Up until Paul Broca's work, the lateralisation of language in the brain was unanticipated. He was the first to ascribe language to the hemisphere contra-lateral to the preferred hand. In fact, this correspondence does not pan out completely; 97% of right handers are left lateralised for language and 75% of left handers are still left lateralised for language. There are more right hemisphere dominant individuals amongst the left-handed population than the right-handed population.

Evidence for right hemisphere involvement in speech and language processes can be seen by studying normal subjects, patients with hemispherectomies, patients with severed commissures, aphasic patients with right hemisphere lesions and patients with right hemisphere lesions.

A right hemisphere infarct does not commonly lead to aphasia. Right hemisphere lesions do not typically result in syntactical or semantic problems. However, language also involves other components. In addition to linguistic processing, there are important gestural and prosodic aspects of human communication, which have been described as para-linguistic, and which link in to the pragmatic or discourse functions of language. The right hemisphere may well preferentially process these.

In tests, RH patients can show problems with semantic comprehension (Lesser 1974) but 'phonological' skills can be spared. Generally straightforward receptive and expressive tasks are OK (Myers 1986). Thus naming, word discrimination, following commands, word and sentence reading and writing can be OK. Where there are language problems, they are mild. Communicative competence is not impaired. The high level communicative impairments seen may contributed to more cognitive-affective problems.

However, the communication skills of patients with RH damage can be 'unusual'. They can lack the understanding of the context of an utterance, following the presuppositions entailed in an utterance, or the tone of an exchange.

Their speech can be: Excessive, Rambling, Inappropriate, Confabulatory, Irrelevant, Literal, Sometimes bizarre (Gardener et al, 1975)

These are roughly divided into cognitive-linguistic aspects of language, and prosodic problems.

Comments can be 'off colour' and humour can be 'inappropriate'. They can focus on insignificant details in conversations, make tangential remarks, and display a reduced range of intonation. They have a problem integrating and organising information that they are presented with. They can have problems interpreting figures of speech and proverbs. They tend to interpret language figuratively. A study by Foldi et al (1983) indicated that RH subjects cannot discriminate between direct and indirect speech, instead they interpret the indirect speech directly, without taking context into account. They can have problems grasping intended or implied meaning embedded in an utterance.

When understanding narratives, RH patients tend to use an analytical style to infer information directly. This meant their processing of emotional, moral and humorous content was literal and abnormal (Gardener et al, 1983). When recalling the narrative, they could recall isolated details, but also made confabulations. They 'violated the story boundary' by including personal experiences and opinions.

In terms of speech perception, patients with RH damage can demonstrate problems with prosodic aspects of speech perception, with possible emphasis on the emotional aspects of speech prosody. Their own speech can demonstrate a 'flat affect' and a monotonous intonation. They can also have problems with interpreting facial expressions of emotion and vocal expressions of emotion in speech or body language. Aprosodia, an impairment of expression of emotion in voice 'melody' is associated with RH damage in the right posterior inferior frontal cortex (the mirror site of Broca's area). Likewise the perception of emotional 'prosody' is associated with posterior temporal/parietal damage on the right. This anterior = production and posterior = perception has the benefit of being somewhat like the neuro-anatomy of language in the left hemisphere, but the extent to which this reflects a selective sensory or a production deficit if emotional prosody is not clear. Thus, it could be that the right temporal lobe is selectively involved in the processing of dynamic pitch variation, and judgements about that (Johnsrude et al, 2000). Thus there could be consequent problems with the processing of intonation, which relies on this. In addition, the issue of non-emotional prosodic information is not always adequately assessed.

The affect expressed by RH patients does not necessarily reflect their mood. Thus flat, monotonous speech does not mean that the person is depressed. Likewise the processing of the emotional prosody can be damaged whilst semantic knowledge of the emotions expressed is intact. In contrast, damage to some sub-cortical regions (e.g. amgydala) can result in deficits of processing non-verbal expressions of emotion (e.g. fear and anger in the face and voice) and this is coupled with an apparent change in the experience of these emotions.

These problems in cognitive-linguistic and affective-prosodic aspects in language 'add up' into a problem with pragmatics. This has been described as a problem in "evaluating the significance of sensory input, associating it with prior knowledge and integrating multiple features of experience into a meaningful pattern or context" (Myers, 1986). This results in their unusual and literal use of language. Using the cookie theft picture from the Boston Diagnostic Aphasia Examination (1972) Myers demonstrated that relative to non brain damaged controls, RH patients used fewer interpretative comments when describing the picture, and more irrelevant comments.

This pragmatic problem can lead to problems in conversation. Compared to LH aphasics, RH patients make similar numbers of correct pragmatic responses, but made a distinct set of pragmatic errors. These included eye gaze, prosody, conciseness and turn-taking mistakes. This has been summarised as LH patients communicating better than they speak, and RH patients speaking better than they communicate.

Studies with 'split brain' patients have shown that there can be language in the right hemisphere. If 'spoon' is flashed into the left visual field of such a patient (the information being then passed to the right hemisphere), they can point at a picture of a spoon but not respond verbally. Thus the RH is not 'word blind' or word deaf. This does not seem to involve sound structures: the meaning of ache and lake could be determined by the right hemisphere, but not that they rhyme.

Sometimes a total hemispherectomy is performed to treat brain disease, especially epilepsy. The results of this vary hugely between children and adults, especially the impact on speech and language. Children can develop essentially normal speech and language following such an operation. It has been claimed that up until puberty, the removal of the (dominant) left hemisphere does not result in aphasia. Several people have argued that the 2 hemispheres are 'equipotential' for language up until the ages of 4-5 years. This suggests that cerebral dominance for language develops with age (although the effects of the epileptic damage may contribute to this).

In contrast, hemispherectomy in late childhood or adult life can have a serious impact on language function. This is infrequently performed, is more likely due to the removal of a glioma, and in this case removal of the left (dominant) hemisphere results in a profound aphasia. If the LH has been damaged since childhood, then the RH has probably already taken over the functions, and the operation does not result in aphasia. There have been reports that the removal of the diseased LH can improve speech and language skills, since there is no longer influence of the damaged LH on the intact RH.

We cannot regenerate nerve cells (neurones). However, neuronal connections can be remodelled (e.g. Hebbian nets).

• This can be seen developmentally. Rats which develop in a 'rich' environment have more synaptic connections than those raised in a sparse environment.
• This can happen due to learning, e.g. the motor representations of the hand can vary depending due to learning to play a fingered musical instrument.
• It can happen due to altered sensory input; e.g. amputees no longer have input from the removed limb. The associated sensory cortex does not wither away; instead there is a re-organisation of the cortical map, e.g. involving greater representations of surrounding regions.
• It can happen when 'new' input is developed, e.g. when a blind person learns to read Braille.

At one level, your brain is synaptically re-organising all the time, as you lay down new memories, as you learn new information and new skills. However the main reason we are addressing this is because it underlies recovery from stroke and brain damage.

There are several possible mechanisms.

• New synaptic connections can be made through dendritic connections.
• There can be axonal regrowth - if the cell body is intact.

Both of these help reform the dense network of cells that make up cortical systems. They, in turn, can lead to functional reorganisation of that area, of an area remote to the lesion site.

Activity dependent modification of synaptic connections and reorganisation of adult cortical areas involve Long Term potentiation (LTP) and Long term depression (LTD), methods of storing information in the mammalian nervous system.

Synaptic plasticity in cortical horizontal connections underlies cortical map reorganisation.

The mechanism can vary across brain regions and due to type of injury sustained. At the level of the cortex, the dense interconnected network of cells lead to a certain resistance to damage. Because the information is shared across the network, then damage to part of the network does not lead to irretrievable damage. Instead, there can be restructuring of the existing network. Thus, some studies show that as long as at least 10-15% of the connecting fibres survive, then there can be (reasonably full) recovery within 2-3 weeks. This is not the case for the whole cortex, however. Recovery is rapid for functions that are subserved by multiple circuits, and slow or impossible in functions supported by a limited number of circuits, e.g. hemianopia., since fewer alternative pathways are available to support the damaged pathways. Thus speech and language, which is so complex a 'function' shows better recovery than memory functions, which do not tend to improve with time.

There can be an impact of type of damage which interacts with this- encephalitis and damage due to surgery has less recovery than TBI.

In some brain regions, up to 95% of the connections can be lost and roughly normal performance maintained. Loose a little bit more, however, and this can be catastrophic. Something like this can be seen in multi-infarct dementia, when there will transpire to have been a history of strokes, with full recovery, before the 'dementia' became apparent in a step-wise fashion. There can be recovery in-between strokes due to this re-wiring, that supports normal function, until at last there is too much damage for the system to cope. This can also be seen in Parkinson's disease, where problems with movement are only clearly detected once 80% of the dopamine cells in the substantia nigra have died. This is due to increased firing of the existing cells.

This is why it is better to have the same amount of damage 'spread' over several small strokes than in one large stroke, and this does predict recovery, as does the interval between infarcts (longer gap is better than shorter). There is the opportunity for re-organisation in-between infarcts, improving the changes of recovery and meaning that the next stroke is not 'additive' in the damage it causes. It also means that the more connections there are there in the first place, the better the outcome will be. Thus, better educated people have more connections, and are to some extent 'buffered' against brain damage. People who suffered brain damage when young, however, are at greater risk from damage when they are older, as the system has already sustained some damage. NB the loss of neurones in the normal ageing brain is compensated for, to some extent, by increased dendritic growth - use it or loose it!

• Lesion occurs - causing neuronal damage.
• Can be followed, peri-infarct, by further loss of tissue,
• The activity of the immune response system
• Due to loss of connections
• Non-use

There are post lesion changes occurring over time, from minutes to months. These can be due to

• Deafferentation
• Removal of inhibition
• Activity dependent synaptic changes
• Changes in membrane excitability
• Growth of new connections
• Unmasking of previous connections

Assuming that the damage area cannot support function, there are three potential locations of recovery.

1. peri-infarct
2. in the homologous contra hemispheric region
3. in a distal brain region

The exact nature of this will vary with the size and location of an infarct.

Brain imaging techniques enable us to unpack some aspects of this.

Some studies have shown peri-infarct activation in recovered aphasics - evidence that brain regions around the edge of the infarct have taken over the function. The amount of peri-infarct acrtivation has been claimed to be a predictor of amount of recovery (Karbe et al, 1998), but this might reflect differing initial damage. Peri-infarct activation can be due to some surviving tissue, rather than actual re-organisation. More often, a mixture of activation is seen.

In Mummery et al, 1999, we demonstrated that the right STG/STS is involved in 'normal' passive speech perception. Also scanned 2 recovered aphasics with mature lesions. Each had recovered some comprehension, though EJ was distinctly better than SS who could understand single words. Both showed extensive right temporal lobe activations when listening to speech, no peri-infarct activation in EJ and only a small residual region of the medial left dorsal temporal lobe responded in SS. Since both aphasic patients had mature lesions, the assumption was that vascular reactivity around the infarct had returned: that is, peri-infarct activation would have been observed if it had been present. Furthermore, in SS activation was observed in the most medial part of the left dorsal temporal lobe, the only residue of superior temporal cortex left after his stroke, and this was directly adjacent to the infarct.

This confirms a role for the right STG/STS in recovery from aphasic stroke, as does the finding that bilateral damage is needed before patients become 'word deaf'.

Weiller et al, 1995. Changes in the organisation of the brain after recovery from aphasia were investigated by measuring increases in regional cerebral blood flow (rCBF) during repetition of pseudowords and during verb generation. Six right-handed patients who had recovered from Wernicke's aphasia caused by an infarction destroying the left posterior perisylvian language zone were compared with 6 healthy, right-handed volunteers.

In the control subjects, strong rCBF increases were found in the left hemisphere in the posterior part of the superior and middle temporal gyrus (around Wernicke's area) in both tasks. During the generation task activation was greater in lateral prefrontal cortex (LPFC) and in inferior frontal gyrus (Broca's area). There were some weak right hemisphere increases in superior temporal gyrus and inferior premotor cortex. In the patients, rCBF increases were preserved in the frontal areas. There was clear right hemisphere activation in superior temporal gyrus and inferior premotor and lateral prefrontal cortices, homotopic to the left hemisphere language zones. Increased left frontal and right perisylvian activity in patients with persisting destruction of Wernicke's area emphasises redistribution of activity within the framework of a pre-existing, parallel processing and bilateral network as the central mechanism in functional reorganisation of the language system after stroke.

Thus Weiller et al, with a more complex task, show that the sensory processing of the input has possibly transferred to the right hemisphere (or has been taken over entirely by the right STG/STS). However the more frontal regions involved in the generation of language, which are distal to the stroke and thus not directly compromised, do not re-lateralise.

In the Mummery et al study, 1999, in addition to extensive right STG/STS activation to speech, EJ also showed right prefrontal activation when listening to speech. This was not seen in the controls or in SS. Strikingly, EJ's recovery is much better than SS's, and she seems to have recruited some prefrontal, 'executive' brain regions in comprehending speech.

The involvement of anterior brain regions in recovery from stroke and brain injury is becoming clearer. The role of the prefrontal cortex is generally in aspects of controlled behaviour that involves decisions, attention and the use of strategies. If this brain region is intact then the chances of recovery are greater - whether due to deliberate strategies, or the recruitment of attention, is not yet known. A better predictor of recovery from aphasic stroke is not post-stroke performance on language tests, but performance on 'prefrontal' tests, like Raven's Matrices. Ian Robertson and colleagues has shown that spared 'sustained attention' predicts motor recovery after stroke. A general good predictor of recovery from stroke if the premorbid cognitive state of the brain - roughly, pre-existing IQ and education. Due probably to both better connections, and 'better' frontal lobes as a result of education.

Drugs

The use of neurotransmitters, e.g. from animal studies, boosting the seretonergic system can facilitate recovery, by directly affecting the damaged hemisphere. The prefrontal cortex could also be targeted.

There can also be intervention to suppress the immune system and reduce the oedema after an infarct, to reduce damage to the surviving tissue.

This can work through helping people to compensate for their difficulties, helping them learn effectively, or through promoting neural re-organisation.

This can be pretty basic, 'mental muscles' stuff, e.g. using repeated presentation of the Token test - Musso et al 1999 showed a benefit, with associated changes in brain function.

Katz and Wetz ran a 3 condition study, with 55 aphasics: computer based training, with either computer reading tasks (visual matching, and reading comprehension), non verbal computer use (games, etc.) or no treatment. Computer reading led to better results than the other 2 conditions which did not differ.

Don't use other systems. Binding the 'good' arm can help recovery in the hemiparetic limb. Can any of these techniques apply to speech?

Stroke clinics. Early mobilisation can help in acute stroke, through reducing secondary thrombitic events, pneumonia and mortality. Admission to a stroke unit is a good way of doing this, with access to a range of qualified staff. Care in such a unit is associated with long term reduction of deaths, and of poor dependency outcomes, independent of patient age, sex or organisation of the stroke unit. It is not know what exact aspects of care are contributing to this, through from the studies described above, targetted therapy in speech and language problems can be beneficial.

1. Memory cannot be recovered, so treatment focuses on compensation.
2. As preserved 'frontal lobe' functions and sustained attention functions can help recovery, so their presence can hinder recovery.
3. Integrity of the rest of the brain, as suggested earlier.
4. Individual idiosyncrasies of brain structure
5. Motivation
6. Emotional factors. This can be a direct result of the damage, or a consequence of the knowledge of the problem; depression is a common outcome after stroke and can impact on recovery.