The Anatomy of the Infra-red Sense Organ in the Facial Pit of Pit Vipers

In a recent communication the properties of the sense organ in the remark-able facial pit of pit vipers have been analysed physiologically (Bullock and Diecke, 1956). It is shown that this structure is highly sensitive to inter-mediate and long infrared radiation to or from the snake. The conclusions are reached that the organ is specialized for the detection of solid objects of slightly different surface temperature from background objects, that the object can be rather small, that there is a certain degree of directional sensitivity, and that the special features are not so much measured by a high temperature sensitivity as a caloric flux sensitivity. The receptors themselves are regarded as warm receptors, a type poorly known heretofore, functionally and anatomically. The three sizeable nervebranches supplying it are found to be nearly pure populations of warm fibres.

These results, together with the unsatisfactory state of the descriptions in the literature, have prompted us to re-examine the facial pit organ histologically.

The facial pit is peculiar to and characteristic of the crotalid snakes and has hence aroused attention for a long time. Lynn (1931) gives an account of the early studies and lists seven theories of its function that have been proposed. We may confine our notice here to the five workers who have undertaken microscopical investigation. Desmoulins (1824) was apparently the first to see the rich innervation and concluded that it was sensory, probably olfactory in function. Leydig (1868) treated it, together with other unfamiliar structures in lower vertebrates, in a memoir on organs of a sixth sense. He provided a rather good description of its histology but described the nerve-fibres as originating in ‘terminal ganglion cells’ in the sensory membrane in spite of recognizing that they are trigeminal nerve-elements, presumably with cell-bodies in the Gasserian ganglion. As we shall see, the strange form of the free nerve-endings does in fact resemble a nerve-cell with several branching processes and it is possible that Leydig saw some of them, only erring in thinking he saw a nucleus. In any case he came closer than later workers for the next 80 years.

West (1900) provided a more complete account and may also have seen some actual endings. But he also believed they were cells and identified as nerve-terminations in sections of embryos what must have been epidermal cells and in sections of adults what we call the parenchymatous cells of the membrane. It is remarkable that these workers made out as much as they did, when we recall that they had only a few, poorly fixed, mostly spirit specimens of these snakes. Not until Lynn (1931) took up this object did a worker with access to fresh material exploit this advantage. With respect to the microscopic structure of the membrane, however, little could be added and the same mistake was perpetuated concerning fifth cranial nerve-fibres with cellbodies in the periphery. This mistake was corrected by Noble (1934) and Noble and Schmidt (1937), who identified the cells of West and Lynn as epidermal cells preparing for shedding and with silver impregnation described free nerve-endings. As we now see, they still did not see the endings. Their impregnations, like many of ours, were incomplete. There is also some reason to doubt their identification of parenchyma as outer epidermis.

Several species of rattlesnake were used, Crotalus atrox, C. cerastes, C. ruber, C. horridus, C. adamanteus, but mostly C. viridis. No consistent differences between the species were noted in respect to the characters under study; it is probable, from incidental observations, that there are differences in the degree of pigmentation of the pit and especially the sensory membrane. This is likely to have little functional significance.

The membrane at the bottom of the facial pit was studied microscopically in its natural position by reflected and by transmitted light. It was removed and observed while fresh. Intravitam methylene blue staining was attempted but without significant success. Overfixing in 1% buffered osmium tetroxide (Palade, 1952) was found to be useful for visualizing the disposition of the nerve-branches in the membrane and following individual fibres to the point where myelin stops abruptly. Alcohol/formaldehyde/acetic, Carnoy, Bouin, and the fixatives required by special silver methods were used to reveal fibreendings in both whole mounts and sections. Sections were cut in low viscosity nitrocellulose or in paraffin or double embedded and stained with Masson’s trichrome stain or silver-on-the-slide methods (Holmes, 1943; Romanes, 1950). A large number of whole mounts of the membrane were treated according to these as well as the silver procedures of Palmgren (1951), Weddell and Zander (1950), and Bodian (1936), and variations of these. A small number of these showed nerve-fibres in various degrees of completeness and a still smaller number showed the palmate expansions beyond the end of the myelin sheath, with their rich aborization of branching processes. We believe a reasonable argument can be made that these preparations show the endings virtually completely, but of course we have no assurance of this.

Snakes in which a nerve or two nerves had been cut were kept for various periods and then the pit membrane was removed and over-fixed in osmium tetroxide to show the pattern of degeneration from that nerve.

One very large specimen yielded a membrane so large as to permit mounting, fresh, in specially made adapters. The transmission spectrum between 2 and i6p, wavelength was recorded with a Baird Associates automatic infrared spectrophotometer, and transmission through the visible into the near ultra-violet was recorded with a Beckman DU and a Beckman IR2.

In order to estimate the fibre-size spectrum in the nerves supplying the pit, these were fixed in Flemming’s fluid (Lillie, 1948), embedded in paraffin, cut, and mounted without staining.

A cutaway drawing of the pit showing the sensory membrane and its relation to the rest of the head is given in the physiological paper (Bullock and Diecke, 1956). Further details of the gross anatomy are provided by the earlier authors mentioned above. The portion of concern here is the thin, richly vascularized and innervated, slackly suspended, dry membrane which forms the floor of the pit. It separates the pit or outer chamber from an airfilled inner chamber which has communication with the outside through a special duct opening into the adnexa of the eye. This means that the membrane is in contact with air on both sides. Its thickness in recently-shed adult rattlesnakes is ro/z or slightly less except for local thickenings, especially where nerve-bundles lie. As other authors have shown, it is thicker in developmental stages and we believe it may be thicker in some of the other genera of the family. At top, it seems to have reached a kind of limit dictated by the size of the erythrocytes. Added to this thickness there will normally be a few microns of the multilayered cornified epidermis preparing to be shed.

On both outer and inner surfaces there is a cornified epidermal layer (fig. 1). This doubtless varies in thickness and number of layers with the stage in the moult cycle; in our sections it is from 0-5 to whereas in a figure given by Noble and Schmidt (1937) it is several times thicker. According to Lange (1931) the layer with which this is continuous in the ordinary epidermis should not be called a cuticle, for he believes it is a proper cellular layer and he applies the term Oberhautchen. We can confirm that the shed skin of the pit consists of two cellular layers but there is also a third layer without cell outlines or nuclei which appears to be a cuticle. For all its thinness, the outer epidermis must have remarkable physical properties since there is virtually nothing else standing between the soft tissues and the dry air. It is not essentially different on the two sides of the membrane except that the inner layers are less cornified.

Under the cornified layer, since it is periodically shed, should be a germinative epidermal layer. This is easily seen in the embryo (Noble and Schmidt, 1937; Lynn, 1931), and in occasional thicker places in the adult membrane. But in the typical adult structure it is so thin that it cannot be traced continuously in our sections. Much flattened-nuclei, 0-5 yu, thick, are encountered at long intervals applied to the underside of the superficial cuticular layer and these we believe represent the basal germinative cell-layer. This layer seems to be even more reduced on the inner surface than on the outer.

Under the epidermal layers on both surfaces is a distinct stratum staining green with the Masson trichrome in contrast to the layers above and below it. This is generally 0-5 to thick or even vanishingly thin on the outer side but i to 2-5 fi or even much thicker in local regions on the inner side of the membrane. We believe these layers to represent connective tissue of the dermis or corium, contrary to Noble and Schmidt (1937), who identify everything from the outer surface through the next and deepest layer as outer epidermis and thus recognize but a single connective tissue-layer, just beneath the inner epidermis. To be sure, the outer layer of connective tissue is so thin in most of the membrane that our only reason for so identifying it is its staining affinity. But the inner has discernible collagenous connective tissue fibres. Both can be followed into thicker regions where they become continuous with typical connective tissue. Fibres which appear to be collagenous connective tissue-fibres are visible in some silver impregnations of whole mounts where they reveal a sparse array of thin, straight fibres running in all directions for long distances. Both outer and inner layers of connective tissue show nuclei though in the thin parts of the membrane these are far apart and very much flattened (fig. 1).

This brings us to the middle and thickest layer which may be called the parenchyma. This is a single-cell layer, generally between the limits of 4 and 12 jit thick. The cells are quite specialized in having, unlike any others in the membrane, an abundant cytoplasm which is coarsely reticular, almost granular, and stains darkly. In Masson’s stain it takes a dull red, distinct from the brighter red of the epidermis in the side walls of the pit. In osmium preparations and in some silver impregnations this cytoplasm takes up considerable colour. The cell-type is not easily identified with any in ordinary skin. It is here regarded as part of the dermis since it lies between connective tissuelayers and is cut up into lobules by the capillary bed which occupies the same level of the cross-section. The nuclei are rich in chromatin and often show two nucleoli, but are mainly distinguished from the epidermal nuclei by being commonly a little irregular in outline rather than smoothly oval as are those of the epidermis of the sensory membrane.

The vascular bed has been figured by Noble and Schmidt (1937) and we can confirm the general character of the blood-supply as described. Being in a single plane, it is easy to visualize in the living membrane in its natural position, and also in prepared whole mounts. The richness of the supply varies considerably and can be suggested by stating that the maximum distance from any point to the nearest capillary is between limits of about 30 and 60 p,.

Aside from the nerve-supply, the only other element in the membrane is a scattering of chromatophores. In the species of Crotalus examined melano-phores are very few ; they are much more abundant in Agkistrodon. However, in silver whole mounts we often see extensive patches of what appear to be pigmentless chromatophores which have taken up a granular metallic impregnation. These have occasionally been recognized in cross-sections as very thin, dark-staining masses immediately beneath the outermost epidermal layer. They may send processes toward both surfaces.

Fig. 2 shows the aspect of the endings as revealed by our best silver-stained whole mounts of the membrane. The axis cylinder, having tapered down to i or zp. and lost its myelin sheath, suddenly expands into a very flat, broad, palmate structure from which 3 to 7, usually 5 or 6 processes spring to branch repeatedly and end as exceedingly fine, free endings. The thinness of the membrane is very favourable for working out the details of these remarkable nerve-endings because it permits whole mounts in which no concern is necessary over processes that go out of the plane of a section, or over apparent endings which may really be created by a knife; and the confinement to a plane of not more than 8p. depth facilitates following the branches. In the drawing every twig is taken from a camera lucida tracing.

The palms exhibit a well-developed reticulum of neurofibrils in reduced silver stain, and these can be followed into the larger processes and for some tens of microns up the axon. The aspect of the palm strongly resembles that of a multipolar neurone with several large dendrites and one axon, but without a nucleus. The branching of the processes is profuse, not dichotomous or regular or with many thorn-like offshoots, but irregular, often at small angles, often quite unequal and often multiple at a given point. Branches characteristically sweep back in an arc toward the palm but at a more superficial plane.

No palm is traceable to an axon which bears another palm. No branch or process is traceable with certainty into two palms, i.e. there is no anastomosis. Processes not uncommonly swell up into minor palms where they branch but this is so obvious that it is not difficult to establish a one-to-one relation between axons and palms. No axon has been seen to end in any other way than as one of these palmate expansions with processes. The palms vary in size but over a limited range of approximately 2 to 4/z across the widest waist between processes, and 20 to 30/x in long dimension between the base of the axon and that of the farthest process. We have not discerned classes based on size or other features and must conclude that, within a range of variation, we have to deal with a single morphological type of sensory ending.

The area of ramification of the terminals of one axon generally does not extensively cross capillaries; it commonly falls entirely inside one mesh of the capillary net, but there is usually more than one palm per mesh. The terminations of the processes seem typically to conform to the outlines of parenchyma cells where these are visible, as at the edge of a lobule of them. The terminations are profuse and end among the parenchyma cells, as well as superficial to them and, in smaller numbers, in the layers below them. Whether there are nerve-terjninals among the epithelial cells we cannot say, but certainly most of them are not in that layer which is so exceedingly thin, while the terminals are found through 6 or 7 μ of thickness of the membrane. The palms lie at the same depth as the capillaries, generally about 5 to 7 μ below the outer surface, but are superficial to the capillaries when they cross them. Over the large nervetrunks the palms are superficial to these. The palms are only about 2-5 μ thick at maximum. The processes mostly proceed toward terminations in a more superficial plane, a large part of them terminating 4 to 5 μ above the palms, therefore very close to the surface. A few processes pass downwards 2 to 3 μ towards the inner surface.

There is not a great deal of overlap of area supplied by adjacent palms. Some interdigitation of terminal branches is seen in every field in good impregnations, but it is, broadly speaking, not very extensive and most of the area of branching of the processes of one palm is not invaded by others. This area is not simple in outline but rather is indented and produced. It is therefore not easy to trace and hence to measure. An approximation of the typical area within which the branches of one axon ramify would lie between 1,000 and 2,000 μ.², but these are not outer limits. Diameters of 30 and of 50/x are not rare. An estimate of the average area, not including overlap, can be obtained by counting the number of palms in a microscopic field of known size. In 18 fields chosen at random and including various parts of the membrane, an average of 520 palms per mm² or 1,920p.² per palm was found. This figure can be regarded as slightly overestimating the area per palm and underestimating the number of palms since it is difficult to allow for the palms which are only partly in the measured field.

The pit membrane is supplied by three divisions of the Vth cranial nerve, as has been fully described by Lynn (1931). The smallest supply comes from the ophthalmic division which approaches the pit from above and divides into a small number of branches which enter the dorsal and posterior quadrant of the membrane. The deep branch of the supramaxillary division which approaches from below medially through the floor of the orbit and divides into 5 or 6 trunks, entering the anterior dorsal side of the membrane, is sometimes the largest nerve. The superficial branch of the supramaxillary division which approaches from behind and divides into 5 or 6 trunks entering the posterior and ventral aspects is perhaps more commonly the largest. Each of these nerves sends fibres also to the skin of the head, lips or roof of the mouth, according to Lynn, but Bullock and Diecke found exceedingly little evidence of action potentials in the last two nerves in response to tactile stimulation of the skin outside the pit and the activity characteristic of them was completely silenced by a drop of cool water placed in the pit. Our own gross dissections and cleared heads with the nerves stained with haematoxylin revealed very few twigs from the three nerves going elsewhere than into the pit membrane.

Certainly we can agree with Lynn’s statement that these nerves are overwhelmingly concerned with this organ.

The nerve-trunks as they enter the membrane are of quite unequal size, io or 12 in number and rather evenly distributed around its periphery (fig. 3, A). They immediately begin to splay out, losing size fibre by fibre and sending off a small number of branches. The larger trunks are discernible almost to the centre of the membrane. They travel on the inner side of the cross-section, between the parenchymatous layer and the inner epidermis and under the capillaries.

As they fan out, the fibres travel short distances and then lose their myelin sheaths (fig. 3, B). Shortly thereafter they expand into the terminal palms but an unmyelinated segment of about 20 to 50 ju. intervenes. The unmyelinated segments come to lie superficially to the capillaries.

The fibres from adjacent trunks do not considerably interdigitate but supply areas of the membrane with sharp, non-overlapping boundaries. This can be seen by tracing from a silver-stained whole mount the palms in the region of a boundary and noting the directions of the axons. These fall cleanly into two groups with a wavy line between the palms belonging to the two. It can be seen most strikingly in the preparations where one nerve has degenerated (see p. 227).

The fibre-size spectrum in a typical case is given in fig. 4. There appeared to be no significant difference between specimens examined of Crotalus atrox, C. horridus, and C. viridis of snout-vent length from 701 to 965 mm. There may be some tapering between a level 10-15 ^mm f^rom the membrane and the sizes close to the membrane, but if so, it is very slight as reflected in the mode. In three ophthalmic nerves the broad mode lay between 4 and 6/u. (uncorrected diameter of myelinated fibres), in one nerve between 3 and 5-5; the largest axons in these small nerves were usually between 7 and 8p. In 5 deep branches of the supramaxillary, the modes lay from 3^-5-7-5 to 5-9-5 and the largest fibres reached 17/x, although only a few per cent, of the fibres are above gp. Four superficial branches of the supramaxillary division lay in the same range and, for comparison, three mandibular nerves supplying ordinary skin gave curves which overlapped completely with these.

In the pit membrane shortly before the myelin sheath stops, the diameters appear to have fallen somewhat but are still typically 3 to 5 p.

We have not attempted to estimate the population of unmyelinated fibres in these nerves. In the silver-stained preparations of the membrane there do not appear to be stained fibres of much smaller diameter or of a different form of termination from the myelinated fibres described already.

The number of fibres in the cross-sections of the nerves is of interest as an independent means of calculating the total nerve-supply to the sense organ.

It should give a reliable maximum figure, although the figure will be too high by an unknown amount due to the component destined to supply the ordinary skin of the anterior part of the head. Counts of representative aliquot parts of each cross-section, corrected to the total cross-section in a 965-mm specimen of C. horridus, gave 815 myelinated fibres for the whole ophthalmic nerve, taken a few millimetres proximal to the pit, 2,921 fibres for the superficial branch of the supramaxillary, and 3,538 for the deep branch, a total of 7,274. In three other snakes totals of 8,254, 6,626, and 7,410 were obtained. Dividing these figures by the approximate areas of the membranes in each case gives 586, 825, 839, and 1,950 fibres per mm². The estimates of areas are particularly subject to error, being made on the excised, concave membrane, so that the real scatter may not be as great as is suggested by these figures.

Five specimens which had had one or two of the nerves cut 1 cm or more from the pit were kept for a few weeks to allow degeneration of the severed axons and then prepared by osmic fixation and staining of the sensory membrane as a whole mount. A period of 2-3 weeks appeared optimal at room temperature for a clear distinction between degenerating fibres and normal ones without too much of the myelin debris being already removed by phagocytes—a process which is beautifully shown in these whole mounts in later stages.

By this means it is possible to map the area of supply of the separate nerves. The nerve-fibre counts in the cross-sections of the nerves have already indicated that the relative contributions of the several nerves vary somewhat. We cannot be sure, therefore, that in the few cases mapped a typical picture of the distributions was obtained. The ophthalmic nerve has not been cut alone but after cutting the other two a very small segment of the membrane, less than one-sixth of its area, remains supplied by undegenerate fibres. After cutting the superficial nerve, a segment of about 60% of the area shows degenerate fibres. The deep nerve, by difference, supplies about half the membrane. These segments do not overlap in the slightest. We have not attempted to preserve accurately the orientation of these segments relative to the body axes.

An astonishing feature of these preparations is that, besides the contrast between normal and degenerating nerve-fibres in the two parts of the membrane, there was found in a few cases a striking contrast between the two regions in the staining affinity of the parenchyma cells (fig. 5, A, B). This means that the whole background in the osmium preparation (cleared whole mount) is either dark (in the normal part of the membrane), or pale (in the part innervated by the now degenerating fibres). Moreover, the separation between these two conditions is sharp. The line is easily followed under the high powers of the microscope because each cell in the region of the boundary is either dark or pale, i.e. they do not intergrade and the line is unbroken: dark cells are not found on the pale side of the boundary or conversely. Curiously, the line does not consistently follow capillaries but it sometimes does. This means that it goes through the middle of the lobules of parenchyma defined by the vascular bed and, therefore, presumably that given nervefibres actually do send terminal branches across capillaries more often than was indicated above from the silver preparations. We have not attempted to go further into the pathological histology to inquire what is the nature of the change in the parenchyma cells. It would seem of considerable interest as a dramatic example of the dependence of peripheral, non-nervous cells for some trophic requirement upon the intactness of the nerve-endings, in this case sensory endings.

Physiological data indicated that the action spectrum for adequate stimulation of the receptors by radiant energy lies essentially between 1-5 and 15 p. or more of wavelength, i.e. in the medium and long infra-red, and that visible light is virtually ineffective (Bullock and Diecke, 1956). This action spectrum must be entirely contained in, i.e. must lie under, the absorption spectrum of the membrane, although they may not coincide. That is, only the energy absorbed can have an effect, although not all energy absorbed need contribute to the stimulation of the receptors. But, according to the hypothesis reached, namely that the receptors are responding to the temperature changes of the tissue resulting from absorption of radiant energy, the absorption spectrum can be expected to coincide with the action spectrum since the heating effect will not discriminate between different wavelengths absorbed. A lack of coincidence could be caused by the absorbing structure being unable to heat the receptors, because of distance between them or loss of heat by transport, e.g. in erythrocytes.

One specimen of Crotalus adamanteus became available which was so large that its two pit-membranes together covered a window of usable dimensions for spectrophotometers, 2 by 15 mm. The freshly dissected membranes were washed free of surface blood, dried for 10 min in a P₂O₅ desiccator under partial vacuum, and mounted by Scotch tape while being held flat with a glass slide. The membrane is naturally curved in both directions, like an orange peel, and cannot be perfectly flattened. Although the light beam consists of parallel rays normal to the plane of the membrane, there is necessarily some loss of light due to reflection and scattering, and this is of unknown amount. For the Baird Associates IR spectrophotometer (NaCl prism), a single membrane in a window 2 by 7’5 mm was used. The results are shown in fig. 6. The main features of interest are the deep absorption maxima at 3 /x and at 6/x, the gradual rise in transmission from 6-6 to lop, and the rather high (approximately 50%) transmission in the long infra-red out to at least 16¼. The latter is not usually associated with tissue or organic substances, especially with high water-content, and must be put down to the thinness of the membrane. Since the long infra-red is the region of most effective stimuli, these being from objects of a temperature only a few degrees different from the snakes, this finding indicates a large percentage inefficiency.

The Beckman DU spectrophotometer showed no large absorption maxima in the near infra-red or visible regions but at wavelengths shorter than 490 μ. absorption rapidly increased and was virtually complete at 380/r. The transmission through the visible was not high (between 40 and 50%), so that either there were sizeable losses by scatter and reflection or there is a discrepancy between the absorption and the action spectrum, since visible light was ineffective in stimulating unless at very high intensity.

It was pointed out by Bullock and Diecke (1956) that the physiological specializations of this organ were not profound. Sensitivity is much higher than in other temperature receptors directly studied by nerve impulse recording, but not higher than that calculated for man. The steady state sensitivity found in the other known temperature receptors is not present here. The unusual forms of response to strong stimuli are not necessarily adaptations. Most of the adaptations of the organ are anatomical.

The primary one of these is the concentration of warm receptor elements into a pure population of that modality. The high density of these elements means that full advantage can be taken of the phenomenon of central summation (in which the threshold is lowered by simultaneous stimulation of many receptors), even though a very small area of the skin is stimulated. This means that the stimulation flux (calories per cm² per sec) is small. Central summation is the phenomenon probably mainly responsible for behavioural threshold, assuring as it does that weak stimuli which give unreliable signals in unit receptors give quite reliable signals in a group of parallel channels.

The thinness of the sensory membrane probably confers a higher sensitivity to caloric flux in a certain range of stimulus durations. If we separate from thinness the closeness of the nerve-endings to the surface, that is, compare a thick membrane with endings equally close to the surface, then thinness will permit a given flux to warm the receptors more, provided that the stimulus is not allowed to act indefinitely and provided that it is allowed to act longer than a very brief period. In the case of a very brief stimulus, the advantage of thinness decreases because there is proportionally less loss of heat to the depths in the thick membrane than there is for longer stimuli. In the case of very long stimuli warming approaches a steady state and, unless there is a large sink in the thick membrane into which heat flows, there will be little advantage in a thin membrane. The range of periods for which the membrane as we see it offers an advantage cannot be accurately calculated without more knowledge of the loss of heat to the blood, but apparently it extends from a second or less to many seconds. This is based on the consideration that in a thin membrane and neglecting losses by reradiation and blood-flow, the change in temperature for a given flux is proportional to the time, while in a thick structure which conducts heat past the receptors, the change is proportional to the square root of the time (see Bullock and Diecke, 1956).

Perhaps more important than the thickness of the membrane is the superficial location of the receptors. We could not give accurate measurements of the depth of the endings, but the average depth is probably no more than 5 μ and possibly only 2 μ from the outer surface. This not only increases the promptness of the response and the ability to detect flickering and brief stimuli, but situates the detector at the most favourable place in the temperature gradient. Actually we can say, from the transmission spectrum, that the adaptation has gone about as far as it is worth going, for the layer between the nerve-endings and the surface is already so thin that it is absorbing something less than half of the energy in the wavelengths chiefly available.

We need not discuss here the special features of the posterior chamber, its outlet and sphincter, the location on the head, the angle of view, the circulation, or the chromatophores (fig. 5, c). But one further feature of the anatomy of the accessories deserves mention. This is that the pit characteristically does not taper inward but actually overhangs, i.e. the mouth is smaller than the membrane at the bottom. This is obvious to the eye and in a small series of randomly chosen specimens of four species, the diameter of the pit opening of 10 pits was from 35% to 50% of that of the membrane. The consequence of this, together with the small depth of the pit, is that most radiating objects unless very large or very close will not illuminate the whole sensory membrane but will cast shadows of the pit margin. This confers the possibility of deriving information from the sense organ about the direction of small objects or of edges of large objects, if the resolution of the sense organ is sufficient. We have already seen that the innervation is rich. There are between 500 and 1,500 axons ending per mm². This compares with about 800 optic nerve fibres per mm² of the retina in man, averaging an estimate of 800,000 fibres into an estimate of 1,000 mm² of retina. The membrane provides enough absolute area and hence nerve-fibres to resolve even fairly poor shadow margins: in snakes between 350 and 1,000 mm snout-vent length the membrane diameters fall near a line connecting 2 and 4-4 mm, corresponding to areas somewhat over 3 and 15 mm² (since the surface is concave). We may conclude that the central nervous system receives adequate information to analyse directionality with a degree of usefulness, especially if the snake scans or if the object is moving.

A final point should be raised concerning the nature of the parenchyma cells of the sensory membrane. These are the only non-nervous cells of the sense organ which are specialized, except for the attenuation of the epithelial and connective tissue-cells. They are not readily identified with any cell-type in ordinary skin. They are polygonal, about 6p thick, with an osmiophil cytoplasm of uniform, dense, reticular structure. Most significantly, they react with a change in staining quality to loss of the nerve-endings and evidently depend on the particular ending intimately located about them; for there is no grading off of the reaction from the boundary between normal and denervated regions (compare Hillarp, 1946). These facts might lead in terms of classical histology to regarding the cells as sense cells. We do not believe they should be so considered unless evidence is obtained that they actually respond to the normal environmental stimulus and mediate the initiation of nerve discharge. In the physiological study Bullock and Diecke were unable to obtain such evidence, although they could not eliminate this possibility. If, as we suppose, the parenchyma are not sense cells, they very probably do have some essential accessory function in connexion with the transducing of temperature change into neuronal activation.

It has been suggested in the literature repeatedly that there may be an analogy between the facial pits of pit vipers and the labial pits of some pythons and boas. No other organ appears to resemble these in structure or function as far as is known. The labial pits are simple depressions, without an inner chamber and hence with a floor receiving the nerve-supply but no membrane (Noble and Schmidt, 1937). Behavioural experiments clearly suggest a function as receiver of warm radiation, like the facial pits of crotalids, although the authors have apparently not realized this and speak of air temperature, measured by a mercury thermometer (Noble and Schmidt, 1937) or air movement resulting from the local heating of the air (Ros, 1935).

As far as we are aware the facial pits are not similar in anatomy to any other temperature receptors among animals, or indeed to receptors of any modality. The nerve-endings themselves seem also to be unique, in the form of the terminal palmate expansion. It does correspond with the conclusions of Weddell and his collaborators that temperature-endings should be free nerve-endings and not necessarily corpuscular. Free nerve-endings have been described in reptilian skin (see Boeke, pp. 859-66), but none which give any special suggestion of being forerunners of these. Bullock and Diecke (1956) could not find evidence of any temperature reception resembling that concentrated in the pit, on searching through ordinary skin-nerves of rattlesnakes and even the homologous branches of the trigeminal in non-crotalid snakes.

If, as we suppose, the receptors are detecting and transducing small temperature changes in the tissue, there is no special significance in the spatial aspect of the ramification of each axonal-ending, that is there is no particular advantage in covering a wide territory as there would be if intercepting photons or mechanical deformation were their function. But we may suggest that the development of an extensive surface, subdivided into quasi-independent regions, enhances the probability of firing to a small change. This is based on the supposition that these fine processes are prone to autorhythmic, sub-threshold, graded changes of state which do not propagate but spread decrementally, not interfering with each other but summating so that at some critical point, perhaps at the emergence of the axon from the palm, all or none impulses are initiated whenever some sharp threshold is exceeded. We would visualize the several processes of the palm each undergoing independently, fluctuations in state, including membrane potential, and these may be quite rhythmic. As they happen to summate and therefore to spread farther they would from time to time trigger impulses in the axon. This would explain, as far as it goes, the non-rhythmic spontaneous background discharge. It also makes it possible to invoke the most exquisitely sensitive mechanism of minute changes in the potential gradient between one part of the neurone surface and another, recently shown by Terzuolo and Bullock (1956), to alter the frequency of firing of already active neurones. This hypothesis of the activity of sensory processes is not uniquely applicable to the present receptors but seems reasonable for many branching terminations of small diameter, especially those with a spontaneous or steady state discharge. It places afferent axonal terminations in a class with dendrites, in not supporting impulses or propagating a disturbance toward the axon, but influencing the initiation of impulses in the axon by small changes in potential gradient along the neuronal membrane (compare Fessard, 1956).

It is a pleasure to thank Dr. Ralph Nusbaum of the University of California Atomic Energy Project at Los Angeles for running the infra-red transmission spectra and for much valuable advice. We are also indebted to Dr. R. B. Cowles of this department and Mr. Charles Shaw of the San Diego Zoo for generous provision of many animals. Financial support from the National Science Foundation is gratefully acknowledged.