The problem that I want to discuss can best be introduced with
the help of an illustration. Let it be supposed that an anti-aircraft
gunner equipped with radar is watching for hostile planes. The
approach or presence of one in the sky is revealed to him by a break
in a line on the screen of a cathode ray oscilloscope that he is
watching. When he sees the break he presses a button and releases
a volley of fire. Here is a chain of successive connected events starting with the aeroplane and passing by cause and effect through the
radar to the operator and on to the guns. At some link along this
chain an act of observation is performed. To locate this act is the
Between the oscilloscope and the pushbutton is the gunner - the
operator. One can apply the term "observer" in a general way to
him; but there are many links in the chain between the eye that
watches the oscilloscope and the finger that presses the button.
There are the afferent nerves that bring messages from the eye to
the brain, a multitude of synapses in the brain itself, efferent nerves
that carry controlling impulses from the brain to the operator's
finger. To avoid vagueness one must discover exactly at which of
the many links the act of observation occurs.
Some of these links can be ruled out as surely as those that are
situated outside the operator's body. What happens on his efferent
side is caused by the observation; it cannot be the observation itself,
for the function of the efferent nerves is to act as instruments of
control; like the operator's finger, the pushbutton, or the gun mechanism. And what happens on the afferent side is not observation
but the cause of the observation. The mechanisms on that side are
instruments of observation, from the oscilloscope right up to and
including the afferent nerves themselves. It is true that one sometimes speaks of "the observing eye" and implies thereby that the
eye is not an instrument of observation but the observer itself. How-
ever, to speak thus is to use language loosely - poetically. The
more precise term is "the observer's eye".
Considerations of function alone therefore lead to the conclusion,
with which few physiologists would disagree, that the act of observation must occur between the afferent and the efferent systems,
between the last link in the chain where the transmission of a received signal ends and the first link where the transmission of a signal
of control begins. The problem is to find precisely what the link
is at which the significant performance known as observation occurs.
Two hypotheses have been put forward and both need to be
considered carefully; the more so as grave objections can be urged
against each of them.
The act of observation is thus seen to be a twofold one, consisting both in the passive reception of a pattern of signals and in the
active interpretation of that pattern. Presented with a set of impulses the observer makes a correct inference. Be the observer a
material brain or a non-material mind the pattern presented for
observation is fundamentally different from the pattern that is inferred. Whether or not an act of interpretation is performed is not
in dispute. Monists and dualists have to agree that there is interpretation. Disagreement occurs only as to what does the interpreting. 1
According to the monist hypothesis, also, the brain is unique;
but for a different reason. While other devices along the chain
between the aeroplane and the guns are regarded as serving the
comparatively simple functions of instruments, either of observation
or of control, the brain is regarded as performing the additional
function of observer. It is therefore also regarded as an interpreter
of patterns. At the junction between the afferent and the efferent
sides, and nowhere else, it is assumed that there is a device quite
unlike all the others. This is supposed to receive a certain pattern
of signals, to interpret it as representing the observed object, to appreciate its significance, and to send out an appropriate pattern of
signals of control to the efferent system. No functions remotely like
these are supposed to be performed by the oscilloscope, or the gun-firing mechanism, or the eye, or the finger. Dualists and monists
may agree that there are many links where a translation of patterns
of signals into other patterns occur; and they must also agree that
there is only one link at which an act basically different from translation, an act of interpretation, occurs.
Defining the Area of Disagreement
A strict analysis of an act of observation thus reveals that the
area of disagreement between the two schools can be defined with
great precision. The dispute turns on the simple question: What
acts as interpreter, brain or mind? But if the analysis simplifies
the form of the question it does not lead to an easy answer. When
the problem is expressed in terms of concrete realities objections to
each of the two possible answers do not vanish but, on the contrary,
become more apparent. This may explain the unwillingness that
thinkers sometimes show to facing a philosophical problem in terms
of concrete realities.
The question what does the interpreting? is a big one, and
the answer cannot be attempted in a short paper. I must be content
only to mention that much could be said both for and against each
of the two possible answers; and it is perhaps not unfair to add that
more can be said against than for each of them. It is important
that all such arguments, the favourable as well as the unfavourable
ones, be squarely faced.
The Robot Model
There I should like to leave the matter, having made such slight
constructive contribution to the problem as is provided by the formulation of an uncompromising question. But consideration of the
question has unfortunately been greatly confused by the introduction
of another and wholly irrelevant one. And I am afraid that attempts
to face those facts that are relevant will be repeatedly frustrated by
the irrelevant ones so long as they are allowed to obtrude.
The irrelevant question is concerned with what may conveniently
be called the "robot model". It is a question about the extent to
which robots could be so designed that they would perform the various activities of living organisms. It has come to be very generally
assumed that the correct answer to this question would also provide
the correct answer to the questions: Where is the observer? What
does the interpreting? It will be shown below that there is no connection between the questions; the belief that there is a connection
results from failure to conform to a very elementary rule, the rule
that in reaching any conclusion all facts used must be tested for
their relevance. When this rule is not followed the result is the error
in logic technically called a non sequitur. In this instance the error
is committed, it will be seen below, not once, but repeatedly.
The reasoning of those who introduce the robot model into the
discussion proceeds on the following lines. The operator of the anti-aircraft guns does not need to interpret the break in the line on the
screen of the oscilloscope as representing an aeroplane in the sky.
It suffices for him to obey an instruction received from the N.C.O.
in charge of the post. He is told to press the button when the break
appears on the screen and the result will be the same whether or
not he understands what he is doing. And so, it is easy to see, there
is no need at all for a human operator. He could be replaced by
some electronic device, by a robot. This would not interpret; it
would not observe. It would merely receive signals from the radar
apparatus, translate their pattern into another form, and transmit
the new pattern to the next link in the causal chain. Its performance
would be basically of the same nature as that of any of the devices
on the efferent side of the operator; and we have already seen that
these do not observe or interpret. The robot would have to be correctly adjusted by the N.C.O. But it would not be required to know
the reason why; only to act.
All that is true enough. There is indeed no limit to the number
of elaborate and difficult performance that can be achieved by a
suitably designed robot. Some of those used in gunnery incorporate
ingenious computing devices. Laymen often believe, erroneously,
that they solve problems in higher mathematics. But the rather
common assumption is not true that this fact has any relevance to
the questions where is the observer?, what does the interpreting?
The purpose of using a robot, it seems to be forgotten, is to replace
a human operator who must observe and interpret by a device that
does not do so. The robot is, by intention, not a true model. The
whole of it is on the efferent side of the person who makes use of
it. Every part of it is designed to serve, like a muscle, a pushbutton,
a gun mechanism as an instrument of control. It can only; be a non
sequitur to reach any conclusion about the acts of observing and
interpreting from the contemplation of a device deliberately designed
not to perform these acts.
Testing for a Non-Sequitur
The best, perhaps the only, way of detecting a non-sequitur is
to attempt to fill in the gap between facts and conclusion by more
closely reasoned steps. Here the facts are that a robot can be designed for a performance identical with that of a human operator.
In these days two distinct and contradictory conclusions are being
erroneously drawn by various writers from these facts. They are:
a that observing and interpreting are performed by a material brain, and
b. that these acts are insignificant and questions about them
By what logical steps is either conclusion reached? The question is not an idle one. For the custom is to present the conclusion
without the intermediate steps, as though it were too self-evident to
need them. And this is unfortunate. It has led to much wasted
effort, which will continue unless the irrelevance is recognised. And
so the above-mentioned test for a non sequitur cannot be avoided.
We must attempt to reconstruct the steps in reasoning that could
be inserted between facts and conclusions, even though this method
must inevitably lead to a line of argument that it seems unkind to
quote. For, no matter how fair one tries to be to those who base
their conclusion on the robot model, every attempted passage from
the facts to the conclusion cannot be made to appear better than a
parody of loose reasoning. The following typical reconstructions
1. "The robot is so designed that (a) no non-material link
occurs in the causally connected chain of events that follow each
other within the structure and (b) it does not observe or interpret.
It follows that, when a human operator does observe and interpret,
these acts are performed by a material link in the chain of events".
2. "The robot has been deliberately designed so as to act
without observing or interpreting. I believe, nevertheless, that it
does these things. There is no non-material link in the causally connected chain of events in the robot; therefore the observing and interpreting that, according to my hypothesis, are performed by it
must be done by a material link. Therefore there is no reason why
observing and interpreting should not be performed by a material
link in the human operator".
3. "If one can design a robot to replace a human gunner
one can design another robot to replace the N.C.O. who gives the
gunner his instructions. One can also design further robots for other
functions. Among these one can, according to my hypothesis, design
robots capable of inventing robots, of manufacturing them, of controlling, adjusting and repairing them; one can design a complete
self-contained robot community. And each individual is designed
so as to act without observing or interpreting. Such a community
would behave, nevertheless, exactly like a human community. This
proves that the human community does no observing or interpreting.
It is therefore meaningless to ask whether these functions are performed by a material brain or by a non-material mind".
Such arguments might serve as classroom exercises, in which
the students would be asked to name, classify and explain the fallacies. There is no need to dilate on them here. But has anyone
provided any closely reasoned steps from the robot model to any
conclusion about observing and interpreting that did not contain
these or similar fallacies? Is it possible to devise across that gap
a row of closely reasoned steps that are free from errors in logic?
I have tried to devise them but have failed. And so the big questions remain as puzzling now. as they were when engineering
devices were not quite so ingenious and elaborate as they are to-day.
In this Paper we find within the field of anatomy and physiology, if not even
of cybernetics, one of the oldest and most classical problems in philosophy and
science: the problem of observation and knowledge. Professor Kapp mentions his
difficulties about observation and shows why the robot model fails to overcome them.
It would not be appropriate in a comment to review the whole ground, but
we should like to mention two aspects that have been emphasised by us elsewhere.
They are respectively the original source of the difficulties mentioned by Professor
Kapp, and the reason why it is only because of the traditional way in which the
problem of observation has been tackled in philosophy that the difficulty arises.
The origin of the difficulty lies far back in time. When the Greeks enquired
into the relation between words and the things designated by the words they sought
an answer, not in unspecified things, and in a specified relation bearing upon them,
but in the things they already were using as words and as named things. That
gave them no opportunity for considering the relation, i.e, the operations of semantisation. On the contrary, that suggested to them the addition of something new
between the words and the named things. It was an activity, that they called
knowing, and was regarded as something passing between the named thing and the
person who was applying the name, and informing him, by means of its differences,
of the differences between one named thing and the others. It followed from this
approach that the activity of knowing could only be described in terms of an
irreducible metaphor and that any attempt at literal description must lead to a
Since knowing is the motion of something, location was attributed to all named
things; and since knowing is information about differences between named things,
all named things were regarded as being differentiated in themselves and by themselves, Hence, all named things, being regarded as occupying different regions of
space by means of the differences between themselves, became physical things.
This view soon led to trouble. Some of the things that we name are not physical things: how are they to be explained? The Greeks did so by attributing location
to these too, be it by locating them in the world of the Gods or in a mind or
consciousness conceived as a point in space, without extension. Those who refused
to attribute this kind of location for fear of being called mystics or, a more
approbrious epithet in some circles, psychologists leaved uncoupled the words of
the things lacking, i.e, leaved these words without semantical relation. The development of the introduction of knowing in philosophy and science since the days
of Thalea, Alcmeon, Empedocles, Lencippus, Democritus, etc. is, however, somewhat
beside the point.
Let us turn instead to the present attitude of those who, perhaps equipped
with the most refined and scientific observational technique, are seeking to understand
the nature of observing regarded as knowing, of observer regarded as knower.
l"or they start with the assumption that the things have already been located and
differentiated, they direct their attention, not to the activities by which the things
are located and differentiated, but to a passive activity, at the best an activity
This is as thought a person were to seek to understand the nature of painting,
not by observing painters at work, but them in a world which is from the beginning a picture gallery.
For these reasons we agree with Professor Kapp when he fails to find the
observer and the act of observing while studying some piece of a machine, not as
a working piece, but as a statical one; and we also agree when he expresses his
dissatisfaction with the solution that has hitherto been offered. But these difficulties only arise because he adopts the traditional view of knowledge, they are
a result of the historical development in philosophy and not intrinsic. Only this
traditional approach leads to the view that a human being could not in principle
be copied by a mechanical model and that machines capable of reproduction are
Those things about man that can be expressed without having to resort to irreducible metaphors or revealing contradictions are in theory capable of being
copied like any other model.
(For an analysis of observing made by our operational School, see for instance my
writing a Consapevolizzazione dell'osservare, Mod. 3. Atti del Congresso di Studi Metodo-logici, Torino, 17-20 dicembre 1952, Edizioni Ramella, Torino, 1954).
REPLY BY REGINALD 0. KAPP
Mr. Ceccato has drawn attention to an all too common error in philosophical
thinking. It is to confuse the meaning of the words to be and to do. Mr. Ceccato
says quite rightly that people in the past have hoped by studying what things are
like to learn about the things that they do and are done to them. This was one
of the basic errors in the once fashionable philosophy of emergence advocated by
Alexander, Lloyd Morgan and others, as I pointed out in my book Science v Materialism . The same error is made by those who pin their faith to the robot analogy. I agree with Mr, Ceccato that a living organism is essentially a machine, that
the difference between a man's body and a life-like robot is not basic. But the
conseguences of this view have to be faced squarely. They are generally dodged and
this perhaps because they are particularly disconcerting to just those followers of
traditional thought whom Mr. Ceccato refers to.
A robot works in a specific way only when specific things are done to it. It
does not do so when left to the random, unspecific influences of its environment. A
man is basically a machine. Therefore, one must infer, he also works in a specific
way only when specific things arc done to him, not if left entirely to the random
unspecified influences of his environment. The agent that does the specific things to
the robot is a human operator; he handles levers, push buttons, rheostats,
adjusting screws and other controlling devices. The agent that does specific things to the human machine is, as just pointed out, not in the material environment. So the conclusion to be drawn from the robot analogy is that this agent is non-physical and
lacks location. One name is "life" but I prefer the less committal term
"a diathete". It is most odd how those who stress the mechanical analogy always argue
in exactly the opposite way. It only shows how necessary is the kind of guidance along logical lines provided by Mr Ceccato’s remarks.
Top of Page
1 We are no more conscious of performing an act of interpretation during
an observation than we are of the nature of the various patterns presented, to the
brain when electrical impulses reach it from the organs of sense perception. But
consciousness is no criterion of reality. That interpretation is performed unconsciously
does not alter the fact that it is performed; for every physiologist knows that there
is a change of pattern between what is presented to the eye by the object observed
and what is presented to the brain by the afferent nerves.
2 There are good reasons why the dualist, if he is consistent, must regard the
whole of living substance as forming a region of interaction between the two component parts of reality, and therefore provided with devices adapted to the function
of interaction. EIsewhere ( Mind, Life and Body Constable) I have described
these devices as "eudiathetous mechanisms". There is, however, no need at the
present moment to consider interaction in such general terms.