By Gary W. Harding
11 October 2002
This story recounts much of the work done on the cortex of the somatosensory/motor system of the cat in the electrophysiology laboratories in the Department of Physiology and Biophysics, University of Washington. It is dedicated to Arnold L. Towe (1927-2002), our mentor and good friend, whose guiding intellect shaped the scientific inquiry for this research. Remembered herein are Russell W. Morse and Fredrick A. Harris who were instrumental in this work and whose untimely deaths sadden us still.
I must point out that the manuscript has been written without giving Arnie an opportunity to convert "Harding's cockney to the Queen's English". Hopefully, I have learned enough about writing from him and others so that a translation will not be necessary.
In addition to the scientific method and the rules of evidence practiced within, there were three rules posted upon the laboratory door:
I. You will not see unless you look!
II. What you see depends upon how you look!
III. Never, ever, forget rules I and II!
Students were advised; to find the answer to a question regarding the cortex, "go and ask the cat". Any student frequently found in the library was dragged by his ear, usually kicking and screaming all the way, into the laboratory.
A question can be answered from many points of view and at many levels. In the cortex, one can examine the tissue from a structural and/or functional perspective. The structural approach can range from gross anatomy to intracellular and membrane architecture. The functional view goes from behavior to membrane properties and cellular biochemistry. The view followed in our journey was functional in the sense that the extracellular electrophysiology of individual neurons was studied. It was also structural in the sense that the connective anatomy in the somatosensory/motor system was its basis. The motivations for this approach to study of the brain were questions regarding the relations between the brain, behavior and evolution.
Much of what we did in studying the nervous system was similar to the approach of a natural historian. We gathered measures of inputs, outputs and the distribution of entities and then tried to make sense out of the relations among them. Along the way, it was natural for these data to be segregated into categories so that classifications could be defined [Mann, '79]. The number of classes so formed produces different views of the system under study. On the one hand, there are "lumpers" who use a small number of classes. They run the risk that important detail may be overlooked. On the other hand, there are "splitters" who define a large number of categories. They run the risk that they may not be able to see the forest for the trees. The ideal number of classes is difficult to determine beforehand. However, in the end, the best taxonomic scheme is the one that imposes the most order in the data [Tyner, '75].
The acquisition of scientific knowledge might be assumed to be an exponential growth process. However, it is much more like a helix. It has been shown that the number of articles published on a particular scientific topic cycles up and down about every 10 years. The same questions are re-examined as new research tools are developed and new researchers enter the field. But, in some cases, finer and finer detail does not really add significant new understanding. Added to this is the hazard of "intellectual phase locking" [see Branscomb, '85; re: When does one know that an experiment is finished?].
One of the tools which can be used to organize the data is to form a model of how one concludes that the system under study works. However, the tendency to attach significance to the model itself must be avoided. It is only a "framework upon which one can hang his hat (ALT)". Models for a particular system can be constructed from many materials. One can use diagrams or mathematical equations or computer simulations or electronic circuits or mechanical devices or even natural language to describe the same system. But, no claim can be made that the model represents how the system is actually built.
The modeling process is sometimes misunderstood. The tendency is to get excited when a model describes a system very well. However, this is not the purpose of a model. To see why this is the case, consider the following steps in the modeling process:
1) The system for study is identified and preliminary data are collected to
measure its properties (natural history).
2) A model is formed which accounts for all of the input-output observations.
3) New experiments are designed to test the model's predictions.
4) The step 3 experiments are conducted on the system under study.
5) The outputs of the system and model are compared.
6) If they agree, then go to step 3.
7) If they do not agree, then go to step 2.
Many would conclude that the desired outcome is to repeatedly go through step 6. However, after a few cycles, this process gets exceedingly dull. Things get interesting when there is a passage through step 7. Why? Because now the model must be amended or even completely replaced. If this is successful, then something new has been learned. In either case, the model is so useful because it serves as an engine for experimental design. Experimental design is often the hardest part of doing research.
It has been said that science is the process of figuring out how to ask the right question. Once the right question is asked, the answer is straightforward. However, one must keep in mind that rule III on the laboratory door applies to the modeling process as well. If the experiments performed through the model and upon the system under study have been conducted without artifactual results and the two outputs do not agree, then it is the model that is in error not the data collected from the system.
In the end, every model loses its utility. Eventually, step 2 or 3 cannot be executed and either the system is completely understood or further investigation must be abandoned. But, the knowledge gained through the process is retained. The model may be revived and carried on at a later time but, new information and/or better research tools are required.
The 'lay of the land' in cat sensory-motor cerebral cortex was first examined from the gross anatomy perspective. The pathways from somatosensory receptors were followed up the spinal cord and through the dorsal and dorsal lateral columns to the cuneate and gracile nuclei. From there, the sensory information was determined to cross to the other side and pass through the medial lemniscus to the venterior-lateral and venterior-posteriorlateral nuclei of the thalamus. The neurons in the thalamus were shown to project to the sigmoid gyrus of the cerebral cortex terminating predominantly in layer IV. Neocortex had been shown to be organized in layers I - VI [von Economo, '29].
The cat sensorimotor cortical region was bounded medially by the sagittal sulcus; posteriorly by the ansate sulcus; laterally by the coronal sulcus and anteriorly by the rostral margin of the gyrus (Fig. 1). The cortical cells sent their axons either locally to other cortical neurons or through the cerebral peduncles and then collected together to form the pyramidal tract in the ventral medial surface of the medulla: These fibers coursed then to the ventrolateral columns of the spinal cord.
Figure 1. Left) Diagram of the left frontal hemisphere showing the sites (1-7) that were sampled. The cytoarchitectonic regions are labeled from left to right. (pcd) - post cruciate dimple. Right) Diagram of cross section of cortex along region labels at left.
The sigmoid gyrus cortex itself was divided into areas depending upon the local distribution of neuron types [Hassler & Muhs-Clement, '64]. There were defined (caudal to rostral) areas 2, 1, 3b, 3a, 4r and 6ab and area 4r was noted to have a pre- and postcruciate portion separated by the cruciate sulcus (Fig. 1).
With the sigmoid gyrus roadmap charted, the next task was to determine the gross functional topography of the region. It was found that the precruciate portion was "motor" in the sense that electrical stimulation of many places in this area would produce peripheral movement. The postcruciate-precoronal region was found to be "sensory" in the sense that electrical stimulation of the soma produced evoked potentials related to the site of stimulation. The general organization of both the motor and sensory responses proceeded from the hindlimbs in the medial part, through the trunk in the middle part and to the forelimbs and neck in the lateral part [Woolsey and Fairman, '46; Woolsey, '58]. The head and mouth region was found in the gyrus lateral to the coronal sulcus.
Work elsewhere in the nervous system with microelectrodes suggested that this technique would be useful in determining the behavior of single neurons in the cerebral cortex. Perhaps, the neuronal circuits which underlay the topography in the sigmoid gyrus could be worked out. Consider the following experiment:
1) Sample single neurons extracellularly within the depths of cortex 2 mm
posterior to the lateral tip of the cruciate sulcus (using a microelectrode oriented
perpendicularly to the cortical surface).
2) Stimulate the contralateral forepaw electrically to find responding neurons
(spike discharges).
3) When a neuron was located, stimulate the pyramidal tract electrically at a high
rate to verify antidromic activation. Record the depth below the pial surface and
the responses to electrical stimulation.
4) Map the peripheral receptive field and modality of the
neuron to natural stimulation.
This experiment made sense because the sampling site had been shown to be topographically related to contralateral forepaw (CFP) input and the anatomy indicated that cortical output was into the pyramidal tract (PYR). When the experiment was conducted in a few cats with this model, the findings were puzzling. Many cells could be found which responded to the CFP electrical stimulation but very few were antidromically driven from the PYR. The receptive fields for the ones that would not respond to the PYR were small and located on the contralateral forepaw. However, the ones that were antidromically driven from the PYR had receptive fields all over the body. The small-field cells were found mostly in layers II and upper III while the wide-field cells were found predominantly in layers lower III & V. The natural stimulus modality was mostly touch for small-field cells and hair deflection for wide-field neurons. In microelectrode penetrations (tracks) with several cells isolated, the modalities were almost always mixed (stoma surface and deep tissue).
These results indicated that the model was wrong so, it was amended and the following was added to the experiment:
5) When a cell is found with CFP stimulation, stimulate the other three paws (IFP, CHP and IHP) as well as the PYR.
When this experiment was conducted in other cats, small-field cells were again found which were not activated by the PYR but, many more wide-field cells were found and about half of them were responsive antidromically to PYR stimulation.
These results lead to the conclusion that there were two populations of cells in the tissue. Further, it appeared that these two sets interacted in some way because the input was not directly connected to the output. The model was updated to reflect this interaction.
The question now was, how would we test the current version of the model? One way would be if we could find a substance which would turn off the wide-field cells so that we could study the small-field cells in isolation. We also needed to find another substance that turned off the small-field cells so that the properties of wide-field cells could be determined. It was found that Nembutal was ideal for turning off wide-field cells. Under this anesthetic condition, the wide-field cells were no longer responsive. Unfortunately, we were unable to identify a substance, that would shut off small-field cells. However, we were able to find the next best thing. If we couldn't turn the small-field cells off, then could we jazz up the behavior of wide-field cells so that the usually subthreshold interactions would then be clearly demonstrated. This substance was alpha-Chloralose. Under this anesthetic condition, not only were wide-field cells jazzed up, but there were many more of them present than had been seen before. Curiosity led to testing not just the somatosensory receptive fields but other sensory inputs too. The wide- field cells were found to respond equally well to visual and auditory inputs!
The use of barbiturate anesthesia in animal studies has been and still is widely accepted. Aspersions were cast, however, upon the use of Chloralose. It has been regarded as a "convulsant" drug since an animal so anesthetized will twitch or jump at the slightest sensory input. But, remember that Chloralose jazzes up the wide-field cells whose axons leave cortex for spinal destinations via the pyramidal tract. Thus, one would expect accentuated motor responses mediated by the intense corticofugal output. Also, observations of neuronal behavior under the Chloralose condition showed that the cell's responses in no way resembled those under known convulsive states [Mann & Towe, '74].
The results from studies with Chloralose anesthesia have been rejected by many because "it is an unnatural state and the brain doesn't normally work that way". We can not accept this argument. In the first place, it is a common trick in science to put a system under study into a perhaps contorted but easily observable state. Even if the system is not in its "normal condition", the relations between its inputs and outputs so determined, do exist and must be accounted for. Secondly, if a system is capable of the responses produced in any state and the mechanism can be discovered, then information about how the system works has been gained. Thirdly, other drugs are commonly used in neuroscience experiments with no attention paid to their state altering properties unless that was the point of the study. Why then, is Chloralose singled out as such a "terrible drug"?
Meanwhile, studies of the cat somatosensory cortex were being conducted elsewhere. They found only small-field neurons and in any one electrode penetration, these all had the same sensory modality. The results from these studies led to the formation of the "columnar hypothesis" of cortical organization [Mountcastle, 1957; Mountcastle et al., 1957]. This notion quickly gained popularity because it satisfied the "simplicity hypothesis" (i.e., among competing explanations for a phenomenon, the simplest one is more likely to be correct). However, our results using putatively the same approach were quite different. We wondered if the disparity could be explained by Rule II on the lab door.
The sequence of events implied by the previous chapter didn't really happen that way. That chronology was presented to make the point regarding Chloralose anesthesia. Our use of Chloralose was actually an historical accident. At that time, the anesthetics of choice for human surgery in the United States were barbiturates. At the same time, the anesthetic used in France was alpha-Chloralose. It so happened that the research into cerebral cortex was begun in our laboratory by a Frenchman by the name of Amassian ['53]. He used Chloralose anesthesia and we continued to use it as a matter of continuity. It wasn't until later that we collected data under barbiturate and awake conditions in order to compare our results with those of other researchers.
Since, under Chloralose anesthesia, a substantial CFP evoked potential was present 2 mm anterior to the lateral tip of the cruciate sulcus [see Harding, Stogsdill & Towe, '79], the early experiments included microelectrode samples from this site as well as the postcruciate site described above [Towe & Kennedy, '61; Kennedy & Towe, '62; Morse & Towe, '62; Patton, Towe & Kennedy, '62; Towe, Patton & Kennedy, '63; Towe, Patton & Kennedy, '64]. Both sites were in area 4r. Large numbers of neurons with sensory responses were found in the supposedly "motor" site. The findings with regard to neuron behavior were much the same at both sites but, the distribution of small- and wide-field cells differed. The experiments included much more detailed measures of cell behavior than implied by the one defined in the previous chapter. These were spike latencies and probabilities, spikes per discharge and stimulus intensity threshold and stimulus frequency following. The measures were used to characterize cell behavior and to establish interaction precedence between neuronal sets.
It became clear that systematically gathered samples would be required to understand what was going on in the neuronal circuits of the tissue. The concept of neuronal population analysis took form as the way to sample the cell's behavior [Towe, '65 & '68]. This process is much like a "Gallup" or "Harris" pole of public opinion. Detailed questions are asked of a sufficient number of individuals to measure the position of the population as a whole as well as sub-populations within it. Interactions between sub-populations are inferred from the data and a model of these relations. However, the data must be corrected for sampling bias before confidence can be placed in the results [Towe & Harding, '74]. (Sample size and sampling bias issues in neuronal population analysis will be discussed in a subsequent chapter.)
Later studies with Chloralose expanded the investigations into other areas of the gyrus. Samples were taken at the pre- and post-cruciate sites topographically related to the contralateral hindpaw in areas 3b and 6ab [Doetsch & Towe, '76; Slimp & Towe, '77; Towe & Slimp, '77]. The tissue in areas 3a and 3b of the contralateral forepaw region was also examined [Towe, Whitehorn & Nyquist, '68]. The same small- and wide-field neurons were found at all sites but their distributions were quite different from site to site. In particular, the PreCor CFP site (area 3b) showed almost all small-field cells and very few wide-field ones [Morse, Adkins & Towe, '65]. In area 3b for the CHP site, the findings were similar to the CFP PostC site (area 4r) [Doetsch & Towe, 76]. Not only that, but it was discovered that part of the wide-field collection at each site was in fact a separate set. These were the neurons which responded to the on-focus paw and its ipsilateral counterpart only (CFP-IFP in forepaw sites or CHP-IHP in hindpaw sites) [Nyquist & Towe, '70]. We had not noticed this until the set was predicted by the developing model of the neuronal circuits. It was later shown that the ipsilateral response was mediated by the homologous tissue in the opposite hemisphere [Tyner & Towe, '70]. Samples were also taken in somatosensory area S II [Morse & Vargo, '70; Slimp & Towe, unpublished]. In order to collect the population samples more efficiently, an automated laboratory was implemented which was used to conduct many of the later studies [Harding & Towe, '76].
Several additional studies were carried out to follow up on specific questions. The details of the influence of small-field cells on wide-field cells [Towe, Tyner & Nyquist, '76] as well as the effect of wide-field neurons on small-field ones [Satterthwaite, Burnham &Towe, 78] were examined. Intracellular recordings were collected for both small- and wide-field neurons [Whitehorn & Towe, '68]. The descending cortical influences on the afferent input [Adkins, Morse & Towe, '66], the response to stimulation of the dorsal columns [Ennever & Towe, '74] and the response to stimulation of the medial lemniscus [Satterthwaite & Towe, '75] were investigated. The effect of several drugs, including strychnine [Mann & Towe, 74; Towe, Mann & Harding, '81], was tested. (These and other studies will be presented in later chapters.)
The same sites surveyed with Chloralose anesthesia were reexamined in the barbiturate and awake states. The picture under barbiturate anesthesia was quite different from that under Chloralose. Precruciate responses disappeared altogether and wide-field PostC neurons were no longer responsive [Harding, Stogsdill & Towe, '79]. The evoked potentials at the other sites were similar. The addition of a small amount of Nembutal while observing a wide-field cell under Chloralose anesthesia, sequentially inhibited the off-focus responses in conduction distance order [Harding, Stogsdill & Towe, '79]. These neurons were thereby either converted to small-field cells or non responsive ones. The wide-field response would return in the reverse order as the barbiturate was metabolized. In the awake but drowsy or quiet condition, cells were found in numbers more than 3 times as great as in either anesthetized condition [Baker, Tyner & Towe, '71; Slimp & Towe, '77; Towe & Slimp, '77]. Their distribution at any site was closer to that of the Nembutal sample than the Chloralose collection. However, when the animal's arousal state was shifted toward very alert, many small fields became much larger and returned to their former small size when the animal was quiet again.
The next chapter will provide a detailed account of what has been described in general terms here. The framework for this will be rule III on the laboratory door, neuron population analysis and the neuronal circuit model as it developed.
From the previous chapter, we saw that the view of cortex depends upon the animal's state and the site sampled. It also hinges upon the cell-behavior classification system, sample size and sampling biases. The taxonomy used here is multidimensional, much like the axes on a hyperspace graph. The major axis is the size and location of the peripheral receptive field (CFP, IFP, CHP and IHP). As we will see, this is called "major" because it imposes more order in the data than any other. Another axis is the depth of the cell below the pial surface. Two other axes are the response to antidromic stimulation (Antidromic - Pt, Not antidromic - ~Pt or Unknown) and the natural stimulus modality of the receptive field (Touch, Hair, T&H, Pressure, Joint, Muscle-Tendon-Joint, Mute or Unknown). In addition, there are axes each in the direction of spike Latencies, Probabilities, Spikes/Discharge and Frequency Following at maximal stimulus intensity and Intensity Threshold. Added to these are the spike Latencies, Probabilities and Spikes/Discharge along an Intensity series, a Frequency series as well as in response to a Conditioning-Testing series. Any of these axes can be used alone to form a minimal number of categories. Many axes can also be used together to define a broader based classification system.
Of course, it was impossible to sample along all of the axes in any one animal. The axes relevant to the question under investigation were used in any one study and an adequate sample was taken to reach stability in the results. However, since many studies shared axes in common, the data from several studies could be pooled. Over a period of 25 years, the behavior of more than 17,500 sensorimotor cortex neurons have been sampled in about 900 cats. Here is what we found out about the sensorimotor cortex by "asking the cat".
Adequate sample size is dependent upon the number of classes in the taxonomy. Sufficient samples are needed in the most infrequently occurring class so that enough data is obtained to achieve stability in its measures. Unfortunately, stability can not be recognized until twice as much data as necessary have been collected. If the population sample is separated by sampling order into two parts by first half, last half; even, odd; and random selection, and there is no significant difference between the measures in each group for each class, then a sufficient number of samples have been collected.
Microelectrodes are biased in the direction of larger cells. Thus, a population sample will have an under-representation of small neurons and an over-representation of large ones. To get an estimate of the degree to which this is the case, we took the distribution of axon diameters in the pyramidal tract, the distribution of antidromic spike latencies in cortex and applied the "size principle". We assumed that the diameter of a pyramidal tract fiber is proportional to the soma size of its origin. Then we converted the fiber distribution to its expected antidromic latency. The ratio of the expected and the observed antidromic latency distributions reflected the sampling bias. Then, given the distribution of cell sizes within the depth of cortex [Ramon-Moliner, '61], we calculated the correction factor required to adjust the sample results for microelectrode sampling bias [Towe & Harding, '70].
If wide-field cells under the Chloralose condition responded to stimulation all over the soma, did they respond to other inputs as well? The answer was yes. They fired equally well to sharp hand claps as they did to electrical stimulation of the paws. Not only that, but they responded to light flashes as well. On-responses, sustained responses and off-responses were commonly found. Auditory inputs were not systematically studied but visual responses were followed up in separate studies involving electrical stimulation of the optic chiasm [Gahery & Towe, '93] and light stimulation of visual fields [Videen, '81]. What about inputs from the gut and reproductive system? Yes, there too, we found responses from electrical stimulation of the splanchnic [Tyner, '79] and pudendal [Slimp & Towe, 80] nerves. These findings were incorporated into the neuronal circuit model to account for the observations.
If wide-field cells in sensorimotor cortex under Chloralose anesthesia responded to auditory and visual stimulation, did primary auditory and visual cortex have wide-field cells which responded to somasthetic inputs? A few natural history experiments were conducted to satisfy our curiosity. The answer was yes, they did. In auditory cortex they had properties similar to wide-field cells in sensorimotor cortex with sound being the on-focus input. In visual cortex, the same thing was found except that the wide-field cells did not respond to light.
One day, we went through step 7 of the modeling process and the only way we could restructure the model to account for the results was to create a new class of neurons. This class had to have the property that cells within it must respond only to CFP and IFP stimulation. We had included these neurons in the wide-field set in the past. When the data were re-analyzed with this set of bilateral-field cells separated out and wide-field cells re-defined as having to respond to at least 3 paws, the wide-field results showed much more order. The properties of the bilateral-field set were much like the small-field set in many respects but were similar to wide-field neurons in some others [Nyquist & Towe, '70]. Having defined bilateral-field class, it was logical to define unilateral-field classes as well (i.e., CFP & CHP). However, these neurons were rare in the population sample.
To examine where the ipsilateral input was coming from in bilateral-field cells, a study was conducted which involved the ablation of the ipsilateral sensorimotor cortex. It was found that the bilateral-field cells were thereby converted to small-field cells [Tyner & Towe, 70]. This implied that the ipsilateral input was arriving via the corpus callosum. The neuronal circuit model was further amended to include these observations.
Figure 2. Modulation of CF latency (Observed/Expected Latency). Observed (solid lines) and expected (dashed line) mean first-spike latencies to 1/sec surpramaximal stimulation of each of the four paws. Vertical hatching shows region of no modulation.
The spike latencies to wide-field inputs were expected to vary together. That is, if the spike latencies to the off-focus inputs were early in the distribution, then the on-focus response should have been early as well and late for late responses. But, the on-focus latencies were not co-variant. Using z-scores we calculated the expected on-focus response latency for each neuron and then the ratio of the observed and expected latencies. This was used as a coadunate index of neuronal behavior. When the data was plotted as a function of the coadunate ratio, a high degree of order resulted. Most wide-field on-focus responses showed facilitation, some showed no change and several showed inhibition. The only explanation for this effect was that small-field cells were modulating the behavior of wide-field neurons [Towe, Tyner & Nyquist, '76].
If there were wide-field cells in visual cortex which did not respond to the on- focus input, were there neurons like these in somatosensory cortex? If so, we would not have seen them because we used CFP stimulation to hunt for responding cells. We had been caught violating rule III. The coadunate model suggested that they would sit just to the right of the cells in Figure 2. Alternate stimulation of the CFP and IFP was used to look for them. And sure enough, they were there in significant numbers. They were called "I" neurons to signify that the on-focus response was inhibited. The curious thing was that this inhibition occurred at maximal stimulus intensity but not at lower intensities. It was argued that this was due to the recruitment of A-delta fibers [Tyner & Miller, '77]. No IFP only responses were found so, CHP and IHP hunting stimuli were not pursued even though this was an obvious violation of rule III. A study was conducted, however, to get an estimate of how many cells we would not see by using CFP alone as the hunting stimulus. This question was examined by using stimulation of the medial lemniscus to hunt for responding cells [Satterthwaite, '76]. The result was that we were activating about 70% of the cells from CFP stimulation. An additional 10% were I cells activated from IFP stimulation. The remaining 20% were likely to have been those responding to deep somasthetic inputs.
What was the mechanism for the interaction between small- and wide-field cells? Were the influences pre- or post-synaptic? A partial answer to this question was found in a Conditioning-Testing (C-T) study. The on-focus response of wide-field pyramidal tract neurons was conditioned by a stimulus to the PYR. The only explanation for the modulated threshold effect seen was that the influence was presynaptic [Mann, Holt & Towe, '77].
If there were I neurons sitting to the right of Figure 2, were there cells to the left as well? These would have been neurons classified as small-field cells but which were actually wide-field neurons whose on-focus response had a short latency and whose off-focus response was subthreshold. A C-T paradigm was used to test for these cells. The conditioning stimulus was off-focus and the test was on-focus. It was discovered that many small-field cells were inhibited or facilitated by wide-field interactions and a few showed mixed effects. The proportions of neurons effected was different at each of the usual four sampling sites but about half of the small-field cells were modulated. Whether these cells were also small-field PYR neurons is not known. Thus, they were a subset of small-field neurons not subthreshold wide-field cells [Satterthwaite, Burnham & Towe, '78]. This, along with the small to wide-field influence noted above, implied a reciprocal interaction between some small- and wide-field cells.
One of the many states that the neuronal circuit model had to account for was the shift to a convulsive condition. The effects of various drugs, including strychnine [Mann & Towe, '74] , bicuculine [Harris & Towe, '76] and (paradoxically) dilantin [Harris & Towe, '78], were tested by topical application and iontophoretic injection as models of convulsive behavior in neurons. Due to the large size of these molecules (thus, a slow diffusion rate) and the rapid development of the convulsive effect, it was concluded from the topical application results that the response was generated in layers I and perhaps upper II of the cortex. Dilantin was being tested as a possible magic bullet that would turn off small-field cells. It didn't. Instead, its convulsive properties were discovered.
Work by others suggested that there were two groups of pyramidal tract neurons - fast and slow Pt's [Takihashi, 65]. We had observed that many wide-field and a few small-field cells were PYR antidromically activated and one group of wide-field cells was located in deep layer III and another in layer V. This raised the question; Was one group fast and the other group slow? This was investigated with direct PYR and cortical stimulation and with cuts through various layers of the cortex [Chen & Towe, '85]. The answer was no and the results from CFP stimulation were similar to what had been seen before.
The discussion so far has dealt with single neuron behavior. Clearly, these responses produced the extracellular evoked potential observed at the surface of cortex as well as within its depth. A model was constructed which described this connection. It was shown that the vertical component of extracellular current flow calculated from the sample of small-field cells matched that calculated from the evoked potentials recorded at 100 micron increments [Towe, '66]. This opened up the possibility that the neuronal circuit model could be investigated from the standpoint of evoked potentials rather than the extremely time consuming population samples of single neurons.
The current flow model was extended to current-source-density analysis which predicted the location and timing of synaptic current sources and sinks. The properties of topically applied strychnine were examined in this way and the results agreed very well with those predicted by the neuronal circuit model [Towe, Mann & Harding, '81].
Through all the years of this work, the results were not consistent with the main tenant of the columnar hypothesis - modality purity [Mountcastle, '57; Mountcastle, Davies & Burman, '57; Mountcastle, '79]. To examine the viability of this hypothesis, samples were taken with a rectilinear array of closely spaced (0.26 mm) tracks in awake cats. If any modality pure columnar organization was present, then this would have been revealed in the data. Arrays of tracks were done in areas 3b and 4r. Cells nonresponsive to field testing were located because they were spontaneously active. Simultaneous observation of 2 to 4 neurons was common. The data showed that the cortex was not columnar in the modality pure sense [Slimp & Towe, '90]. The pattern of modalities and field sizes was chaotic. The random modality distributions were present with a two category (surface and deep) classification system as well as with a more specific eight class one.
It has been said that, in the nervous system, everything is connected to everything else. The sigmoid gyrus of the cat is no exception. A11 of the nervous system structures that it might be in communication with (either directly or indirectly) would be hard to determine. However, the direct CFP pathway from the periphery includes two major nuclei - the cuneate nucleus in the brainstem and the VPL nucleus of the thalamus.
Several studies were conducted to examine the properties of neurons in the VPL. Samples from nearly 1600 cells were made in the Chloralose, Nembutal and awake conditions. Both small- and wide-field neurons were found in proportions similar to those in the sigmoid gyrus under the same anesthetic states. A study conducted in the laboratory next door to ours found similar small- and wide-field cell distribution differences in the lateral reticular nucleus under Nembutal and Chloralose anesthesia [Crichlow and Kennedy, '67] to those seen in the cortex and thalamus.
Other studies were done to look at the makeup of the cuneate nucleus [Jabbur & Towe, '61; Towe & Jabbur, '61; Blum et al. '75; Bromberg et al. '75; Bromberg et al., '81; Surmeier, '83; Weinberg, '83]. The major study among these used a regular 0.5 mm grid sampling paradigm to cover the whole nucleus. Nearly 1800 neurons and fibers were sampled with the descending influence of cortex and thalamus stimulation tested [Bromberg, Blum & Whitehorn, '75]. Specific questions regarding the behavior of several types of cells were followed up as well [Jabbur & Towe, '61; Towe & Jabbur, '61; Morse, unpublished; Bromberg et al., '81; Surmeier, '83; Weinberg, '83]. A total of 3200 cuneate neurons was studied over a period of 25 years.
Though not directly related to the somatosensory system, some studies were conducted elsewhere in the cat's nervous system. These included 257 cells in the cerebellum [Talbot, '66] and 171 fibers in the cat's tail [Reid, '68]. Thus, in addition to the 17,500 neurons studied in the sigmoid gyrus, 5200 cells (and fibers) were sampled in the cuneate, thalamus and other sites in the cat.
Late at night, while piloting microelectrodes through the murky cortical waters, we often recited poems from our favorite volume. This collection of nursery rhymes was written by a nuclear physicist for his children, but it turned out to be much more about the conduct of science. Here is an example:
Hey Diddle Diddle
Distribute the Middle
The Premise Controls the Conclusion
The Disjunctive Affirms
That the Diet of Worms
Is a Borbetomagic Confusion
[Winsor & Parry, '58]
Although data collection wound down in the mid 1980's, Arnie and colleagues continued to try to publish their findings through the 1990's. The paradigm shift from "systems" to "molecular" research made this difficult and frustrating. The columnar hypothesis has always been controversial [e.g., Towe, '75; Jones & Wise, '77; Mountcastle '79 & '97; Dykes & Gabor, '81; White, '89], but we found now that scientific debate had died; a situation that Arnie often lamented over the past decade. A number of manuscripts have remained unpublished because they were "fatally flawed" according to the reviewers (but they did not indicate what the flaw was). His colleague, friend and superb scientist J. Walter Woodbury put the predicament this way: "Every scientist deserves the opportunity to hang himself in print, provided that he does so with clarity" (personal communication to ALT, circa. 1960).
We miss you Arnie, we miss you terribly. Our thoughts are also with Arnie's wife, companion and friend, Laurie.
Adkins, R. J., R. W. Morse and A. L. Towe. 1966. Control of somatosensory input by cerebral cortex. Science 153:1020-1022.
Amassian, V. E. 1953. Evoked singe cortical unit activity in the somatic sensory area. Electroenceph. clin. Neurophysiol. 5:415-438.
Branscomb, L. M. 1985. Integrity in science. Am. Sci. 73:421-423.
Chen, Z. and A. L. Towe. 1985. Influence of molecular layer on pyramidal tract neurons. Exp. Neurol. 88:215-228.
Chen, Z., G. W. Harding and A. L. Towe. 1983. Effect of strychnine on the cutaneous responsiveness of wide-field cerebral neurons after depression by pentobarbital. Exp. Neurol. 81:770-775.
Crichlow, E. C. and T. T. Kennedy., 1967. Functional characteristics of neurons in the lateral reticular nucleus with reference to localized cerebellar potentials. Exp. Neurol. 18:141-153.
Doetsch, G. S. and A. L. Towe. 1976. Response properties of distinct neuronal subsets in hindlimb sensorimotor cerebral cortex of the domestic cat. Exp. Neurol. 53:520-547.
Dykes, R. W. and A. Gabor. 1981. Magnification functions and receptive field sequences for submodality-specific bands in SI cortex of cats. J. comp. Neurol. 202:597-620.
Ennever. J. A. and A. L. Towe. 1974. Response of somatosensory cerebral neurons to stimulation of dorsal and dorsolateral spinal funiculi. Exp. Neurol. 43:124-142.
Gehary, Y. and A. L. Towe. 1993. Effect of optic nerve stimulation on neurons in pericruciate cortex of cats. Exp. Brain Res. 94:272-278.
Harding, G. W. and A. L. Towe. 1976. An on-line real-time laboratory for single neuron studies. Comput. Biomed. Res. 9:471-501.
Harding, G. W., R. M. Stogsdill and A. L. Towe. 1979. Relative effects of pentobarbital and chloralose on the responsiveness of neurons in sensorimotor cerebral cortex of the domestic cat. Neuroscience 4:369-378.
Harris, F. A. and A. L. Towe. 1976. Effects of topical bicuculline on primary evoked responses in pericruciate and precoronal cortex of the domestic cat. Exp. Neurol. 52:227-241.
Harris, F. A. and A. L. Towe. 1978. Effects of topical application of diphenylhydantoin on gross evoked responses and single-neuron activity in pericruciate and precoronal cerebral cortex of the domestic cat. Exp. Neurol. 62:521-538.
Harrison T. A. and A. L. Towe. 1986. Antidromic response to medulary pyramid stimulation in rats and its relation to that in cats. Brain Behav. Evol. 29:143-161.
Hassler, R. and K. Muhs-Clement. 1964. Architektonischer aufbau des sensomotorischen und parietalen cortex der katze. J. Hirnforsch. (Heft 6) 6: 377-420.
Jones, E. G. and S. P. Wise. 1977. Size, laminar and columnar distribution of efferent cells in the sensory-motor cortex of monkeys. J. Comp. Neurol. 175: 391-438.
Kennedy, T. T. and A. L. Towe. 1962. Identification of a fast lamnisco-cortical system in the cat. J. Physiol. (London) 160:535-547.
Mann, M. D. 1979. Sets of neurons in somatic cerebral cortex of the cat and their ontogeny. Brain Res. Rev. 1:3-45.
Mann, M. D., W. S. Holt and A. L. Towe. 1977. Afferent modulation of the excitability of pyramidal tract fibers. Exp. Neurol. 55:414-435.
Mann, M. D., and A. L. Towe. 1974. Effect of strychnine on single neurons of the pericruciate cerebral cortex. Exp. Neurol. 42:388-411.
Morse, R. W. and R. A. Vargo. 1970. Functional neuronal subsets in the forepaw focus of somatosensory area II of the cat. Exp. Neurol. 27:125-138.
Morse, R. W., R. J. Adkins and A. L. Towe. 1965. Population and modality characteristics of neurons in the coronal region of somatosensory area I of the cat. Exp. Neurol. 11:419-440.
Mountcastle, V. B. 1957. Modality and topographic properties of single neurons of the cat's somatic sensory cortex. J. Neurophysiol. 20:408-434.
Mountcastle, V. B. 1979. An organizing principal for cerebral function: The unit module and distributed system. In "The Neurosciences. Fourth Study Program" F. 0. Schmidt and F. G. Worden, (eds.) MIT Press. Cambridge pp21-42.
Mountcastle, V. B. 1997. The columnar organization of the neocortex. Brain 120:701-722.
Mountcastle, V. B., P. V. Davies, and A. L. Berman. 1957. Response properties of neurons of cat's somatic sensory cortex to peripheral stimuli. J. Neurophysiol. 20:374-407.
Nyquist, J. K. and A. L. Towe. 1970. Neuronal activity evoked in cat precruciate cerebral cortex by cutaneous stimulation. Exp. Neurol. 29:494-512.
Patton, H. D., A. L. Towe, and T. T. Kennedy. 1962. Activation of pyramidal tract neurons by ipsilateral cutaneous stimuli. J. Neurophysiol. 25:501-514.
Ramon-Moliner, E. 1961. The histology of the postcruciate gyrus in the cat. I. Quantitative studies. J. comp. Neurol. 117:43-62.
Satterthwaite, W. R. 1976. Feline pericruciate cerebral neurons activated by electrical stimulation of the ventral medulla. PhD Dissertation, University of Washington. 251pp.
Satterthwaite, W. R., J. A. Burnham, and A. L. Towe. 1978. Wide-field conditioning effects on small-field neurons in the posterior sigmoid gyrus of domestic cats. Exp. Neurol. 60:603-613.
Slimp, J. C. and A. L. Towe. 1977. Characteristics of somatic receptive fields of neurons in postcruciate cerebral cortex in awake-restrained and two anesthetic conditions in the same cat. Neurosci. Abs. 2:492.
Slimp, J. C. and A. L. Towe. 1980. Effects of pudendal nerve stimulation on neurons in pericruciate cerebral cortex of male domestic cats. Exp. Neurol. 67:181-204
Slimp, J. C. and A. L. Towe. 1990. Spatial distribution of modalities and receptive fields in sensorimotor cortex of awake cats. Exp. Neurol. 107:78-96.
Takahashi, K. 1965. Slow and fast groups of pyramidal tract cells and their respective membrane properties. J. Neurophysiol. 28:908-924.
Towe, A. L. 1965. Neuron population analysis in the cerebral cortex, pp. 143-156. In P. W. Nye (ed.). "Proceedings, Symposium on Information Processing in Sight -Sensory systems'". California Institute of Technology, Pasadena, California.
Towe, A. L. 1966. On the nature of the primary evoked response. Exp. Neurol. 15:113-139.
Towe, A. L. 1968. Neuronal population behavior in the somatosensory systems. pp. 552--574. In D. R.- Kenshalo (ed.) "The Skin Senses". Thomas, Springfield, Illinois.
Towe, A. L. 1975. Notes on the hypothesis of columnar organization in somatosensory cerebral cortex. Brain Behav. Evol. 11:16-47.
Towe, A. L. and G. W. Harding. 1970. Extracellular microelectrode sampling bias. Exp. Neurol. 29:366-381.
Towe, A. L. and T. T. Kennedy. 1961. Response of cortical neurons to variation of stimulus intensity and locus. Exp. Neurol. 3:570-587.
Towe, A. L. and R. W. Morse. 1962. Dependence of the response characteristics of somatosensory neurons on the form of their afferent input. Exp. Neurol. 6:407-425.
Towe, A. L., M. D. Mann, and G. W. Harding. 1981. On the currents that flow during the strychnine spike. Elecroenceph. clin. Neurophysiol. 51:306-327.
Towe, A. L., H. D. Patton, and T. T. Kennedy. 1963. Properties of the pyramidal system in the cat. Exp. Neurol. 8:220-238.
Towe, A. L., H. D. Patton, and T. T. Kennedy. 1964. Response properties of neurons in the pericruciate cortex of the cat following electrical stimulation of the appendages. Exp. Neurol. 10:325-344.
Towe, A. L. and J. C. Slimp. 1977. Organization of postcruciate neurons with respect to somatic modality and receptive field in a awake-restrained cat. Neurosci. Abs. 3:493.
Towe, A. L., C. F. Tyner, and J. K. Nyquist. 1976. Facilitatory an inhibitory modulation of wide-field neuron activity in postcruciate cerebral cortex of the domestic cat. Exp. Neurol. 50:734-756.
Towe, A. L., D. Whitehorn, and J. K. Nyquist. 1968. Differential activity among wide-field neurons of the cat postcruciate cerebral cortex. Exp. Neurol. 20:497-521.
Tyner, C. F. 1975. The naming of neurons: Application of taxonomic theory to the study of cellular populations. Brain Behav. Evol. 12:75-96.
Tyner, C. F. 1979. Splanchnic nerve activation of single cells in the cat's postcruciate motorsensory cortex. Exp. Neurol. 63:76-93.
Tyner, C. F. and M. G. Miller. 1977. Selective inhibition of some wide-field sensorimotor cortex neurons by high intensity skin stimuli. Neurosci. Abs. 3:72.
Tyner, C. F. and A. L. Towe. 1970. Interhemispheric influences on sensorimotor neurons. Exp. Neurol. 28:88-105.
Videen, T. 0. 1981. Visual input to small-field and wide-field neurons in the postcruciate cortex of domestic cat. Exp. Neurol. 71:341-355.
von Economo, C. F: 1929. "The Cytoarchitectonics of the Human Cerebral Cortex". Oxford Medical Publications, London, England. 186 pp.
White, E. L. 1989. "Cortical Circuits: Synaptic Organization of the Cerebral Cortex; Structure, Function, and Theory" Birkhauser, Boston. 223pp.
Whitehorn, D. and A. L. Towe. 1968. Postsynaptic potential patterns evoked upon cells in sensorimotor cortex of cat by stimulation at the periphery. Exp. Neurol. 22:222-242.
Winsor, F. and M. Parry. 1958. "The Space Child's Mother Goose." Simon and Schuster, New York, New York.
Woolsey, C. N. 1958. Organization of somatic sensory and motor areas of the cerebral cortex. pp. 63-82. In: H. F. Harlow and C. N. Woolsey (eds.) "Biological and Biochemical Basies of Behavior". University of Wisconsin, Madison.
Woolsey, C. N. and 0. Fairman. 1946. Contralateral, ipsilateral, and bilateral representation of cutaneous receptors in somatic areas I and II of the cerebral cortex of pig, sheep, and other animals. Surgery 19:684-702.