editor - VIBE score

B - Beat Score

What is it? Why did we make it? How to get the most out of it.

Sentence Burstiness

What it is

Measures the prose “music”: healthy variation in sentence length and syllable mix so it never feels metronomic.

What it looks for

• A natural mix of short/medium/long sentences
• Diversity in syllables per sentence (not all choppy, not all dense)

Why it matters

Predictability slows readers; controlled variation sustains attention and improves scan comfort (ties to expectancy/surprisal in reading).

Research anchors

Levy (2008): expectation-based processing cost—more predictable sequences are read differently; we adapt the idea to sentence-level cadence.

Scoring gist:

How to improve (tactics)

Merge two medium sentences; make one short punch, one longer unpack.

Break “and/which/that” chains into two sentences.

Vary openings (not every sentence starting with “You/We/It”).

Pacing Entropy

What it is

Evaluates how predictable your tempo is. We want interest, not chaos or clockwork.

What it looks for

An “interesting” rhythm—neither repetitive nor wildly erratic.

Why it matters

Predictability vs. surprise affects processing; mid-range unpredictability keeps readers engaged.

Research anchors

• Expectation/surprisal frameworks connect predictability to reading effort; we proxy at sentence-length level.

Scoring gist:

• Bucket lengths (S/M/L/XL), model transitions, compute entropy.
• Mid-range entropy = best; extreme order or chaos both score lower.

How to improve

• If you see a run of medium sentences, insert a short one.
• Avoid mechanical short/long/short/long patterns.
• Add a transitional clause to smooth a jump from very short to very long.

Lexical Rhythm

What it is

The “breathing” layer: micro-pauses (punctuation) + macro-breath (paragraph lengths).

What it looks for

• Pause density around an optimal band (~11–15 per 100 words, tunable)
• Paragraph lengths clustering near a comfortable range (≈35–75 words)

Why it matters

Scannability improves with well-formed paragraphs; over/under-punctuation strains reading.

Research anchors

• NN/g classic work on scannability and web reading behavior.
  

Scoring gist:

• 80% weight: Gaussian around optimal pause density
• 20% weight: “Breathability” bell on paragraph length distribution

How to improve

• Add one comma or dash in long, clause-heavy sentences.
• Split walls (>140 words) into two paragraphs.
• Merge long runs of one-liners into a coherent paragraph.

Flow Consistency

What it is

Keeps cognitive load steady—no whiplash from very simple to very dense.

What it looks for

Similar readability across adjacent chunks; if complexity rises, it ramps.

Why it matters

Consistent difficulty supports comprehension; readability metrics are a practical proxy.

Research anchors

• Flesch/F-K: word/sentence factors and perceived difficulty. Limitations noted, but still useful as a signal.

Scoring gist:

• Slice into 100-word chunks; compute readability per chunk.
• Penalize high variance; lower variance = higher score.

How to improve

If a section must be technical, add a plain-language setup sentence.

Replace bursty jargon clusters with one term + quick definition.

Keep sentence length roughly steady within a dense section.

Skimmability

What it is

How fast a scanner can grasp the outline and locate what they need.

What it looks for

• Structure slice (e.g., 40% of Skim cap): H2–H4 distribution, lists, images/figures, tables, pull-quotes, and constrained inline emphasis.
• Breathability slice (e.g., 30%): paragraph air (same breathability01).
• Diminishing returns so you can’t “spam” headings or images.

Why it matters

Users scan pages. Headings, lists, and well-spaced blocks dramatically boost usability and recall; relevant visuals help.

Research anchors

• NN/g: concise + scannable layouts raise usability (47–124%).
• Houts et al. (2006); Kools et al. (2006): pictures paired with text increase attention, recall, and task performance.

Scoring gist:

• Doc-level “sweet-spot”: ideal spacing targets (e.g., ~1 H2 per 175 words; 1 image per 800–1200 words), scored with a bell around the ideal.
• Window anomalies: 300/60 windows get a light positive/negative bump if a local block is much better/worse than the doc baseline.
• Add breathability per window. EWMA-smooth to a final score

How to improve

• Aim for 3–7 H2s; add H3s only when a section truly splits; keep H4s rare.

• Use lists for 3–7 related points; convert list-dumps back to prose.

• Include one useful figure/table per long section (caption tables).

• Keep paragraph length in the 35–75-word band; avoid 140+-word walls.

• Use bold sparingly for key terms; avoid paragraphs that are mostly bold/italic.