Cognition, Vol. 4, No. 1

Editorial

Hiroshi Faculty Tokyo Hongo, Tokyo,

Azuma of Education, University, Bunkyo-ku, Japan

Richard Cromer MRC Develovmental Psychoiogi Unit, Drayton House, Gordon Street, London, WCIH OAN, Gt. Btitain

Paul Bertelson Laboratoire de Psychologie Exphimentale, UniversitP Libre de Bruxeiies I1 7, Au. Adolphe Buyl, B-IO.50 Bruwelles, Belgique

Peter D. Eimas Waiter S. Hunter Laboratory of Psychology, Brown University, Providence, R.I. 02912, U.S.A.

Ned Block Dept. of Philosophy, M.I. T., Cambridge, Mass. 02139, U.S.A.

Gunnar Fant Lab. of Speech Transmission, Royal Institute of Technology, S-10044 Stockholm 70, Sweden

T. G. R. Bower Dept. of Psychology, University of Edinburgh, 60, Pleasance, Edinburgh EH8 9 TJ, Gt. Bn’tain Franqois Bresson Laboratoire de Psychologie 54, bud. Raspail, F- 75006 Paris, France Roger Brown Dept. of Psychology, Harvard University, Cambridge, Mass. 02138,

U.S.A.

Jerry Fodor Dept. of Psychology, M.I.T. EIO-34 Cambridge, Mass. 02139, Kenneth Dept. of Monash Clayton,

U.S.A.

Forster Psychology, University, Vie. 3168, Australia

Merrill Garrett Department of Psychology, M.I.T. EIO-034, Cambridge, Mass. 02139, U.S.A.

board

David T. Hakes Department of Psychology, University of Texas, Austin, Tex. 78712, U.S.A. Henry Hecaen Directeur d’Etudes. Ecole Pratique des Hautes Etudes, Unit6 de Recherches Neuropsychologiques. I.N.S.E.R.M., 2, rue d ‘Al.&ia, F-75014 Paris, France Michel Imbert Laboratoire de Neuropl~~~siologie. College de France, I I, place Marcelin Berthelot, F- 75005 Paris, France Barbel Inhelder Fact&P de Psychologie et des Sciences de I’Education, Universitt de GenPve, CH-1211 GenPve 14, Suisse Marc Jeannerod Laboratoire de Neuropsychologie ExpCrimentale, Dojien L&pine, F-69500 Bron, France James Jenkins Center for Research and Human Learning, University of Minnesota, Minneapolis, Minn. 55455,

Jerome S. Bruner Dept. of Experimental Psychology, University of Oxford, South Parks Road, Oxford, OXI 3UD, Gt. Britain

Lila Gleitman Graduate SchoolofEducation, University of Pennsylvania, 3700 Walnut Street. Philadelphia, Pa. I9 104. U.S.A.

Peter W. Carey 2730 Dwight Way, Berkeley, Calil: 94704,

Pierre Greco Laboratoire de Psychologie, 54, bud. Raspail, F- 75006 Paris, France

Philip Johnson-Laird Laboratory of Experimental Psychology, Centre for Research on Perception and Cognition, Sussex University, Brighton BNI 9QC, Gt. Britain

Jean-Blaise Grize I, Chantemerle, Neuchdtel, Suisse

Daniel Kahneman Dept. of Psychology, The Hebrew Universitv of Jerusalem, Jerusalem,-Is&e1

U.S.A

Noam Chomsky Dept. Modern Languages and Linguistics, M. I. T., Cambridge, Mass. 02139. U.S.A.

U.S.A

Jerrold J. Katz Dept. of Philosophy, M. I. T., Cambridge, Mass. 02139, U.S.A. Edward Klima Dept. of Linguistics, La Jolla, University of California, San Diego, Calif 92037, U.S.A. Alexei Leontiev Faculty of Psychology, University of Moscow, 13, Frunze Street, Moscow G.19, U.S.S.R. Wilhelm Levelt Psychological Laboratory, Numegen University, Erasmuslaan 16, Nijmegen, Netherlands A. R. Luria Faculty of Psychology, University of Moscow, 13, Frunze Street, Moscow G.19, U.S.S.R. John Lyons Dept. of Linguistics, Adam Ferguson Building, Edinburgh EH8 9LL, Gt. Britain Humberto Maturana Escuela de Medicina, Universidad de Chile, A. Sanartu 1042, Santiago, Chile David McNeil1 Department of Behavioral Sciences, Committee on Cognition and Communication, University of Chicago, 5848 South University Avenue, Chicago, IN. 60637, U.S.A. John Morton Applied Psychology Unit, 15, Chaucer Road, Cambridge CB2 2EF. Gt. Britain George Noizet Laboratoire de Psychologie Experimentale, F-I 3 Aix en Provence, France

Daniel Osherson Department of Psychology, University of Pennsylvania, 3813-15 Walnut Street, Philadelphia, Pa. 19174, U.S.A.

Robert Shaw Center for Research and Human Learning, University of Minnesota, Minneapolis, Minn. 55455,

Domenico Parisi Istituto di Psicologia, Consiglio Nazionale delle Richerche, Piazzale delle scienze 7, Rome, Italy

Dan I. Slobin Department of Psychology, University of California, Berkeley, Calif 94720, U.S.A.

Michael Posner Dept. of Psychology, University of Oregon, Eugene, Ore. 9 7403, U.S.A. Zenon Pylyshyn Departnient of Psychology, The University of Western Ontario, London 72, Ont., Canada Martin Richards Unit for Research on Medical Applications of Psychology, University of Cambridge, 5, Salisbury Villas, Station Road, Cambridge CBI 2JQ? Gt. Bn’tain Steven Rose Biology Department, ‘The Open University, Walton Hall, Milton Keynes MK7 6AA, Gt. Britain Nicolas Ruwet Dept. de Linguistique, Centre Univ. de Vincennes, 12, Rue de Tourelle, F- 75012 Paris. France

U.S.A.

Jan Smedslund Institute of Psychology, Universitet i Oslo, Box 1094, Oslo 3, Norway Sidney Strauss Department of Educational Sciences, Tel Aviv University, Ramat Aviv, Israel Alina Szeminska Olesiska 513, Warsaw, Poland Yoshihisi Tanaka Dept. of Psychology, University of Tokyo, Bunkyo-ku, Tokyo 113, Japan Hans-Lukas Teuber Dept. of Psychology, M.I.T. EIO-034, Cambridge, Mass, 02139,

U.S.A.

Virginia Valian Ph.D. Program in Psychology, C.U.N. Y. Graduate Center, 33 West 42nd Street, New York, N. Y. 10036, U.S.A.

Harris B. Savin Dept. of Psychology, University of Pennsylvania, Philadelphia, Pa. 19104, U.S.A.

Peter Wason Psycholinguistics University College London, Research Unit, 4, Stephenson Way, London NW1 2HE, Gt. Britain

Tim Shallice Psychology Department, The National Hospital for Nervous Diseases, Queen Square, London, WCI, Gt. Britain

Hermina Sinclair de Zwart Centre d’EpistCmoZogie Gtnetique, Universite de GenPve, CH-121 I Geneve, Suisse.

Editorial For the past four years we have been editing this journal both to print the most interesting advances in the field, and to discuss the epistemological, social and political stresses that are the scientists’s lot today. In the course of this endeavor, we have run into many situations which have obliged us to reflect on issues explicitly that we ordinarily might have ignored. For example, we have had to decide whether or not to publish manuscripts that originate in places where the political chaos would hardly appear to allow anyone to concentrate on affairs as abstract as those which generally interest us; we have had to reject and accept papers on the basis of criteria, which though stringently adhered to in our reviewing procedures, are not very clear in substance. Finally, we have had to develop a position as to what Cognition as a field is and what it should be. In doing such things as these, we have accumulated so many unsolved questions that we believe it best to come straight out and talk candidly about our perception of the problems that hound us as editors and scientists. It is obvious that we work within the framework of a cultural setting that obliges us to define what we think is valid and what is not. In our case, for example, we have tried to present papers that are theoretically relevant rather than papers that emanate from interesting observations but which are not presented in a wider frame of reference. Consequently, we have had to juggle such notions as interesting, relevant, theory and frame of reference: Our practical problem is how best to evaluate such notions. Of course, we use the usual methods for doing this, such as having referees who seem to us to be the most appropriate. We adhere strictly to the procedure of soliciting two reviews for each manuscript - two positive reviews are necessary for acceptance of a manuscript regardless of who the author may be or the nature of the subject matter. We also send a paper out to a third reviewer whenever the first two readers do not agree. All in all, we apply the standard methods which should allow us to proceed with some semblance of good conscience. Yet we remain uneasy. The problem that we are raising concerns the implicit nature of the procedures that have been with us all for a great many years. Psychology, like other human sciences, has been influenced greatly by the cultural background in which it has developed. For instance, it is now clear that during the years when positivism and behaviorism dominated the field, other approaches were by and large ignored in publication. Clearly, explicit censorship was rarely used, nor was it needed. Rather, rigorous format monetary correlation accompanies this phenomenon. What is “good”

8

Editorial

standards were established by editors in such a way as to favor manuscripts that originated in the “acceptable” scientific schools. A cultural provincialism developed rendering manuscripts that followed the accepted standards apparently more professional, better written and more comprehensible. As a result, the better known journals all began to look very much alike. Even after many scientists had given up the behaviorist dogma, they continued to mold their papers in the standard way, often masking their philosophical maturation behind a facade of useless procedures and empty statistics. We embarked on our venture, Cognition, largely in reaction to this state of affairs at a time when the pattern was already beginning to change. In fact, the rigidity of the standards and the monotony of the many atheoretical studies led to a reaction which began in the sixties and has been emerging ever since. Cognition has succeeded in part thanks to this reaction. Our problem now is how to avoid being imprisoned by our success. We now go through the same motions as did harassed editors before us. We have our own hobbyhorses and reviewing procedures to protect us. Our dilemma is this: We directly rejected the monotonous format of the standard journals in favor of an open style that would allow us to focus on content. We have discovered in the process that much of what traditionally passed for content was, in fact, form. With so many cognitive “theories” and so few direct empirical implications of any theory, how do our reviewers and we decide what to publish? We must grapple openly with the fact that the atmosphere surrounding the judgment of the worth of an article has become so rarified that we had better go back and take another look at the whole issue of value judgments. Accordingly, we think it may be clarifying to consider the process of criticism in art. This is an area in which at least initially, utilitarian considerations were not at stake as they are in one way or another in most branches of science. If we manage to gain a little understanding of the situation currently existing in art, it is possible that we may shed a little light on what is occurring closer to home. It is highly probable, that if questioned concerning which single painting in the world should be saved from the hecatomb, anyone having finished a secondary education would opt for the Gioconda. It is also equally likely that no consensus would result if the same population was asked which painting in the world they would most like to have hanging on their own wall providing, of course, they had the money to afford it. This observation, though admittedly rather superficial, would appear to demonstrate the fact that society obliges us to have a set of official standards and value judgments often far removed from our personal ones. Naturally, a certain

Editorial

is unquestionably more expensive than what is not. However, it should be noted that there is generally very little a priori justification for our implicit value judgments. Our evaluations lead to no far-reaching theories on aesthetics and human nature. On the contrary, what generally occurs is the irrevocable division of society into those who produce art and those who consume it, with little interaction between the two. The producers quickly draw others into their group and the public soon follows, either to be in the vanguard of a new cultural trend, or for more venal considerations such as potential investment value. Thus, self defined market value determines a good part of the motivation for preserving certain works of art and not others. Our own feeling is that there may well be an independent notion of “good” and “bad” art, based on a covert, aesthetic theory. Nonetheless, this does not affect our argument in one way or another. Most of the time our own feelings need not be closely related to the value judgments reigning in our community. Given the expertise and generally elitist properties required in order to appreciate the more dominant forms of art, there is very little relation between what we like personally, and the value that we as members of society attribute to works of art. Obviously, as academics, it would be hard for us to manifest the same disapproval or contempt for the dominant forms of art that we bestow on certain forms of popular art regardless of what our subjective judgment may be. In a wholly unalienated society it is possible that some form of universal aesthetic theory might be helpful in understanding the dynamics of this situation. However, in a system as plurally determined as ours, we doubt that it would do much good. What should be retained, however, is that aside from putative universal theory, a cultural pattern exists which controls us by fabricating a hierarchy of “good” and “bad” artists whom we select or reject in economic terms, thereby participating openly in the meritocratic system. That a meritocracy serves certain purposes cannot be denied. It automatically limits active participation in any one field to a chosen few while relegating the majority to the role of a passive audience. However, in so doing it also often defeats the fundamental purpose of art, which is to give pleasure by expanding and gratifying our aesthetic senses. This state of affairs is particularly effective in societies where there is a need to use the production and consumption of art as a symbol for the distinctions that are the form and content of elitist political and economic structure. But our point is not to enter into a discussion of social injustice so we will not pursue this point any further. What we would like to stress is that artistic production and criticism have from the start been a form of value judgment aimed at drawing a clear distinction between those in control and those controlled.

10

Editorial

With the passage of time, of course, the concept of what art is has undergone severe changes. It is possible, for example, to claim that the Beatles made an artistic contribution every bit as real as that of any composer working in the modern musical tradition. Schools and trends and their accompanying sets of value judgments are breaking up in all directions. Slicing one’s ear with a razor blade is called “body art”, while leaving a canvas blank is described as “minimal” art. This change in criteria1 attributes can be traced to the fact that society, with the dominance of technology in the last century, has shifted from emphasizing one symbol of social structure (art), to stressing another (science). Art has been freed of its social function and has in consequence been freed of the roots that constrained it. The artist today has few of the characteristics of his or her predecessors. Leonardo, or even, for that matter, the “pointillistes”, were exploring domains which were to be collectively shared. Today, artists generally restrict themselves to the expression of idiosyncratic regularities of form and structure: It is the person of knowledge who has the power of representing our aims and delusions. We are the cultural heros and it is our production that is subjected to societal ratings of “good” and “bad”. In that sense, then, it is far from surprising that the current unconscious race for primacy of value-systems is being fought out in the sciences rather than in the arts. In the light of this shift, it is necessary to explore the way in which we, as scientists, proceed in our daily endeavors, while at the same time acting as standard bearers for the characteristics that symbolize and ensure the hierarchical structure of our society. Science deals with the way in which we construct models of the world. This, in turn, allows us to predict and to control our environment. In this sense, therefore, it should be relatively easy to distinguish between those who contribute productively to such knowledge and those who do not. Although this was an acknowledged fact, the institutionalization of science has resulted in the proliferation of competitive domains, each of which has produced a long list of articles which will fall into oblivion without leaving any imprint on the environment or on the edifice of science. It is increasingly apparent that the selection of the domains to be studied can only partially be attributed to those who conduct the studies. Many intermediary organizations and social structures exert at least as much influence in determining research as any group or individual scientist. Finally, journals and other means of disseminating information have a great influence in shaping the direction of research. These considerations lead us full circle back to our concern about current behavioral science and our role in it as teachers, researchers, authors and editors. The main difficulty that we see is that the true intended product of

Editorial

11

behavioral science, knowledge about humanity, has become clearly alienated from the means which are supposed to produce this product. That is, scientific publications are presumed to lead to increased knowledge; but they serve so many other purposes that are more directly manifest, that their production and consumption has become an independent value. This is reflected in a variety of facts some of which are well-known, but some of which have surprised us as we have discovered them. It is obvious that this editorial has raised issues related mostly to judgments of value. If this is so, it is because it is easy to discern a certain unrest in the scientific community over the dichotomy which prevails between the ways in which we think and feel and the ways in which we actually proceed in everyday life. It would be possible to demonstrate that our comments apply not only to publications and editors but also to the conditions of our employment, promotion or income, the treatment meted out to us by funding agencies, our selection to committees, the pressures of specialization, etc., etc. In fact, although it is rarely discussed as such, the whole issue of value judgments underlies much of our everyday life. It is never apparent how the scientist’s desire to produce new knowledge and the more down to earth considerations such as advancement, power needs, the desire to secure the highest possible evaluation and the most praise, interact to determine our behavior. This being said, we have not raised the questions we have, in order then to be able to enumerate one hundred and eleven rules and regulations for solving the problems of the scientific community, nor to pass judgment on motivations or lack of them. It is our contention that no individual or single organism can do this without playing into the hands of the existing system. Alternatively we feel that it is only by raising a general awareness to the ways in which our feelings and desires diverge from our practices and how the latter are shaped and molded, that a collective form of change that does not contain the vices of the current system may be defined. It is only natural then, that we should invite you to join in a discussion and debate of all or some of these issues in the coming numbers of this journal. J. MEHLER T. BEVER S. FRANCK

Cognition, 4 (1976) 13-30 0 Elsevier Sequoia S.A., Lausanne

- Printed

1

in the Netherlands

Some attributes of adjectives used by young children*

KATHERINE Yale

NELSON

University

Abstract Analysis of 24 spontaneous speech samples from children at 24 and 30 months revealed a correlated progression in the form, function and meaning of modifiers used with increased language development. Predicate adjectives were used to comment on transitory states of objects and prcdominated in the early speech samples. Attributive adjectives were used to subdivide classes or to specify among particular instances within classes on the basis of physical characteristics such as size, by use of a conventional type name, or by) use of a possessive. Attributives predominated in the later speech samples. Differential use of a particular form appears to be a function of its cognitive and communicational usefulness rather than itssyntactic derivation. Adjectives in adult language serve a variety of functions: They describe something about an aspect of a situation, specify which of a number of alternatives is being referred to, or sub-classify within a larger class of things. For example, (1) Elizabeth

was a beautiful

child.

(2) It was the tall blond man who pulled the gun. (3) Dan bought an electric typewriter. We can see that in the first case the adjective describes a particular referent (Elizabeth), while in the second case it specifies which of a number of possible particular referents (several different men) is meant. In the third case, however, the adjective does not modify a particular *This research was supported by a grant from the Carnegie Corp’oration of New York. Data collection was made possible by a NICHD postdoctoral fellowship to the author. Requests for reprints should be sent to the author, Department of Psychology, Yale University, New Haven, Connecticut 06520.

14

Katherine Nelson

referent but a referent class (typewriter) which can be subdivided into the classes of manual and electric typewriters (Bolinger, 1967). Do young children learn and use adjectives in this way? Although children acquire a number of adjective types very early - some appear among their earliest words (e.g., ‘hot’,‘broken’) - there is little research on how they are used, with the exception of a small set of spatial adjectives (e.g., Donaldson and Wales, 1970; Sinclair-de-Zwart, 1969). The use of modifiers (adverbs and possessives as well as adjectives) by young children is of interest for several reasons, however, in addition to the fact that they constitute an early form of predication. (According to Brown’s [ 19731 summary of Stage I speech, when mean length of utterance - MLU - equals approximately 1.75 morphemes, attributives accounted for 6% of the multi-word utterances and possessives for 870.) First, they serve an important conceptual service. They enable the child to make distinctions among referent objects and classes of objects on the basis of both general and specific properties. That is, they provide him with a linguistic means to generate new reference classes. They are thus basic to lexical productivity. In addition, the adjectives used by the child specify the properties of objects, people and other referent entities, properties which are important to studies of concept formation, attention, memory and logical thought. The question of which properties are coded in the language of the young child is therefore a potentially revealing one. The research described here is an attempt to begin the investigation of this problem by examining the use of modifiers in the spontaneous speech of beginning speakers. In this analysis, the syntactic roles, meanings, and functional relations to modified nouns of early modifiers will be considered. Both adjectives and possessives are included here as modifiers. While possessives relate people to things possessed, and adjectives refer to states and properties of things, they share similar semanticsyntactic relations to nouns, and both are used by young children to make distinctions among objects. Adverbs (except for locatives) were rare in the present samples and were not included in the analysis. Special problems arise when spontaneous speech samples are analyzed for the purpose of revealing functional use that can be overlooked when the emphasis is only upon linguistic analysis. When children are learning language speech may contain a good deal of ‘form practice’ and elicited or imitated utterances in addition to functional statements. These learning forms may be revealing of linguistic structure (see Slobin, 1968; Weir, 1962, for example) but are misleading for a functional analysis in that whether they are representative of the child’s own knowledge or interest

Some attributes of adjectives used by young children

15

cannot be known.* Learning episodes have their own conceptual importance, but they are different in kind from the functional use of speech; they cannot be considered true speech acts. A second problem in attempting to determine what the child knows from what he says is that negative evidence cannot be evaluated. If a child does not say something we cannot conclude that he could not. This is especially true for relatively brief samples of speech. At the extreme, it is exemplified in those children who understand what is said to them but do not themselves produce language, whether for physical or psychological reasons. The antithesis of the point is that we cannot conclude that the child means what he says, even when he says it. That is, the child’s meaning may differ from the adult’s even when the same forms are used. At the extreme this is exemplified by the use of ‘empty’ or rote forms that have no reference for the child. ‘Why’ is used by many children in the pre-school period before they acquire an adequate understanding of causation, for example (Blank, 1974). Much recent analysis of developmental semantics (e.g., Clark, 1973) has been devoted to the exploration of the ways in which the child’s meaning for certain terms differs systematically from the adult’s. Still, except for empty forms, it can be argued that most terms used by the child share some of the central meaning structure of the adult terms; if they did not, they would not be acquired, understood or survive in the child’s speech (cf., Fodor, 1972 on this point). Therefore, if detailed claims are not made about the meaning structure of particular terms, the analyst is on safe grounds in assuming that terms and forms that are used frequently and by many child speakers reflect, at least in part, the knowledge that we would normally attribute to an adult speaker who used the same term in the same context. These preliminaries indicate the problems encountered in this type of analysis and serve as a cautionary guide to the interpretations presented here. Analysis of spontaneous speech is necessary to identify phenomena of interest in language use, but it will need to be followed by more systematic verification methods.

*This learning factor probably accounts for some of the discrepancy in the frequency use of object labels (see Bloom, 1973; Nelson, 1973). Children, when first learning (or any other) term tend to repeat it endlessly, but once it is practiced thoroughly, it out of use, if it is not of continuing importance to the child, while remaining in the vocabulary and available as a label when needed. Much the same thing happens later color terms are acquired, for example.

of early an object may drop receptive on, when

16

Katherine Nelson

Sampling Method The data for this analysis were derived from speech samples of one or two hours in length taken from 18 children at 24 months and from 16 of the same children at 30 months (the data are described in more detail in Nelson, 1973). For the present analysis, the first 100 utterances from each of 24 of the resulting transcripts were used. Some of the children whose speech will be analyzed here were earlier identified as primarily Expressive (E) rather than Referential (R) speakers, on the basis of the words that they learned and used at the one-word level (see Nelson, 1973 for details). R speakers learned predominately object-words. seeming to indicate an interest in things and their categorization, while E speakers used more social-personal forms. A question of interest in the present analysis was whether the differences in speech patterns between E and R children would persist at more advanced levels. To the extent that adjective use is related to categorization of objects it might be expected that R children, whose earliest language was more object-oriented and thus appeared to be related more directly to the child’s categorization of the world, would tend to use more adjectives than E children, or to use them in a more advanced way. The average age at which the identification of R versus E speech types was made for each child (based on the percent of object words in the first 50 words learned) was 19 months, 5 months prior to the 24-month speech samples and 11 months prior to the 30month samples used here. Transcripts were selected on the basis of the mean length of utterance (MLU), age, sex and vocabulary group (R or E) of the child. MLU was derived from the entire speech sample for each child and was based on Slobin’s (1967) rules. Seven MLU levels with a range of 0.5 morphemes were defined from 1.0 to 4.5. For most analyses, the levels were collapsed to two groups: Low proficiency (MLU = 1.0 - 2.5) and Hi& proficiency (MLU = 2.5 - 4.5). These cover roughly the. same levels as from less than Stage I to Stage II for the low group, and Stage III to V for the high group in terms of Brown’s (1973) Stages. The low group here may then be considered roughly equivalent to Stage I speech, for which there is a good deal of comparable data, although the range in this study is broader. Six transcripts from each MLU group (High or Low) and each vocabulary type (R or E) were analyzed. To the extent possible, sex of the child was balanced within each sub-group. Transcripts were analyzed from 10 boys’ and 14 girls’ records. Some but not all of the 18 children were represented at both ages, while a few of the total group of children were not represented at all. Age could not be separated from proficiency in this

Some attributes ofadjectives used bJ1young children

1I

sample; all of the High-E group were 30 months of age and all of the LowE group were 24 months, while only two of the younger R group were in the High level and two of the older were selected at the Low level. Thus, we cannot speculate to what extent age (and experience) rather than language or conceptual structure may account for the differences between High and Low proficiency speakers. The distribution across MLU levels was kept as balanced and evenly distributed as was possible within the total sample. Each of the seven levels was represented at least once for each vocabulary group and only one (R 2.0 - 2.5) was represented more than twice (3 times). The 2400 utterances selected in this manner (both single and multimorpheme) were analyzed in terms of noun-verb relations, presence of modifiers, pronouns, quantifiers, distinguishers and locatives. Aspects of the data other than modifier use will be reported eslewhere (Nelson, 1975). Of the total sample of utterances, 13.8% contained a nominal modifier; that is, a modifier was contained in every 7.21 utterances, on the average, for a total of 333 cases. As noted above, a modifier for this purpose might be a possessive noun or pronoun as well as an adjective. Terms that were used only in response to the mother’s elicitation (e.g., “what color is that?“) were not included in the final analysis.

Analysis and Discussion Several different analyses were undertaken; description of the method, results and discussion will be presented for each in turn. At the most general level, Table 1 presents the relation between language status and use of modifiers, in terms of the correlations of MLU (7 levels) with various’indices. It is surprising to note that there are few strong relations of modifier use with speech maturity apparent in this table. Although total number of adjectives and possessives show a positive relationship, they do not reach statistical significance. The number of different adjective and possessive types are, however, significantly related to MLU level; that is, with increasing maturity a greater variety of forms is used. Neither the type-token ratios nor the ratio of modifiers to nouns or to each other show any overall relation to MLU status, howev.er. This indicates that these classes occur in the same proportions with development - each increasing in use to the same degree. Differences within the class of modifiers were apparent, however, as shown below.

18

Katherine Nelson

Table 1

Correlation

of lnodificr

Total

number

of adtectivcs

Total

number

of possessives used

0.219

Total

number

of modifiers

0.35

of adjective

Number

of possessive types type-token

types

Adjcctive~-noun Modifiers-noun

correlations.

ad

0.245 --0.006

ratio

levels of MLU

Forms

ratio

0.1 09

ratio

*Seven

1

0.4 1o** --0.018

ratio

moment

0.335

0.436**

ratio

Possessive type-token Possessive-noun

used used

Number Adjective

Adjective

use variables with MLU level*

0. I29 from **p

I .Oto 4.5

were used. Correlations

are Parson

product

< 0.05.

I*‘urlctiorls

A basic distinction among adjective constructions is that between predicate adjectives and attributive or prenominal forms. Each of the examples above (1 - 3) involve attributive adjectives, that is, those appearing in prenominal position. The predicate forms (e.g., ‘Elizabeth is beautiful’) appear usually to be descriptive of an entity or comments on a topic. Such forms can be found even as single word utterances, for example, ‘broken’. In this case, some object which is understood from the prior linguistic or non-linguistic context is the topic and the comment refers to a property or state of the thing. Attributive forms precede the noun, as in ‘the big truck’, regardless of what role the noun phrase plays in the total utterance. It is obvious therefore that single-word modifiers can only be a reduced form of predicates and cannot be attributives, because these require the presence of a modified noun. Attributives have generally been analyzed as derived from the predicate adjective construction, thus involving an embedded sentence of the form ‘the truck is big ’ in the deep structure of ‘the big truck’. Several authors (for example, Bolinger, 1967 and Vendler, 1968) have shown, however, that for many adjectives this derivation will not work. Bolinger claims that the attributive, not the predicate, is the basic form, while Vendler has analyzed a number of different types of adjectives that are derived from different underlying predications. Brown (1973) has noted that attributives appear earlier than predicate adjectives in children’s sentences and therefore questions the developmental reality of the view that the former are derived from the latter. Although the question of the

Some attributes of adjectives used by young children

19

correct grammatical derivation is not of direct concern here, the related distinctions that have, been made in regard to different types of adjectives, as revealed in these linguistic analyses, are of interest. In the present analysis, all adjectives appearing in the pre-nominal position were classed as attributives, while all others, whether used alone (e.g., ‘broken’) or in a predicate phrase (e.g., ‘it’s broken’) were classed as predicates. This distinction relies on word order in child utterances in the same way that it does for adult utterances. Of the predicate adjectives, less than half (44%) were used in single word constructions. In this sample, the correlations of MLU with total number of attributive adjectives was positive and significant (Y = 0.54, p < 0.01) while the number of predicate adjectives was negatively related to MLU (Y = -0.375, n.s.). Further, the correlation of MLU with the proportion of all adjectives that occur in the attributive position was highly significant (Y = 0.667, p < 0.001). Table 2 s1rows that attributive modifiers occur more frequently overall, and this finding alone would tend to support Brown’s (1973) statement of the sequence of emergence of these forms. However, as indicated by the high correlation of MLU with proportion of attributives, attributives increase in relative frequency with development, regardless of language type, while predicates decrease. The two types are of approximately equal frequency in the Low samples, but attributives are about three times as frequent in the more mature samples. These facts tend to contradict the claim of attributive priority, although they might be explicable in terms of the prevalence of one-word utterances in the Low samples, which by definition are predicate types. However, while more of the predicate adjectives used by Low speakers were one-word utterances (62%) than those of the High speakers (21%), there was a proportional decline in use of predicate adjectives with language status whether all adjectives (including single words) were considered (from 62% to 27%) or only multi-word utterances were included (from 38% to 23%). While it appears from Table 2 that R speakers used more adjectives at the High levels than did E speakers, this comparison was not significant (Language Type main effect and Interaction with MLU were both greater than p = 0.15 in an Analysis of Variance on this variable). The MLU main effect was only marginally signficant (F = 3.649, p = 0.07). These are reflections of the fact that the inter-child variability in number of adjectives used was very great. Proportional analyses of adjective types were therefore deemed appropriate. It may also be noted in Table 2 that both the absolute numbers and the proportions of the two modifier types are very similar for the two types of language learners (Referential and Expressive). An analysis of variance

20

Katherine Nelson

Table 2.

Number and percent of attributive and predicate adjectives by language group. Language

Mean number of Mean number of Mean number of Percent predicate Percent attributive

adjectives predicates attributives adjectives adjectives

group

R-Lo

R-Hi

E-Lo

E-Hi

Total

6.17 4.00 2.17 66.7 33.3

13.33 2.83 10.33 23.3 75.3

6.33 3.17 3.17 56.6 43.4

7.50 1.50 6.00 13.7 86.3

8.33 2.88 5.42 40.1 59.6

on the proportion of attributive adjectives showed a significant main effect for MLU status (F,,20 = 13.299, p < 0.002) but no effect for language type or interaction of language type and MLU status (F < 1.OO). Clearly, while the number of predicate modifiers remains fairly stable with language development, the number of attributives increases. sharply. As noted above, predicate adjectives can be considered comments on a topic, whereas attributives may be considered to specify the referent or the referent class more completely. The predicate adjectives are apparently always descriptive of the referent in the traditional sense in which the adjective describes a person, place or thing. Attributives, even when used descriptively [see (1) above], perform a role subordinate to the modified noun. Bolinger (1967) has noted that attributives may specify something about the referent [see (2) above] or they may specify something about the reference class (3), in which case they serve as subclassifiers or types. He notes that, while the former can be derived from predicate adjectives, the latter cannot be. If an attributive modifies a particular referent, it can be easily related to the predicate adjective both in terms of its construction and in terms of its function. One can imagine a child saying, for example, ‘That is a bear. That bear is big. That is a big bear.’ Clearly, ‘big’ here refers to the referent, this particular bear, which is big for bears, or perhaps, big for pictures, or big for toys, if these constitute the immediate perceptual array. Consider, however, what looks like an equivalent sentence: ‘That’s a teddy bear’. Here the adjective serves to subclassify the referent into a new reference class. The adjective modifies the reference class, not the particular referent. One cannot say ‘That bear is (a) teddy’, as Bolinger points out. The general class term must be added: ‘That bear is a teddy bear’. Here the referent is not being described or specified, but identified by class or classified.

Some attributes of adjectives used by young children

Table 3.

21

Percent of referent and reference class modification by language group

Percent attributive referent Percent attributive class Total percent referent (Attributive and predicate)

R-Lo

R-Hi

E-Lo

E-Hi

5.0 28.3 71.7

21.8 48.5 51.1

13.0 30.4 69.6

10.1 16.2 23.8

In the adult language this distinction between referent and reference class modification can be made successfully according to type of adjective. Here, however, one of the issues of interest is what types of adjectives are used by young children for these different purposes. It is not always easy to determine which of these relations is being employed by the beginning talker. The young child may use size, for example, to subdivide a larger class of, say, dogs. Thus ‘big dog’ may be used as a classification as well as a comment on the size of a particular referent. Some uses of this kind are standard, for example, ‘little girl’. For this analysis, the context of the utterance, including emphasis and the preceding and following conversation, was used to determine whether the child appeared to be emphasizing the referent (e.g., ‘Where’s the big truck?‘) or utilizing the adjective to specify a sub-category (e.g., ‘Give me a big stick’). As this example shows, the preceding definite and indefinite articles are also indications of referent and class use, respectively. In addition, apparent intention and paraphrase were used to disambiguate the category. All classifications were made prior to further analysis. The results of this classification are shown in Table 3. An analysis of variance on the proportion of adjectives used as classifiers showed a significant main effect of MLU level (F,,20 = 6.92, p = 0.02) but neither the Language Type nor the interaction effects were significant. The distribution between the two types of attribution (reference classification and referent description) by language status clarifies the difference between the more advanced and less advanced speakers that was apparent in their differential use of predication and attribution as shown in Table 2. The difference in relative adjective functions as language develops is shown in summary form in Figure 1. Adjective

types

The referent-reference the types of words

class distinction used in different

just discussed constructions.

is apparent also in Consider first the

22

Katherine Nelson

Figure 1.

Comparative use of different 0

i

0’

Pred. Ref. Att

n

Class

of adjectives

at two MLU levels

Att

70 -

a

60

-

;

50

-

k k =

40

-

30

-

20

L

z w a

m

types

LOW

MLU

HIGH

MLU

predicatives. Each adjective was categorized into the following types (all words included in the category are listed in parentilcses): Descriptive properties of the object (big, bigger, white, colored); Evaluative terms (smarter, nice, no good); Transient object states (cold, hot, missing, dirty, all gone, whole, broke(n), tix(ed), open, on (sound), off (sound), stuck); and Animate states (hurt, dead, stinks, finished, sleepy). When crossclassified by MLU status the distribution according to these categories is shown in Table 4 with States collapsed into one category. (Because of the small numbers involved in some categories and because the proportions were similar for R and E speakers, the data for these distributions were grouped together.) It is obvious from Table 4 that the vast majority of predicate adjectives were used to comment on object or animate states, and that only the more advanced speakers use this form to comment on descriptive properties of objects, such as size or color. Consider now the distribution for attributive adjectives used for referents also shown in Table 4. Here the categories involved are: Descriptive properties (big, little, baby); Evaluative (bad, yummy, pretty, different); Object and animate states (new, bare, broken, whole). The small numbers for this group make the data difficult to interpret, particularly for the less advanced talkers. It appears, however, that more object properties are used in this position for description or specification of the referent than are used in the predicate form.

Some attributes

Table 4.

Number

of words

(percent

of total in paren theses)

Descriptive Properties Predicate

type

used by young

used as referent-specijjing

I HYPOTHESES

t DATA

THEORY PREDICTION

,.

-

DATA

OBSERVATION

+-

REALITY

PREDICTION)

adopted in linguistics some decades ago but has now been abandoned. Instead, generative methods are used, largely as a consequence of the work of N. Chomsky. The generative method may be illustrated as in Figure 1. Observations are made and data are collected serving as the basis for the scientist’s formulation of hypotheses. These hypotheses are integrated into a tentative rule system, or theory capable of generating data, or predictions. If the predictions agree with the observations, the rule system used for the predictions is a good description of reality, and knowledge is obtained. If not, the rule system contains errors which should be corrected. In this way, a generative rule system offers a unique possibility of testing hypothetical descriptions scientifically. Generative methods in descriptions of musical styles have recently been advocated, particularly by linguists (cf. e.g., Winograd 1968, Ruwet 1972, Nattiez 1974). The present paper attempts to describe facts in musical theory by means of generative rule systems. Two rule systems will be presented, one for some Swedish nursery tunes, another one for a set of different versions of an old Swedish folk song which are all felt to be melodically similar. The rule systems show striking similarities with the generative phonology of English as proposed by Chomsky & Halle (1968). The significance of these similarities will be discussed.

Generative theories in language and music descriptions

Figure 2.

Tree diagram of’ the constituent his sister”.

structure of’ the sentence

NOUN /\

101

“John beats

PHRASE

\ NOUN PHRASE

NOUN

/ VERB

/\ PRONOUN

NOUN

John

beats

his

sister

Formalization

of sentences

Let us start with recapitulating the major properties of generative grammar by considering the linguists’ description of certain aspects of spoken sentences. Take the utterance John beats his sister In a traditional grammatical description of this sentence the following features might be mentioned: First, the whole string of words makes up a sentence (S). beats his sister is a verb phrase (VP) which consists of the verb (V) beats, and the noun phrase (NP) his sister. The latter unit is made up by the pronoun (Proun) his and the noun (N) sister. John is an NP that consists of a noun. This information on the constituent structure of the sentence can be represented in the form of a tree diagram as shown in Figure 2. The constituent structure of an utterance is reflected in the prosodic aspects of pronunciation, i.e., in intonation and stress pattern. A procedure for relating prosody and constituent structure has been proposed by Chomsky & Halle (1968). It can be diagrammed as in Figure 3. Applied to the sentence: John beats his sister this procedure works as follows. We can represent the distribution of stresses that might be observed in speech under normal, idealized conditions, as Join

be&s his si&ter

where the heaviest, main or primary

stress = 1, secondary

= 2, and tertiary

102

Johan Sundberg and &G-n Lindblom

Block

Figure 3.

diagram of the Chomsky

CONSTITUENT

PROM I NENCE

DURATIONS,

FUNDAMENTAL

prosody.

CONTOUR

FREQUENCY,

BEATS

JOHN

of speech

STRUCTURE

Cyclic derivation of prominence

Figure 4.

& Halle model

HIS

AMPLITUDE,

VOWEL

QUALITY

contour of sentence. SISTER RULE:

OUTPUT:

= 3.

((

1

jN ((

1

jv

((

1

jN ((

’

)” (

((

’

IN (

2

I

(

2

3

1

2

3

I

(( 1 jPRON(

1

I

STRESS )N)NP)VP)SLEXICAL PLACEMENT

)NP)VP)S )VP’S

)s

RIGHT

PRIORITY

RIGHT

PRIORITY

Within the framework of Chomsky & Halle’s theory of generative phonology it has been suggested that a sequence of stresses such as that proposed for John beats his sister be generated in the manner illustrated in Figure 4. Here the constituent structure is represented in terms of 1abeEed brackets which is exactly equivalent to a tree diagram. In short, the computation of the stress contour begins inside the innermost pair of brackets. We start out by applying rules that assign stresses to the individual words. The deepest constituents are his and sister. The latter receives stress on the first syllable. The former is a pronoun and does not receive

Generative theories in language and music descriptions

103

stress. Next we erase the innermost pair of brackets and again look for the deepest constituents. This time we find (John)N(beats)v and (his sister)Np. At this level also John and beats are dealt with according to the rule for word stress. Erase innermost parentheses once more and we get the second line of the figure. The verb phrase, the VP, contains two main stresses. Since it is a phrase, a right-priority rule (Nuclear Stress Rule) is applied, that is, the rightmost stress remains and the other one is lowered by one degree. After this operation brackets are erased and we now have the sequence (1 2 1 )s_ The right priority rule gives 2 3 1 and that is exactly what we intended to get. Two things should be observed here. One is that the prominence contour is derived automatically from the constituent structure by means of a rule system. The second is that the prominence values determine a number of important properties in the acoustic code. Next it will be demonstrated that this seems to hold also in our melodies. Formalization

of melodies

Figure 5 shows an overview of the occurrence of various phenomena in meter, harmony and tone assignment in a series of Swedish 8-bar nursery tunes written by the composer Alice Tegner. These melodies were composed in the late 19th century and they are familiar to and sung by most Swedish children even today. With respect to meter we observe that in even-numbered bars long durations occur and short notes are avoided. The chords are divided into three groups according to their harmonic functions: Introductory chords which present the tonic, target chords, chords preceded by their dominants, and anticipatory chords, dominant chords followed by its own tonic. Introductory chords are found in bars I and V, and target chords may occur at the end of every bar. Anticipatory chords are not found at the end of bars IV and VIII. With melody we distinguish between chord notes and non-chord notes. The suspension is regarded as a special case of a non-chord note. Suspensions seem to be avoided in approximately the same places where short durations are avoided. By and large non-chord notes occur in the same positions as suspensions, but have a greater number of forbidden positions. These positions show a symmetrical and regular pattern. With each of these three kinds of data we see traces of a segmentation of the tune. Thus, bar VIII, particularly, but also bar IV, and bars II and VI, are treated in quite a special way. Periodicity occurs in all the diagrams. It is noteworthy that these regularities are associated with conceptions employed by musicologists for a long time: Beats, subphrase, phrase, and period.

104

Johan Sundberg and BjGrn Lindblom

Figure 5.

Schematical demonstration of the distribution of metric, harmonic, and tonal features within a corpus of Swedish nursery tunes. Shaded areas indicate the position in which the feature listed to the left has been observed to occur. Position is shown with reference to the bars of the period (vertical dotted lines). BAR XCURRENCE d AND

OF

J

,

1

HARMONY INTRODUCTORY CHORD

m

TARFET

-_I-

CHORD

ANTICIPATORY CHORD

NON

CHORD

,,,

----

MELODY NOTE

SUSPENSION

Model Figure 6 shows an attempt to integrate the observations made into a generative model. At the deepest level we find the syntactic structure. Through a prominence contour this structure determines the choice of durations, chords, and pitches. We now describe in more detail the most important properties of the generative model. A syntactic structure of the type shown in Figure 7 can be ascribed to a typical 8-bar melody. The entire melody is a period consisting of two phrases. Each of them consists of two subphrases made up of bars, feet, and beats. The problem is to relate this constituent structure to durations, and to harmonic and tonal data. In order to do this we transform the syntactic information to a string of prominence values. This procedure is exactly the same as in the case of the linguistic example mentioned before. The procedure is illustrated step by step in Figure 8. The topmost line shows the set of beats, each embedded in pairs of parentheses so that the syntactic structure is reflected in the distribution of parentheses. In erasing the parentheses, a left priority is applied up to the bar level and right priority at higher levels. The final result is given in the lowermost line in the figure. It gives the prominences of the beats along a scale of

Generative theories in language and music descriptions

Figure 6.

105

Block diagram of the generative model. I

I

L-l--J SYNTAX

CONSTITUENT

STRUCTURE

PROMINENCE

PROMINENCE

CONTOUR

TIMING RULES

CHORD RULES

TONAL RULES

HARMONIZED MELODY

integers where numbers refer to rank of prominence. This prominence contour provides a method of relating the syntactic structure to the metric, harmonic and tonal properties of the melody. It gives the raw material for the derivation of meter and harmony and for the assignment of pitches in a melody. We should now explain in more detail the nature of the various components of the rule system. Let us do this by starting with the following

:

JJ

14

/B\

J1

of prominence

J1

=6

2

71 =7

=g

PERIOD

/Y =9

((3

(4

PERIOD:

(((2

PHRASE:

((((1

FOOT:

5

4

3

2)

4

3

2

(1

5

4

3

2)

4

5

4

3

2

3

2)

(1

3

2

1

(1

5

4

3)

2))

4

3

(2

((1

/p\

41

5

4

3

2)

4

3

2

(1

=I1

/& =12

CLOSING

r\ =I3

11

J1

1

1

J

4

2)

5

4

3

4

3

2

(1

2)

5

4

3

2

14)

1

(1

5

3))

2)))

4

(3

((2

(((1

2)

5

4

3

4

3

2

(1

2)

5

4

3

4

3

2

(1

1

5

4

3

2)

4

3

2

1

(1

4

5

4

3)

2))

d

4

3

(2

((1

1

5

4

3

2)

J

4

3

2

(1

1

5

4

3

2)

4

4

3

2

(1

J

5

4

3

2)

J

15)

1

4))

/Y F16

RIGHT

RIGHT

RIGHT

LEFT

PRIORITY:

F15

2))))

d

4

1 3)))

(1

d

F14

lP\

PHRASE

values for the beats in the period shown in Fig. 7.

=I0

3

/\

ofan 8-bar period in 4/4 time.

JJJJJJJJJJJJJJJJJJJJJJJJ

F5

7i

/\ PHRASE

strucCure

(((((l)(l))((~)(l))((l)(l~)((~xl))x((l)(l))((~)(l))((l)(l))((l)(l))))((((lx~)x(l)(l)x(lxl)x(~xl)))(((l)(~)x(lxl)x(lxl)x(l)(~)))))

44

SUBPHRASE:

BEAT:

INPUT:

Cyclic derivation

Figure 8.

/\

JJJJJJJJ

:

OPENING

Tree diagram of the constituent

BEAT:

FOOT:

BAR:

SUB PHRASE

PHRASE

Figure 7.

Generative theories in language and music descriptions

107

108

Johan Sundberg and Bj&-n Lindblonz

question: Have we found that system of rules which automatically and completely enumerates Tegner’s nursery tunes? If we have, the rule system should be able to generate all of Tegner’s &bar melodies. How do we choose among the various alternatives, made available by the rules at every step of the generation, in such a way as to produce a melody that approximates as closely as possible a given Tegner melody? The first step is to generate the meter. An overview of the metric rules is given in the right column of Figure 9. These rules are applied in the order shown in the figure. In the Ieft part of the figure the derivation can be followed step by step. The uppermost line shows the input to the rules constituted by the string of prominence values that has just been developed in Figure 8. First, the sequence-pattern is determined. This shows which bars are to be metrically similar and which dissimilar. Two types of metrical pattern may occur, A and B, and the metrical pattern of the first four bars is obligatorily repeated in the last four bars. The next step offers three alternative choices: Delete prominence rank 5, (a) when it follows prominence rank 1, (b) when it follows prominence ranks 1 and 2, (c) when it follows prominence ranks 1, 2, and 3. We call this deletion procedure catalexis, borrowing a term from the theory of verse. In the next step we may insert additional prominences of rank 6 after prominence ranks 4 and 5 in the first bars marked A and B. Though not done in this case, it is possible at this point to dot notes immediately followed by minima in the prominence contour. After this, the insertions and dottings are transferred to the other bars in accordance with the sequence-pattern. The final step is to transform the modified prominence contour into durations. Every prominence except those of rank 6, and those which have been dotted, is assigned a duration of one beat. Figure 10 presents an overview of the rule system used for the chordal interpretations of prominence ranks. Here, too, the rules are applied in the order shown in the figure. Their function is to determine how often and in which way the chords change. The derivation is illustrated in Figure 11. Once again the topmost line shows the input material, i.e., the string of prominence values derived in Figure 8. An obligatory rule deletes all items of prominence rank 5. After this, some items of prominence rank 4 may be deleted. Next, introductory chords, I:s, and target chords, M:s, are distributed. It would be possible to return to the introductory chord at the beginning of the closing phrase, but this has not been done in the melody considered. As a fundamental principle of the harmonic function of chords we propose that anticipatory chords, DOM:s, must alternate with rest chords, V:s. Given the function of a chord, its harmonic distance from the tonic is determined by its relationship to the following target

Generative theories in language and music descriptions

Overview of the harmonic rules.

Figure 10. HARMONIC

1:

DELETION

OF 4’S

2:

DELETION

OF

TEMPO 5’S

___-__---------___--_----_-____-_

__-__-_-----3:

CADENCE

109

INTRODUCTORY

4:

TARGET

5:

INTRODUCTORY

6:

EXTENDED

CHORDS

(OBLIGATORY

CHORDS

(OBLIGATORY

RULE) RULE)

STRUCTURE CHORDS

INTRODUCTION

7:

HARMONIC

FUNCTION

8:

HARMONIC

DISTANCE

_____--------------------------REALI7ATION

RULE)

(OPTIONAL

RULE)

-_----------

________-__---__-_-__-----------ABSTRACT HARMONIC REPRESENTATION

(OPTIONAL

FROM --

9:

CHORD

THE

TONIC

-

SYMBOLS

chord. These distances lie along a scale of integer numbers and in the figure they are given as indices. The abstract representation of the harmony obtained is then transformed, step by step, to chord symbols in a completely automatic way: When the harmonic functions and distances from the tonic have been determined for a set of chords, the chords themselves can be chosen automatically. The output of the metric and harmonic rules provides the material required for the generation of pitches. The rules applied to determine melody are those given in Figure 12. First a sequence pattern is assigned, determining which bars are to be melodically similar. After this, pitches are assigned in an order corresponding to the order of the prominence ranks. The various lines in Figure 13 show, step by step, the stages of the derivation. After the sequence-pattern has been determined, the pitches for prominence ranks 1, 2, and 3 are assigned in turn. The main rule allows nothing but chord-notes in these cases. After this, pitches are assigned for prominence ranks of four. Here there are three guiding principles: (1) Harmonic implication (the pitch signals the underlying chord); (2) tonal adjacency or the “short-cut-principle” (when two pitches have been assigned intermediate notes will receive intermediate pitches) and (3) the sequence-pattern. At this point it is possible to insert suspensions, which represent a special case of the principles of harmonic implication and tonal adjacency. In the present example this possibility has not been used.

OUTPUT:

INPUT:

Figure 11.

!

:

;:

Sp

1

j

SP

i

T

I

I

1

V2

1

T

I

1

1

V2

I

I

I

V2

DOM!

)

V

I

1

I

D7

D7

D7

DOMl

DOMl

41

31

II

41

4

I

b

1-I

5 14 52

31

3

5 4

41

14

5 14 5 3 5 b

I

f4

4

14 5 4 2

I

(4

514

T7

IS

SP

SP

DOM3 (S

V2 V2

1

V2

V

31

31

3

1

1 D7

D7 1

1

I

D7

I

14

514

I

DOMl

DOM

4

4

4

5 4

I D-f1

/

1

14

5 3 514

DOM3 1

“OM3

(4

514

DOM3

4

5 4 5

T

T

T

T

M.

M

DISTANCE

FUNCTION

11 CHORD

11 CHORD

11 CHORD

SYMBOLS

SYMBOLS

11 CHORDSYMBOLS

11 HARMONIC

II HARMONIC

AND

4’S

5’S

TARGET

FROM

INTRODUCTION

11 INTRODUCTORY EXTENDED

11 DELETION

OF

OF

1

11 DELETION

M

1

1511

THE

TONIC

CHORDS

Generative theories in language and music descriptions

Figure 12.

Overview of the tonal rules. CONTENTS

OF

4

HARMONIC

IMPLICATION

4

HARMONIC

IMPLICATION

ORDER 1.

PATTERN

2.

PROMINENCES

3.

PROMINENCES

OF

RULES

SEQUENCING > =

TONAL

ADJACENCY (SHORT-CUT

4.

111

PROMINENCES