Speech-Language Sample Analysis and Informal Assessment

Speech-language sample analysis and informal assessment is the descriptive evaluation of a person's communication as it occurs in conversation, narration, or play, rather than under the fixed conditions of a standardised test. A representative sample of speech and language is recorded, transcribed, and analysed for measures such as utterance length, grammatical accuracy, vocabulary diversity, intelligibility, and narrative organisation, giving a picture of functional language use.

Definition

Language sample analysis is the elicitation, transcription, and systematic analysis of a representative sample of a person's spontaneous speech and language to describe communication abilities using measures such as mean length of utterance, grammatical and lexical indices, and narrative structure.

Scope

This topic covers the rationale for sampling natural communication, the procedures of eliciting, transcribing, segmenting, and coding a sample, the common quantitative and descriptive measures derived from it, and the strengths and limits of informal assessment relative to standardised testing. It is presented as a reference account of a method, not as a procedure for evaluating an individual.

Core questions

• How is a sample made representative of a person's everyday communication?
• Which measures derived from a sample most reliably reflect language ability, and how is reliability affected by sample length?
• What can descriptive sampling reveal that a standardised test cannot, and vice versa?
• How are transcription and coding decisions kept consistent enough to support comparison over time?

Key concepts

• Representative sampling of natural communication
• Elicitation contexts (conversation, narrative, play, expository)
• Transcription and utterance segmentation
• Mean length of utterance (MLU)
• Lexical diversity and grammatical measures
• Narrative macrostructure and microstructure
• Intelligibility analysis
• Reference databases for descriptive comparison

Mechanisms

A clinician elicits connected speech in one or more contexts, records it, and transcribes it into segmented utterances. From the transcript, quantitative indices (for example, mean length of utterance, measures of lexical diversity, and counts of grammatical structures) and descriptive analyses (such as narrative macrostructure) are computed and may be compared against reference databases. The validity of these measures depends on the sample being representative and of sufficient length; shorter transcripts yield less stable estimates of some measures, so length is itself a methodological variable (Heilmann, Nockerts, & Miller, 2010). Narrative scoring schemes extend sampling beyond sentence-level metrics to discourse organisation (Heilmann, Miller, Nockerts, & Dunaway, 2010).

Clinical relevance

Language sampling provides functional, context-rich information that complements standardised scores and can be especially informative when norm-referenced tests are unavailable or ill-suited to a person's background. This entry describes how such samples are gathered and analysed and what they can and cannot show; it is a reference orientation and not a protocol for assessing an individual.

Evidence & guidelines

Studies of language sample measures show that their stability depends on transcript length and on consistent transcription and coding, and that reference databases allow descriptive comparison without the constraints of formal norms (Heilmann, Nockerts, & Miller, 2010; Heilmann, Miller, Nockerts, & Dunaway, 2010). Sampling is often recommended as a complement to standardised testing, particularly given documented psychometric limitations of some formal tests (McCauley & Swisher, 1984).

History

Descriptive analysis of children's spontaneous language has a long tradition in developmental study, and clinical language sample analysis was systematised in the later twentieth century through structured transcription and coding conventions and, later, computer-assisted transcript analysis and reference databases. Research has since refined which measures are reliable and how sampling conditions, including transcript length, affect them (Heilmann, Nockerts, & Miller, 2010).

Debates

How long must a sample be to yield stable measures?

Some indices derived from language samples stabilise only with sufficiently long transcripts, so very short samples can misrepresent ability; the trade-off between feasibility and measurement stability is an active methodological consideration.

Key figures

• Jon Miller
• John Heilmann
• Ann Nockerts
• Robin Chapman

Seminal works

• heilmann-2010-length
• heilmann-2010-databases

Frequently asked questions

How does language sampling differ from a standardised test?

A standardised test compares fixed responses with norms under controlled conditions, whereas a language sample describes how a person actually communicates in natural contexts, capturing functional use that structured tests may miss.

Why are transcript length and consistent coding emphasised?

Because some sample-based measures only become stable with adequate length and because transcription and coding choices affect the numbers, controlling these factors is necessary for the measures to be reliable and comparable over time.