* * *
Lets continue from Part 1 (link)
The definitions of an “Edge” according to Lynch are: “Linear elements not considered as paths” (62) “Boundaries between two kinds of areas” (62) “[Edges are] visually prominent, … continuous in form and impenetrable to cross movement” (62)
“Edges, whether of railroads, topography, throughways, or district boundaries, are a very typical feature…and tend to fragment [the environment].” (63) “Edges are often paths as well”. (65)
If we consider edges, which are neither paths nor the boundaries of districts, are there spatial and visual qualities to an edge, which are unique to the concept of an edge and are definable in terms of the spatial descriptors used in space syntax? Consider one of Lynch’s definitions of the edge, that edges are usually “Visually prominent,… continuous in form and impenetrable to cross movement” Let us consider what this might mean in terms of isovist properties. What are the attributes of an isovist, which could be “visually prominent,… continuous in form and impenetrable to cross movement”? Obviously, we are talking about surfaces bounding spaces, and to use Benedikt’s terms (Benedikt, 1979) we are distinguishing between the ‘real surface perimeter’ and those portions of an isovist’s boundary, the ‘occluding radials’, that constitute the rest of the isovist’s perimeter. If we imagine an isovist consisting of a high number of lines of sight radiating from a single point in space (‘isovist radials’), we can ‘unfurl’ the isovist and plot the distribution of its radial lengths. See Figure 10 for an example of three such isovists and the graphs showing the distribution of their radial lengths. An edge, in isovist terms can be defined as a specific property of the distribution of these isovist radials, namely where there occurs a smooth or uniform increase or decrease in the radial lengths. This is the case in the first two examples shown in Figure 10; these two examples illustrate isovists with visually prominent edges constituting a major proportion of the visual boundary. Where the chart demonstrates this characteristic of a regular rate of increase or decrease in radial lengths (the slope of a portion of the graph), then it can be held that there is a flat, occluding surface bounding the space. This is directly equivalent to Benedikt’s ‘real surface perimeter’.
Where there is a sudden ‘jump’ in the distribution of radial lengths in the graph, this indicates the presence of an ‘occluding radial’ in Benedikt’s terms. This reflects Gibson’s definition of a edge, “An occluding edge is usually but not necessarily projected as a… discontinuity in the gradient of binocular disparity (not when vision is with one eye).”9 This characteristic can clearly be seen in the rightmost example of Figure 10. The graph is not smooth and continuous, but is irregular and disjointed. A sudden increase in the radial lengths represents a line of sight that shoots past the corner of one occluding surface, continuing until it terminates at another occluding surface some distance from the first surface (a line in the ‘all line axial map’ and an possibly an E-partition). Since any edge can be defined as the real surface perimeter of an isovist, the question must be asked, what degree or amount of real surface perimeter constituting an isovist must be present for that edge to be perceived as a “visually prominent” boundary. We suggest that if a continuous section of an isovist’s real surface perimeter constitutes a high proportion of an isovist’s perimeter, then this would be perceived to be a visually prominent boundary or an ‘edge’ in Lynchian terminology.
In Lynch’s Boston study, however, there seem to be very few edges, and this indicates that something more than their own visual property might come into play in determining whether visually prominent boundaries act as edges in a cognitive map. As Lynch points out, paths and boundaries of districts may often act as edges. In Lynch’s maps of Boston, the subjects identify only two edges—the harbour to the east and the Charles river to the west and both are areas where the axial map ends abruptly. The only path that is picked-up as an edge is the elevated central artery, and that too only by his trained observers. Interestingly, this is also a situation of a sharp transition in the axial map. What is puzzling is that the water-front in the North End does not appear as an edge. Once again, the axial map provides clues for this phenomenon. There is no strong axial line parallel to the water-front here, as in the other two cases. In other words, the edge in an urban environment seems to depend to a great extent, not only on its own visual (isovist) properties, but where it occurs with respect to the main paths of movement (the structure of the axial map).
7.2 Relationship between intelligibility and hierarchy
This paper examined the relationships between what we have termed first order elements (structurally distinctive) and second order elements (visually distinctive) demonstrating how both sets can be defined by Lynch and through space syntax terminology. In doing so, we increasingly observed that the dependencies connecting them did not suggest a mutual mapping. We conclude that there is a dependency of the Lynchian elements upon the basic space syntax spatial descriptors but no such dependency exists in reverse. This hypothesis can be summed up by the following statement: all imageable cities must be intelligible , but all intelligible cities need not be imageable
The potential for visual differentiation arises and can only arise from an already existing structure and hence hierarchy of use. It is true, as Lynch claims, that a visually differentiated and ordered landscape (an imageable landscape) is characteristic of a functional city (as opposed to a dysfunctional or pathological one), but the visual differentiation can only arise from a well-developed structural hierarchy.