it is possible to predict from this study lution.

By contrast, solvent treatment of These results suggest that unsaturation

that single-kernel differences would be dried seeds altered the mobility of the in the whole kernel can be measured in
heritable and that breeding programs, oil sufficiently to provide very good times as short as 10 seconds.
such as mass selection and single re- signal-to-noise ratios and adequate T. F. CONWAY
current selection which make use of resolution to separate the olefinic and George M. Moffett Research
high heritability, would probably be glyceride methylene protons (Fig. 1C). Laboratories,, Corn Products Company,
effective for modification of corn oil Proper integration of spectra from single Argo, Illinois
quality. Consequently, a high-resolution seeds permits determination of iodine L. F. JOHNSON
NMR selection technique for unsatura- value and average molecular weight Instrument Division, Varian
tion may prove as advantageous as the by means of the Johnson-Shoolery (8) Associates, Palo Alto, California
wide-line NMR single-kernel technique calculation. These data indicate that
in speeding the development of new References ad Notes
sample drying and solvent treatment
corn strains. are necessary to obtain usable spectra. 1. T. F. Conway and R. J. Smith, in Develop-
ments in Applied Spectroscopy, J. R. Ferraro
In 1962 Johnson and Shoolery (8) Variations in NMR spectra resolution and J. S. Ziomek, Eds. (Plenum Press, New
described a high-resolution NMR meth- have been observed between seeds York, 1963), p. 115; D. E. Alexander, L.
Silvela S., F. I. Collins, R. C. Rogers, J.
od for determining the average molecu- which are attributed to differences in Amer. Oil Chem. Soc. 44, 555 (1967).
2. 1. Zupan6i6 et al., Acta Chem. Scand. 21, 1664
lar weight and iodine value of natural magnetic susceptibilities between the (1967).
fats dissolved in carbon tetrachloride. oil and the corn germ matrix as well 3. L. F. Bauman, T. F. Conway, S. A. Watson,
Science 139, 498 (1963); Crop Science 5, 137
With this method the NMR integral as between the seed and surrounding (1965).
signal from C(1) and C(3) glyceride solvent. Minimizing these should im- 4. J. Dumanovic and V. Trifunovic, Contemp.
Agr. (Novi Sad) 748, 521 (1964).
protons provides an internal standard prove spectra resolution. 5. L. Silvela S., "Improvement in recurrent se-
from which the olefinic and total num- The spectra shown in Fig. 1 were ob- lection methods for oil in maize through nu-
clear magnetic resonance analysis," thesis,
ber of hydrogen atoms are measured. tained with 250-second sweep time. University of Illinois, Urbana (1965).
From these, an average molecular Through use of NMR radio-frequency 6. B. BrimhaU and G. F. Sprague, Cereal Chem.
28, 225 (1951).
weight and iodine value are calculated. pulse techniques and Fourier transfor- 7. C. G. Poneleit, "Diallel analysis of fatty acids
Agreement between NMR and Wijs mation of the data (10), it is possible in corn (Zea mays L.) oil," thesis, Purdue
University, Lafayette, Ind. (1968).
iodine values is remarkably good. to greatly reduce instrument time. 8. L. F. Johnson and J. N. Shoolery, Anal.
To demonstrate viability under the Studies with the Varian FT-100 Fourier Chem. 34, 1136 (1962).
9. A. L. Elder, D. M. Rathmann, T. F. Conway,
conditions of our method, we dried transform accessory (Fig. 2) show a Quial. Plant. Mat. Veg. 15, 67 (1964).
whole kernels of 25 commercial corn 10-second examination of an excised 10. R. R. Ernst, in Advances In Magnetic Reso-
nance, J. S. Waugh, Ed. (Academic Press,
hybrids to 4 percent moisture and sub- corn germ which was fitted into a 5-mm New York, 1966), vol. 2, p. 108.
jected them to 95°C for periods up to tube and surrounded by Freon-1 13. 9 January 1969
90 minutes. Others were soaked in
Freon-113 for periods up to 6 days.
Subsequent germination tests showed a
high degree of tolerance to low moisture
heating (94 percent survival) and to the Simultaneously Recorded Trains of Action Potentials:
solvent treatment (98 percent survival). Analysis and Functional Interpretation
For NMR experiments proposed
earlier (9), individual corn kernels or Abstract. A new kind of statistical display, the joint peri-stimulus-time scatter
excised germ were mounted on a Teflon- diagram, facilitates the analysis and interpretation of two or more simultaneously
glass pedestal, transferred to 12-mm recorded trains of action potentials. The display is a generalization of the cross
tubes and covered with Freon-113 to correlation and the peri-stimulus-time histogram, and it reflects specific under-
minimize the large discontinuities in lying neuronal interactions. The technique yields quantitative measures of inter-
magnetic susceptibilities within the action in terms of effectiveness of synaptic connections.
sample cell. Samples were examined
with a Varian model HA-100-15 high- Much of the known picture of the method of performing such analyses
resolution spectrometer equipped with nervous system has been obtained from for experiments in which a repeated
a wide gap magnet. Typical spectra single neurons individually studied for stimulus is presented to a system of
from kernels of commercial-dried (15 an extended period of time. On the neurons and in which two or more
percent corn moisture), laboratory- other hand, anatomical, physiological, spike trains are simultaneously moni-
dried (4 percent corn moisture), and and behavioral evidence overwhelm- tored. This method is more powerful
laboratory-dried corn soaked in Freon- ingly suggest that integrative functions and sensitive than that suggested by
113 for 6 days are shown in Fig. 1 with in the nervous system are performed Perkel et al. (2, p. 429) and permits a
NMR chemical shift assignments. The by simultaneous activity in groups of relatively direct functional interpreta-
spectrum from the commercial-dried neurons. It has become possible to re- tion.
kernel (Fig. lA) was not usable because cord several trains of action potentials Our basic data consist of the times
of the broad absorption signal. The (spike trains) simultaneously and to of occurrence of three sets of events:
laboratory-dried kernel (Fig. 1B) ex- analyze their temporal relationships. a train of periodic (identical) stimuli,
hibited a poor NMR spectrum; resolu- The results of these analyses may, in and two trains of action potentials (A
tion and signal-to-noise ratio were less turn, be interpreted in terms of con- and B). Let the time of occurrence of
than satisfactory. Surprisingly, heating nections which underlie the integrative the ith stimulus be Si; that of the jth
the dry sample up to 950C to alter the process in the group of neurons under spike of train A be Aj; and that of the
mobility of the fat did not improve reso- observation (1, 2). We describe here a kth spike of train B be Bk. We con-
828 SCIENCE, VOL. 164
struct a scatter diagram of the joint tion; the width and structure of the A final principal class of interaction
occurrences of spikes in trains A and B band depend on the details of the ex- is represented by a shared source of
relative to the times of stimulation. citatory mechanism. If the stimulus input to A and B which is independent
Thus, the ordinate of each point plotted produces a fixed latency decrease in of the stimulus. This gives rise to a di-
corresponds to the time between a stim- firing probability of neuron A (for ex- agonal band which is difficult to distin-
ulus event and a spike in A (for exam- ample, due to synaptic inhibition) there guish qualitatively from the diagonal
ple, A - S1), and the abscissa corre- will be a horizontal band of lower point band produced by a direct excitatory
sponds to the time between the same density than the general background pathway from A to B or from B to A.
stimulus event and a spike in B (for density in the scatter diagram. If the However, for a given strength of syn-
example, Bk- S). For each stimulus stimulus affects neuron B and not neu- apse, shared excitatory input produces
event, a point is plotted for each com- ron A, then only a corresponding set a much less dense and a wider band
bination of S-A and S-B intervals, both of vertical bands will be produced. than a direct synaptic connection be-
of which fall within a specified range. A direct synaptic pathway from neu- tween the two neurons. The density
Thus, in general, a spike in train A will ron A to neuron B will produce a 450 difference and other structural details
give rise to as many points in the scatter band of different density lying below of the band usually enable a choice to
diagram as there are spikes in train B the principal diagonal at a distance pro- be made between the alternatives (2, p.
within the specified time range about portional to the latency of the interac- 431). If such a source of shared input is
the stimulus (and vice versa). This tion. A synaptic pathway from neuron itself modified in a time-locked way by
computation and plot have been done B to neuron A will produce a similar the stimulus, these modifications will be
with a Linc computer as well as with band lying above the principal diag- reflected in density changes along the
considerably larger machines. onal. diagonal band.
We call this plot the joint peri-stim- If the synaptic pathways between A If combinations of these elementary
ulus-time (PST) scatter diagram (3) and B can be affected by the stimulus influences occur, each of the corre-
(Figs. 1 and 2). In fact, the scatter (as by stimulus influence on an inter- sponding bands will ordinarily appear
diagram represents an estimate of an neuron), then the diagonal band will in the scatter diagram (Figs. 1 and 2).
underlying joint PST density; the two show changes in density along its length. The scatter diagram can yield quanti-
PST histograms estimate a related set For example, if the stimulus inhibits an tative estimates of synaptic effective-
of marginal densities (4). excitatory interneuron between A and ness in the inferred connections be-
The scatter diagram represents a B, the diagonal band will be reduced in tween neurons if the points within
generalization of the PST histograms density over a segment corresponding particular bands and band intersections
of each of the two spike trains, and of to the latency and duration of the in- are counted.
the cross-correlation histogram. If we hibition (Fig. 1). We may estimate the contributions
impose a square grid of convenient
spacing over the scatter diagram, then
the set of tallies by columns is related 0.5 sec 0.5 sec
to the PST histogram of train B, and
the corresponding set of tallies by rows
I
is related to the PST histogram of train
A. Furthermore, tallies along the direc-
tion parallel to the principal diagonal (a) A*

are proportional to the cross-correlation

histogram, with a suitable normalization
of the time axis. The scatter diagram
contains information beyond that pro-
vided by the two PST and the cross- - 0.5 sec 0.5 sec
correlation histograms and cannot be
reconstructed from these histograms 0.5 sec 0.3 sec
alone, no matter how fine the resolution.
Each of the several possible types of
functional connection to and between
the observed neurons produces a char-
acteristic signature in the scatter dia- (c)
gram. For example, two independently -. .:
; 0...
firing neurons which are totally un-
affected by the stimulus will produce
a uniform scatter of points. If, in addi-
tion, the stimulus produces an increase
in firing probability at fixed latency in
I - 0.5 sec
I 3 sec
Fig. 1. Joint PST scatter diagrams in computer-simulated neural networks. (a) Neurons
neuron A only (such as would be pro- 1 and 2 with common excitatory followed by inhibitory input from a periodic stimulus.
Duration of the stimulus is indicated by dark bars. Note vertical and horizontal bands
duced by an excitatory synaptic path- of higher density. (b) Same situation as (a) but with added excitatory interaction from
way), then a horizontal band of higher neuron 1 to neuron 2. Note diagonal band of higher density. (c) Same situation as
point density will be seen in the scatter (b) except that the excitatory interaction from neuron 1 to neuron 2 is itself inhibited
diagram. The location of the band is by the stimulus. In contrast to (b) the diagonal band is interrupted during the latter
portions of stimulus. (d) Enlargement of lower left region of (c). Details of each band
a measure of the latency of the excita- structure allow inference about various synaptic input sequences.
16 MAY 1969
829
of only the direct stimulus effects to stimulus train, each spike train can be ate histogram. Examination of the
the scatter diagram by constructing the assigned its own Cartesian coordinate. histograms can serve as a guide to the
cross product of the marginal proba- The resulting N-dimensional scatter detailed interpretation of particular fea-
bility densities (obtained from the two diagram, although presenting some dif- tures in the scatter diagram. This com-
PST histograms). The resulting bivari- ficulties in display, can be interpreted in bined approach should be particularly
ate density, which contains no diagonal a completely analogous manner to that applicable, for example, to experiments
bands, serves as a control. It represents of the two-dimensional diagrams. For which seek to associate plasticity of
the hypothetical case in which the two example, a concentration of points behavior with changes in neuronal in-
neurons fire independently except for about a line parallel to the principal teractions.
shared input caused by the stimulus. diagonal of the coordinate system cor- GEORGE L. GERSTEIN
Comparison of this control with the responds to a set of interactions, or Departments of Biophysics and
observed joint density facilitates the shared input unrelated to the stimulus, Physiology, University of Pennsylvania,
detection of direct interactions and involving all of the monitored neurons. Philadelphia 19104
shared input from sources other than The scatter diagram described here DONALD H. PERKEL
the stimulus. is a supplement to rather than a substi- Department of Mathematics,
The method described can be extend- tute for the PST and cross-correlation RAND Corporation, Santa Monica,
ed as follows: When more than two histograms. Weak interactions may be California 90406
neurons are monitored together with a more readily detected in the appropri-
References and Notes
1. G. P. Moore, D. H. Perkel, J. P. Segundo,
0.5 sec 0.2 Ann. Rev. Physiol. 28, 493 (1966).
2. D. H. Perkel, G. L. Gerstein, G. P. Moore,
Biophys. J. 7, 419 (1967).
3. We propose "peri-stimulus-time" instead of
the older restrictive term "post-stimulus-time."
The initials PST are thereby preserved.
4. The joint PST density may be defined by the
equation
(a) (b)
Gab (t, U) = lim P
At-O At AU
U*0
where P equals probability [spike in A in
3," (v + t, v + t + At) and spike in B in
(v + u, v + u + Au), given the stimulus
event at v]
I - 5. C. C. Bell and R. J. Grimm, personal comn-
0.5 sec 0.2 sec munication.
6. Supported by PHS grants NB 05606, NB 07325,
FR-15, and by United States Air Force Project
RAND.
0.5 sec . . 0.5 e sec 6 February 1969
tWAS

(c) I'J c-
s
.- (d) Scanning Electron Microscopy
of Cells
Abstract. The scanning electron
energy-analyzing microscope uses char-
acteristic energy losses to provide pic-
ture contrast. At different levels of
I 0.5 sec energy loss particular structures are
-~0.5 sec distinguished with high contrast in an
Fig. 2. Joint PST scatter diagrams. Units in auditory cortex of unanesthetized cat
were 20-msec noise bursts, two per second, and are indicated by the dark bmars. (a)
trStimuli unstained section of a cell.
Units 1 and 3: strong vertical and horizontal bands represent the stimuluis-caused
activity of each unit individually. The different latencies, time courses, and fil ne struc- Electron microscopy of ultrathin
tures of the two responses are apparent. The narrow diagonal band represents a direct sections of biological tissue generally
excitatory interaction between the two neurons with a latency of 3 msec. TIrhe wide requires electron-opaque staining to
diagonal band of increased density in central and upper right portion is most Iplausibly obtain suitable contrast for detailed
attributed to a weak shared excitatory input. The absence of this wide diagonal
the lower left portion of scatter diagram indicates inhibition of the inferred common analysis. A heavy atom that selectively
source for some time after the stimulus. (b) Enlargement of lower left portior of (a); attaches to certain cell components
differences in vertical and horizontal structure of region of high density and variation (such as nuclei, mitochondria, and
of density along the thin diagonal band, presumably reflecting stimulus-related rreduction lysosomes) is typically used to stain the
of firing, are shown. (c) Units 2 and 3, same preparation. The wide diagoi ces. The specimen. The contrast provided by
again indicates that the stimulus inhibited shared input from nonstimulus sour such staining for examination with a
shared input reaches unit 3 after unit 2, in that the wide diagonal band lies atbove the
principal diagonal. There is no evidence for a direct synaptic connection. I(d) Two conventional electron microscope de-
-Purkinje cells in the vermis of anesthetized cat cerebellum during an auditory stimula- pends primarily on specimen density.
tion by clicks (one click per second). There are no significant vertical or h oraizontal An ideal mechanism for obtaining
features, indicating that these units did not respond to the stimulus. Multiple tdiwithal contrast would be one in which dif-
bands are visible and presumably represent shared input and time delays associi
parallel-fiber input (2). ferent structures could be shown suc-
830 SCIENCE, VOL. 164