BRAIN AND LANGUAGE 2, 281-303 (1975)
Spontaneous Speech of Aphasic Patients: A Psycholinguistic Analysis 1 gRIN WAGENAAR
Aphasia Foundation of the Netherlands AND C A T H E R I N E S N O W AND R O N A L D PRINS
Institute for General Linguistics, University of Amsterdam An attempt was made to develop an objective and complete scoring procedure for describing the spontaneous speech of aphasia patients. A factor analysis performed on 30 items produced six significant factors: Fluency, Telegraphic Speech, Grammatical Errors, Articulation, Verbal Paraphasia, and Empty Speech. Of these, Fluency was by far the most important. A classification of patients into fluent and nonfluent types could be made on the basis of two variables, Speech Tempo and Mean Length of Utterance. Suggestions are made concerning the items which should be included in a shortened scoring procedure.
INTRODUCTION
Analysis of spontaneous speech has received considerable attention in the recent literature concerning diagnosis and classification of aphasic patients (Benson, 1967; Goodglass & Kaplan, 1972; Goodglass, Quadfasel & Timberlake, 1964; Howes, 1964; Howes & Geschwind, 1964; Kerschensteiner, Poeck & Brunner, 1972; Wepman & Jones, 1966), undoubtedly because disruption of spontaneous speech is normally the most distressing symptom of aphasia, and often the first observed. Spontaneous speech production is also important theoretically because it provides the most subtle and complete reflection of language abilities, at least if it is analyzed in a sufficiently detailed and linguistically interpretable manner. This research was supported by the Praeventiefonds in The Hague, The Netherlands. The authors express their gratitude to Helene Bakker, Marjo van Beek, Frieda Blokker, Marjan HOhle, Willie Malkus, Herbert Minter, Herman Olde, Leo Spaans, Eefje Visch, and Diek Wenkebach for their work in developing and using the scoring procedure, to Leo van Herpt and Jan Blom of the Institute for Phonetic Sciences, University of Amsterdam for their help with the computer analysis, to Dr. H. R. van Dongen for his careful reading of an earlier draft of this manuscript, and to Dr. B. G. Deelman, Drs. H. Richter, and especially Professor Dr. B. Th. Tervoort for their continued help and support in the planning of the research. Authors' mailing address is: Institute for General Linguistics, University of Amsterdam, Spui 21, Amsterdam, The Netherlands. 281 Copyright© 1975by AcademicPress, Inc. All rightsof reproductionin any formreserved.
282
WAGENAAR, SNOW AND PRINS
Spontaneous speech analysis has increased in importance as doubts have arisen concerning the basis for the classical diagnostic distinction of aphasics into 'motor' types, whose speech production is disturbed but whose language comprehension is relatively intact, and 'sensory' types, who show the complementary pattern of more or less intact production with disturbed comprehension (Kleist, 1934; Lichtheim, 1885; Wernicke, 1874). Although this classical distinction continues to be widely used (often with new terminology, e.g., Brain, 1965; Jakobson, 1970; Luria, 1966), clinical and experimental reports suggest that speech comprehension is disturbed in all aphasics, independent of the degree and character of the expressive disability (De Renzi & Vignolo, 1962; Goodglass, 1968; Poeck, Kerschensteiner & Hartje, 1972; Shewan & Canter, 1971). These findings cast doubt, not only on the traditional basis for classification, but also on the efficacy of traditional languagecomprehension tests in the diagnostic process. Analysis of spontaneous speech production becomes therefore that much more important for the evaluation of aphasia patients. One group of researchers from the so-called Boston school who have made spontaneous speech analysis a central point in their evaluation of aphasics have arrived at a translation of the classical motor-sensory distinction into a new distinction based primarily on aspects of speech production (Goodglass et al., 1964; Howes, 1967; Benson, 1967; Goodglass & Kaplan, 1972). In their description the classical motor patient is referred to as "nonfluent" because his speech production is difficult, slow, typically characterized by poor melody, many pronunciation mistakes, short utterances, simple grammatical structure, and fewer function words and grammatical markers than in normal speech. The classical sensory patient is called "fluent" because he speaks with normal, or even excessive, speed, normal melody and articulation, often in long and grammatically complex sentences. His disability often reveals itself in the presence of literal and verbal paraphasias (sound and word substitutions) and in that the speech seems to lack content; nouns and verbs are relatively infrequent, many words are either paraphasic mistakes or neologisms, and the utterances as a whole convey little or no information. Various subtypes within both the fluent and the nonfluent populations can be identified (Goodglass & Kaplan, 1972) while a certain number of patients are unclassifiable because their symptomatology shows a mixture of the typically fuent and typically nonfluent characteristics. A usable means of analyzing spontaneous production is crucial to the postulation and defense of the fuent-nonfluent distinction, and to the diagnosis of patients by those who operate with the distinction. However, the analyses of spontaneous speech production which have been
S P O N T A N E O U S SPEECH ANALYSIS
283
published are deficient in several ways: (1) The items are scored in a subjective and nonquantified manner, which makes replication or application of the same scales to other samples impossible. For example, Benson (1967) has analyzed spontaneous speech with ten language measures, each of them a judgement given in terms of a three-point scale. The ten measures used were: (1) rate of speech, (2) prosody, (3) articulation, (4) utterance length, (5) effort, (6) pauses, (7) press of speech, (8) perseverations, (9) word choice, (10) paraphasia. The scales were so defined that the typical nonfluent patient would receive a rating of 1 on each item, or 10 total, whereas the typical fluent patient would receive a total of 30. This analysis is of course highly subjective, especially since rating the 10 items truly independently of each other is hardly possible except with an elaborate double blind procedure. The same criticism applies to Kerschensteiner et al. (1972), who replicated Benson's study and confirmed his results by way of a mathematically more refined analysis. The subjectivity of the ratings is not intrinsic; all of the items except possibly prosody, articulation, and effort are easily quantifiable. Goodglass & Kaplan's (1972) six language measures (melody, utterance length, articulation, grammatical variety, paraphasia, and word finding) are just as subjective as those used by Benson and Kerschensteiner et al., though a more extended, 7-point rating system was used with more precise definitions for each point. (2) The analyses do not attempt a characterization of the grammatical structure of the speech produced. Goodglass & Kaplan (1972) used one item having to do with grammatical variety; beyond that, no one has incorporated the clinical impression that fluent aphasics' speech is grammatically complex into his spontaneous speech analysis. Perhaps the most striking example of disinterest in the syntactical aspects of aphasics' speech is Howes' (1964, 1967) analysis, based primarily on the distribution of word frequencies, rate of speaking and results of word association tests. (3) The situation in which the spontaneous speech is elicited is too variant from patient to patient to allow strict comparison. In general, no information is even offered concerning the speech samples analyzed, where and how they were collected, what the topic was, etc. Yet it is known that variables such as the speech-situation (informal vs formal conversation, picture description vs free narrative), the emotional state of the patient or the topic discussed can greatly influence the amount and character of the speech output of aphasic patients (Luria, 1966; Goodglass & Kaplan, 1972). (4) Information concerning the distribution of the patient population into the subgroups identified is absent or incomplete, as is information
284
WAGENAAR, SNOW AND PRINS
concerning the range of scores on a given item to be found within any subgroup. For example, Wepman & Jones have used the highly quantified method of counting the frequencies with which different word classes occur in aphasics' speech. The relative frequencies of the various classes provide the basis for classification, as well as a measure of the seriousness of the disability (Wepman & Jones, 1966b; Jones & Wepman, 1967). Unfortunately, no information is offered concerning the percentage of patients in each subgroup, nor concerning the distribution of scores on a given item for each subgroup. Similarly, Goodglass and Kaplan (1972), whose method of speech analysis is probably the most extensive, well-defined, and usable to be found, give only profiles of typical patients in a given group, but no information about the frequency of occurrence of the subtypes. An analysis of the spontaneous speech of 107 Dutch-speaking aphasics was undertaken with the intention of satisfying at least some of the criticisms directed at previous studies. We have attempted to provide a linguistically-sophisticated analysis which is as complete and as quantified as possible and which reflects type as well as degree of aphasia. A factor analysis was applied to the patients' scores so as to provide information concerning (a) the factors necessary for a complete description of the spontaneous speech of aphasics, and (b) the minimal number of items which can be used to reflect those factors adequately. Furthermore, we attempted to use the scoring procedure so developed as a basis for differentiating fluent from nonfluent aphasics, and for identifying various of the subtypes of aphasics that have been discussed in the literature. METHOD
Subjects Out of the complete population of patients being treated for aphasia in seven treatment centers, consisting of 156 patients, all those were selected who satisfied the following criteria: (1) aphasia developed as a result of a cerebrovascular accident, (2) aphasic symptoms had been present for at least three months at time of first testing, and (3) the patients were willing and able to participate in a test program lasting one year. The spontaneous speech of only 74 of the 107 patients meeting these criteria could be fully analyzed. Of the 107 patients, five had no spontaneous speech at all; five produced only incomprehensible (and therefore unscorable) muttering; 17 produced fewer then 10 words/min, mainly "yes," "no," a few single words or word attempts, and/or some neologisms. Five patients produced only stereotypics like "wow-dow," "tata," etc. One patient could not be analyzed because he produced only incomprehensible phonological jargon. Table 1 presents information concerning the sex, age, history, and therapy of all the patients included in the study.
285
SPONTANEOUS SPEECH ANALYSIS
TABLE 1 AGE, SEX, HISTORY, THERAPY AND SOCIOECONOMIC CLASS OF PATIENT SAMPLE
Age Sex Onset of" aphasia Receiving therapy Occupati°nb Sch°°lingb
mean range male female mean range yes no mean range mean range
Analyzed group n = 74
Unanalyzed group n = 33
59.2 18-89 36 38 20.5 3-100 61 13 4.06 1-7 2.87 1-7
62.3 34-85 15 17 21.5 4-99 31 2 3.86 1-7 3.09 1-7
a Expressed in number of months prior to testing of spontaneous speech. b Expressed in terms of a 7 point-scale developed by the Dutch Federal Employment Bureau. 1 represents the lowest and 7 the highest possible score.
Testing Procedure The spontaneous speech samples were collected by the same test assistant who performed the other language tests with the patient. Occasionally in the course of testing the assistant would conversationally ask one of three standard questions. The assistant was instructed to say as little as possible during the patients' responses. The replies were tape recorded. The topic proposed was pursued for a minimum of 2 rain. The questions asked were: "What do you usually spend the day doing?," "How did your speech problems start?" and "Tell me something about the place you live." The two rain of conversation following each question were transcribed using standard spelling the same day by the assistant, who added explanatory notes where necessary.
Scoring Procedure The texts and accompanying tapes were analyzed by one of six two-person teams of judges. The 12 judges had themselves helped to develop the scoring procedures, and had spent four months scoring practice material in order to achieve high interjudge reliability. The judges worked with the help of an extensive instruction manual which had been written on the basis of difficult cases in the practice scoring sessions. Each 2-min speech sample was scored separately; the patient's score was a mean or a total of the three separate scores, depending on the item. The scoring form consisted of 40 items, organized in five categories. The scored items which were subjected to statistical analysis are listed in Table 2. (1) Evaluation. In order to reflect as many aspects of aphasic speech as possible, it was necessary to include a number of items which were scored subjectively; this subjectivity was minimized by providing extensive and detailed scoring instructions for these items, and by checking interjudge reliability. Each sample was rated on a 7-point scale for (1) communicative capacity, (2) melody, and (3) articulation. In addition, each utterance in the
TABLE 2 THIRTY VARIABLES SELECTED FOR THE FACTOR ANALYSIS
Variable Speech tempo Communicative capacity Syntactic complexity Melody Articulation Utterance production Utterances shorther than Five-words Mean length of utterance Mean length of three longest utterances Complex utterances Seconds incomprehensible Self-corrections Automatisms Imitations Literal paraphasias Verbal paraphasias Neologisms Literal perservations Verbal perseverations Function-word substitutions Function-word deletions Content-word deletions Syntactic mixtures Content-word/Functionword Ratio Nouns Personal pronouns Pronouns Word-order mistakes Tense mistakes Unclassified mistakes
Method of calculation Number of words produced in 6 rain Average of the evaluations of each 2-min sample Average of the evaluations of each utterance Average of the evaluations of each 2-min sample Average of the evaluations of each 2-min sample Number of utterances produced in 6 min Number of utterances shorter than five words expressed as percentage of total number of utterances Number of words divided by number of utterances Number of words in three longest utterances divided by 3. Number of complex utterances expressed as percentage of total number of utterances Number of seconds of speech which were incomprehensible in 6 min Number of self-corrections expressed as percentage of total number of utterances Number of automatisms expressed as percentage of total number of utterances Number of imitations of the test-assistant expressed as percentage of total number of utterances Number of literal paraphasias expressed as percentage of number of content words Number of verbal paraphasias expressed as percentage of number of content words Number of neologisms expresses as percentage of number of content words Number of literal perseverations expressed as percentage of number of utterances Number of verbal perseverations expressed as percentage of number of utterances Number of substitutions of function words expressed as percentage of number of function words Number of deletions of function words expressed as percentage of number of utterances Number of deletions of content words expressed as percentage of number of utterances Number of syntactically confused structures expressed as percentage of number of utterances Ratio of number of content words to number of function words Number of nouns expressed as percentage of total number of words Ratio of personal pronouns to number of nouns Ratio of all pronouns to number of content words Number of word-order mistakes expressed as percentage of number of utterances Number of tense mistakes expressed as percentage of number of utterances Number of all other kinds of grammatical mistakes expressed as percentage of number of utterances
286
287
SPONTANEOUS SPEECH ANALYSIS TABLE 3 INTERJUDGE RELIABILITIES FOR RATING SCALES Scale
Reliability~
Communicative capacity Syntactic complexity Melody Articulation
.914 .941 .864 .883
Significance p p p p
< < <
Spontaneous Speech of Aphasic Patients: A Psycholinguistic Analysis 1 gRIN WAGENAAR
Aphasia Foundation of the Netherlands AND C A T H E R I N E S N O W AND R O N A L D PRINS
Institute for General Linguistics, University of Amsterdam An attempt was made to develop an objective and complete scoring procedure for describing the spontaneous speech of aphasia patients. A factor analysis performed on 30 items produced six significant factors: Fluency, Telegraphic Speech, Grammatical Errors, Articulation, Verbal Paraphasia, and Empty Speech. Of these, Fluency was by far the most important. A classification of patients into fluent and nonfluent types could be made on the basis of two variables, Speech Tempo and Mean Length of Utterance. Suggestions are made concerning the items which should be included in a shortened scoring procedure.
INTRODUCTION
Analysis of spontaneous speech has received considerable attention in the recent literature concerning diagnosis and classification of aphasic patients (Benson, 1967; Goodglass & Kaplan, 1972; Goodglass, Quadfasel & Timberlake, 1964; Howes, 1964; Howes & Geschwind, 1964; Kerschensteiner, Poeck & Brunner, 1972; Wepman & Jones, 1966), undoubtedly because disruption of spontaneous speech is normally the most distressing symptom of aphasia, and often the first observed. Spontaneous speech production is also important theoretically because it provides the most subtle and complete reflection of language abilities, at least if it is analyzed in a sufficiently detailed and linguistically interpretable manner. This research was supported by the Praeventiefonds in The Hague, The Netherlands. The authors express their gratitude to Helene Bakker, Marjo van Beek, Frieda Blokker, Marjan HOhle, Willie Malkus, Herbert Minter, Herman Olde, Leo Spaans, Eefje Visch, and Diek Wenkebach for their work in developing and using the scoring procedure, to Leo van Herpt and Jan Blom of the Institute for Phonetic Sciences, University of Amsterdam for their help with the computer analysis, to Dr. H. R. van Dongen for his careful reading of an earlier draft of this manuscript, and to Dr. B. G. Deelman, Drs. H. Richter, and especially Professor Dr. B. Th. Tervoort for their continued help and support in the planning of the research. Authors' mailing address is: Institute for General Linguistics, University of Amsterdam, Spui 21, Amsterdam, The Netherlands. 281 Copyright© 1975by AcademicPress, Inc. All rightsof reproductionin any formreserved.
282
WAGENAAR, SNOW AND PRINS
Spontaneous speech analysis has increased in importance as doubts have arisen concerning the basis for the classical diagnostic distinction of aphasics into 'motor' types, whose speech production is disturbed but whose language comprehension is relatively intact, and 'sensory' types, who show the complementary pattern of more or less intact production with disturbed comprehension (Kleist, 1934; Lichtheim, 1885; Wernicke, 1874). Although this classical distinction continues to be widely used (often with new terminology, e.g., Brain, 1965; Jakobson, 1970; Luria, 1966), clinical and experimental reports suggest that speech comprehension is disturbed in all aphasics, independent of the degree and character of the expressive disability (De Renzi & Vignolo, 1962; Goodglass, 1968; Poeck, Kerschensteiner & Hartje, 1972; Shewan & Canter, 1971). These findings cast doubt, not only on the traditional basis for classification, but also on the efficacy of traditional languagecomprehension tests in the diagnostic process. Analysis of spontaneous speech production becomes therefore that much more important for the evaluation of aphasia patients. One group of researchers from the so-called Boston school who have made spontaneous speech analysis a central point in their evaluation of aphasics have arrived at a translation of the classical motor-sensory distinction into a new distinction based primarily on aspects of speech production (Goodglass et al., 1964; Howes, 1967; Benson, 1967; Goodglass & Kaplan, 1972). In their description the classical motor patient is referred to as "nonfluent" because his speech production is difficult, slow, typically characterized by poor melody, many pronunciation mistakes, short utterances, simple grammatical structure, and fewer function words and grammatical markers than in normal speech. The classical sensory patient is called "fluent" because he speaks with normal, or even excessive, speed, normal melody and articulation, often in long and grammatically complex sentences. His disability often reveals itself in the presence of literal and verbal paraphasias (sound and word substitutions) and in that the speech seems to lack content; nouns and verbs are relatively infrequent, many words are either paraphasic mistakes or neologisms, and the utterances as a whole convey little or no information. Various subtypes within both the fluent and the nonfluent populations can be identified (Goodglass & Kaplan, 1972) while a certain number of patients are unclassifiable because their symptomatology shows a mixture of the typically fuent and typically nonfluent characteristics. A usable means of analyzing spontaneous production is crucial to the postulation and defense of the fuent-nonfluent distinction, and to the diagnosis of patients by those who operate with the distinction. However, the analyses of spontaneous speech production which have been
S P O N T A N E O U S SPEECH ANALYSIS
283
published are deficient in several ways: (1) The items are scored in a subjective and nonquantified manner, which makes replication or application of the same scales to other samples impossible. For example, Benson (1967) has analyzed spontaneous speech with ten language measures, each of them a judgement given in terms of a three-point scale. The ten measures used were: (1) rate of speech, (2) prosody, (3) articulation, (4) utterance length, (5) effort, (6) pauses, (7) press of speech, (8) perseverations, (9) word choice, (10) paraphasia. The scales were so defined that the typical nonfluent patient would receive a rating of 1 on each item, or 10 total, whereas the typical fluent patient would receive a total of 30. This analysis is of course highly subjective, especially since rating the 10 items truly independently of each other is hardly possible except with an elaborate double blind procedure. The same criticism applies to Kerschensteiner et al. (1972), who replicated Benson's study and confirmed his results by way of a mathematically more refined analysis. The subjectivity of the ratings is not intrinsic; all of the items except possibly prosody, articulation, and effort are easily quantifiable. Goodglass & Kaplan's (1972) six language measures (melody, utterance length, articulation, grammatical variety, paraphasia, and word finding) are just as subjective as those used by Benson and Kerschensteiner et al., though a more extended, 7-point rating system was used with more precise definitions for each point. (2) The analyses do not attempt a characterization of the grammatical structure of the speech produced. Goodglass & Kaplan (1972) used one item having to do with grammatical variety; beyond that, no one has incorporated the clinical impression that fluent aphasics' speech is grammatically complex into his spontaneous speech analysis. Perhaps the most striking example of disinterest in the syntactical aspects of aphasics' speech is Howes' (1964, 1967) analysis, based primarily on the distribution of word frequencies, rate of speaking and results of word association tests. (3) The situation in which the spontaneous speech is elicited is too variant from patient to patient to allow strict comparison. In general, no information is even offered concerning the speech samples analyzed, where and how they were collected, what the topic was, etc. Yet it is known that variables such as the speech-situation (informal vs formal conversation, picture description vs free narrative), the emotional state of the patient or the topic discussed can greatly influence the amount and character of the speech output of aphasic patients (Luria, 1966; Goodglass & Kaplan, 1972). (4) Information concerning the distribution of the patient population into the subgroups identified is absent or incomplete, as is information
284
WAGENAAR, SNOW AND PRINS
concerning the range of scores on a given item to be found within any subgroup. For example, Wepman & Jones have used the highly quantified method of counting the frequencies with which different word classes occur in aphasics' speech. The relative frequencies of the various classes provide the basis for classification, as well as a measure of the seriousness of the disability (Wepman & Jones, 1966b; Jones & Wepman, 1967). Unfortunately, no information is offered concerning the percentage of patients in each subgroup, nor concerning the distribution of scores on a given item for each subgroup. Similarly, Goodglass and Kaplan (1972), whose method of speech analysis is probably the most extensive, well-defined, and usable to be found, give only profiles of typical patients in a given group, but no information about the frequency of occurrence of the subtypes. An analysis of the spontaneous speech of 107 Dutch-speaking aphasics was undertaken with the intention of satisfying at least some of the criticisms directed at previous studies. We have attempted to provide a linguistically-sophisticated analysis which is as complete and as quantified as possible and which reflects type as well as degree of aphasia. A factor analysis was applied to the patients' scores so as to provide information concerning (a) the factors necessary for a complete description of the spontaneous speech of aphasics, and (b) the minimal number of items which can be used to reflect those factors adequately. Furthermore, we attempted to use the scoring procedure so developed as a basis for differentiating fluent from nonfluent aphasics, and for identifying various of the subtypes of aphasics that have been discussed in the literature. METHOD
Subjects Out of the complete population of patients being treated for aphasia in seven treatment centers, consisting of 156 patients, all those were selected who satisfied the following criteria: (1) aphasia developed as a result of a cerebrovascular accident, (2) aphasic symptoms had been present for at least three months at time of first testing, and (3) the patients were willing and able to participate in a test program lasting one year. The spontaneous speech of only 74 of the 107 patients meeting these criteria could be fully analyzed. Of the 107 patients, five had no spontaneous speech at all; five produced only incomprehensible (and therefore unscorable) muttering; 17 produced fewer then 10 words/min, mainly "yes," "no," a few single words or word attempts, and/or some neologisms. Five patients produced only stereotypics like "wow-dow," "tata," etc. One patient could not be analyzed because he produced only incomprehensible phonological jargon. Table 1 presents information concerning the sex, age, history, and therapy of all the patients included in the study.
285
SPONTANEOUS SPEECH ANALYSIS
TABLE 1 AGE, SEX, HISTORY, THERAPY AND SOCIOECONOMIC CLASS OF PATIENT SAMPLE
Age Sex Onset of" aphasia Receiving therapy Occupati°nb Sch°°lingb
mean range male female mean range yes no mean range mean range
Analyzed group n = 74
Unanalyzed group n = 33
59.2 18-89 36 38 20.5 3-100 61 13 4.06 1-7 2.87 1-7
62.3 34-85 15 17 21.5 4-99 31 2 3.86 1-7 3.09 1-7
a Expressed in number of months prior to testing of spontaneous speech. b Expressed in terms of a 7 point-scale developed by the Dutch Federal Employment Bureau. 1 represents the lowest and 7 the highest possible score.
Testing Procedure The spontaneous speech samples were collected by the same test assistant who performed the other language tests with the patient. Occasionally in the course of testing the assistant would conversationally ask one of three standard questions. The assistant was instructed to say as little as possible during the patients' responses. The replies were tape recorded. The topic proposed was pursued for a minimum of 2 rain. The questions asked were: "What do you usually spend the day doing?," "How did your speech problems start?" and "Tell me something about the place you live." The two rain of conversation following each question were transcribed using standard spelling the same day by the assistant, who added explanatory notes where necessary.
Scoring Procedure The texts and accompanying tapes were analyzed by one of six two-person teams of judges. The 12 judges had themselves helped to develop the scoring procedures, and had spent four months scoring practice material in order to achieve high interjudge reliability. The judges worked with the help of an extensive instruction manual which had been written on the basis of difficult cases in the practice scoring sessions. Each 2-min speech sample was scored separately; the patient's score was a mean or a total of the three separate scores, depending on the item. The scoring form consisted of 40 items, organized in five categories. The scored items which were subjected to statistical analysis are listed in Table 2. (1) Evaluation. In order to reflect as many aspects of aphasic speech as possible, it was necessary to include a number of items which were scored subjectively; this subjectivity was minimized by providing extensive and detailed scoring instructions for these items, and by checking interjudge reliability. Each sample was rated on a 7-point scale for (1) communicative capacity, (2) melody, and (3) articulation. In addition, each utterance in the
TABLE 2 THIRTY VARIABLES SELECTED FOR THE FACTOR ANALYSIS
Variable Speech tempo Communicative capacity Syntactic complexity Melody Articulation Utterance production Utterances shorther than Five-words Mean length of utterance Mean length of three longest utterances Complex utterances Seconds incomprehensible Self-corrections Automatisms Imitations Literal paraphasias Verbal paraphasias Neologisms Literal perservations Verbal perseverations Function-word substitutions Function-word deletions Content-word deletions Syntactic mixtures Content-word/Functionword Ratio Nouns Personal pronouns Pronouns Word-order mistakes Tense mistakes Unclassified mistakes
Method of calculation Number of words produced in 6 rain Average of the evaluations of each 2-min sample Average of the evaluations of each utterance Average of the evaluations of each 2-min sample Average of the evaluations of each 2-min sample Number of utterances produced in 6 min Number of utterances shorter than five words expressed as percentage of total number of utterances Number of words divided by number of utterances Number of words in three longest utterances divided by 3. Number of complex utterances expressed as percentage of total number of utterances Number of seconds of speech which were incomprehensible in 6 min Number of self-corrections expressed as percentage of total number of utterances Number of automatisms expressed as percentage of total number of utterances Number of imitations of the test-assistant expressed as percentage of total number of utterances Number of literal paraphasias expressed as percentage of number of content words Number of verbal paraphasias expressed as percentage of number of content words Number of neologisms expresses as percentage of number of content words Number of literal perseverations expressed as percentage of number of utterances Number of verbal perseverations expressed as percentage of number of utterances Number of substitutions of function words expressed as percentage of number of function words Number of deletions of function words expressed as percentage of number of utterances Number of deletions of content words expressed as percentage of number of utterances Number of syntactically confused structures expressed as percentage of number of utterances Ratio of number of content words to number of function words Number of nouns expressed as percentage of total number of words Ratio of personal pronouns to number of nouns Ratio of all pronouns to number of content words Number of word-order mistakes expressed as percentage of number of utterances Number of tense mistakes expressed as percentage of number of utterances Number of all other kinds of grammatical mistakes expressed as percentage of number of utterances
286
287
SPONTANEOUS SPEECH ANALYSIS TABLE 3 INTERJUDGE RELIABILITIES FOR RATING SCALES Scale
Reliability~
Communicative capacity Syntactic complexity Melody Articulation
.914 .941 .864 .883
Significance p p p p
< < <
