Computer-assisted analysis of spontaneous speech: quantification of basic parameters in aphasic and unimpaired language

Abstract

Although generally accepted as an important part of aphasia assessment, detailed analysis of spontaneous speech is rarely carried out in clinical practice mostly due to time limitations. The Aachener Sprachanalyse (ASPA; Aachen Speech Analysis) is a computer-assisted method for the quantitative analysis of German spontaneous speech that allows for a detailed assessment by means of linguistic basic parameters in an acceptable amount of time. It has previously been proven sensitive for monitoring changes over time. In this study, we present data of 52 aphasic participants whose spontaneous speech was analyzed retrospectively before and after an intensive therapy program. The measured changes are evaluated with reference to normative data of 60 non-brain-damaged speakers. Results confirm good sensitivity to document changes over time. Clinical relevance of changes is assessed with reference to critical score ranges derived from the normative data. Findings provide further evidence of the clinical applicability and usefulness of ASPA.

Keywords:

• aphasia
• linguistic analysis
• normative data
• clinical application
• German

View correction statement:

Erratum

Acknowledgments

This research was supported by the START program of the Faculty of Medicine, RWTH Aachen University, Aachen. Some of the results were presented as work in progress at the “7.Jahrestagung der Gesellschaft für Aphasieforschung und –behandlung” (poster presentation and corresponding short printed report, see Hussmann et al., 2008). We thank Roelien Bastiaanse and one anonymous reviewer for helpful comments and suggestions for improvement.

Declaration of Interest: The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.

Notes

1 We regard the terms spontaneous language and spontaneous speech as equivalent in the context of this study; we use the latter term following the definition of Prins and Bastiaanse (2004).

2 The type-token-ratio is known to be a parameter sensitive to different length of transcripts (Wright, Silverman, & Newhoff, 2003); but as it is also a well-established, easily administered parameter indicating lexical richness, we decided to include it tentatively in this study being aware of the required caution when interpreting the results.

3 One patient had suffered from a cerebro-vascular accident in the right hemisphere. As he was ambidextrous and presented with typical aphasic symptoms, he was included in this study.

4 Score 0 in more than one of these rating scales would characterize most severe language impairment with few or even no analyzable utterances so that the speech sample would not be long enough to be analyzed at all. A score of 4 more than once would indicate only mild to very mild impairment. Whether the ASPA parameters are sensitive enough to measure changes in such mild impairments will be the subject of further studies.

5 Transcription of 30 CLUs led to 183 words (open and closed class) on average which we considered as sufficient (acceptable length in manageable amount of time).

6 As the data were analyzed retrospectively and no detailed information about the participants' educational level was present, their respective classification to the high- or low-level group was concluded from the participants' last occupation before onset of illness.

7 Critical differences were calculated according to the formula 1.645 × SD × √(2 × (1 − Cronbach's alpha)). For details of the procedure, see Grande et al. (2008).

8 Type I error level of 10% as commonly used in individual diagnostic decisions.

9 SINGLIMS.EXE is a statistical program for comparison of an individual's score on a single test with the score of a normative or control sample; download: http://www.abdn.ac.uk/~psy086/dept/psychom.htm#conflims.