you're in home | identity | who I am | the book


Connectionism and Meaning: from truth conditions to weight representations

This book was published in 1996 by the academic publisher Ablex.

As befits a research monograph, the book is based largely upon the research I conducted during the course of my doctorate.

Two additional chapters were also included to accommodate subsequent research findings, most notably the "weight-space" formalism for visualising network computation.

The following is an extract from the first, introductory chapter.


Computationally driven, philosophically inspired, and psychologically motivated, this book is a theoretical investigation into meaning, examining and combining concepts from formal and psychological semantics, the philosophy of mind and computational theory within the framework of connectionist cognitive science. The investigation begins by examining the invention of formal calculi, late in the 19th century, and ends by proposing how atomic representations of computational engines can be hooked with what they are representations of - a broad sweep of concerns indeed.

The book is largely about ideas: of what semantics is; of what meaning is, in a general sense, of how language signs have meaning; of how the meanings of words and sentences in language might be mentally represented; and more basically, of how meanings might arise at the ineffable boundary between representations and represented. These, I believe, are fascinating questions, and I hope that the book conveys some of the intellectual excitement generated in considering them.

The book is also, unashamedly, connectionist, drawing upon the novel properties of networks of simple computing elements as its computational metaphor. Although connectionist cognitive science is a young discipline, effectively less than 10 years old, it is an exciting and vigorous one, with the potential of actually delivering what symbolic cognitive science has long promised but never really achieved - a comprehensive theory of mental life. This is not, I believe, misplaced optimism. Again and again, connectionists have been forced to recognise seemingly insurmountable problems with their novel class of computing machinery; that they are merely associative devices, that they lack computational power, that they are unable to support structured representations, that they merely implement symbolic architectures, and so on, and so on. In each case, these seemingly insurmountable problems have been overcome with the resultant generation of truly innovative ideas in the process.

The investigations of meaning that I undertake in this book are in the same pioneer spirit. The reader will find established wisdom hand in hand with novel ideas, but at all times anchored to the actual computational properties of connectionist networks. It is my fervent wish that the book provokes reaction, whether it be surprised agreement or irritated disagreement: both are indications that the interest of the reader has been engaged in the subject matter, which is all that an author can reasonably expect. Much worse than vociferous criticism is bland apathy: being branded boring is indeed a terrible thing.

Identity: who I am
Identity: Who: The Fud
Identity: Who: The Book
Email on yellow