Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
CLOZE PROCEDURE AS A TECHNIQUE FOR THE INVESTXGATION OF SOCIAL CLASS DIFFERENCES IN LANGUAGE USAGE W.P. ROBINSON University of Hull Cloze procedure was used to investigate the nature and extent of the differences in verbal behaviour of working and middle class boys. Words were deleted in sentences taken from ‘formal ’ and ‘ informal ’, middle and working class letters and from‘middle and working class oral utterances. k h e results showed that the middle class boys used a wider range of words and preferred different words in this situation. The working class boys showed more conformity in their responses than the middle class boys, especially for the written materials. \ Fruitful lines for further research on ‘restricted ’ and ‘elaborated ’ codes are discussed. PROBLEM Bernstein (1961, 1962a, 1962b) has drawn a distinction between ‘ restricted ’ and ‘ elaborated’ codes. He has formulated rules for separating them empirically and commented upon the functional significance of this differentiation. The evidence he has collected is consistent with his theoretical premises. There are many types of ‘ restricted ’ code, but Bernstein is particularly interested in the one to which many working class people may bc currently confined. He suggests that working class language serves mainly to change or maintain social relationships and that its structure fits it for this function. Middle class people also use ‘ restricted ’ codes, but in addition they have an ‘ elaborated ’ code which is structured to serve as a major means of communicating information about the external world and personal intentions and feelings. An initial problem is to specify the differences between the two codes and subsequent problems are to find out if the working class has no ‘ elaborated ’ code, why it should have none, and what the significance of this would be. Bernstein has said that the ‘restricted ’ code of the working class will differ from the ‘ elaborated ’ code of the middle class in the following ways : (1) The lexicon will be different. (2) The grammatical structure will be different. (3) Utterances will have low “information” content even for someone who is familiar with the code : this means that the structure and content of utterances will be predictable. As a result of this it might follow that working class utterances will be fluent, fast and may well contain long phrases. (4) There will be more introductory and final phrases in working class language which serve to test that the desired effect is being registered on the listener. Bernstein’s evidence relates to each of these propositions, but it is based upon a single study in which small groups of working or middle class boys discussed capital Downloaded from las.sagepub.com at PENNSYLVANIA STATE UNIV on May 17, 2016 43 This is a situation in which it would be expected that the middle class boys would utilize an ‘ elaborated ’ code. It is not quite clear what the working class would do. If they had no ‘ elaborated ’ code, then they would use a ‘ restricted ’ one. if they had an ‘ elaborated ’ code available, there might still have been situational factors which would make them use a ‘restricted ’ code. They may have seen the situation differently from the middle class boys. Another possibility is mentioned later. This does not vitiate the dserences found in the speech samples from the two groups, but it does leave open the possibility that the working class boys had an ‘ elaborated ’ code available, but chose not to use it. Studies relevant to this issue are in progress, but the present study seeks to examine the validity of only two of the four propositions (1 and 3) in situations where the probabilities of using an ‘ elaborated ’ or ‘ restricted ’ code are varied. To what extent are middle and working class subjects using the same lexicon ? Are working class utterances more predictable than middle class utterances? At the same time it is hoped that the efficacy of ‘ cloze ’ procedure for use in this field can be tested (Taylor, 1953). Prima facie the technique should be ideal j subjects are given sentences with one word omitted and asked to fill the gap. This makes the situation reasonably natural interesting as well as manipulable. There are difficulties in posing the theoretical problems in empirical terms. That people may treat the same situation differently and choose to use a particular code makes it difficult to interpret Bernstein‘s results as being the consequence of the working class not having an ‘ elaborated ’ code. Like Bernstein’s work this study will concern itself with vocabulary and grammar in use and not the vocabulary and grammar which could have been used. Hence the results will tell us more about how working class boys behave than how they might behave. However, it will not be assumed here that these boys have no ‘elaborated’ code, but only that any such code they may have will be less developed than that of middle class boys. Whether or not working class utterances are more predictable than middle class utterances raises at least three questions : how utterances are to be collected, what is to a u n t as a correct prediction at a linguistic level and whether ‘correatness’ should be dillerently defined for different samples of subjects. The efficiency with which words omitted from sentences can be guessed will have many determinants and it will be important to control for all those not relevant to the problem in hand. ‘ Function ’ words (e.g. prepositions) are generally more predictable than ‘ content ’ words (e.g. nouns) and some control must be made for this. Similarly there is variability across situations and across people regardless of social class and some comparability of situation and persons for deriving the utterances is nceded as a control. The ambiguity of the meaning of predictability lies in the diaculty of deciding what word filled into a gap in a sentence should count as correct. It is unsatisfactory to restrict correct responses to the original word deleted, since that was but one instance of a response. Hence the measure should be based on the variability of the responses given by a large number of relatively homogeneous subjects giving one guess Downloaded from las.sagepub.com at PENNSYLVANIA STATE UNIV on May 17, 2016 44 Cloze Procedure for Investigation of Differences in Language Usage or a small number giving many. The smaller the variability the greater is the predictability. Whether two different word tokens with the same external referent and whether variations in the tense or mood of verbs should be counted as similar must also be resolved. The third problem is to consider who is predicting. Familiarity with a code will be a major factor affecting efficiency of prediction. If middle class children are more familiar with their own ‘ elaborated ’ code than they are with the partidar ‘ restricted ’ code of the working class children, then there may be less variation in their responses for the ‘ elaborated ’ code. This would give information about the subjects, but little about the codes and their relative redundancy. Further, it is premature to assume that the ‘restricted’ codes of the middle and working class children are similar linguistically. T o meet this complication the hypotheses must be elaborated. In general, utterances m ‘ elaborated ’ code wiU be less predictable than utterances in ‘ restricted ’ codes. This will be true of middle and working class children separately (assuming that the working class children have an ‘ elaborated ’ code). However, in comparisons across social class this may not be true because of differential familiarity with the codes. Working class children should find middle class language in ‘ elaborated ’ code more unpredictable than working class language in ‘ restricted ’ code, but the predictability for middle class children in this interaction cannot be predicted. Further, working class children should find it easier to predict words omitted from working class samples of language than from middle class samples, but again the middle class children’s performance cannot be predicted. METHOD Subjects. The subjects were twelve pairs of 13 year old boys matched on verbal (Moray House Verbal Reasoning Test, mean of two scores) and non-verbal (Raven’s Progressive Matrices, 1938) intelligence test scores, but contrasted on social class. One set of boys had fathers whose occupational ranking on the Hall-Jones scale (Hall arid CaradogJones, 1950) was 4 or above and the others were all 5 or below. An attempt to maximize the social class contrast between the groups was limited by the size of the sample from which these groups could be obtained. This sample comprised the 70 boys in the top five streams of a Comprehensive School. Some 10% of the children in this school’s catchment area went to a very restricted Grammar School. The children lived either in the country or in a market town. The characteristics of the sample are summarized in Table 1. The mean difference score for the I.Q.’s was calculated without regard to the direction of the difference. Materials : Vocabulary Test. Twenty-four words were selected from the responses of the top two forms. These words were judged by the experimenter to be relatively uncommon. Ten of these with Thorndike Lorge Word Counts of 20 or less were used to test the children’s recognition vocabulary. These were : entracing (labelled A, but this seems unlikely in current usage), disperse (II), lease (14), r d e m (3)’ receipt (20), strenuous (9),picturesque (18), accumulate (17), aggravate (6), cafdogue (12). Downloaded from las.sagepub.com at PENNSYLVANIA STATE UNIV on May 17, 2016 W.P. Robinson 45 TABLE 1 Matching of Subjects MIDDLECLASS n = 12 Mean Social Class (1-7) Mean Verbal I.Q. Mean Non-Verbal I.Q. Mean Da. Verbal LQ. Mean DB.Non-V&bal I.Q. WORKING CLASS n = 12 3.2 107.7 42.2 5.9 106.6 42.3 2.7 1.4 C b z e procedure. The set of sentences listed in Table 2 was collected. Two major sources were used, letters and group discussions. The letters were of two types : formal letters written to a school governor justifying an a w d of a prize for a holiday and informal letters written to a named good friend who had been away from school for a fortnight. It was hoped that the formal letters would tend to be in ‘ elaborated ’ and informal letters in ‘ restricted ’ code. The deletions comprised nouns (4), adjectives (4), verbs (4), prepositions (4) and conjunctions (2), selected in pairs from woiking and middle class boys’ letters with qua! numbers from formal and informal letters. A more complete coverage of the grammatical categories was not possible in the time available for testing. The spoken utterances were collected from two tape-recorded four person discussion groups, one of middle class and one of working class bog. Each discussion was in two parts. In the first part the groups were asked to discuss ‘ what sort of things are wrong and why do you think they are wrong ’. They were given a list of twenty items, e.g. cheating, boasting, selbhness, to guide their discussion. In the second half they were asked to talk with each other about what they had been doing in the last week and the conversation was allowed to drift. There was no obvious difference in either group between the formal and informal conversation. Four grammatically matched deletions were selected, but the remaining items chosen were simply those which the experimenter felt might yield interesting ditTerences between the subjects. Again time available as well as inadequacy of raw material prevented a wider coverage. This meant that for written language four types of utterances were available : middle class ‘ elaborated ’ code, middle class ‘ restricted ’ code, working class ‘ restricted ’ code and working class ‘ elaborated ’ code. The last category may be falsely labelled. The situation was such that an ‘ elaborated ’ code was most suitable for discussing the problem. One additional item included was item 24. This was of interest s h t ~ Bernstein has shown a preference for working class boys to use ‘ we ’ rather than an impersonal ‘ you ’ or ‘ one ’ (1962b). For spoken language, no attempt was made to separate the formal and informal discussions. Four items matched on grammatical unit were taken from each group. Downloaded from las.sagepub.com at PENNSYLVANIA STATE UNIV on May 17, 2016 Cloze Procedure for Znvestigatwn of Differences in Language Usage TABLE 2 Sentences used for ‘‘ cloze ” procedure PART OF SPEECH OMnTED On Sunday I went there on my sledge, it was.. I know the numbers are limited and.1 hope you can. ..me in. We are going for a bike ride some time next week, so if you sent your address we . . . come and see you. When it comes in it comes rushing through door, round table, straight . . . kitchen and Gob, Gob Gob its tea down. I would be able to watch the athletes while they limber up and perhaps they might give me a few tips which they have . . Don’t you think it’s a bit . . here, they have the-if you’re late-you’re just a bit behind the bell, they put you down as late. I still went to the . . and afterwards I saw Chris who told me what was wrong with you. I am sure I shall not be alone with this wish for the . . . would be both enjoyable and educational. It can only benefit the boys . . . this school. There hasn’t been much happening here except that it’s still. . .down with snow. I would consider this to be an excellent opportunity to extend and satisfy my knowledge. . . the subject. And then when you . . .you think to yourself, ooh blimey, why the heck did I start up this. You get a lot more . . . out of a thing if you’ve saved up for it for a long time. I would also like to go . . I could mix with the Americans and get to know their habits. Whilst I was there I saw a . .. fight between two boys. One of them received a cut eye but it was not too bad. . the hill they were selling tea, and coffee, and sandwiches, they were doing a roaring trade. . . . . .. . CLASS ORIGIN COMMUNI- FORMALITY CATION adj. M I L verb W F L verb’ M I L prep. W - S verb M F L adj. M - S noun 1 M I L noun prep. M W F F L verb W I L prep. M F L verb M - S noun h.1 - S conj. W F L adj. W I L prep. W I L Downloaded from las.sagepub.com at PENNSYLVANIA STATE UNIV on May 17, 2016 L W . P. Robinson 47 TABLE 2 (CONTINUED) PART OF SPEECH OMITTED ORIGIN FORMALITY COMMUNICATION conj. M F L noun W F L pronoun W - S Prep. M - S noun W I L Prep. adv. W - I L S pronoun M W verb - F L L This course is called a holiday, but it certainly is no rest . . you will have walked, climbed, and struggled over 200 miles by the time you have finished. Thank you for letting me have the . , , for this and I gladly exspt. Some people do-most people- . . . and him don’t. I got these shoes under the impression that they were for best, but I’ve ended by wearing them. . . school. A lot of your . . . were against having it shot, but it had to be. . . . the holidays I am going to spend a few days train spotting. You can’t say you . . .said it now. Then on leaving the summit go back down the slope and relax on the side of one of the lakes. I shouldn’t have . . . my trap, should I ? To know a place one must be able to make a careful study of . . . places. You’re rude you are and you shouldn’t be rude, ’cos it’s . . . I would also like to go up to the top of Ifell tower because you get a . .look of France and also the river Sein. Boy 1 “ Do you know what I spend my pocket money on ? ’’ Boy 2 “ What ? No ” Boy 3 ‘‘ . . . you don’t bet any.” If you back out of a fight, it’s best to back out. You want to watch out you don’t get. . . . . . . . .’that hurt that did !3 CLASS . .. . . M = middle class W = working class F = formal I = informal M adj. M F L adj. W - S adj. W F L conj. W - S nounz W - S S = speech L = letter ‘ I t was mi experimenter’s error to include an awiliary part o j a verb rather than a verb stem. ‘ M o s t answers mere p a r participles passive. This item was given the context : Spoken-Boy to self (dropping something on his foot). Downloaded from las.sagepub.com at PENNSYLVANIA STATE UNIV on May 17, 2016 48 Cloze Procedure for Inaestigaiion of Differences in Language Usage Four additional items were included : Item 23 involved the possible use of a double negative believed to be a characteristic of working class speech. Fries (1940) found this to be rare, but more rare in middle class letters. In item 19, ‘me’ as the subject of a sentence was omitted. Item 31 was invented to see whether the subjects were prepared to give honest replies and what social class differences might emerge. Item 30 offered the choice between a passive participle and a noun. The word deleted was ‘stick’. ‘ T h e ’ is often omitted in the local speech although there is a glottal stop preceding the noun. There was a chance that working class boys would tend to use the noun, especially as the sentence was in fairIy common use. Procedure. Subjects performed four tasks in two school periods. In the first period they wrote a formal or informal letter and then performed the Cloze task. In the second period the other letter was written and afterwards three classes were given‘a short vocabulary test and two classes re+xamined their answers for the Cloze task and rank-ordered their earlier replies in terms of best answers. Znstrudions for Clo8.e tusk : “You each have a booklet in front of you. Each page has a sentence on it with one word missing shown by some dots like this. You have to read what it says at the top of each page where it says whether the sentence was spoken or written and it also tells you to whom it was said. All you have to do is decide which words you think could be put into the gap. Zmugine you have said or written the sentence. What word might you have used ? Under (a) put the first word you think might fit, under (b) another word. The order you write them in will be the order in which you think of them. We will have 1 minute for each sentence. I’ll tell you when to turn the page. Put down as many as you like. Short words, long words, any sort of words. This is not a test of any sort-just like Iast time. I’m the only person who will see what you write. Don’t look at each other’s. I’ll just do an example.’’ The example was done by the experimenter with tbe class helping. Instructions for second Ctoze task : “Yesterday you filled in gaps in these sentences. If you look at the set of words you wrote, you may now think some of your answers do not fit. The first thing to do is put a X by answers that don’t fit. The second thing I want you to do is pick out the most likely word of the set and put a 1 beside it, then a 2 by the next most likely and a 3 by the third. You will have half a minute for each page and I will tell you when to turn over.” A demonstration was given, questions were answered and the task was completed. Results from this task are not reported here. Instructions for Vocabulq Test : “ Here are ten words. I would like you to write down what they mean if you know. If you don’t know, don’t WOKY and don’t copy. This is not a test ”. Downloaded from las.sagepub.com at PENNSYLVANIA STATE UNIV on May 17, 2016 W. P. Robinson 49 TREATMENT OF RESULTS Lexicon. Two methods were used to examine the extent to which both groups of subjects were sampling from the same vocabulary : (1) The total lists of different and same responses for each item were compared across groups. (2) The total lists of different first responses for each item were compared across groups. These measures give some indication of the range of words which might be used by the two groups. The next stage was to devise a way of assessing whether or not the two groups :ended to choose different responses within this sample. Only first responses were Jsed. For each item the nature and frequency of the most common response of each social class group was noted and this was compared with the incidence of this s m e response in the other group.. The final problem was to construct an index which would indicate the information content of the gaps in the item for each group. A measure of redundancy could have been obtained in information theory units, but it was considered that the number of subjects available was too small to make this a reliable estimate. Two other possibilities were : T o count the most frequent response for any item in each group and contrast it with the most frequent item in the other group regardless of whether it was the same word. If one group is much more agreed on the most common response than the other, this would suggest greater redundancy for the gap for that group. This first measure does lose some information which might be useful. For instance one group might have a most frequent response of 6, with all the other responses being different. The other group might have 5 responses of 2 each and 2 of 1. This measure takes no cognizance of the paired responses and would show Group 1 to be much higher on conformity than Group 2. T o take account of this a second measure was devised which would count the total number of different ' types ' given by each group. For the example above, this measure would now give equal answers No. of types - 1 which gives for the two groups. The actual formula used was 1 No. of tokens values varying between 0 and 1 and may be termed a coefficient of conformity. RESULTS Vocabulary Test. When the Wilcoxon hlatched Pairs Signed Ranks Test was applied to those matched pairs of subjects who were given the vocabulary test, no significant difference between the groups was found. This suggests that there may be no sigruficant difference between the ' recognition ' vocabulary of the two groups. Since verbal intelligence test scores correlate highly with measures of vocabulary this is not surprising. The mean vocabulary score of the working class group was 7.5 and of the middle class group 7.0. Group differences in range of vocabitlary rued. A list was made of all the different responses given by each group as possible words to fill the gaps. A count was made Downloaded from las.sagepub.com at PENNSYLVANIA STATE UNIV on May 17, 2016 so Cloze Procedure for lnvestigation of Differences in Language Usage TABLE 3 Summary of total number of different responses to word gaps by social class (items grouped by part of speech omitted) OCCURRENCESOFDIFFERENTRESPONSE PERCENTAGE Middle class Working class SAME RESPONSE Item No. 8 18 7 21 13 30 nouns 26 28 1 15 6 27 adjectives 5 2 3 10 12 25 YY >Y YY YY YY YY YY Y> YY YY verbs YY >Y YY >Y YY 11 9 22 16 20 4 prepositions 24 19 23 31 pronoun YY YY YY YY YY YY adverb expletive 34 42 40 31 29 31 31 21 34 32 20 30 14 14 28 33 29 20 30 24 27 38 37 27 28 20 26 25 22 23 21 29 32 21 20 28 21 18 8 26 24 30 14 12 5 15 17 21 35 25 63 17 21 38 13 16 11 14 12 17 17 11 11 16 14 11 36 35 47 17 22 16 15 10 14 7 13 19 33 28 40 Downloaded from las.sagepub.com at PENNSYLVANIA STATE UNIV on May 17, 2016 58 53 33 W . P. Robinson 51 of how many words the two groups had in common and how many were peculiar to one group (Table 3). Strict criteria for differentiating words were used ; tense or mood variations were scored as separate responses. Some of the responses were in fact quite unsuitable words for the gaps and the differences between the used vocabularies of the groups is exaggerated. A test of whether the two groups are sampling from the same vocabulary cannot be made. Different subjects have contributed different numbers of words and these are not necessarily independent. In addition the question about sampling from different vocabdaries cannot be answered in statistical terms. It is only possible to calculate the percentage of responses common to the groups and comment on this measure of agreement. For content words, nouns, verbs and adjectives, the common vocabulary used is generally between 20% and 30% (median 25%). For function words, prepositions, conjunctions and the auxiliary part of a verb (item 3) the agreement is higher (median 39%). If it may be assumed that the different responses within a group are independent of each other, then an analysis shows that over the thirty items (item 31 excluded) middle class boys are making more different responses than working class boys. They give more different responses to item 25, fewer to item 4 and the same to item 1. This difference is highly significant (Binomial Test, = 4.2, p < 0.00006). Group differences in relative frequency with which words ure used. Although the total range of vocabulary used in this situation for the two groups differs, the relative incidence of certain words among first choices may be different or the same. The differences found in vocabulary may be a function of different choices among lower level choices. A simple test was made to investigate this. For each sentence for each group the most commonly used word among first choices was noted and its frequency counted. The incidence of this same word in the other group was counted. For the first comparison a list was obtained of the incidence of the most frequent choice of the middle class group across all items. The incidence of these words in the working class responses was contrasted with the first figure and items were classified in three way : middle < working, middle = working, working < middle. This procedure was repeated with the most frequent working class responses as standards for comparison. These data were subjected to the Binomial Test and the results showed that middle class boys differed from working class boys in the relative frequency of usage of certain words in this situation. The word with the most frequent first response in the middle class was more frequent in this class than in the working class on 18 of the 25 words ( 2 = 2.00, p < 0.05) : 6 gaps had more than one most frequent word and were not included. Conversely the most frequent working class first choice words had a higher incidence in their own group than the middle class group. Two were not testable, 23 were greater ( z = 2.98, p < 0.003). Group conformity of response in Cloze procedure. Sentences were separated into the sets shown in Table 4. Two indices of group conformity were calculated for each sentence and the groups were contrasted by a Binomial Test. The first index was a count of the most frequent word used to fill a gap within each group. The most frequent word for that gap in the other group was counted and a table drawn up Downloaded from las.sagepub.com at PENNSYLVANIA STATE UNIV on May 17, 2016 52 Cloze Procedure for Investigation of Differences in Language Usage TABLE 4 Cloze procedure : Coefficients of conformity in first responses for middle and working class boys Origin : Middle Class Formal Item MC wc 8 0.00 0.10 26 0.18 0.27 5 0.46 0.55 0.64 0.46 11* 0.55 0.64 17 24 0.60 0.73 Origin : Working Class Formal Item MC wc 18 0.00 0.36 28* 0.64 0.55 2 0.46 0.73 9 0.55 0.73 14 0.27 0.55 Middle Class Informal 7 0.36 0.46 1* 0.46 0.36 3* 0.82 0.73 0.64 0.73 22 Working Class Informal 21 0.46 0.73 15* 0.55 0.46 10 0.36 0.46 16 0.82 0.82 Middle Class Spoken 0.46 0.64 13 6* 0.27 0.09 0 . 0 9 0.18 12 20 0.73 0.73 Working Class Spoken 0.46 0.36 30* 27* 0.64 0.36 0.64 0.50 25* 0.64 4 0.55 29 0.36 0.60 ’ 31 Extra 0.36 0.55 Binomial Test : All items z = 2.22, p = 0.0132 WC > MC Written items z = 1.84, p = 0.0329 WC > MC Spoken items z = 0.00 WC = MC * Items deviant from the majority. classifying sentences showing a greater, lesser, or the same degree of conformity for the middle class group. The second index was the variant of the type-token ratio mentioned earlier ; the coefficient of conformity. It was expected that the working class should show greater conformity of response than the middle class boys and this should be particularly evident in the working class language samples. A reversal might appear with middle class language samples if the familiarity with the code’s nature becomes more important than the code’s redundancy. Both indices gave the same result. There is an overall greater conformity of response among the working class boys. Downloaded from las.sagepub.com at PENNSYLVANIA STATE UNIV on May 17, 2016 W . P. Robiiisori 53 For the first index, working class boys conformed more than,middle class boys in their most frequent first response (Binomial Test, z = 2.01, p = 0.0222, one tailed test). This is mainly due to the written items (a = 1.54, p = 0.0618, one tailed test). For the second index the conformity of the working class was greater at a significant level (p = 0.0132, one-tailed test) and this again was mainly due to the written items (p = 0.0329). Further examination within types of language renders this simple confirmation more dBcuIt to interpret. For the gaps in working class language, the working class boys did not show significantly greater conformity than the middle class boys and this seems to be due to the results with the spoken working class utterances where three out of five items show greater conformity for the middle class group. Certain qualitative observations may be made for future investigations. Item 24 : Bernstein has suggested and found that the working class use ‘ we ’ where the middle class would use ‘ one ’ or ‘ you ’. 6 working class boys used ‘ we ’)but only 3 middle class boys. For ‘ you ’ the figures are 2 and 3. Items 14 and 29 : Bernstein showed that words like ‘ because ’, ‘ so ’, ‘ as ’, occur more frequently in working class speech and the same was found here. In the two items combined, working class boys gave 10 responses of this type, middle class boys 3. Item 23 : Although the double negative was rare in both groups, working class boys (3) showed less aversion to it than middle class boys (1). Degrees of conjomtiiy in f o r d und informul Zunguuge. If the ‘restricted’ code of informal language is more predictable than the ‘ elaborated ’ code of formal language, then the coefficientsof conformity for filled gaps in samples of that language should be higher. 8 formal items matched for the part of speech omitted and social class origin could be contrasted with 8 informal items (8, 26, 5, 11, 18, 28, 2, 9, 14 versus 7, 1, 3,22,21, 15, 10, 16). When this was done for the combined coefficients of conformity of middle and working class subjects, 6 of the 8 pairs of items showed more conformity for the informal items. A 7/1 split would have achieved statistical significance, but this result suggests that a more satisfactory test should be fruitful. DISCUSSION The results show that with these groups of subjects differences in Cloze procedure performance were not correlated with ditferences in recognition vocabulary. The control for verbal intelligence test scores should have insured this and the vocabulary test showed this presumption to be valid. Differences found were therefore differences of vocabulary in use. The &st set of results showed that middle class boys were in general using a larger range of words than working class boys. The concordance in the vocabulary of the two groups was greater for ‘function’ words and less for ‘ content ’ words. T o find if working class and middle class boys are sampling from different vocabularies more information would be required. Further samples of subjects of each class would permit a comparison within and across groups and the data could be taken beyond the bald statements of concordance given here. Downloaded from las.sagepub.com at PENNSYLVANIA STATE UNIV on May 17, 2016 54 Cloae Procedure for Investigation of Differences in Language Usage Further analysis showed quantitative as well as qualitative differences between the groups. The most frequently used response in one group was not used by the other group to the same extent, although working class boys showed greater conformity of response. Since the boys were matched on verbal and non-verbal intelligence test scores, were of similar age and in the same forms of a single school the differences found are surprising. Social class, crudely defined by father’s occupation is sufficiently relevant to produce a differentiation. Apart from that most dangerous source of information ‘ common knowledge ’, facts about social class difference in vocabulary used are rare. Fries (1940) adysed a large body of letters written to government agencies (i.e. formal letters) and produced a long list of social class differences. The usefulness of this work may be unfortunately limited by the absence of a control for I.Q. and the omission of data analysis beyond simple tabulation. However, no student of qualitative differences in language could d o r d to ignore either the facts or the linguistic conceptual framework of this mammoth study. Bernstein (1%2a, 1962b) has found qualitative and quantitative differences consistent with the results here. The results on the redundancy of the gaps in sentences support the view that working class children exhibit greater overall conformity in response in Cloze procedure on two alternative measures. Although these differences are not large, the close matching of the subjects again adds to their value. The number of sentences from different origins was too small to allow adequate tests of differential in social class conformity across the written-spoken, forrnal-informal dimensions. There was a hint that the generally greater conformity of working class responses was not evident in the sample of working class spoken utterances. . There are several possible reasons for this. It could be that working class spoken utterances are treated differently by the two groups. While working class boys may put in words that they might use, the middle class boys may be using a stereotype ; they may see the sentence as “working class” and make the word fit their idea of working class language. If there is a cultural stereotype available this could lead to the middle class showing greater conformity for this type of language, regardless of whether the stereotype is exaggerated or inaccurate. This methodological problem could occur in most techniques currently used in this field, but the problem can be investigated independently. A second possibility is that differences would emerge if the presentation of the material and the responses were oral rather than written. The use of edited tape recordings individually administered would resolve this. A third possibility is that the sample of spoken utterances only contained utterances in ‘ restricted ’ code and the social classes do not differ in their ‘ restricted ’ codes. This could be tested by increasing the number of spoken middle class items to see if working class boys are as effective at predicting gaps in these as middle class boys. If they are, then this would support the parsimonious view that there is only one ‘ restricted ’ code used by the two groups in this situation. A further study should also extend the range of spoken utterances. No distinction could be made here between formal and informal discussion. It may be that boys Downloaded from las.sagepub.com at PENNSYLVANIA STATE UNIV on May 17, 2016 W.P. Robinson 55 of this age are not really capable of using an ‘elaborated ’ code in discussion, but it does not seem possible to do more than define the code by the nature of the situation in which the utterances are made. If repeated samples from such discussions fail to differfrom utterances in informal discussion for either social class group, the distinction could not be maintained for the oral communication of boys of this age. Such a procedure would also enable us to decide on empirical grounds whether or not working class boys are abIe to operate with an ‘ elaborated ’ code. This elaboration of the scope of the study would also permit a better evaluation of the differential redundancy of formal and informal letters. It may well be that writing will show differences, but speech not. In conversation there is considerable pressure to act quickly to maintain the flow and it is not so easy to amend speech as writing. If the working class boy is confined to a ‘ restricted ’ code in speech this may only mean that he is unable to produce an ‘ elaborated ’ code under certain conditions, i.e. when under pressure to act quickly. His writing may well be in ‘ elaborated ’ code. He is given considerable information about the ‘ elaborated ’ code and has much practice in writing it. School gives him little practice in speaking with it, although this is what he most urgently requires. It should be added that these studies and Bernstein’s are best viewed as a necessary prelude to the major tasks. One final intention is to examine the extent to which social class differences in the units and structure of language are relevant to thinking. (A second is to describe the social class variation in social psychological terms.) Some social class differences will be found to be irrelevant to thinking (e.g. “ while ” in the East Riding often replaces “ until ” elsewhere, but this is probably to be explained in sociological or historical terms and has most significance for these disciplines). The specification of the differences should make it easier to move into the more difficult problem. Cloze procedure seems to be a useful way of exploring the nature of the language differences. I should like to thank Dr. Bernstein and Mr. Young of the Institute of Education, London, and my colleagues at Hull for helpful discussions. I should also like to acknowledge the facilities provided by the East Riding County .Council and the Headmaster of Longcroft Sch~ol,Beverley. REFERENCES BERNSTEIN, B. (1961). Social structure, language and learning. Educ. Res., 3, 163. B. (1962a). Linguistic codes, hesitation phenomena and intelligence. Imnguage ad BERNSTEIN, Speech, 5, 31. BWSTEIN, B. (1962b). Social class, h g u k t k codes and grammatical elements. Jhnguuge and Speech, 5, 221. FRIES, C. C. (1940). American English Grammar (New York). HALL, J. and JONES, D. C. (1950). Social gradings of occupations. Bnf. 3. SoCid., 1, 31. TAYLOR, W. L. (1953). ‘ Cloze procedure ’ : a new tool for measuring readability. ~ m d i m Qumlerly, 30, 415. Downloaded from las.sagepub.com at PENNSYLVANIA STATE UNIV on May 17, 2016