CHILDREN'S COMPREHENSION OF DESCRIPTIONS OF VERTICAL RELATIONSHIPS: THEROLES OF EXTENSION AND REFERENCE POINT A Dissertation for {In Degree of DE. D. MICHIGAN STATE UNIVERSITY Lisa Friedenberg I977 This is to certify that the thesis entitled Children's Comprehension of Descriptions of Vertical Relationships: The Roles of Extension and Reference Point presented by Lisa Friedenberg has been accepted towards fulfillment of the requirements for PhD degree in lsychology (3/ re; s. .7 Major pr es I Date July 18, 1977 0-7639 ABSTRACT CHILDREN'S COMPREHENSION OF DESCRIPTIONS OF VERTICAL RELATIONSHIPS: THE ROLES OF EXTENSION AND REFERENCE POINT BY Lisa Friedenberg Two experiments assessed how Well young children understand descriptions of vertical spatial relationships. The first used adjectives to describe static relationships (high, low, deep, shallow), the second verb phrases to describe dynamic relationships (rising away from, falling toward, falling away from, rising toward). Although the words represented different form classes, both groups referred to four basic types of vertical relationships. These resulted from the classification of spatial locations on the perceptual dimensions of "extension" (how near to or far from the reference point the location was) and "reference point" (whether the location was above or below the reference point). The purpose of these experiments was to illustrate how variables such as extension and reference point influence the learning of different description types in the same way. This prediction was derived from analyses of the importance of these attributes Lisa Friedenberg in the construction of spatial knowledge and in the acqui- sition of spatial terms (e.g., H. Clark, 1973; Fillmore, 1971). These theorists, among others, have proposed that perceptual dimensions are encoded semantically in spatial descriptions. Descriptions which encode the same attributes should be learned in a similar fashion. Two hypotheses formed the basis of each study: (a) that children would understand words for extended locations before words for unextended locations; and (b) that children would understand words for locations above a reference point before words for locations below a reference point. These were derived from studies of how the mapping of perceptual attributes onto lexical items influences children's comprehension (e.g., E. Clark, 1972, 1974). In the first experiment, high and low were used to describe extended and unextended locations abgye a refer- ence point. Deep and shallow were used for analogous locations below a reference point. Objects in these loca- tions were displayed to children between 2% and 5% in a paired-comparison task. On each trial, two felt squirrels were placed in different locations on a felt board. Each location was twice paired with each other location, and was always displayed the same way. Children were asked to point to one squirrel in response to a sentence, such as: "Show me the squirrel in a high place." Lisa Friedenberg In the second experiment, rising away from and falling toward were used to describe extended and unex- tended locations above a reference point, and falling away from and rising toward to describe analogous locations below a reference point. These dynamic relationships were presented to children between 3% and 6% years in a dif— ferent paired-comparison task. On each trial, two mobile airplanes "flew" up or down in relationship to two stationary airplanes. Each location was paired with each other location, and always displayed the same way. Chil— dren were asked to point to one of the moving airplanes in response to a sentence, such as: "The blue airplane is rising away from the green airplane." Errors in pointing responses were analyzed in each experiment. Both indicated that children understood extended words first. Fewer errors occurred on high-deep and rising-falling away from sentences than on sentences using the unextended words. Little support for the reference point hypothesis occurred in either experiment, although there was a slight trend towards fewer errors on high-low than deep-shallow in Experiment 1. Age-mates from the two experiments were compared to determine whether performance was better on either the adjectives or verb phrases. No significant difference occurred. Overall, children appeared first to learn the meaning of words for extended locations and then the meaning of words for unextended locations. Lisa Friedenberg The results are discussed in relation to theories of lexical development, the importance of perceptual space characteristics in the development of spatial knowledge, and the study of spatial terms. CHILDREN'S COMPREHENSION OF DESCRIPTIONS OF VERTICAL RELATIONSHIPS: THE ROLES OF EXTENSION AND REFERENCE POINT BY Lisa Friedenberg A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Psychology 1977 «n DEDICATION This research is dedicated to the children, parents, teachers, and administrators without whom this dissertation would not have been possible. ii ACKNOWLEDGMENTS I would like to thank Drs. Terrence Allen, Helen Benedict, and Gordon Wood for their helpful comments and criticisms; Drs. Hiram Fitzgerald and Gary Olson for their assistance, patience, and encouragement; and my family and friends, whose continual support was so important throughout my work on this dissertation. This research was supported in part by an NIH Institutional National Service Award lT32-MH 14622—01. LIST OF LIST OF TABLE OF CONTENTS TABLES . . . . . . . . . . . . FIGURES . . . . . . . . . INTRODUCTION . . . . . . . . . . . . . Chapter I. II. LANGUAGE AND SPACE . . . . . . . . The Acquisition of Lexical Items . . . Semantic Feature Theory . . Generalization Theory . . . . . . Concept-Formation Theory . . . . 'Process' versus 'Structure' Issues The Structure of Perceptual Space . . References Planes and Points . . . . Reference Directions . . . . . . The Study of Spatial Terms . . . . . . The Lexical Marking Approach . . . . Problems with This Approach . . . . Incorporating Perception and Cognition. The Neglected Issue: Between-Pair Differences and Similarities . . . General Research Hypotheses . . . . EXPERIMENT l . . . Method . . . . . . . . . . . . Subjects . . . . . . . . . . Materials . . . . . . . . . Design . . . . . . . . . . . Procedure . . . . . . . . Results and Discussion . . . Analysis of Variance (ANOVA) . . . . Analysis of Position Biases . . . . Age Differences and Age Changes . Differences Attributable to Extension . iv Page vi vii l4 16 20 23 25 3O 35 36 37 38 4O 41 44 48 48 49 50 51 52 52 57 58 59 Chapter Page Differences Attributable to Reference Point . . . . . . . 64 Differences Attributable to Pairing Conditions . . . . . . . . . 66 III. EXPERIMENT 2 . . . . . . . . . . 68 Method . . . . . . . . . . . . 75 Subjects . . . . . . . . . . 75 Materials . . . . . . . . . . 76 Design . . . . . . . . . . . 78 Procedure . . . . . . . . . . 79 Results and Discussion . . . . 80 Analysis of Variance (ANOVA) . . . . 80 Analysis of Position Biases . . . . 84 Age Differences and Age Changes . . . 85 Differences Attributable to Extension . 87 Differences Attributable to Reference Point . . . . . 89 Differences Attributable to Pairing Conditions . . . . . . . . . 91 Differences in the Acquisition of Adjectives and Verb Phrases . . . 92 IV. SUMMARY AND CONCLUSIONS . . . . . . . 94 Conclusions . . . . . . . . . . 96 Lexical Development . . . . . . . 97 Perceptual Space . . . . . . . . 100 Spatial Terms . . . . . . . . . 102 APPENDICES A. Letter to Parents for Experiment 1 . . . 107 B. Summary Tables for EXperiment 1 . . . . 109 C. Letter to Parents for Experiment 2 . . . 115 D. Summary Tables for Experiment 2 . . . . 117 REFERENCES . . . . . . . . . . . . . . 124 LIST OF TABLES Table Page 1. Characteristics of Subjects in Experiment 1 . . . . . . . . . . . . . . 49 2. Pairing Conditions Used in Experiment 1 . . . . . . . . . . . . . . 51 3. Conditional Proportions in Experiment 1 . . . . . . . . . . . . . . 66 4. Characteristics of Subjects in Experiment 2 . . . . . . . . . . . . . 76 5. Pairing Conditions Used in Experiment 2 . . . . . . . . . . . . . . 78 6. Conditional Proportions in Experiment 2 . . . . . . . . . . . . . . 90 Vi LIST OF FIGURES Figure Page 1. The theoretical reference directions of perceptual space . . . . . . . . . . . 31 2. The significant interaction of extension, reference point, and pairing condition in Experiment 1 . . . . . . . . . . . 55 3. The significant interaction of extension, pairing condition, and age in Experiment 2 . . . . . . . . . . . . . 82 INTRODUCTION This dissertation demonstrates that young children use certain perceptual attributes to organize portions of their lexicons. The order in which they learned different spatial descriptions (adjectives and verb phrases) indicates that words expressing certain perceptual attributes are consistently understood before words expressing other attributes. This order of acqui- sition reflects how these descriptions are classified on the variables of extension and reference point. In fact, children of the same age performed equally well on adjectives and verb phrases which represent the same forms of extension and reference point. To identify a spatial location, the perceptual attributes defining that location must be recognized. When language is used to describe the location, the relationship between the perceptual attributes defining the location and the semantic attributes defining the words must be recognized. This relationship is often referred to as a 'mapping' of perceptual attributes onto lexical items. One of the most interesting characteristics of language is that a given situation may be described in a variety of ways. Spatial descriptions are well-suited 1 for studies of similarities in meaning. Perceptual attributes may be clearly identified, and descriptions which must contain those attributes may be generated easily. ‘By examining the order in which children learn different spatial descriptions, it is possible to assess the relationship between perceptual and semantic features. Fillmore (1971),IL Clark (1973), and others pro- pose that perceptual attributes differ in saliency and complexity. These differences have been attributed to our egocentric reliance on the body's orientation in space during the construction of spatial knowledge. These theorists also propose that spatial descriptions differ in complexity according to the perceptual attributes they encode semantically. In the present eXperiments, the Spatial descriptions learned early contained salient, less complex perceptual attributes as elements of meaning. Thus, the dissertation demonstrates that theoretical dif- ferences in the complexity of perceptual attributes pre- dict data from language comprehension, suggesting that these theoretical differences represent a system which children employ in the learning of spatial descriptions from different form classes. Each eXperiment presented Spatial locations which were either above or below an established reference point (the freference point' variable) and either near to or far from that reference point (the 'extension' variable). The first experiment used the adjectives high, low, deep, and shallow to describe static relationships. The second used the verb phrases rising away from, falling toward, falling away from, and rising toward to describe dynamic relationships. In each case, there was one lexical item for each of four types of vertical spatial relationships: high and rising away from refer to loca- tions far above a reference point; low and falling toward refer to locations above but near to a reference point; deep and falling away from refer to locations far below a reference point; and shallow and rising toward refer to locations below but near to a reference point. Both experiments employed paired-comparison pro— cedures. Preschoolers and kindergarteners were shown objects in two contrasting locations, and were asked to choose which object represented the location described by a sentence. Each location was twice paired with each of the three remaining locations in each eXperiment. This allowed for identification of different response char— acteristics, such as scoring correctly on an item by eliminating the "wrong" choice (a nonlinguistic strategy), confusing words for the opposing ends of a dimension, or confusing words for locations from different dimensions which represent the same form of extension. The dissertation is organized in the following manner. The first chapter reviews literature in three areas: lexical development, the structure of perceptual space, and the study of spatial terms. Research findings and proposals within these areas are used to derive the experimental hypotheses. Chapters II and III present the two experiments, followed by a general discussion and the overall conclusions in Chapter IV. CHAPTER I LANGUAGE AND SPACE As the child progresses from a preverbal state to mastery of language, there is a multitude of information which must be acquired and organized. The child must learn about semantic relationships, those inherent in the meaning of a word, and syntactic relationships, those into which a word may enter according to the rules of grammar. Despite the apparent rapid development of language, cer— tain properties of these relationships affect even adult language comprehension. One property of semantic rela- tionships has been extensively illustrated with words from different form classes. When antonym pairs are used in comprehension tasks, one pair member is typically understood before its Opposite (e.g., E. Clark, 1972a; Donaldson and Wales, 1970; Friedenberg and Olson, 1977) and is easier for adults to comprehend (e.g., H. Clark, 1969a, 1969b; H. Clark and Chase, 1972, Friedenberg, 1976). Early theories explained this phenomenon with strictly linguistic criteria, but more recent research suggests that it results from mapping of differences in perception and cognition onto lexical items (e.g., H. Clark, 1973; Fillmore, 1971). The majority of these demonstrations have used spatial and temporal word pairs. It is easy to see how differences in perception could affect concepts of space, which in turn could influence comprehension of spatial words. Expression of these processes for temporal con— cepts could be accomplished by attention to the iso- morphism between spatial and temporal terms, sometimes referred to as "time as a spatial metaphor" (H. Clark, 1973). Differences in perception of space have been attributed to several factors: the reference plane for the dimension, the normative direction for the dimension, the location of dominant sensors within the dimension, and other physical assymmetries of the human body along dif— ferent dimensions (e.g., H. Clark, 1973). The question now becomes: Are the observed differences in language comprehension congruent with hypothesized differences in perception of space? The literature reviewed in this chapter seeks to provide a description of language acqui- sition and of perceptual space in which this could occur. The Acquisition of Lexical Items There are many different approaches to the study of lexical development, and it is not possible to present a complete picture of this area in a few pages. Instead, this review focuses on theories of how aspects of the environment, including perceptual attributes, are represented in word meanings, how these attributes are acquired in the process of learning words, and how these attributes influence conceptual development. Additional information about lexical development is available in such sources as: Anglin (1970), Bowerman (1973), Bloom (1973), Braine (1976), Brown (1973), E. Clark (1973a, 1974), MacNamara (1972), and Nelson (1973, 1974). Semantic Feature Theory Semantic feature theory describes both how aspects of the environment are represented in word mean— ings, and how these attributes are acquired in the process of learning words. Lexical development is viewed as the acquisition of 'semantic features,‘ which are considered to be the basic elements of word meaning (e.g., E. Clark, 1972a; 1973a). The meaning of any word may be represented as a hierarchy of semantic features, with each feature encoding a different critical defining attribute of that word. When children learn a word, they learn that any situation in which that word can be used must contain these critical defining dimensions. Thus, feature theory provides a means for directly representing perceptual attributes as elements of meaning in spatial words. According to feature theory, the defining features are abstracted by the child from the context in which the word is used. However, feature theory makes several Claims about which features are abstracted at different points in lexical development. To understand these claims, some familiarity with the feature system is needed. Semantic features are arranged hierarchically, with general features, like 'living' at the tOp of the hierarchy, and specific features, like 'married' at the bottom. Each feature is coded as positive or negative. Positively coded features indicate that the referent must contain a specified attribute; negatively coded features indicate that the referent must lack that attribute. The categories of 'living' and 'not living,’ then, would be represented as (+1iving) and (—living). (This coding system reduces the number of features needed to define words, and allows for easy identification of relationships between words, such as the identification of antonyms.) Children begin by acquiring only a few features defining a given word and continue to acquire additional features until the entire set is learned.‘ Two criteria determine which features are among the first few learned (e.g., E. Clark, 1972a). First, children learn features which encode salient perceptual attributes such as size, shape, movement, sound, and texture before they learn other types of features. Second, children learn positively coded forms of these perceptual features before they learn their negatively coded forms. According to Rosch (1973; Rosch, Mervis, Gray, Johnson and Boyes-Braem, 1976), features which differen- tiate classes of objects and events co—occur in predict- able ways. By learning that an animal has feathers, one knows it is more likely for that animal to have wings than an animal with fur. When children learn features which differentiate objects and define words, it is most efficient to focus on those features which carry the most information and lead to the greatest differentiation among referents. E. Clark (1972a, 1973a) contends.that perceptual features meet this criterion. Perceptual features encode prOperties of the environment which are salient to young children during their early interactions with the world, and which are useful to them in distin- guishing primitive categories. Furthermore, it is simple to note that co—occurrence of salient perceptual features (e.g., feathers and wings) and thereby to lessen the number of features which must be acquired to define a new but related referent. Bierwisch (1967) makes an even stronger statement. Humans are predisposed by their biological and functional characteristics, he suggests, to apprehend these attributes. Perceptual features are universal, then, to the extent that humans are biologically and functionally the same. Regardless of whether the early acquisition of perceptual features results from a biological predisposition, or an interaction of our 10 information-seeking processes and the structure of the environment, there are important consequences of this pattern. By learning the intermediate-level perceptual features first, children obtain the most useful informa— tion about referents in the most simple and efficient manner possible. Positive features are less complex than negative features because they indicate the presence of a quality. Since it is logical to perceive the presence of a quality before noting its absence, positive features are acquired before negative ones. There is a variety of data used to support the claims of semantic feature theory. Major substantiation of the gradual acquisition of features, beginning with general, perceptual features, is found in studies of productive corpora. It is frequently noted that young children use words in ways which differ from adult fluent speakers. They sometimes loverextend words to referents which adults would label with different words, or under— extend words by failing to apply them in appropriate con- texts. E. Clark (l972a, 1973a) suggests that over- and underextensions Ixnnrrt from partial feature acquisition. In her analyses of productive corpora (E. Clark, 1974), she found that most overextensions were based on general, perceptual attributes (e.g., calling all four-legged 11 animals "dog"). She further noted that the particular features learned vary from child to child within these constraints. The acquisition of positively coded features before negatively coded features has been examined extensively in empirical tasks. Donaldson and Balfour (1968) found children unable to differentiate the words "more" and "less." Up to four years of age, children tended to use both words to refer to the 'more' situation. Palermo (1973, 1974) found this pattern to persist up to seven years of age. Wales and Campbell (1970) and Donaldson and Wales (1970) identified a similar phenome— non with pairs of adjectives including: "big-wee," "long-short,‘I "tall-short," "thick-thin," "high-low," "fat-thin," "deep-shallow," and "broad-narrow." In all cases, children demonstrated more accurate comprehension and production of the positive pair member. These results have been replicated by E. Clark (1972a) with an even wider variety of spatio-temporal and dimensional pairs. The pattern persisted until children acquired the negatively coded feature needed to differentiate the pair members. E. Clark further demonstrated (Klatzky, Clark and Macken, 1973) that nonsense syllables used to refer to dimensional concepts such as length, width, and height were learned in the same fashion. 12 Two cautions must be mentioned at this point. First, not all words are learned by gradual acquisition of semantic features. As pointed out by Nelson (1973), some words are learned in highly specific, even ritualized situations (e.g., "hi," "bye—bye"). Second, not all words learned in this fashion are misused (e.g., overextended). The presence of an overextension, for example, implies the lack of a word for the referent of the overextension. It also implies that the child is more sensitive to similari— ties between referents than to their differences. If attentive to their differences, the child may realize that an apprOpriate word is lacking, and not use an already acquired one. Semantic feature theory has many critics, and there are data to refute its claims about the learning of semantic features. As early as 1957, Danzinger presented evidence regarding the acquisition of kinship terms which suggested that children learn concrete, specific features first. However, E. Clark was able to demonstrate (Haviland and Clark, 1974) that this conflict resulted from different conceptualizations of the arrangement of defining semantic features. Data from her study support the early acquisition of general features. Research by Anglin (1970), to be presented shortly, also indicated that specific features are acquired first. His research employed the same types of spatial, temporal, and 13 dimensional words as the early studies cited in these pages. More recently, Brewer and Stone (1975) found it more common for children to confuse words from different dimensions which contained similar specific features, than to confuse words from the same dimension but with different specific features. This indicates that the general, dimensional features were learned after the specific, polarity features——a direct contradiction of semantic feature theory. At a more theoretical level, feature theory has been criticized as a reductionist approach which does not consider the psychological processes underlying communi- cation, the role of context in modifying meaning, or the development of creative language abilities such as meta- phor (Palermo, 1976). Nelson (1974) Views semantic feature theory as an 'updated' version of the abstraction theory formulated by John Locke in 1690. She objects to the lack of specification of what constitutes a semantic feature and the apparent equation of semantic features with perceptual attributes. Finally, she notes that semantic feature theory completely ignores the role of cognition in the child's organization of the environment. Feature theory presents an extremely one-sided View of lexical development, focusing only on how concepts are abstracted from lexical items. Despite these objections, 14 semantic feature theory remains a dominant model of lexical development in current research. Generalization Theory Anglin (1970) also prOposes that words are com— posed of semantic features, but his generalization theory presents an alternative View of lexical develOpment. In this approach, children are believed to be initially more sensitive to concrete, specific features, and to move developmentally toward acquisition of higher-level, more general, abstract features. Anglin agrees that children are more idiosyncratic in their initial definitions of words, and tend to overextend words on the basis of superficial similarities among referents. However, the basis of these early definitions are low-level, concrete features. Anglin also constructs a hierarchy of semantic features to define each word, but has a different View of what constitutes a semantic feature. While E. Clark concentrates on perceptual attributes, Anglin includes perceptual, semantic, and syntactic attributes as defining features. During the early phases of lexical development, children zero in on a specific attribute of a referent, and use that feature (or those few features) to define a word. The child who responds to all men with moustaches as "Daddy" has selected a salient but specific attribute 15 of "Daddy" as the basis of its definition. There are at least as many examples of this type of overextension in early productive corpora as there are examples such as using "dog" for all four-legged animals. Anglin supports his generalization theory with a series of experiments involving synonyms, antonyms, and superordinates-subordinates (Anglin, 1970). The same words were presented to children and adults in several different tasks--sorting tasks, free recall tasks, and free association tasks. In all cases, the developmental trend from young child to adult was toward identification of more general, superordinate relationships among words, implying that these features are acquired last. There is a critical difference, though, between the research supporting generalization theory and that supporting semantic feature theory. Researchers utilizing feature theory have concentrated on lexical development during the preschool period, while researchers utilizing generalization theory have concentrated on lexical develOpment in school—age children. However, it is dif- ficult to say whether this difference could account for the difference in direction of lexical growth between the two theories. Most criticisms of generalization theory focus on the use of component features, rather than the direction of lexical growth. As such, critics of generalization 16 theory point to the same weaknesses in this approach as in feature theory: the lack of independent specification of what constitutes a semantic feature, the reductionist nature of componential analysis, and the isolation of lexical develOpment from the rest of conceptual develop- ment (e.g., Nelson, 1974; Palermo, 1976). In addition, since the two theories propose diametrically Oppose directions of lexical growth, there are as many studies refuting the claims of generalization theory as there are studies supporting them. Concept-Formation Theory According to Nelson (1974), the problem with the preceeding theories is that they rely on language struc— ture as the key to lexical development. It is more reasonable, she prcposes, to view language learning from the perspective of the child's cognitive structure, and to include lexical develOpment in the general scheme of concept formation. Both generalization theory and semantic feature theory prcpose that children can construct concepts as they learn words, by abstracting common seman- tic features (e.g., all words which contain the feature (+living) form the class of living things). Neither theory, though, provides an adequate account of how chil— dren can encode previously acquired concepts with language. (Nelson's objection is not completely accurate. The next 17 section of this chapter demonstrates how differences in the salience of aspects of perceptual space may be used to predict differences in the acquisition of spatial words. The intermediate step in this process is a speci- fication of how differences in perception produce differ- ences in concepts of space, allowing for language items to be mapped onto concepts as the concepts develop.) The influence of language on conceptual develOp- ment has also been addressed by Blank (1974). Blank con- tends that in some situations, language plays a minimal role in concept formation because other representational skills may be used to construct concepts. However, in certain situations language plays in important role in concept formation, because other representational skills alone are not sufficient to derive concepts. As an example, Blank compares the learning of shape discrimina— tions with the learning of temporal discriminations. In the first case, visual information is available to differ— entiate the stimuli, and the use of language to label the shapes did not lead to a great improvement in performance. In the second case, where children were asked to discrimi- nate between one and two flashes of light, the use of language dramatically improved their performance. Blank concludes that in the learning of less tangible concepts, language can play a major role in concept formation. 18 Nelson (1974) outlines a concept—formation theory which includes specification of how children label concepts with lexical items. Throughout their development, children are involved in the simultaneous processes of analyzing the environment into attributes, and synthesizing these attri- butes into concepts. The attributes on which children focus are not static, perceptual attributes. Instead, Nelson proposes that children identify aspects of the environment as dynamic, functional relations. An object is not 'heavy,‘ for example, but 'hard to lift.’ These functional relations form the basis of both conceptual and lexical development. They are represented as elements of early core concepts, and as elements of early word mean- ings. Children's initial use and misuse of words reflect the mapping of a few, salient, functional relations onto lexical items. Nelson even cites E. Clark's identification of movement and sound as salient features to support her claims about functional relations (Nelson, 1974). How are these functional relations acquired? Nelson presents a four stage model of concept formation. In the first stage, children must identify 'wholes' in the environment (i.e., objects, events, etc.). These 'wholes' are the basis of primitive categorical concepts. In the second stage, children must be able to identify the important functional relations associated with these 'wholes.‘ Initially, some relations which are irrelevant 19 to the core concept may be included as concept-defining elements. As children continue to interact with the environment, they learn which relations are concept— defining and which are not. In the third stage, children identify new instances of these 'wholes' by noting stable, salient, functional relations. This helps refine concept definitions further via exclusion of irrelevant relations. In the fourth stage, children can attach names to the resultant concepts. The fact that lexical acquisition occupies the fourth position in this model does not dic- tate that words cannot be learned before concepts are fully developed. The child does not merely pass through these stages once in the formation of a concept. A label may be attached to a core concept which still includes certain irrelevant relations, and still lacks certain critical defining relations. Thus, there are concommitant changes in word meanings as different relations are mapped onto lexical items. Nelson's own research supports the hypothesis that early concepts and word meanings reflect the importance of functional relations. Additional confirmation of her View is mounting. For example, Bloom, Lightbown and Hood (1975), found that children first encode dynamic, action events, and then static, stative events. Their data also indicated that children encode dynamic, nonlocative rela- tions before static, locative relations. Sinclair (1976) 20 contends that cognitive develOpment can only proceed by acquisition of dynamic, functional relations. According to her Piagetian perspective, children's interactions with the world are always dynamically and functionally based. Beginning with the sensorimotor period, all knowl- edge is represented as action schemes. While even early concepts include information about perceptual character- istics, it is unlikely that this information is repre— sented as static attributes. It is more likely that the information is represented as it was acquired--as func- tional relationships (e.g., 'hard to lift,‘ rather than 'heavy'). 'Process' versus 'Structure' Issues On the surface, these three theories seem to be quite different. However, they share two characteristics. First, each is concerned with the process by which children acquire lexical items, and the relationship between this process and the derivation of concepts. Second, each attempts to derive a set of semantic structure elements to characterize lexical meaning. Regardless of the type of elements prcposed (semantic features or functional rela- tions), the elements represent the encoding of aspects of the environment which differentiate objects, events, etc. 21 Familiarity with these 'process' and 'structure' issues is essential to an understanding of the present dissertation research. The focus of the present experi- ments was the acquisition of semantic structure elements which differentiate words for spatial locations. The purpose of this approach was to identify a framework within which different spatial descriptions were learned, a framework which directly corresponded to theorized dif- ferences in our perception and organization of space. As such, the experiments address neither the process by which these elements are learned, nor their actual character (semantic features or fUnctional relations). The labels given to the semantic dimensions under study could indicate either semantic features or functional relations. The critical factor is the overall structure or framework within which percepts, concepts, and lexical items are organized by young children. Whereas much attention has been paid to the acquisition of syntactic structures (e.g., Brown, 1973), far less research has assessed the acquisition of semantic structures. This includes research which attempts to relate the acquisition of such structures to cognition. Sinclair (1971) has presented an interesting account of how the acquisition of syntactic structures reflects sensorimotor cognitive processes. Yet researchers are 22 just beginning to look for isomorphisms between semantic and cognitive structures. Spatial words are well—suited to the examination of isomorphisms between semantic and cognitive structures. The critical dimensions encoded by these words may be easily identified, and it is possible to generate words from many different form classes which encode these dimensions. By studying the comprehension of spatial adjectives and verb phrases, it is possible to determine whether differences in their complexity are independent of (a) the form class under study; and (b) the level of language ability of the participant (e.g., child or adult). If complexity differences transcend these restrictions, they must indicate differences in the complexity of basic semantic structure elements. Furthermore, if these dif- ferences are congruent with hypothesized differences about our perception of space, they imply a general, organiza- tional framework which guides the learning of spatial percepts, the construction of spatial concepts, and the learning of spatial descriptions. Factors underlying differences in our perception of space are discussed in the next section, including inferences about how these differences affect the develOp— ment of spatial concepts. Beginning with the end of infancy, when children are able to internally represent aspects of the environment, children construct concepts 23 based on their interactions with the environment. The more salient a particular aspect is, the more experience children have with the aspect, the more highly defined and differentiated the resultant concept will be. In addition, as put forth by Nelson (1974), the rudiments of a core concept must be present before a label can be acquired. Thus, concepts which develOp earlier may potentially be linked to lexical items at earlier points in develOpment, although this is not necessarily the case. By studying the acquisition of lexical items, it is possible to trace the expression of developing concepts as they are mapped onto language. If a child understands a location when referred to one way (e.g., The duck is higher than the fish), but cannot identify the location when asked for in a different way (e.g., The fish is lower than the duck), it is not possible to conclude that the child lacks conceptual understanding of that location. However, it is possible to infer that the former presenta— tion of that relationship is a form which is more congruent with the way the child typically conceptualizes the rela- tionship. The Structure of Perceptual Space To derive the predictions about the framework within which these spatial descriptions are learned, we must examine the structure of perceptual space. Perceptual 24 space is composed of a coordinate system of several axes, and the language of space corresponds to dimensions and points within those dimensions. The relationship between spatial words and physical space is most obvious in the study of Spatial antonyms. Each antonym pair refers to a perceptual dimension, with each pair member indicating a location on opposing sections of that dimension. Children must deal with spatial words in many con- texts, including reading and writing (e.g., 'up' or 'down' on a page, the 'above' section, etc.), learning parts of the body (e.g., 'left' and 'right'), and orientation of the self, other peOple, and objects and events (e.g., deictic words such as 'here' and 'there,' the inherently relational nature of words like 'more' and 'less,' etc.). Children must also learn that any given spatial relation- ship can be described in a variety of ways. Adjectives (e.g., higher—lower than), prepositions (e.g., above— below), and verb phrases (e.g., rising—falling away from) may all be used to describe two objects which are verti— cally oriented, depending on the spearker's preference or point of emphasis (e.g., whether or not motion is important to convey to the bearer). Perceptual space consists of a set of reference planes, points, and directions with which spatial phenomana are organized and spatial relationships are conceptualized. It is frequently proposed that particular attributes of 25 space are more salient than other attributes, and are more likely to be used by children and adults in the construc- tion of spatial knowledge. For example, Palermo (1976), prcposes that these organizational characteristics reflect the abstracting characteristics of the human organism, and thus constitute natural prototypic concepts from which other concepts may be derived. Rosch (1973) suggests a similar phenomenon in her description of 'natural cate- gories.‘ These prcposals are reminiscent of Bierwisch's contention (1967) that certain features of the environment are universally important because of the biological con- struction of humans and their functional interactions with the environment. Perhaps the strongest statements about the exist— ence of a bias in our construction of spatial knowledge come from researchers like H. Clark (1973) and Fillmore (1971). These individuals contend that the specific attributes of space used to organize the perceptual world reflect an egocentric consideration of the body's orienta- tion in space. References Planes and Points When one watches the sun set over the horizon, a vertical relationship is experienced. The location of the sun relative to the horizon (its reference plane) gradually changes. When someone enters the room and approaches you, 26 a horizontal relationship is experienced. The location of that person relative to you (the reference plane) changes as the person approaches. These two situations exemplify the standard primary reference planes described by H. Clark (1973). In the first case, a vertical rela- tionship involving a horizontal reference plane (the horizon or 'ground level') was eXperienced. In the second case, a horizontal relationship involving a vertical reference plane (your body) was experienced. The most commonly experienced reference planes for vertical and horizontal relationships, according to H. Clark, are the ground level and one's body respectively. Sometimes, a relationship involves only a point within one of these planes. Yet in all spatial relationships, some reference location is present. H. Clark further distinguishes between primary and secondary reference locations. In addition to the primary system just described, secondary reference locations may be established by objects which are located relative to primary ones. For example, judging the location of a star relative to the moon involves conceptualizing the moon as a secondary reference point. The intermediate step, which is rarely considered, involves conceptualizing the location of the moon relative to the primary reference plane of the ground. Thus, the ability to understand relationships 27 involving secondary reference locations presupposes an intuitive grasp of the primary system. However, there is a difference in our experience of reference locations for horizontal and vertical rela- tionships, a difference which could affect the develOpment of percepts, concepts, and labels for them. As pointed out by Fillmore (1971), horizontal relationships are often first experienced relative to the 'self' (i.e., in situa- tions where one's own body is the reference plane). Because of this initial egocentric conceptualization of horizontal relationships, children can approach relation- ships with independent reference locations egocentrically, and identify the 'self' with the independent location. As Piaget and Inhelder (1956) demonstrated, young children often contend that a person sitting at the Opposite side of a table sees a display from the same perspective as their own. Piaget and Inhelder attribute this tendency to a general cognitive inability to decenter and shift per— spective. Fillmore (1971) proposes that this is a problem peculiar to any study of horizontal relationships. While he does not refer to the Piaget and Inhelder findings specifically, he contends that the study of horizontal relationships is always potentially confounded with this type Of egocentric identification. This problem should not occur in the learning Of vertical relationships if they are initially experienced 28 with the ground as the reference plane. The body is generally considered to be vertically oriented. The ground level is a horizontal plane. In theory, it should be easier to identify with other vertically oriented objects, such as other peOple, trees, buildings, etc., than to identify with horizontally oriented Objects. Thus, it is possible that secondary reference locations for vertical relationships can more readily retain their independent, non-egocentric character. There is some empirical support for the contention that vertical and horizontal relationships are conceptualized in different manners. In the learning of words for motion on these dimensions, words for horizontal motion toward a reference location are learned before words for horizontal motion away from a reference location (e.g., Clark and Garnica, 1974; Fillmore, 1966; Fillmore, 1971). Assessments of words for vertical motion have been lacking, but the research in this dissertation indicates that words for vertical motion away from a reference location are learned before words for motion toward that location. (This point will be discussed in greater detail after the presentation of the research findings.) Piaget and Inhelder (1956) contend that the learn- ing of relationships where the 'self' is the reference point is ontogenetically prior to learning relationships with an independent reference location. If it is in fact 29 true that vertical relationships typically involve non- egocentric reference locations, this would imply that children learn about vertical relationships after learn- ing about horizontal relationships. However, literature reviewed by Miller and Johnson-Laird (1976) indicates that even very young children are capable of non-egocentric space perception when tasks are presented appropriately. Thus, there is no a priori reason to assume that either type is learned before the other. In fact, the vertical dimension seems to be more salient to young children than the horizontal dimension. Neonates scan past a horizontal line entering their visual fields, but immediately fixate their gazes on a vertical line and scan back and forth across it (Kessen, Haith and Salapatek, 1972). Children as old as five years define the word "big" primarily in terms of 'vertical extent' (Maratsos, 1973). Even when differences in the horizontal (length) dimension are grossly exaggerated, children choose the taller of two objects when asked which is "bigger." Gibson (1969) comments that it is far easier to distinguish aspects of the vertical dimension (e.g., 'Hqfl'and"down") than to distinguish aspects of the horizontal (e.g., 'left' and 'right'). Nelson (1974) cites research by a colleague who identified the word "up" among the first five words used by a one-year-Old 30 child to indicate any vertical motion in space. No such word for horizontal motion was present at that time. It appears, then, that there are important differ— ences between our conceptualization of horizontal and vertical relationships. This is complicated further by the fact that there are two horizontal axes in Space-— the front/back and the 1eft/right--and only one vertical axis--the up/down. This complication is quite obvious in the delineation of reference directions for each of these dimensions. Reference Directions Figure 1 presents a schematic drawing of the three spatial axes. Assuming that there is an origin or national reference point, there are four possible reference direc— tions for each dimension. Two of these are vectors which originate at the reference plane, and extend infinitely from it. The remaining two are vectors which terminate at the reference location, originating somewhere in space. The vectors which originate at the reference location are typically referred to as 'forward—backward,‘ 'to the left- to the right,‘ and 'upward-downward.‘ The vectors which terminate at the reference location do not have standard labels, and are referred to in a variety of ways (e.g., 'come' and 'bring' for the horizontals, 'to' or 'toward' for all three dimensions, etc.). This fact alone is of 31 1|: +—————————-—- ooooeoooooqoo......0.-QOOOOCOOO I / FRONT ........ vectors which originate at the reference point ._ - vectors which terminate at the reference point ———-———-§ ‘o....o 0......oooo coo-coo f DOWN 32 interest: That English provides Specific vocabulary for one set of directions, but lacks precise vocabulary for the other set. According to H. Clark (1973), as children become aware of their orientation and movement in space, they experience certain directions more readily than others. These directions are commonly used in the initial organiza- tion of space. For example, our standard Vector of loco- motion moves forward on the horizontal. Thus, 'forward' is easily used as the reference direction for the front/ back dimension. H. Clark also notes that our perceptual apparatus is distributed asymmetrically along this dimension, since it is located exclusively on the front side of the body. It is more difficult to determine a reference direction for the left/right dimension. There is no loco- motive vector on which to rely, and the perceptual appara— tus is symmetrically distributed along this axis. However, various societal conventions (e.g., reading, writing, driving) and certain functional abilities (e.g., handed- ness) emphasize one direction over the other, and may operate to induce a dominant reference direction (e.g., Olson and Laxar, 1973, 1974). The lack of a clearly experienced reference direction may be responsible for the difficulty children have in learning to distinguish left from right (e.g., Corballis and Beale, 1976). 33 The reference direction for the vertical dimension, according to H. Clark (1973), is upward, determined by our perpendicular stance and the location of dominant sensors at the upward end of the body. While the forces of gravity define a downward vector, which could be the reference direction for this dimension, research indicates that 'upward‘ is more readily conceptualized than 'downward.‘ Studies in progress by the author indicate that, when pre- sented with two vertically oriented shapes, adults are faster and more accurate in detecting changes in the top shape than changes in the bottom shape. Studies of lan- guage comprehension with children (e.g., Friedenberg and Olson, 1977) and adults (e.g., Friedenberg, 1976) indicate that words for 'upward' locations are learned earlier and processed more rapidly than words for 'downward' locations. Thus far, only vectors which originate at the reference location and extend infinitely from it have been discussed (the dotted lines in Figure 1). Theoretically, 'forward' is the normative direction for the front/back dimension, and 'upward' is the normative direction for the up/down dimension. (Because of the ambiguous nature of the left/right dimension, it will be omitted from future discussion.) One of the major hypotheses of this dissertation was that words for locations above a reference point would be acquired before words for locations below a reference point. According to H. Clark's analysis of the 34 vertical dimension, this prediction is consistent with the theorized normative character of the 'upward' vector. If this direction is more readily experienced and more easily conceptualized than the 'downward' direction, then words for locations along the 'upward' vector should be acquired first. The second major prediction Of the dissertation research was that words for extended locations would be acquired before words for unextended locations. However, this prediction is derived more from research on language acquisition than from research on the structure of per- ceptual space. (This research will be presented in the next section.) As is frequently the case in research on language and space, proposals about differences in percep— tion are based on Observations about differences in language. For example, there are two vectors in Figure 1 which extend infinitely along the vertical dimension, and two vectors which terminate at the reference point. Loca- tions which are characterized as 'extended' would lie far from the reference point at spots which could be the terminal locations Of the infinitely extending vectors. high and dggp represent two extended locations which con- form to this pattern. Locations which are characterized as 'unextended' would be near to the reference point, at spots where vectors originating somewhere in space would terminate. Low and shallow represent this type of 35 location. The relationship between these vectors and spatial locations is most apparent in the case Of verb phrases. An object rising away from a reference point actually traverses the path of an infinitely extending vector, as does an object falling away from that point. An Object falling toward or rising toward the reference point traverses the paths of vectors which terminate at that point. If spatial locations which are extended are learned first, in both their static and dynamic forms, then vectors which increase extension must be less complex than Vectors which decrease extension. The Study of Spatial Terms The preceeding analysis has identified the follow— ing characteristics of vertical relationships: (a) The primary reference plane is the ground level, and any Object oriented with respect to this primary plane may serve as a secondary reference point for vertical rela- tionships; and (b) upward is the normative direction for the vertical dimension, and is more easily conceptualized than downward. The research presented in this section seeks to illustrate how these characteristics influence the acquisition of spatial terms. The study of spatial terms has been almost exclusively approached with componential feature analysis. Spatial terms lend themselves to such analysis, since they 36 are easily arranged in antonym pairs. This allows researchers to exert control over the specific feature differences under study-—each antonym pair represents a change in the coding of one or a very few semantic features. The identification of numerous cases in which one pair member is learned before the other has maintained this approach, since the source of such complexity differ— ences must lie in the different coding of semantic features. There has been a shift, though, from emphasis on a strictly linguistic basis for these differences to an appreciation of what these differences imply for per- ceptual and cognitive processes. The Lexical Marking Approach Application of semantic feature theory to the study of spatial terms may be traced to lexical marking analyses of antonymous adjectives. Bierwisch (1967), in his discussion Of German adjectivals, proposes that adjectives are grouped in antonymous pairs according to their eXpression of dimensional properties. For example, high and igy both refer to height, but they differ in the type Of height relationship they express. High refers to a location possessing height, while igg refers to a loca— tion lacking height. high is therefore coded as a positive polarity item (+height), and igg as a negative polarity item (-height). Furthermore, there is a 37 derivational relationship between the word high and the name for the dimension, height. This allows the word high to be used 'nominally,‘ without presupposing that the referent possesses or lacks the quality, as in asking: How high was his drive on the golf course? (His drive could have been high or igw,) When used nominally, the adjective high is unmarked with respect to polarity. According to H. Clark (1969a, 1969b), positive polarity features are linguistically less complex than negative polarity features. The nominal use Of high is least complex of all, since no polarity feature is present. Early studies of the acquisition of spatial antonyms (e.g., Donaldson and Wales, 1970; Wales and Campbell, 1970) attributed the acquisition of one pair member before the other to differences in polarity (e.g., H. Clark, 1970a). Problems With This Approach This analysis of strictly linguistic differences has been modified for several reasons. First, as dis- cussed previously, children appear to rely heavily on perceptual attributes as the basis for early word mean- ings. The use of a polarity contrast does not lend itself to an appreciation of this relationship. Second, differ- ences in the acquisition of members of antonym pairs, and in the processing of these pairs by adults, have been identified in form classes other than adjectives. 38 Intra-pair differences are present in adverbs (e.g., E. Clark, 1970), prepositions (e.g., H. Clark and Chase, 1972; Friedenberg, 1976; Friedenberg and Olson, 1977), and verbs (e.g., Clark and Garnica, 1974; Friedenberg, 1976; Friedenberg and Olson, 1977). In none of these cases is there a derivational relationship between the earlier acquired member and the scale name, a relationship which underlies the positive/negative polarity difference. Third, the class of spatial terms differs from the class of other dimensional terms in a fundamental way. Most dimensional pairs are contrastive pairs-— dichotomous polar adjectives (e.g., good/bad, happy/sad). Spatial (and temporal) terms are relative pairs, which may be Of any form class and for which the scale and its endpoints vary according to the context of application (Nelson and Benedict, 1974). Herein lies the main source of difficulty with the lexical marking approach: While it may be apprOpriate for contrastive pairs, where the positive polarity item shares a derivational relationship with the scale name, it may be inapprOpriate for relative pairs. Incorporating Perception and Cognition Even while lexical marking was the dominant model for studies of spatial terms, researchers were aware of the relationship among aspects of perception, cognition, 39 and semantics. H. Clark (1970a) noted that unmarked/ positive polarity items always refer to extension along a dimension, although there is no 3 priori reason why this should be so. In his account of the findings of Donaldson and Wales (1970), H. Clark proposed that children focus on situations exemplifying the 'greatest extent' of a dimensional quality, and use this as the feature to initially define both members Of an antonym pair (H. Clark, 1970a). Thus, children often confuse the two pairs members, using both to refer to the 'greatest extent' situation. This is particularly obvious in the case of "more” and "less" described earlier-~children often pass through a stage of using both to mean 'more' until they learn that 'extent' includes both greater and lesser amounts (e.g., Donaldson and Balfour, 1968). Extension is a perceivable attribute of spatial relationships, and the use of semantic features of this nature emphasizes the continuity between perception, cognition, and language. Extension as a feature may be coded positively or negatively, and fits directly into the framework of semantic feature theory. Predictions about which member of an antonym pair is acquired first are the same--high is still a positive polarity item,-but the source of this positive coding has been made more explicit. The advantage of using constructs such as extension is the ability to explain the acquisition of a 4O broader range Of lexical items. By describing the meaning of spatial words with attributes like extension and reference point, the expression of differences in the acquisition of varied lexical items can be more directly related to our perception and organization of space. The Neglected Issue: Between— Pair Differences and SimilafIties The issues just discussed apply only to the study of intra-pair differences. The comprehension of spatial terms is rarely approached from the perspective Of com- paring pairs for the same or closely related dimensions. Even the moSt exhaustive studies of children's compre- hension (e.g., E. Clark, 1972a; Donaldson and Wales, 1970; Wales and Campbell, 1970) have not attended to the order of learning words for different dimensional concepts. As mentioned previously, one Of the most interesting characteristics of language is that there are multiple descriptions available for any given Spatial relationship. Recent research by the author indicates that there are consistent patterns in the way children learn related words, and that these patterns are also reflected in how easily adults can process these terms. Friedenberg (1976) used the descriptions higher—lower than, above-below, and rising—falling away from to describe a plus and a star in a vertical relationship. Adults were better able to match 41 descriptions with an 'upward' perspective (higher, above, rising) to pictures than descriptions with a 'downward' perspective, regardless of the form class used. Children asked to construct two Object relationships in response to these words (Friedenberg and Olson, 1977) appear to learn the upward descriptions before the downward ones, again, regardless of grammatical class. These studies demon- strated that the 'normative' character of the upward vector influences the acquisition and processing of very different description types. However, these studies also indicated that adjectives were easier to learn and process than preposié tions, which in turn were easier than verb phrases. The generality of this finding is weak, though, since the verb phrases were applied as descriptions of static relationships. One reason for comparing adjectives and verb phrases in this dissertation research was to assess complexity differences between these types with appro- priate visual displays. General Research Hypotheses The present studies examine the roles Of extension and reference point in children's acquisition of spatial descriptions to determine if these variables affect the learning of different descriptions in the same manner. In addition, the research is designed to allow for 42 comparison of complexity differences in the acquisition of these items to theorized differences in our perception Of space. The normative role of vectors which originate at the reference plane, as compared to vectors which terminate at that plane, is examined by comparing children's comprehension of words for extended and unextended location. The normative role of the upward moving vector, as compared to the downward moving vector, is examined by comparing children's comprehension of words for locations above a reference point to words for loca- tions below that point. The roles of these variables are assessed in situations where located objects are sta- tionary, and in situations where located Objects actually traverse the paths of the vectors themselves. The use of these two situations also provides a means of comparing the acquisition Of adjectives and verb phrases, to deter- mine if one description type is generally acquired before the other. The following general hypotheses form the basis Of these studies (each will be restated specifically in each experiment, according to the lexical items under study): 1. The extension hypothesis: It is predicted that children will learn words or phrases for extended locations (far from the reference point) before learning words or phrases for unextended locations (near to the reference point). 43 The reference point hypothesis: It is predicted that children will learn words or phrases for locations above a reference point before learning words or phrases for analogous locations below that point. The description type hypothesis: It is predicted that children will learn adjectives, descriptions of static relationships, before verb phrases, descriptions of dynamic relationships. CHAPTER II EXPERIMENT 1 The purpose Of this experiment was to assess young children's comprehension of the adjectives: high, igy, gggp, and shallow, when used to describe the location of a stationary object relative to the ground (the primary reference plane for vertical relationships). This situa- tion differs somewhat from the use of such adjectives in their comparative forms (e.g., higher-lower than). In the latter case, the located Object may be infinitely far from the reference point when it is either above or below that point. In the present experiment, such "infinite" extension is only possible for high and gggp locations; since the reference point is the ground level, igh and shallow are restricted in extension and border on the reference plane itself. According to the extension hypothesis, high and dggp_should be learned before igw and shallow. The lin- guistic basis for this prediction is derived from semantic feature theory (e.g., E. Clark, 1972a; H. Clark, 1970a). high and dggp are positive polarity words, as may be inferred from their derivational relationships 45 with their scale names (i.e., height and ggpih). ESE and shallow are negative polarity words, indicating the absence of height and depth respectively. According to semantic feature theory, positively coded words are acquired before their negatively coded counterparts. The modification of feature theory to represent this polarity difference as a difference in extension (e.g., E. Clark, 1974) predicts the same pattern. Children are initially more sensitive to the "greatest extent" Of any given dimension, and therefore learn words which encode exten- sion before they learn words which encode the lack of extension. However, there is an additional dimension differ- entiating these words. high and igg refer to locations above a reference point, while gggp and shallow refer to locations below a reference point. According to the reference point hypothesis, high and igg should be acquired before h2g2 and shallow. There is a linguistic basis for this prediction, again drawn from semantic feature theory. According to H. Clark (1973), gggp and shallow are linguistically more complex than high and igh because of syncretization. Natural languages often differentiate antonym pairs on the basis of different types of usage. For example, when high-low is compared to tall-short, a difference in the type Of vertical information encoded emerges. High-low 46 refers to the position of objects within the vertical dimension, while tall-short refers to how far along the vertical dimension Objects extend. High—low is a posi— tional pair, and tall-short an extensional pair. However, the vocabulary of English for analogous relationships below a reference point is lacking. Deep-shallow must serve both positional and extensional uses. This syncre- tization, or combination of two senses within one pair, increases the linguistic complexity of deep-shallow. As a result, deep-shallow should be acquired later than high- igh. Syncretization is a rather obscure point in lin- guistic analysis, and a weak point on which to base the reference point hypothesis. Nevertheless, there are cognitive-perceptual reasons why deep-shallow might be acquired later. In his analysis of reference directions for the vertical dimension, H. Clark (1973) proposed that 'upward' from the reference point is normative. It is the most salient vertical direction, it is most frequently experienced, and therefore it is most readily conceptual- ized. If children are more likely to have concepts about spatial relationships above a reference point than con- cepts about relationships below that point, it should be easier for them to map the items high-low onto spatial concepts than it is to map the items deep-shallow. 47 In summary, the following specific hypotheses regarding children's comprehension of the four adjectives are proposed: (1) Older children will make fewer errors on all target words than younger children, since they have more perceptual experience with spatial relationships, better conceptual development relative to spatial relationships (and, in general), and therefore should show evidence of superior language comprehension. (2) The extension hypothesis is restated as the prediction that all children should generally make fewer errors on high/deep than on low/shallow since the former pair refers to extended locations. Extended locations have been shown to be perceptually more salient, easier to conceptualize, and represented by linguistically less complex words. However, the magnitude of this difference should decrease with age as children learn the meanings of the more difficult items. (3) The reference point hypothesis is restated as the prediction that all children should generally make fewer errors on high-low than on deep—shallow, since the former pair refer to locations above the reference point. Locations above the reference point have also been shown to be perceptually more salient, easier to conceptualize, and indicated by linguistically less complex forms. The 48 magnitude of this difference should also decrease with age as children learn the meanings of the more complex forms. (Children should show one of the following pat- terns of decreasing error rates: high, igh, gggp, shallow; or high, deep, low, shallow. Both patterns indi— cate better comprehension of words for extended locations and of words for locations above a reference point; i.e., these orders represent confirmation of hypotheses 2 and BJ Method Subjects Preschool children from the MSAU Day Care Center (Michigan State University) and the Eastminster Day Care Center (East Lansing) were recruited for the study. Each family received a Parent Letter describing the project (see Appendix A) and returned a Signed permission card prior to testing. Twenty-one children in each of three age groups participated. All were attending a day care program at least twice a week, and were from families of lower to upper middle class. The breakdown of subjects according to age is presented in Table 1. Each family received a copy of the group results at the conclusion of testing. Individual results were treated confidentially and with anonymity. 49 Table 1 Characteristics of Subjects in Experiment 1 Group h X Age Age Range 3 yr olds 21 36.1 mos 31-42 mos males 9 35.8 mos 31-42 mos females 12 36.3 mos 31—42 mos 4 yr olds 21 49.1 mos 43—54 mos males 10 49.2 mos 44-54 mos females 11 49.0 mos 43-53 mos 5 yr olds 21 60.1 mos 55-68 mos males 9 58.4 mos 55-65 mos females 12 61.4 mos 55-68 mos Materials Testing materials included a felt board and felt cut-outs. The 42.5 cm by 55 cm felt board was divided into four equal sections. The two upper quadrants were covered with light blue felt to represent the sky; the two lower quadrants were covered with dark green felt to represent the ground. A 20 cm long brown tree (three branches, no leaves) and a 20 cm long brown hole were placed on the board as locations for squirrels. The base of the tree touched the midsection of the board, as did the top edge of the hole. However, each object was located slightly to one side of the center of the board so that the tree was in either the upper left or upper right quadrant, the hole in the opposite lower quadrant. Two white felt squirrels, each 8.75 cm long, were placed on 50 the board for each trial. The high squirrel sat on the top branch of the tree, the igh_squirrel at the base of the tree (bordering the midsection), the hggp squirrel at the bottom of the hole, and the shallow squirrel adjacent to midsection in the upper part of the hole. Sentences used to describe the locations of squirrels included one of the target items (high, low, deep, shallow) in the frame: Show me the squirrel in a place. Design The 3 x 2 x 2 x 3 design involved three age groups, two types of extension (extended ==high, deep; unextended = low, shallow), two types of reference point (above reference point = high, igh; below reference point== h3g2, shallow), and three pairing conditions (consisting of three different displays used to test comprehension of each target word). (Two presentations of each pairing condition for left/right position counterbalancing were used--correct location on the right side and on the left side of the felt board.) A summary of the major features of each pairing condition is presented in Table 2. Each child saw a total of twenty-four displays consisting of two presentations of the three pairing conditions for each target item. 51 Table 2 Pairing COnditions Used in Experiment 1 . . . Independent Vari— Board Cond1t10n D1splay able Contrast Positions 1 high squirrel & reference point opposite sides deep squirrel (one in tree, -Or- one in hole) low squirrel & shallow squirrel 2 high squirrel & extension same side low squirrel (both in tree) deep squirrel & shallow squirrel (both in hole) 3 high squirrel & extension & Opposite sides shallow squirrel reference point (one in tree, —or— one in hole) low squirrel & deep squirrel Procedure Each child was tested individually by either the principal experimenter or an undergraduate research assistant. All testing was performed in a separate room at the day care center, and generally lasted 15 minutes. The child was seated next to the experimenter, facing the upright felt board. Different chairs were available to insure that the child's eye level was approx- imately at the midsection of the board. The cut-outs were presented, and the relationships among the various parts were discussed. The blue Sky and green grass were pointed out, and the child was asked to pretend this was a picture 52 of a sunny day outside. The tree and the hole were placed appropriately on the board, one on each side, and the child was encouraged to describe where a squirrel could sit in this scene. The experimenter verified that the child understood the tree to be standing above the ground and the hole to have been dug below the ground. Testing began by locating two squirrels on the board and proceeded according to a predetermined order. All words in all pairing conditions were first tested with one left/right orientation Of the tree and hole, and then with their 1eft’ right positions reversed. Within this constraint the order Of testing words was random. The initial left/right positions of the tree and hole were also randomized. On each trial the child was asked to "Show" the experimenter a particular squirrel by pointing to it. Both squirrels were then removed from the board and placed in two new locations simultaneously. Errors in pointing responses were recorded as the dependent measure. Results and Discussion Analysis of Variance (ANOVA) A four-way mixed analysis of variance was per- formed on the mean error rate data for the variables of age, reference point, extension, and pairing condition. The following main effects and interactions reached at least an .05 level Of significance: 53 (1) the main effect of age (h(2,60)=11.32, p1.0005), with Older children making fewer errors overall than younger children; (2) the main effect Of reference point (h(l,60)= 7.99, p1.006), with children making fewer errors overall on words for locations above the reference point (high- igg) than on words for locations below the reference point (deep-shallow); (3)-the main effect of extension (h(l,60)=62.86, p1.0005), with children making fewer errors overall on words for extended locations (high/deep) than on words for unextended locations (low/Shallow)? (4) the main effect Of pairing condition (h(2,120)= 5.69, p1.004), with children making fewer errors overall on pairing condition 2 (squirrel in an extended location, high or h3g2, paired with squirrel in an unextended loca- tion, igh or shallow) than on either of the other two pairing conditions: (5) the interaction of extension with pairing con- dition (h(2,120)=4.37, p1.015), in which children's per- formance on the three pairing conditions was quite similar for the extended words (high/deep), but quite different for the unextended words (low/shallow); (6) the interaction of reference point with pair- ing condition (h(2,120)=15.77, p1.005), in which perfor- mance on the three pairing conditions varied according to 54 whether the target word indicated an above- or below- reference point location; and (7) the triple interaction of extension, reference point, and pairing condition (h(2,120)=21.9, PA.0005). Appendix B presents a table of the raw data, and the summary tables from the overall ANOVA and the series of simple main effects ANOVAS performed on the data from the three significant interactions. The main effect of age can simply be explained as evidence of improvement in comprehension across ages. The youngest children averaged 7.87errors the 4—year—Olds 4.66, and 5-year—olds 3.8 errors overall. However, the other three significant main effects are involved in both two-way and three-way inter- actions, and must be explained in light of these inter- actions. The purpose of the simple main effects ANOVAS was to determine which error rates within these inter- actions were actually significantly different. It is the triple interaction which most clearly reveals the structure Of the data (see Figure 2). Dif— ferences due to the extension variable were most pervas- ive. Children performed significantly better on high_than igg in pairing conditions 1 and 3, and performed better on gggp than shallow in conditions 1 and 2. This produced a significant interaction Of extension and pairing condi- tions in both the overall ANOVA and the simple main effects ANOVAS. The interaction is most clearly seen by comparing TOTAL ERRORS TOTAL ERRORS 55 60 50 40* 30 20 10 T V high [extended] low [unextended] ABOVE-REFERENCE POINT 60 5O 4O 30 20 1O p I. / / 2 / I , P1 f" / /’ I ’ I / I I ’ P3 - / , , ’ I P1- — -.contrast in reference point _ / P2_ ._ contrast in extension - p3 contrast in both reference point and extension 1 1 deep [extended] shallow [unextended] BELOW-REFERENCE POINT 56 the lines labeled P1, P2, and P3 in Figure 2. The two lines labeled Pl show almost identical slopes, indicating that the difference between children's error rates on high— low was about the same as on deep-shallow. The lines labeled P2 and P3 each display one significant and one non-significant difference. Children made fewer errors on high than on igh in P3, but made about the same number of errors on gggp and shallow. The pattern is reversed in P2: Children made fewer errors on h§§p_than on shallow, but made about the same number of errors on high and igg. The similarity in performance for P1 is confirmed by the fact that the interaction of reference point and extension was significant for pairing conditions 2 and 3, but not for condition 1. No predictions were made about differences among the pairing conditions, yet they Obviously affected chil- dren's performance. The ANOVAS indicated that children performed equally well on all pairing conditions for high and gggp, the extended words. This cannot be attributed to floor effects, since children did make errors on both words in all conditions. However, pairing conditions did affect children's performance on igy and shallow. Pairing condition 1 was of intermediate difficulty for both items, but the difficulty of P2 and P3 was reversed. Tukey tests for g posteriori comparisons (Winer, 1971) indicated that for low, the conditions were ordered: P2 1 Pl 1 P3 (P1 v 57 P2, q=24.1; Pl v P3, q=14.2; P2 v P3, q=38.3; =2.8 for q.05 each), and for shallow, the conditions were ordered: P3 1 P1 4 P2 (P1 v P2, q=4.96; Pl v P3, q=7.8; P2 v P3, q=12.76; q 05:2.8 for each). Hypotheses as to the source Of these differences, which also underlie the differences attributable to extension, will be offered presently. Effects due to reference point were scarce. Chil- dren did not differ in performance on high and gggp, the extended items for the two types of reference point loca- tions. Error rates on igh_and shallow did differ signifi- cantly, but only in pairing conditions 2 and 3. This produced a significant interaction of reference point and pairing conditions, which expresses the same contrast as the interaction of reference point and extension: Children made fewer errors on 19E than on shallow in pairing condi- tion 2, but made more errors on igh than on shallow in pairing condition 3. This is also clearly seen in Figure 2. Analysis of Position Biases One reason for using the paired—comparison pro- cedure was to allow for inferences about developing comprehension from the types of errors children made. For example, E. Clark (1972a) found it most common for children to confuse words for the two opposing ends of a dimension, while Brewer and Stone (1975) found it most common for children to confuse words of the same extension 58 but of different dimensions. Before any such inferences can be made, the data must be examined for simple position biases which reflect nothing about developing comprehen- sion. There are several position biases which could be present in the data. Children could always choose the object on the left side of the board, on the right side of the board, on the top section Of the board, or on the bot- tom section Of the board. When children were classified according to these patterns, almost 85% in each age group showed a very mixed pattern of errors. About 5% made errors on the right side only, about 7% on the left side only, and about 3% made no errors at all. Thus, it is unlikely that any of the significant error rate differences reflect nonlinguistic position biases. Age Differences and Age Changes When possible, developmental studies should be examined for evidence of age changes and age differences. This experiment is unusual as a language comprehension experiment because no significant age changes occurred. Although older children were more accurate at identifying all locations than younger children, they showed the same relative error patterns. It is common in the study of antonym acquisition to find that as the age of the chil- dren increases, the difference between their error rates on extended-unextended items decreases (e.g., E. Clark, 1972a; Donaldson & Wales, 1970; Friedenberg & Olson, 1977). 59 The ANOVAs indicated that the relative difficulty of extended-unextended words, and of above-below reference point words, remained the same across age groups in this experiment. Differences Attributable to Extension Although the main effect of extension reached sig- nificance in the overall ANOVA, the expression of errors on igh_and shallow was modulated by pairing condition. Children made fewer errors on high than on igg in condi- tions 1 and 3, but made the same number of errors on the two words in condition 2. Comprehension of gggp waslmatter than shallow in conditions 1 and 2, but performance in condition 3 was equivalent. In each case, two conditions support the extension hypothesis, and one condition does not. By examining the types of errors made by children in each age group, it may be possible to determine which pat- terns actually reflect the status of comprehension of igg and shallow. The equivalent error rates on high and igh in condition 2 may be attributed to a nonlinguistic elimina- tion strategy. This condition presented 12E and high squirrels (the extension contrast), and children could improve their performance on 12H sentences by eliminating the unnamed high squirrel as the "wrong” choice. This hypothesis is supported by the fact that children in all three age groups showed almost identical error rates on 60 high and on igh in this condition (see Appendix Bl). Thus, condition 2 cannot be used to disconfirm the extension hypothesis regarding acquisition of high and igg. This raises the question of whether or not this elimination strategy was also employed on trials where gggp or shallow was paired with high. hggp and high were paired in condition 1 (the reference point contrast). The youngest children made many errors on hggp in all three pairing conditions, while the two Older groups made prac— tically no errors on gggp in all three conditions. There is no evidence that any elimination of the high squirrel as the "wrong" choice occurred. However, such a strategy would only be needed by the youngest children: The two older groups performed as well on gggp as they did on high. It is unclear as to why the youngest children did not employ the high elimination strategy with hggp as they did with igh. Shallow and high were paired in condition 3 (extension and reference point contrast). This condition is also of interest because it represents a disconfirma- tion of the extension hypothesis: Children performed equivalently on h3g3 and shallow in condition 3. Close examination of the data in Appendix Bl indicates that the age groups differed in performance on this condition, although the differences were not large enough to reach significance in the ANOVAS. The youngest children 61 performed equally poorly on gggp and shallow, the 4-year- olds performed better on gggp_than on Shallow, and the 5-year-Olds performed equally well on the two words. It is not surprising for the 3-year-olds to perform so poorly on both words--they are below-reference point words which theoretically are more difficult to learn. The fact that performance improves on hggp in all conditions, but improves on shallow only in condition 3, lends support to the prediction that gggp is learned before shallow. Chil— dren could improve their performance on shallow in condi- tion 3 by eliminating the high squirrel as the "wrong" choice. Since the Oldest children were only good on shallow sentences in condition 3, it is unlikely that their performance reflects accurate comprehension of the word shallOw. The question of why neither of the younger groups employed the elimination strategy cannot be answered from the data. However, the data do indicate that condition 3 should not be taken as strong disconfir- mation of the prediction that hggp is learned before shallow. Thus far, the extension hypothesis generally is confirmed by data on high-low and deep-shallow which does not reflect a nonlinguistic strategy. However, there are additional facets of the experimental design which bear on support for differences attributable to extension. It is curious that children always made many errors on low 62 sentences in condition 3. This condition presented igg and hggp squirrels, and was always the most difficult condition for igh sentences. It is not necessary to examine these error rates relative to errors in condition 2 because the latter condition is likely to reflect a non— linguistic elimination strategy. However, performance on low in condition 1 (low—shallow squirrels) was always better than performance in condition 3. Close examination of the data by subjects indicated that, of the 75% of children who made at least one error in this experiment, 29% made one error on the ighfgggp pairing, and 33% made two errors on this pairing. (Surprisingly enough, slightly more children in the Oldest group made two errors on this pairing than in either of the two other groups.) Why was this condition so much more difficult than condition 1? Consider the perceptual display in condition 3. One squirrel is at the base of the tree (near the "ground"), while the other is at the base of the felt board (in the bottom of the hole). If children momentar- ily forgot that the midsection Of the board represented the ground, they could judge "lowness" in an absolute sense as distance from the top of the board. This would allow for choosing the gggp squirrel in response to igh sentences. Thus, the ambiguity of the display used in condition 3 could have resulted in an inflated error rate 63 on 12! sentences, producing a disparity between perfor- mance on conditions 1 and 3. To summarize, the following account Of differences attributable to extension is Offered. For higheigh, condition 2 did not disconfirm the extension hypothesis, because children's performance probably did not reflect accurate comprehension of the word igg. It is more likely that performance reflected the elimination of the high squirrel as the "wrong" choice. This was supported by the fact that children did make errors in the other two conditions, while hardly ever making errors in condition 2. Condition 3 was probably the most difficult because the perceptual display was ambiguous. Without this ambiguity, performance might equal performance on condition 1. Nonetheless, even allowing for inflation Of the error rate in condition 3, children made more errors on igg than on high, as they did in condition 1. The data generally support the prediction of the extension hypothesis. For deep—shallow, condition 3 did not fully dis- confirm the extension hypothesis for two reasons. First, children did make more errors on Shallow than on gggp once performance on gggp improved (between 3 and 4 years). Also, the dramatic improvement on shallow sentences in condition 3 at 5 years probably reflected the same elimina- tion of the high squirrel previously discussed. Taken with children's superior performance on deep in the other 64 two conditions, the data generally support the extension hypothesis. Differences Attributable to Reference Point The only support from the ANOVAs that children learned above-reference point words before below-reference point words involves igg and shallow. Children performed better on igh than on shallow in condition 2, as predicted by the reference point hypothesis. However, it has been proposed that the error rate on low in condition 2 was depressed via elimination of the unnamed high squirrel as the "wrong" choice. It cannot be concluded, then, that the significant difference supports the reference point hypothesis. In condition 3, children made fewer errors on shallow than on igh, in direct contradiction of the hypothesis. However, it has been suggested that chil- dren's performance on igh in condition 3 was hampered by an ambiguous perceptual display. Thus, this difference also cannot be used to address the reference point hypothesis. It is only from condition 1, then, that perfor- mance on igh and shallow reflected the status of compre- hension. Although fewer errors were made on 12H sentences in this condition, the difference was not large enough to reach significance in the ANOVAs. The trend is in the predicted direction, but is not robust. 65 There is another nonsignificant trend in the data supporting the reference point hypothesis. At 3 years, children performed significantly worse on gggp than on high (see Appendix Bl), although the performance of the two Older groups was the same on the two words. It is possible to derive some support for this reference point difference in another way. The preceding pages use children's errors to support the experimental hypotheses. However, the data may be approached differently. If high is acquired before gggp, for example, one would expect expect children to first score perfectly on high, and then score perfectly on gggp. More children should Show this pattern than the reverse pattern——scoring correctly on hggp before scoring correctly on high. Table 3 presents conditional propor- tions of children scoring correctly on each word before scoring correctly on each other word. The entries indi- cate that .875 of the children who were correct on h2g2 were also correct on high, while .583 who were correct on high were also correct on gggp. From these asymmetries in the conditional probabilities one can infer an acquisi- tion order. Careful examination of Table 3 indicates that the most common order of scoring correctly on all four words was: high, deep, low, shallow. This order supports both the extension and reference point hypotheses. 66 Table 3 Conditional Proportions in Experiment 1 Word Acquired Word Acquired Second First High Deep Low Shallow High ... .875 .938 .857 Deep .583 ... .625 .857 Low .313 .313 ... .214 Shallow .250 .375 .188 ... Note: Each entry represents the prcportion of all children making no errors on the row item, given that none were made on the column item. Differences Attributable to Pairing Conditions While no predictions were made about differences due to pairing conditions, children did differ in their error rates on ABS and shallow in conditions 2 and 3. Condition 1, in which locations were contrasted in refer- ence point (low-shallow comparison), was consistently intermediate in difficulty. This type of error, confusing words of the same extension but of different dimensional sections, is the same type of error identified as most common by Brewer & Stone (1975). It is not the form of error characteristic of feature theory acquisition (i.e., confusing words for the two opposing ends of a dimension). Condition 2, the extensional contrast, represents the typical feature theory type of error. The Tukey tests previously presented indicated that this type of error was 67 most frequent for shallow sentences, and least frequent for igg sentences. However, the dearth of errors on igg in condition 2 has been attributed to a nonlinguistic strategy, and errors in this situation should not be included in discussions of confusion among words. Condi- tion 3, which presented both an extensional and a reference point contrast, was identified by Tukey tests as the most difficult condition for igh sentences, and the easiest for shallow sentences. This condition represents neither of the error types usually identified in antonym research. Since both variables are contrasted, children needed only to identify one dimension of difference on which to make their locational selection. Thus, it is not unreasonable for this condition to be the easiest one. The fact that condition 3 was identified as the most difficult condition for igh sentences may be attributed to the perceptual con- fusion previously proposed. CHAPTER III EXPERIMENT 2 The preceding experiment demonstrated how exten- sion and reference point influenced the learning of descriptions of static vertical relationships. The pur- pose Of the present experiment was to assess the roles of these variables in the learning of phrases for vertical motion in space. The verb phrases were chosen so as to maximize the similarity between the encoding of these variables in the different lexical items: Rising away from was used as an extended above-reference point phrase, analogous to high; falling toward was used as the unex- tended above-reference point phrase, related to low; falling away from, similar in character to deep, was used for the extended below-reference point location; and rising toward, like shallow, represented the unextended below-reference point location. If these variables represent dimensions with which children organize spatial phenomena, they should influence the acquisition of these verb phrases in the same way as they influenced the learning Of the adjectives in the first experiment. Accordingly, the extension hypothesis predicted that the two away from phrases, representing extended locations, would be acquired before the two toward phrases, representing unextended locations. The 68 69 reference point hypothesis predicted that the two above- reference point phrases, rising away from-falling toward, would be acquired before the two below-reference point phrases, falling away from-rising toward. On the basis of the results of the first experiment, differences attrib- utable to extension should be more pervasive than differ— ences attributable to reference point. Since effects due to pairing condition were erratic in the first experiment, no predictions regarding this variable were made in the present experiment. There are several reasons why these experiments as a unit are of interest. First, in the demonstration of similar functioning of extension and reference point in the two cases, there is support for the notion that the semantics of the verb phrases must be analyzed by consid— ering the phrases as units. The alternative would be to regard them as composed Of the lexical items rise, fall, away—from (which could be regarded as a unit analogous to "from"),and toward. Second, there is support for H. Clark's (1973) proposals about complexity differences in reference directions. The objects depicting the verb phrases actually traverse the paths of Clark's vertical reference directions, while the objects depicting the adjectives rest at locations along these paths. If in both cases, the order of acquisition Of these items paral- lels the theoretical complexity differences in reference 70 directions, there can be little doubt as to the influence of these normative vectors in the learning of spatial phenomena. Third, there is the opportunity to determine whether similarities among the semantic structure compon- ents of the adjectives and verb phrases are more important in language acquisition than the differences attributable to form class. Children from the two experiments may be compared to determine whether the adjectives and phrases are learned around the same time, or whether one type is generally learned before the other. Fourth, the two experiments present slightly dif- ferent types Of reference points, although both are examples of common reference points for vertical relation— ships. The first experiment presented a pictorially represented "ground", a primary reference plane. The second experiment employed an airplane as the reference Object, which is Obviously a secondary reference point. The experiments can therefore determine if this variable also influences the acquisition of these descriptions. It is difficult to determine the linguistic basis for the predictions of either experimental hypothesis. The words are not adjectives, and cannot be approached from a lexical marking standpoint. Furthermore, semantic feature theory has not addressed the question of phrase learning. Assessments of the learning of phrase 71 components are of limited value since there is no real mechanism for determining how to combine the information from the components to describe the phrases. However, some inferences about the pair rising away from-fallihg toward may be made from previous assessments of the simple verbs rise-fall. When a speaker refers to an object as rising,it is assumed that the object is rising away from the ground. As the primary reference plane for vertical motion, the ground level functions as the 'default' value when no reference point is specified. Likewise, when a speaker refers to an Object as falling, it is assumed that the Object is falling toward the ground. In theorizing about the complexity of the verbs rise-fall, it is common to propose that ii§§_is less complex and a more likely candidate for early acquisition (e.g., H. Clark, 1973; Miller, 1972; Miller & Johnson-Laird, 1976). This is because high refers to upward motion, which is positively coded as a semantic feature. If the phrases rising away from-falling toward are viewed as explicit forms of the simple verbs rise-fall, the early acquisition of the extended away from phrase could be predicted from this analysis. It is far more appropriate, though, to derive the extension hypothesis from cognitive-perceptual concerns. The motions described by these four phrases comprise the four vectors characteristically associated with the 72 vertical dimension (e.g., H. Clark, 1973). The away from phrases represent the vectors which originate at the reference plane, while the toward phrases represent the two vectors which terminate at that plane. As described in Chapter I, the vectors which originate at the plane underlie extended spatial locations, while the vectors which terminate at the plane underlie unextended locations. If children are initially more sensitive to the presence of extension than to its absence, as was indicated in the first experiment also, then phrases for extended locations should be acquired first. It is even more difficult to determine the lin— guistic basis for the reference point hypothesis. The phrases falling away from-rising toward do not have any simple relationship to the verbs rise-fall, and for this reason one could predict that they would be acquired later than rising away from-falling toward (the explicit forms Of the verbs rise-fall). Our only other choice for a linguistic analysis is to decompose the phrases into individual lexical items. A brief example demonstrates the fruitlessness of this approach. If high is coded positively, and iaii is coded negatively (as previously suggested), then each pair of above- and below-reference point phrases contains one positive and one negative phrase component. All previous analyses of away from and toward have dealt with horizontal motion, an Obviously 73 different situation, and are inappropriate for our present purpose. Suppose that away from is coded as positive because it is extended, and toward as negative because it is unextended. This would support earlier acquisition Of rising away from than of falling away from-~the former is twice positively coded, the latter contains one positive and one negative coding. Yet the system breaks down for the other two phrases. Falling toward is twice negatively coded, but is predicted to be acquired before rising toward, which contains one positive and one negative coding. Obviously, this type of decomposition is not the answer. However, attention to cognitive-perceptual concerns may again be used to derive the reference point hypothesis. Of the two vectors which originate at the reference plane, the one extending upward is viewed as normative (H. Clark, 1973). Locations along this vector (e.g., rising away from-falling toward) could therefore be learned earlier than locations along the downward vector (e.g., falling away from-rising toward). A third prediction was made in this experiment: That children would learn these phrases after learning the adjectives. There are a variety of reasons why this should occur. Simple adjective forms typically appear in children's vocabularies quite early, often among the first fifty words (e.g., Nelson, 1973). Phrases using the present progressive tense (-ing form) typically emerge 74 during the stage Of two-word utterances, when the child's average sentence length (MLU) is about 2.3 units (Cazden & Brown, 1976). However, differences in the comprehension of these types have not been examined. And if, as Nelson suggests (1974), children are sensitive to dynamic prop- erties of the world around them, it would be quite reasonable to discover that they learn the verb phrases at about the same time as the adjectives. In summary, the following specific hypotheses regarding children's comprehension of the four verb phrases are proposed: (1) Older children should make fewer errors over— all than younger children On all phrases, since they have more perceptual experience with these relationships, and are conceptually and linguistically more advanced. (2) The extension hypothesis is restated to pro- pose that all children should make fewer errors on the away from phrases than on the toward phrases, since the former refer to extended locations. If extension is a salient organizing principle in the acquisition of nominal adjectives, it is logical to infer that this same scheme will be employed in the learning of the verb phrases. This difference should diminish with age. (3) The reference point hypothesis is restated to propose that all children should make fewer errors over- all On rising away from-falling toward than on falling away 75 from-rising toward, since the former pair refers to loca- tions above a reference point. Although reference point did not emerge as a strong organizational scheme in the learning Of the nominal adjectives, differences attribut- able to reference point are present in that data. This difference should also diminish with age. (Children should show one of the following patterns of decreasing errors: rising away from, falling toward, falling away from, rising toward; or rising away from, fallinggaway from, falling toward, rising toward, if hypotheses 2 and 3 are confirmed.) (4) Children of the same age participating in the two experiments will demonstrate better comprehension of the adjectives than the verb phrases. These children will be compared as to their error rates on the two types of lexical items in each of the four locational categories: extended above—reference point, extended below—reference point, unextended above-reference point, and unextended below-reference point. Method Subjects Preschool children from the Laboratory Preschool (Michigan State University) and kindergarten children from the Sycamore School (Holt, MI) were recruited for the study. A copy of the Parent Letter sent to each 76 participating family is included in Appendix C. A signed permission card was obtained for each of the twenty chil- dren in each age group. Families of children tested ranged from lower to upper middle class. The breakdown Of subjects according to sex and age is presented in Table 4. Table 4 Characteristics of Subjects in Experiment 2 Group h X Age Age Range 4 yr olds 20 47.3 mos 42-51 mos males 11 47.3 mos 44-51 mos females 9 47.3 mos 42-51 mos 5 yr olds 20 55.9 mos 52-61 mos males 8 55.0 mos 52-59 mos females 12 56.6 mos 52-61 mos kindergartena 20 males females aAge information was not Obtained for these children. Each family received a copy Of the group results at the conclusion of testing. Individual results were kept both confidential and anonymous. Materials Testing materials included a motorized apparatus and a set Of colored plastic airplanes. The apparatus 77 measured 54 cm tall by 48.75 cm wide by 31 cm deep. The bottom 8.75 cm consisted of an enclosed base housing the motors and pulleys. Attached tO this base were two vertical posts, each 45 cm tall by 3.75 cm wide by 2 cm deep, placed about 16 cm apart. Each pole contained three protruding metal rods, each 2.5 cm long. One rod on each post was stationary, located at midsection (22.5 cm from the base). The rod above and the rod below midsec— tion on each post was attached to a motorized pulley. By pushing a toggle switch in the appropriate direction the mobile rod would move up or down the post within its 22.5 cm pathway. With four separate motors and pulleys, each mobile rod could be operated independently. All motors ran at the same constant speed. Several sets of colored airplanes were used, with the constraint that the airplanes at midsection of both posts were of the same color (e.g., blue) and the airplanes attached to mobile rods of both posts were a different color (e.g., green). By using biplanes, airplanes could be attached to the rods by inserting the rod between the top and bottom wings of each airplane. Sentences used to describe the relationships used one of the target phrases rising away_from, falling toward, falling away from, rising toward) in the frame: The green plane is the blue plane. Design The variables used in the 3 x 2 x 2 x 3 design were identical to those used in the first experiment: three age groups, two types of extension (extended = rising-falling away from; unextended = rising-falling toward); two types Of reference point (above—reference point = rising away from-falling toward; below-reference point = falling away from-rising toward); and three pairing conditions. (Two presentations of each pairing condition for position counter- balancing, as in Experiment 1, were used again.) Each child saw 24 displays consisting of two presentations Of each pairing condition for each target item. The displays used in each pairing condition, and the contrasts they presented to children, are presented in Table 5. Table 5 Pairing Conditions Used in Experiment 2 ,. . . Independent Var- Directional Cond1t1on . Display 1able Contrast Contrast 1 Rising away from & Reference Rising v, falling away from point falling falling toward & rising toward 2 Rising away from & Extension Away from v. falling toward toward & -or— rising v, falling away from&: falling rising toward 3 Rising away from & Extension & Away from v. rising toward reference toward -or— point falling away from& falling toward 79 Procedure Each child was tested individually by the principal experimenter in a 15—minute session in a separate room at the day care center or school. The apparatus was placed on a table between the child and experimenter. Different chairs were used to insure that the child's eye level was approximately at the midsection of the posts. While fac- ing the apparatus, the child was shown how the rods could be moved up and down, and was asked to name the colors of the airplanes to be used. A game was suggested in which the experimenter would "name" (describe) one of the moving airplanes, and the child would identify it by pointing to the proper post. One color of airplanes was chosen for the stationary rods, another for the mobile rods. The stationary airplanes were attached, and testing began with the first moving relationships. In order for the moving airplanes to be able to travel as far as possible, the rods were appropriately oriented between trials, before their airplanes were attached. Thus, each trial began with only the stationary airplanes visible. The mobile planes were attached, and the sentence was pronounced as the planes began to move. Planes were moved simultaneously by the experimenter hitting two switches. The order of testing all pairings was randomized for all children. 80 Results and Discussion Analysis of Variance (ANOVA) A four-way mixed analysis of variance was per- formed on the mean error rate data for the variables of age, reference point, extension, and pairing condition. The following main effects and interactions were found to be significant: (1) the main effect of age (h(2,57)=13.56, p1.0005), with Older children making fewer errors overall than younger children; (2) the main effect of extension (h(l,57)=37.44, PL.0005), with children making fewer errors overall on phrases for extended locations (away from phrases) than on phrases for unextended locations (toward phrases); (3) the main effect Of pairing condition (h(2,1l4)= 17.82, p4.0005), with children making the fewest errors overall on pairing condition 2 (airplane in an extended location, rising away from or falling away from, paired with an airplane in an unextended location, falling toward or rising toward) than on the other two pairing conditions; (4) the interaction of age with extension (h(2,57)= 5.47, p4.007), indicating that children's comprehension of the phrases for unextended locations (toward phrases) improved across ages; (5) the interaction of extension with pairing con- dition (h(2,114)=6.98, p4.001), in which the difference 81 between performance on extended (away from) and unextended (toward) items varied according to pairing condition; and (6) the triple interaction of extension, pairing condition, and age (h(4,ll4)=2.56, p1.042), revealing that the difference between performance on extended (away from) and unextended (toward) items varied according to both pairing condition and age. The ANOVA Summary Table describing these results is presented in Appendix D, along with the results of three simple main effects ANOVAS and a table of the raw data. Since each of the significant main effects is involved in at least two interactions, all must be explained in the context of those interactions. Again, it is the triple interaction which provides the most insight to the structure Of the data (see Figure 3). The three age groups differed significantly on away from sentences only in pairing condition 1 (simple main effects ANOVA). Tukey tests indicated that the 4-year- olds performed worst, followed by the 5-year-olds, with the kindergarten children performing best (4 v. 5 years: =2.81 for each). q=2.88; 4 v. K: q=6.9; 5 v. K: q=4.03; q.05 The age groups also differed significantly on toward sentences in all three conditions, although the pattern of age differences identified via Tukey tests varied: (1) In pairing condition 1 all three age groups differed significantly, with 4—year-Olds performing worst TOTAL ERRORS TOTAL ERRORS 82 40 30 1 20- o l 1 away from toward [extended] [unextended] P1 40 30 y- ’,4 ’I I 20 - ,I' l I” /5 10 P I, / , / , / ’/ / 1.... away from toward P3 , , _=:_——=-“‘ .fi away from toward P2 P1: rising-falling discrimination __ both toward-away from and ' P2- rising—falling varying P3: toward-away from discrimination 4""'4year olds 5——5 year olds K—‘Kindergarteners 83 and kindergarteners performing best (4 v. 5: q=4.03; 4 v. K: q=6.9; 5 v. K: q=2.88; =2.81 for each). q.05 (2) In pairing condition 2 the 4-year-Olds per- formed significantly worse than both of the other groups (4 v. 5/K: q=8.63; =2.81). (1.05 (3) In pairing condition 3 all three age groups differed significantly, with 4-year-Olds making the most errors and kindergarteners making the least errors (4 v. 5: q=10.35; 4 V. K: q=15.53; 5 v. K: q=5.18; q 05:2.81 for each). The difference between error rates on away from (extended) and toward (unextended) sentences varied according to age and pairing condition. In condition 1, children in all three age groups made fewer errors on away from than on toward. In conditions 2 and 3, only the 4-year-Olds made fewer errors on the away from sentences. NO significant differences attributable to refer— ence point occurred in the data. However, there were again some differences due to pairing conditions. The three pairing conditions differed significantly for agay ihgh sentences only for 4-year—olds. Tukey tests revealed that P1 was most difficult, followed by P2, with P3 as the easiest (P1 V. P2: q=4.82; Pl v. P3: q=7.83; P2 v. P3: q=3.01; q.05=2'8 for each). The pairing conditions dif- fered significantly for toward sentences at all three ages: 84 (l) AIL4 years, conditions 1 and 3 were equally difficult, but condition 2 was significantly easier (Pl v. p2: q=9.04; P3 v. P2: q=8.43; =2.3). q.05 (2)A¢.5 years, condition 1 was most difficult, followed by condition 3, and then condition 2 (Pl v. P2: q=13.86; Pl v. P3: q=7.23; P2 v. P3: q=6.63; q 05:2.8 for each). (3) For the kindergarten children, condition 1 was most difficult, followed by conditions 2 and 3 (Pl v. P2: q=10.84; P1 v. 93: q=9.64; =2.8). q.05 Analysis of Position Biases As in the first experiment, children's errors were classified according to possible simple position biases such as choosing the moving airplane on the left or right side, or above or below the reference airplane. Almost 66% Of the children showed mixed patterns of errors, 17% made no errors, % made errors on the left only, and 15% made errors on the right only. It could be argued that there was a position bias of choosing the airplane on the right side since so many children showed this pattern. However, the bias is unlikely to have produced any of the Significant differences in the data. About 10% of these children made only one error, with the remaining 5% making only two errors. Furthermore, those children making two errors never made both errors on the same target item. 85 Age Differences and Age Changes The present experiment evidenced both age differ- ences and age changes. The three age groups differed significantly in overall performance on all target items, with the 4-year—Olds making the most errors and the kinder- garten children the least. In addition, there was a sig- nificant interaction of age with extension. (Although the triple interaction of these two variables with pairing condition was significant in the overall ANOVA, the simple main effects ANOVA indicated that the interaction of age with pairing condition was non-significant.) The three age groups differed significantly on away from sentences only in condition 1 (reference point contrast, presenting rising-falling away from displays). Floor effects eradi- cated any potential differences among the groups in the other two conditions. On toward sentences, the three ages differed significantly in conditions 1 and 3. In condition 2, only the 4—year—olds performed significantly worse than the other two groups. Again, the lack of significant dif- ferences among the other groups can be attributed to floor effects. Closer examination of the data (see Appendix Fl) reveals some interesting patterns. In condition 1, although the three ages differed in performance on EHEX ihgh phrases, each age group made about the same number of errors on rising away from and falling away from sentences. 86 However, the differences among these groups on toward sentences in condition 1 represents an improvement in performance on falling toward sentences, rather than on rising toward sentences. Between 4 and 5 years, children reduced by half their error rates on the former type, while maintaining a high error rate on the latter type. This trend supports the contention of the reference point hypothesis that children should learn falling toward before risihg toward. In condition 2, floor effects obscured any potential differences among the three ages for away from sentences, and between the two Older groups for toward sentences. The data revealed that children improved on toward sentences between 4 and 5 years in this condition. The source Of this change was actually improve- ment on rising toward sentences, rather than on falling toward sentences. Four-year-olds already made practically no errors on falling toward sentences in condition 2. This also supports earlier acquisition Of falling toward, as compared to rising toward. In condition 3, all chil— dren made few errors on either away from phrase, but a dramatic improvement occurred on both toward phrases between 4 and 5 years. It appears, then, that the majority of age changes occurred between 4 and 5 years, and that the changes occurred somewhat earlier on the above-reference point phrase falling toward than on its below-reference point 87 counterpart rising toward. It is only in condition 1 (reference point contrast, presenting rising-falling away high displays) that performance continued to improve through kindergarten. The data on age changes may also be used to infer children's ability to discriminate rising—falling (con- trasted in condition 1) and toward-away from (contrasted in condition 3). At 4 years, children only performed well on condition 2, in which both of these dimensions varied. At 5 years, children performed best on condition 2, and next best on condition 3 (toward-away from). Condition 1, the risihg-falling discrimination, was most difficult for 5-year olds. Kindergarten children performed equally well on conditions 2 and 3, but still had some difficulty with the rising—falling discrimination of condition 1. It appears that the ages differ most in ability to differen— tiate away from-toward between 4 and 5 years, but that the ability to discriminate rising-falling continues to change through kindergarten. Differences Attributable to Extension The expression of differences due to extension was modulated by both age and pairing condition. In condi- tion 1, the most difficult condition, children always made fewer errors on away from sentences. They were better at discriminating rising-falling away from than rising-falling toward. This is not surprising if the away from phrases 88 are learned first. In condition 2, the easiest condition, only the 4-year—Olds made fewer errors on away from phrases. The lack of a difference between the 5-year-Olds and kindergarteners can be attributed to floor effects-- both groups made practically no errors on any phrases in this condition. Furthermore, the Significant extension difference for the youngest children actually represents a difference between performance on the two away from phrases and performance on rising toward sentences. Chil— dren in this group made the same number of errors on fall- ing toward sentences in condition 2 as they did on both away from types. In condition 3, the condition of intermediate dif- ficulty, both the 4-year-olds and 5-year-olds were better on away from than on toward sentences, but the differences only reached significance at 4 years. This implies that the error rate for the 5-year-olds on toward sentences in condition 2 may be somewhat depressed by an elimination strategy. Since both the rising-falling and toward—away ihgh directions varied in condition 2, children could improve their scores on the difficult toward sentences by eliminating the "wrong" choice. This would obscure the trend seen in both other conditions for the 5—year-olds to perform somewhat better on away from sentences. It is 89 difficult to determine the exact nature of this elimination strategy because both dimensions varied. However, since the 5-year-Olds were more adept at discriminating away from—toward than rising-falling, it is likely that 5—year— olds eliminated the away from relationship as the "wrong" choice in condition 2 to improve their scores. However, the data do indicate that children understood the aygy EEQE phrases before the toward phrases, as predicted by the extension hypothesis, and that their performance on toward phrases gradually improved with age. Differences Attributable to Reference Point NO support for the reference point hypothesis emerged from the data. The only situatiions demonstrating comprehension differences attributable to reference point were (a) the trend of improvement with age on falling toward (unextended, above-reference point) sentences, and the lack of concurrent improvement on rising toward sen- tences (unextended, below-reference point) in condition 1; and (b) the higher error rate of the 4-year-Olds on rising toward, as compared to falling toward, in condition 2. Effects due to reference point were weak in Experiment 1, and their disappearance in the present experiment is not surprising. The present task employed 90 an airplane as a secondary reference point, while the first used a pictorially represented primary reference plane. The disappearance of reference point effects could reflect this difference in reference point charac- ter, but further research is needed to substantiate this. As in the first experiment, conditional propor- tions were computed to determine the most common order of scoring correctly pn all four phrases. These propor- tions are presented in Table 6. Table 6 Conditional Proportions in Experiment 2 Phrase Acquired Second Phras; Acquired rising falling rising falling 1rst away from away from toward toward Rising away from ... .795 .792 .826 Falling away from .756 ... .625 .826 Rising toward .463 .385 ... .565 Falling toward .463 .487 .542 ... Note: Each entry represents the proportion of all children making no errors on the row item, given that none were made on the column item. 91 These proportions should be interpreted in the same manner as those in Experiment 1. The entries confirm that chil- dren's comprehension of rising away from and falling away from developed at about the same time, and that comprehen— sion of rising toward and falling toward developed later but at about the same rate. The trends toward reference point differences cited in the above paragraph are not truly reflective of children's comprehension. Differences Attributable to Pairing Conditions While no predictions were made about differences among the pairing conditions, children did perform dif- ferently on the same sentences in different conditions. Condition 1, the reference point contrast, was consis- tently the most difficult condition for children. How- ever, it is unlikely that its difficulty was due to the reference point variable. It is more likely that its difficulty reflects the problems children had in discrim- inating rising-falling. Condithn3 was the next most difficult condition. Children improved on this pairing between 4 and 5 years, and by kindergarten made practically no errors in this situation. Condition 3 contrasted loca— tions in both reference point and extension, but it is more likely that the improvements reflect the ability to discriminate toward-away from. Condition 2, the extension contrast, was consistently the easiest condition. This is 92 the typical feature theory type of error, and was the least frequent error observed in all age groups. It is likely that children's superior performance in this con- dition reflects the fact that both the rising3falling and toward-away from dimensions varied in the displays. Chil- dren needed only to identify one dimension of difference to make a correct response. Differences in the Acquisition of Adjectives and Verb Phrases The final issue in evaluating these data is whether or not children of the same age performed better on adjectives than on verb phrases, according to each locational type. Since each experiment used a group of 4-year-Olds and a group of 5-year-Olds, the performance of these children may be compared to address this issue. To simplify the analysis, one 4-year-old and one 5-year-old were randomly excluded from the children in the first experiment, so that both groups of 4's and 5's would have twenty children each. Four 2 x 2 between—subjects ANOVAs were performed on these data, one for each locational type. In the com- parison of extended above—reference point items, children at both ages performed significantly better on high than on rising away from (h(1,76)=6.69, p 1.05). However, this was the condition supporting the predicted difference. The extended below-reference point comparison indicated 93 that 5-year—olds performed better than 4—year—Olds on both deep and falling away from (h(l,76)=4.32, p 1.05). The comparison of unextended above-reference point items revealed a significant interaction of age and item (h(l,76)=4.3l, p 1.05). A simple main effects analysis indicated that the source of this interaction was in improvement on falling toward sentences between 4 and 5 years, with no concurrent change in performance on low. The comparison of the unextended below-reference point items shallow and rising toward did not reveal any signif- icant differences. The data indicated that children generally per- formed as well on the verb phrases, descriptions of dynamic relationships, as on the adjectives, descriptions of static relationships. CHAPTER IV SUMMARY AND CONCLUSIONS The purpose of these experiments was to evaluate the roles of extension, reference point, and description differences in children's comprehension Of spatial terms. The experiments differed somewhat in their support of the three general hypotheses advanced in the first chapter. Both experiments indicated that children are sensitive to the degrees of extension underlying spatial locations. Data which reflect comprehension of the target items, rather than the operation of nonlinguistic strategies, indicated that extended items were learned before unextended items. Children showed earlier acquisition of the extended relationship for words describing both static (high and ggpp) and dynamic (rising away from and falling away from) situations. Additional support for the early salience of extension comes from the nonlinguistic strat- egies identified in the data: In all cases where children were able to identify and eliminate an unnamed relation- ship as the "wrong" choice, that relationship represented an extended location. The experiments differed, though, in their support of the reference point hypothesis. In the first experi- ment, children's performance indicated that some differ- ences existed in the learning of above- and below- 94 95 reference point phrases. This support included (a) significant differences in children's performance on iph and shallow sentences; (b) large differences in the pro- portions of children scoring correctly on high before hgpp versus hpgp before high, and in analogous proportions for igh and Shallow; (c) differences in the youngest chil- dren's performance on ggpp versus their performance on high; and (d) the fact that an elimination strategy was employed when the high location was the "wrong" choice, but not when the gpgp location was the "wrong‘I choice. ‘Although some of these differences did not reach signifi— cance, the trends did support earlier learning of high-low as compared to deep-shallow. In the second experiment, no support for differences due to reference point emerged. Although two trends in the data did differentiate the acquisition of falling toward and rising toward, the above- and below—reference point unextended phrases, neither was strong enough to produce any overall reference point effect. It is possible that the disappearance Of reference point effects in the second experiment reflects the difference in type of reference point (primary versus secondary), but additional research is needed to assess this question. With a secondary reference point, the entire dimension under study is actually above the primary reference plane of the ground level. Relationships relative to this secondary point 96 could therefore all be treated as within the portion of the vertical dimension which is above the ground. Since that portion of the vertical dimension is least complex and most easily conceptualized, all of the relationships within this dimensional portion could be learned before any relationships below the primary plane. In this View, all Of the phrases in the second experiment could be classified as above the primary reference plane, while the adjectives deep-shallow in the first experiment would retain their below-reference point character. This would allow for expression of some reference point differences in the first experiment, and their absence in the second one. The experiments did not demonstrate that children understand words for these locations in their static states before phrases for these locations in their dynamic states. Only in a comparison of the extended above— reference point items high and rising away from did any support for this hypothesis emerge. Conclusions The dissertation demonstrated that young children use certain perceptual attributes to organize portions of their lexicons. The order in which they learned different types of spatial descriptions was independent of descrip- tion type differences, and indicated that words expressing certain perceptual attributes were consistently understood 97 before words expressing other attributes. The data from these two experiments may be used to draw certain con- clusions regarding lexical development, the organization Of perceptual space, and the study of spatial words. Lexical Development The focus of these studies was not the mechanism by which children map characteristics of perceptual space onto lexical items, but rather hhigh characteristics they map onto those items at different points in development. However, some inferences about the claims of different theories of lexical development may be made from these data. Semantic feature theory (e.g., E. Clark, 1972a) and generalization theory (Anglin, 1970) both claim that perceptual attributes are represented as semantic features in the meaning of spatial words. Nelson's concept forma- tion theory (1974) contends that aspects of meaning are represented as dynamic, functional relations. While the data do not determine which of these perspectives is closer to the actual mechanism used by young children, the data do reveal that children learned descriptions for the dif— ferent relationships in their dynamic states (i.e., the verb phrases) at the same age as they learned descriptions for these relationships in their static states (i.e., the adjectives). Bloom, Lightbown and Hood (1975), in their study of productive language, found that children encoded 98 action events before stative events. This has been used to support the importance of dynamic, functional relations. The present data address this phenomenon in comprehension. Children did not learn to decode descriptions of action events before descriptions of stative events, though. Instead, they were equally adept at decoding both types. Semantic feature theory claims that children first learn the dimension to which an antonym pair refers, and then learn to apply each pair member to a different end of that dimension (e.g., E. Clark, 1972a). Because of this, it is common for children to confuse words on the basis of their implied extension within a dimension. Critics of feature theory (e.g., Anglin, 1970; Brewer & Stone, 1975) counter this proposal with the contention that specific, extensional features are learned before general, dimensional features. Thus, it should be more common for children to confuse words of the same extension, but of different dimensions. The data from the present experiments more closely support this latter View. The most frequent type of error made by all children on shallow sentences was confusion of gpgp and shallow. This is the standard feature theory type of error. However, this pattern did not occur on igy sentences, and was also not seen in the learning of the verb phrases in the second experiment. The absence of errors on iph sentences when igh and high squirrels were paired is not surprising: High is learned 99 so early that it is unlikely to be confused with any other location. The absence of confusions of rising away from and falling toward, or of falling away from and rising toward, may be similarly explained. These pairings were particularly easy because both the rising-falling and toward-away from dimensions varied. Children needed only to identify one portion of the motion to choose the correct relationship. The scarcity of feature theory-type errors, then, could be an artifact of the experimental design. However, it was quite common for children to con- fuse words of the same extension but of different dimen- sions, according to the location of the reference point. The most common confusion children made in identifying ipg locations was to confuse igh and shallow. This type of error was second most frequent in performance in shallow sentences, and was most frequent in the comprehension of all the verb phrases. Children in the second experiment had the most difficulty discriminating rising-falling away from, and rising-falling toward. Confusion Of the away ipgh phrases is confusion of extended items from different dimensional portions, while confusion of the toward phrases is confusion of unextended items from different dimensional portions. Thus, the data imply that the extension of spatial locations is learned before their dimensional properties. 100 This conclusion is also supported by the condi- tional proportions computed in both experiments. In both cases, children first learned extended items, then unextended items. Only in the first experiment was there evidence that the dimensional portion to which the words referred influenced the order of their acquisition. Perceptual Space The data from both experiments lend some support to H. Clark's (1973) analysis of perceptual space into dominant reference directions, certain of which are more easily conceptualized than others. The normative role Of the vector extending upward from the ground, as compared to the vector extending downward from the ground, was the basis for the reference point hypothesis. As has been previously discussed, this hypothesis was not strongly supported by the data. In the first experiment, there was some support for earlier learning of static locations along this vector (high—low), as compared to analogous locations along the downward vector (deep-shallow). The trend was not overwhelming, though, and foreshadowed the results of the second experiment. In the latter case, no support for differences in the learning of dynamic locations occurred. As has been suggested, the discrepancy between the two sets Of data may reflect differences in the types of reference points used--primary versus secondary. In addition, it is possible that the use of a two-dimensional 101 display in the first experiment affected the strength of the reference point differences. It would be interesting to assess the same locations with a three-dimensional display to determine if the reference point effects become more robust. However, both experiments imply that vectors which originate at the reference plane, underlying extended loca— tions, are more readily conceptualized than vectors which terminate at that plane, and underlie unextended locations. AS mentioned in the introduction, it is difficult to derive an a priori basis for this prediction based purely on per- ceptual concerns. Yet it is possible to infer differences in the conceptualization of these underlying vectors from children's language performance. The fact that children consistently learned words for extended locations before learning words for unextended locations, regardless of their static/dynamic states, implies that children are more sensitive to the presence Of extension in a spatial location than to the lack of extension. As many theorists claim (e.g., H. Clark, 1973; Nelson, 1974), it should be easier to acquire a lexical item for a more readily con— ceptualized phenomenon. The case for the cognitive- perceptual simplicity of vectors which originate at the plane is most strongly supported in the second experiment, where moving Objects actually traversed the paths of H. Clark's theoretical vectors. 102 Spatial Terms The research in this dissertation supports current explanations of the acquisition of spatial terms which rely on an integration of cognitive, perceptual, and lin— guistic concerns, rather than explanations which focus on strictly linguistic differences. While the early lexical marking approach could be used to account for differences in performance in the first experiment, it cannot be applied to the verb phrases in the second one. Further— more, concentration on linguistic differences obscures the beauty and simplicity of the semantic system. Even in the acquisition of quite different lexical items, some of which these children had probably never heard before (e.g., falling toward), the order of learning the words and phrases was generally the same. In addition, the order of learning these items supported theoretical dif- ferences in the complexity of perceptual attributes. The experiments indicate that the order of acqui- sition of spatial terms may be predicted on the basis of their extensional characteristics, regardless of the type of description in question. Reference point may be of some value as a predictor, but additional research is needed to verify its usefulness. As hypothesized by H. Clark (1970a), children are quite sensitive to dif— ferences in degrees of extension. However, children in the present experiments did not demonstrate much confusion 103 of the words for opposite ends of a given dimension. It was more common for children to confuse words of the same extension, but of different dimensional portions. The present research also clarifies the research of Friedenberg and Olson (1977) on the acquisition of adjectives and verb phrases. While that study indicated that children better understood the adjectives higher- lower than than the verb phrases rising-falling away from, children were required to match the verbal descriptions to static displays. In the present experiments, verb phrases were used to describe dynamic relationships, and differ— ences between the acquisition of adjectives and verb phrases disappeared. These experiments raise additional questions about children's comprehension of spatial terms. First, the experiments established that children learn away from phrases before toward phrases when applied to vertical motion in space. This finding contradicts studies of other spatial verbs used for motion on the horizontal dimension. For example, children typically demonstrate accurate com- prehension of 99mg before g9, and bring before take_(e.g., Clark & Garnica, 1974; Fillmore, 1966). In both of these cases, the verb representing a decrease in extension between the moving object and its reference point is learned first. Fillmore has suggested (personal communi- cation) that egocentricity plays a role in these studies 104 of horizontal motion. In cases where motion is relative to the "self", as in come-go and bring-take, it may be more important to first learn about motion toward the "self". Such motions are important in an appreciation of impending personal danger, the increase in apparent size of an approaching object, etc. Thus, children.may more readily conceptualize horizontal motions toward the self, as compared with motions away from the self. This would allow for earlier learning of verbs like 99mg and bgigg. These horizontal motions should be assessed in situations where an independent reference point is used, to determine if the character of the reference point (egocentric versus independent) is the source of the difference between the two dimensions. Also, it would be of interest to determine whether the two—dimensional nature of the displays used in the first experiment affected the expression of reference point differences. The dissertation demonstrated the importance of appropriate perceptual displays, as in the case of the Friedenberg and Olson (1977) study, and it is possible that reference point differences were weakened by the use of a felt board. Finally, it is curious that no differences occurred in children's learning of the phrases rising-falling away from and rising-falling toward. The results imply that children learn the words rising and falling at about the 105 same time, although they learned away from at least a year before learning toward. Differences in adults' compre- hension of rising-falling away from have been demonstrated (Friedenberg, 1976), and in children's ability to apply these phrases to static displays (Friedenberg & Olson, 1977). Studies of the simple verbs rise-fall have also shown differences in adult performance (Miller, 1972). It is advisable to replicate the findings of the second experiment to determine if such a difference is present. APPENDICES 106 APPENDIX A LETTER TO PARENTS FOR EXPERIMENT 1 APPENDIX A LETTER TO PARENTS FOR EXPERIMENT 1 Michigan State University Department of Psychology Graduate Division Dear Parent(s), I will be conducting a research project at the MSAU Day Care Center, and would like your permission for your child to participate. This study has been approved by my doctoral dissertation committee, the Early Childhood Studies Committee and the Parent Board of the day care center. From past experience with tasks such as this one, children generally enjoy their participation. The task is simple and fun to do. I am interested in how the patterns children show in learning spatial words are related to their concepts of space. This project uses the words: high, low, deep and shallow. All of these refer to locations on the vertical dimension. Each child will be seen individually for about 15 minutes. I will have a felt board with a cut-out of a tree and a wishing well, to be used as locations for felt squirrels. I will place two squirrels in two different places on the board, and read a sentence describing the location of one of them (for example, Point to the squirrel in a high place.) I will repeat this procedure with several different pairs of locations, recording children's point- ing responses as correct or incorrect. Answers will be totalled for three age groups: 2% — 3%, 3% — 4%, 4% - 5% years. This information will give me a pattern of language comprehension according to which words are difficult and easy for young children to understand. All testing will be performed by myself or my undergraduate research assistant. Please indicate your permission/non-permission for your child's partici- pation on the enclosed card, and return it to the day care center. I will forward a copy of the general findings of the study to each par- ticipating family, as soon as they are available. Individual results will be kept both confidential and anonymous. If you have any question about the project, please feel free to call me any evening at 332—3811. Thank you for your interest and cooperation. Sincerely, Lisa Friedenberg Ph.D. Candidate Developmental Psychology 108 APPENDIX B SUMMARY TABLES FOR EXPERIMENT 1 109 IIIIIIIIiIIiIIIIIIIIIIIIIIIIIllIIlll::::T--————————_______r APPENDIX B SUMMARY TABLES FOR EXPERIMENT 1 APPENDIX TABLE B1 Total Errors for Each Condition According to Age Group Deep Shallow A e Grou Jail—— ”EL— 9 9 P1 P2 P3 P1 P2 P3 P1 P2 P3 P1 P2 P3 Total // 2%-3% yrs 7 6 8 l4 4 21 15 13 17 19 24 17 165 3%—4% yrs 0 1 4 l6 1 17 4 5 5 15 15 15 98 4%-5% yrs 0 2 1 10 1 22 2 4 3 14 16 5 80 Totals 7 9 13 40 6 60 21 22 25 48 55 37 343 29 106 68 140 Totals Note: N=21 in each group 110 APPENDIX TABLE B2 ANOVA Summary Table Source SS df MS F p 4 Between subjects Age (A) 15.92 2 7.96 11.32 .0005 Subjs in groups 42.21 60 0.70 Within subjects Reference Point (R) 7.05 7.05 7.99 .006 A X R 2.83 1.41 1.61 NS R X subjs in groups 52.87 60 0.88 Extension (E) 29.37 1 29.37 62.86 .0005 A X E 1.69 0.84 1.80 NS E X subjs in groups 28.03 60 0.47 Pairing Cond. (P) 3.68 1.84 5.69 .004 A X P 0.47 0.12 0.37 NS P X subjs in groups 38.84 120 0.33 E X R 0.03 0.33 0.08 NS A X E X R 0.08 0.04 0.09 NS E X R X subjs in grps 25.97 60 0.43 E X P 2.30 1.15 4.37 .015 A X P X E 0.73 0.18 0.69 NS E X P X subjs in grps 31.64 120 0.26 R X P 10.61 2 5.30 15.77 .0005 A X R X P 1.03 4 0.26 0.76 NS R X P X subjs in grps 40.36 120 0.34 E X R X P 10.11 2 5.06 21.91 .0005 A X E X R X P 1.85 4 0.46 2.01 NS E X R X P X subjs 27.69 120 0.23 Total 375.38 755 111 APPENDIX TABLE B 3 Simple Main Effects ANOVA Summary Table for Extension X Pairing Condition Source SS df MS F p_£ Within Subjects Extension at P1 14.29 1 14.29 43.08 .01 Extension at P2 3.57 1 3.57 10.77 .01 Extension at P3 13.81 1 13.81 41.66 .01 Pooled error 180 0.33 Pairs for extended words 0.42 2 0.21 0.71 NS Pairs for unextended words 5.57 2 2.79 9.49 .01 Pooled error 240 0.29 112 APPENDIX TABLE B4 Simple Main Effects ANOVA Summary Table for Reference Point X Pairing Condition Source SS df MS F pp; Within Subjects Reference Points at P1 1.92 1 1.92 3.71 NS Reference Points at P2 15.26 1 15.26 29.47 .01 Reference Points at P3 0.48 1 0.48 0.92 NS Pooled error 180 0.52 Pairs for above reference point words 13.39 2 6.69 20.33 .01 Pairs for below reference point words 0.89 2 0.45 1.36 NS Pooled error 240 0.33 113 APPENDIX TABLE B5 Simple Main Effects ANOVA Summary Table for Extension X Reference Point X Pairing Condition Source SS df MS F p 1 Within Subjects Pairs for high 0.30 2 0.15 0.51 NS Pairs for 19w 23.66 2 11.83 40.99 .01 Pairs for deep 0.14 2 0.07 0.24 NS Pairs for shallow 2.61 2 1.31 4.53 .05 Pooled error 480 0.29 Reference points for high-deep at P1 1.55 1 1.55 3.82 NS Reference points for high—deep at P2 1.34 l 1.34 3.29 NS Reference points for high—deep at P3 1.14 l 1.14 2.80 NS Reference points for low—shallow at P1 0.51 1 0.51 1.25 NS Reference points for low-shallow at P2 19.05 1 19.05 46.69 .01 Reference points for low-shallow at P3 4.20 1 4.20 10.29 .01 Pooled error 360 0.41 Extension for high—low at P1 8.64 1 8.64 27.45 .01 Extension for high-low at P2 0.07 l 0.07 0.23 NS Extension for high-low at P3 17.53 1 17.53 55.69 .01 Extension for deep- shallow at P1 5.79 1 5.79 18.38 .01 Extension for deep- shallow at P2 8.64 1 8.64 27.45 .01 Extension for deep- shallow at P3 1.14 1 1.14 3.63 NS Pooled error 360 0.32 R X E at P1 0.14 1 0.14 0.48 NS R X E at P2 5.14 1 5.14 17.25 .01 R X E at P3 4.86 1 4.86 16.30 .01 Pooled error 180 0.29 R X P for high-deep 0.02 2 0.008 0.03 NS R X P for low-shallow 20.70 2 10.35 36.50 .01 Pooled error 240 0.28 E X P for high-low 10.56 2 5.28 21.36 .01 E X P for deep—shallow 1.86 2 0.93 3.76 .05 Pooled error 240 0.25 114 APPENDIX C LETTER TO PARENTS FOR EXPERIMENT 2 115 APPENDIX C LETTER TO PARENTS FOR EXPERIMENT 2 Michigan State University Department of Psychology Graduate Division Dear Parent(s), I will be conducting a research project at the Laboratory Pre-School/ MSAU Day Care Center, and would like your permission for your child to participate. The study has been approved by my doctoral disserta- tion committee, the Early Childhood Studies Committee and the Parent Board. From past experience with tasks such as this one, children generally enjoy their participation. The task is simple and fun to do. I am interested in what the patterns children show in learning spatial words tell us about children's concepts of space. This project uses the verb phrases, rising away from, falling toward, falling away from, and rising toward. A11 refer to locations on the vertical dimension. Each child will be seen individually for about 15 minutes. I have a wooden box with two upright poles, on which I can move airplanes up and down by motorized pulleys. I will place two different colored airplanes on each pole, and read a sentence like: The blue airplane is rising away from the red airplane, asking the child to point to the pole that shows this relationship. I will repeat this procedure with several different pairs of moving relationships, recording children's pointing responses as correct or incorrect. Answers will be totalled for age groups and used to determine which verb phrases are easy and difficult for young children to understand. Please indicate your permission/nonpermission on the enclosed card and return it to the center. We must have a signed card from each family contacted, even if the child (children) are not to participate. I will forward a copy of the general findings of the study to each participating family as soon as they are available. Individual results will be kept both confidential and anonymous. If you have any questions about the project, please feel free to call me any evening at 332-3811. Thank you for your interest and cooperation. Sincerely, Lisa Friedenberg Ph.D. Candidate Developmental Psychology 116 APPENDIX D SUMMARY TABLES FOR EXPERIMENT 2 117 APPENDIX D SUMMARY TABLES FOR EXPERIMENT 2 APPENDIX TABLE D1 Total Errors for Each Condition According to Age Group Rising Falling Falling Rising Age Group away from toward away from toward P1 P2 P3 P1 P2 P3 P1 P2 P3 P1 P2 P3 Total 3%-4% yrs 8 3 2 18 5 15 8 5 1 13 11 15 104 46-56 yrs 6 1 2 10 0 6 5 3 2 14 1 6 56 Kinder- 3 o 1 8 o 2 1 o 1 11 1 1 29 garten Totals 17 4 5 36 5 23 14 8 4 38 13 22 189 Totals 26 64 26 73 Note: N = 20 in each group 118 APPENDIX TABLE D2 ANOVA Summary Table Source SS df MS F p_£ Between subjects Age (A) 12.025 6.012 13.56 .0005 Subjs in groups 25.28 57 0.444 Within subjects Reference Point (R) 0.168 0.168 0.585 NS A X R 0.053 0.026 0.092 NS R X subjs in grps 16.363 57 0.287 Extension (E) 10.035 1 10.035 37.435 .0005 A X E 2.936 2 1.468 5.477 .007 E X subjs in grps 15.279 57 0.268 Pairing Cond. (P) 12.033 2 6.016 17.823 .0005 A X P 0.316 4 0.079 0.234 NS P X subjs in grps 38.483 114 0.337 E X R 0.169 1 0.169 0.816 A X E X R 0.019 2 0.009 0.047 NS E X R X subjs/grps 11.729 57 0.206 E X P 2.977 2 1.488 6.975 .001 A X E X P 2.188 4 0.547 2.564 .042 E X P X subjs/grps 24.333 114 0.213 R X P 0.544 2 0.272 1.743 NS A X R X P 0.822 4 0.205 1.316 NS R X P X subjs/grps 17.8 114 0.156 R X E X P 0.444 2 0.022 0.135 NS A X R X E X P 1.105 4 0.276 1.686 NS R X E X P X subjs 18.683 114 0.164 Total 213.3875 719 119 APPENDIX TABLE D 3 Simple Main Effects ANOVA Summary Table for Age X Extension Source SS df MS F 1 Within Subjects Extension at 4 yrs 10.416 1 10.416 38.868 .01 Extension at 5 yrs 1.35 l 1.35 5.037 .05 Extension at Kind. 1.204 1 1.204 4.493 .05 Pooled error 57 0.268 Ages for away from 1.872 2 0.936 2.631 NS Ages for toward 13.088 2 6.544 18.395 .01 Pooled error 114 0.356 120 APPENDIX TABLE D4 Simple Main Effects ANOVA Summary Table for Extension X Pairing Condition Source SS df MS F p l\ Within Subjects Extension at P1 7.704 1 7.704 33.258 .01 Extension at P2 0.204 1 0.204 0.881 NS Extension at P3 5.104 1 5.104 22.033 .01 Pooled error 171 0.232 Pairings for away from 2.372 2 1.186 4.305 .05 Pairings for toward 12.638 2 6.319 22.938 .01 Pooled error 228 0.275 121 APPENDIX TABLE D5 Simple Main Effects ANOVA Summary Table for Age X Extension X Pairing Conditions Source SS df MS w ro h\ Within Subjects Pairs for away at 4 yrs 2.15 Pairs for toward at 4 yrs 2.866 Pairs for away at 5 yrs 0.816 Pairs for toward at 5 yrs 6.166 Pairs for away at Kind. 0.2 Pairs for toward at Kind. 4.866 1.075 3.902 .05 1.433 5.202 .01 0.408 1.482 NS 3.308 12.008 .01 0.1 0.363 NS 2.433 0.832 .01 Pooled error 22 0.275 Ages for away at P1 1.816 0.908 3.005 .05 Ages for away at P2 0.8 0.4 1.323 NS Ages for away at P3 0.05 0.025 0.083 NS Ages for toward at P1 1.816 0.908 3.005 .05 Ages for toward at P2 4.266 2.133 7.058 .01 Ages for toward at P3 8.716 4.358 14.418 .01 0.302 2.813 12.141 .01 1.013 4.371 .05 8.45 36.477 .01 2.113 9.119 .01 0.113 0.486 NS 0.8 3.454 NS 2.813 12.141 .01 0.013 0.054 NS 0.013 0.054 NS 00 d> Pooled error Extension for Extension for Extension for Extension for yrs at P1 2.813 yrs at P2 1.013 yrs at P3 8.45 yrs at P1 2.113 Extension for yrs at P2 0.113 Extension for yrs at P3 0.8 Extension for Kind. at P1 2.813 Extension for Kind. at P2 0.013 Extension for Kind. at P3 0.013 U1U1011~bnl>~sb Pooled error 17 0.232 A X P for away from 0.794 0.199 0.721 NS A X P for toward 1.711 0.428 1.553 NS Pooled error 22 0.276 A X E for P1 0.033 0.016 0.072 NS A X E for P2 0.933 0.466 2.014 NS A X E for P3 4.159 2.079 8.976 .01 Pooled error 17 0.232 E X P for 4 yrs 1.858 0.929 4.015 .05 E X P for 5 yrs 1.675 0.838 3.924 .05 E X P for Kind. 1.633 0.816 3.826 .05 nhUJNN l—‘NNN con-4:. i—‘I—‘l—‘l—‘f—‘l—‘l-‘Hl—‘H NNNNNNN CONNNNNN 0.214 H [—4 Pooled error 122 REFERENCES 123 REFERENCES Anglin, J. M. The growth of word meaning. Cambridge: MIT Press, 1970. Bierwisch, M. Some semantic universals of German adjectivals. Foundations of Language, 1967, 3, 1-36. Blank, M. The cognitive functions of language in the preschool years. Developmental Psychology, 1974, 10, 229—245. Bloom, L. One word at a time: The use of single word utterances before syntax. The Hague: Mouton, 1973. Bloom, L., Lightbown, P. & Hood, L. Structure and variation in child language. Monograph of the Society for Research in Child Development, 1975, 40 (No. 160). Bowerman, M. Early syntactic development: A cross- linguistic study with special reference to Finnish. Cambridge, England: Cambridge University Press, 1973. Braine, M. Children first word combinations. Monograph of the Society_for Research in Child Development, 1976, 41 (No. 164). Brewer, W. F. & Stone, J. B. Acquisition of spatial antonym pairs. Journal of Experimental Child Psychology, 1975, 19, 299-307. Brown, R. A first language: The early stage. Cambridge, Mass. Harvard University Press, 1973. Cazden, C. & Brown, R. The early development of the mother tongue. In E. Lenneberg’ & L. Lenneberg (Eds): Foundations of language development, Vol. 1, 1976, 299-310. Clark, E. Some aspects of the conceptual basis for first language learning. In R. L. Schiefelbusch & L. Lloyd (Eds): Language perspectives: Acquisition, retardation, and intervention. University Park Press, 1974, 335-345. Clark, E. What's in a word? On the child's acquisition of semantics in his first language. In T. E. Moore (Ed): Cognitive development and the acquisition of language, Academic Press, 1973(a), 65-110. Clark, E. Nonlinguistic strategies and the acquisition of word meanings. Cognition, 1973(b), 2, 161-182. 124 125 Clark, E. On the child's acquisition of antonyms in two semantic fields. Journal of Verbal Learning and Verbal Behavior, 1972(a), 11} 750-758. Clark, E. Some perceptual factors in the acquisition of locative terms by young children. Paper presented at the 8th regional meeting of the Chicago Linguistic Society, 1972(b). Clark, B. 'How young children describe events in time. In G. Flores d'Arcais & W. Levelt (Eds): Advances in psycholinguistics. North Holland Publishing Co., 1970, 275—285. Clark, E. & Garnica, 0. Is he coming or going? On the acquisition of deictic verbs. Journal of Verbal Learning and Verbal Behavior, 1974, 19, 559-572. Clark, H. Space, time, semantics and the child. In T. E. Moore (Ed): Cognitive development and the acquisition of language, Academic Press, 1973, 27-64. : Clark, H. The primitive nature of children's relational concepts. In J. R. Hayes (Ed): Cognition and the development of language, John Wiley & Sons, 1970(a), 269-278. Clark, H. On comprehending comparatives. In G. Flores d'Arcais & W. Levelt (Eds): Advances in psycho— linguistics, North Holland Publishing Co., 1970(b). Clark, H. Linguistic processes in deductive reasoning. Psychological Review, 1969(a), 19, 387-404. Clark, H. The influence of language in solving three term series problems. Journal of Experimental Psychology, l969(b),l99, 205-215. Clark, H. & Chase, W. On the process of comparing sentences 'mapictures. Cognitive Psychology, 1972, 91, 513—521. Corballis, M. C. & Beale, I. L. The psychology of left and right, Halstead Press: Hillsdale, N. J., 1976. Danzinger, K. Children's understanding of kinship terms. Journal of Genetic Psychology, 1957, 91, 213-232. Donaldson, M. & Balfour, G. Less is more: A study of language comprehension. British Journal of Psychology, 1968, 99, 461-471. 126 Donaldson, M. & Wales, R. On the acquisition of some relational terms. In J. R. Hayes (Ed): Cognition and the development of language, John Wiley and Sons, 1970, 235-268. Fillmore, C. J. Santa Cruz lectures on deixis (1971). Indi— and University Linguistic Club reproduction, 1975. Fillmore, C. J. Deictic categories in the semantics of "come". Foundations of language, 1966, 9, 219-227. Friedenberg, L. Linguistic and conceptual processes in the comprehension of spatial antonyms. Paper presented at the 84th Annual Convention of the American Psychological Association, 1976. Friedenberg, L. & Olson, G. M. Children's comprehension of simple descriptions of vertical arrays. Child Development, 1977, 49, 265-269. Gibson, E. J. Principles ofgperceptual learning and development, New York: Appleton-Century—Crofts, 1969. Greenberg, J. Language universals. The Hague: Mouton, 1966. Kessen, W., Haith, M. & Salapatek, P. The visual response of the human newborn to linear contour. Journal of Experimental Child Psychology, 1972, 19, 9-20. Klatzky, R., Clark, E. & Macken, M. Asymmetries in the acquisition of polar adjectives. Journal of Experi- mental Child Psychology, 1973, 19, 32—46. MacNamara, J. Cognitive basis of language learning in infants. Psychological Review, 1972, 19, 1-14. Maratsos, M. Decrease in understanding of "big" in pre- school children. Child Development, 1973, 44, 747-752. Miller, G. A. English verbs of motion: A case study in semantic and lexical memory. In A. W. Melton & E. Martin (Eds): Coding processes in human memory, Winston Press, 1972, 335-372. Miller, G. A. & Johnson-Laird, P. N. Language and Perception. Harvard University Press, Cambridge, Mass., 1976. Nelson, K. Concept, word, and sentence. Psychological Review, 1974, 91, 267-285. Nelson, K. Structure and strategy in learning to talk. Monograph of the Society for Research in Child Development, 1973, 99. 127 Nelson, K. & Benedict, H. The comprehension of relative, absolute, and contrastive adjectives by young children. Journal of Psycholinguistic Research, 1974, 9, 333-342. Olson, G. M. & Laxar, K. Asymmetries in processing the terms "right" and "left". Journal of Experimental Psychology, 1973, 100, 284-290. Olson, G. M. & Laxar, K. Processing the terms "right" and "left": A note on left-handers. Journal of Experimental Psychology, 1974, 102, 1135-1137. Palermo, D. S. Semantics and language acquisition: Some theoretical considerations. Paper presented at the Psychology of Language Conference, University of Sterling: Edinburgh, Scotland, 1976. Palermo, D. S. Still more about the comprehension of "less". Developmental Psychology, 1974, 19, 827—829. Palermo, D. S. More about less: A study of language comprehension. Journal of Verbal Learning and Verbal Behavior, 1973, 19, 211-221. Piaget, J. & Inhelder, B. The child's conception of space (1948). Translated by F. J. Langdon & J. L. Lunzer, London: Routledge & Kegan Paul, 1956. Rosch, E. On the internal structure of perceptual and semantic categories. In T. E. Moore (Ed): Cognitive development and the acquisition of language, Academic Press, 1973, 111—144. Rosch, E., Mervis, C. B., Gray, W. D., Johnson, D. M. and Boyes-Braem, P. Basic objects in natural categories, Cognitive Psychology, 1976, 9, 382-439. Sinclair, H. The role of cognitive structure in language acquisition. In E. Lenneberg & E. Lenneberg (Eds): Foundations of language developemnt, Vol. 1, 1975, Sinclair, H. Sensorimotor patterns as a condition for the acquisition of syntax. In Huxley & Ingram (Eds): Language Acquisition: Models and Methods, 1971. Wales, R. & Campbell, R. On the development of comparison and the comparison of development. In G. Flores d'Arcais & W. Levelt (Eds): Advances in psycholin— guistics, North Holland Publishing, 1970, 373-396. Winer, B. Statistical principles in experimental design. McGraw—Hill Book Co., 1971. llHlllllllHHIIIHINIIHIHIWIIIIHHIIHUHIHHIIIHill