Hl’: ”Tl‘I'III‘ WW IIIII It: 1" “I TIbl" WIVIIIIIIIII‘ ':l ' i .I . .. . ‘ v H '. .. I II'III III III. 'III':-iI.'II' «II In. I " ’ :I. ‘. ‘ ' ._.I'I:- .l I ‘1? J r -,-.. ‘ o d 3..- - .. o , . _ " __1 .. ”...... .--. 4: 'I II"."‘ II IIIIQI.‘ I'I'I""II-I.I'II!II?I'I- . 1' . . ... "r -' f“: :4..3I'?:—--:':..-_> "I "'1' II' IIIIII I‘zIIIII 'II‘IIIII' "IIEI" . .' ' . '_I‘ -li ~ » - . _. ,II I I 'iIII'I I'IlIIIII‘II'IIIIJ'III.‘."IIIIVI-r. . . -. . -- . I ”I. . ~- III‘II'IIH -I‘“"III'I5':5III-""-‘ I"? I I I I I 1' g - 53.94 "'II'III" flf: . I IIII lg... I - “I. . .I.IIIIllImII-‘IIII.III‘II .I."II'I I I'..'.""",€'I’ 0 III I'I'I'I .I 'IIIII‘IIIII IIIIIfI‘III " I I'll" II ”(‘1 'III “I“ .I. 1'” III 'III;III I" ' |.-’||'~' If\’1!] ”II. I‘! I I“. .‘|‘\ ... . 3 I'II III III ".I';! III” "'II II. IIILIIIII iI'II IIII" I III III'IIII'Il I I I III” “III; Il’IIIIIIIIIIIIIII . lIlfl In; I. l I. Ill “BI ll 'gl' lI W} rm! Pl't ”I lIIII'w I IIILl lIlII [I .lw I I VII hIII lIIlIMI IlII Ml I SIIW . fiI:Ifif1:f‘Ilw 'lL IIIIIII IIII .I’ IJIHI II I (III III.“ “1‘l'1 ' LI III‘ I I II I I III III IIIII'IIIII IIII IIIIIIlIIIIII IIIllIQ IIIIII ‘III (1;:an ' ‘Iigth-IEI j Sifigtif: o . I _ - _ ‘5 . . ‘ .0 ‘ _ ’ . Til ‘ aging" : l ‘ _ . . . II 'I 51' ; _.I. u I‘: :I‘I (:ch ‘ l 1 '1'. I r- ' . .._ 1‘. I.“ -..,I« - . . -.. I "42...: I ‘ it; . .1 ' . I‘ ...!I. O " . J 0“ 1 ': 'lafi‘ 13" I p I ' . | L— '0 ’l.'. . I F P‘" A(.b.- L. r;” ' i I a ‘ I I ' o I 0 II ' .‘I':." . g . . ‘ I :r' 0‘ . :1 ufi‘ ' '-' \r ' ‘n . \‘l A - hf: Jig. "I‘l 5"! I I. - c t “I IV’I . 'I | . _ III:I‘:II.‘Irb-;.:"; :‘ - ‘ ' h ‘.‘5 ‘I ‘ “'1' . .‘6' . I ' 3 ‘ v. . : .. ‘ i .' 0;. u q ‘ Il Io‘ I a . I \ O 1 D t...:';I ' .. u I‘ ' ‘ I . ‘ I . 3 . . ‘3 ° _ .»’.‘l ’ ‘1 .'OI..I ’ '-'\.. 'G‘ ’ , . .1. L:' :f’. : '.’ ., i . , b. 1 J : . ‘ ‘ ..x. ‘ . . .- -’ Q 7|! IlIH i} '0 I HII IL Iv 'l. . . ..I. II .. II IIIIEI 'II ”III: . :3 .I--. .. «III IIIIII II'II.' I.-' In . I: 3.3m IIIIIIIII‘..£.~ IIII’. .. I; ' I' may.” I”. I III' III; IIII‘IIIIIlII. III I - I . . I I .. ‘ . .: . II'I‘IIII'IJIII . IIII'I. I I’IIII Ir" ,IIIIl IEIII II :II-[III l} IIIfIJhlIII‘II 3." I -T . II: III1L""21 .II ..?°I--'.III.IIII"I'IIIIIIIEITI'Lfitz.‘ 37":II " ., .II . . . . I. .I: .. .... . III. .I... 'I IIIII'II'II":I II’ III III IlI'III'I' ' ..EIIIIIfiIITIIJP‘II ..’=." . .. . . . . , Ill! ‘ III' ...,” {I 0 l". ;\ II.“ !I;j.:.|' ’ _. {.15}: t ”(I o. ' .. I ;‘ .wk . . .II. . . . . .3, :1) 'E‘ ’3‘. ,2: .. 2.1:: :L; I ;. . ....III. ..I......II‘:I I IIzlzs—Ir-III I I. II. . -- . '1 .- .- "I ' I Ivy. I3.” :3 III In... I. III ‘ l 3' .' .' .. .. I" . II‘I III :I'II’I .EIj'I III” II IL’. ':'_ "I” II" ”lib? .' I'., Q IIII I I I ' II IIIIII ' I ‘ I I o I I. : . o i I- I I'1;III.I‘I 'I I---'-‘II’III} {I IIIiIII IIIIIIIII ‘I'IIIE IEII:I§t: ”l h“, Il 1' 'I‘ I ' II 'I‘II IIIII I'II'IIIIIIIII‘IEIIIII 'IIII"II'IIIII2:::II'I“.I-.‘I’IIIIIIVIII' I"? . .. I-L .f;,3 . I "’l'-I.'I'I'.I;'I2'III~I ”I... IlII'ETIIII‘IIIIII‘IIIII I I II I lIII .‘ {Ir .z‘I. “’1. Ir MI'II “. , . .0 I:' . IIIII r‘ .IIII I; -.I'll" IIIlI' I'III EIQI'IlI-‘tlIIil 'I' "II I '. Il‘.‘ ‘ nul- II'IIl' II... IIIIIIIJIIIIRII‘I'I :1 1‘3: l ' IV . .'If I:II| ‘ I': II'f'I,:0VI| 0.:3'.l"l:.10.l ‘m.' 'n I 3" I. . 1 _ J I I I III: ;I‘I III II I:,"I f IIIIIII'II‘lIIl:I II Ql:I ;gI ¢;.'.I' II I. I . , I 'I .I; I' :j'I'I 3;; men,” ::..I: bI‘kIlszI-lIfIlIinI iE ’ I [:5. I I II1I' . l'l.,.::. .I.[' l .INJ. 2""!.| ”.3!” [I I.' I" . g 0. I'. .I . .I.lo.1,.1l.‘ l' IhuI ' 4. 'IlII.‘I . .. . . A - I H" . v.‘ :1 "I.“ ' .fIit‘“ .‘I tr: ' .. ....I I. - ... ... ‘I I' '- ;. '."?.I;I#'I..'I Ii-IIIIII '.. ' ~ ‘IIZ‘M '1 j j I . f“... g; Ié'II‘I. I 1; :i'IIijl, . ,1 ':._"-I'I-I II“. J 'I I' .' "I I I .-H‘- IIII-I' - .'II ’r‘ J1 - 'I- * 1-? . . :.: I-.. .' T’IIII'IFII" '.-..III:::.. ' I... .. =- “H.“ .m‘ I I‘, III ..‘fij-jfiIjI-fI IlEIIIII 11"l‘l3'3l; I - I' '. IlI ‘- I 14;?” ........ ' r. I'II -| I. . ' .;I 'lII" -‘- ‘ .I‘I’ HI. I. . II v 1 I - - . . . . III: I I I... II - . : “1 .«r...-.:I.....I..:I‘I:‘;II "I‘ I - I . IIIIII. I.” .l'. IIII:III.‘IIII.I'I . .II} I :I..; HI. . g- 1': f. I .‘ '_ 1 . _ ’I I":..'.I I.l IIII ' I ‘I l ILII 11 I”, ‘ I Q; . i . v I I ' , ' I I ' I . I .II.‘ III I . :_ -.’ I. ‘I I a H“ ...... ‘. ‘ ... .. I Y' t .l . III. II. 'II‘IIII I2}: ..Il I‘III- .. .I'II-I I .I.I . , I“I;II’I III l . ' . .II “0 'I_Ilill"' II I II. ; .T .' -."I‘:I':I. I .;:‘.°.--1 IiI"'I:I.IIII *I' . ' ’ '. . ‘ . 0 II ‘. '.° . I . ,.' I II" ' 'I ’ I g s . . .II'I ‘6 .II., I I."l ’II I. :IIIIII'I'I.IIIl ‘.I.l:'IIIII I ' u ‘.{U£:I t? . ., I .' I ' I. I" :I'I‘It'f II III‘IIIlII " IIIIIIIIII'I ‘ I . ‘I I, I.I I II III'..0:'I"°' ‘-|3 ' .. :- I -’. I'.'_ I'.‘. I III- ..lIIIl '[II *I lII (I IIIlII'IIlIII . .. . ..l; .II’ l II. .' .. .II'II'I 3' . IIIwI III IiI"II"I" . .. II ' ' 1 ' .‘ IQI I I' _I I‘III'I II " . ' .. l . I’;"’. I I' I In. a . ;. '. I I _. I. .IIIIIIII II ,I I.‘f IIIIII.. 'Ill lIlI. I I I ”II ‘II ' l I f , . , '., lllI‘IIfI-l lI Il‘llIII I.’ I llIIIoI . . '3 III'.I-;I III III I II ‘ " ' ‘ "II‘ II IIIII II‘ IIIIIl‘III "II I I . o h I III lIl llI IlIlII III! II‘II I I] I“ ll'l.l']'I'I.l'l"l 'tI '. a I .I l .. I II. - I ' I III .I I III .II]. IlIl' I‘d! IllIIH lIIII ll.“ I”; I' I' II'. I I I ". It I ;' - I II . I_, Il l.I_ l l , I I Il I ' l.I-Ill : II 'I . - _ . I . II I .I ‘. I . I I I III. I' ‘ I I H II {III IIIII.I III I -* ' . I "I'- ""I'III'I' I. III III ' . .. I I I“ III H IIIHI II'II I“ I II ‘I'J‘IMIII I I |*l I‘ “I I .I i. n I 'I I 1 1‘ ‘I LIBRARY Mich‘rvm State This is to certify that the thesis entitled HUMAN SPATIAL LEARNING: THE EFFECT OF MOVEMENT PATTERNS ON KINESTHETIC AND VISUAL REPRESENTATIONS OF A KINESTHETICALLY EXPERIENCED SPATIAL LAYOUT presented by MICHAEL SEAN DENNY has been accepted towards fulfillment of the requirements for PA ' D. degree in Mt] '32“ jor professor Date W 0-7 639 _r.—A.fi-—.-'_.‘ - N... 'v_.—. -fi. -0 -.- -1q- 'TAO' . -'-- - -..---" HUMAN SPATIAL LEARNING: THE EFFECT OF MOVEMENT PATTERNS ON KINESTHETIC AND VISUAL REPRESENTATIONS OF A KINESTHETICALLY EXPERIENCED SPATIAL LAYOUT BY Michael Sean Denny A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Psychology 1977 ABSTRACT HUMAN SPATIAL LEARNING: THE EFFECT OF MOVEMENT PATTERNS ON KINESTHETIC AND VISUAL REPRESENTATIONS OF A KINESTHETICALLY EXPERIENCED SPATIAL LAYOUT BY Michael Sean Denny The study investigated the acquisition and transfer of Spatial knowledge gained from varied kinesthetic exPerience. The kinesthetic experience was structured to facilitate; (1) serial or route learning (uniform condition), or (2) simultaneous or map learning (random condition). The acquisition task was a series of blind hand movements between six points arrayed in a planar space. The sequence of cued movements was either: constant between the six points, and thus, totally predictable (uniform condition); or it was semi-random, consisting of three different move- ments to each location (random condition). Following 96 or 192 movement trials, a new movement sequence was introduced to 80 subjects. This transfer sequence, in part, consisted of previously unexperienced interpoint movements. Another 100 subjects were shown a series of visual images depicting the Spatial layout of points as a configuration of six Michael Sean Denny small circles. The images were projected onto the training surface in congruency with actual kinesthetic point loca- tions; or they were projected onto a remote vertical screen. Visual recognition of the original spatial layout was measured by the accuracy and speed of detecting the one misplaced circle (point) in each image. There was no evidence that route learning occurred. The mean interpoint movement times during acquisition and kinesthetic transfer (new sequence) indicated that kines- thetic point locations, and not Specific hand movements, were learned. The locations were learned equally well under either sequence. The visual transfer results also failed to support a route learning hypothesis. Recognition- detection errors emphasized an egocentric contraction of point locations, especially following the uniform kinesthe- tic experience. The finding was interpreted as an appli- cation of egocentric reference to kinesthetic data. It was preposed that the reference system was preserved under visual translation and systematically distorted the visual representation of the Spatial layout. Evidence for map learning was found in acquisition and visual transfer results. Correlations of movement times to the various points revealed more integrated and predictable performance on the random training sequence. The same subjects made less errors on the visual transfer test and, after 96 acquisition trials, showed no distortion in their visual representation of the Spatial layout. The Michael Sean Denny findings were interpreted as an application of field reference to visually translated kinethetic data. It was proposed that both movement sequences led to the Simultane- ous development of kinesthetic and visual representations of the space; yet, only the random experience led to a visual Spatial schema sufficiently undistorted to guide interpoint movements. TABLE OF CONTENTS Page LIST OF TABLES . . . . . . . . . . . . . . . . . . . . vi LIST OF FIGURES . . . . . . . . . . . . . . . . . . . Vii INTRODUCTION . . . . . . . . . . . . . . . . . . . . . 1 Serial Versus Simultaneous Representations . . . . 3 Sensory System Specificity . . . . . . . . . . . . 7 Information Specificity . . . . . . . . . . . . . 10 Experimental Objectives . . . . . . . . . . . . . ll METHOD 0 O O O O O O O O O O O O O O O I O O O O O O O l 6 Subjects . . . . . . . . . . . . . . . . . . . . . 16 Apparatus . . . . . . . . . . . . . . . . . . . . 16 Design . . . . . . . . . . . . . . . . . . . . . . 20 Procedure . . . . . . . . . . . . . . . . . . . . 22 Kinesthetic task . . . . . . . . . . . . . . . 24 Visual task . . . . . . . . . . . . . . . . . 25 OVERVIEW OF RESULTS . . . . . . . . . . . . . . . . . 28 Summary of Acquisition . . . . . . . . . . . . . . 28 Trials . . . . . . . . . . . . . . . . . . . . 28 Sequence . . . . . . . . . . . . . . . . . . . 28 Conditional probability of segments . . . . . 3O Interpoint distance of segments . . . . . . . 30 Points . . . . . . . . . . . . . . . . . . . . 31 Summary of Transfer Tests . . . . . . . . . . . . 32 Kinesthetic transfer . . . . . . . . . . . . . 32 Visual transfer . . . . . . . . . . . . . . . 32 RESULTS AND DISCUSSION . . . . . . . . . . . . . . . . 35 Acquisition . . . . . . . . . . . . . . . . . . . 35 Segment probability . . . . . . . . . . . . . 37 Segment distance . . . . . . . . . . . . . . . 50 Correlation of performance among points . . . 60 Performance profile for points . . . . . . . . 66 ii Ph‘ JL. Page Kinesthetic Transfer . . . . . . . . . . . . . . . 78 Reaction time . . . . . . . . . . . . . . . . 79 Movement time . . . . . . . . . . . . . . . . 85 Processing time: an aside . . . . . . . . . . 89 Visual Transfer . . . . . . . . . . . . . . . . . 93 Displacements . . . . . . . . . . . . . . . . 97 Reference systems . . . . . . . . . . . . . . 104 Analysis by point . . . . . . . . . . . . . . 108 Individual differences . . . . . . . . . . . . 117 Remote visual test . . . . . . . . . . . . . . 124 GENERAL DISCUSSION . . . . . . . . . . . . . . . . . . 129 Location or Route? . . . . . . . . . . . . . . . . 129 Integration of Location Information . . . . . . . 135 Visual or Kinesthetic? . . . . . . . . . . . . . . 137 Routes and Visual Schemata . . . . . . . . . . . . 140 Other Issues . . . . . . . . . . . . . . . . . . . 147 A Model . . . . . . . . . . . . . . . . . . . . . 149 APPENDIX . . . . . . . . . . . . . . . . . . . . . . . 155 Apparatus Construction and Operation . . . . . . . 155 Signal light and tone . . . . . . . . . . . 156 Enclosure base and kinesthetic field . . . . . 157 Projection system and visual field . . . . . . 160 Movement sequences . . . . . . . . . . . . . . 162 Instructions and Procedure . . . . . . . . . . . . 164 LIST OF REFERENCES . . . . . . . . . . . . . . . . . . 168 iii 10 12. 14. Table 10. ll. 12. 13. 14. LIST OF TABLES Group Size for Experimental Conditions . . Acquisition Block Means by Sequence . . . Acquisition Block Means for Random Groups by Segment Probability . . . . . . . . . MANOVA of Reaction Time (sec) on Uniform and P.5 Random Segments . . . . . . . . MANOVA of Movement Time (sec) on Uniform and P.5 Random Segments . . . . . . . . MANOVA of Reaction Time (sec) on P.5 and P.25 Random Segments . . . . . . . . MANOVA of Movement Time (sec) on P.5 and P.25 Random Segments . . . . . . . . Acquisition Block Means for Random Groups by Segment Distance . . . . . . . . . . MANOVA of Reaction Time (sec) on Short, Medium and Long Random Segments . . . . MANOVA of Movement Time (sec) on Short, Medium and Long Random Segments . . . . Average Interpoint Correlations of Acquisition Scores . . . . . . . . . . . Acquisition Means for Points by Sequence and Length of Training . . . . . . . . . MANOVA on Profile of Reaction Time (sec) to Points Averaged Over Early and Late BIOCRS O O O O O O O O O O O O 0 O MANOVA on Equality of Reaction Time (sec) to Points Averaged Over Early and Late Blocks . . . . . . . . . . . . . . iv Page 23 38 39 42 44 45 47 52 54 57 62 67 70 71 Ta: 17 all. 15 22 «(a fi/s 24 26 27 28 29 Table Page 15. MANOVA on Profile of Movement Time (sec) to Points Averaged Over Early and Late BlOCkS O I O O O O O O O O O O O O O O O O O O 75 16. MANOVA on Equality of Movement Time (sec) to Points Averaged Over Early and Late Blocks . . . . . . . . . . . . . . . . . . . . 76 17. Performance on Kinesthetic Transfer Test 0 O O O O O O O O O O O O O O O O O O O O 80 18. ANOVA of Reaction Time (sec) on Kines- thetic Transfer Trials . . . . . . . . . . . . 81 19. ANOVA of Movement Time on Kinesthetic Transfer Trials . . . . . . . . . . . . . . . 87 20. Performance on DiSplacementS in the Visual Transfer Test . . . . . . . . . . . . . 98 21. ANOVA of Error Rate on Displacements in the Visual Transfer Test . . . . . . . . . . . 99 22. ANOVA of Reaction Time (sec) to Displace- ments in the Visual Transfer Test . . . . . . 101 23. Performance on Points in the Visual TranSfer Test 0 O O O O O O O O C O O O I O I 109 24. Partial ANOVA of Error Rate on Points in the Visual Transfer Test . . . . . . . . . 110 25. Partial ANOVA of Reaction Time (sec) to Points in the Visual Transfer Test 0 O O O O O O O O O O O O O O I O O O O O 111 26. Overall Performance on Visual Test by Sex of Subject . . . . . . . . . . . . . . . . 120 27. Partial ANOVA of Error Rate on Visual Test by Sex of Subject . . . . . . . . . . . . 121 28. Partial ANOVA of Reaction Time (sec) to Visual Test by Sex of Subject . . . . . . . 122 29. Comparison of Overall Remote and Proximal Visual Transfer Performance . . . . . . . . . . . . . . . . . 126 Table Page 30. ANOVA of Reaction Time (sec) on Remote and Proximal Visual Transfer Tests . . . . . . 127 31. ANOVA of Error Rate on Remote and Proximal Visual Transfer Tests . . . . . . . . 128 32. Relative Position of Points from Field Center . . . . . . . . . . . . . . . . . . . . 159 vi '1) H. ‘0 IN) Figure 1. 2. 3. 10. 11. LIST OF FIGURES Four Views of the Apparatus . . . . . The Layout of Points Acquisition Trend of Uniform and Random ments 0 O O I O 0 Acquisition Trend of Uniform and Random ments 0 O O O O 0 Acquisition Trend of Distance Segments Acquisition Trend of Distance Segments Acquisition Trend of Reaction Times to Probability Seg- Movement Times on Probability Seg- Reaction Times to Movement Times on Interpoint Corre- lations of Performance Scores Sepa- rately for Uniform and Random Sequences Profile of Reaction Times to Points Averaged Over Early and Late Blocks of Uniform and Random Acquisition Trials . . . . . . Profile of Movement Times to Points Averaged Over Early and Late Blocks of Uniform and Random Acquisition Trials . . . . . . Reaction Time to Kinesthetic Transfer Trials Relative to Acquisition Per- formance Under Uniform and Random Conditions . . . . Movement Time on Kinesthetic Transfer Trials Relative to Acquisition Per- formance Under Uniform and Random Conditions . . . . vii Page 17 19 40 43 53 56 63 68 74 83 86 Figure Page 12. Profile of Error Rate on Points in the Visual Transfer Test Following Differ— ent Acquisition Conditions . . . . . . . . . . 113 13. Profile of Reaction Time to Points on the Visual Transfer Test Following Different Acquisition Conditions . . . . . . . 114 14. A Model of Processing Kinesthetic Infor- mation Which Achieves Memorial Repre- sentations at Different Levels of Spatial Reference . . . . . . . . . . . . . . 150 15. A Model of the Selection Process for Retrieving Stored Spatial Information Leading to a Movement Response . . . . . . . . 153 viii :71 D) U) I'h In INTRODUCTION The memorial representation of spatial arrays is an ubiquitous concern among psychologists--often for quite different reasons. Cognitive psychologists like Posner and Attneave have been concerned with spatial representation as an example of the transformation or encoding of information which occurs during learning and memory processing (Attneave and Benson, 1969; Posner, 1967). Coming from a different direction, Piaget has consistently emphasized the central position of spatial representation as a dimension of cog- nitive development. Even verbal learning theorists refer to spatial representations as a factor in the organization of memory, particularly in the context of serial learning and mnemonic strategies (Huttenlocher, 1968; Luria, 1968). Such representations are also held as indicative of the way humans operationally structure their environment, be it from a behavioral (e.g., Downs and Stea, 1973) or a cogni- tive theoretic viewpoint (e.g., Palmer, 1975). Spatial cognition as a component in animal maze learning was con- sidered by Woodworth (1938) and vigorously pursued, empir- ically and theoretically, by Tolman (see Tolman, 1948). Cemparative and ethological psychologists have similar con- cerns about how animals assimilate the Spatial features of their habitat (e.g., establish and maintain reference to nesting and feeding sites). Researchers like Luria and Semmes have been con- cerned with the neurophysiological locus of spatial repre- sentations, and the general concept has been important in differentiating hemispheric function in the human brain (Luria, 1969; Semmes, 1968). Work is also progressing on the animal level (see Olton, 1977). On a more applied level. Spatial cognition has become an increasingly important factor in urban and architectural planning. The construct also has been applied to the study of normative and patho- logic behavior in habitual and institutional environments (e.g., Hall, 1966; Ittelson, Proshansky and Rivlin, 1970). The representational process also has been of concern in the study of instrument design and human performance, the creative process (e.g., Arnheim, 1966), and in efforts to facilitate the mobility of the blind. The present study considers some of the major questions currently being raised in connection with spatial cognition, particularly those relating to the formation of stable representations of serially experienced stimulus arrays. The study was initially motivated by a concern with how environmental information is acquired and processed to form a unitary representation of the spatial layout of elements in the environment. The representation was pre- sumed to reflect not only the actual geometry of a space but idiosynchrasies in experiencing the space. Efficacy of the experiential variable is a rather old issue (e.g., Bartlett, 1932) which is being considered in a number of current approaches to spatial cognition. Serial Versus Simultaneous Representations It has been suggested in a recent review of spatial cognition that the spatial qualities of an environment are initially organized by the sequential linking of the ele- ments within the environmental array (Hart and Moore, 1973). Thus, in the early course of learning, a spatial layout becomes represented as a serial configuration of points. Hart and Moore consider this form of spatial representation to be a product of the processing involved in the perceptual analysis of a sequenced exposure to the space. This assump- tion requires that the processing involved be sufficiently integrative to provide the contiguous associations between points necessary for the formation of a stable represen- tation. Whether the associations between points are medi- ated by movement responses, in keeping with S-R theories, or whether the associations are direct perceptual links, as field theories would have it, is not specified. A strictly serial representation, however, is more easily predicted from an S-R interpretation. It is clear that there is always, to some degree, a temporal—geometric serialization involved in experiencing a spatial array, be it walking through a metrOpolitan area or viewing a picture postcard. an th in SC ma th Thus, while Hart and Moore's suggestion appears to reflect an S-R viewpoint, it is primarily the physical constraints of the environment and human sensory machinery that dictate such a position. The serial nature of spatial representations, of course, is not merely an artifact of the learning paradigm. This style of organizing information is particularly con- sonant with the demands usually made on such a represen- tation. For example, one needs to know what is connected to what, or how to get from here to there. On the other hand, the serial information which can guide one through a space may be accompanied by a simultaneous representation of the unitary space. It is assumed that the detail of this representation, or "spatial schema," is highly variable and may be insufficient to guide one's operations within that space. However, as experience with the environment increases, so does the quality and consolidation of the schema (Stea and Blaut, 1973), so that eventually the schema may serve the needs of the operator more efficiently than the serial representation. Indeed, the schema can provide (code) geometric information more efficiently and be more easily accessed than a serial representation (see Attneave, 1972). The concept of schema used in this paper was sug- gested by Lee (1973) and refers to the memorial structure expressing the "collective unity of a temporal sequence of actions" or "spatial patterns" (Lee, 1973, p. 98). Thus, a spatial schema integrates spatial data derived from dis- crete perceptual contacts with the space over time. If, as the course of learning progresses, the spatial representation is inferred to shift from a serial to a simultaneous nature, two possibilities exist. Either the two kinds of representations develop in parallel (with or without interaction) or the simultaneous is a direct consequence of the serial antecedent. In their review on the development of spatial cognition, Siegel and White (1975) opt for the latter possibility. They contend that all spatial representations are functionally "landmarks- connected-by-routes" which progress from association to structure in a unidimensional manner. Thus, for them, spatial representations vary only in their degree of inte- gration. Ultimately, given sufficient sensorimotor expo- sure, all spaces become uniquely represented as network- like schemata. Theirs is a rather strict interpretation of Tolman's idea that the transition from a serial to a simul- taneous representation is an integrative process where the "comprehensive-map" of an experienced environment is built up from "strip-maps" of serial exposure (Tolman, 1948). This postulation derives some support from the fact that this transition is analogous to the presumed development of general spatial knowledge in the child as the cognitive structure of his environment is formed (Piaget and Inhelder, 1956). The serial-to-simultaneous View, at this point, is oversimplified. Early human maze studies have shown that an initial stage of general orientation learning is common (Perrin, 1914; reported in Miles, 1931) which is followed by response sequencing (Brown, 1932). Whether the initial spatial schema is exploited and elaborated is a subject variable (Warden, 1924) and a function of the heterogeneity of responses required by the maze's complexity (Scott, 1936). The development of a schematic, Simultaneous repre- sentation is most common when the task is truly spatial and not ameliorated by verbal or other serial process mediation. In the face of meager data it seems safer, though less parsimonious, to reject the idea of representational singu- larity in favor of parallel, but not independent, repre- sentations. This approach still does not explain the transition, if it occurs. One possibility is that the use of a particular representational form is a strategy decision affecting performance but not necessarily learning. In a maze study, for example, both a serial representation and a schema of the maze pattern might develop over trials. Yet, if a subject forms a bias by consciously attending to one representation, then, during a task where the other representation is more effective, his performance will suffer. Recognition of the inappropriateness could result in a shift of attention to the other representational form. Sensory System Specificity It has been suggested, on the perceptual level (Gibson, 1966) and the cognitive level (Pylyshyn, 1973), that the nature or code language of the spatial represen- tation is relatively independent of sensory system codes. These authors propose that the representation exists at a processing stage which is integrative and abstract to the extent that distinguishing sensory characteristics have been obscured or discarded. This position is not without its problems. For example, it is at odds with the findings of Posner who has demonstrated different memory functions for kinesthetic and visual spatial information (Posner, 1967). In general, attempts to demonstrate amodal coding of stimulus information have been inconclusive. For example, in a well controlled study, Shaffer and Ellis (1974) found only minimal stimulus learning transfer in a vision-to-touch form recognition task. Another approach to the issue of whether spatial representations are characterized by modal specificity has been forwarded by Attneave and Benson (1969) and Connolly and Jones (1970). In separate studies a ready transfer from tactual or kinesthetic sensation to a "visual" repre- sentation was found. The transfer was considered to be a nearly immediate process of mapping or translating the input into a spatial-visual form. The major difference between this approach and the aforementioned one is the level at which the transformation occurs. This dictates such features as the "visualness" of the spatial repre- sentation and the modal specificity of the coded space as represented in Short-term memory. In other words, if the representation is derived from stimulus attributes which are initially coded invari— antly across sensory systems (amodal code), then the infor- mation in short-term memory should be amodal as well. This means that the content of Short-term memory is no more like a visual trace than it is like a kinesthetic trace or a haptic trace. On the other hand, if the stimulus is coded in terms of information specific to the sensory system being activated and then translated into other forms before a representation is formed and held in short-term memory: then the content of Short-term memory is more like a visual trace than anything else, regardless of input modality. If the translation does not occur it is more like the modal trace associated with the sensory system being activated. Note that the last interpretation assumes that the operational form of a Spatial representation is essentially visual. The, once popular, notion that the core of spatial representation is visual recognition memory has been much qualified in recent years (e.g., see Siegel and White, 1975). The original conclusion was based, in part, on the assumption that gestalt, or "survey," representations require by definition the simultaneous quality of hen .3 m: sem file] 636 1111 the and information organization that typically is assigned only to the visual system. This assumption, however, is easily challenged. For example, current neuropsychological theorizing on the contrasting roles of the two neocortical hemispheres does not restrict simultaneity to the visual system. The parieto-occipital and temporal regions of the right hemisphere are strongly implicated in spatial oper- ations (see reviews by Harris, 1975a, 1975b; Milner, 1971). Semmes (1968) and Levy-Agresti and Sperry (1969) assert that gestalt processing occurs in the left hemisphere, and serial or analytic processing occurs in the right. Semmes goes on to hypothesize that the right hemisphere is functionally endowed with a diffusely organized synthetic processing capacity to deal with heteromodal integration. If the hypothesis is accepted it becomes easy to explain the existence of spatial gestalts in other than visual representational terms. That is, if all information that reaches these areas of the right hemisphere is susceptible to synthetic integration, the simultaneous amodal repre- sentation may be a rather automatic consequence of hemis- pheric organization. Extending the argument suggests that a serial analytic representation would exist concurrently in the left hemisphere. This notion is consistent with the Hart and Moore view and less consistent with Siegel and White's unidimensional interpretation. The idea of 10 nonvisual schemata has notable proponents (e.g., Bartlett, 1932; Neisser, 1967), and experimental support is growing. For example, Schmidt (1975) presents evidence for motor schemata existing in both recall and recognition memory systems. Information Specificity Regardless of how nonspecific or heteromodal the representation of space is, the quality, or information content, of the coded spatial relations is undoubtedly subject to certain constraints imposed by the internal organization of the processing mechanisms. What losses or distortions occur during processing are not known, nor is the basic structure of the resulting schema. Attneave, in an attempt to deal with the latter point, has suggested that the schema, in its highest form, is isotropic or analogue (Attneave, 1972). The work of Shepard and Metzler (1971) has provided an example of what Attneave meant as analogue. They have Shown that the time required to match rotated block designs in three dimensional perspective is a linear function of the degree of relative rotation, regard- less of whether the rotation is in the picture plane or in depth. They hypothesize that the subjects are represen- tationally imposing a continuous rotation on the visual design (image). The notion has met with a certain amount of skepticism about the probability that a continuous (analogue) operation can be applied to data which even at 11 the most primary levels of visual processing are charac- teristically discontinuous (Newell, 1973). More recently, Shepard (1975) has suggested that the rotation process may be less than analogue, in consideration of the likelihood that the image is less than isomorphic with the three— dimensional form it represents (see, Shepard, 1975, for a discussion of second-order isomorphism). A comprehensive treatment of schema.structure within a developmental framework has been made by Piaget and Inhelder (1956). They propose a series of cognitive stages in each of which they assign a different structure to the schema (e.g., topologic vs. projective vs. Euclidean). Piaget and Inhelder assume, not unreasonably, that if these distinctions are real, more than one of the structures will appear in the spatial information processing used by adults. The pr0posed coding systems for spatial relations, however, may not be distinct or meaningful on a mathematical level. For example, the existence of a real topological system is in doubt because empirically a uniquely ametric spatial representation is unsubstantiated (see Petkovich, 1974). Experimental Objectives The questions listed below are central to the issues of Spatial representation raised in the introduction. It is believed that, when formulated as hypotheses, these questions predicate a rational experimental study of the acquisition and structure of spatial representations. The 12 questions presuppose that the space is experienced in only one sensory modality by movement responses between discrete units of the space. In this way the response becomes iso- morphic with a spatial relation. That is, the relationship between two points in space is defined by the nature (angular direction, speed, duration) of the movement response. If we assume that a spatial representation must be, in its best form, a structured set of relations, then the acquisition of this representation can be determined by the acquisition of a set of movement responses. The most important questions, then, are about the factors which control the acquisition and transfer of space-related responses. For example: 1. What are the effects of response frequency on acquisition? 2. What are the effects of the serial order of responding on acquisition? 3. What determines the efficiency of the intramodal transfer of acquisition? What roles do frequency and serial order play in the generation of novel responses? 4. Does the acquisition of space-related responses yield consistent response patterns in other sensory modalities? What determines the efficiency of this intermodal transfer? 13 Let us consider, in the context of these questions, a strictly planar space defined by the geometric relations among a limited number of discriminable points, bounded by a loosely defined perimeter, and otherwise unpatterned and homogeneous. The present study was designed to (l) analyze the course of learning this planar Space as the spatial array is sequentially SXperienced, and (2) at the same time be sensitive to performance factors indicative of schema consolidation. These objectives were achieved by limiting the kinesthetically—based spatial experience during acquisition to specified pathways requiring the generation of novel responses during testing. Further, the objectives were met by isolating components of response performance (reaction time and movement time) indicative of place learning and route learning. The experiment was designed also to (3) investigate the structure and content of the resulting schema. This objective was achieved by restrict- ing the input of spatial information during acquisition to kinesthetic cues and limited exteroceptive signals, before introducing a visual recognition task designed to identify schema organization. To contrast route or serial learning with place or schema learning, two patterns of movement through the Space were established. In a "uniform" condition, a simple serial order of movements between points in the space was used. In a "random" condition, the movement pattern was more 14 complex and not readily predictable. If route learning occurs, it should be pronounced in the more uniformly experienced space. On the other hand, the less constrained random pattern should facilitate place learning and schema formation. In one study, better transfer on a kinesthetic task was achieved after multiple versus single patterns of training (Duncan, 1958). To investigate the idea that Spatial representations progress from serial forms to schemata, the amount of experience gained by moving about the spatial layout was varied. In addition, the visual or visual-like component of the representation was examined for evidence of spatial referencing between adjacent points along a "route" and referencing between a point and the subject's position relative to the field of points. If the space is represented in visual terms, a test for dif- ferent reference systems should provide a sensitive method for contrasting different forms of learning. It is postu- lated, for example, that route learning specifically involves creating spatial links or axes of reference between adjacent points in the pattern of movement between points. It follows that the visual representation of these routes should be more precise or more clearly defined than the representation of other relationships within the space. Admittedly, the experimental task used in this study represents an abstract version of everyday environ- mental experience. In one respect, the difference between 15 the task and everyday behavior is critical and reflects a major constraint on the generalizability of the findings. Specifically, in this study, the subject remained station— ary--moving about in a space solely by arm movements. This introduces a system of reference, namely egocentricity, which is not constant when one's whole body moves through a space. Although this Special feature unquestionably con— strains extrapolations to real-life environments, it does not compromise the ability to answer the basic questions about the representational process of encoding spatial information, given that the information exists with a fixed relationship to the person (body position). Answering some of these questions requires analysis at the motor level of responding. For theoretical purposes, it is suggested that the structure of spatial cognition can be induced from facets of motor learning and behavior. It is believed that this reductionistic approach can expose important stages of processing spatial information which may determine learning strategies and spatial reference systems used by subjects. METHOD Subjects One hundred and eighty students at Michigan State University served as subjects. The 63 males and 117 females were enrolled in introductory psychology classes and received credit toward course requirements for their participation. Subjects were randomly assigned to one of 10 groups under the constraint that the ratio of males to females in each group was comparable. Each subject partici- pated individually in an experimental session that lasted about 50 minutes. All subjects were right-handed to the extent that they expressed a preference for using their right hand to perform the experimental (hand-movement) task. Apparatus The experimental apparatus pictured in Figure l was essentially a rectangular box (.91 m wide by .64 m high by .86 m deep) mounted on a .74 m high table of the same width. The base of the enclosure was covered with white Formica and formed the interior floor of the enclosure. This sur- face served as a planar space (.56 m2) on which a six-point field could be represented either kinesthetically or visually. The Formica floor could be shielded from a sub- ject's view by a removable horizontal panel with a curved 16 17 .chnce COWOOONOHQ M0 kmwb Hmmm uucmwm .ewmwk Hmsmwa M0 newuomhoum Hmexoum "omen .20wuomh0Nm muosmm How embosmm Hmboo access ”annex .mOmHm cw 363m. fie... 36.9 “sea "ammo .msumummm¢ may no m30w> mochil.a .mwm 18 front edge. Centered on the upper surface of this panel was a .28 m radius circle drawn in white. Beneath the Formica floor six magnetic reed switches were fixed and determined the location of the six kinesthetic points. The six points or locations on the Formica surface were undetectable except when a 3 cm diam- eter magnetic disc was positioned over one of the magnetic switches (point locations) producing a brief tone. A switch would activate if the center of the disc was within 1.5 cm of the switch's axial intersection with the Formica surface. The configuration of points is shown in Figure 2. Each point was identified by a color which could be dis- played on the screen of a signal light mounted at approxi- mate eye level on the back wall of the enclosure. The colors used to identify points were pink, yellow, red, green, white, and blue. The interior of the enclosure was always illuminated by low level indirect lighting. A projection tunnel (1.53 m in length) was attached to the top of the enclosure and extended back through a partition in the experimental room. A carousel slide pro- jector was located at the rear of the tunnel. A mirror at the front of the tunnel reflected the light path of the slide projector through a hole in the top of the enclosure onto the Formica surface of the base (with shield panel removed). By removing a front cover and the mirror from the projection tunnel, an image also could be projected 19 .muoflom mo usommq 0581:.m .mam utzz. \ \ \ / ” \\ \ x; / .u / sz/O ‘ , ./ 7.52383 1.4 \ u / . SEzMmeu ....-- \ m / ... L, 20 onto a projection screen (1.52 m square). The removable screen was attached to the wall of the room facing the front of the enclosure. Photographic slides were produced which, when pro- jected onto the Formica surface, produced an image of six colored circles (3 cm diameter). Each correctly positioned circle appeared directly over the position of one of the magnetic reed switches which designated the location of a point. The area of the circle was congruent with the area of sensitivity for the switch. The color of the circle corresponded to the color identifying that point. The colors of the circles matched the colors displayed on the signal light screen. When the slides were projected onto the remote screen, the image size was doubled, relative to the field, and the pattern of points was viewed in the vertical rather than the horizontal plane. Operation of the apparatus, including recording of response times, was controlled by a PDP-8 computer. Further details on the apparatus and its Operation are given in the Appendix. Design All subjects participated in a two-stage procedure consisting of initial acquisition and a transfer test on a six point space. Six points were selected because the number was great enough to configure a space which could not easily be interpreted as a familiar form (e.g., a 21 regular polygon), yet small enough not to exceed the capacity of Short-term memory (span). Acquisition con- sisted of either 96 (EEEEE training) or 192 (lggg training) movement trials between the six point locations on the Formica base. Movements were always from one point to another point. The order of the points was determined by the training sequence (uniform or random) which determined the number of different point-to-point movements made to each point. The transfer test was either kinesthetic or visual. The kinesthetic test was a continuation of the acquisition task under a new sequence. The change in sequence meant that previously untrained point-to—point movements had to be made. The visual test was a pattern recognition task for which subjects had to detect a mis- placed point in visual presentations of the kinesthetic field (layout of points). Experimental conditions were formed from the factorial combination of the two two-level training factors (uniform versus random sequence, and short versus long length of training) and the two-level transfer factor (kinesthetic versus visual). In addition to these eight conditions, two conditions were formed from the levels of the sequence factor combined with long training and a variant of the visual transfer test (remote test versus the proximal test used in the other visual transfer conditions). Ten independent groups were assigned to the experimental cond in I give tic: the ject colc 22 conditions. The design, including group size, is summarized in Table 1. Procedure The subject was seated in front of the enclosure and given a set of instructions. During the course of instruc— tions, the subject was familiarized with the apparatus and the signal light colors used to identify points. The sub- ject was told that the six points, each assigned a different color, were located somewhere on the Formica surface under- neath the circle marked on the Shield. Subjects were told that the goal of the task was Simply to learn the locations of the points. They were not told of a change in sequence or that a visual test of locations would follow. The sub- ject's hand, holding the magnetic disc, was positioned at a starting point on the Formica surface before the room lights were turned off and the series of acquisition trials was started. Kinesthetic transfer groups continued with the kinesthetic task until the end of the session. Visual transfer groups were given additional instructions at the end of acquisition while the apparatus was adjusted for proximal or remote projection. Subjects were told that they would see a series of visual patterns, composed of six colored circles, intended to represent the location of points learned during acquisition. Within two minutes of the last kinesthetic trial, the first visual trial was 23 Table 1 Group Size for Experimental Conditions Acquisition Transfer Test Sequence Length Kinesthetic Visual Visual Remote uniform short 20 20 -- random short 20 20 -- uniform long 20 20 10 random long 20 20 10 Note: total n = 180 SD of 24 presented. At the end of the experimental session, all subjects were informally debriefed. The instructions read to subjects and other details of the procedure can be found in the Appendix. Kinesthetic task. The subject sat in a chair with his chest a few centimeters from the front edge of the shield. The position allowed free movement of the hand over the Formica surface of the enclosure base. The sub- ject could not see the surface or his arm. The task was to move the magnetic disc, held in the right hand, from one point location to another in the minimum amount of time. Each trial began with the presentation of one of the six colors on the signal light screen. The signal directed the subject to locate (move to) the point corre3ponding to the color shown. A successful move required the subject to position the disc at the correct location for at least 500 msec. A tone sounded during the 500 msec interval to signal correct positioning of the disc. Once the correct position was maintained, the signal light changed color initiating the next trial. The reaction time (from signal onset until movement) and the movement time of the location response were recorded on each trial. Until color identity and general location were learned, the location task involved moving over the entire field in search of the 'point. After this initial exploration and learning, the task became one of directed movement. pat eve bot 981 IE] me: Se CC 36 u:- 25 The sequence of color signals that determined the pattern of movements between points repeated itself after every 6 trials (uniform) or every 24 trials (random). In both cases each of the six colors appeared an equal number of times. Thus, in the uniform condition, a point was always preceded in the sequence by the same point. In the course of the 24 trials of the random sequence, each point was preceded by three different points. One of the pre- ceding points was the same as that which preceded the point in the uniform sequence. This meant that the random sequence was composed of 18 different two-point segments, generating 18 different point-to-point movements. The remaining 6 sequence segments were replicates of the seg- ments common to both sequences. Segments were classified by the sequence-wide conditional probability of their two- point precession. The common segments had conditional probabilities of 1.0 in the uniform sequence (Pl segments) and .50 in the random sequence (P.5 segments). Segments unique to the random sequence had conditional probabilities of .25 (P.25 segments). The 30-tria1 kinesthetic test series was composed, in part, of two replicate sets of the common uniform segments and two replicate sets of Six novel segments. The latter generated six different, previously untrained, point-to-point movements. Visual task. The subject sat in front of the enclosure, as during acquisiton, but with the shield 26 removed. The proximal task was to rapidly identify which one of six colored circles, projected onto the Formica surface, did not coincide with the true position of the kinesthetic point identified by that color. Twenty-four patterns of points were presented, with 5 sec delays between each presentation. A brief tone preceded each presentation by 1 sec. The projection remained on until the subject responded by pressing a hand-held key and reporting a color, or until 30 sec passed without a response. The verbal response and reaction time were recorded on each trial. The remote visual task was the same as the proximal except the patterns were projected in the vertical onto a screen viewed from 1.8 m, by subjects, after their seating position was rotated by 180°. From this position the circles, as projected, maintained the left—right orien- tation of kinesthetic points. In this condition the task was one of identifying relative position rather than true position. Each projected pattern consisted of five correctly positioned circles and one displaced circle. Displacements of the circles (points) were along two axes: the "egocentric" axis--either toward (negative) or away from (positive) the subject's position in front of the enclosure; or the "lococentric" axis--either toward (negative) or away from (positive) the preceding point in the sequence forming the uniform segment (P1 or P.5). The series of 24 patterns 27 represented each point diSplaced in four directions. Points were displaced by 6 cm true scale (proximal condition). acq pro eac sep rat and tri OVERVIEW OF RESULTS Summary of Acquisition The variables considered in the analysis of the acquisition data included sequence, conditional segment probability, movement distance and points. The effect of each variable on the course of acquisition will be reviewed separately. First, however, the general performance over trials will be summarized. Trials. Reaction time and movement time dropped quickly over the first four blocks of trials with decele- rating rates. The drop in these scores was more gradual and approximately linear over the last four blocks of trials. Sequence. Uniform and random groups were compared on the common segments (P1 and P.5 segments, respectively). Reaction times were consistently longer on the random sequence beginning in block 3 and continuing throughout acquisition. The initial slopes of the acquisition curves were comparable, however, reaction times on the uniform sequence showed a Significantly greater slope than random times across the last four blocks of trials. The final mean reaction time for uniform and random groups differed 28 29 by 242 msec i 25 msec after four blocks of training and by 340 msec : 20 msec after eight blocks of training. An estimate of the maximum processing time required to retrieve location information was calculated to be 362 msec i 22 msec. The average correlation between reaction times to the six points, plotted across trial blocks, revealed a pattern of stronger relationships under random conditions. Pearson correlations of about .48 characterized the termi- nal EEEEE and long random groups while the uniform groups Showed correlations of about .25 and .10 on block 4 and block 8, respectively. There were no differences in movement time across the eight blocks of acquisition. The slopes of the uniform and random curves were comparable. The final times on all segments averaged 6.317 seconds i 324 msec after four blocks of training and 4.381 seconds i 254 msec after eight blocks of training. The average correlations between movement times to the six points were not so clearly differentiated as the reaction time correlations. While the correlations for the random sequence leveled off in block 3 at about .65, the uniform correlations continued to vascilate with signifi- cant peaks on block 5 and block 8. The average corre— lation across all blocks was not reliably greater under the random condition (.52 versus .56). 30 Conditional probabilityfiof segments. The random sequence afforded an opportunity to compare performance on segments that had conditional probabilities of .50 and .25. Overall, reaction times were found to be significantly lower on P.5 segments. The difference was most reliable about three fourths of the way through the looo group's training where the difference amounted to 69 msec i 13 msec. While the linear trends over these later trials were comparable, the trend over early trials was significantly steeper on P.5 segments. There was an initial and rather large difference in movement time between P.5 and P.25 segments favoring P.25 segments. This difference, however, disappeared after the first block of acquisition. The rapid deterioration of the initial discrepancy resulted in a significantly steeper linear trend on P.5 segments over the first four blocks of acquisition. Interpoint distance of segments. The random sequence also afforded the opportunity to contrast per- formance over three levels of interpoint distance. The effect of distance on reaction time emerged in block 2 and was at least marginally maintained across all subsequent blocks. The average reaction time for blocks 5 to 8 demon- strated that times were lowest on medium—distance segments, greater on long-distance segments and highest on 31 Short-distance segments. The linear trends were com- parable for all three distance levels. Overall, there was no reliable distance effect on movement time. However, an initial effect limited to the first block of trials tended to differentiate the medium and long-distance segments. Points. Analysis of the pattern of performance across individual points revealed no differences between sequence conditions. In terms of reaction time, a fairly uniform response to the various points was found on both early and late trial blocks. Reaction times to yellow were elevated while times to blue were depressed on early and late blocks. The differences tended to diminish over blocks. The most drastic difference was the reaction time to yellow--some 174 msec above the grand mean (t 25 msec) for long blocks. By contrast, movement time was highly variable with a range of 5.542 seconds (i 584 msec) averaged over early trial blocks. The range on later blocks was 3.016 seconds (i 288 msec). In both cases the mean movement time was high on pink and low on blue. In addition, the mean move- ment times to white on early blocks and to green on late blocks were Significantly elevated. 32 Summary of Transfer Tests Kinesthetic and visual transfer tests were executed after four and eight blocks of training. Kinesthetic transfer. The kinesthetic test com- pared groups on their response times to new and old seg- ments. The only important difference between sequence conditions on this transfer test was a greater rise in the reaction time to new segments for the oniform groups. Movement times on test segments, while they drOpped from EBEEE to looo conditions, were not affected by the sequence of training. Visual transfer. The visual transfer performance was analyzed on two displacement factors--Centricity and Direction--as well as an analysis by points. The sequences were found to differ on overall performance for both per- formance scores. The random groups responded quicker and with fewer errors than did uniform groups. The poorer overall performance of uniform groups broke down into significantly higher error rates and reaction times on negative versus positive displacements when compared to random groups. For reaction time the difference broke down further showing that the direction effect for uniform groups was greater on egocentric displacements. In general, reaction times to displacements on the egocentric axis were Shorter. 33 Length of training did not affect the overall transfer performance, but when the centricity by direction interaction for error rate was considered there was a reli- able increase with further training in the direction effect on egocentric displacements as compared to lococentric displacements. When the results were analyzed against points, significant interpoint variability was found. Transfer performance on the pink point was better than the average, while the blue point showed poorer transfer than the average. The pattern across points differed among groups, essentially showing a flatter curve for random groups, especially for the logo group when pink and blue were not considered. A remote visual test was given to two additional groups after eight blocks of acquisition trials. Compared to the proximal visual test these subjects did as well overall and actually demonstrated a 17% lower error rate (marginally significant). The relationship between acquisition performance and visual transfer performance was assessed by looking at the correlations between scores on the last block of acquisition and the visual test scores. The correlation between the reaction time scores was significantly positive for random groups (.47 and .67 for EEQEE and logo groups, respectively) and significantly higher than uniform groups. 34 This was also true for the correlation between movement time and error rate (.37 and .30 for EEQEE and logo groups, respectively). The only Significant relationship demonstrated by uniform subjects was a negative correlation between movement time and the reaction time to visual trials for the looo groups (E = -.59). An analysis of the overall visual transfer per- formance of female and male subjects failed to show any differences. Controlling for the initial kinesthetic performance during acquisition did not alter the finding. RESULTS AND DISCUSS ION This section is divided into three parts. In the first part the acquisition performance is described and the important between-group performance trends are con- trasted. In addition, the effects of two factors imbedded in the random sequence are assessed. .In the second part the kinesthetic transfer performance is described and related to the trends revealed during training. The visual transfer data are considered in the final part and an analysis of reference systems is discussed. Acquisition The acquisition data were collapsed into block scores where times were averaged across a complete run of the sequence series. By blocking reaction time and move- ment time data into 24-trial mean scores the training sequences could be contrasted without regard to the serial position of the points within the sequence. Scores for the first four blocks included data from both EEEEE and looo_groups (o's = 90) while the later block scores were based on data from the logo groups (o's = 50). Before proceeding to these results, the general strategy used for the analysis of acquisition data should be explained. The means across blocks were analyzed by 35 36 MANOVA based on procedures outlined by Morrison (1976). Each block score, for the particular measure under con- sideration, was treated as a multivariate contributing to the overall test of effect represented by the multivariate F. The analysis included an assessment of the effect for each block and the linear trend across blocks when the multivariate §_was Significant. The equality of EEQEE and looo groups across the first four blocks of trials was confirmed for uniform and random group scores under separate MANOVAS (all F (4, 85)'s < 1.0). Except in the tables where both BIS are presented, the reported multi- variate F is for the eight blocks. Subsequent analyses combined EEEEE and logo groups. Again, because of the change in o, the trend analyses testing the parallelism of the curves were carried out independently on early and late blocks. For each linear trend and within—block comparison Scheffé's post hoc critical range for the Simultaneous tests at each inde- pendent level of training was obtained. The tests for independent groups and repeated measures are multivariate extensions of the Scheffé technique based on Roy's Union- intersection principle (see Morrison, 1976). When the multivariate F was significant, the critical range Statistics for block and trend comparisons were included in the table summarizing the analysis. The critical range, Stated for the 95% confidence interval, was always for the specific 37 contrast of means based on its unique variance. Regardless of the outcome of the multivariate test, mean sum of squares and degrees of freedom of the effect and error terms were presented for each block. The overall block means for uniform and random groups are presented in Table 2. The movement time score always includes the 500 msec rest period over the target point. It is obvious from the block means that performance improved quickly over early trials, particularly movement times, and leveled off in later trials. The general trend characterizes both training conditions; however, obvious differences exist. Comparison of the performance trends was reserved for scores which allowed the two sequences to be equated on sequence segments. For example, by consider- ing only the scores from segments of the random sequence that were common to the uniform sequence (i.e., the P.5 segments), any differential effect by segment could be eliminated. Segment probability. It should be recalled that the random sequence was composed equally of the common P.5 segments and the remaining P.25 segments. The reaction time and movement time block means for the segments are presented in Table 3. These block means were based on 12 trials (equally distributed throughout the original series of 24 trials). The trend of reaction time on the P.5 seg- ments is compared with the uniform trend in Figure 3. 38 una.e vom.v mmh.¢ vmm.m wha.m vmm.h mmo.oa mem.mm Eoocmm mwh.v m¢>.v «ha.m mom.m ham.m mmv.h mmm.oa www.mm Snowflco Aommv mafia uo05m>oz «mm. Hmm. mmm. Nvm. Hao.a ~¢o.a mom.~ mwc.a Eoocmm vmm. mam. mew. mow. new. mam. who.a mvm.a enemas: Aommv mEHu coauommm m h o m v m N H mocmoomm xoon mocmoomm an mama: xooam SOAHAmHSood N magma 39 omm.m nev.v Nhh.v vma.m hmm.m «no.5 mom.m Hma.o~ m~.m vmm.v Hum.¢ mmb.v vmm.m m~¢.w mmv.h mnm.oa mam.¢m m.m Aoomv oEHu ucmEm>oz cam. mom. 5mm. Hmm. mmo.H Hmo.a mma.a emv.a mm.m mmm. mmm. mmm. mom. mam. Nmo.H va~.a mom.a m.m Aommv maeu cofluommm m h m m v m N H mafiaflamnoum ucmammm xoon xuwaflomnoum vooemmm an mmoouw Eoocmm How mono: xoon cowuwmeswom m OHQMB 4O .mucmeomm muwaflnmooum Eoocmm occ Escudo: ou mmawa coauocmm mo comma coHuwmflooomni.m .mam 32:. Co $.35 w n v m mm IN — u q q u m. .IIIIII.I.II..I.I’ I mmm“x&|\\\-.JOEIIIluuwnfiuuunAEv Eovcom .... .. .... crate: imw. 1.F. -m. a. . «0 im m. 10.. s i... 1N; in; l./ [IVO 1 ¢\_ r 41 The lower reaction time on these segments under uniform training, apparent in Figure 3, was confirmed by the MANOVA (E (8, 91) = 5.63, o < .001) which is summarized in Table 4. While the early trends for uniform and P.5 segments were comparable (-456 msec versus -387 msec), a Significantly steeper slope under uniform conditions was apparent on later blocks (~129 msec versus —20 msec). From block 3 through the end of training, the uniform groups reacted significantly faster and, as the linear trend indicates, progressively so over the last four blocks of training. These differences are all at the .05 alpha level (see Table 4). The trend of movement time on P.5 segments is con- trasted with the uniform trend in Figure 4. Movement time on the common segments did not appear to be affected by sequence. The multivariate F from the MANOVA summarized in Table 5 failed to detect a significant difference between the Pl (uniform) and P.5 (random) segments across the eight blocks of training (F (8, 91) < 1.0). The preceding results tentatively suggest that the higher the probability of a segment, the quicker the reaction time will be on that segment. If so, it would be expected that the reaction time on P.5 segments would be quicker than the reaction time on P.25 segments. This, in fact, is what was found. The overall test from the analy- sis summarized in Table 6 was significant (F (8, 42) = 42 Table 4 MANOVA of Reaction Time (sec) on Uniform and P.5 Random Segments —— Multivariatea: _ (4, 6.48 91) = 5.63 Source of MS Critical Rangeb Block 1 1 1.1569 .265 Error 178 .3931 Block 2 l .9053 .181 Error 178 .2496 Block 3 1 2.0363 .137* Error 178 .1028 Block 4 1 2.6407 .140* Error 178 .1074 TrendC 1 .2255 .186 Error 178 .1905 Block 5 1 .9481 .174* Error 98 .0898 Block 6 1 1.0920 .152* Error 98 .0689 Block 7 1 1.6897 .157* Error 98 .0737 Block 8 1 2.8829 .169* Error 98 .0845 Trendc 1 .2946 .076* Error 98 .0173 aSeparate {'8 are for short and long groups, respectively; p's < .001. bScheffé critical range at .05 alpha level; starred values are exceeded by the contrast of means from Table 2 and Table 3. cLinear trend over the preceding four blocks of trials. 43 .mucmamwm >uflaflomooum Eoocmm one EAOMSSD co moEwa ucoEm>oz mo ozone coauflmwooocil.w .mflm flat... E 9.35 m h m n v m N . q _ q q 4 u .i nu e. i¢ in 6...... «a 15% nP Lms 1m 10. 602.3. uuuuu X i: 6.32:: I z ’0 “W _N 44 Table 5 MANOVA of Movement Time (sec) on Uniform and P.5 Random Segments Multivariatea: g (4, 75) < 1.0 91) < 1.0 Source df MS Block 1 1 172.341 Error 178 112.330 Block 2 1 .169 Error 178 30.165 Block 3 l .207 Error 178 13.603 Block 4 1 1.290 Error 178 12.189 Block 5 l .578 Error 98 6.988 Block 6 1 4.431 Error 98 4.613 Block 7 l .873 Error 98 2.833 Block 8 1 4.531 Error 98 4.907 aSeparate {‘5 are for Short and long groups, respectively. 45 Table 6 MANOVA of Reaction Time (sec) on P.5 and P.25 Random Segments Multivariatea: g (4, 36) = 2.80 g (8, 42) = 3.70 Source SE MS Critical Rangeb Block 1 l .2289 .093 Error 89 .0654 Block 2 l .0208 .065 Error 89 .0450 Block 3 l .0173 .035 Error 89 .0133 Block 4 l .0848 .032* Error 89 .0112 TrendC 1 .3336 .060* Error 89 .0393 Block 5 1 .1494 .057* Error 49 .0177 Block 6 l .1192 .039* Error 49 .0084 Block 7 1 .0557 .044* Error 49 .0108 Block 8 1 .0342 .032* Error 49 .0056 Trendc 1 .0255 .044* Error 49 .0107 aSeparate {'3 are for short (2 < .05) and long (p < .01) groups, respectively. bScheffé critical range at .05 alpha level: starred values are exceeded by the contrast of means from Table 3. cLinear trend over the preceding four blocks of trials. 46 3.70, o < .01), and there was a significantly steeper slope for P.5 segments across the first four blocks of training (~387 msec versus -300 msec). Figure 3 shows that the initially higher reaction time on P.5 segments quickly dropped below the P.25 level (block 3) and remained so through the rest of acquisition. The curves are essen- tially parallel over later trials (-20 msec versus -52 msec) where the P.5 reaction time was significantly lower than times on P.25 segments. The movement times on the two probability com- ponents of the random sequence showed some differentiation as evidenced by the smaller means of P.25 segments (see Table 3 and Figure 4). Overall, the difference was signifi- cant under the MANOVA summarized in Table 7 (multivariate F (8, 42) = 2.44, o < .05). The most obvious difference between the segments occurred on block 1, dissipating noticeably on later blocks. The linear trends for the early blocks were, indeed, different with P.5 segments showing the greater lepe (-12.8 sec versus -10.l sec.). These findings suggest that the P.25 segments initially, and thus intrinsically, represented configurations that in some way facilitated the movement response. Given that the P.25 segments were easier to track, the ordinal interaction seen for reaction time on the early trials is easily explained. That is, the advantage of a higher probability for P.5 segments had to overcome the initial advantage of 47 Table 7 and P.25 Random Segments MANOVA of Movement Time (sec) on P.5 Multivariatea: g (4, 36) = 2.66 g (8, 42) = 2.44 Source 9E MS Critical Rangeb Block 1 1 845.282 2.783 Error 89 83.253 Block 2 1 51.233 1.322 Error 89 18.804 Block 3 1 7.880 .645 Error 89 4.474 Block 4 1 111.277 .595 Error 89 3.803 Trendc 1 332.093 1.779 Error 89 34.023 Block 5 1 6.767 .711 Error 49 2.801 Block 6 l .009 .558 Error 49 1.720 Block 7 l .327 .430 Error 49 1.022 Block 8 1 3.484 .530 Error 49 1.555 TrendC 1 .118 .601 Error 49 1.997 respectively; o's < .05. aSeparate F's are for short and long groups, Scheffé critical range at .05 alpha level: starred values are exceeded by the contrast of means from Table 3. trials. cLinear trend over the preceding four blocks of 48 easier movements required by P.25 segments which it is assumed also would lead to Shorter reaction times. The effect of route segment probability appears to have been consistent across training conditions. Overall, we have seen that the uniform (Pl) segments yielded lower latencies than the P.5 segments which yielded lower latencies than the P.25 segments. On the other hand, there was no evidence that segment probability affected the movement response, excepting the anomalously early lower movement time on P.25 segments. In general, then, the greater the conditional probability of a segment, the lower its reaction time. Megaw (1972) has also found reaction time to be similarly influenced by the number of possible directions a movement response could take. When comparing across sequences (Pl versus P.5), this finding is hardly surprising because there is an obvious advantage under the uniform condition in terms of the predictability of the sequence. The serial order of segments is always the same and the subject, early on, can correctly anticipate the upcoming point. This is not true for random groups. Although their sequence does have an order to it, subjects were, without exception, unable to identify the order during post-experiment debriefing. Even in the rare case when a subject reported that he felt that some points were more likely to follow one another, he was unable to accurately identify the more probable pairs 49 (segments). Thus, the advantage of P.5 segments appears not to be a function of an awareness of the segments' greater frequency. The subjects' level of anticipation seems limited to a preparation for determining the correct response which does not extend to actually determining a motor response. The issue of premotor processing will be taken up after the kinesthetic transfer results are described, but a response—dependent interpretation can be considered now. The response-dependent interpretation considers the effect of familiarity on the movement response. The fact that the P.5 segments occur twice as often as P.25 segments means that the amount of practice on the specific movements tracking these segments is also doubled. It would be expected that these specific movements would benefit from the greater practice and, as they became more established, would become more automatic. This would reduce the time necessary for determining or programming the appropriate motor response. It is assumed that this savings would be reflected in shorter reaction times. The problem with this interpretation is that if the movement response to P.5 segments were better established, it would follow that the movement time on these segments also would be Shorter. There was no evidence of shorter movement times, however. The fact that movement times are not lower for the more frequently practiced segments is, by itself, a 50 provocative finding. Essentially, the between-group (P1 versus P.5) and within-group (P.5 versus P.25) results mean that regardless of whether a particular movement is enacted 32, 16, or 8 times (loog groups) the movement time achieved at the end of training is the same. This strongly suggests that the space, composed of all points rather than movements among points, is being learned simultaneously. That is, any set of movement trials among points tends to increase the proficiency of moving about in the space, in general. Further support for this notion comes from the fact that performance on P.25 segments was substanti— ally better after 8 blocks, yielding 8 enactments of the specific movement, than it was on P1 segments after 4 blocks, yielding 16 enactments of the specific movement. For example, the mean movement time on P1 segments in block 4 is 6.547 seconds versus 3.990 seconds on P.25 segments in block 8. This difference is highly significant (2 (138) = 26.64, p < .001). Another way of stating the effect is that during acquisition the positions of the points within the Space (field of points) are what is being learned, and not specific motor responses. Segment distance. Another feature of the training sequences was the distance of, or required length of movement for, each of their segments. The random sequence afforded a differentiation of distance that was not con- founded by points. Each of the route segments was 51 classified into one of three distance categories. Each category classified eight segments distributed across the constituent points. The distance categories were defined as: "short," 16-22 cm segments; "medium," 26-30 cm seg- ments; and "long," 34—44 cm segments. The mean reaction time and movement time scores for each distance category over blocks of training are presented in Table 8. Inspection of the curves shown in Figure 5 suggests that the pattern of differences in reaction time among the different distance segments is established early and continues throughout acquisition. It is apparent that the medium segments yielded the Short- est reaction times and the short segments yielded the longest reaction times. The MANOVA on these means, sum- marized in Table 9, indicated that, overall, the distance effect was significant (multivariate F (16, 182) = 3.78, o < .001). The linear trends for early and late blocks were not significantly different (-322 msec versus -335 msec versus -374 msec, and -40 msec versus -41 msec versus -27 msec, respectively), substantiating the observation that the separation of the distance curves was maintained across blocks. The by-block comparisons revealed that reaction time on medium segments was significantly less than that on short segments throughout acquisition, except block 1 and block 7. 52 wmv.v oom.e mom.v o¢¢.m mma.m ASN.5 omm.oa hme.vm moon mmm.m mne.e hmm.e mmo.m mmo.m avo.h Nvm.m omn.o~ Eoeomz mam.v mmm.v mmo.n mon.m Hmm.» 5mm.h mwh.m HNN.N~ uuonm Aoomv 05H» ucoEm>oz mam. omm. mam. omm. mam. omo.H HSH.H Hom.a moon omm. mmm. mmm. omm. vmm. hao.a omH.H mmv.a scape: vmm. mom. Hmm. hoo.a emo.a mmo.a mmm.a bmv.a unocm Aommv mafia cofluomom m n m m e m m H mocmumwo ucwEmmm sooam oocmomfio ucmEowm an monouw Eoocmm now made: xoon coaufimwoo0< m manna 53 m .mucoEmmm mocmumeo on needs ooHuommm mo ozone coeuflmfiooo .10). The fact that movement time is not a positive monotonic function of distance means that this score is not a measure of motor performance. If the score were tapping primarily motor performance, then by basic principles of motor behav- ior, movement time would have to be, at the worst, a positive monotonic function of distance and at the best, a 56 .mucofimmm mocmumflo co mmEHB unmEm>oz mo ozone GOAHHmHovo¢II.m .owm 23:. +0 8305 h m n v m N . . fl . T . 1 1 - em iv in 1m .km nv 1M 1 L : 9 K: 57 Table 10 MANOVA of Movement Time (sec) on Short, Medium and Long Random Segments Multivariatea: g (8, 150) = 1.71 g (16, 182) = 1.19 Source SE MS Block 1 2 377.518 Error 178 166.309 Block 2 2 16.544 Error 178 18.017 Block 3 2 5.525 Error 178 8.266 Block 4 2 .648 Error 178 9.539 Block 5 2 5.770 Error 98 3.977 Block 6 2 6.359 Error 98 3.113 Block 7 2 .046 Error 98 1.920 Block 8 2 3.859 Error 98 1.711 aSeparate F's are for short and long groups, respectively; p's > .10. 58 linear function of log2 (2L) where L equals length (Fitts' Law: Fitts, 1954; Pitts and Peterson, 1964). The alter- native is that movement time is primarily a measure of spatial knowledge (location of the points within the field). While the slight advantage of medium segments over short segments did not persist in terms of movement time, reaction time performance continued to demonstrate the medium-long-short hierarchy. Previously, reaction time has been described as a function of anticipation which is obviously not a feature of the segment distance classifi- cation. It would appear that the determination or pro- gramming of movement responses on medium-distance seg- ments or long segments is less time consuming than on Short segments. Why preparation for the movement on short segments Should take the most time is not entirely clear. Con- sideration of the general topography of the movement response, however, suggests one explanation. A number of subjects were observed during the acquisition phase of the experiment and the typical movement response was identified. The response consisted of the following: the hand rapidly moved from the initial point toward the end-point slowing noticeably as it approached the point; at this juncture, unless the original movement put the subject directly over the point, the response shifted to a series of rapid or methodical back-and-forth motions centering around the 59 location where the original linear movement was broken off; the length of the back-and-forth movements was pro- gressively extended to cover more area until the point was located. Although other response topographies were evident, they either occurred exclusively during early trials or only after repeated failures to locate a point. These results suggest that subjects tended to rely on the second component (random search) of their response to provide the precision required to locate a point some distance away. Conversely, for those segments where the point separation was hardly any greater than the length of the back and forth movements of the random search component, the subject may have concentrated on the first component (linear movement). The reasoning, here, is that sufficient precision could be achieved without the random search when the initial movement was short. Relying on the first component would force the subject to be as precise, in angle and length of the movement, as possible. This attention to programming the response would consume more time, yielding higher reaction times. The argument can be extended to the long segments if we assume that reliance on the random search component is equivalent for medium and long segments. This means that to achieve the same pre- cision from the first movement component, more attention to angle and distance would be required for the long segments as the absolute error for each increases with distance. 60 Again more attention to programming aspects of the motor response would yield higher reaction times. Marteniuk (1973) and Klapp (1975) report similar evidence of motor programming being limited to short movements. Correlation of performance among points. One of the major premises of this research was that a more random sequence of exposure to the points constituting the space would lead to a more integrated representation of the space. That is, more global features of the field would be learned or learned more quickly under these conditions than under a repetitious uniform sequence of exposure. The analyses, so far, have not directly tested this notion although they have provided evidence that spatial location rather than a set of specific motor responses is being learned. This result allows one to interpret the movement time score as an index of how well the locations of the points have been learned. It follows that if movement time reflects the accuracy or resolution of knowledge about the location of a point, reaction time will reflect the accessibility of this information. For example, if the memorial representation is a well—integrated schema, the resolution of each point should be essentially the same, and the location performance for any point in the field should predict the location performance for any other point. This means the correlations among the movement times on the individual points Should be positive and 61 relatively high. Further, if the location information for every point is being accessed from the same representation, the correlations among reaction times to individual points should be positive and high. If the form of the repre- sentation is stable, the inter-point variability in access and retrieval time should be consistently lower than the variability in the accuracy of location information across the set of points. Thus, the correlations among the reaction time scores would be higher than the correlations among the movement time scores. The Pearson correlations among the point scores were averaged for each block of trials and are presented in Table 11. It is evident in Figure 7 that the reaction time correlations were higher. The contrast between uniform and random training is also evident in Figure 7. Of particular interest is that the uniform and random curves tend to separate in block 3 for both movement time and reaction time. The interpoint correlations of the random movement time scores Showed the expected higher values throughout the rest of acquisition. Using Fisher's l transformation the movement time correlations were com- pared at block 4 and block 8. The correlations were deter— mined to be significantly higher under random conditions at the .05 alpha level (3 = 1.83 and o = 1.68, respectively). Unlike movement time, the differentiation between sequence groups was not consistent for reaction time 62 «we. wee. mmv. one. mmv. mmm. Hmm. omm. Eoocmm Hma. ona. one. mma. NmN. nmm. «mm. ohm. ShOmch mag» ucoEm>oz mom. vow. mew. vac. hmm. mew. mmv. evm. Eoocmm won. mvm. mmm. mmw. one. mam. mmv. com. EHOMHSD mafia cofiuommm m n m m e m N a oocmooom xoon mouoom coauwmwoood mo mcoflumaouuou ucfiooumucH mmmuw>¢ Ha OHQMB 63 .mmocmoomm Eoocmm com EMOHacD How mamumummmm mmuoom moccEHOMHmm mo mcoflumamuuoo mowedumuca mo comma cofiuwmwooodni.n .mwm flat... *0 8.02m m N. w n v m N _ fl - E 1 ll nN—O W «9 nu SE. coEu>o \s u \o .... p s— so. 1 ¢. 3 \\ ll, \ m 0“ ’OI-"SIO", 1 no N .. m. m. .\O|.IZI|AU\ . “W 0.5... 5:251 Eovcem .... ...... .. 632:: IIIII .. oozaoe 64 correlations. A test of the correlations at block 4 and block 8 revealed a Significantly higher correlation for random groups on block 4 but not on block 8 (g = 1.81 and o = .02, respectively). The vacillation across blocks in the uniform interpoint correlations is quite remarkable. The reliability of the abrupt rise in the correlations at block 5 and block 8 was confirmed using the l transfor- mation procedures for related samples (5 = 1.97 for the block 4 against block 5 contrast and o = 1.96 for the block 7 against block 8 contrast). The erratic pattern, in the context of consistently lower movement time correlations, is best explained by changes in the heterogeneity of the representational form used to store location information. The low movement time correlations suggest that the accuracy of location infor- mation is independent for each point. This does not, necessarily, imply that points are not stored in a common representational form, but merely that the representation is composed of discrete informational units derived inde- pendent of one another. In the case of a nonintegrated representation, movement time correlations would be low while reaction time correlations could be high by virtue of the representational commonality. On the other hand, locational representation for each point could be inde- pendent units in memory. This means that the separate representations would not be bound to a common information 65 mode or domain. The issue of information domain will be taken up in the next section and in later sections after the transfer results are discussed, but it is possible at this point to briefly examine one way independent repre- sentations could develop. The process was suggested by post—experimental debriefing which revealed that subjects in the uniform groups, by report, relied more on their sense of arm position and arm movement than random subjects did. The fact that uniform subjects were more aware of the kines- thetic feedback associated with each point is reasonable considering there were only 6 specific movement responses for uniform groups while the random sequence required 18 different movement responses. The relevance of this dif- ference in strategy to the reaction time correlations is that subjects could learn the end position of the 6 uniform movement responses as separate components of the space without representing the entire space as a unit. Acqui- sition would be relatively independent for each point or segment, reducing the interpoint correlations of the per- formance times. The unique reaction time events in block 5 and block 8 could reflect attempts to integrate the spatial information into a unitary representation which would align the access times without, necessarily, differentially affecting the resolution of location information. 66 Performance profile forgpoints. The analysis of association among points raises the question whether the spatial information obtained for each point was equivalent in terms of location accuracy. While there were no spe— cific a priori notions about the relative ease of locating the different points, there was no reason to believe that each point could be located within the field with the same facility. It was proposed that the interaction between the geometry of the field and geometric features of the refer— ence system adopted by a subject would be a principal determinant of any differences among points. The concern, in this study, was whether the patterns of point location performance would differ between the two training sequences. The movement and reaction time scores for each point were averaged across the first four and last four blocks of acquisition trials. The means for each sequence are presented in Table 12. The mean reaction times on the points are shown in Figure 8 with the left to right order corresponding to the order in the uniform sequence. The uniform and random curves are obviously similar on both early and late stages of acquisition. For example, the only extreme reaction time for either condition is on yellow, and then, only for the early stage of acquisition. These results were analyzed using a multivariate profile analysis described by Morrison (1976). The pro- cedure sets up repeated pair-contrasts to characterize 67 Hmo.m mmo.m mao.m mea.e msm.e maa.m macs Soeamm mam.m Hem.m mme.e kmm.e omm.m mma.e macs success ooo.m mmm.mo maH.~H mom.m Hem.oa HSH.4H unonm Soecmm mme.m sem.~a cam.~a Hea.HH meo.aa omm.ma unoam Socoaco Rummy mEHu usmEm>02 Hmm. 0mm. Hmm. Gem. mum. mmm. mace soecmm Ohm. 8mm. emm. mam. «we. cam. mace SuoLaaS neo.a aeH.H meH.H NSH.H mem.a HSH.H unonm sesame ass. ems. mNo.H was. OSH.H cam. uuoam Snoooas Aommv mEHu ooHuommm 65am when: ammuo ems Sofiamw scam macho pcooa mcflowmua mo numcmq com mocmoomm an muowom MOM mommz coHufimwooom NH magma 68 mxoon mung com maumm um>o ommmum>4 mucwom ou mmEHB cofluommm mo maemoumll.m .mamwua cofluwmfloood Eoocmm com EHOMHSD mo mph—Om 03.0 0:53 cameo omm 30__m> SEE . _ q q 5 q HO m. nYIIIIIIIIIIII Iiu.‘-|.\lIn._|I._ll._.I...V.lu..llllllll.Au.II....||_l.......-|.._...Au l mY s .. n. mg; m. 0.! '''' 0' I. I ‘IIOI . .1, I’lO‘“... ”” Illulvo Lm / ’0‘“ 10.— »tom . all 117.0: ...o. i . _ III}..- / X... A. 622.3. uuuuu xztxx - m _ Eton—E: i T. .mam spuooas 69 the curves and then compares the contrast matrices of the two independent conditions of the sequence factor. If the comparison is null, the curves are considered parallel and additional analysis follows. Specifically given paral- lelism, the curves are combined and the effect of points (equality hypothesis) is assessed by a multivariate com- parison of the separate point effects against the grand mean as a constant. It should be noted that these pair— contrasts correspond to the segments that compose the uniform sequence. Analysis of the reaction time data is presented in Table 13 and Table 14. Mean sums of squares are pro- vided for the separate contrasts. Table 13 summarizes the analysis of the profile or interaction effect, and Table 14 summarizes the inequality (points) effect. Analysis con- firmed that the separate sequence curves were parallel (multivariate g (5, 174) = 1.56 and F (5, 94) = 1.90 for the early and late acquisition stages, respectively; o's > .10). The hypothesis of equality of points, with combined sequence groups, was rejected in both cases (multivariate F (6, 173) = 16.65, p < .001; and F (6, 93) = 4.18, o < .01; for early and late trials, respectively. The lack of a significant difference between the reaction time profiles, particularly in the late stage of acquisition, ensures that the sequence differences in reaction time during acquisition reflected a general effect 70 Table 13 MANOVA on Profile of Reaction Time (sec) to Points Averaged Over Early and Late Blocks Source of MS Blocks 1-4 Multivariate g (5, 174) = 1.56a Pink—Yellow 1 .1967 Error 178 .1481 Yellow-Red l .0701 Error 178 .1152 Red-Green l .0199 Error 178 .1387 Green-White l .1157 Error 178 .1588 White-Blue l .1773 Error 178 .1098 Blocks 5-8 Multivariate g (5, 94) = 1.90a Pink-Yellow l .0043 Error 98 .0501 Yellow-Red 1 .0634 Error 98 .0672 Red-Green 1 .0340 Error 98 .0669 Green—White l .1552 Error 98 .0552 White-Blue 1 .1027 Error 98 .0309 2'8 > 71 Table 14 MANOVA on Equality of Reaction Time (sec) to Points Averaged Over Early and Late Blocks Source .QE MS Critical Rangea Blocks 1-4 Multivariate g (6, 173) = 16.65b Pink 1 .7049 .072 Error 178 .0625 Yellow 1 5.4795 .068* Error 178 .0561 Red 1 .0154 .064 Error 178 .0489 Green 1 .0273 .073 Error 178 .0645 White 1 .2434 .067 Error 178 .0537 Blue 1 1.1004 .065* Error 178 .0507 . . _ b Blocks 5-8 Multivariate F (6, 93) — 4.18 Pink 1 .0077 .056 Error 98 .0198 Yellow 1 .3181 .064* Error 98 .0261 Red 1 .0228 .066 Error 98 .0277 Green 1 .0464 .066 Error 98 .0280 White 1 .0400 .048 Error 98 .0147 Blue 1 .2664 .050* Error 98 .0161 aScheffé critical range at the .01 alpha level; starred values are exceeded by the contrast of means in Table 12. 82 < .001 and E < .01 for early and late blocks, respectively. 72 independent of specific movement requirements. On the other hand, the significant inequality effect suggests the access of location information was not constant across points. This finding makes it difficult to accept the notion that all points are represented by a unitary schema-- that is, referenced simultaneously within a common system. The representation of the field may breakdown by point for at least two reasons. Some points may be uniquely referenced to extra-field features of the environment which are not used to reference other points. Such referencing would effectively exclude these points from an integral representation of the field and, because of separate memory storage, could require more or less accessing time. A second, and more important possibility, is that some points may be referenced within a different informational domain. Information domain, here, is conceived as a con- struct similar to sensory modality where data coding pro- cesses are discretely determined for each level of input. It is proposed that location information on some points may be processed differently, or more extensively, than on other points. Consider two possible information domains: a kinesthetic, body reference system; and a visual, field reference system. The latter domain obviously requires additional processing to arrive at visual-like relations from kinesthetic input. The final informational codes for these two domains suggest different memory storage 73 requirements, in turn suggesting different access times. The significance of information domains as a central factor in location processing will be discussed further after movement time results are described. Variations in the mean movement times presented in Table 12 were markedly greater than reaction time vari- ation. The pattern across points, depicted in Figure 9, fails to suggest any consistent relationship between the relative reaction time and movement time performance on the points. As found for reaction time, however, the movement time patterns are similar for different sequence and length of training conditions. The MANOVA is presented in Table 15. In early and late trials the uniform and random patterns were judged parallel (F (5, 174) = 1.86 and F (5, 94) = 1.91, respectively; o's > .10). The inequality of points was confirmed for both stages of acquisition (multivariate g (6, 173) = 28.21 and F (6, 93) = 38.46, for early and late trials, respectively; o's < .001). The summary of the analysis in Table 16 includes post hoc tests of each point. Overall, pink and blue represented the extreme performance levels. Blue was consistently located rapidly while movement time on pink was protracted, indicating poor location ability in all groups on this point. One level at which the points can be contrasted is their relative relationships to field and _self (the subject's body). Blue is the closest point to 74 .mamwna SOAuemwoood eoocmm ocm EhOmch mo mxoon mama pom Nahum um>o ommmumbm mucwom cu mmfiwe pomEm>oz mo mawmoumni.m .mwh mPZ_Om 03m 0:53 c0050 .31 3o__m> :51 d d d _ . 4 AV 0/ mm 26.. c m m h 0/ m Eoocom 11...... m 0 o E. 2:: O. _ _ N _ 75 Table 15 MANOVA on Profile of Movement Time (sec) to Points Averaged Over Early and Late Blocks Source of MS Blocks 1-4 Multivariate g (5, 174) = 1.86a Pink-Yellow 1 20.672 Error 178 61.012 Yellow-Red 1 ‘ 170.527 Error 178 42.001 Red-Green 1 125.679 Error 178 31.719 Green-White 1 43.494 Error 178 50.562 White-Blue 1 66.739 Error 178 51.143 Blocks 5-8 Multivariate g (5, 94) = 1.91a Pink-Yellow 1 13.392 Error 98 15.259 Yellow-Red 1 39.855 Error 98 7.634 Red-Green 1 2.310 Error 98 8.111 Green-White 1 2.252 Error 98 7.404 White-Blue 1 9.254 Error 98 3.358 a 2'3 > .10. 76 Table 16 MANOVA on Equality of Movement Time (sec) to Points Averaged Over Early and Late Blocks Source SE MS Critical Rangea Blocks 1-4 Multivariate g (6, 173) = 28.21b Pink 1 893.669 1.540* Error 178 28.380 Yellow 1 67.665 1.279 Error 178 19.586 Red 1 86.443 1.153 Error 178 15.919 Green 1 104.421 .947 Error 178 10.748 White 1 478.055 1.535* Error 178 28.210 Blue 1 1976.161 1.061* Error 178 13.484 Blocks 5-8 Multivariate g (6, 93) = 38.46b Pink 1 131.691 1.154* Error 98 8.456 Yellow 1 62.682 .739* Error 98 3.474 Red 1 13.383 .729 Error 98 3.374 Green 1 151.264 .797* Error 98 4.031 White 1 32.047 .634 Error 98 2.547 Blue 1 318.982 .549* Error 98 1.909 aScheffé critical range at the .01 alpha level; starred values are exceeded by the contrast of means in Table 12. bo's < .001. 77 the subject and slightly to the right of the body center line conforming to the right-shift bias introduced by using the right hand for movements (see Siddall, Holding, and Draper, 1957). In terms of the two informational domains, blue is the most likely candidate for body refer- encing (kinesthetic). Pink, on the other hand, is more removed from the subjects' position and lies at about the center of the field. This point, minimally distant from all other points, has the most geometric significance if all point locations are defined configurationally. Thus, pink is the most likely candidate for field referencing (visual). If the preceding propositions are true, the move- ment time scores mean that kinesthetic referencing as a strategy is a more accurate location system. On the other hand, the strategy may be appropriate for only a few points which fall into a restricted space contiguous with a certain body axis. For example, a perpendicular drawn from the inside of the right shoulder of an "average" subject (1.7 m, 68 kg, male) intersects blue and yellow, two points showing the shortest movement times. The line is also oriented to the "natural" resting position on the table top of the right hand. This notion, that some points are more kinesthetically referenced and less visually referenced than others, will surface again when the visual transfer results are discussed. 78 Kinesthetic Transfer The extent to which the spatial information acquired during training has been integrated into a schema repre- senting the field of points should have some obvious con- sequences on the general ability to accurately move about in the field. This would include the ability to generate novel movement responses, that is movements between points not previously occurring sequentially. The relative per- formance or transfer on these new segments represents a useful test of the functional value and generalizability of the schema, in whatever form it may take. Before discussing the results of the kinesthetic transfer phase of this study, the derivative nature of memorial schemata must be examined. Previous discussions have assumed that a schema is the simultaneous represen— tation of data (Spatial information) depicting a specific structure or relationship among the data that is system- atically related to elements of the stimulus complex (space). This assumption requires that the data be com- mensurable, that is, derived from essentially the same encoding process terminating at some discrete level of encoding common to the schema. The possibility remains, however, that because of features intrinsic in the space and because of intentional or more automatic selection of strategies for processing the kinesthetic feedback, data are derived by different routes. This means that for some 79 points in the field, information may exist at a different encoding level than for other points, preventing the for- mation of a unitary integrated representation of the entire field. Thus, schemata may be incomplete in terms of their point complement, and multiple schemata, at different levels of encoding, may exist. With the possibility of multiple coding in mind we continue with an examination of the kinesthetic transfer. The degree of transfer achieved under the two acquisition sequences was assessed by contrasting the mean performance on "new" and "old" route segments. From the 30-trial test sequence which terminated the kinesthetic session, 12 trials were common to both acquisition sequences (uniform segments). The average score on these segments became the old score. Scores on the 12 trials which yielded segments not used in either acquisition sequence were averaged to become the new score. If transfer were perfect, no difference between new and old scores would be expected. Reaction time. The mean reaction time on transfer trials after four and eight blocks of training are presented in Table 17. Regardless of sequence, latencies appeared to be higher on new segments. A 2 x 2 x 2 ANOVA on these data, where sequence, length of training, and segment novelty were each represented on two-level factors, is sum- marized in Table 18. The analysis revealed a significant 80 Table 17 Performance on Kinesthetic Transfer Test Test Segments Group Old New Reaction time (sec) Uniform Short .790 1.017 Random short .863 .971 Uniform long .822 1.056 Random long .839 .931 Uniform short Random short Uniform long Random long Movement time (sec) 5.692 5.337 4.336 4.189 5.122 5.276 5.012 4.415 81 Table 18 ANOVA of Reaction Time (sec) on Kinesthetic Transfer Trials Source SS of MS F 2 Sequence (8) .0166 l .0166 <1.0 Length (L) .0001 1 .0001 HumHmm mamflue Hmmmcmua owumcummcflx on mafia coeuommmii.oa .oflm Eat... *0 .305 :6... to as. 5m 5m 51. 4 - J .“5 7 LJLJ which L m, spuooas 33.: .826: 302 u i ”.2: .829: 20 o Eoecom .... in :3: 5:33.84 6 883:: 84 segment probability. The new segments have a functional probability of zero relative to the acquisition sequence, and the times would be expected to be very high. Reaction times on the old segments, however, require further expla- nation. When performance on the old test segments was compared with the performance on the segments during the last block of acquisition trials in Figure 10, an apparent increase was noted. The 43 msec rise for the EEEEE groups was not significant (E = .95, df = 19, o > .10) but the jump of 288 msec after eight blocks of training was signifi- cant (E = 8.23, df = 19, o < .001). This difference indi- cates that a factor in addition to segment probability is affecting performance. While reaction times have been explained as a function of the conditional probability of the individual segments, it is apparent that the occurrence of the entire sequence was also important. Disruption of the anticipated sequence, after eight blocks (32 cycles of the uniform sequence), slows the reaction time of uniform subjects to the level attained by random subjects. Random subjects had not been able to anticipate the entire random sequence and, therefore, showed no rise in the reaction time to old segments. An interesting aspect of the conditions responsible for this equality is the significantly longer time to respond to new segments after random training. When the old score from the transfer test of the short random group 85 is replaced by the mean reaction time to P.25 segments in block 5 (981 msec), the effect disappears (E = .22, df = 68, o > .10). The comparison demonstrates that while new segments affect performance relative to old P.5 segments in the test, the effect is not different from that of the familiar P.25 segments after the same amount of training. Movement time. The movement time results are presented in Table 17. The disruption in performance evident in reaction time scores was not present in the movement time scores. Figure 11, in fact, depicts a slight drop in movement time on the test trials, after four blocks of training on either sequence. After eight blocks, per- formance remains relatively stable without a drop in move- ment time. A 2 x 2 x 2 ANOVA, described previously for the reaction time analysis, is summarized in Table 19. The analysis showed that the average movement time transfer scores were not significantly different for uniform and random groups (sequence and sequence by length E (1, 76)'S < 1.0). The average transfer scores, however, were Significantly lower when taken after eight blocks of acquisition (F (1, 76) = 4.99, o'< .05). There were no reliable differences between old and new segments (test and sequence by test E (l, 76)'s < 1.0). While the effect of test segments on random groups' movement performance was obviously inconsequential, the uniform results were more ambiguous. Transfer trials 86 .cheuaocoo Eoocmm ocm EHOMNGD Hmoca moccEHomumm oofluwmfiooom ou m>wumamm macwue Hmmmcmua oeumnummcwx :0 mafia acmem>ozii.aa .me& 32.... .8 .305 amok to 58. 5m 5m 5.. - q . H. d, ‘Hhv An - v S 1 3 0 i m w p L S i w ”.2: .328: 302 .. .. A. $3.: .229: 20 6 Soccer -11.. 1 Eat. 5:33.34 6 .5225 87 Table 19 ANOVA of Movement Time on Kinesthetic Transfer Trials Source SS of MS F E Sequence (5) 2.2310 1 2.2310 .05 S X L X T 2.3003 1 2.3003 1.27 >.lO Subjects X T/S X L 137.8345 76 1.8136 Total 639.0259 159 88 for the uniform short group yielded a 570 msec lower mean movement time on new segments. This difference for the uniform long group was reversed yielding a mean score on new segments that was 676 msec higher than on old segments. This effect contributed to an overall length by test effect that was marginally significant (E (1, 76) = 3.24, p < .10) but was not significantly different from the random pattern (sequence by length by test, 3': 1.27, p > .10). The equality of the results across sequence con- ditions is weakly challenged by the unique pattern of uniform transfer means on the new and old test segments. The uniform results are especially interesting because the movement time on new segments following four blocks of trials is actually lower than the movement time on block 5 by 684 msec, but again this difference was not significant (E_= 1.37, df = 68, p’> .10). Under extended acquisition (uniform long group) subjects showed no improvement over block 8 on new test segments. These results do not show a deficiency in performance on new segments, either within- group relative to old segments, or between—groups relative to random scores. It must be concluded that the location information generated under the two sequence conditions is equally effective when novel response movements are required. Again, kinesthetically defined location infor- mation and not movement coding best describes the apparent nature of spatial memory. Further, these results provide 89 no indication that the random condition fosters a more integrated or schema-like representation of the space. Processing time: an aside. The concept of an anticipation process has been used repeatedly to satisfy the demands of movement performance data. What the subject achieves with the time gained by knowing the next point in the sequence has not been elaborated and deserves attention. The potential sources of gain are proposed in the following chronology of events antecedent to the movement response. First, the signal is perceived and identified. Then, the location of the identified point is retrieved from memory. This stage has two steps; first, the memorial represen— tation of the space (or a piece of the space) is accessed, and second, the location of the point is derived from this representation. If the representation is a schema relating all points to the space, the schema, given sufficient training, is likely to be under nearly constant access or can be quickly accessed (cf., "effortless retrieval" in Posner, 1973). In terms of processing time, it is assumed that for a schema the bulk of this stage is given over to searching memory for the location of the point and not to accessing the schema. Once the location information is retrieved from memory it must be translated into the appropriate movement response. This means that motor response parameters are determined such that the movement takes on the correct 9O direction and, to a lesser degree, the correct extent. This stage is also conceived as having two separable steps. If the location information has been stored in a code unlike the original kinesthetic feedback, a decoding process is needed to translate the information into kinesthetic terms. Once the information exists in a code interpretable by the motor system, the second step is possible. To some degree this step involves programming the motor response (Keele, 1968), which carries with it particular constraints on the initial conditions. Keele (1968) and Pew (1974) have pointed out that a motor program requires pre-response proprioceptive information about the state of the muscular system (e.g., position of the arm at rest over a point). Thus, in the present experimental task, the subjects would have to reach the point and come to a stop before the motor program for the next point in the sequence could be gene- rated. Alternatively, it is possible that the location information exists at the level of motor or kinesthetic coding. In this situation no translation would be necessary and the proprioceptive information needed to satisfy the initial conditions would exist in memory, eliminating the need to be at rest over the preceding point in the sequence. The results from acquisition and the transfer test are consistent with the notion that location information is kinesthetically coded. In general, this step determines the response and may or may not involve motor programming, 91 as the movement can easily be executed under feedback control and conform to Fitt's law. What this means to the anticipation process is that given kinesthetic coding all processing steps that precede the actual initiation of the movement could be enacted while in the process of locating the preceding point. The advantage gained by being able to reliably predict the upcoming point lies in the result of the anticipation process. That is, the subject is primed to make the appropriate response when the signal is presented. The primed state entails retrieval of the location information and can be seen in both uniform and random subjects (reaction time as a function of segment probability). Another condition can be hypothesized for uniform subjects-- the fixed state--where not only is location information retrieved ahead of time but the movement response is deter- mined or programmed in advance. In the fixed state the subject is committed to making the movement before the signal is presented. A concomitant effect of the fixed state could be a shift of control away from external sig- nals and toward internalized sequence and temporal repre- sentations. Evidence exists which graphically demonstrates this possibility. Examination of the raw scores revealed that reaction times of less than 200 msec-~the recognized ‘unit reaction time for visual information--began showing up in block 8 of uniform trials. Sixty-eight percent of 92 the subjects demonstrated at least one instance of a reaction time below 200 msec in this block and approxi- mately 7% of the trials overall, were below this minimum. These times can be explained only if the subject's antici- pation process includes fixed states that have come under temporal control. In other words, subjects on these trials must be initiating the movement response before the signal is perceived. The unique events in block 8 present an opportunity to roughly assess the functional significance of the reaction time score. Specifically, the simple difference in reaction time between the random P.25 segments and uniform segments can serve as an estimate of memory process time. This requires that the mean performance in block 8 represented the consistent use of primed states in respond- ing to the uniform sequence. Since a majority of uniform long subjects displayed only incipient temporal control, it is assumed that performance on this block is predomi- nantly a function of the primed state control of responding. Subtracting the 7% of scores below 200 msec from the mean affords a reasonable estimate of primed state reaction time (548 msec). The P.25 reaction time (910 msec) less this estimate yields a memory processing time estimate of 362 lmsec with an associated standard error of 22 msec. While ‘the quantitative validity of this estimate is uncertain, .it seems to be realistic. That is, it affords adequate 93 time to enact all hypothesized stages of preresponse activity. It is proposed that the bulk of this estimate reflects memory search time. This component in the present experiment appears equivalent to what Teichner and Krebs (1974) have labeled the "stimulus-response trans- lation" in various choice reaction time studies. Their derived time estimate of this processing stage for four alternative stimuli and corresponding responses was about 300 msec with a comparable amount of training. Visual Transfer The visual transfer data from the recognition- detection task were examined under two classification schemes. First, test trials were grouped on the factors of Centricity (orientation of displacement axis) and Direction (direction of axial displacement). These cate- gories cut across the constituent points so that every point was represented in a category. In the second part of this section the transfer performance on the constitu- ent points will be examined independent of how the point was displaced on test trials. The 24 trials of the visual recognition test were collapsed into four categories based on the types of dis- placement involved: lococentric displacements away from (positive) and toward (negative) the preceding point making up the uniform or P.5 segment and egocentric displacements (away from (positive) and toward (negative) the subject's 94 position relative to the field. Within each category the proportion of incorrect point identifications, or error rate, and the mean latency to respond with an identifi- cation were tabulated. A few assumptions were made before the results were analyzed to provide a framework for interpreting the data. First, it was assumed that a subject would reference points with some sort of straight line geometry which primarily relied on the angle and secondarily on the length of "lines" connecting points with other points or with some positional elements within the immediate environment (e.g., the medial axis of the person's body). Following from this, if these lines are generated by a system char- acterized by either lococentric or egocentric reference, displacements along the corresponding axis and those along its quasi-perpendicular should represent different problems of orientation and error detection. Specifically, it is assumed that because angle is considered the salient cue for orientation, displacements off the axis of reference (i.e., along the quasi-perpendicular) should be easier to detect. Displacements along the axis of reference would have to be detected primarily on distance information. If an axis is to be considered as a reference axis, based on a demonstration of better error detection of dis- placements off-axis, the error rate on the quasi-perpen- dicular axis must be equal for positive and negative 95 displacements. This is because the angle of displacement and the distance of the reference point would be the same in either direction. When the detection of error varies with the direction of displacement on one axis, it will not be possible to identify the quasi-perpendicular axis as an axis of reference without abandoning the assumption of angle saliency. On the other hand, an axis with differential error rates may be, itself, a reference axis. If system- atic unilateral error occurs, meaning subjective point locations are contracted or extended along an axis, the schema can be considered distorted by virtue of inaccurate distance referencing on an angle-salient axis. A cautious application of this interpretation requires the assumption that the distance metric, on the psychological level, is symmetrically scaled about the true position of the point. The practical consequence of these assumptions is that if either of the tested axes is relevant to the reference system used by a subject, one of the following conditions will be met. (1) If error rates for negative and positive displacements on one axis are equal and lower than the average error rate on the other axis, then the quasi-perpendicular axis can be interpreted as a reference axis with angle saliency. (2) If the error rates differ yet the error rate for one direction is lower than the average error rate on the other axis, then the axis, itself, can be interpreted as a reference axis with 96 distorted distance coding. (3) If both axes are relevant within a reference system immune to distance distortion, no conclusion can be drawn about angle-salient reference based on the data collected. (4) If the reference system induces systematic distance distortion, however, both axes can be interpreted as reference axes when differential detection of positive and negative displacements occur on both axes. Another assumption made prior to analysis bears upon the interpretation of the two performance measures. In some respects error rate and reaction time would be expected to mirror one another, yet there are obvious qualitative differences in the two scores. In relationship to a cognitive schema of the space to which the projected points can be compared, the error score is assumed to reflect the accuracy or congruency of the schema. The reaction time would also, to some degree, reflect the accuracy of the image, but is thought to more closely tap the "consolidation" of the schema. Consolidation, in this context, can refer to the clarity, stability, or accessi- bility of the image. This in turn could translate into a construct of the subject's response confidence or perceived reliability of his schema. For example, if the error rate was high yet the reaction time was low, performance would be interpreted as indicating an inaccurate but well 97 consolidated schema. If the converse were true, an accu- rate but poorly consolidated schema would be indicated. Displacements. The mean scores for displacement categories from each training group are given in Table 20. Each score reflects data from six trials of the test. The data were analyzed with 2 x 2 x 2 x 2 ANOVAs where (training) Sequence and Length (of training) were inde- pendent group factors and Centricity and Direction were two-level repeated measures factors. The overall effects of the displacement factors, independent of Sequence and Length, will be considered first. From the mean error rates in Table 20 it can be calculated that the overall error rate on displacements in a positive direction was .25 compared to a rate of .36 on negative displacements. The ANOVA summarized in Table 21 confirmed that this direction effect was significant (F (l, 76) = 10.87, p < .001). The advantage evidenced for positive displacements was primarily a consequence of the more accurate detection of these displacements on the ego- centric axis. While the egocentric negative error rate was twice the error rate of .19 on positive displacements, the rates were essentially equivalent on the lococentric axis (.33 versus .32). This contrast is represented by the significant centricity by direction interaction (2 1, 76) = 11.03, E < .001). 98 mmm. ham. mmm. «as. mamcflmumz mom. now. «am. Nvm. mnH. macs Eoecmm mhm. mom. New. New. med. mcoH Enemflco mom. now. «am. omm. New. uuonm soscmm mmm. has. New. use. ham. usogm anodes: mums Houum Hom.v Hoe.a mmm.v mm~.q mamcflmumz mae.m mem.m mmo.q omm.m mam.m mcoa eoncmm koo.m mom.q sam.m Hmm.m mm¢.v macs shotgun Hom.q mom.¢ mmm.q mmo.e meo.v uuoem soccmm mom.q ooa.m mme.v am~.m nmm.v uuoam anodes: Aommv TEH# COHUUmmm m>flummmz m>fiuflmom m>Hpmmmz m>HuHmom mamcflmpwz msouw UHHDCOOCOOA oauucmoomm “mm? HQMWCMHB HMDmfl> ®£# Cfl m¥C®E®OMHQmHQ CO QUCMEHOWme om OHQMB 99 on moom.oH Hmuos ammo. mo oomo.o H x m\o x o x muomnnsm oH.A oe.H oooH. H oooH. a x o x H x m Ho.v Ho.m mmmo. H -mv. o x o x H oH.A om.H mmmo. H mmoo. a x o x m Ho.v mo.HH homo. H sumo. a x H oooo. on ooHH.m H x m\o x muommns m oH.A o~.H ommo. H ommo. a x H x m o.Hv omHo. H omHo. a x H Ho.v mH.oH mmHo. H mmHo. a x m Hoo.v so.o~ memo. H memo. “no qupoouH a Hhoo. mm momm.m H x mxo x muommnsm o.Hv moHo. H ooHo. o x H x m o.Hv oooo. H osoo. o x H o.Hv omoo. H omoo. o x m oH.A mo.H «moo. H oooo. loo muHoHuucmo oomo. on HmoH.o H x m\mpomnnsm o.Hv momo. H mono. H x m oH.A mH.H ommo. H ammo. AHV apocmH Ho.v Ho.oH moom. H ~oom. Ame mocmswmm m m m2 mm mm mousom ”:me memGMHn—w HMDmH> on» em mucmemomammma so mumm uounm mo ¢>oz¢ HN OHQMB 100 The reaction times showed a somewhat different pattern as supported by the analysis in Table 22. The advantage for detecting egocentric displacements was reli- able regardless of direction (E (l, 76) = 4.99, p < .05). Again the advantage, this time in speed of responding, was pronounced on displacements in a positive direction. This effect, represented by the centricity by direction inter- action, was significant at the .01 alpha level (E (1, 76) = 9.33). It should be noted, here, that the means are based on data from all trials including those where the subject's response was incorrect. Thus, if some proportion of errors were the result of responding too quickly, a negative relationship between the reaction time and error rate scores would be expected. There was no evidence of this occurring, and a number of previous reports (e.g., Cooper and Shepard, 1975) have indicated that in visual matching tasks true and false reaction times are comparable. To what extent the patterns of visual transfer performance were determined by the training sequence will be considered next. Returning to Table 20, it can be seen that the random groups had a lower error rate overall (.26 versus .35) on the visual test and responded approximately 800 msec faster than did uniform groups. Both differences represented significant main effects (E (l, 76) = 10.01, p < .01; and F (1, 76) = 6.70, p < .05; for error rate and reaction time, respectively). 101 on omo.omoH Hmuoa omH.H on moH.mo H x m\o x o x muommnsm o.Hv HoH. H HoH. o x o x H x m oH.A oo.~ oom.m H ohm.~ a x o x H mo.v NH.o moo.o H omm.o a x o x m Ho.v mm.m ohm.oH H oom.oH o x o ooo.~ on oHo.omH m\o x muommnsm oH.A om.H oom.~ H moo.m o x H x m o.Hv omm. H omm. a x H mo.v H~.m ooo.oH H moo.oH o x m o.Hv mom. H ohm. Loo :oHuomuHo oom.H on moo.oo m\o x muomflnsm o.Hv mmH. H meH. o x H x m o.Hv moo. H moo. o x H o.Hv Hom. H Hom. o x m mo.v oo.o omo.m H woo.o loo suHoHuHcmo Hom.k on mnm.m>m H x m\muomnnsm oH.A mm.m om~.oH H o-.oH H x m oH.A om.H oom.m H omo.o AHV apocmH mo.v oo.m mvo.om H mom.om Ame mocmsomm m m m: Mm mm mousom ‘IDI pmme uwmwcmue Hmsmfl> may CH mummEmomammHQ ou Hommv mEHB cofluomwm mo «>024 NN mHQwB 102 The superior transfer achieved by the random groups was consistent with the initial hypothesis that this experi— ence would lead to a more consolidated and more accurate image or schematic representation of the space. Another experimental hypothesis was that detection of displacements along the egocentric axis, as the quasi—perpendicular to the lococentric reference axis, would be favored by uniform training. The sequence by centricity effect, however, was not significant for either error rate or reaction time. Instead, the uniform groups demonstrated differential reaction times for direction on the egocentric axis with positive displacements detected some 650 msec faster than any other displacements. The random groups, overall, showed longer reaction times to positive displacements on both axes (310 msec average difference). This difference was significantly different from the uniform contrast for displacement direction (F (1, 76) = 5.21, E,< .05). The reliability of the unique uniform pattern of reaction times is demonstrated by the significant sequence by centricity by direction interaction (F (l, 76) = 4.12, p < .05). This pattern was also present when performance was assessed by the error data. That is, the .18 error rate on egocentric positive displacements after uniform training was at least 48% less than the error rates on any other displacement type. Unlike the case for reaction times, the combined random groups showed a similar though weaker 103 pattern (at least a 22% lower rate for egocentric positive displacements). The difference between the uniform and random patterns was not significant (5 (1, 76) = 1.58, o > .10). The groups, however, did respond differently to the direction of displacements where the uniform groups showed, on the average, a 40% greater error rate to negative displacements and the random groups showed a negligible 11% difference between negative and positive displacements. The length of training on the sequences was sur- prisingly inconsequential for visual transfer. Reaction times were not affected by length of training at any level. The detectability of displacements, however, did shift in an interesting way when error rates were contrasted between EEQEE and loog training conditions. Essentially the pattern most associated with the uniform training, that is, the marked accuracy of detecting egocentric positive displace- ments, was predominant when sequence groups under the logo training condition were combined. Here, the direction differential was 64% compared to 30% under the EEQEE training condition. The direction differential on the lococentric axis went from a positive 21% to a negative 16% with additional training. As Table 21 indicates, length by centricity by direction interaction was significant at the .01 alpha level (5 (l, 76) = 7.81). 104 Reference systems. Training on the random sequence seems to have provided subjects with a more accurate and better consolidated visual schema of the layout of points, as this condition yielded fewer errors and lower reaction time on the visual transfer test. The schemata appear to have been established early, as additional training did not improve transfer performance in either the random or uniform conditions. The kind of errors that were made during the visual test, however, did depend on the sequence and length of training. Relying on planned contrasts among the four types of displacements, separately for each treatment group, allows a simple depiction of these effects. The within-subjects contrasts were generated by the assumption about reference axes discussed earlier in this section. Specifically, the axes (centricity) were com- pared across direction, and the directions of displacement were compared with centricity held constant. The pooled variance across groups was used in each case. Comparisons of the means in Table 20 revealed the following patterns where differences exceeded the critical range for the 95% confidence interval. The uniform short group responded with the same accuracy and speed on both axes. However, direction differentials, with better accu- racy on the positive displacements, were present on each axis. This differential was apparent also in the reaction time to egocentric displacements. The uniform long group 105 continued to show comparable performance across axes, but the direction differential was present only on the ego- centric axis and only in terms of accuracy. There was no evidence that the random short group detected one type of displacement more accurately or more quickly than another type. The random long group did show a direction differ- ential on the egocentric axis where, as in the uniform condition, few errors were made on positive displacements. The critical range for the separate direction contrasts was .085 for errors and 945 msec for reaction time. For the combined centricity contrasts, the critical range was .065 for mean error rate and 526 msec for mean reaction time. The lack of a centricity effect on the separate groups indicated that neither axis, alone, developed into a predominant angle-salient reference which was capable of preserving accurate distance information. Whether both axes simultaneously developed this capability was difficult to ascertain. The significance of the random short group having met the partial criteria for this condition will be discussed later. On the other hand, there was good evidence that the axes were used as angle-salient refer- ences which distorted the distance metric. This is particularly true for uniform groups where point displace- ments along the egocentric axis in a positive direction ‘were detected about twice as often as negative displacements 106 after four blocks of training and about four times as often after eight blocks of training. Here, the saliency of this axis appears to be increasing, although no specific test of the comparison was built into the analysis. In a Scheffé post-hoc test of the effect, the mean difference of .20 did exceed the critical range at the .05 alpha level (the critical range using the pooled variance term was .170). Thus, this reference axis appears to be established early for uniform subjects who tend to rely on it more as their training progresses. These subjects appear to be systematically dis— torting the space by contracting locations along the ego- centric axis. That is, their schema of the space appears to have pulled the points in closer to the subject's posi- tion, front and center of the field. Thus, negative dis- placements appear correctly positioned and positive dis- placements are maximally discrepant. For uniform short subjects the spatial schema is also foreshortened along the lococentric axis. This general type of distortion is not surprising if one considers representational effici- ency to be achieved when the distances among points or between points and an extrinsic reference point are mini- mized. This conceptualization rests on the idea that the representation of the space is isomorphic to the space (Attneave, 1974; Shepard, 1975) and isotropic to itself (Arbib, 1972). The rationale is that a uniform contraction 107 of the space preserves the geometric relations while reducing the amount of low information "empty space." While the results are fairly conclusive when dis- tance was distorted by the reference axes, whether both axes were used as components of an isometric angle-salient system is more equivocal. The random short group responded with equivalent accuracy to both egocentric and lococentric displacement regardless of the direction of the displace- ments. Two possibilities exist, given that accurate visual transfer was achieved. One is that neither axis was salient and some axis or geometric feature of the field environment, orthogonal to both axes, was the reference for generating angular information. The other is that both axes were angle-salient. It can be tentatively concluded that random EEQEE subjects used a reference system which relied on or subsumed both axes in generating angular information, as it is difficult to conceive of an alternative reference axis which would be orthogonal to both axes in the sense that neither axis was differentially affected by orien- tation along this axis. One candidate, however, does exist if we consider a reference system composed of a set of axes radiating from a fixed point in the field. This yields something like a compass with its center positioned near the center of the field. The orientation of the radial system must also be fixed, for example, by aligning one axis with the body midline. In this case, the equivocal 108 centricity effect could be explained because, in general, radial axes would bisect the angles formed by the ego- centric and lococentric axes. On the other hand, it would do nothing to explain the direction differential, particu- larly on the egocentric axis, because for half of the points a displacement in one direction (positive or nega- tive) would be in the opposite direction on the radial axis. Some support for the radial system will be discussed in the next section. Analysis by point. The overall transfer perform- ance by point was also analyzed. These results, as error rate and reaction time scores, are presented in Table 23. An analysis by point is important for two reasons. If the effect of sequence is shown to be an overall effect independent of which point is under consideration, then this rules out the possibility that the route(s) taken to that point during training specifically determined the difficulty of visually locating the point. This independence would give the training factor broader generality as its effect is distributed across all points. The sequence by point effect in the ANOVAs presented in Table 24 and 25 are tests of this generality. The second reason that the analysis is important is that some feature of the point's position may make it intrinsically more difficult to locate under any conditions. This possibility is tested by the point effect in the ANOVAs. 109 oom. mom. mNm. ohm. mom. Hmm. HmnHoumz omm. mmm. mom. mow. oom. ooH. ocoH sooamm moo. omm. mNo. mom. omm. mum. ocoH auoHHoo mmm. omm. mow. omm. mum. mNN. uuonm saunas omo. mHm. mom. mmm. omm. omo. Huonm enoMHco wumh HOHHm ~oo.o ooo.o Hoo.o mmm.o mom.o «mo.m HmcHoumz o-.o moh.m oms.m mom.m mos.m «HH.m maoH soocmm ooo.o mms.o mm~.m on.m Hmm.o oom.o ocoH suoHHco ooo.o moo.o omo.o mom.o Nom.o o-.o uuosm soocom Hom.m onm.o ooo.m Hom.o omm.m omo.m uuonm suocho Hommv mam» sowuommm msHm mHan :mmuo com onHm» xcHa msouo mucwom Ummfi HmeCMHB HMDmH> 0£H CH mHGHOQ GO @UGMEHOHHGQ mm manna 110 Table 24 Partial ANOVA of Error Rate on Points in the Visual Transfer Test Source SS of MS 3 2 Point (P) 1.2000 5 .2400 7.12 <.001 Sequences (S) X P .1510 5 .0302 .10 S X L X P .1907 5 .0381 1.13 >.10 Subjects X P/S X L 12.8010 380 .0337 Totala 21.6903 479 aTotal includes independent group and error terms not shown: refer to Table 21 for ratio equivalents of these terms. 111 Table 25 Partial ANOVA of Reaction Time (sec) to Points in the Visual Transfer Test Source 88 of MS F E Point (P) 38.789 5 7.758 5.34 .001 Sequence (8) X P 2.288 5 .458 <1.0 Length (L) X P 6.774 5 1.355 <1.0 S X L X P 19.285 5 3.857 2.65 .05 Subjects X P/S X L 552.366 380 1.453 Totala 1607.025 479 aTotal includes independent group and error terms not shown: refer to Table 22 for ratio equivalents of these terms. 112 The performance profiles across points (Figure 12 and Figure 13) show distinct differences among points but not between groups. Under the ANOVA the main effects of the training factors are the same as in the previous analyses and are not repeated. Of the remaining effects there was a significant point effect on error rate and reaction time (F (5, 380)'s = 7.12 and 5.34, respectively, p's < .001). There was no reliable sequence by point or length by point interaction on either score. There was, however, a significant sequence by length by point inter- action on reaction time scores (§_(5, 380) = 2.65, o < .05). The point effect, without an interaction by sequence, sug- gests that some points are more difficult to locate based on their position in the field and not on how they were accessed (routes during acquisition). The same inter- pretation was drawn from the kinesthetic transfer results. In combination, results from the two transfer tests are compelling evidence of a processing system which trans- lates prOprioceptive data directly into location infor- mation. It is not suggested that this is the only system implicated for spatial representation; others will be considered in the last section. As stated previously, there was no a priori basis for predicting the effect of field position on location performance. During acquisition blue, the point closest to the subjects' right-biased midline, was located with the .msowuwcsou coauwmwswod ucmuommwn mcflsoHHom umma ummmsmua Hmsmw> map :H musmom so ovum Houum mo mHHuoumnn.~H .mHm 113 2.2.9. 2.5 2...; :85 Ba :2...» 8...... H» H H H H H nv Au .. 0.. m. Au . m x” . a. m. III II III-I. \I I. I06. 4 On. m. nu 1 u... L 0v. 93. 53:3. Oullo :2.» E2231 0.3:... 93. 832:: I 1 On. :21 6.63.5 I .m20mumpsoo acmummmsvo< usoummman msH3oHHom umoa Hmmmsmua Hmsmm> 0:» cm muswom on mama soHuommm mo uHHmoumlu.ma .mHm 32.0.. 35 £23 5.5 2m 32.; ....E 114 H H H H H H (r0 1 .0 u mm \\ \ .L \ .\x. L ‘Il‘n‘l‘O'lunlllnlO'nl'l-‘O‘ 0000\hv .JfiV b3 \\ a 0‘ II In. a \Lvllnl .I fly 1“. “H .\.&.II n? \‘ o I o l s .\1\ I'll . 28. 58:3. 0......o toga 53:3. 7...... 9.2 63:5 To 1 w to... 5.5351 115 fastest movements while times to the central pink point were considerably slower. A scan of the marginal means in Table 20, however, revealed that the visual location per- formance was reversed on these points. Both performance scores were high on blue and low on pink. Using Scheffé's method, pink and blue were found to differ significantly from the overall mean on reaction time and error rate. The critical ranges were 290 msec and .05, respectively. This finding is predicted by the distinction between ego- centric kinesthetic reference and field-based visual reference introduced to explain the kinesthetic transfer results. That is, transfer performance on points referenced by visual location information, under visual recognition conditions, would exceed performance on points referenced with kinesthetic location information. Earlier the possibility was discussed that sub- jects used a radial reference system when location infor- mation was predominantly visual. For this system to be effective it requires that a central point be accurately fixed within the field. The finding that pink, only a few centimeters from the field's center of gravity, was the more accurately placed point supports the notion of a radial system by providing a fixed central point, relative to the set of remaining points. The fact that blue was the less accurately placed point may reflect the egocentric contractions suggested previously. If, indeed, egocentric 116 reference relies on kinesthetic location information, then the expected contraction would be greatest on points most obviously referenced this way. Before leaving the results of the visual transfer test, the issue of consolidation versus accuracy in the visual schema deserves further attention. Contrasting the reaction time and error rate patterns across analyses was less informative than had been expected. In general, error rate appeared to be a more sensitive measure of schemata differentiation, and differences have been discussed primarily in terms of location accuracy. While reaction times uniquely accounted for some of the differences observed between location performance on the lococentric and egocentric axes, the measure was not affected, overall, by length of training. It was presumed that the consoli- dation process would be a function of trials and this find- ing suggests that the process must have occurred early in training. The schema or schemata, however, may not be asymptotically consolidated by block 4. When individual points were considered, the significant sequence by length by point interaction indicated that the effect of length depended on sequence for at least some of the points. Separate interaction contrasts ((random EEEEE - random loog)-- (uniform EEEEE.‘ uniform logo)) for each point yielded the following combined reaction time means: pink = 1821 msec, yellow = 173 msec, red = 1996 msec, green = 1415 msec, 117 white = 451 msec, and blue = 29 msec. In every case the drOp in reaction time was greater for random subjects, yet the difference varied considerably across points. A Scheffé test of the comparison revealed that pink and red exceeded the critical range (1804 msec, o_< .05). Thus, the apparent consolidation of the schema for these points deteriorates with extended uniform training and improves with extended random training. One explanation for the opposing trends in consolidation would be the reliance on different representation domains as training progresses. Uniform subjects, by developing a representation mode which was primarily kinesthetic for some points, would rely less and less on a visual schema during acquisition. It is proposed that the lack of attention to a concurrent visual representation would leave the schema open to decay. Individual differences. Differences in visual memory for point location have been described in terms of group means and relative quantitative effects without regard to subjects' individual transfer performance pat- terns. Individual subject results were not compared with the respective groups pattern and, obviously, the findings specify the performance of some subjects better than others. For example, the degree of kinesthetic coding and ego— centric referencing undoubtedly varied among uniform sub- jects, and some may have developed accurate visual repre- sentations by structuring the space with a form of 118 exocentric reference immune to the systematic distortions which have been suggested for the visualization of ego- centrically referenced locations. Specifically, subjects less inclined to attend to kinesthetic feedback may have de-emphasized egocentric reference and relied on a system of fixing points relative to an external frame of refer- ence. A few subjects reported that they relied primarily on relating points to the rectangular perimeter of the enclosure. Conversely, some subjects in the random con- dition may have failed to exploit triangulation of visual- ized locations and relied on egocentric referencing. These possible variations are similar to variations discussed earlier to explain differences in the representation of selected points. It might be expected that performance on the visual transfer test would depend, to some extent, on an individ- ual's general spatial abilities and especially an individ- ual's nonanalytic visual spatial ability. Performance on tasks classified as nonanalytic visual spatial (after Maccoby and Jacklin, 1974) has revealed that, for adoles- cents and adults, spatial ability is not equally weighted between the sexes. Spatial tasks are typically more facile for males (see Maccoby and Jacklin, 1974). Some explanations for this difference are rooted in the neuro- logical and physiological differences found between the sexes (e.g., hormone balance: Broverman et a1., 1968; 119 Andrew, 1972: brain lateralization: Knox and Kimura, 1970; Buffery and Gray, 1972). Other explanations attribute the difference to socialization (see Kagan and Kagan, 1970; Sherman, 1967). A comparison by sex of the visual transfer per- formance indicated a slight advantage for males in terms of lower reaction times and error rates (see Table 26) by 357 msec and 6%, respectively. The lower reaction times were more evident on uniform segments while the lower error rates were more apparent after eight blocks of train- ing. Partial ANOVAs contrasting the Sequence and Length factors with the Sex factor were conducted and are summar- ized in Table 27 and Table 28. Neither the main effect nor the interaction effects were significant. Thus, though the differences were in the expected direction, the transfer task was not reliably easier for males. When the effects were adjusted for initial acqui- sition performance (block 1), an ANCOVA did not alter the null conclusions. The multiple regression effect (within cells) of acquisition movement time and reaction time accounted for 3.4% and 14.4% of the error variance for visual reaction time and error rate, respectively. The reason no sex differences were found on the visual test is not immediately clear. When the nature of the acquisition task is considered, the result, at first look, is perplexing. The presence of a perceptual motor (o— . —.—-—- —— ———.———-—..—. ...-..-————. ...—“*- an: _ “m...— .____._....—-.-.H-.HH..-——.-..-— wfi‘- cum—A 120 Table 26 Overall Performance on Visual Test by Sex of Subject Group Female Male Reaction time (sec) Uniform short 5.046 4.538 Random short 4.505 4.692 Uniform long 5.432 4.217 Random long Eiilé 3.785 4.665 4.308 Error rate Uniform short .31? .335 Random short .253 .285 Uniform long .390 .347 Random long _llol .222 .315 .297 E i! ii i? '9 121 Table 27 Partial ANOVA of Error Rate on Visual Test by Sex of Subject :1."- — w—m .HH- ————-— v -- ...-.— .o-o —-— . — - .. -—< nun—-- Source SS of MS F 2 Sex .00741 1 .00741 <1.0 Sex X Sequence (S) .00113 1 .00113 <1.0 Sex X Length (L) .03422 1 .03422 2.51 >.10 Sex X S X L .00247 1 .00247 .lO Sex X Sequence (S) 4.1159 1 4.1159 2.18 >.10 Sex X Length (L) .6909 1 .6909 .05 Sequence (S) .1993 l .1993 16.81 <.001 T X S .0037 l .0037 mmso¢ £OH£3 coaumEH0msH owumoummswx moommmooum mo Hoooz ¢I|.vH .mmm - rose: .23; 23 H ..s...m .5... 8:23am 22m 2km rose-2 , 2.2303 8.2.3.; o 2.2.325. o H H 3:: 3.29.2.1 toe-z toes: I :02 0060550th1 .OhOQF—OF £0051 bO-Oz xo02 151 body space corrdinates and visual space coordinates. This spatial information is stored in separate memories. This satisfies the findings of Diewert (1975) that location information suffers from both kinesthetic and visual inter- ference (filled interval) without postulating a combined kinesthetic—visual memory stage. There is no supporting evidence for such a memory stage which Diewert postulated under a misinterpretation of what Connolly and Jones meant by an "integrated store." This store is essentially the same as the visuo- motor store in the present model. The store has a long history of support beginning with Lashley (1917). It merely maps motor space onto visual space and is believed to be, to some extent, hard-wired (e.g., from the work with one to three week old infants by Bower, Braughton, and Moore (1970)). The dual simultaneous coding and storage of spatial information is considered automatic and not subject to attentional veto. This idea cpposes the Connolly and Jones model which asserts that the subject's knowledge about the upcoming transfer test determines which storage modality is used. In the present experiment, however, subjects were not informed of any transfer test (all reported that the visual test was unexpected). The coded visual information is not necessarily, as suggested above, immediately stored. As it comes out of the first short-term memory it is egocentrically referenced 152 and, under conditions described previously, may be suited to relational referencing processes such as triangulation. This precedes entry into the final visual memory stage. How the information residing in the various memo- ries of the model is used by the subject is depicted in Figure 15. When signaled to move to a particular point in the space, the subject "selects" a particular memory to access, retrieves the location, and generates the appropri- ate movement. The "attentional switch" is considered to be influenced by the general cognitive set or strategy scheme characterizing the subject at this point in time. The important factors are postulated to be the subject's memory of the past relative performance for this point, internal strengths of the two memory systems, and the level of con- fidence once a location is retrieved (arrow up from second short-term memory). The visuomotor store is tenta- tively positioned to allow translation of visual location data into motor space coordinates. However, this could be an internal feature of the response programmer. Input from the motor memory has been limited to this final stage of processing because it is believed that (1) this information is not available to central processes and (2) the infor- mation would most resemble motor codes associated with response programming. Assuming this information is used, the question whether or not it is switched is open to speculation. In any case, the information probably serves 153 SIGNAL I Stimulus 5 STM J Serial J Store f 1 Store [Performancol LMemory Kinesthetic Memory Visual Memory J I Visuomotor] L Store MOVEMENT Fig. 15.--A Model of the Selection Process for Retrieving Stored Spatial Information Leading to a Movement Response. 154 a secondary function augmenting location information only under specific demand situations such as speed versus accuracy. Whether the experimental findings and this model anticipate the character of everyday spatial learning in the macrospace will be determined only after further study. The current experiment has suggested some directions such research might take. Keeping in mind the stated qualifi— cations of the experiment, the following points may have particular relevance to learning macrospace: (l) the multiplicity of sequenced exposures to the space may deter- mine the quality of spatial knowledge; (2) this knowledge, in the form of visual schemata, may direct movement behavior within the space; (3) visual routes within the space may have no special representation in the schema, especially in terms of distance parameters; and (4) direct visual information may not be crucial to the development of visual-like representations of the space. The last point particularly applies to the special problem of the blind. In fact, all points bear on the issue of map training for the blind with minimum extrapolation from the actual con- ditions of the study. APPENDIX APPENDIX Apparatus Construction and Operation The general features and dimensions of the appara- tus shown in Figure l have been described in the Method section. As can be seen in Figure l, the sides of the box actually extended (6.5 cm) beyond the forward edges of the top and bottom. The box itself was constructed of 2 cm pine stock and covered with 3 mm perforated masonite panels. Each interior wall had a light fixture mounted at its mid- line near the top (44 cm above the base). The 18 cm long fixtures each held a four watt daylight fluorescent tube shielded by an aluminum hood. These lights provided an even, low level, illumination of the interior of the enclosure including the Formica floor. A panel of 2 mm phenolic sheet was used to shield the Formica surface from view. The panel was semicircular in shape with a radius of .66 m, then cut equally at each side, at right angles to the straight edge, to fit a width of .90 meters. The panel was held rigid by thin pine bracing extending the full width of the panel. The panel protruded about 5 cm beyond the front edge of the field. This ensured that no part of the field could be seen by the subject while the curvature allowed unrestricted arm movement. The panel was supported 155 156 20 cm above the base by lengths of 2 cm square pine fixed along the inside walls. Attached along the midline of the top of the enclo- sure a 43.2 cm wide by 31.8 cm high tunnel extended 1.53 m back from its forward point (8.3 cm back from the forward edge of the top). The tunnel was constructed of 2 cm ply- wood except for the top (6 mm masonite) and the removable front cover (6 mm plywood). All surfaces except the white formica were painted flat brown. The cooling fan of the slide projector mounted in the tunnel was always operated at high speed and provided a moderate level of background noise which served to mask extraneous sounds. The fan was left running throughout an experimental session. The apparatus was housed in two subrooms connected by a door in the plywood partition which divided the 2.75 m x 8 m room. The walls and ceiling were painted flat black and the floor was black linoleum. The enclosure was positioned in one of the subrooms so that the rear of the projection tunnel extended through an Opening in the partition into the adjoining subroom. The opening was sealed around the tunnel with black felt. The PDP-8 com- puter was housed in a neighboring room. Signal light and tone. A 3 cm diameter hole was located on the midline of the back wall, 3 cm above the base. Mounted behind the hole, so that its frosted plexi- glass screen was flush with the wall, was an inline 157 projection device manufactured by Electronic Engineer Corporation. The one-plane readout, Model 10R02, was fitted with 1820 lamps fed by a regulated 28 volt power supply. Six of the cells were wired and fitted with Roscolene filter material (frost #801, dark lemon #807, light red #821, bright pink #827, light blue #856, and light green #871). A 7.5 cm loudspeaker connected to a 1000 hertz audio oscillator was also mounted on the outside of the backwall near the inline projector. The lights were individually switched electronically. The transistor switching circuits for each light and the tone oscillator were wired to separate remote lines from the output of a PDP-8 minicomputer (Digital Equipment Corporation). The lights could also be switched by a hand- held control box (19 cm by 12.7 cm by 33 cm) connected to the computer junction box at the rear of the enclosure by a 5 m cable. The control box, in addition, had a tone switch and L.E.D. indicators for the state of each light and the tone. Enclosure base and kinesthetic field. The base of the enclosure, flush with the table top, was a shallow box structure .91 m wide by .7 m deep by 5 cm high with a 6 mm plywood bottom and a flat-white Formica top. The t0p was kept flat and stiff with a number of lengths of 2.5 cm by 5 cm pine bracing attached to the plywood bottom. The six magnetic reed switches positioned underneath the Formica 158 were held in place by 'Z' shaped aluminum brackets fitted with rubber grommets. The long axes of the switches were held perpendicular to the top and the distance between the switch and the top could be adjusted easily. This arrange— ment allowed a precise calibration to be made of each switch's sensitivity pattern. Each switch was connected by a remote line to the input of the PDP—8 minicomputer. The actual floor area of the enclosure was .87 m wide by .64 m deep. Centered within this area a pattern of points was devised which met a number of criteria, some of which had to be met jointly by the pattern of points and the pattern of movement between the points (or movement responses). The later aspects were considered in the Method section. The pattern was centered on the midline of the field but slightly closer to the front of the field (28 cm from the front edge of the base) with no point more than 22 cm from the center. In addition, the points were distributed so that the interpoint distances were unequal but never shorter than 16 cm. The position of the points, relative to the field center, is given in Table 32. The signal light on the back wall is taken as 0° and the angular component of the position is based on clockwise rotation. The operation of the apparatus was controlled by the PDP-8 computer which could be programmed to execute the entire experimental procedure. When the apparatus was 159 Table 32 Relative Position of Points from Field Center Point Angle (°) Distance (cm) Yellow 30 14.6 Green 84 21.5 Blue 153 16.7 White 240 20.6 Pink 289 6.7 Red 305 25.6 160 operated as a kinesthetic field, the color appearing on the screen of the signal light determined the function of each magnetic reed switch. When a color was switched on, the action gated all open switches except one. When this reed switch was activated by the magnetic disc the tone generator was switched on for as long as the disc remained in the switch's area of sensitivity for up to 500 msec. At the end of 500 msec of constant activation, the tone was switched off and the color of the signal light was changed according to the predetermined sequence. Activations of less than 500 msec had no effect other than momentarily switching on the tone. A clock was restarted whenever the signal light changed color. Subsequently, the clock was read twice; first, when the magnetic disc was moved out of the area of sensitivity of the closed reed switch which had instigated the color change, and again after the disc was held for 500 msec within the area of sensitivity of the reed switch for the point corresponding to the color cur- rently displayed. These times (reaction time and movement time, respectively) were automatically recorded and punched onto paper tape. Projection system and visual field. At the front end of the projection tunnel a 40.5 cm square back-surfaced mirror was mounted at a 45° angle. The mirror faced the rear of the tunnel and the Formica surface through a 40 cm wide by 32 cm deep hole cut in the top of the enclosure. 161 The bottom edge of the mirror came to the forward edge of the hole which was 16.5 cm back from the forward edge of the t0p of the enclosure. The mirror slid into shallow grooves cut in the tunnel sides. Forward of the mirror a 43 cm square of 6 mm plywood was fastened to the front of the tunnel with pressure fit pins at each corner. Both the mirror and the front cover could be removed easily. A Kodak Ektagraphic slide projector, Model B-Z, fitted with a 750 mm Ektanar lens, was positioned in the rear of the tunnel. The projector, mounted on a 10.8 cm high plywood box, was positioned with the center of the front lens 14.8 cm above the middle of the tunnel floor and .86 m from the center of the mirror face. The lens was fitted with an Uniblitz electronic shutter, Model 225LA35 (Vincent Associates) connected to a Model 100 shutter drive unit. The projector advance mechanism was electronically switched and the quartz-halogen light source was powered through a Type 116 Powerstat (Superior Electric Company) so that image brightness could be controlled. Both the electronic switch and the shutter drive unit were connected by remote lines to the output of the PDP-8 computer. The projection system circuitry included a subject— operated response switch connected by a remote line to the input of the PDP-8 computer. The hand-held device con- sisted of a push button momentary contact switch mounted 162 in a small plastic cylinder at the end of a cable running from the exterior of the enclosure. The projection system was also controlled by the PDP-8 computer. Depressing the response key, closed the shutter (blanking the display), read a clock, advanced the projector, initiated a 5 sec delay before the shutter was again Opened to project the next pattern in the series, and restarted the clock. The reaction time was recorded and punched onto paper tape. The photographic slides were produced using the same color filter material used in the signal light system. Each positive transparency consisted of a pink, yellow, red, green, white, and blue circle on a black background. Movement sequences. One cycle of the uniform sequence was defined as the series; pink-yellow-red-green- white-blue. One cycle of the random sequence was defined as the series; pink-yellow-green-white-yellow-red-white— blue—red-green-blue-pink-green-white-pink-yellow-white- green-yellow-red-blue-pink-red-green. Sequences were segmented into overlapping pairs of points by taking the last point and the first point in the cycle as a segment, and then every point with the preceding point as a segment. While segment always referred to a two-point connectedness, the connection could be con- sidered in two distinct ways. First, the connection was probabilistic. Given that a point occurred in the sequence as the first point of a given segment, there was a 163 sequence-wide probability that the succeeding point would be the end point of the segment. This meant that segments and their frequencies within a sequence defined the con- ditional probabilities of one point following another. The connection also could be viewed as geometric, in that segments defined how two points were related in space. That is, a segment represented a directed movement between two points generated by the sequence, and this line of movement had a certain length and orientation in the two- dimensional plane. The uniform sequence had the geometric feature that the orientations of route segments were markedly horizontal in the plane.“ This is clear in Figure 2 where the subject- to-point vectors are diagrammed. It should be noted that except for pink and yellow, the angle between segments and the subject-to-point vector were large (71° to 86°). This relationship was an essential design feature of the experi- ment. The random sequence was designed to maximize the randomness of the sequence while preserving the pairwise and unidirectional components of the uniform sequence. In addition, the sequence contained no segment reversals (i.e., reversing the order of points in P.25 segments did not duplicate any P.5 segments), and no three-point section of the sequence occurred twice in the same cycle. The 30-trial sequence used for the kinesthetic transfer test was: pink-yellow-blue-green-white-yellow- 164 red-pink-white-blue-red-green-yellow-blue—pink-green-white- red-pink-yellow-white-blue-green-yellow-red-blue-pink-white- red-green. This sequence preserved all of the features mentioned for the random sequence while introducing six new segments. Instructions and Procedure The subject was seated in front of the enclosure, handed an instruction card and asked to read along as the experimenter read aloud the following instructions. In this experiment you will be learning the location on this table-top of six points. All the points are located within the bounds of the white circle marked on the panel above the table-top. Each of the points is assigned a different color which is shown by the light on the back wall of the cubicle. The six colors are goo, GREEN, PINK, YELLOW, BLUE, AND WHITE. I will now show you the colors and ask you to name them. If you have any difficulty in telling them apart, please let me know. The room lights were turned off. After all colors were presented the procedure was repeated (order of presentation was random) until all colors were accurately named. The lights were then turned on, and the following instructions were given. You are to explore the area under the circle by moving this disc across the table-top. Because of the panel you will not be able to see the surface or your hand as you move the disc. Your task is to discover the location of each point when its assigned color appears. This is done by moving the disc around until a tone is heard. The tone means that you are on the correct point. That is, you are on the point that has been assigned the color being shown on the back wall light. When you find the point you must remain on the spot for # second. This means the tone will remain on for 165 5 second. After remaining on the point for % second, the tone will go off and the light will change color. You are then to begin searching for the next point which is assigned this new color. Again, the tone will sound when you are over the correct point and you must remain on the spot for 8 second before the light will change to the next color. I want to stress, at this time, that for each color there is one and only one point and that this point will always be in the same spot. You are to continue moving the disc around the table— tOp in search of the different points, as indicated by the color of the light, until the light finally goes dark. There is not necessarily any consistent order to the series of colors that will be shown and, at first, some colors may not appear as often as other colors. It is important that you pay attention to the colors and learn which color goes with which point on the table-t0p. You should move as quickly and as accurately as you can from one point to the next point, but be sure the light has changed color before you move off the point. Also, be sure to keep the disc in contact with the table-top surface. The session will take about 8 hour. At first, it may seem to be taking a long time to find the correct point for each color. If it does, do not feel dis- couraged—-it will become easier and your movements will become more direct as the colors are repeated and you have more practice. The subjects were asked whether they had any questions about the procedure before they were positioned in front of the enclosure. A subject sat with his chest less than 6 cm from the edge of the enclosure shield and his right hand placed over the magnetic disc positioned on the Formica. The subjects were told to keep the other hand in their lap throughout the experimental session. The experimenter then left the room telling the subject that he would be on the 166 other side of the partition. The room was darkened through— out kinesthetic trials. Following the last trial for visual groups, the experimenter entered the room, handed the subject another instruction card, and proceeded to read the following instructions. Now I would like to find out what kind of "mental image" of the arrangement of points you have formed. A series of slides will be projected on the table-tOp. Each slide will consist of six colored circles, each circle appearing similar to the light on the back wall. There will be a colored circle for each of the points you have learned. The color of the circle at each point will be the same as the color that was assigned to that point when you were learning its location. However, in each slide, ONE OF THE CIRCLES WILL BE MISPLACED. That is, its posIEion will not coincide with the true position for this point. When a slide is projected, I want you to determine which circle is out of position as quickly as you can, and then push this button and tell me which color appeared misplaced. A brief tone will come on about one second before a slide is presented. Once the slide is projected and you think you know which color is misplaced, do not delay in FIRST pushing the button and then reporting the color to me. If you are unsure about which circle is misplaced you should guess. Once you push the button, the slide will go off and a short delay will occur before the next slide is presented. Remember, there is always just one circle that is not in the right place. ' The instructions were varied slightly for the remote visual groups to accommodate the remote nature of the image. The second and third paragraphs of the instructions were changed as follows. 167 O A series of slides will be projected in front of you. Each slide will consist of six colored circles, one for each of the points you have learned. The position and color of the circles will be the same as the position and color of the points. Thus, the six circles will form a pattern similar to the pattern of the points on the table-top. However, in each slide the pattern of the circles will not be exactly the same as the pattern you learned for the location of the points. Instead, ooo ol the circles will be misplaced in relation to the other five circles. That 15, its position will not coincide with the cor- rect position for this point. After subjects had an Opportunity to ask questions, the room lights were turned off, and the first display was presented. During the session, the experimenter remained seated at a small table in the corner of the room, and, with the aid of a shielded table light, recorded the sub— ject's verbal responses. LI ST OF REFERENCES LIST OF REFERENCES Adams, J. A. A closed loop theory of motor learning. Journal of Motor Behavior, 1971, 3, 111-150. Andrew, R. J. Recognition processes and behavior with special reference to the effects of testosterone on persistence. In D. S. Lehrman, R. A. Hinde & E. Shaw (eds.), Advances lo the study of behavior (Vol. 4). New York: Academic Press, 1972. Appleyard, D. Styles and methods of structuring a city. Environment and Behavior, 1970, 2, 100-118. Appleyard, D., Lynch, K., & Myer, J. The view from the road. Cambridge, Mass.: MIT Press, 1964. Arbib, M. A. The metaphorical brain. New York: Wiley, 1972. Arnheim, R. Visual thinking. Berkeley: University of California Press, 1969. Attneave, F. How do you know? American Psychologist, 1974, lg, 493-499. Attneave, F. Representation of physical space. In A. W. Melton & E. J. Martin (eds.), Coding processes lo human memory. Washington, D.C.: V. H. Winston, 1972. Attneave, F., & Benson, B. Spatial coding of tactual stimulation. Journal o; Experimental Psychology, Bartlett, F. C. Remembering. Cambridge: Cambridge Univer- sity Press, 1932. Bower, T. G. R., Broughton, J. M., & Moore, M. K. Infant response to approaching Objects: an indicator of response to distal variables. Perception and Psychophysics, 1970, 9, 193-196. 168 169 Broverman, D. M., Kaiber, E. L., Kobayashi, Y., & Vogel, W. Roles of activation and inhibition in sex differ- ences in cognitive abilities. Psychological Review, 1968, lo, 23—50. Brown, W. Spatial integrations in a human maze. University o£ California Publications lo_Psychology, 1932, o, 123-134: Bruner, J. S. On cognitive growth. In J. S. Bruner, R. R. Olver, & O. M. Greenfield (eds.), Studies lo cognitive growth. New York: Wiley, 1966. Buffery, A. W. H., & Gray, J. A. Sex differences in the development of spatial and linguistic skills. In C. Ounsted & D. C. Taylor (eds.), Gender differences: their ontogeny and significance. Edinburgh: Churchill L1v1ngstone, 1972. Carr, S., & Schissler, D. The city as a trip: perceptual selection and memory in the view from the road. Environment and Behavior, 1969, l, 7-36. Connolly, K., & Jones, B. A developmental study of afferent- reafferent integration. British Journal o: Psychology, 1970, El, 259-266. COOper, L. A., & Shepard, R. N. Chronometric studies of the rotation of mental images. In W. G. Chase (ed.), Visual information processing. New York: Academic Press, 1973. Diewert, G. L. Retention and coding in motor short-term memory: a comparison of storage codes for distance and location information. Journal o: Motor Behavior, 1975, 1, 183-190. Downs, R. M., & Stea, D. Image and environment. Chicago: Aldine, 1973. Duncan, D. P. Transfer after training with single versus multiple tasks. Journal o: Experimental Psychology, 1958, 22' 63-72. Edgington, E. S. Kinesthetically guided movements of head and arm. Journal ol Psychology, 1953, lo, 51-57. Fiori, N., Semjen, A., & Requin, J. Analyse chronometrique du pattern preparatoire a un mouvement spatialement oriente resultats preliminaires. £3 Travail Humain, 1974, ll, 229-248. 170 Fitts, P. M. The information capacity of the human motor system in controlling the amplitude of movement. Journal o£_Experimental Psychology, 1954, ll, Fitts, P. M., & Peterson, J. R. Information capacity of discrete motor responses. Journal o: Experimental Psychology, 1964, 911 103-112. Gibson, J. J. The senses considered oo perceptual systems. Boston: Houghton-Mifflin, 1966. Hall, E. T. The hidden dimension. New York: Doubleday, 1966. Harris, L. J. Sex differences in spatial ability: possible environmental, genetic, and neurological factors. In M. Kinsbourne (ed.), Hemispheric asymmetries of function. Cambridge: Cambridge University Press, 1975(a). Harris, L. J. Neurophysiological factors in the develop- ment of spatial skills. In J. Eliot & N. J. Salkind (eds.), Chlldren's s atial development. Springfield, 111.: Thomas, 19 5(b). Hart, R. A., & Moore, G. T. The development of spatial cognition: a review. In R. M. Downs & D. Stea (eds.), Image and environment. Chicago: Aldine, 1973. Held, R. Exposure history as a factor in maintaining stability of perception and coordination. Journal o: Behavior and Mental Diseases, 1961, 132, 26-32. Howard, I. P., & Templeton, W. B. Human spatial orien- tation. London: Wiley, 1966. Huttenlocher, J. Constructing spatial images: a strategy in reasoning. Psychological Review, 1968, 12' 550-560. Ittleson, W. H., Proshansky, H. M., & Rivlin, L. G. The environmental psychology of the psychiatric ward. In H. M. Proshansky, W. H. Ittleson & L. G. Rivlin (eds.), Environmental s cholo . New York: Holt, Rinehart and Winston, E§)0. Kagan, J., & Kogan, N. Individuality and cognitive per- formance. In P. H. Mussen (ed.), Carmichael's manual o; child psychology. New York: Wiley, 1970. 171 Keele, S. W. Movement control in skilled motor performance., Psychological Bulletin, 1968, 12, 387-403. Klapp, S. T. Feedback versus motor programming in the con- trol of aimed movements. Journal oi Experimental Ps cholo : Human Perception and Performance, 1975, 104 147-153. I Kosslyn, S. M., Pick, H. L., & Fariello, G. R. Cognitive maps in children and men. Child Development, 1974, lo, 707-716. Laabs, G. E. Retention characteristics of different repro- duction cues in motor short-term memory. Journal of Experimental Psychology, 1973, 100, 168—177. Lashley, K. S. The problem of serial order in behavior. In L. A. Jeffrees (ed.), Cerebral mechanisms lo behavior. New York: Wiley, 1951. Lashley, K. S. The accuracy of movement in the absence of excitation from the moving organ. American Journal o£ Physiology, 1917, ll, 169-194. Levy-Argresti, J., & Sperry, R. W. Differential perceptual capacities in major and minor hemispheres. Pro- ceedings o: the National Academy o: Science, 1968, 61, 1151. Luria, A. R. The neuropsychology o£_memory. New York: Halsted Press, 19 6 Luria, A. R. The mind of the mnemonist. New York: Basic Books, 1968. '—_ Luria, A. R. Neuropsychology in the local diagnosis of brain damage. Cortex, 1964, l, 1-18. Maccoby, E. M., & Jacklin, C. N. The s cholo o; sex differences. Stanford, California: Stanford University Press, 1974. MacNeilage, P. F. Motor control of serial ordering of speech. Psychological Review, 1970, 11, 182-196. Marteniuk, R. G. Retention characteristics of short-term memory cues. Journal o; Motor Behavior, 1973, 2' 312-317. Marteniuk, R. G., & Roy, E. A. The codability of kines- thetic location and distance information. Acta Psychologica, 1972, lo, 471-479. 172 McGuiness, D. Sex differences in the organization of per- ception and cognition. In B. Lloyd & J. Archer (eds.), Exploring sex differences. New York: Academic Press, 1976. Megaw, E. D. Directional errors and their correction in a discrete tracking task. Ergonomics, 1972, lo, 633-643. Miles, W. Unidirectional mazes for studying human learning. Journal oi General Psychology, 1931, 2, 269-275. Milner, B. Interhemispheric differences in the locali- zation of psychological processes in man. British Medical Bulletin, 1971, ll, 272-277. Morrison, D. F. Multivariate statistical methods. New York: McGraw-Hill, 1976. Neisser, U. Cognitive psychology. New York: Appleton, Century, Crofts, 1967. Newell, A. You can't play 20 questions with nature and win: projective comments on the papers of this symposium. In W. G. Chase (ed. ), Visual information processing. New York: Academic, 1973. Olton, D. 8. Spatial memory. Scientific American, June 1977, 236(6), 82-98. Pailhous, J. Influense de l'ordre de presentation des donnees sur la constitution de l'image spatiale. lo_Travail Humain, 1972, 2;, 69-84. Palmer, S. E. Visual perception and world knowledge: Notes on a model of sensory-cognitive interaction. In D. A. Norman & D. E. Rumelhart (eds.), . Explorations in cognition. San Francisco: W. H. Freeman, 1975. Perrin, F. A. C. An experimental and introspective study of the human learning process in the maze. Psychological Monographs! 1914, lol70). Petkovich, M. Development of the fgrammar action": an analysis of movement sequences in children' s draW1ngs 5f geometric forms. Unpublished master' 3 thesis, Michigan State University, 1974. 173 Pew, R. W. Human perceptual motor performance. In B. D. Kantowitz (ed.), Human information processing: tutorials lo porformance and cogoition. Hillsdale, N.J.: Erlbaum, 1974. Piaget, J., & Inhelder, B. The child's conception of space. London: Routledge & Kegan Paul, 1956. Posner, M. I. Characteristics of visual and kinesthetic memory codes. Journal of Experimental Psychology, 1967, 12' 103-107. Posner, M. I. Cognition: go introduction. Glenview, Ill.: Scott-Foresman, 1973. Pylyshyn, Z. W. What the mind's eye tells the mind's brain: a critique of mental imagery. Psychological Bulletin, 1973, 1-24. Rand, G. Some Copernican views of the city. Architectural Forum, 1969, 132, 77-81. Schmidt, R. A. A schema theory of discrete motor skill learning. Psychological Review, 1975, fig, 225-260. Schmidt, R. A., & Wrisberg, C. A. Further tests of the Adams' closed-loop theory: response produced feedback and the error-detection mechanism. Journal of Motor Behavior, 1973, 2, 155-164. Scott, T. C. Imagery--to what extent is it a function of maze pattern? Journal o£_Genetic Psychology, 1936, 12! 271-278. Semmes, J. Hemispheric specialization: a possible clue to mechanisms. Neuropsychologia, 1968, o, ll-26. Shaffer, R. W., & Ellis, H. C. An analysis of intersensory transfer of form. Journal of Experimental Psychology, 1974, 102, 948-953. Shemyakin, F. N. Orientation in space. In B. G. Anan'ev, G. S. Kostyuk & others (eds.), Psychological science in the USSR (Vol. 1). Washington, D.C.: Office of_T§EHnicaI Services, 1962. (Report #62,11083). Shepard, R. N. Form, formation, and transformation of internal representations. In R. L. Solso (ed.), Information processing and cognition: The Loyola Symposium. Hillsdale, N.J.: Erlbaum, 1975. 174 Shepard, R. N., & Metzler, J. Mental rotation of three- dimensional Objects. Science, 1971, 171, 701-703. Sherman, J. A. Problems of sex differences in space per- ception and aspects of intellectual functioning. Psychological Review, 1967, 1:, 240-299. Siddall, E. J., Holding, D. H., & Draper, J. Errors of aim and extent in manual point to point movement. Occupotional Psychology, 1957, ll, 185-195. Siegel, A. W., & White, S. H. The development of spatial representations of large-scale environments. In H. W. Reese (ed.), Advances in Child Development (Vol. 10). New York: Academia Press, 197*. Slinger, R. T., & Horsley, U. Upon the orientation of points in space by the muscular, arthrodial, and tactile senses of the upper limbs in normal indi- viduals and in blind persons. Brain, 1906, 32, 1-27. Smothergill, D. W. Accuracy and variability in the local- ization of spatial targets at three age levels. Developmental Psychology, 1973, o, 62-66. Stea, D., & Blaut, J. M. Notes toward a develOpmental theory of spatial learning. In R. M. Downs & D. Stea (eds.), Image and environment. Chicago: Aldine, 1973. Stelmach, G. E., Kelso, J. A. S., & Wallace, 8. A. Pre- selection in short-term motor memory. Journal o; Ex erimental Psychology: Human Learning and Memogy, 1975, l, 745-755. Teichner, W. H., & Krebs, M. J. Laws of visual choice reaction time. Psychological Review, 1974, §l 75-98. Tolman, C. Cognitive maps in rats and men. Psychological Review, 1948, 22' 189-208. Warden, C. J. The relative economy of various modes of attack in the mastery of the stylus maze. Journal of Experimental Psychology, 1924, 7, 243-275.—_ _ Welford, A. T. Fundamentals o: skill. London: Methuen Press, 1968. 175 Wilton, R. N., & File, P. A. Knowledge of spatial relations: a preliminary investigation. Quarterly Journal of Experimental Psychology, 1975, 2 , 51-257. Woodworth, R. D. Experimental psychology. New York: Holt, 1938. Worchel, P. Space perception and orientation in the blind. Psychological Monographs, 1951, o§(332). MMMMM "I"iii/uiiiiiiiiimn‘“