How many words make a sample? Determining the minimum number of word tokens needed in connected speech samples for child speech assessment.
Yvonne Wren, Jill Titterington, Paul White
PMID 33021116WHAT IT FOUND
For most speech measures, a connected speech sample of 75 words is as reliable as 100.
However, samples of 50 words are too unstable. For complex measures like pMLU or error counts, stick to 100 words to ensure accuracy.
Key findings
01Speech samples of 75 word tokens showed little difference from samples containing 100 words for most single score measures.
02The stability of speech measures is lost once the sample size drops to 50 word tokens.
03pMLU, percentage substitutions, and percentage distortions showed non-ignorable trends with sample size, requiring a higher threshold of 100 word tokens.
STILL TO COME
How it was doneWhat they foundWhat it means for SLPs
Read the rest of this summary
You get three full summaries a month, free, and we do not ask for a card. Search, the TL;DRs and your library stay unlimited either way.
What it does not show
The study used data from a normative population (mostly typically developing children). Results may differ for children with severe Speech Sound Disorders, who might produce shorter samples or different error patterns. The speech samples were collected under a strict 20-minute time limit. Clinicians who allow more time for probing and conversation might obtain different sample characteristics. The analysis compared groups of children who produced different numbers of words. Children who produced fewer words may have had lower language skills, which could influence the error patterns observed (e.g., fewer substitutions in longer samples might reflect better language skills rather than just sample size).
Declared interests
The study was supported by the UK Medical Research Council, Wellcome, the University of Bristol, and the National Institute for Health Research. The authors declared no conflicts of interest.
The easy way to misread this
Do not assume that 75 words is sufficient for every speech measure. If you are calculating pMLU or analysing specific error types like substitutions and distortions, a 75-word sample will yield unstable results; you must collect 100 words for these metrics.