Golledge has provided us with a valuable collection of essays which bring together a variety of researchers on both human and nonhuman wayfinding behavior. This collection promises to be a strong reference po int for future researchers in this area. The following is a critical commentary on the essays in Part II of Golledge's collection: "Perceptual and Cognitive Processing of Environmental Information." The three chapters in Part II investigate, in a variety of contexts and through a variety of approaches, human navigation relative to the environment. I will comment on each chapter separately. I think that any multi-disciplinary project such as this demands clarity more than other projects because of the varied nature of the audience. Hence, one or two of my comments are concerned primarily with the clarity of the exposition.
1. In chapter 5, Loomis et al, discuss the process of human navigation by path integration. Navigation is defined as 'the planning of travel through the environment, updating position and orientation during travel, and, in the event of becoming lost, reorienting and reestablishing travel toward the destination' (p. 125). Path integration is defined as 'the process of navigation by which the traveler's local translations and rotations, whether continuous or discrete, are integrated to provide a current estimate of position and orientation within a larger spatial framework' (p. 129). Path integration, then, involves the integration of sensory (and perhaps some non-sensory) information available to the subject during translations and rotations, but it does not involve prior planning of the route by the subject.
2. The discussion in chapter 5 is well organized and informative about the sorts of studies (animal and human) to date on path integration. It is also informative about the predominant terminology used to discuss path integration, the current models of path integration, and the methods (perceptual and representational) available to subjects for updating their location relative to some point of origin. Moreover, this chapter describes how the subject relies on both allothetic and idiothetic inputs for path integration. Allothetic inputs include sensing optic flow and acoustic flow, and idiothetic inputs include efference copies of muscle commands, afferent proprioception, and vestibular signals from the otoliths and semicircular canals. Chapter 5 is informative about the studies which have attempted to assess the different contributions from each of these input sources to path integration.
3. One difficulty I had with chapter 5 concerned the clarity of the explanation of the evidence for a particular 'encoding error' model. This model is proposed to account for systematic error in the data of an experiment by Loomis et al. (1993). This is a path completion experiment. The subjects are guided along a path which consists of two straight legs. The subjects are either blindfolded, congenitally blind, or adventitio usly blind. Once guided along two legs of a path, the subjects are asked to find their way back to the starting point. None of the subjects performed well on the experiment. 'The average subject turned too little when large turns to the origin were called for and overturned when small turns were called for.' Moreover, on their return trip the subjects overshot short distances and undershot long distances. The encoding error model is proposed to explain these systematic errors.
4. The problem that I have is with the authors' explanation for the following assumption underlying the model: 'The primary assumption of the model is that all of the systematic error in subject's performance is the result of error in sensing the outbound path and representing that information in memory' (p. 139). In other words, it was assumed that 'when subjects attempted to return to the estimated position of the origin, they did so without systematic error. 'All systematic error, then, occurs on the outbound trip and is thus encoded in the subject's memory for the return trip. The authors promise to give a justification for their assumption later, and they do. This assumption is justified by the result of another experiment.
5. The other experiment is on perceptually directed action (Loomis, et al. 1998, experiment 3). In this experiment the subject is allowed to view a visual target in a large open field. Then he or she attempts to walk to the target without additional perceptual input during the traverse. The study found little systematic error due to path integration in these experiments. The subjects performed very well in finding the targets. The authors then state: 'Incidently, it is the absence of systematic error... cited above that motivated the critical assumption of the encoding-error model of no systematic error in executing the return to the origin' (p. 147). The problem is that this is all we are offered as an explanation of the justification of a critical assumption. And it is not immediately obvious why this result justifies the encoding-error model assumption. More explanation would have been helpful here. I do not disagree that there is evidence for the assumption. I just think that evidence for a critical assumption should be made clearer.
6. I have a couple of other points to make about the author's depiction of path integration. The authors' analysis of the data in path completion experiments leads them to suggest that path completion is incompatible with moment-to-moment updating by the subject and that it is compatible with configural updating. 'In moment-to-moment updating, the traveler is continually estimating current position and orientation on the basis of trave l velocity and the prior estimate of position and orientation.' (p. 132) In contrast, '[i]n configural updating, the traveler maintains some more elaborate representation of the traveled path and every so often updates the estimate of current position and orientation based on the prior estimate and the stored representation of the path.' (p. 132) The difference between the two sorts of updating is that in the moment-to moment updating there is no representation kept of the path traveled. Moment-to-moment updaters maintain only a representation of their position and orientation relative to the origin, but the configural updaters maintain a representation of the path traveled, and in virtue of this path-traveled representation are able to maintain another representation of their position and orientation relative to the origin.
7. The problem with the view that path completion is incompatible with moment-to-moment updating. For it would seems that configural updating presupposes some sort of moment-to-moment updating. Since one has no knowledge of the path before being led outward by an experimenter, one must perform some moment-to-moment updating if one is going to be able -- as the configural updater must -- to acquire one's initial 'more elaborate representation of the traveled path.' In other words, one cannot acquire an elaborate representation of the traveled path until one has traveled some of the path. For example, at least along the first leg of the path, one is going to need to monitor one's velocity moment-to-moment if one is going to have the raw materials necessary to form, at a later time, an elaborate representation of one's distance from the origin. Consider also that for the first leg of the outward path any representation of the path traveled is going to be identical to a representation of one's position and orientation relative to the origin. Since a representation of the path traveled cannot be built until the path has been traveled a bit, moment-to-moment updating seems essential for the possibility of configural updating. Even for the second straight leg of the path, it seems that moment-to-moment updating with respect to the starting point of the second leg would have to occur in order for the configural updater to have the raw materials needed to add the next leg to their representation of the path traveled, and so on for each additional leg. Therefore, if it is true that configural updating takes place, then so must moment-to-moment updating. The reverse, of course, does not necessarily follow.
8. The authors of chapter 5 do give evidence against the traveler using moment-to-moment updating. However, their explanations (p. 141) are very short and therefore difficult to follow. In any case, it's hard to see, based on their definitions of configural and moment-to-moment updating, how configural updating could get along without moment-to-moment updating. However, the authors do point out a reason to be a bit suspicious of configural updating. On the one hand they notice that configural updating would be helpful for humans who traverse straight legs of path turning every so often. This would be helpful because it would allow humans to retrace their path if the need arises. On the other hand, they recognize that in the case where humans traverse paths which are continuously varying, configural updating is not likely to be helpful in the same way. Unfortunately, they do not say why this is the case, except that 'such a strategy [of retracing one's path according to one's configurally updated representation of the path - my parenthetical inserted for clarity] cannot succeed in the general case, for it does not seem to apply naturally to paths that are varying continuously' (p. 142). More explanation of this explanation would have been helpful. For, it seems to me that one could continuously update one's representation of one's traveled path and occasionally use this to update one's position and orientation relative to the origin. This would still count as configural updating (by the definition given) as long as what is continually updated is the representation of the path traveled thus far. Nevertheless, there is probably need for more study of what happens when the traveler's path is continuously varying; which is probably the typical kind of path that organisms traverse in nature.
9. In chapter 6, "A Neurocognitive Approach to Human Navigation," Michel-Ange Amorim argues for combining cognitive psychology with neuroscience in order to 'figure out the basis of human wayfinding' behavior and in order to 'improve spatial behavior' (p. 167) Amorim defines the method of cognitive psychology as making 'extensive use of reaction times (Luce 1986) and errors as performance measurements in order to infer elements of mental organization' (p. 152). Unfortunately, Amorim nowhere defines neuroscience or neurocognition. Neuroscience is the attempt to map areas of the brain to all sorts of features of human perception, experience, body parts, cognition and behavior (pain, feet, arms, tongue, spatial organization; see Paul Churchland 1984). Neurocognition is a subdivision of neuroscience, and it maps areas of the brain specifically to cognitive and behavioral functions (neurocognition is the same as neuropsychology, Churchland 1984). However, we cannot be certain of precisely what Amorim takes these terms to mean because the definitions are never given. Definitions of neuroscience and neurocognition would have been very helpful here, given that the proposed approach is to combine neuroscience with cognitive psychology.
10. Nevertheless, Amorim states two goals. First, Amorim is interested in validating a model of visuo-spatial cognition proposed by Kosslyn (1991) and Kosslyn, Van Kleek, & Kirby (1990). This model describes the 'psychological transformations by which an individual acquires, codes, stores, recalls, and decodes information about the relative locations and attributes of phenomena in the everyday spatial environment' (p. 153). This model is supposed to be applicable to spatial processing of all sorts, and Amorim attempts to validate it for the less general task of human navigation by discussing two studies. One study 'investigates the role of reference frames in computing locations in space; the other compares two different processing modes for the updating of an object's location and orientation' (p. 153). In addition to his goal of validating Kosslyn's model for path integration, Amorim is interested in applying the neurocognitive approach to the study of pathological cases of environmental cognition, that is, to 'topographical disorientation.' Topographical disorientation is disori entation in one's path integration capacities.
11. Although I think that Amorim's essay makes an important contribution to the literature, I'm not convinced that the contribution is, in the first instance, a contribution to what we know about neurocognition, as suggested by the title of Amorim's article. It is not very surprising that a neurocognitive approach to human navigation is promising. Consider that Amorim, in his section which introduces the neurocognitive framework (p. 153), concentrates on explaining the particular processes involved in Kosslyn's cognitive psychology model of visuo-spatial cognition. And in one sentence toward the end of this introduction, Amorim adds, '[t]he neural structures underlying each of these processes of high-level vision as well as their associated pathologies are rather well defined (for a review, see Kosslyn & Koenig 1992)' (p. 154). This seems to suggest that we already have a neurocognitive model which is compatible with Kosslyn's visuo-spatial processing model. The rest of this section on the neurocognitive framework provides examples where areas of the brain have been associated with aspects of this model. For example, it is pointed out that studies on primates have shown that cognitive information on objects and their locations are associated with the occipital-temporal pathway and the occipital-posterior parietal pathway respectively (Mishkin, Ungerleider, & Macko 1983).
12. What I think about this article is that the significance of its contribution is obscured by too much discussion about neuroscience. The significance of the article, as far as I can tell, is its validation of Kosslyn's general model of visuo-spatial cognition for the more specific case of human navigation. According to Amorim, Kosslyn's model is already compatible with a lot of what neuroscience tells us about spatial processing. In other words, according to Amorim, it seems as though findings in neuroscience confirm that the relatively general cognitive mechanisms posited by Kosslyn's model have associated neural structures. What we do not know is whether Kosslyn's general model of visuo-spatial cognition can handle the sort of spatial cognition that is needed to facilitate human navigation. What Amorim attempts to show is that Kosslyn's model is able to handle the sort of spatial processing used in human navigation. This is where the contribution of Amorim's article lies. For if we already know that findings in neuroscience confirm that the relatively general cognitive mechanisms posited by Kosslyn's model have an associated neural structures, and we discover that Kosslyn's model can handle human navigation, then it follows that findings in neuroscience will confirm that cognitive mechanisms relevant to human navigation will have associated neural correlates.
13. This is not to suggest that we should not then search for new (and point out old) sources of the confirmation that there are these neural mechanisms underlying the spatial cognitive mechanisms involved in human navigation, and Amorim does this in the last section of his paper on topographical disorientation. The point is just that what opens the way for the promise of a neurocognitive approach to human navigation is threefold: (1) there needs to be a reliable general framework in spatial cognition; (2) this general framework must be compatible with what neuroscience tells us about the neural mechanisms involved in visuo-spatial processing; and (3) the work on the cognitive processes underlying human navigation must fit nicely into this general framework. According to Amorim, we already have 1 and 2. Therefore, what is needed is to show 3. My criticism is that Amorim's contribution is obscured by concentrating, at many points throughout the article, on the details of what is already known about neurocognition; this ought to be secondary to validating Kosslyn's model for the case of human navigation. However, in the article it seems as though validating Kosslyn's model is merely secondary to some point about neurocognition in general.
14. In Chapter 7, John J. Rieser proposes a model for the relationship between action, perception, and representation. According to the model, a link couples perception and action, and a link couples representation and action. Between action and perception the link is a covariance between afferent and efferent motor input (during action) on the one hand, and perceptible changes in perspective. Efferent motor inputs precede and guide action. Examples include higher level information about the entire route, lower level information about the next step, etc. Afferent motor input is information specifying self movement such as 'joint and muscle input signaling limb movements,' vestibular input from the semi-circular canals, otoliths, and optical flow. The link between action and representation is a covariance between action (efferent and afferent motor inputs) and what subjects remembers -- rather than what they perceive -- about their surroundings.
15. Rieser's primary thesis is that 'the coupling of perception and action is the basis for the coupling of representation and action' (p.172). The idea is that through visual and nonvisual information about the surroundings, people learn about these reliable co-variances. What they learn serves as the basis for the coupling of action and representation. And this coupling of action and representation 'accounts for the awareness of changing spatial orientation relative to the surroundings while walking without environmental information' (p.172).
16. Briefly, Rieser sets out three implications of the thesis that there is this dependency relationship between the two sorts of links. Rieser and colleagues design ingenious experiments to test the thesis. And the results of the tests confirm the thesis. All of this is laid out very clearly and I have no comments on these experiments. I am, however, interested in commenting on Rieser's use of the concept of directness when referring to the link between action and representation.
17. Rieser states that '[m]ost research about the perception of self- movement has assumed that self movement is perceived relative to a body-centered frame of reference, whereby people perceive changes over time in their bodily positions, which are not themselves linked to the environment' (p. 176) Rieser states that this implies a two-stage process 'whereby people first perceive the velocities and accelerations from instant to instant, and then integrate those body-centered perceptions with their knowledge of the surrounding environment' (p.176). However, Rieser does not think that a two-stage process is involved. He claims that there are two reasons to think that the process one-staged. First, it would be inefficient for there to be two stages when one stage could do the job. And Second, 'people report perceiving their self-movement directly in environmental terms.'
18. Although Rieser is generally very careful to define all of his terms throughout the article, the terms are not well defined in this discussion about the directness of coupling. First, it is unclear what, precisely, distinguishes two-stage from one-stage processes. It seems to me that it could be a spatial, a temporal, or a methodological distinction of some kind. Second, the reasons Rieser provides for thinking that there is only one process seem untenable. First, it is not clear why Rieser thinks that prima facie differences in efficiency tells us anything about what processes exist de facto. According to evolutionary theory, traits are only as efficient as they need to be for adaptive purposes. I'm not sure if he was thinking of evolutionary theory, but nothing about evolutionary theory, prima facie, go against a two-stage process. Second, it is not clear why we should trust people's first person reports since it is, by now, well known that people are not conscious of all the processes involved in their cognition and behavior. Perhaps their reports of perceiving their self-movement directly in environmental terms are reports about the output of an unconscious two-stage process.
19. More important, however, it is not clear why it matters to Rieser whether the process is two-staged. His thesis seems to be that the representation-action link is dependent on, and shaped by the perception-action link. What does it matter how many stages it takes to implement these links as long as the coupling in both cases respects this dependency relationship?
20. I would like to conclude, about Rieser's article, that I found it to be, on the whole, very well organized and argued. Moreover, it provides a good degree of evidence for an ecological approach to understanding perception and cognition (Gibson 1979). The role of the environment, even in mere memory, according to Rieser's results, plays a very significant role in guiding behavior, and hence can play an important role in our understanding of wayfinding behavior. Incidently, Rieser's experiments suggest that work on the ontology of the environment may also have a significant role to play in understanding wayfinding behavior (The beginnings of such work can be found in Roger Barker 1968, 1978; G. Evelyn Hutchinson 1978; and Barry Smith and Achille C. Varzi 1999).
Barker, Roger. (1978) Habitats, environments, and Human Behavior (San Francisco, Washington, and London: Jossey-Bass Publishers).
Barker, Roger. (1968) Ecological psychology: Concepts and Methods for Studying the Environment of Human Behavior (Stanford, CA: Stanford University Press).
Churchland, Paul. (1984) Matter and Consciousness (Cambridge, MA and London: The MIT Press).
Gibson, J. J. (1979) The Ecological Approach to Visual Perception (Boston: Houghton-Mifflin).
Golledge, R. (1999a) (Ed.) Wayfinding Behavior: Cognitive Mapping and Other Spatial Processes. John Hopkins.
Golledge, R. (1999b). Precis of: Wayfinding Behavior: Cognitive Mapping and Other Spatial Processes. PSYCOLOQUY 10(036) ftp://ftp.princeton.edu/pub/harnad/Psycoloquy/1999.volume.10/ psyc.99.10.036.cognitive-mapping.1.golledge http://www.cogsci.soton.ac.uk/cgi/psyc/newpsy?10.036
Hutchinson, G. Evelyn. (1978) An Introduction to Population Ecology (New Haven and London: Yale University Press): Chapter 5.
Kosslyn, S.M. (1991) "A Cognitive Neuroscience of Visual Cognition: Further Developments," In R.H. Logie & M. Denis (eds.), Mental Images in Human Cognition (Amsterdam: Elsevier Science): pp. 351-381.
Kosslyn, S.M., Koenig, O. (1992) Wet Mind: The New Cognitive Neuroscience. (New York: Free Press).
Kosslyn, S. M., Van Kleek M. H., and Kirby K. N. (1990) "A Neurologically Plausible Model of Individual Differences in Visual Mental Imagery," In P. J Hampson, D. F. Marks, & J. T. E. Richardson (eds.), Imagery: Current Developments (New York: Routeledge): pp.39-77.
Loomis, J. M., Golledge, R. G., Klatzky, R. L. (1998) "Navigation System for the Blind: Auditory Display Modes and Guidance. Presence, 7, pp. 193-203.
Loomis, J. M., Klatzky, R. L., Golledge, R. G., Cicinelli, J. G., Pellegrino, J. W., & Fry, P. A. (1993) "Nonvisual Navigation by Blind and Sighted: Assessment of Path Integration Ability. Journal of Experimental Psychology: General, 122 (1), pp. 73-91.
Luce, R. D. (1986) Response Times: Their Role in Inferring Elementary Mental Organization (Oxford: Oxford University Press).
Mishkin, M., Ungerleider, L. G., & Macko, K. A. (1983) "Object Vision and Spatial Vision: Two Cortical Pathways. Trends in Neuroscience, 6, pp. 414-417.
Smith, Barry and Achille C. Varzi. (1999) "The Niche," Nous, vol. 33, (2), pp. 214-238.