Understanding the Brain p. 6 Josephson
9. Language
The outcome of representational redescription is the development of a partly separate abstract level of activity,
mirroring activity itself (in accord with appropriate correspondence schemes), and thereby supportive of it,
instantiated by its own growing collection of systems specialised for such activity. These higher level systems
take a particular abstract form, and search for valid content consistent with the demands of that form. Language
can then be characterised as an extension of this redescription process, a further derivative level of activity
supportive of existing levels. The present approach leads us to frame the question in terms of a collection of
interrelated model systems, within which the processes of language are defined. Language can be thought of as
an idea made up out of many ideas, the workings of which we may, as research workers, come to understand
through our analyses of how language works. Nature instead simply discovered empirically, through the
mechanism of natural selection, processes implementing the ideas concerned.
Jackendoff s proposals concerning the evolution of language provide a good starting point for the application of
the present approach. He argues provisionally for a particular sequence of strategies through which the present-
day form of language could have evolved, examples of which are the use of symbols in a non-situation-specific
manner, the use of symbol position to convey basic semantic relations, hierarchical phrase structure, and a
system of grammatical functions to convey semantic relations. In addition to these specifically linguistic
functions, imitation plays an important role in making language system to a first approximation a shared
system rather than one deriving from an individual.
All of these processes have to be implemented by physical hardware. Proposals indicative of the general form
of the architecture, including specification of the various data types involved in language, are suggested by
Jackendoff, but the present scheme can support much more detailed specifications, involving a specification of
all the processes associated with a particular abstraction, together with the hyperstructure and frame-creation
mechanisms for creating new structure. This paper will not attempt an exhaustive analysis, but rather focus on a
few important themes.
We consider what are essentially prelinguistic stages, where the fact that language makes reference has not
entered into the picture. Perception and imitation are the relevant themes here. As emphasised by Arbib
(2000), mirror systems play an important role in imitation, representing the perceived behaviour of others in a
format suitable for making a copy of the observed behaviour, which representational system as far as spoken
language is concerned can be identified with the phonology box in Jackendoff s scheme. However, there is no
reason for this component of the system to be linked exclusively with speech, since imitation is equally relevant
for written language and for sign languages. Each modality must have its own forms of representation in the
mirror system, but the target process viz. a match between the product of one s own activity and the perceived
activity of another, is the same in every case. This well illustrates our theme that it is the abstraction that
counts: imitation involves a common system (as well as systems dependent on the modality), which may be
used in a standard way in higher level activities. A typical use of the phonological system (or equivalent in
other forms of language) is the development of systems to represent language in the standard forms of the
language, i.e. words or other standard units.
One of the targets of the language learner is to build up such a system for representation and imitation. The
move towards use of these representations in connection with reference and meaning provides an example of the
moving target concept, the scenario addressed moving beyond perception and imitation to the connections
between signs and their referents in a communicative context. While a linguistic sign could function as a
trigger for action, the link is more typically indirect, which we interpret as saying its effects go via an abstract
level of representation, allowing greater flexibility in meaning, in particular Jackendoff s non-situation-specific
use of signs. The hyperstructure scheme demands that any tentative links, e.g. based on observing correlations,
be tested by usage, so that for example when a familiar linguistic unit is perceived the corresponding referent
system is activated, confirmation being given by some positive consequence of this activation, this being the