The Relationship Between Verbal Ability and ...

Scand Audiol 1992; 21: 67-72

THE RELATIONSHIP BETWEEN VERBAL ABILITY A N D SENTENCE-BASED SPEECHREADING Bjorn Lyxell’,2and Jerker Ronnberg’ From the ’Deparrmen! of Psychology, University of Umed, Umed, and 2Depar!menr of Education and Psychology, Linkoping University, Linkoping, Sweden

Scandinavian Audiology Downloaded from informahealthcare.com by 90.228.209.200 on 12/05/13 For personal use only.

ABSTRACT Relarionship berween verbal ability and sentence-based speechreading. Lyxell. B. and Ronnberg, J. (Department of Psychology, University of Umeb, Umeb, and Department of Education and Psychology, Linkoping University, Linkoping, Sweden). Scand Audiol 1992; 21: 61-12. Eighteen hearing-impaired subjects participated in the present study. The purpose was to investigate one general question: The nature of the relationship between verbal ability and speechreading. Verbal ability was assessed by two types of measure: a test of vocabulary size, and four tests of lexical access speed. The results demonstrated that lexical access speed was related to speechreading performance. Vocabulary size was not found to be directly related to the speechreading criterion: rather. its influence was in an indirect fashion via its relation to lexical access speed. It was concluded that lexical access speed could be used as a diagnostic tool, such that when an individual demonstrates lexical access that is unreasonably slow, it could be taken as an indication to suggest that rehabilitation programs should emphasize alternatives to speechreading. A general implication o f the present results is that absence of relation between a predictor variable and the speechreading criterion does not necessarily imply absence of relation between the two. There is still a possibility that the predictor variable might be indirectly related to the speechreading criterion. Key words: speechreading, verbal ability, vocabulary size, lexical access speed.

INTRODUCTION Speechreading (lipreading) is a complex informationprocessing task and it is well known from both clinical observation and experimental studies that individuals vary widely in their ability to speechread. However, it is not so well known how and to what extent different subject variables contribute to this variation in performance (Gailey, 1987; Lyxell, 1989). In the present study, we examine how one such subject variable (verbal ability) is related to sentence-based speechreading performance. According to Hunt (1 978; 1985), individuals denoted ‘high verbal’ are characterized to “know lots of

words and being facile at recognizing them”. The role of word knowledge (or vocabulary size) in speechreading seems obvious: It would be impossible to speechread a word without having its referent stored in the mental dictionary (or long-term memory). However, no relationship has been reported between speechreading performance and vocabulary size when the population has been constituted by normal-hearing or hearing-impaired subjects (with post-lingually acquired impairments; Lyxell, 1989; O’Neill. 195 1 ; Simmons, 1959), whereas a significant correlation is usually found when other populations are examined (e.g., deaf or hearing-impaired children; Quigley & Frisina, 1961 ; Lowcll, 1960). The interpretation of these results is that word knowledge is not critically related to speechreading performance, given that the individual’s word knowledge is within the normal range (Lyxell, 1989; Oyer & O’Neill, 1985), and that the sentences to-be-speechread are ‘normal’ from a linguistic point of view (e.g., simple syntax and absence of low-frequency words). Nevertheless, word knowledge has proved to be a ‘good’ predictor of general verbal ability (Hunt, 1985; Matarazzo, 1972; usually correlation coefficients around 0.80 are reported between vocabulary size and overall performance in various intelligence tests). A plausible role of word knowledge (i.e., verbal ability) in speechreading could be that it serves as a ‘support variable’. That is, it is not directly related to speechreading performance; rather, its impact on speechreading might go via other important variables, such as inference-making (Lyxell, 1989; Lyxell & Ronnberg, 1989). Hunt (1978, 1980, 1985; see also Sternberg, 1985) has proposed that part of the variability in verbal ability could be explained by individual differences in the speed with which information is accessed from long-term memory (i.e., lexical access). Intuitively, this should also apply to speechreading, as there is no ScandAudiol21


68

B. Lyxell and J . Riinnherg

second chance to ‘re-read’ once an utterance has been made (Ronnberg, 1990). A more precise hypothesis about the role of lexical access in speechreading is that a relatively rapid access would free resources for other necessary processes in speechreading (e.g., inference-making; Lyxell, 1989). In the present study, a set of different lexical access tests will be employed, varying in complexity regarding what kind of information that is to be retrieved, from relatively simple matching of symbols to deciding whether words belong to different semantic categories (Ronnberg, 1990; Ronnberg et al., 1989). The lexical access tests used in the present study were chosen since they have previously demonstrated that they play critical roles in both listening and reading comprehension (Baddeley et al., 1985; Gordon-Salant, 1987). Speechreading ability was measured by a sentencebased speechreading test developed by Lyxell & Ronnberg (1 987 a, b, 1989, 1991 a, h; Ronnberg, 1990; Ronnberg et al., 1989). Specifically, this test comprises sentences of different length (Clouser, 1976; 1977; Lyxell & Ronnberg, 1987a), and different amounts of contextual information (Smith & Kitchen, 1972). MATERIAL AND METHOD Subjects Eighteen (9 males) hearing-impaired subjects (bilateral hearing loss) participated in the present study. Each subject was paid the equivalent of $7 for their participation. Their mean age was 46. All subjects used hearing-aids. The pure-tone average hearing-loss was 51 dB ( I S 0 389; calculated for the better ear over the frequencies 500, 1000 and 2 000 Hz). The audiograms belonged to most recent medical records for each subject. The average number of years as handicapped was 28 years. The average number of years which the subjects had used their hearing-aid was 10 years. All subjects were able to receive spoken instructions from the experimenter. Finally, none of the subjects had any previous experience of formal speechreading practice. Materia1.y and procedure All subjects participated in two experimental sessions. In session one, all subjects were given two tests: ( I ) a speechreading test and (2) a vocabulary test. Each subject was tested individually on each of the two tests. Experimental sessioii 1 Sentence-based Speechreading test. To measure the subjects’ sentence-based speechreading ability, 24 different sentencesto-be-speechread were presented (i.e., same test as in Ronnberg et al., 1989; Lyxell & Ronnberg, 1989, 1991 a; Johansson et al., 1991). These 24 sentences were divided

into three blocks, each with eight sentences. All eight sentences in a given block were presented together before any sentence from another block was presented. Prior t o the presentation of the sentences in each of the three blocks, the subjects were instructed to read an outline story which was printed on the answer-sheet and informed the subjects about a scenario where a hypothetical two-way conversation took place. Three different scenarios were used (i.e., a ‘train scenario’: a ‘restaurant scenario’; and a ‘shop scenario’), each with a unique outline story. More specifically, the outline stories instructed the subjects to take a particular role within this two-way conversation (i.e.; a conductor on a train in the ‘train scenario’: a shop-hand i n the ‘shop scenario’: and, finally, a waiter/waitress in the ‘restaurant scenario’) and to speechread a counterpart (e.g., a passenger on the train). For half of the sentences the subjects were offered an extra contextual cue. This cue informed about the semantic content of the sentence-to-be-speechread (e.g., a sample contextual cue for the ‘train scenario’ could be “this sentence is about ticketchecking”). The contextual cue was printed on the answersheet and the subjects were instructed to read this prior to the presentation of the particular sentence. The sentences in the test were according to the author’s best judgement. simple declarative Swedish sentences. The number of words within each sentence was either 3 words, 6 words, 9 words or 12 words. The number of syllables varied for each sentence length with 4-5 syllables. 9-10, 13-15 and 18-22 syllables, respectively. Each of the three outline stories consisted of eight sentences, where each sentence-length was replicated once. The eight sentences within each outline story were then subjected to a randomization procedure. Four different randomization orders were created. each according to the following principles: if one sentence of a certain length was presented with a contextual cue. the replicate sentence of the same sentence-length in the block of eight sentences had to be presented without a contextual cue. Furthermore, if a sentence of a certain length was presented with a contextual cue in the first of the four randomization orders, the same sentence had to be presented without such a cue in the second order, and consequently, if a sentence was presented uithout a contextual cue in the first randomization order, the same sentence had to be presented with such a cue in the second order. Randomization orders 3 and 4 were the reverse of orders 1 and 2, both within and between blocks. The materials were then video-taped. The details of each sentence presentation were as follows: First, the videoscreen displayed a red colour for a period of 5 s; following this, the male actor (B. L.) appeared on the videoscreen. The actor was silent for a period of 3 s, before and after the actual presentation of the sentence. The presentation was recorded with sound, but presented without sound. After the presentation, the response interval started (25 s), and during this period the TV-screen displayed a grey colour. This procedure was repeated for all sentences in the test. At the recording session, great care was taken t o prevent shadows o n the speaker’s face. Only the head and shoulders of the speaker were allowed to appear on screen. In the test session a 26“ Luxor colour TV-set was used, which created an almost life-size image of the speaker. During the test session the subjects were placed at a distance of 3 m from the screen. The subjects were instructed t o write in a verbatim manner

Verbal abilify and speechreading


the words that they had been able to speechread, and they were also encouraged to guess the rest of the message. Scoring of the responses was carried out with respect to both correctness of the word and their correct position in the sentence. Credit was not given for a correct word in a sentence if the sentence was unrelated to the frame-history and contextual cue (when this was presented). A correct word in the wrong position in the sentence was not classified as correct, even if the sentence was related to the outline story and contextual cue. The number of correctly recalled words per sentence was expressed as proportions for each sentence and subject when included in the statistical anaysis (ANOVA). Test ofvocabulary size. The test of vocabulary size used in the present study was a test of knowledge of antonyms. This test was a part of an intelligence test (‘F-testet’, published by Psykologiforlaget. Stockholm). The test consisted of 70 items (one target word and four other words, one of which was an antonym to the target word). The subjects were allowed 14 min to solve the task.

One hundred items were used in the test, half being real words. and consequently, half were not. The real words used in the present test were all, according t o AllCn (l970), familiar words. The same presentation order as for the matching tests was used. Semantic decision-making. The subjects’ task was t o decide whether a word belonged t o a certain pre-defined semantic category or not (cf. Shoben, 1982). Four trials were used with 24 to-be-categorized items, of which 12 times belonged to a semantic category and 12 items were lures. The four respective trials were: ‘colours’, ‘occupations’. ‘diseases’, and ‘parts of the body’. Presentation characteristics were the same as for the other lexical access tests. For all long-term memory access tests, latency data were based on the average of each subject’s y e s h o responses.

RESULTS Table I gives the mean performance and standard deviation for each test in both sessions in the present study. For the speechreading test, the results are similar to previously reported results using this speechreading test. That is, there is significance for the contextual support variable such that performance increases when contextual support is offered (t(17) = 2 . 2 6 , p t 0 . 0 5 ; Lyxell, 1989; Lyxell & Ronnberg, 1989, 1991 a , b; Ronnberg, 1990; Ronnberg et al., 1989; see also Garstecki & O’Neill, 1980; Samuelsson & Ronnberg, in press). The resuIts obtained for the lexical access tests replicates previously reported data. That is, physical matching is faster than name matching (Posner & Mitchell, 1967; see also Perfetti, 1985), semantic decision making is faster than lexical deci-

Experimental session 2 In experimental session 2 the subjects were given four tests o f lexical speed: ( I ) a semantic decision test, (2) a lexical decision test, and (3) Posner and Mitchell’s physical and (4) name matching tasks. General procedure for the lexical access tests The administration followed the same procedure as in the Ronnberg et. al. study (1989; Lyxell, 1989; Ronnberg, 1990). This means that the lexical access tests were administered by means of a computer programme: TIPS (Text-lnformationProcessing-System). With respect to reaction-time, the same subroutine is used in all kinds of access tests: and hence, no source of variation in precision emanates from the software per se. The order of test presentation is automatically randomized for each subject by TIPS, and thus, unique for each subject. The stimulus material used in this study is exactly the same as the material used by Ronnberg et al. ( 1 989).

Lexical access tests Physical matching. The subjects’ task was to judge whether two simultaneously presented letters had the same physical shape (e.g., A-A), or not (A-a). The letter-pairs were presented by means of a computer (Apple’s Lisa), and the subjects were to respond by pressing predefined buttons for ‘yes’ and ‘no’ answers. Half of the pairs were identical and half were not. Four different pairs of letters (i,e., a-b, n-h, d-e and r-t) were used in four trials. Eight pairs of letters were used in each trial. The letter-pairs were (one at a time) presented in a fixed-text window (in the middle of the screen) for 2 s. The reaction time was measured from the onset of the 2 s interval. After the response a 2 s interval started before the next letter-pair was presented. Name matching. Instead of matching letter-pairs for their physical identity, the subjects were asked to match the same letter-pairs as in the physical matching test for their name. That is, if the letter-pair ‘A-a’ was presented on the screen the correct answer was ‘yes’. Sixty-four responses were collected in four trials of 16 responses. The same letter-pairs were used as in the test of physical identity. Lexical decision-making. The subjects’ task was to judge whether a string of three letters constituted a real word or not.

69

Table I. Means and SDs for each fesf in the presenf study Type of test

Mean

SD

Speechreading tasks

,

Overall performance Contextual support No contextual support Short sentences Long sentences

.31 .37 .24 .30 .31

Lexical acce.m tasks Physical matching Name matching Lexical decision making Semantic decision making

.79 .92 .83 .72

.I 9 .2 I .I0 .I5

.76

.09

Vocabulary test Overall performance

Scand Audio1 21

70

B. Lyxell and J. Ronnberg

Table 11. Correlation coefficients between performance on the speechreading test and the verbal tests Test

Sprechreading test 1. Overall 2. Short 3. Long 4. Context 5. Nocontext

2

3

4

5

6

7

8

9

.96'

.95" .83"

.93" .93" .85"

.92" .85' .92" .7ZU

-.I1 -.03 -.20 -.07 -.I4

.OO .09 -.09 -.03 .05

-.34 -.20 -.47' -.38 -.25

-.29

.12

-.16

.o1

-.42h -.30 -.24

.22 .I3 .09


Lexical access tests 6. Physical matching 7. Name matching 8. Lexical decision making 9. Semantic decision making

.82"

.6gh .62h

10

.49b .43"

-.23

.75u

-.42h

.oo

- .47b

Vocabulary test 10. Antonyms p