April 3, 1955

Dr. John von Neumann
Institute for Advanced Study
Princeton, N.d.

Dear Dr. von Neumann:

Thank you for your letter of March 14, referring me to the Hixon
Symposium. By a curious coincidence, I had just then "apontanesously"
stumbled on the book, and have had some chance to read your article
in the interval. I also note the account by John Kemeny in the last
mamber of the Selentific American.

Under separate cover, I did send an article of my own, "Cell Genetics
and Hereditary Symbiosis" which may serve rather to illustrate perplexity
than to illuminate concept.

In your treatment, I am particularly impressed by the way in which
one can evade the notion of a "self-reproducing particle", for you empha-
size that it is the entire assmmbly alone that has that property. In dif-
ferent lahguage, I have been groping for the same inference, simply on
the basis that genes, or even nuclei, are ineapable of producing anything,
mach less coples of themselves, when isolated from the whole machine. But
there are still some diffheulties, for tha geneticist would still like
to abstract, from the entire organism, the least structire that will still
perpetuate the genetic function. The non-germinal or somatic elements of
higher organiams are generally more conspicuous than the germ, but even
within the scope of a single "macromolecule", a similar differentiation
ean be seen, for example in the way in which terminal threonine residues
have been split from tobacco mosddc virus without impadring the ability
of the partiéle to engender further generations of typical virus. You
have indicated an analogy between the genes and the "information tape",
but I would be interested to know the explicit criteria by which to tell
how an intracellular organelle corresponds to one or wre of the elemnts
of your assembly. I shall, in fact, be surprised if your conceptual analysis
has a structural representat&on, or if this was intended by yourself.

As you need hardly be told, my own thoughts on this subject are still
amorphous. I am still trying to see what can be salvaged of the notion of
a "self~reproducing particle",if anything. I am more concerned how a sys-
tem such as you postulate can have evolved, and am therefore still interested
in more strictly autocatalytic processes, which may be useful in preliminary
model building. I am hopeful that, once our ideas are more precisely developed,
it may already be technically possible to build chemical models which may
exemplify som: reproductive processes.
In your article, you hint that a dozen kinds of elementary parts
would suffice for a s-r machine, but I am afraid I still do not underatand
what you mean by a part, and would be grateful for a clarification, and
for som notion how your inventory is derived.

In reviewing the whole article, I also wondered if your concluding sec-
tion did not contradict the suggestion that nature has not relied on digital
coding: the linear chromosome must be one of the most elegantlgxepded sequences,
having baffled even Gamow's cryptography. Perhaps you were referring more
narrowly to neural mechanisms: the main point is obviously that a digital
code needs a detailed structure on which the "tape” can be oriented, either
temporally or spatially. And we have perhaps some hints of this in (some
versions of) the acoustic sensory mechanism, though you are perhaps excluding
codes other than those involving a single tranmaigssion line. For these, the
greatest problem may be that of orientation and registration, which would
haw to be temporally regulated. The problem of keeping the CNS in phase with
the external receptops may be biologieally insuperable in view of such pertur—
bations as result, for example, from tamperature variation, whose consequences
are, for the reasons you outline, far more disastrous the less the redundancy.
Coding on a geometric (as opposed to temporal) scaffold is far safer, and not

infrequently (indeed, in a mjgumk macromolecular chemical context, invariably)
found.

Yours sincerely,

Joshua Lederberg
Professor of Genetics