Click for next page ( 236


The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 235
HUMAN PERFORCE I SSUES l

OCR for page 235

OCR for page 235
INTRODUCTI ON Jo Ann Kinney I believe that I represent for this conference the personification of institutional memory, and I am taking my position seriously by pre- senting a brief review of night vision testing over the past 25 years. In 1961, I was asked to do a review of the literature on night vision testing for the Committee on Vision (Kinney, l962~. There was extensive material to review, since the years during and following World War II produced a large number of different tests of night vision and a great deal of research on their reliability and validity. Fortun- ately, William Berry summarized this work in 1949. However, from his assessment he took a rather dim view of night vision testing, and I quote some of his conclusions (Berry, 1949~: One generalization is unfortunately that tests of night vision are not very reliable. Visual acuity, contrast sensitivity, absolute sensitivity etc. have not been convincingly demonstrated as co-varying. Is night vision ability of any sort sufficiently important to the Armed Forces to warrant night vision testing efforts? Two new tests appeared in the 1950s: the Naval Medical Research Laboratory Night Vision Sensitivity Test, for which I was primarily responsible for the Navy, and the Army Night Seeing Tester, done by the Personnel Research Branch of the Adjutant General's Office, which was actually a test of mesopic acuity. My assessment in 1961 was much more upbeat: that night vision testing rested upon a firmer foundation of knowledge than did the war- time tests; that, one by one, important variables were being identified and understood; and that, in general, to answer one of Dr. Berry's questions, night vision testing was feasible. In 1968, I was again asked to assess the measurement of night vision for a meeting of the Committee on Vision; at that time I could find no organized program of night vision testing. In the armed forces, essen- tially nothing was going on; interestingly, the Navy test was on loan to the Army and the Army test to the National Aeronautics and Space Administration. In order to have something to say, I did some new 237

OCR for page 235
238 research on the correlations among mesopic acuities and scotopic sensitivity, and my review of the literature consisted of that (Kinney, 1968). These data are included in Appendix A. My experience exemplifies night vision testing over the years. Regularly the necessity for night vision testing was disparaged r as each new electronic aid or superior sensing system emerged. And just as regularly it was revitalized as some new combat need surfaced. This cycle was repeated every 5 to 10 years, with the result that very little progress in testing occurred. In the meantime, basic knowledge advanced by leaps and bounds. In 1961, I thought we had learned a lot--if I could have foreseen 1985' In the papers that follow the reader will find not only a wealth of new knowledge but also a very effective and complete summary. Andre Sanders covers the broad topic of visual search, including the major determiners of search effectiveness: structural or display factors and strategic factors. His analysis, which is based upon data obtained primarily from photopic levels of illumination, emphasizes the research needed to apply these results to search at low levels of illu ~ m~nat~on. Chris Johnson provides an overview of peripheral function, organ- ized under topics of detection sensitivity, temporal contrast sensi- tivity, motion sensitivity, and a variety of suprathreshold functions. This summary provides an excellent example of our new knowledge: for example, spatial and temporal contrast sensitivity were virtually un- known 25 years ago. Dr. Johnson also adds a section on the equipment that would be needed to measure these functions and the areas for which additional research is needed. Cynthia Owsley presents a survey of current knowledge on the effects of aging on night vision which includes senile miosis, increased lenti- cular density, elevated dark adaptation functions, loss of acuity, and increased sensitivity to glare. Dr. Owsley notes that much of our know- ledge on aging and vision comes from investigations conducted at pho- topic levels and that there is a great need for better determinations of the losses at low levels of illumination. Andrew Watson illustrates the advantages of modeling for the study of vision with a number of examples, such as temporal sensitivity, con- trast sensitivity, and motion sensing. He suggests that modeling might also be used effectively in the study of night vision and includes a discussion of the equipment that would be needed to attempt this. Throughout these papers, one is impressed by both the amount of new knowledge of vision and the recurring themes of gaps in this knowledge. Particularly evident among the latter, in all of the summaries, is the sparsity of information on the size of individual differences in the various night vision functions and the fact that we know, in each area, so much more about photopic vision than scotopic vision. The challenges will now be to effectively utilize the knowledge to test night vision, to determine the areas crucial to further understanding, and to pursue these goals despite cyclical variation in the apparent need for night vision testing.

OCR for page 235
239 REFERENCE S Ber ry, W. 1949 Review of Wartime Studies of Dark Adaptation, Night Vision Tests, and Related Topics. Armed Forces, National Research Council Vision Committee. Washington, D.C.: National Academy of Sc fences . K inney, J .A. S . 1962 Review of literature on night vision testing. Pp. 3-11 in M.A. Whitcomb, ea., Visual Problems of the Armed Forces. l Washington, D.C.: National Research Council. 1968 Clinical measurement of night vision. Pp. 139-152 in M.A. Whitcomb, ea., Visual Problems of the Armed Forces. - tiashington, D .C .: National Research Council.

OCR for page 235
VI SUAL SEARCH IN VIG I LANT PERK ORMANCE A.F. Sanders In a recent summary of the literature Monk (1984) has correctly noted that the term visual search appears as a loose label for a variety of phenomena and experimental paradigms that certainly do not share a common denominator. In this summary I will follow Monk's sug- gestion to limit the discussion to situations that are characterized by "spatial uncertainty reduction and target uncertainty to a greater or lesser extent" (Monk, 1984, p. 294~. The first property is self- evident. The second property is meant to imply that subjects may either know or not know whether a target is present during a certain time period. In the case of a brief time period--usually defined as the duration of a discrete trial--certainty about the presence of a target leads to an experimental paradigm in which a subject searches for the target until it is actually found. heisser's (1963) classical studies on target search in a letter matrix are prototypical for this paradigm. When subjects are uncertain about the presence of a target, the experimental paradigm becomes one of target search in tachistosco- pic recognition (e.g., Rabbitt, 1967), with probability of detection or detection time as a measure. Rabbitt's studies on ignoring irrelevant information in target search and later work by Shiffrin and Gardner (1972), Shiffrin (1975), and Shiffrin and Schneider (1977) on visual search for predefined targets with a display size between one and four elements are again among the prototypical studies. Alternatively, if the task does not consist of discrete trials but is open-ended and continues for a longer period of time, there is con- tinuing search, resulting in successes and failures of target detection on those occasions that a target is visible. The latter situation con- tains the characteristics of the traditional vigilance situation, as first studied by Mackworth (1950~. Yet it is not limited to situations where targets are relatively rare. For instance, when driving a car there is a continuous search for relevant information which may vary from a monotonous infrequent target situation on an empty motorway to the overload of information in busy city traffic (e.g., Naetanen and Summula, 1975~. In the last case there are almost always targets, and the subjects are faced with the problem of deciding which targets have the priorities for action. It could be argued that the latter situa- tion does not require visual search in the proper sense of the word. Yet, when faced with several targets at once (e.g., Yntema and 240

OCR for page 235
241 TABLE 1 Major Distinctions of Categories of Visual Search Tests Task type Signal Presence Discrete Trial Openended Trial Certain Searching lists for a Keeping track of several target (Neisser, 1963) several things at once (Yntema, 1962) Uncertain Target detection in Vigilant search for limited search infrequent targets (Rabbitt, 1967) (Mackworth, 1950) Schulm.an, 1967) there can be a visual search problem in that targets at certain positions may be favored in comparison with targets at other positions. Apart from this subdivision (see Table 1) there are various other major distinctions between categories of visual search tests. One important distinction concerns the continuous versus discrete character of the search display. Prototypical examples of the first case are car driving and sonar display inspection, which are characterized by the fact that targets may be found anywhere (e.~., Baker, 19581. An example of the second case is the discrete display inspection, as exemplified in the work of Senders (1984~. A second distinction concerns the size of the display field, where possibly relevant signals may appear. Although the usual visual search studies have dealt with fairly limited displays--as applied to sonar (Vallerie and Link, 1968) or to aerial maps (e.g., Enoch, 1959--there are various other conditions where the search can extend beyond the eye field (Sanders, 19701; this includes head and body movements. A third distinction concerns the extent to which a display is filled with confusing nontargets or, in other words, the extent of clutter. Usually the effects of clutter are more impor- tant in continuous than in discrete displays but they need not be absent in discrete displays, in particular not as the number of alternative signal sources is larger. A final distinction is not so much concerned with the type of search but rather what kind of search modes are per- mitted in the actual experiment. On the one extreme-there are strict instructions to fixate the eye in an attempt to construct the lobe, a probability contour concerned with target detection under conditions of spatial uncertainty at various eccentricities. On the other extreme there are free search conditions usually characterized by analysis of eye movements. The first type of experiment starts from the reasonable assumption that visual search is guided by the limiting conditions of the visual system, and in particular by the properties of peripheral vision. The common f inding that intake of information is limited to f ixations and that there is no useful vision during saccadic eye move- ments--e.~. , Matin (1974--is at the basis of visual studies with the f ixated eye.

OCR for page 235
242 The studies with fixated eyes have usually centered around issues of structural constraints, while the free search studies have been predom- inantly concerned with questions about visual search strategies. Thus, there is either a relative emphasis on bottom-up type of constraints or on top-down type of analysis. It is interesting to note this difference in emphasis when comparing the proceedings of a symposium on search and the human observer (Clare and Sinclair, 1979) with Rabbitt's (1984) re- cent account (Parasuraman and Davies, 1984~. It may be clear from this attempt toward classification of issues on visual search that this paper would be faced with an impossible task when aiming at a detailed review of the results on each cross-section and discussion. I will limit myself , therefore, to a marginal discus- sion of the issues that are treated in more detail in various recent reviews (e.g., Monk, 1984; Rabbitt, 1984; Wiener, 1984; Megaw and Bellamy, 1979; Bouma, 1978~. In addition I will elaborate some issues that might be relatively neglected in these papers. I will follow a fairly arbitrary distinction between structural-display-related and strategic-related factors in visual search, although a strict separa- tion is hard to maintain. Attentional and cognitive factors can be shown to play an important role in situations that, at first sight, may seem to be largely determined by stimulus. Alternatively, visual search strategies are, of course, never free but always subject to cer- tain display-determined constraints. I will conclude the paper with a brief summary of what I consider to be some relevant future research issues. STRUCTURAL AND DI SPLAY FACTORS A structural analysis of visual search starts with the observation that in all visual scanning the eyes are steady for relatively brief periods of time--200 to 400 ms, perhaps extending to 800 ms under ex- treme conditions--after which there is a rapid shift to a new position. Oculomotor factors appear to require a minimum 200 ms to stop and start (Salthouse and Ellis, 1980), while processing demands constitute a second determining factor. Both components may operate in parallel so that effects of processing demands are only found when the minimal dura- tion is exceeded (Vaughan and Graefe, 19771. In support of this view Sanders and Reitsma (1982) found that stimulus processing starts imme- diately upon fixation--even in the case of a combined eye-head shift where the initial part of the fixation consists of a compensatory eye movement. The most relevant question with regard to visual search is related to the determination of the next fixation during the preceding one. The excellent accuracy in aiming at the next fixation, together with the limited saccadic movement times--about 100 ms for a 40-degree movement (e.g., Sanders, 1963--render a closed-loop explanation of saccadic eye movements quite unlikely. Hence, the new aiming point is supposed to be preprogrammed during the previous fixation, or in other words, the "where to look" of the next fixation is decided during the previous one. Consequently, the properties of peripheral viewing have

OCR for page 235
243 ~ m sti mulus response stimulus response FIGURE 1 Measurement of the inspection time of the let t signal, the saccadic movement time, and the inspection of the right signal. received a good deal of interest in the visual search literature. This starts with visual acuity prof lies of the common Landolt ring type in conditions without either temporal or spatial uncertainty, which is then extended to conditions with spatial uncertainty (e.g., Michon and Kirk, 1962a; Corbin et al., 1958~. Recent research in this direction concerns covert orientation of attention, which has shown convincing evidence that visual attention can be shifted to a position other than the line of sight. Reaction times are shorter and the probability of detection is improved when a signal occurs at an expected as compared with an unexpected position (Posner, 1980~. Some studies (Poaner et al., 1980; Shulman et al., 1979) suggest that covert orientation can be conceived of as an internal spotlight moving in an analog fashion across the visual field. Presumably there are relations between the intake of peripheral information and the determination of the next saccade. Evidence about peripheral acquisition of information comes from studies like those of Edwards and Goolkasian (1974 ~ and Antes and Edwards (1973 ~ which sug- gest that information load in the per iphery is the most important factor in determining visual performance. Another example concerns work on the functional visual field (Sanders, 1963, 19;0) in which a nonlinear relation was observed between performance and the display angle at which a visual task is carried out. In a typical experiment two signals are presented at equal distances to the left and the right of the subject's meridian. At the start of a trial the left signal is always fixated, followed by a shift to the right signal and a same or a different reaction. This setup enables separate measurement of the inspection time of the left signal (tl), the saccadic movement time (tm), and the inspection of the right signal (tr) (see Figure 1~. In a number of studies (Sanders, 1963; Sanders and Reitsma, 1982; Houtmans and Sanders, 1984), it was consistently found that tr was considerably shorter when signals constitute an eye field rather than a head field display. In the eye f ield an eye movement is suff icient to cover the angle between the left and the right signal, while a supplementary head movement is needed in the head f ield. The interpretation of the

OCR for page 235
244 reduced tr in the eye field--which was recently confirmed in a number of additional studies (e.g., Sanders and Houtmans, 1984)--is that, while fixating the left signal, subjects acquire a hypothesis about the right signal, as long as the visual angle does not exceed the eye f ield. This hypothesis is checked during the subsequent fixation of the right signal--an activity that takes less time than when a full new percept needs to be formed, which is supposed to occur in the head field. The implication of these results is that as long as search is lim- ited to the eye field, more or less pronounced hypotheses are obtained about all signals that are present in the eye field. me hypotheses would allow direct shifts of the eye to the most relevant signal for additional close inspection. In the head field the processing mode undergoes a basic change since no parallel hypotheses about all pre- sent signals are obtained. In line with the work on covert attention, Houtmans and Sanders (1984) found evidence that the acquisition of peripheral hypotheses does not run off automatically but involves controlled processing. Yet, it should be fully clear that these considerations are at best a small part of the visual search story. One of the main limitations is that the work discussed so far is concerned with a largely empty visual field, while, as mentioned above, more structured visual dis- plays are more common. In a review Bouma (1978) has discussed several structural constraints of more cluttered visual fields, including the effects of lateral inhibition of surrounding items. Although the effect of lateral inhibition has been known for a long time, systematic inves- tigations have not been carried out before the 1970s. Figure 2 shows the pronounced effects of lateral inhibition on visual performance in the periphery. At present there is much more detailed knowledge about various parameters affecting the size of the effect, including angular distances between target and noise letters, the extent of eccentricity, the number of noise letters, the right versus left visual field, and shape differences between target and noise. Engel (1977) has put for- ward the idea that subjects carry out about random saccadic shifts until the target is in the area where it can be peripherally detected. Subsequently, a rapidly directed accede brings the target into foveal vision and, hence, to detection. This was tested by presenting sub- jects with a background of identical disks, except for two disks which deviated in size. First the lobe--the area of display around the cen- ter of fixation, within which a target can be detected with some proba- bil~ty--was determined for each individual subject by tachistoscopic recognition, "hereafter they searched for targets in a search study during which saccades were recorded. The functions did not fully coin- cide, but they still had a fair degree of common variance, to suggest that this type of model has promise. It is doubtful whether the pre- detection search is really random or continuously tests the most likely hypothesis available (e.g., Bloomfield, 1972, 1975; see also Cohen's model, 1981~. In addition to the factors mentioned above that affect lateral interference, the size of the lobe is affected by a range of display variables. Display density; display size; the number of nontargets (e.s., Drury and Clement, 1978~; the degree of homogeneity of

OCR for page 235
JO ~8 245 ~j ~ ~T ~1 I' ~ /a/ 0 O 2 ~ cocontrldty (degrees) \,0` _I 6 8 10 12 FIGURE 2 The pronounced effects of lateral inhibition on visual perfor- mance in the per iphery. nontargets; shape, size, and color; and the regularity of the display (Bloo~r.f ield, 1972) are factors that are discussed in more detail by Monk (1984). Color effects have received the most interest (Green and Anderson, 1956; Von Wright, 1970; Noble and Sanders, 1981) and turn out to be the most powerful cue in visual search. It should be noted, though, that to be effective, the targets should always have the same color. If targets have another color, their detection is either impoverished or the color cue becomes ineffective as a means of Sating information (Posner, 1964; Noble and Sanders, 1980). The properties of the nontargets are quite relevant in determining the efficiency of visual search. The effect of physical similarity between nontargets and targets was convincingly shown in the classical studies of Neisser (1963). Effects of clustering of nontargets have been found by Banks and Prinzmetal (1976), while Rabbitt (1967) was the first to demonstrate the effect of the constancy of nontarget patterns. Subjects do not only learn how to search for targets they also learn how to ignore irrelevant information, and this is most easily achieved when the nontarget items are characterized by constancy in spatial loca- tion and content in relation to the targets (see also Prinz, 1979). In other words targets, as defined at some trials, should preferably not appear as nontargets at later trials and vice versa. This principle is also at the basis of the work of Shiffrin and Schneider (1977) on auto- matic and controlled processing. In the case of consistent mapping of targets and nontargets, automatic detection responses to targets develop in a parallel processing mode of all items in the display set. Alter- natively, in the case of variable mapping of targets and nontargets, Schneider and Shiffrin (1977) found evidence for a sequential controlled search through the items of the display set.

OCR for page 235
292 a particular location, spatial frequency, orientation, and spatial phase. These samples are then examined by an ideal observer to deter- mine whether a stimulus was present, or which of several stimuli was present. The sensors incorporate three of the aspects of visual sensi- tivity noted above: spatial contrast sensitivity, spatial inhomogene- ity, and spatial channels. First, the channel aspect is implicit in the spatial tuning of each sensor, which responds to a one octave band of two-dimensional spatial frequency. Figure 4 shows at the top the receptive field of a sensor tuned to a particular spatial frequency and orientation; below this is shown its two-dimensional frequency spectrum. Second, the sensors scale in size with distance from the fovea, which gives rise to the proper variation in contrast sensitivity with eccentricity. Finally, the gain of each sensor is set by an overall contrast sensitivity function that is unique to each observer. With this simple structure it has been found that many of the outstanding features of spatial sensitivity and discrimination can be explained. There are also some cases where it fails abysmally, but these are of great value, because they can provide guidance toward new and better models. Generally, the failures are for fairly complex dis- criminations rather than for predictions of sensitivity (Nielsen et al., 1985~. Lately work has been done on making the model more general, com- pact, portable, and easy to use (Watson, 1985~. Motion Model Another prominent item on the list is motion. Motion sensitivity involves both spatial and temporal dimensions, hence the motion model used has borrowed heavily from the preceding models of temporal and spatial sensitivity. In fact, each of the individual motion sensors that make up the model is constructed from the building blocks provided by those models. The spatial tuning of each sensor is given by the spatial sensor, and the temporal tuning is provided by the linear fil- ter of the temporal model. A few components are added to endow the sensor with direction selectivity, and a second processing stage is introduced to compute actual image velocities from the sensor outputs. The result is a model that can both predict whether the observer will detect a moving target, and which way it appears to move (Watson and Ahumada, 19851. The model accepts any visual input that can be repre- sented as a sequence of digitized images. Thus one can present iden- tical stimuli to human and model observers. AN APPLICATION: SYMBOL DESIGE Although I have only briefly discussed these examples, my purpose is to convey the spirit rather than the details of these models. How- ever, I will present here at least one example of how such models might be applied. Andrew Fitzhugh and I are now engaged in one project, which we call SYMBOL, that is an attempt to create a system for automatic

OCR for page 235
FIGURE 4 Receptive f ield weighting function of a spatial sensor and its f requency spectrum. evaluation of the legibility of alphabetic fonts, or~ more generally, the discriminability of arbitrary sets of visual symbols. The basis of the system consists of a simplif fed version of the spatial model. The out- put can be either a prediction for the target application, for exan~ple, determination of the probabil ity that these two symbols will be confused when presented at this viewing distance, or a simpler distance measure, which relates to the perceptual distance between any pair of symbols. The project has only just begun, so I cannot tell you whether it will succeed, but the type of ca1 culations that are performed can be des- cr ibed .

OCR for page 235
294 At the top of Figure 5 are shown three characters. The next line shows the characters after they have been passed through a spatial fil- ter that mimics human contrast sensitivity. The third line shows the autocorrelations, and the last line shows the cross-correlations between each pair. These are used in the computation of a perceptual distance between any two characters. Figure 6 shows a distance matrix for an entire font of 26 uppercase letters. Each cell depicts the distance measure between the two letters indicated by row and column. The key runs from zero or very small distances, such as between a letter and itself (along the diagonal), to very large distances. Figure 6 shows a very small distance, or highly confusable pair, which is in fact the I-T shown in Figure 5. Finally, the histogram of interpair distance can be compared for complete fonts. Figure 7 shows examples for three fonts, and it is clear that one font generates generally larger perceptual distances, and hence higher legibility. This illustrates one example of how it is hoped that models of human visual sensitivity will be used to optimize the design of visual displays. A NOTE ON TOOLS One of the purposes of the papers presented in this volume is to give suggestions for the design of a night vision laboratory. Since I am promoting the notion of model and theory as an integral part of such a laboratory, the question arises as to what tools will best serve the vision theorist. First, since so much of visual theory can and should be cast in the language of digital image processing, the scientist should be provided with a personal workstation capable of processing and displaying images in color at high resolution. The workstation (perhaps it should be called a theory station) should be served by massive amounts of f ile space (on the order of 0.1-1 gigabyte/user). The workstation should be capable of very fast floating point operations and should, if possible, be equipped with a fast (15 millions of floating point operations per second) array processor. It would be helpful to have a data link to a supercomputer, such as a GRAY, for models that are beyond the computa- tional capacity of the local workstation. Also useful would be a video camera and framegrabber for digitizing images to serve as model input. An ideal would be a workstation that is also capable of conducting psychophysical experiments, that is, of generating displays and collecting responses. As important as the hardware, however, is the need for appropriate software. This should include extensive libraries of mathematical sub- routines, image-processing routines, and statistical procedures. Also essential are graphics routines for the interactive study of complex data structures. As is well known, all of this is hard to find in a single package.

OCR for page 235
295 s U2 s I: U2 ~ o 3 ~ - O : S O eQ ~ ~ a) 3 G. O S al O o S ED X Z .,, In S X ~ O L. - S U] 3 O ~ S O In o 3 O :D O S ED ~ - U] H 3 _ o S A) U] 3 O ~ O - o - ~ - U] S - ED no, so . I: - : ~ O - U] - U) ~ >~ -1 AS U] ~ U] 0 3 o S U] 3 3 C: U] H Jo U] A' 3 a) Q - ~ O S - ~ - In rl _ C~ O _ o :^ - - - U] o o . - Q~ o O - O - U] o 1 ~n U] o S~ U) o _1 S" o 1 U] U] o 3 O

OCR for page 235
A JO 3 Q o ~5 3 o ~4 U] :3 to a, ~ O U] O ~ -

OCR for page 235
297 . no_. - . ~ . . ~ :{ .~. ..... . FIGURE 7 Histograms of letter pair distances for the three different fonts illustrated in enlarged form. The rightmost histogram should correspond to the most legible font. It is also hoped that vision models will be developed that adhere to certain standards that enhance portability and ease of use. This will promote the sharing of models among scientists, which will greatly increase the productivity, as well as the reliability, of individual models. Perhaps most critical is the need for someone, other than the sci- entist, to purchase, integrate, and perhaps program the theory station. Scientists are curious by nature, and there are few things more arous- ing of curiosity than a fancy computer, and just as curiosity killed the cat, so may the challenge of building the theory station consume the energy and time of the vision scientist. REFERENCES Nielsen, K.R.K., A.B. Watson, and A.J. Ahumada, Jr. 1985 Application of a computable model of human spatial vision to phase discrimination. Journal of the Optical Society of America A 2:1600-1606. Robson, J. 1966 Spatial and temporal contrast sensitivity functions of the visual system. Journal of the Optical Society of America 56:1141-1142. Watson, AeBe 1983 Detection and recognition of simple spatial fortes. In O.J. Braddick and A.C. Sleigh, eds., Physical and Biolog ical Processing of Images. New York: Springer-Verlag.

OCR for page 235
298 1985 Image transforms for visual modeling. Investigative Ophthal malogy and Visual Science 26~3) (Suppl.) :83. 1986a Window of visibility: Psychophysical theory of fidelity in Journal of the Opt ical - time-sampled visual motion displays. Journal of the Optical Society of America A 3 (3) :300-307. 1986b Temporal sensitivity. In R. Boff, L. Kaufman, and J. myomas, eds., Handbook of Perception and Human Performance. New York: John Wiley & Sons. Watson, A. B., and A .J. Ahumada, Jr. 1985 Model of human visual-motion sensing. Journal of Optical Society of America A 2: 322-342. Watson, A.B., and L. Miller 1986 Tests of a working model of temporal sensitivity. Unpublished pape r .

OCR for page 235
GENERAL D I SCUSSI ON KINNEY: I'm going to take the chairman's prerogative and ask the first question. Before I do that, I want to give you a little back- ground for my question. Some of you may not be aware that there were dozens of tests of night vision developed during World War II. Refer- ring back to my introductory remarks concerning my reviews of the early literature on night vision testing, two points stand out in my mind. For my first review, there was a wealth of material to evaluate from the wartime studies; however, the tests turned out to be not very reli- able, the validity scores were poor, and the many tests did not even correlate with one another. The second strong impression is of the lack of new data that I found to review a few years later. This epitomizes what has been going on in night vision testing ever since: a flurry of interest followed by a period of complete inactivity. The result is that, while there has been a 9 reat deal of basic research on night vision, there has been very little new in the applied aspects of testing. So my question, f inally, to the panelists is this: What factors f rom the basic work would you now include in a test of night vision--factors that you believe would improve on our past performance in the testing of night vision? SANDERS: Spatial and temporal uncertainty--with the traditional psychophysical methods, but without the subject's knowing where and when the signal will appear. The uncertainty factor is a very essen- tial aspect of functioning at night. I realize that this is a depar- ture f rom testing sensory functions ~ n its purest sense, but I think you approach more reality. The more you extend the tests with essen- tially nonsensory functions, the more you get to add effects of cogni- tion. The beauty of pu re sensory systems--most of you who are in sensory psychology or sensory physiology will recognize this-- is that you are working with a well-clef ined anatomical and physiolog ical substrate. I t is very tempting, therefore, to study sensory functions in i solation. In the 1960s, I came across a metatheoretical psychology book about the relationship between real life and testing of performance. The author defended laboratory research by saying, "Suppose you want to test visual ecu ity, and you ask 'Does the result of the Landolt r ing test really transform to real life? ' Where are we interested in visual acuity in 299

OCR for page 235
300 real life?" Well, perhaps in the dark we are interested that we don't hit a lantern pole when we are walking on the street. No one will then defend a so-called Lantern pole test" where you bring people into a f ield study in the street and register the probability of them hitting a lantern pole while making a stroll. No, we'll go in the laboratory, test the visual acuity, and if needed, prescribe glasses. You presume that that will lower the probability of hitting a lantern pole while walking on the street, although that has not been proven. In the 1960s, we thought we could develop similar simple and small paradigms that would extend to real life as measures of higher cognitive processes and skilled performances. This belief has been considerably shaken in the 1970s and the 1980s. m erefore, the better you simulate reality, the better you are. If you say well, in night vision functioning in reality there is always spatial and temporal uncertainty--so we'll introduce spatial and temporal uncertainty. There is considerable probability that if you don't, your measures do not predict real-life functioning as well as you might expect. MASSOF: I think it might be premature to even ask that question. It's not clear to me, even after 2 days of listening to the talks, what the problem is that needs to be solved. I think it would be important first to determine what you want to do with the data before you collect it. What is it that needs to be known? What are you screening for? Obviously screening for night vision pathology is well worked out, it's easy to do, and night vision disorders are sufficiently rare that the discovery of undiagnosed pathology is not going to happen that often. But In terms of discriminating among people who are Normal and find- ing the "supernormals," I think that what you're going to have to focus on is not sensory equipment but, as Dr. Sanders pointed out, the task. What are we asking them to do? Cognitive factors may be a lot more important than the sensory factors. JOHNSON: I think there are a couple of things that have emerged since 1961 which are worth looking at. One of them is oculomotor func- tion under low illumination. Even though it was reported in the 1800s, the bulk of research on the dark focus of accommodation and dark conver- gence has been obtained within the past 20 years. Contrast sensitivity is another new approach that has emerged since 1961. Neither of these topics represents a panacea for night vision problems, but they are certainly new areas that are worthy of further investigation. With regard to Dr. Sandersl comment about spatial uncertainty' I would like to mention that several clinical automated devices for evaluating peripheral visual function (automated perimeters) are now designed to perform tests with a high degree of spatial uncertainty. They minimize the likelihood that people will be able to predict where or when a target will appear in the peripheral visual field by random- izing target presentation. m is is done for two reasons: (1) to re- duce the number of spurious responses, and (2) to minimize the occur- rence of eye movements. Spatial uncertainty in terms of automated visual field testing has been developed and is used quite widely. By implementing modest alterations to these test procedures, appropriate night visual search tasks or related tests incorporating spatial un- certainty could be readily accomplished.

OCR for page 235
301 OWSLEY: I just want to emphasize a point that Bob Massof made, and then add something. I think that we would probably be more helpful as basic researchers in coming up with this battery if we understoca better the task requirements of the military. Not only from a standpoint of the cognitive factors but also it would be helpful to know a lot more about the visual requirements of these tasks along the lines that Ralph Haber was talking about. TREDICI: In your experience, Chris [Johnson], do you think the sensitivity of the size of the central 30-degree field is different enough in individuals to be worth studying that in normals? JOHNSON: There's considerable individual variation in the normal population, especially for the over-60 age groups. But even in the younger population there are large individual differences. I certainly feel that these individual differences in the normal population should be investigated, especially with regard to task performance and the effects of practice and training. MASSOF: About the idea of individual variability of test perfor- mance--I think from classic high-threshold theory there's a tendency to assume that these individual differences can be attributed to individual differences in the sensory apparatus, and from detection theory princi- ples we would argue that a large source of the individual difference could be simply due to performance variables. That is, they can be attributed to criterion shifts or criterion differences among observers.

OCR for page 235