| { |
| "paper_id": "J75-3003", |
| "header": { |
| "generated_with": "S2ORC 1.0.0", |
| "date_generated": "2023-01-19T02:41:00.121730Z" |
| }, |
| "title": "", |
| "authors": [], |
| "year": "", |
| "venue": null, |
| "identifiers": {}, |
| "abstract": "", |
| "pdf_parse": { |
| "paper_id": "J75-3003", |
| "_pdf_hash": "", |
| "abstract": [], |
| "body_text": [ |
| { |
| "text": "as...!lbull sessi~ns.\" among three participants that, in the case of the children, were pot into motion by a friendly adult who faded into the background after the convers.ation warmed up, and in the case of the \"adults,\" took place in what was purported to be an experiment in small group process. Conversations were tape-recorded and then transcribed both in regular orthography and in phonemes. Various types of printed material were also analyzed, to represent \"written language\": school readers at several grade levels, trade books rated as liked by children, and adult material from a previous study by Newman and Waugh (Information and Control, 1960, 3 , 141-153).", |
| "cite_spans": [ |
| { |
| "start": 626, |
| "end": 659, |
| "text": "(Information and Control, 1960, 3", |
| "ref_id": null |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "", |
| "sec_num": null |
| }, |
| { |
| "text": "The principal mode of analysis--in fact essentially the sole mode of analysis--was inspired by information t?reory, and concentrates on the phonemes and graphemes of the corpus. Extensive tables (pp. give data on first-order frequencies and probabilities of letters and phonemes, and some of the higher-order sequential frequencies and probabilities (up to \"triphones\"* for phonemes, and word-initial and word-final \"tetragrams\" for graphemes). The discussion of these atis is tical tables occupies pages 23-45.", |
| "cite_spans": [ |
| { |
| "start": 350, |
| "end": 353, |
| "text": "(up", |
| "ref_id": null |
| } |
| ], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "", |
| "sec_num": null |
| }, |
| { |
| "text": "The book itself is a handsome production, printed in a tasreful style on high-quality stock and nicely bound in clolh On closer examination oE its contents, however, one is tempted to conclude that the authurs, having completed their manuscript and handed it over to the printer in about 1968, proceeded to dioest themselves of any further responsibility for its editing and publication. Only in this way can one explain the many egregiods typographioal errors, inexcusable editorial changes, and glaring omissions of important materials that are t~ be noted in the book.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "", |
| "sec_num": null |
| }, |
| { |
| "text": "The defects are in many instances serious enough to destroy much of the book's potential usefulness.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "", |
| "sec_num": null |
| }, |
| { |
| "text": "I infer that the manuscript was completed in 1968 or there-", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "", |
| "sec_num": null |
| }, |
| { |
| "text": "1 1", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "", |
| "sec_num": null |
| }, |
| { |
| "text": "abouts because a reference to a 1968 article is cited as in press, t t and there are no references to ahy publications subsequent to that year. Along with illiterate spellings and typographical errors such as pulication, concensus, idiosyncracies, and diagrams (for digrams, p. vii), we find inconsistent mathematical notation (pp. 18-20) and incorrect plotting of data points (in Figure 3. Are the authors c o r r e c t , however, i n saying they a r e studying differences between \"spoken language: and \"written language\"?", |
| "cite_spans": [ |
| { |
| "start": 327, |
| "end": 338, |
| "text": "(pp. 18-20)", |
| "ref_id": null |
| } |
| ], |
| "ref_spans": [ |
| { |
| "start": 381, |
| "end": 390, |
| "text": "Figure 3.", |
| "ref_id": null |
| } |
| ], |
| "eq_spans": [], |
| "section": "", |
| "sec_num": null |
| }, |
| { |
| "text": "Evidently the abthors mean t o draw the d i s t i n c t i o n n o t between \"written\" language and \"spoken\" language a s such (for any sample of language can be e i t h e r w r i t t e n down o r spoken alqud), but bet t", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "", |
| "sec_num": null |
| }, |
| { |
| "text": "tween \" i n \u20ac ormal language\" and \"formal language, i. e . edited language. That i s , they a r e concerned with how language i s generated.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "", |
| "sec_num": null |
| }, |
| { |
| "text": "They apparently f e e l tha't the most \" n a t u r a l , genu5ne1' language i s generated i n informal conversational s i t u a t i o n s where t h e speakers have l i t t l e i f any pressure, t o make t h e i r speech conform t o a r t i - Analysis. As mentioned previously, all analyses of the material were at a phonemic or graphemic level. The intention was to use the \"powerful tools of information theory\" to trace t h e development of \"redund8ncy. \" This. mission was certainly enough to occupy the authors throughout the course of their study; it was apparently their intention to leave other types of analysis to future workers: One may raise the question, however, whether inkormation theory analysis was really so \"powerful\", and indeed how infoamative it .was when applied solely to zero-and higher- It seemed to me, having done these computations, that they meant very little. I tabulated zero-order, first-order, and second-order sequential probabilities, esfLmates of information (H), and the like, b u t it seemed t o me that all that was being shown was that certain phonemes and combinations of phonemes were more frequent than others because of their appearance in words or sequences of words having the higher frequencies. people generate language, I", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "", |
| "sec_num": null |
| }, |
| { |
| "text": "reasoned, not by selecting phonemes but by selecting words and sequences of words; therefore, the frequencies of phonemes and their combinations were mere epiphenomena. Tabulatkons of these sequences mighc conceivably have some uses in designing stimulus material for psycholinguistic experiments, to control for t h e frequenci'es These authors muse have been disappoifitedwwith their findings on the zero-order distributions of letters and phonemes. They find that for the distributions of letters in the conventionally printed version of their conversational samples t h e r e is very little change over age. As they say, \"the largest change is in m. tain \"noise Gords\" like-\"urn\", \"well,\" etc. Yet the authors take pains to tompare their results to those of other investigators of phoneme distributions, claiming that \"the highest correlations usually occur between phonemic systems derived from material closest to nat~ral~speech, whereas the lowest correlations occur with phonemic systems based on material furthest from natural speech\"", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "", |
| "sec_num": null |
| }, |
| { |
| "text": "which", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "", |
| "sec_num": null |
| }, |
| { |
| "text": "(p. 29). Their argument is not convincing, however, for they seem to underestimate the effect of different systems of phonemic transcriptions used in various studies, and also the effect of tlie \"editing\" that occurs as one goes from Highly informal speech 1 1", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "", |
| "sec_num": null |
| }, |
| { |
| "text": "(with its noise\" words) to more highly edited speech (e.g., contrived speech in high-school plays).", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "", |
| "sec_num": null |
| }, |
| { |
| "text": "What the authors make most point of are t M differences", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "", |
| "sec_num": null |
| }, |
| { |
| "text": "among various types of material, spoken or written, in \"redundamy\" or \"relative sequential constraint\" as defined by informationtheoretic statistics. Actually, there are no differences between first-grade speech and the \"adult\" speech samples in phoneme redundancy--the curves of relative sequential constraint across second-symbol positions (Figure 3 : 3 . 1 The authors also pay some attention to words and sentence lengths, b'oth in the printed material and the phonemic transc'rip-", |
| "cite_spans": [], |
| "ref_spans": [ |
| { |
| "start": 342, |
| "end": 359, |
| "text": "(Figure 3 : 3 . 1", |
| "ref_id": null |
| } |
| ], |
| "eq_spans": [], |
| "section": "", |
| "sec_num": null |
| }, |
| { |
| "text": "EQUATION", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [ |
| { |
| "start": 0, |
| "end": 8, |
| "text": "EQUATION", |
| "ref_id": "EQREF", |
| "raw_str": "tions ,", |
| "eq_num": "1 1" |
| } |
| ], |
| "section": "", |
| "sec_num": null |
| }, |
| { |
| "text": "Their \"phonemic word\" is defined only in terms of a prosodical feature, specifically a pause in the f1.o~ of sound:\" They find phonemic words to be three times as long, on the average, fl as lexical words. They suggest, \"Insofar as, the units of spoken language and written language are different, the learning of written language (reading) will be difficult,\" but do not explore the implications of this suggestion further.", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "", |
| "sec_num": null |
| }, |
| { |
| "text": "Final evaluation. The authors at several places state that the results presented should be used \"with great caution. \" I would say that this caveat must be taken to apply to the whole work. Some linguists, and psychologists, and educators may find ", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "", |
| "sec_num": null |
| }, |
| { |
| "text": "t h a t i n t e r e s t i n g differences exist between \"informal\" and formal\" speech, bur the a u t h o r s ' a n a l y s i s has n o t rlevealed them.P p . x i v + 646 $ 2 5 . 0 0 .", |
| "cite_spans": [], |
| "ref_spans": [], |
| "eq_spans": [], |
| "section": "", |
| "sec_num": null |
| } |
| ], |
| "back_matter": [], |
| "bib_entries": {}, |
| "ref_entries": { |
| "FIGREF0": { |
| "uris": null, |
| "num": null, |
| "type_str": "figure", |
| "text": "By the authors' reckoning, the resulting carpus, from 58 peergroup sessions involving 174 different individuals, contained 84,164 lexical words, 313,694 alphabetic letters (in the conventionally spelled form) and 251,360 phonemes (in the phonemically transcribed form). If wotd and sentence marks (spaces and periods) are included, the figures are 405,906 alphabetic characters and 282,240 phonemes. The total corpus thus gathered occupies pages 57 through 439 of the book; the conventionally printed form and the phonemically transcribed form are on facing pages. No information is given as to whether the corpus is now available in machine-readable form, although it must have existed in that form at some stage of its analysis." |
| }, |
| "FIGREF1": { |
| "uris": null, |
| "num": null, |
| "type_str": "figure", |
| "text": "able o r f e a s i b l e ." |
| }, |
| "FIGREF2": { |
| "uris": null, |
| "num": null, |
| "type_str": "figure", |
| "text": "f i c i a l norms of correctness, grammaticality, or even communicative e f f i c i e n c y . I t i s probably f o r t h i s reason t h a t they t i t l e d t h e book Informal Speech. They were a l s o i n t e r e s t e d i n development P of speech generation, a t l e a s t over a c e r t a i n age range rom the f i r s t graae t o t h e junior college l e v e l . Note, h o w~v e r , tt895 they d i d nbt saaple the i n f o h a 1 ( r e l a t i v e l y unedited) speech of highly educated, mature speakers such a s might be found i n a mngressional cloakroom o r the salons of an ivy-league f a c u l t y club. I t i s q u i t e possible t h a t speech sampled i n such circumstances would conform fairly closely t o t h e norms o f e d i t e d , w r i t t e n language, a t l e a s t i n some respects (ana probably i n t h e r e s p e c t s studied by t h e authors) . I f one bears i n mind, t h e r e f o r e , the l i m i t a t i o n t h a t i t -i s not speech a s such (as opposed t o writ-ing) t h a t these authors have studied, but r a t h e r unedited speech of r e l a t i v e l y immature speakers, the work has considerable unique value by virtue of its presentation of extensive samples of such speech. The authors do not really analyze, however, the ways in which unedited language differs from edited language, nor the ways in which speakers develop strategies of editing their speech. Another objective of the authors was to use \"the rather powerful tools of information theory in the description of informal speech over the age range, in an effort to trace the role of redundancy in shaping language as a person uses it and presumably understands it in discourse. \" I will have more to say about the authors' use of information theory below, Data collection procedures, For their purposes, the procedures were excellent--certainly superior to procedures (interviews, con-t~ived play situations, classroom discourse) used by other investigators, for the procedures certainly elicited informal, unedited speech full of idterrupted sentences, hesitations, false starts, etc. The content covered a wide range of topics. Nevertheless, it was all conversation; the participants were merely exchanging ideas,, and declarative and interrogattive utterances abound. They were not directing each others' physical activity; thus, there appears to be a low incidence of imperatives, requests, etc. The corpus is certainly large enough for the kinds of analysis employed by the authors at a phonemic or graphemic level, but it might not be sufficiently large or representative for certain types of lexical or syntactical an,alysis . Transcription procedures, Transcription of the corpus was a fqrmidable and time-consuming project, not only in terms of a con-ventional printed form but also and particularly so, in terms of a phonemic verseon, One can o n l y s a y that t h e a u t h o r s made approximately the best of a very difficult job, They found it impossible always to identify speakers, and decided to omit any speaker ide~tifications, showing o n l y changes of s p e a x e r s , For the phonemic t x a n s c r i p r i o n , they used a modification of the Trager-Smith transcription suggested by Peter Ladefoged, but found it hard to get hired \"phonetirian~\" to adhere to the s y s t e m consistently, The system used was admittedly only p a r ' t i a l l y phonemic; for example, the glottal stops that were recolfded may or man n o t be phonemic. One wonders how consistently such distinctions as those between faydownow/, /aydannow/, and /aydownnow/ were observed, The treatment of pause phenomena was pa~ticularly bothersome. Pause phenomena were represented. in the phonemic transcriptions only by spaces and periods; thus, the phonemic transcriptians c o n t a i n a high proportion of very long \"phonemic words\" like / n a k t o w v~r~m r l k b a d~l~n n~y w a r g o w~~~n~& r . / , transcribed in graphemic form as \"knocked over a milk b o t t l e when they were going in there\" ( p p . 312-313), B u t i n t h e p r i n t e d version the location of the junclturos between these \"phonemic words\" is unfor-Tunatelp n o t shown, although it would have been fairly simple to have done so, perhaps by the use of slashes ( ) For certain purposes, it 4 s unfortunate, also, that certain kinds of material were omitted from the transcriptions, e . g . repetitions of words when in answer to wh-questions, and certain k i n d s of interruptions in continued sentences. It .is curious t h a t p r o p e r names were gene r a l l y deleted in the ptinted version b u t l e f t i n t h e phonemic trapscription ( e -g . compare pages 432-433); the aurhors apparently felbthat privacy would be preserved in the mystique of a phonemic branscription but not in conventional spelling. One can only guess as to hat s t r e s s and intbnational phenomena occurred in-the conversations. The printed version contains no question or exclamation marks, and the phonemic transcription contains no indication of intonations. Whether th@ tape-recorded material is archived somewhere, available for further analysis, is not stated." |
| }, |
| "FIGREF3": { |
| "uris": null, |
| "num": null, |
| "type_str": "figure", |
| "text": "of h a b i t p a t t e r n s , b u t beyond t h a t they would b e of little-interest either linguistically or psychologically. When Lee Hultzen requested use of my material for his analysis (Hultzen, Allen, and Miron, Tables of transitional frequencies of English phonemes, University of Illinois Press, 1 9 6 4 ) , I was only too happy to turn it over to h i m . Now,.what do we f i n d in the work of Oarterette and Jones?" |
| }, |
| "FIGREF4": { |
| "uris": null, |
| "num": null, |
| "type_str": "figure", |
| "text": "Ior the tzranscrril)ed speech samples, but the limitations of 91 these samples--particularly ir t h e way i n whtch they a r e presented--must be borne in mind. One can conceive uses f o r the statistical tables, perhaps by p s y c h o l o~i s t s seeking ways to control experimental stimulus material for phoneme f r e q u e n c i e s . I n genera l , howeyer, one wonders whether i t was worthwhile to pursue the statistical analyses and tabulations o f phoneme and grapheme f r equencies to t h e extremes reached by Carterette and J o n e s . I t i s l i t t l e wonder t h a t these authors seemed to abandon t h e i r interest in t h e i r r e s e a r c h a f t e r completing the manuscript represented in t h i s strangely unfinished book. But more importantly, the a u t h o r s have not p e~s u a d e d me that \"spoken language\" is d i f f e r e n t from \"writtew language\" in any interesting ways. I t is conceivable" |
| }, |
| "TABREF0": { |
| "html": null, |
| "content": "<table><tr><td/><td>81</td></tr><tr><td colspan=\"2\">selves (pp. 479-5135) no such symbols are to be found. Apparently</td></tr><tr><td colspan=\"2\">the original manuscript contained the symbols, but the printer</td></tr><tr><td colspan=\"2\">converted all of them to blank spaces and left-justified the</td></tr><tr><td colspan=\"2\">entries. Thus, rhere is no indication as to whether the most</td></tr><tr><td colspan=\"2\">frequent trigram, printed as \"th\", is to be taken as/TH or TH/</td></tr><tr><td colspan=\"2\">In this particular case, it is almost certainly to be taken as /TH</td></tr><tr><td colspan=\"2\">(from word-initial % I -in frequent determiners and some content</td></tr><tr><td colspan=\"2\">words), but what about a less frequent trigram such as what is</td></tr><tr><td colspan=\"2\">printed as \"en\": is this /EN or EN/ ? Fortunately a similar</td></tr><tr><td colspan=\"2\">error does not occur in Che tables of \"triphones\" (pp. 537-613)</td></tr><tr><td colspan=\"2\">where the printer indicated the space character as a carat ( A ) ,</td></tr><tr><td colspan=\"2\">although the authors intended use of the slash (1).</td></tr><tr><td colspan=\"2\">The gross omissions +re of certain summary statistical tables</td></tr><tr><td colspan=\"2\">that were, according to the authors' text discussion (pp. 43-45),</td></tr><tr><td colspan=\"2\">supposed to accompany the tables of triphones and trigrams. Tables</td></tr><tr><td colspan=\"2\">8.7.1-8.7.4 are not, as promised, preceded by tables giving fre-</td></tr><tr><td colspan=\"2\">quency distributions of trigrams; similarly, the tables in the 8.9</td></tr><tr><td colspan=\"2\">series are not accompanied by the promised frequency distributions.</td></tr><tr><td colspan=\"2\">I gather also that these omitted tables contained reports of cer-</td></tr><tr><td colspan=\"2\">tain information-theoretic statistics such as H or H ' . With much</td></tr><tr><td colspan=\"2\">effort, a user of this book could construct the frequency distri-</td></tr><tr><td colspan=\"2\">3.1 butions from the detailed tabulations, but it is still inexcusable</td></tr><tr><td colspan=\"2\">the points do not everywhere increase monotonically, although they for the printer to have omitted them.</td></tr><tr><td colspan=\"2\">must, in theory, and do, by the values aecorded in Table 7.6, p. 451). But the printer (or the publisher) is not to be blamed for</td></tr><tr><td colspan=\"2\">everything. There are also problems with the manuscript, and the</td></tr><tr><td colspan=\"2\">research that lay behind it. Questions must be raised about the</td></tr><tr><td>purposes of the work,</td><td>'-procedures, and the methods of analysis.</td></tr></table>", |
| "type_str": "table", |
| "num": null, |
| "text": "Purpose of the work. At the time the work was undertaken, probably in the early 19601s, it may have been correct to say, as the authors do, that all previous statistical analyses of language had been based on written or printed material. Nevertheless, one can think of exceptions: even the authors cite the study of eelephone speech byFrench, Carter and Koenig (Bell System Technical Journal, 1930, 9, 290-324), although this study has its limitations." |
| }, |
| "TABREF1": { |
| "html": null, |
| "content": "<table/>", |
| "type_str": "table", |
| "num": null, |
| "text": "order phenomena such as word distributions, distributions of syntactical patterns, etc.? Perhaps I can illustrate my attitude byrelating my own experiences with such analyses on the phonemic level. In 1951, in connection Qith an interdisciplinary seminar of psychologists and linguists, a group of the participants decided to investigate the information-theory characteristics or sequences p f phonemes in American Eng12sh speech. Not having readily available any authentic samples of speech, we decided to make a phonemic transcription of a series of one-act plays8 written for high school student performances. One of the linguists, ( F r e d Agard) transcribed some 20,000 phonemes from this corpus and subsequently I did a number of information-theory analyses of the data. The results were incorparated in a mimeographed report that, incideptly, was cited by Carterette and Jones in their reference list as, however, \"not seen. \" (A copy of the report c o u~d easily have been made available to them if they had asked me for it.)" |
| }, |
| "TABREF2": { |
| "html": null, |
| "content": "<table><tr><td>t is \"partly accounted For by-a decrease in zhe use of urn' as a</td></tr><tr><td>noise word, with the comcomitant [sic] rise in the use of 'you</td></tr><tr><td>know. ' \" (p. 23). Furthermore, the distribution of letters is</td></tr><tr><td>about w h~t many other investigators have found for samples of</td></tr><tr><td>printed English. Etaoin Shrdlu can still be the linotyper's</td></tr><tr><td>friend! Changes iri zero-ordbr distributions over different age</td></tr><tr><td>levels seem rhainly toreflect changes in the frequencies of cer-</td></tr></table>", |
| "type_str": "table", |
| "num": null, |
| "text": "decreases steadily from first grade to adult speech. \" This" |
| }, |
| "TABREF3": { |
| "html": null, |
| "content": "<table><tr><td>) tare virtually identical,</td></tr><tr><td>leveling off at about . 3 0 . I would interpret this to mean merely</td></tr><tr><td>that both first-graders and adults are using the same (Southern</td></tr><tr><td>California English) language, and that the same system o f phonemic</td></tr><tr><td>transcription has been used in the two cases. I would be much</td></tr><tr><td>surprised to learn that first-grade and adult speech samples could</td></tr><tr><td>not be differentiated in many ways--in lexical selection, in gram-</td></tr><tr><td>matical patterns, etc.</td></tr></table>", |
| "type_str": "table", |
| "num": null, |
| "text": "Sequential constraint of phonemes is probably not a sensitive way of indexing any thin^ useful or interesting about lmguage samples. It is at least misleading for the authors to state that \" [ i l n terms of simple sound pattern redundancies, therefore, 6-year-old speech is already adult\" (p. 30).The case i s slightly different when the redundancy statistics are applied to letters (graphemes) in the transcriptions of speech, or in printed materials. First g m d e speech is shown to be slightly more redundant khan adult speech; I w o~i l d think that a large part of the difference could be traced to differences in lexical distributions--differences that show up in letters but not in phonemes because lexical boundaries are observed in the letter statistics but are rarely preserved in the phonemic transcriptions.First grade readers are also much more redundant in letter distributions than even first-grade speech; b u t it is well known that first-grade readers are typically highly redundant in their lexical distributions. Redundancy statistics based on grapheme distributions, apparently reflect these lexical distributions, but hoy reliably, it is difficult to cell. There is the suggestion, arising from these results, that redundancy statistics based on grapheme distributions and their sequences could be a useful surrogate for other types o f indexes based on incidence and sequences of words, grammatical patterns, etc. But the authors' suggestion that differences thus revealed between'natural speech and written material are somehow important to take into account in the teaching of .reading seems rather forced and gra-" |
| } |
| } |
| } |
| } |
