CLN Wiki · Memory & Cognition
Pre-Attentive Processing
= 200ms · before reading begins @ Treisman & Gelade 1980 ~ 5 sections
01// The Mechanism Neuroscience
// ##mech-def · Definition =Pre-attentive processing — visual feature extraction occurring before conscious attention is directed##pa-def .latency: 100–200ms — faster than a voluntary eye movement (150–200ms) .parallel — entire visual field processed simultaneously, not serially .automatic — cannot be suppressed; happens even when not looking for it @Treisman & Gelade (1980) — Feature Integration Theory // ##mech-popout · The pop-out effect .a single red item among 100 blue items is found in ~50ms regardless of array size .search time is flat — not proportional to n .a conjunction target (red + circle, among red squares + blue circles) scales linearly →single-feature targets pop out; conjunction targets require serial attention →CLN operators are single features — each pops out independently
02// Four Feature Types Reference
●
Color
hue · saturation · luminance
▲
Shape
orientation · curvature · closure
▮▮▮
Size
length · width · area
↕
Position
x/y · depth · column
// ##feat-cln-mapping · How CLN uses each feature type =Color — each operator has a distinct hue; ! red, = gold, → green →"find all risks" = scan for red — no reading required =Shape — operators are distinct glyphs: = vs . vs → vs ! vs ~ →shape discrimination works even at small size or low contrast =Position — operator column is fixed at column 0 (or indent level) →eyes can scan the operator column vertically without horizontal saccades =Size / weight — operator glyphs are bold; content text is normal weight →luminance contrast makes operators visible in peripheral vision
03// CLN Design Application CLN
// ##design-semantics · Semantic encoding via visual features →CLN maps semantic types onto pre-attentive features — not arbitrary decoration##sem-enc =gold = → definition — stable, grounding =red ! → risk — danger color, universal signal =green → → conclusion — forward, resolved =purple ? → unknown — cognitive dissonance color =pink ~ → tension — neither resolved nor dangerous →color choices are not random — they exploit pre-existing color-emotion associations
=
Define
.
Action
→
Result
!
Risk
~
Tension
?
Unknown
// ##design-violations · What breaks pre-attentive scanning !inconsistent operator use — = sometimes means "action" — color loses meaning !no operator column — mixed indentation destroys the vertical scan path !too many operator types per block — conjunction search replaces pop-out !colored prose without operators — decoration without semantics; noise
04// Prose vs CLN Scan Path Demo
Prose — serial scan

The system was updated last week and there are several things to consider. First, the performance has improved but there is a critical issue with the database connection that needs immediate attention.

Additionally, the team has confirmed that the new API is working correctly. However there remain some open questions about edge cases.

↓ eyes scan every word left→right to find the risk

CLN — operator column scan
.system updated last week
+performance improved
!DB connection — critical← found in 80ms
→new API confirmed working
?edge cases open
// ##scan-path · What the eye actually does .Prose — horizontal saccades, line by line; semantic scan requires reading every word .CLN — single vertical saccade down the operator column; color identifies type at each fixation →"find all risks" in 100-line prose: ~40 fixations, ~8 seconds →"find all risks" in 100-line CLN: ~5 fixations, ~1 second — scan the red column @Rayner (1998) — average fixation duration 200–250ms; saccade 20–200ms
05// Eye-Tracking Evidence Research
// ##eye-findings · What eye-tracking studies show =structured documents with visual hierarchy reduce total fixation count @Nielsen (1997) — users scan in F-pattern; headings + first words dominate fixations =color-coded categories reduce time-to-target by 50–70% vs monochrome @Ware (2004) — "Information Visualization: Perception for Design" =consistent positional anchors reduce cognitive load during re-reading @Sweller (1994) — extraneous cognitive load from inconsistent layout // ##eye-limits · Limits of the evidence ~studies on structured docs ≠ studies on CLN specifically — no direct CLN eye-tracking data yet ~benefits assume operator consistency — irregular CLN use may increase load vs prose →CLN is designed on the correct principles — empirical validation is an open research task // ##eye-open · Open questions ?does CLN operator column create a measurable vertical scan path in eye-tracking? ?minimum operator count before color-coding becomes noise rather than signal? ?does rendered CLN (HTML) outperform plain-text CLN on scan speed? @→ Spatial Memory — positional encoding and place cells @→ Block Annotations — chunking and episodic boundaries @→ CLN Specification v2.1 — full operator reference