short, energetic, closed, getting sharper. A name whose words fight each other on every dimension reads as incoherent even when each word is fine alone.
This is also where the analysis connects to scoring. In its first release the contour is shown as insight, not as points — it explains *why* a name feels the way it does without changing the number. Once the model is validated against known names, contour congruence becomes a candidate scoring factor in its own right. The sequence is deliberate: measure first, prove it tracks human judgment, then let it move the score.
Two things stay true to how PhonoPair already works.
First, it is deterministic. The six dimensions are computed from documented phoneme-to-feeling mappings, not generated by a model. The same name produces the same profile every time, and every value traces back to a specific sound and a specific piece of research. There is no language model anywhere in this.
Second, it consolidates rather than bolts on. The existing bouba/kiki signal is not left running beside a second, parallel sound-feel heuristic — that would risk double-counting the same effect. Instead, bouba/kiki becomes a special case of the six-dimension engine: one composite of sharpness and roundedness, computed from the same source of truth as everything else. Fewer moving parts, one explanation, no contradictions between two systems measuring the same thing.
The phonosemantic profile is a lens on the existing phonetic pillar, made legible — not a new pillar competing with it.
1. Decide the feeling before you judge the name. A wellness brand and a fintech brand want opposite profiles. Know which arc you are looking for, then read the contour against it.
2. Watch the landing syllable. The last syllable disproportionately shapes the impression a name leaves. A soft open ending and a sharp closed ending are different promises.
3. Check that the words agree. In a multi-word name, words that fight each other on sharpness or brightness produce a name that feels incoherent even when each word tests well in isolation.
4. Use the contrast deliberately. "Prozac" works because of its arc, not in spite of it. A perfectly uniform name is often a forgettable one.
5. Say it, then read the profile. The fastest gut-check is still your own ear — but the syllable breakdown tells you *why* your ear reacted the way it did, which is what you need when you have to defend the choice to someone else.
See the phonetic profile of your own name candidates