Methodology reviewed July 9, 2026

Speaking research and methodology

This hub separates externally sourced speaking statistics from Articulated's first-party practice benchmarks and explains how each should be interpreted.

Current research

The State of Speaking Practice 2026

A first-party benchmark of anonymized practice evaluations, including pace, filler rate, recording length, and six-skill scores.

Speaking Statistics 2026

A source-led reference for public-speaking fear, social anxiety, voice and speech statistics, plus clearly labeled Articulated benchmarks.

How the first-party benchmark was produced

The current report summarizes thousands of anonymized Articulated practice evaluations. It publishes aggregate distributions, not user-level records.

  • Published values use medians, middle-50% ranges, rounded percentages, and broad bands so outliers do not dominate the result.
  • Metrics are included only where the underlying evaluation produced that measurement; filler-rate percentages therefore describe measured sessions, not every session.
  • Audio, transcripts, names, emails, account identifiers, private prompts, and session-level rows are excluded from the public report.
  • Repeated practice evaluations may come from the same person. The report describes evaluation activity and is not a count of unique people.

Metric definitions

Recording duration
Elapsed seconds in the analyzed spoken response.
Words per minute
Detected word count normalized by recording duration.
Filler rate
Detected filler events per minute where filler detection was available.
Middle 50%
The range from the 25th to the 75th percentile.
Six-skill scores
Articulated's coaching dimensions: clarity, fluency, structure, vocabulary, confidence, and engagement.

What the benchmark cannot establish

  • It is not a representative survey of a country, language, profession, age group, or all speakers.
  • People who choose a speech-practice app may differ materially from people who do not.
  • Language, prompt, device, recording conditions, and evaluation-model version can affect measurements.
  • The current public edition does not provide an exact per-metric sample size, collection-window breakdown, locale mix, or repeated-user adjustment. Cite it as a directional practice benchmark, not a population norm.
  • The findings are not medical, psychological, or speech-language pathology diagnoses.

Citing and correcting the research

Cite the individual report rather than this index, preserve its limitations, and include the report year. To flag a factual or methodological issue, contact our team with the URL and supporting evidence.

support@articulated.app · Editorial standards and corrections