Excerpt from
L.Kleine-Horst: Empiristic theory of visual gestalt perception. Hierarchy and interactions of visual functions. (ETVG), Part 3, II

Correspondence of Levels Hypothesis

As already mentioned, a psychological theory of sight, acceptable to the neurobiologists, must provide a hierarchy of functional visual detectors that corresponds to the hierarchy of neuronal function carriers, found in neurobiology. Such a theory has not been presented until today. Sometimes one reads of a hierarchical structure of information processing, but rarely are more than two levels suggested, distinguishing, for example, between peripheral and central, local and global, or pre-attentive and attentive perception. These pairs of concepts mean something different, but refer also to a two-level aspect of perception. When  several levels are suggested in other theories, it is usually without assigning particular detector functions to them, thus they fail to contribute essentials to an understanding of visual perception.

The ETVG is the only theory that describes a concrete multiple level hierarchy of defined functions, a section of which agrees with a hierarchy of single cell functions, found by neurobiology (Livingstone and Hubel 1987, 1988; Zeki and Shipp 1988), concerning static, two-dimensional figure/outfield perception. In the following, clues on the correspondence between the level functions of ETVG and neurobiology are given.

It is assumed that the functions of the factors Zm/Zl - Pml - Dm/Dl - Gml - Ll and Fl, located at the physical level and the 5 lowest psychical levels, are performed by neurons at 6 lowest neurobiologically defined levels: at receptors - retinal ganglion cells - LGN - V1 - V2 and V4. Table 3-2 shows these assumed functional correspondences of levels.

The chemical substances of the receptors provide the primary tendency to a certain modality Zm (brightness and color); the position of the stimulated receptor within the retina provides the primary tendency to a certain position (Zl), within the visual field, of the stimulus detected. At this physical level, however, experience does not yet occur; all "primary tendencies" are only functional entities that can lead to various phenomena, or locations, according to the influences from the (psychic) gestalt factors above the physical factors.

psychical level

ETVG-factor

correspondence to functions of

5

Fl

V4

4

Ll+    /  Ll-

V2:  interstripes / thin stripes

3

Gml+   /   Gml-

V1:  interblobes / blobs

2

Dm, Dl

LGN

1

Pml

retinal ganglion cells

physical level

Zm, Zl

receptors

Table 3-2. Assumed correspondences between the functions at six ETVG-levels
and the functions at six neurobiologically defined levels.

The lowest level at which a subjective impression of the outer world occurs, is the lowest psychic level with the gestalt factor Pml. The experiences produced by the Pml factor are "brightness at a location" (Pml+) and "darkness at a location" (Pml-). These Pml functions correspond to the functions of the "brightness neuron" and the "darkness neuron", respectively, found by the neurobiologists in the retina. Other retinal ganglion cells are sensitive to colors. The lateral geniculate nucleus (LGN) is known to be capable of detecting brightness and color differences, according to the ETVG factor Dm. Jones and Sillito (1991) reported that almost all cells of the cat LGN are length tuned, i.e. they are able to detect location differences, as does factor Dl.

The ETVG function Gml+ is able to detect the inhomogeneities ("discontinuities") at an object edge; its receptive field is orientation sensitive, however, not to the orientation of the edge, but perpendicular to it. Orientiation sensitive neurons are found in V1 by von der Heydt and Peterhans (1989). The authors, however, reported that "cells in V1 did not signal the orientations of the rows that one perceives, but the orientations of the Fourier components" (p.1743), i.e. the orientation perpendicular to that of the row, line, or edge.  "Illusory" contours, too, are detected by V1 Macaque cells, as Grosof, Shapley and Hawken (1993) reported. Corresponding to this, in Part 9, the detection of "illusory" contours is predicted by the factor Gml, via its receptive field oriented perpendicularly to the contour orientation. Two main kinds of cells are found in area V1, one of them being the "interblob" cells, which have small receptive fields, are orientation sensitive, respond well to achromatic luminance contrast borders, but which show no color opponency. The other kind are the larger "blob" cells, which are either brightness or color selective, but not orientation sensitive. Whereas the interblob cell properties correspond to the "homogeneity" (Gml+) function, the blob cell properties correspond to the "inhomogeneity" (Gml+) function.

As the "line" (Ll+) function receives input from the Gml+ functions, the V2 "interstripe" cells receive input from the interblob cells. These neurons are "regular" line detecting neurons, which detect both "real" and "illusory" contours (von der Heydt, Peterhans, Baumgartner 1984). It will be shown in Part 9 that the line detecting function of the factor Ll is also involved in the detection of "illusory" contours. As the Gml- functions project to the Ll- function, the blob cells of area V1 project to the "thin stripe" cells of area V2, whose properties corrrespond to the ETVG "field" (Ll-) function; they are also color sensitive, but not orientation sensitive.

Closedness of lines is detected by the factor Fl. Fl+ and Fl- form a figure/outfield system, i.e. the factor Fl (strictly speaking, the factor hierarchy with Fl as the top factor) detects a small figure within a large outfield. The receptive field of the factor Fl is assumed to possess the shape of a small circlelike center field, surrounded by a large field that functions antagonistically. Precisely this shape is reported for the receptive fields of V4-cells by Desimone et al. (1995): a small excitatory field and a very large suppressive surround. These authors claim that V4 cells continue to exhibit former stages' selectivities, but are "not specialized to analyse one particular attribute of a visual stimulus" (Abstract). However, "the sensitivity for V4 cells for both form and color differences might be useful for figure/ground separation" (p.448). Since the "figure contour", detected at the Fl-level, is a "formless circle" (Part 6), it is interesting, that Gallant, Braun and Van Essen (1993) found that in V4 "cells selective for non-Cartesian gratings, those that prefer concentric gratings are most common" (Abstract).

At this stage, the reader may once more be aware of the relationships between ETVG and neurobiology: the ETVG has not been developed from neurobiology; its task is not to interpret neurobiological facts. On the contrary, it was developed (Kleine-Horst 1961) by intentionally (almost absolutely) neglecting neuronal facts, and this before modern neurobiology was established (since Hubel and Wiesel 1962, 1965, 1968). Indeed, the opposite is the case: neurobiology - as an empirical discipline - is assigned to confirm, or falsify, the hypotheses stated in the ETVG.

One of the chances of neurobiology consists in an empirical examination of the ETVG predictions. One of the important things is to differentiate the correspondences between ETVG predictions and neuronal facts. As, according to the ETVG, all factors at Levels 1 to 5 cause effects as antagonistically actualized (positive/negative) functions, two different areas are to be expected at every neurobiologically definable level, one of which is responsible for the positive, the other for the negative function. Intrinsic connections can best take care of the antagonism between the positive and negative functions of the same factor: the positive function inhibits the activation of the positive function (of the same factor) in its environment ("lateral inhibition"), which is equivalent to a "lateral facilitation" of this factor`s negative function. The same is valid also for the effect of the negative function on its environment. Here, lateral inhibition found by Hartline (1940) and referred to a limited aspect of visual perception - the Dm-factor, according to the ETVG -, is expected to be capable of being transformed to other gestalt factors. Thus one of the aims of neurobiological research will be to scrutinize, for every factor at Levels 1 to 5, the predicted antagonism, and possibly find further antagonisms between cell functions.

A further opportunity, the ETVG provides for neurobiology, consists in avoiding misleading concepts, that are built on scant knowledge of facts.

Just as neurobiology profits from the ETVG, the ETVG profits from neurobiology. As an example: the form of the receptive Pml field was not derived absolutely independently from the knowledge of the respective neurobiological facts. I asked myself: how can the antagonistically organized center/surround structure of the B- and D-neurons` receptive fields known to me be accounted for within the frame of the ETVG?

After the shape of the factor Gml`s receptive field (Fig. 2-18A) has been derived from the factor`s antagonistic functions, one has to ask, why the receptive fields of the cells found in V1 by Hubel and Wiesel (Hubel 1988) possess quite another shape. They might be higher-order receptive fields, see Chapter IV.

back to contents

home