The evidence that learning induces specific molecular and structural change in certain structures of the chick brain (which are also unusually active during learning) is compelling. Such changes have been shown to occur only if the opportunity to learn is given, to correlate with the degree of learning, and to be blocked if amnesia is induced after learning (Chapters 8, 10). A wide range of controls for general effects, which might be associated with the learning situation and which might change neuronal properties, have been carried out.
It is thus possible to begin to study the chain of events which leads to the establishment of a permanent memory in the chick. It is true that the story may turn out to be far more complicated than the sketches which are at present possible. Several brain structures are certainly affected by the bead task (Chapters 10, 11); no doubt more remain to be identified for both tasks. Plasticity of neuronal functions may follow different routes in different structures, or during different ways of processing information. Horn (Chapter 8) notes that changes within one structure, the IMHV, seem to differ after different tasks.
A crucial distinction is drawn by Rose (Chapter 10; see also Bateson, Chapter 4) between changes which ‘enable’ the later processes which lead, through protein synthesis, to the permanent storage of information, but which are not themselves accessible to retrieval mechanisms, and the changes which do allow retrieval during the earlier parts of memory formation. It would be possible for enabling changes to be quite independent of the substrate for early memory: the two might even be held in separate populations of neurons in separate brain structures. Conversely, the two might be tightly coupled, so that each stage on the way to the production of the substrate for permanent memory might itself be responsible for current memory: neurons affected by enabling changes would also be labelled for retrieval. The issue is considered more fully in Part V.
Key events in the route to permanent memory which are already demonstrated in the chick may begin with up-regulation of transmitter receptors (e.g. cholinergic or NMDA: Chapters 8, 10). An early event which has been shown to be essential for permanent memory formation, in that its blockage results in amnesia, is presynaptic change in the phosphorylation of a specific protein (Chapter 10). By 30 minutes an increase in both a particular gene message, and in RNA polymerase marks (p.216) the earliest steps leading to protein synthesis (which involves both glycoproteins and tubulin); again, blockage of these steps has been shown to produce amnesia (Chapter 10).
It seems very probable (see also Chapter 7) that such protein synthesis leads to the structural changes which have also been shown to be learning induced. These include both synaptic changes and changes in spine density and shape (Horn, Chapter 8; Stewart, Chapter 11), which can at least partly be understood as facilitating transmission (Chapter 11).
The IMHV has been independently identified as crucial in two very different learning tasks. The possible anatomical correspondence between the HV, of which it is part, and parts of cortical areas such as prefrontal, orbitofrontal, cingulate, and temporal (Section 2; Horn 1985) makes understanding the functioning of the IMHV in learning and memory a matter of great importance. Studies in which the IMHV was lesioned, reviewed here by McCabe (Chapter 9) and Davies (Chapter 12), have led to a model by which, after imprinting, the left IMHV holds permanent memory in the left hemisphere, but the right IMHV acts as a buffer store, establishing permanent memory elsewhere (S’) in the right hemisphere, after a substantial delay (c. 6 hours). This agrees entirely with the extensive evidence that learning-induced change occurs predominantly in the left IMHV. It explains such a variety of evidence from lesion studies that it must clearly be the preferred hypothesis.
However, the model continues to be extended and changed. Both Davies and McCabe (Chapters 9 and 12) note that unilateral lesions of the left IMHV before training prevent either acquisition or retention of the aversive bead task; lesions of the right IMHV have no effect. They conclude that in this task learning involves only the left IMHV (unlike imprinting, where either IMHV by itself suffices for acquisition; note that brain structures other than the IMHV are likely to be involved in both hemispheres in the aversive bead task). It is thus of great interest that bilateral IMHV lesions only 1 hour after the bead task have no effect on retention. Clearly this implies, on the hypothesis under discussion, that the left IMHV has here acted as a buffer store, and after 1 hour has set up a permanent store elsewhere. One possibility is that this permanent store (S’?) is in the left hemisphere; this idea has the attraction that it would imply that memory formation is similarly organized in the two hemispheres, except that it takes far longer to establish S’ in the left hemisphere. Another possibility is suggested by the hypothesis (McCabe, Chapter 9) that the left IMHV causes the establishment of S’ in the right hemisphere: it might do this by transferring information to the right hemisphere.
The idea that in both hemispheres the IMHV functions predominantly as a buffer store meets an obvious and perhaps insurmountable obstacle in the striking concentration of learning-induced change in the left IMHV; (p.217) in the case of the bead task, both Rose and Stewart (Chapters 10, 11) stress the surprising scale of such change at both molecular and structural levels. It is worth noting, however, that permanent learning-induced change in the young chick could have a number of functions. These range from: (i) a detailed record of training as a unique experience (‘episodic’ memory); (ii) a record of key features of the experience, selected after comparison with records of relevant earlier experiences, and their association over time with unpleasant taste. Such ‘mediational’ memory (Horn, Chapter 8) or ‘semantic’ memory, is necessary to allow subsequent control of behaviour; (iii) changes in perceptual mechanisms (e.g. the development of complex feature detectors: Chapter 4); (iv) changes in mechanisms necessary for learning (e.g. the establishment of functional connections between reinforcing and other mechanisms; maturation of inputs from arousal mechanisms: Chapter 8). It is possible that a structure which is predominantly used as a buffer store might show permanent changes in categories (iii) and (iv), when it is used for the first time in learning.
It is possible that the description of the IMHV as a buffer store only covers part of its functions. Horn (Chapter 8) compares chicks in which the IMHV have been lesioned with human beings who have become amnesic due to medial temporal lobe/diencephalic damage. Such amnesics are not only able to make use of past learning but also continue to learn, particularly when comparable experiences recur repeatedly; what they lack (or at least deny that they possess) is the ability to recollect specific events. In chicks, bilateral lesions of the 1MHV do not affect retention of either imprinting or the bead tast, so long as lesioning occurs sufficiently long after learning. Acquisition of the same tasks is, however, blocked; in contrast, a discrimination task involving unambiguous and repeated pairing of one stimulus with reinforcement and another with its absence can be learned. It is possible that this involves learning comparable with the acquisition of new information by human amnesics as a result of repetitive experience. However, it is also possible that effects of age may be important. Both imprinting and the bead task have been studied in the first 2 days of life, whereas the discrimination tasks used older chicks. Some learning is certainly possible in young chicks with bilateral lesions of the IMHV: they will learn to move on to a pedal to see an imprinting object (Chapters 8 and 12). Nevertheless, it might be interesting to carry out a wide range of tasks in lesioned chicks which are all of the same age.
A final striking feature of learning-induced change in the chick is its asymmetry. In general before training, the biochemical and structural features which change due to training are less well developed in the left IMHV (and probably the left LPO and PA), and training reduces or abolishes this asymmetry by causing the left hemisphere to catch up (Chapters 8, 9, 11). There is increased metabolic and neural activity in (p.218) right hemisphere structures associated with learning (IMHV, LPO, PA), and all show some more permanent change (Chapter 10), so that they are certainly involved to some extent in both imprinting and the bead task.
Two features of cerebral lateralization in the chick offer potential explanations for such asymmetry. The first is the differing specialization of the two hemispheres for the processing of information (Chapter 21). The right hemisphere appears to be especially concerned with, and good at analysing and storing detailed records of experiences. The left hemisphere is apparently more concerned with the categorization of stimuli for purposes of response. It is likely that categorization is only possible after a certain amount of perceptual learning, based on single stimuli and experiences, has been completed. The more advanced state in the right hemisphere of structures central to learning processes could then be understood in two ways (which are complementary rather than mutually exclusive). Perceptual learning of a variety of types before training might bring about the more advanced state. On the other hand, the advanced state might be a preparation for such learning, rather than a consequence of experience.
The second feature of chick lateralization which may be relevant is that cerebral dominance appears to change with age and task (Chapters 6, 20, 21). If, for either reason, the left hemisphere were predominantly responsible for the processing of information and the control of behaviour in young chicks during imprinting or the aversive bead task, this might explain the asymmetry in learning-induced change. These issues are considered further in later sections.