54
M. J. WELLS
apparent that the results were difficult to interpret in the absence of
this background. The few really successful experiments-successful in
the sense that it has been possible to use them as tools for further
investigation of the animals’ nervous systems-are those that by
chance or design happened to mimic some natural learning situation.
Thus, for example, the reward and punishment visual discrimination
technique devised by Boycott and Young (1950) for training octopuses
is successful probably because it tests the animals under conditions
similar to those that they would have to contend with in the sea.
Octopuses normally live in “ homes ” from which they emerge to
attack passing prey. The tank situation teaches them which prey is
worth attacking, a type of discrimination they would probably have to
learn anyway. In the sea they must relearn repeatedly as they grow up
and become capable of preying upon animals a t first too large and
dangerous for them to tackle.
It is possible, though the data are at present scanty, that
animals can be divided into two great categories in respect of the things
that they can learn. The distinction rests on whether they can or
cannot take into account the positions of parts of their own bodies
when they learn. Octopuses at least, which have an otherwise impressive performance, seem unable to do this. Arthropods, by contrast,
seem to learn to make discriminations based on kinaesthetic cues
relatively easily. It is very noticeable that arthropods and vertebrates
are the only animals that have been shown to learn mazes beyond the
complexity of a straightforward T. It has been argued elsewhere
(Wells, 1963b, 1965b) that this may be because the movement of softbodied animals, unrestricted by joints, is necessarily organized hierarchically, so that proprioceptive inputs never penetrate to the uppermost
parts of the CNS. With the restrictions imposed by joints, central
control of movement becomes feasible and with this the possibility of
taking details of movement into account in learning. In this respect the
annelids may represent an intermediate condition in which segmentation has produced pseudoarticulation and with it the possibility of
some central control of movement on a kinaesthetic basis. If these
speculations are correct, the range of things that it is reasonable to try
and teach a soft-bodied animal is limited.
The more one learns about invertebrates the more it becomes
apparent that many, perhaps most of them, arrange their activities
mainly in response to chemical cues. These, then, are the stimuli that
should be used for most invertebrates in experiments on learning.
And these, unhappily, represent the one class of stimuli that we
ourselves find hard to measure and impossible to classify.
4.
5.
M. J. WELLS
apparent that the results were difficult to interpret in the absence of
this background. The few really successful experiments-successful in
the sense that it has been possible to use them as tools for further
investigation of the animals’ nervous systems-are those that by
chance or design happened to mimic some natural learning situation.
Thus, for example, the reward and punishment visual discrimination
technique devised by Boycott and Young (1950) for training octopuses
is successful probably because it tests the animals under conditions
similar to those that they would have to contend with in the sea.
Octopuses normally live in “ homes ” from which they emerge to
attack passing prey. The tank situation teaches them which prey is
worth attacking, a type of discrimination they would probably have to
learn anyway. In the sea they must relearn repeatedly as they grow up
and become capable of preying upon animals a t first too large and
dangerous for them to tackle.
It is possible, though the data are at present scanty, that
animals can be divided into two great categories in respect of the things
that they can learn. The distinction rests on whether they can or
cannot take into account the positions of parts of their own bodies
when they learn. Octopuses at least, which have an otherwise impressive performance, seem unable to do this. Arthropods, by contrast,
seem to learn to make discriminations based on kinaesthetic cues
relatively easily. It is very noticeable that arthropods and vertebrates
are the only animals that have been shown to learn mazes beyond the
complexity of a straightforward T. It has been argued elsewhere
(Wells, 1963b, 1965b) that this may be because the movement of softbodied animals, unrestricted by joints, is necessarily organized hierarchically, so that proprioceptive inputs never penetrate to the uppermost
parts of the CNS. With the restrictions imposed by joints, central
control of movement becomes feasible and with this the possibility of
taking details of movement into account in learning. In this respect the
annelids may represent an intermediate condition in which segmentation has produced pseudoarticulation and with it the possibility of
some central control of movement on a kinaesthetic basis. If these
speculations are correct, the range of things that it is reasonable to try
and teach a soft-bodied animal is limited.
The more one learns about invertebrates the more it becomes
apparent that many, perhaps most of them, arrange their activities
mainly in response to chemical cues. These, then, are the stimuli that
should be used for most invertebrates in experiments on learning.
And these, unhappily, represent the one class of stimuli that we
ourselves find hard to measure and impossible to classify.
4.
5.