Brain–Computer Interfaces - Index of

Brain–Computer Interface in Neurorehabilitation 159
my throat”, “I would like to see my friend” etc.) are not followed by the anticipated
or desired consequence. Therefore, extinction takes place within the first weeks or
months of the complete locked-in state. Such a negative learning process can only
be abolished and voluntary, operant control can be reinstated if reliable contingencies
re-occur. Without the knowledge of the patients’ goal-directed thoughts or its
electrophysiological antecedents a reinstatement seems impossible. We therefore
recommend early initiation of BCI training before entering the complete locked-in
state.
These negative results for the completely locked-in and the “thought extinction
hypothesis” illustrate the failure of operant conditioning of autonomic responses in
the long-term curarized rat studied by Neal Miller and his students at Rockefeller
University during the 60s and 70s. Miller [25] reported successful operant conditioning
of different types of autonomic signals in the curarized rat. He proposed
that autonomic function is under voluntary (cortical) control comparable to motor
responses. The term “autonomic” for vegetative system responsivity does not seem
to be the appropriate term for body functions which are essential under voluntary
control. Miller’s argumentation opened the door for the learning treatment and
biofeedback of many bodily functions and disorders such as heart disease, high
blood pressure, diseases of the vascular system such as arrythmias and disorders of
the gastrointestinal system. But the replication of these experiments on the curarized
(treated with curare to relax the skeletal muscles) rat turned out to be impossible
[12] and the consequent clinical applications in humans in training patients with
hypertension or gastrointestinal disorders were also largely unsuccessful. The reason
for the failure to replicate the curarized rat experiments remained obscure. The
extinction-of-thought hypothesis [7, 8] tries to explain both the failure of operant
brain conditioning in completely paralyzed locked-in patients and the failure to
train long-term curarized artificially respirated rats to increase or decrease autonomic
functions. In two of our completely locked-in patients, we tried to improve
the signal-to-noise ratio by surgical implantation of an electrode grid subdurally
and tried to train the patient’s electrocorticogramm in order to reinstate communication.
At the time of implantation, one patient was already completely locked-in
for more than a year. After implantation, no successful communication was possible
and no voluntary operant control of any brain signal was possible, supporting
the extinction-of-thought hypothesis. A second patient with minimal eye control
left was implanted recently in our laboratory with a 120 electrode grid epidurally
and was able to communicate by using electrocorticographic oscillations from 3 to
40 Hz. This patient had some motor contingencies still left at the time of implantation
and therefore the general extinction of goal directed behavioral responses
was not complete, allowing the patient to reinstate voluntary control and social
communication after complete paralysis.
At present with such a small number of cases any definite recommendation is
premature but we strongly recommend invasive recording of brain activity if EEGbased
BCI fails and there are still some behavior-environment contingencies left.
The application of brain–computer interfaces in disorders with severe restriction
of communication such as autism, some vegetative state patients and probably in