Tacit knowledge and the effect of pattern configuration on mental scanning

Transcription

1 Memry & Cgnitin 1983, Vl. 11(2), Tacit knwledge and the effect f pattern cnfiguratin n mental scanning STEPHEN K. REED and HOWARD S. HOCK Flrida Atlantic University, Bca Ratn, Flrida and G. R. LOCKHEAD Duke University, Durham, Nrth Carlina Tw ppsing hyptheses are that estimated distances may be used t predict scanning times in mental scanning experiments (Mitchell & Richman, 1980; Pylyshyn, 1981) and that SCAnning times may be used t estimate distances n maps (Thrndyke, 1981). We tested these hyptheses by having peple bth estimate lengths and mentally scan diagnal lines, spirals, and mazes. There were large and cnsistent differences in the rate f SCAnning the three cnfiguratins, regardless f whether peple scanned percepts r images. These differences culd nt be accunted fr by differences in length estimates r by peple's tacit knwledge f their scanning rates. We prpse that peple actually d mental scanning in a mental scanning task rather than use length estimates t generate scanning time and that they d nt rely n scanning times t estimate length in a length estimatin task. The study f visual imagery has emerged as ne f the central tpics f cgnitive psychlgy. Althugh there nw appears t be extensive evidence that peple use visual images t perfrm many cgnitive tasks (Ksslyn & Pmerantz, 1977), critics have ccasinally questined whether sme f this evidence might have been prduced by subjects wh anticipated the results and generated apprpriate data withut using visual images. The tacit knwledge view, prpsed by Pylyshyn (1981), suggests that subjects use their tacit knwledge f what shuld ccur in reality t simulate peratins n images withut actually perfrming the peratins. A specific example is the Ksslyn, Ball, and Reiser (1978) fmding that the time it takes t scan between tw bjects in a visual image is a linear functin f the. distance between the tw bjects. Ksslyn et a1. interpret this fmding as evidence that peple can image a dt mving at a cnstant rate frm ne bject t anther. In cntrast, the tacit knwledge view suggests that peple culd simulate this result by using their knwledge that it takes an bject lnger t mve thrugh a greater distance. The results might therefre be based n estimated time intervals rather than n actually imagining a dt mve acrss an image (Pylyshyn, 1981, pp ). This alternative t image scanning was suggested previusly by Mitchell and Richman (1980), wh bserved that the Ksslyn et a1. (1978) fmdings culd have been prduced by the demand characteristics f the experiment. If the hypthesized relatinship between distance and scanning time was bvius t the subjects, they may have generated the expected results withut actually scanning an image. In rder t investigate subjects' tacit knwledge abut the scanning task,mitchell and Richman asked subjects t estimate hw lng it wuld take them t scan between tw bjects. The subjects prduced a high crrelatin between distance and predicted scan time, demnstrating that they culd anticipate Ksslyn et al.'s results. The Mitchell and Richman (1980) fmdings supprt the plausibility f the tacit knwledge interpretatin f visual scanning. If peple can predict the relative time required t scan ver different distances, they may respnd n the basis f internally generated time intervals rather than n the basis f internally generated images. Neither Mitchell and Richman (1980) nr Pylyshyn (1981) argue that subjects' ability t predict the image-scanning data caused the riginal results. Hwever, bth crrectly argue that if the image-scanning paradigm is t be a useful tl in the study f imagery, it is necessary t distinguish between the amunt f variance due t the nature f imagery and the amunt due t tacit knwledge and the demand characteristics f the experiment. We are grateful t Nelly Saavedra fr testing subjects. Requests fr reprints shuld be addressed t Stephen K. Reed, Department f Psychlgy, Flrida Atlantic University, Bca Ratn, Flrida Varying Pattern Cnfiguratin Althugh peple can successfully predict that scanning time increases with distance (Mitchell & Richman, 1980), they may nt be able t predict hw cnfigura- 137 Cpyright 1983 Psychnmic Sciety, Inc.

2 138 REED, HOCK, AND LOCKHEAD Line / / / / / // Spiral f@ Maze @) Figure 1. The three sets f line cnfiguratins. tin influences scanning time. T examine this, we used a variatin f the mental scanning task in which peple scanned three different cnfiguratins. Figure 1 shws the three cnfiguratins, which are labeled line, spiral, and maze. If pattern cnfiguratin influences scanning time, but peple are unaware f this effect, the data wuld supprt the view that peple are actually scanning the patterns rather than using tacit knwledge and an internal clck t generate the results. In investigating hw cnfiguratin influences perfrmance in scanning experiments, we are interested in tw parameters. The fust parameter is the slpe f the best-fitting linear regressin equatin describing hw scanning time increases as a functin f length. This measures the rate f scanning. The secnd parameter is the y-intercept f that linear regressin equatin. This is particularly relevant when peple are asked t estimate scanning time, because it may reflect tacit knwledge abut cnfiguratins that is manifested by an additive cnstant that is independent f length. Fr example, peple may have tacit knwledge that they will scan mazes mre slwly than lines because they have t slw dwn fr crners. Estimating lengths. The tacit knwledge hypthesis prpses that peple use their estimates f length t predict scanning time. In cntrast t this, Thrndyke (1981) has prpsed that scanning time is used t estimate length. His task required that subjects estimate distances n maps. Their distance estimates increased as a functin f the number f pints (such as the names f cities) alng the rute. T accunt fr this fmding, Thrndyke prpsed that subjects perceptually scan a rute r an image f a rute and use scan duratin t estimate path distance. The mdel assumes that intervening pints slw dwn scanning, because subjects must stp and retrieve the name f the city t determine whether it is their destinatin pint, and this results in estimates f greater distance. Thrndyke did nt directly test this hypthesis by measuring scanning time. The cmpeting hyptheses, that distance estimates are used t predict scannning times (Mitchell & Richman, 1980; Pylyshyn, 1981) and that scanning times are used t estimate distances (Thrndyke, 1981), suggest that we have an incmplete knwledge f hw scanning time and distance are related. The fllwing experiments investigate this relatin by requiring that subjects participate in bth a mental scanning and a length estimatin task. In the fust experiment,subjects were instructed t estimate the lengths f the 24 patterns shwn in Figure 1 and t mentally scan the patterns s we culd measure scanning time. In ne cnditin, each pattern remained n the screen until the subject respnded. This allwed fr scanning f the pattern. In a secnd cnditin, the stimulus was n the screen fr.5 sec, which is nt sufficient time t perfrm this task, and subjects were instructed t base their respnse n a visual image f the pattern. In the secnd experiment, subjects' tacit knwledge was examined by requiring that they estimate their scanning time fr each pattern. EXPERIMENT 1 Methd Subjects. The subjects were 24 undergraduates at Flrida Atlantic University wh received $3 fr participating. There were 12 subjects in the perceptin grup and 12 subjects in the image grup. Stimuli. The stimuli cnsisted f the 24 patterns shwn in Figure 1. They were rear-prjected nt a 14 in. (height) x 21 in. (width) translucent screen placed at eye height. The patterns appeared in the center f a rectangular slide frame that was 5.5 in. high and in. lng. The eight different lengths f each pattern were similar acrss patterns:.8, 1.7, 2.8, 3.5, 4.3, 5.1, 5.9, and 7.0 in. fr the diagnal line;.8, 1.7, 2.8, 3.4, 4.2, 4.8,5.9, and 6.8 in. fr the spiral; and.7, 1.5,2.6,3.3,4.0, 4.9, 5.7, and 6.7 in. fr the maze. Subjects viewed the patterns frm a distance f 3 ft. At this distance, the largest pattern in each series subtended a visual angle f 8 deg fr the diagnal line, 2.4 deg fr the spiral, and 1.6 deg fr the maze. Prcedure. Subjects in bth the perceptin and image grups participated in tw tasks: a visual scanning task and a length estimatin task. Fr the perceptin cnditin, six f the subjects did the scanning task fust and the ther six did the length estimatin task fust. Each f the 24 patterns ccurred twice during each task. The patterns ccurred in a randm rder, except fr the cnstraint that the secnd presentatin f a pattern did nt ccur until all patterns had been presented. Subjects read the fllwing instructins fr the visual scanning task: "This task was designed t study hw well peple can

3 TACIT KNOWLEDGE AND MENTAL SCANNING 139 mentally scan bjects. We will be shwing yu three different kinds f bjects: diagnal lines, curved spirals, and straight spirals. The bjects will appear, ne at a time, n the screen in frnt f yu. We want yu t scan the bject by imaging a little black speck mving frm ne end f the line t the ther. Begin by fcusing n the left end f the diagnal line (if a diagnal line is presented) r the end pint inside a spiral. When yu are ready t begin scanning, press the buttn n the switch and immediately image the speck mving alng the line. The speck shuld mve as quickly as pssible while still remaining visible. As sn as yu reach the end f the line, push the buttn again. The length f the line will vary fr each f the bjects. Try t mve the speck at the same speed even thugh the line lengths and bjects will change. Yu will have nine practice trials befre we begin." The fllwing instructins were used fr the length estimatin task: "This task was designed t study hw well peple can estimate the lengths f lines. We will be shwing yu three different kinds f lines: diagnal lines, curved spirals, and straight spirals. The lines will appear, ne at a time, n the screen in frnt f yu. We want yu t estimate the length f each line in inches. Yu may make yur estimate t the nearest tenth f an inch if yu wish. Yu will have nine practice trials befre we begin." The nine practice trials cnsisted f the shrtest, the lngest, and an intermediate length fr each f the three patterns. Subjects did nt receive any feedback during the practice trials but were encuraged t use a faster scanning rate if they exceeded 6 sec n any f the practice patterns. After cmpleting bth tasks, subjects answered several questins abut strategies. Fr the image cnditin, the ther 12 subjects fllwed the same prcedure as the perceptin grup, except that the patterns were presented fr nly.5 sec and the subjects were instructed t base their judgments n a visual image f the pattern. They were infrmed that the purpse f the experiment was t study hw well peple can scan visual images and use visual images t estimate the length f lines. Results Figure 2 shws the relatin between estimated length and actual length fr each f the three patterns (right clumn) and shws the scanning time fr each pattern (left clumn), when the subjects scanned the pattern (tp rw), and when they scanned their image r memry f the pattern (middle rw). Each pint in these figures is the average f 24 judgments (12 subjects X 2 judgments). Length estimatin. The grup crrelatin between actual and estimated length was 0 fr each f the three patterns in the perceptin cnditin, in which the pattern remained n the screen until the respnse. The average f the 12 individual subject crrelatins was.99 fr the line,.97 fr the spiral, and.98 fr the maze. These results shw that peple are cnsistent in their length estimates and that this cnsistency is maintained fr cmplex patterns, such as spirals and mazes, as well as fr straight lines. Fr the image cnditin, in which the pattern appeared fr nly.5 sec, the crrelatins based n grup averages were 0 fr the line,.99 fr the spiral, and 0 fr the maze. The average f the individual crrelatins was.98 fr the line,.97 fr the spiral, and.96 fr the maze. This extends the results f the perceptin cnditin by shwing that peple can maintain a high level f cnsistency in their length estimates even when they view a pattern fr nly a brief time. The slpes f the regressin lines in the perceptin cnditin were 3 fr the line,.92 fr the spiral, and 5 fr the maze. The slpes in the image cnditin were.87 fr the line,.69 fr the spiral, and.75 fr the maze. A 2 (grups) by 3 (patterns) analysis f variance revealed significant main effects fr bth grup [F(1,22) ::: 4.52, p <.05] and pattern [F(2,44)::: 4.31, p <.05], but the Grup by Pattern interactin was nt significant [F(2,44) < 1]. The significant grup effect was caused by the underestimatin f length in the image cnditin. The underestimatin is cnsistent with the results f Myer, Bradley, Srensen, Whiting, and Mansfield (1978), wh reprted that remembered size is less than perceived size when peple judge length, area, r vlume.! The significant effect f pattern was caused by the underestimatin f spiral length relative t line length. A Newman-Keuls analysis revealed that the slpe fr the spiral was significantly less (p <.05) than the slpe fr the diagnal line, but neither pattern differed significantly frm the maze. If peple used length estimates t simulate mental scanning, scanning times fr the spiral shuld be less than fr the diagnal line. As shwn in Figure 2, this did nt ccur. Scan time. The grup crrelatins between scan time and line length in the perceptin cnditin (tp left panel in Figure 2) were.95 fr the line,.99 fr the spiral, and.99 fr the maze. The average f the individual crrelatins was.86 fr the line,.93 fr the spiral, and.94 fr the maze. The clse relatin between scan time and physical length when the figure is scanned was maintained when subjects were asked t scan images (middle left panel in Figure 2). The grup crrelatins were.99 fr the line,.97 fr the spiral, and.99 fr the maze. The average f the individual crrelatins was.81 fr the line,.91 fr the spiral, and.94 fr the maze. Perhaps the mst striking characteristic f the results was the effect f pattern n scan rate. As measured by the slpes f the regressin lines, the rate f scanning in the perceptin cnditin was.17 sec/in. fr the line,.30 sec/in. fr the spiral, and.44 sec/in. fr the maze. The crrespnding values were.23,.37, and.47 fr the image cnditin. A 2 (grups) by 3 (patterns) analysis f variance indicated significantly different scan rates acrss patterns [F(2,44)::: 46.14, p <.01]. All three scan rates differed Significantly (p <.01) frm each ther accrding t a Newman-Keuls analysis. Neither the effect f grup [F(1,22) < 1] nr the Grup by Pattern interactin [F(2,44) < 1] was significant. Scanning rates were essentially the same when subjects scanned a perceptual pattern as when they scanned an image f the pattern. Table 1 summarizes the crrelatins and regressin parameters fund in Experiment 1. The table als shws the results f an experiment (labeled Experiment 1A) that we cnducted befre Experiment 1. Since the

4 140 REED, HOCK, AND LOCKHEAD u; 3.5 "t:> GJ V> 2.5 i= 2.0 z (f) 0.5 Maze Y=.44x +.99 Spiral Y =.30x +.81 Line Y=.I7x +.76 '" 6.0.c; '" u c;; I aj ' w!:i 3.0 i= 2.0 Mze Y=I.5x +.66 Splrl Y=.92x +.50 Line Y=I.3x +.36 ID D a D u; 3.5 "t:> b 2.5 i= 2.0 z (f) ID 0.5 Maze Y=.47x +.84 Spiral Y=.37x +.69 Line Y =23x + 9 '" 6.0 u c;; I b aj 4D ---' w t:i 3.0 i= 2.0 Mze Y=.75x +.51 Splrl Y=.69x +.39 Line Y =.87x D 3.0 4D 5D 6D 7.0 LENGTH (inches) 2D D 6D 7.0 LENGTH (inches) U; 4.5 "t:> c 84.0 GJ V> ;: 3.5 i= 3D z «25 [ w 0.5 Mze Y=A2x+.93 Spiral Y=.37x+.68 Line. Y=.39x +.48.c; '" 6.0 u I f-- t9 aj ' w!:i 3.0 t;; w 2.0 Mze Y=LOOx+.70 Spiral Y=.96x +.42 line Y=I.02x D 4D 6D 7[) 2D 3D 4.0 6D 7D Figure 2. Scanning times and estimated lengths fr the perceptin (tp rw) and imagery (middle rw) cnditins f Experiment 1. The bttm rw shws estimated scanning times and estimated lengths in Experiment 2.

5 TACIT KNOWLEDGE AND MENTAL SCANNING 141 Table I Crrelatins and Regressin Parameters fr Scan Time and Length Estimates Task Scan Time Length Estimate Experiment Cnditin Stimulus G 1 Perceptin Line I Perceptin Spiral Perceptin Maze I Imagery Line I Imagery Spiral Imagery Maze IA Perceptin Line IA Perceptin Spiral IA Perceptin Maze Imagery Line Imagery Spiral Imagery Maze Nte-G = grup, I = individual. r Slpe Intercept G Slpe Intercept prcedure was identical t the prcedure fllwed by the perceptin grup in Experiment 1 and prduced the same pattern f results, we have summarized these results withut reprting them in detail. The cmbined results demnstrate that the cnfiguratin f the line being scanned influences scanning time, and it appears that these cnfiguratins have little influence n length estimatin. A tacit knwledge hypthesis based nly n time and extent cannt accunt fr these results, since perceived length des nt predict scanning time acrss cnfiguratins. Fr that hypthesis t be supprted, it wuld be necessary fr peple t knw hw variatins in the pattern f the lines influence their scan time. This is cnsidered in Experiment 2. EXPERIMENT 2 A cnsistent rmding in Experiment 1 was that pattern cnfiguratin influenced scanning rate. Subjects scanned diagnal lines significantly faster than spirals and spirals significantly faster than mazes. In rder fr the tacit knwledge hypthesis t accunt fr these results, peple wuld have t anticipate hw pattern cnfiguratin wuld influence scan time. Althugh peple can anticipate that scan time increases with linear distance (Mitchell & Richman, 1980), it is nt clear whether they can anticipate the differential cnfiguratin effect. The purpse f Experiment 2 was t evaluate this by asking them t estimate scan time fr the 24 patterns used in Experiment 1. 2 Methd Subjects. The subjects were 12 undergraduates at Flrida Atlantic University wh were paid $3 fr their participatin. Prcedure. The prcedure was identical t that used in the image cnditin f Experiment 1 fr bth the length estimatin and scanning tasks, except that subjects estimated scan time during the "scanning" task. The near identity f the tw prcedures was enhanced by initially giving subjects the same instructins used in the image cnditin f Experiment l. After they read the instructins, subjects were given the fllwing mdified instructins fr the scanning task: ''The instructins that yu read were used in a previus experiment. In this experiment we want yu t estimate hw lng yu wuld take t scan an image withut actually scanning it. After the pattern ges ff the screen, estimate hw lng it wuld take yu t scan the image if yu attempted t fllw the previus instructins. Yu may make yur estimate t the nearest tenth f a secnd if yu wish. Yu will have nine practice trials befre we begin." Results The bttm right panel in Figure 2 shws the relatin between the actual length and the estimated length. The crrelatin between the tw variables was.99 fr the line,.99 fr the spiral, and 0 fr the maze. The means f the individual crrelatins were.95 fr the line,.97 fr the spiral, and.97 fr the maze. The slpe f the regressin line was 2 fr the line,.96 fr the spiral, and 0 fr the maze. The slpes indicate that the average length judgments were very accurate, in cntrast t the underestimatin f length fund fr the imagery cnditin in Experiment 1. The differences in slpe acrss patterns were nt significant [F(2,22) < 1]. The bttm left panel in Figure 2 shws that estimated scan time increased with length, as was als reprted by Mitchell and Richman (1980). The grup crrelatins were 0 fr the line,.99 fr the spiral, and.98 fr the maze. The average individual crrelatins were.80 fr the line,.77 fr the spiral, and.82 fr the maze. Unlike the scanning results, these estimated scan rates did nt differ significantly acrss patterns [F(2,22) < 1]. The estimated scan rate was.39 sec/in. fr the line,.37 sec/in. fr the spiral, and.42 sec/in. fr the maze. Hwever, there was a significant difference in the y-intercepts [F(2,22) = 5.18, P <.05]. The intercept

6 142 REED, HOCK, AND LOCKHEAD values f.48 sec fr the line,.68 sec fr the spiral, and.93 sec fr the maze may reflect that peple scan the line mre quickly than the spiral and the spiral mre quickly than the maze. Peple may have tacit knwledge abut the relative time it takes t scan different cnfiguratins, which is reflected by an additive cnstant (y-intercept) rather than by scan rates (slpe differences). These differences amng the y-intercepts generally crrespnd t thse fund in Experiment 1. The values fr the line, spiral, and maze were.76,.81, and.99 fr the perceptin cnditin and.69,.69, and.84 fr the image cnditin. Thse times were nt significantly different [F(2,44) = 2.77, p <.08], but the pattern f the fmdings and significantly different values f.73,1, and 1.17 in Experiment la [F(2,22) = 9.67, p <.01] may suggest a respnse bias due t tacit knwledge, in additin t the actual time it takes t mentally scan a pattern. DISCUSSION The results f these studies can best be accunted fr by assuming that length estimatin and visual scanning are separate and independent tasks. We did nt fmd any supprt either fr the hypthesis that length estimates are used t simulate mental scanning r fr the hypthesis that scanning time is used t estimate lengths. The results are mst cnsistent with the hypthesis that peple actually scan in the mental scanning task and d nt rely n scanning time t estimate lengths. We cnsider the length estimatin task first and then the scanning task. Length Estimatin The argument that scanning time did nt influence length estimates is supprted by the lack f relatinship between the tw variables. There were large and cnsistent differences in scanning times acrss the three cnfiguratins. Scanning was fastest fr the line, intermediate fr the spiral, and slwest fr the maze. Thus the scanning time hypthesis shuld predict that the line wuld be judged the shrtest and the maze the lngest. Instead, the spiral was always judged the shrtest and the line was judged similar t the maze. Anther argument against the scanning time hypthesis is that the crrelatins between length estimates and length are higher than the crrelatins between scanning time and length. This is particularly evident fr the diagnal line. The average individual crrelatins between scan time and length were.86 and.81 in the tw cnditins f Experiment 1. The crrespnding crrelatins between estimated length and length were.99 and.98. The crrelatins raise the questin fr the scanning time hypthesis f hw a less cnsistent judgment (scan time) can serve as the basis fr a mre cnsistent judgment (estimated length). In rder t give subjects the pprtunity t reprt n their strategies, we designed a questinnaire in which we asked subjects t rate hw ften they used each f several alternative strategies. The ratings were 0 (nt at all), 1 (seldm), 2 (abut half f the time), 3 (mst f the time), and 4 (all f the time). The ratings f the 24 subjects tested in the perceptin cnditin (Experiments I and IA) and the 24 subjects tested in the imagery cnditin (Experiments 1 and 2) suggest that peple usually estimate line length by mving a unit f measurement alng the line. The average ratings fr this strategy were 2.7 in the perceptin cnditin and 1.7 in the imagery cnditin. The average ratings fr using scan time t estimate length were.3 in the perceptin cnditin and.7 in the imagery cnditin. The strategy f estimating length by mving a standard alng a line has als received sme supprt in previus investigatins. Hartley (1981) argued that subjects used this strategy because the time it tk them t estimate length increased linearly with length. Hwever, this fmding is als cnsistent with the scanning time hypthesis. Thrndyke (1981) prpsed that subjects use scanning time t estimate length, but his subjects reprted that they estimated length by determining hw many times a standard wuld fit a line and multiplying by the length f the standard. Althugh Thrndyke argues that these reprts are cnsistent with his timing mdel, it is nt clear why subjects wuld base their estimates n scanning time if they cmputed the answer by superimpsing a unit f measurement. Perhaps subjects superimpse a standard but als have knwledge f hw lng it takes them t cmplete this prcedure. They may then calculate the answer based n this strategy but slightly mdify their answer accrding t hw lng it tk them t calculate it. This time bias wuld have t be task specific, because it did nt ccur in ur experiments. Anther pssibility is that Thrndyke's (1981) results are unrelated t scanning time. Further research, including the measurement f scanning time in his paradigm, is necessary t reslve this issue. Mental Scanning Supprt fr the suggestin that tacit knwledge had very little influence n the utcme f the scanning experiments cmes frm the finding that predicted scanning times are different frm the times prduced when subjects are instructed t scan the cnfiguratins. Althugh peple crrectly predict the effect f length n scanning time, they did nt crrectly predict the effect f cnfiguratin n scanning time. Any tacit knwledge that cnfiguratin influences scan time was instead represented by differences in the y-intercepts f the regressin lines. Subjects crrectly predicted that scanning a maze shuld take lnger than scanning a spiral and scanning a spiral shuld take lnger than scanning a line, but estimates differed by an additive cnstant that was independent f length. In cntrast, when subjects are asked t actually scan patterns, their respnses reveal differences in the rate f scanning acrss cnfiguratins, as measured by significant differences in the slpes f the regressin lines. These slpes did nt

7 TACIT KNOWLEDGE AND MENTAL SCANNING 143 differ significantly when subjects were asked t estimate scanning time. A secndary finding that may be difficult t explain in terms f tacit knwledge is hw pattern cnfiguratin influences the cnsistency f scanning times. The average individual crrelatin between scanning time and length was always less fr the lines than it was fr the spiral and maze (Table 1). It is difficult t fresee hw the tacit knwledge hypthesis culd accunt fr this finding unless peple were als less cnsistent in estimating the lengths f lines than in estimating the lengths f spirals and mazes. This, hwever, did nt ccur. Althugh the tacit knwledge hypthesis appears t be an inadequate explanatin f hw peple scan cmplex cnfiguratins, it is pssible that peple might rely n their tacit knwledge t avid scanning linear distances. Hwever, Finke and Pinker (1982) recently btained evidence that suggests peple mentally scan linear distances even when they were nt instructed t scan a mental image. Their paradigm required subjects t judge whether an arrw at an unexpected lcatin pinted at any dts in a previusly expsed dt pattern. Reactin times were directly prprtinal t the distance separating the dts and the arrws, suggesting that the subjects used mental scanning t make inferences abut spatial relatins. Our argument that the tacit knwledge hypthesis cannt accunt fr ur results supprts the mre general argument made by Finke (1980) that it is unlikely that peple culd anticipate the results f many imagery experiments. Finke cites many specific examples in which he claims that the utcme is nt bvius, usually because peple are ding the task fr the first time and their tacit knwledge is incrrect r incmplete. Cnsistent with that argument, it appears frm the data reprted here that tacit knwledge is inadequate t accunt fr all mental scanning data. REFERENCES FINKE, R. A. Levels f equivalence in imagery and perceptin. Psychlgical Review, 1980,87, FINKE, R. A., & PINKER, S. Spntaneus imagery scanning in mental extraplatin. Jurnal f Experimental Psychlgy: Learning, Memry, and Cgnitin, 1982,8, HARTLEY, A. A. Mental measurement f line length: The rle f the standard. Jurnal f Experimental Psychlgy: Human Perceptin and Perfrmance, 1981, 7, INTONS PETERSON, M. J. Imagery paradigms: Hw vulnerable are they t experimenters' expectatins? Jurnal f Experimental Psychlgy: Human Perceptin and Perfrmance, in press. KERST, S. M., & HOWARD, J. H. Memry psychphysics fr visual length and area. Memry &I Cgnitin, 1978,6, KOSSLYN, S. M., BALL, T. M., & REISER, B. J. Visual images preserve metric spatial infrmatin: Evidence frm studies f image scanning. Jurnal f Experimental Psychlgy: Human Perceptin and Perfrmance, 1978, 1, KOSSLYN, S. M., & POMERANTZ, J. R.lmagery, prpsitins, and the frm f internal representatins. Cgnitive Psychlgy, 1977,9, MITCHELL, D. B., & RICHMAN, C. L. Cnfirmed reservatins: Mental travel. Jurnal f Experimental Psychlgy: Human Perceptin and Perfrmance, 1980,6, MOYER, R. S., BRADLEY, D. R., SRENSEN, M. H., WHITING, J. C., & MANSFIELD, D. P. Psychphysical functins fr perceived and remembered size. Science, 1978,200, PYLYSHYN, Z. W. The imagery debate: Analgue media versus tacit knwledge. Psychlgical Review, 1981,88, THORNDYKE, P. W. Distance estimatin frm cgnitive maps. Cgnitive Psychlgy, 1981,13, NOTES 1. Althugh the results f Experiment 1 and Myer et al. (1978) indicate that length is underestimated in the imagery cnditin, ther fmdings differ. Kerst and Hward (1978) fund that estimated distance was slightly greater fr the imagery cnditin. In additin, ur results frm the imagery cnditin in Experiment 2 were very similar t the length estimates btained frm the perceptin cnditin in Experiment The subjects were tested by an undergraduate wh was unfamiliar with previus wrk n visual imagery and was unaware f the hyptheses that mtivated this research. She did nt see the results f the experiments until the entire prject was cmpleted. We therefre avided the experimenter-bias effects described by Intns-Petersn (in press). (Received fr publicatin June 21, 1982; revisin accepted September 27, 1982.)