get access to the configurational prey-category by weighting p and c (Figure 2.8A) according to the equations shown in Table 2.5.
Table 2.5 Constants α and β determining essential traits of the features (p,c)-relating-algorithm; k1 and k2 depending on other stimulus parameters and the toad’s prey-catching motivation. RB = prey-catching orienting activity [responses/30s]. (Ewert 1984).
Variations in other stimulus parameters, e.g., movement direction (Figure 2.4), velocity, motion pattern, or background pattern (cf. Movie A2) influence a toad’s general prey-catching activity, but the basic effects of the configurational features p and c (Figure 2.8A, charts p,c) are invariant to those changes (Burghagen & Ewert 1982, 1983).
Developmental studies showed that the features-relating-algorithm emerges—without prey-catching experience—after metamorphosis with transition from aquatic to terrestrial life (Traud 1983; cit. Ewert 1984). The principle is common to terrestrial anurans but shows species specificities (Burghagen 1979; Ewert & Burghagen 1969; cit. Ewert 1984).
What does the eye tell its brain?
In leopard frogs, Rana pipiens, Barlow (1953) and Lettvin et al. (1959) recorded spike activities of retinal ganglion cells—from axon terminals in the optic tectum—toward objects traversing the center of a cell’s visual receptive field, RF (Figure 2.9A). The four classes of retinal ganglion cells differ in sensitivities to dimming/brightening, and to an object’s contrast, velocity, and size. The latter is correlated with the diameter of the excitatory RF ranging among different cell classes from 2 to 16 degrees visual angle.
Figure 2.9 Neuroimaging toad’s visual system. (A) Dorsal view of toad’s brain. A stimulus (S) traverses the receptive field (RF) of a retinal ganglion cell (G) of the right eye (E), whose optic nerve (ON) projects to left optic tectum (T) and pretectal thalamus (TH). R, receptor cells; DT, dorsal tectal lobe; VT, ventral tectal lobe; MP, telencephalic ventro-medial pallium; M, medulla oblongata. (B) Functional neuroimaging: 14C-2DG-uptake in brain transverse sections at levels a-d. (see also Suggested Reading, Movie A1). (a) After hand-conditioning of a right-eyed toad, left MP showed 14C-2DG-uptake toward the conditioned stimulus (Finkenstädt & Ewert 1988). (b) Toad escaping a predator stimulus showed strong 14C-2DG-uptake in DT and TH. (c) Toad stiffening toward a threat-like moving stripe presented to the right eye showed moderate 14C-2DG-uptake in left TH and less so in DT. (d) Toad binocularly snapping toward a prey-like stripe showed strong 14C-2DG-uptake bilaterally in VT (Finkenstädt et al. 1985.).
The pioneering work by Barlow and Lettvin and coworkers made us realize that retinal ganglion cells—the output neurons of the retinal network—perform a first-stage analysis of visual input. Subsequent quantitative investigations, however, showed that retinal processing is not sufficient to recognize prey, as suggested formerly. Applying the configurational stimulus paradigm, no correlation was found between retinal neuronal and prey-catching activities in common toads (Ewert & Hock 1972; cit. Ewert 1984). Consequently, configurational feature analysis requires further processing by retina-recipient neurons in the brain.
In search of brain structures involved in feature detection
The visual field of toad’s retina is mapped—via retinal ganglion cell axons along the optic nerve—inter alia mainly in the contralateral optic tectum and pretectal thalamus (Figure 2.9A ). If, in the absence of retinal input, a locus in the tectum of a free-moving toad was excited by trains of electrical impulses delivered by an implanted electrode, the toad responded with orienting or snapping. Probably, the electrical stimulus excited neurons mediating information on prey recognition (Ewert 1974, 1984). In the pretectal thalamus, focal electrical stimulation elicited avoidance, such as ducking or jumping or freezing associated with secretion of skin poison glands—all-up behaviors known to be released by airborne or ground predators.
A neuroimaging technique allows one to check the regional neural activities in response to prey or predator stimuli (Finkenstädt et al. 1985). If 14C-labeled 2-deoxy-D-glucose, 14C-2DG, was administered systemically to the toad, active neurons were confusing the 2-deoxy-D-glucose with glucose, hence taking it up, but failing in decomposing it like glucose. The more active neurons were, the greater the storage of 14C-2DG and thus the radioactivity measured in brain sections later on (Figure 2.9B; see also Suggested Reading, Movie A1).
Figure 2.9 Bd shows a color-coded autoradiographic image of a transverse section through the midbrain of a toad snapping toward a prey-like stripe moving in the binocular field. Strong radioactivity was focused bilaterally on the ventrolateral tectum: “snapping-evoking areas” (Figure 2.9 Bd, VT). In a toad escaping from a moving large square, the overall radioactivity was high, strongest in tectal and pretectal/thalamic structures (Figure 2.9 Bb, DT and TH). This substantiates Tinbergen’s prediction that “neural orchestration” of the whole brain may participate in a stimulus-response. In a toad becoming stiffened toward a threat-like stripe moving in the right visual field, moderate radioactivity in the corresponding left pretectal thalamus was stronger than in the optic tectum (Figure 2.9 Bc).
Configurational object perception involves parallel processing streams and their interaction
Optic tectum and pretectal thalamus are involved in prey catching and predator avoidance. At the neuronal level of analysis, extracellular recordings from toad’s optic tectum reveal monocular T5-type neurons. One type T5.1 is sensitive to extension of feature p. A second type T5.2 is selective in that its activity—according to prey-catching activity (Table 2.5)—increases with extension of p (Figure 2.8 Bp), within limits, but progressively decreases with extension of c (Figure 2.8 Bc) The activity of type T5.2 in response to different configurational objects reflects the probability that an object fits the prey category (Figure 2.8, cf. A, B).
Recordings from pretectal thalamus reveal various types of TH-type neurons, among them monocular neurons TH3 responsive to extension of c or p and c. All these neurons are integrated in a feature-analyzing network.
The “window hypothesis” (Figure 2.10A–C) suggests that feature p is analyzed in a retinotectal processing stream originating in certain classes of retinal ganglion cells (classes R2 and R3) and continuing in tectal prey-selective T5.2-neurons. In parallel, feature c is analyzed in a retinopretectal/thalamic processing stream originating in partly other retinal ganglion cells (classes R3 and R4) and continuing in TH4-neurons that are selective to predatory objects (see also Suggested Reading, Movie A1).
Figure 2.10 “Window hypothesis” of configurational feature analysis in toads. Illustrative schemes of feature-sensitive/selective neurons (symbolized by circles) integrated in a neuronal network (Ewert 1974, 2004). For explanations see text.
In detail, the evaluation of an object as belonging to the prey category results from convergence of both processing streams on tectal T5.2-neurons that weight feature p by excitatory tectal input and feature c by inhibitory pretectal/thalamic input (Figure 2.10A). Predator evaluation results from convergence of both processing streams on pretectal/thalamic TH4-neurons that are weighing p and c by excitatory pretectal/thalamic and excitatory tectal inputs (Figure 2.10C). We speak of parallel distributed interactive processing