Three to five minutes a session. The map takes weeks.
No account · No payment · No data collected
No outside requests: the typefaces are served from this site. Your sessions never leave the device.
Most of the signal you receive in a day is about other people: their stories, their agendas, their content. Without a fixed reference point you drift with it, and after a while you mistake the noise for your own voice. V3KTOR builds the reference point. Measured, not narrated.
One word at a time. Positive, tap right. Negative, tap left. Your thumb answers before your story does, and the time it takes is the measurement.
Every word carries the crowd's values for how positive, how intense and how dominant it feels (NRC VAD Lexicon, National Research Council Canada). Your answer is set against them. Where you agree is the shared floor. Where you part ways is you.
The engine adapts. It serves more from what you react to quickly, spaces its follow-ups so one answer can't pull the next, places a turning point after a dark run, and remembers which words it chose for you so they count less than the ones you met cold.
The timing is corrected for what is known to distort it on a phone: how you hold it, the length of the word, the pull of the previous answer, fatigue within the session, a foggy day. What remains is read through two standard models of decision-making: how clearly the evidence came, how much you demanded before committing, how far you lean toward yes.
Before each session you say how you feel, in four taps and an optional line of context. The test records what your thumb does. The gap between the two is kept, because it has a shape.
One session is a reading. Each one after it raises the resolution. Words met again confirm or contradict. Contexts that flip a word show themselves. Foggy sessions count less toward speed; what you answered on them counts in full. Sittings on different days in different states count most.
The map checks itself. Every ten sittings close a block; from twenty on, each new block is checked against the one before it, and the map tells you in one line how much to trust it. Before that, a provisional figure stands in. The thresholds can't be lowered. A map that takes months to confirm is one you have watched form.
What comes out: what you affirm, what you refuse, the exceptions that hold, and a file that carries its own confidence. Paste it into a language model for a reading, or show it to an AI so it stops guessing who it is talking to. The map is yours. Nothing here is built for a machine to steer by.
Not a clinical tool, and not a copy of anyone. None of the parts are new: the word values, the frequency norms, the reaction-time paradigm, the decision models were all published, and are used here as published. What is new is the assembly, for one person, over months, on a phone: a map of what you answer before you explain, the layer under the story you tell others and yourself, measured rather than narrated. No one has tested the whole. That test is the one you run by using it.
NRC VAD Lexicon v2.1 (Mohammad, 2018) · the valence, arousal and dominance value of every word; the crowd your answers are set against. 516 word values, scaled to 0–1, used with attribution under NRC's terms of use for research and non-commercial purposes; the lexicon itself is not redistributed here, obtain it from NRC
SUBTLEX-US (Brysbaert & New, 2009) · how common each word is; common words are read faster whatever they mean, and that is removed before speed is read
Pleasure–arousal–dominance (Mehrabian & Russell, 1974; Russell, 1980) · the three axes
EZ-diffusion (Wagenmakers, van der Maas & Grasman, 2007) · splits your timing into how clearly the evidence came, how much you demanded before committing, and plain reading-and-thumb time
Signal detection theory (Green & Swets, 1966; Hautus, 1995) · how sharply you separate positive from negative, and which way you lean
Cohen's kappa (1960) · whether two blocks of sittings agree beyond what chance would give; the number behind STABLE
Weighted sampling without replacement (Efraimidis & Spirakis, 2006) · how the engine chooses which words you meet, by what the map still needs
Median absolute deviation (Hampel, 1974; Leys et al., 2013) · how a lapse is told from an answer
Abramowitz & Stegun (1964) · the inverse-normal approximation inside the detection numbers
findings the design answers to
Lexical decision: length and frequency (Balota et al., 2007; New et al., 2006) · why reading time is corrected before a fast word is called a reflex
The Implicit Association Test and its retest critiques (Greenwald, McGhee & Schwartz, 1998; Lane et al., 2007) · why the thumbs are measured separately, and why agreement is corrected for chance
Affective priming (Fazio et al., 1986; Klauer & Musch, 2003) · the previous word colours the next; here that is measured, not designed away
Post-error slowing (Rabbitt, 1966; Dutilh et al., 2012) · the answer after a conflict, and after a slip
Automatic vigilance (Estes & Adelman, 2008; Larsen et al., 2008) · dark words are slow for everyone; the baseline measures you against your own speed in each direction
Sequential effects in two-choice tasks (Kirby, 1976; Soetens et al., 1985) · carry-over from one answer to the next
Repetition priming (Scarborough et al., 1977) · why a word met an hour ago is not met again today
Acute stress and decision style (Starcke & Brand, 2012; Hales et al., 2016) · the premise of pressure mode
Adult trait stability (McCrae & Costa, 1994; Roberts & DelVecchio, 2000; Vecchione et al., 2016) · the premise that there is a floor to find
Timing precision on the web and on touchscreens (Pronk et al., 2020; Anwyl-Irvine et al., 2021) · why the clock runs from the word being fully shown to finger-down, and why the file says how much to trust the milliseconds
Jittered intervals against rhythmic responding (standard practice; Luce, 1986) · why the pause before each word varies
Brightness, valence and a fixed mapping (Meier, Robinson & Clore, 2004; Meier, Fetterman & Robinson, 2015; Lakens, Semin & Foroni, 2012; Schneider & Shiffrin, 1977) · dark reads as bad and light as good, most of all with both in view, and a side that never changes becomes reflex; why negative is the dark side, positive the light, and they never swap
Hand and hemisphere (Poffenberger, 1912; Marzi et al., 1991; Johnstone & Carey, 2016) · each thumb is run by the opposite half of the brain, crossing between halves costs a few milliseconds, and which thumb starts sooner can't be read from handedness; why each thumb is timed before the words
studies of the whole
None. The assembly is untested. Every file it produces says how much of it has been confirmed, and by what.
Your call.
tap anywhere to advance
V3KTOR
// Mapping the Self
Free · No ads · No sign-in · No tracking
HISTORY · tap misread if you answered a different word than the one shown
BEFORE EACH SESSION
Four rows: how light or heavy you feel, how calm or wired, how much on top of things, how much energy. Tap one tile in each. It takes five seconds.
This is what you say. The test measures what your thumb does. The gap between the two is one of the things the map is built from.
The line underneath is for context the rows can't hold: three coffees, slept badly, won the argument. Optional. It is never scored; it is there so whoever reads the session, you included, doesn't have to guess.
RIGHT NOW
Tap one in each row · We compare it to your reflex · Optional
READ BEFORE STARTING
NOT A DIAGNOSIS
This tool is not a medical or psychological instrument. It does not replace professional consultation. Results are indicative only and should not be used as a basis for health decisions.
YOUR DATA IS YOURS
No data is transmitted. Everything stays on your device. You can erase everything at any time.
IF IN DISTRESS
You are not alone. Contact a health professional or crisis helpline. This tool is not designed to respond to a crisis situation.
By tapping CONTINUE, you acknowledge that you have read and understood this information, and that you are using this tool of your own accord.
tap to continue
SETTINGS
TEST
NUMBER OF WORDS60
60 words is the shortest sitting the baseline accepts; 100 is the default. Longer sittings meet more of the pool per session.
WORDS PER SET30
BREAK DURATION5s
MAX RESPONSE TIME2000ms
2000 ms for a first session; 1200–1500 ms once answers are routine. Answers that miss the window are kept as unanswered, which is its own category.
RESEARCH MODE
Extended capture: top 10 per direction instead of top 5. Locks parameters: 120 words min, 60 per set, 1700ms max. Meant for after the first few sittings.
ESTIMATED DURATION—
APPEARANCE
BRIGHTNESS50
Increase if your screen looks too dark.
PREFERENCES
LOCAL STORAGE
Storage is disabled. Your sessions will not be saved.
TUTORIAL AND INTRO
The intro and tutorial will be skipped. The test starts directly after the countdown.
MOTOR CHECK
Eight arrows before the first word, tapped on the side they point to, measure each thumb's plain movement time. Off, the yes/no speed difference can't be split into hand and decision.
FIXATION DOT
The pause before each word varies in length so the thumb can't fall into a rhythm. On, a small dim dot marks the centre during it; off, the screen is blank.
PRESSURE MODE
The engine chases what you react to, with no spacing and a looser cap (its queue can hold up to 60% of the session, against 30%) — the way the first version did. Same words, same timing. Sessions are tagged PRESSURE, kept in the log, and left out of the baseline and calibration.
DATA
Your sessions, baseline and settings live on this device. Export to keep a portable copy. Import to restore on another device — or after a wipe. The app asks your browser to keep this storage durable, so it isn't cleared when space runs low; nothing is sent anywhere. The app makes no request to any other site: its typefaces are served from v3ktor.app itself.
JSON = full backup (every field of every session, baseline, settings). CSV = sessions only, the main columns: enough for Excel / R / pandas, not a full backup. COPY / PASTE = the CSV, through the clipboard, for when downloads are awkward. To move everything to another device, use JSON. Encryption passphrase is optional for all — leave blank for plaintext.
FAQ
YOUR 4 MOST STABLE SIGNALS
BASELINE
CLEAR HISTORY
RESTORE CLEARED LOG
NEGATIVE
POSITIVE
3
// TEST PAUSED
10
SESSION COMPLETE
YOUR RESULTS
All data stays on your device. Nothing is transmitted anywhere.