Two Questions About the Origins of Connectionism

The connectionism that most people know about become popular in the 1980s.  It is defined, more or less, as the attempt to explain or model cognitive functions using artificial neural networks.  The name “connectionism” for this movement goes back at least to a 1982 article by Feldman and Ballard (“Connectionist Models and their Properties, Cognitive Science 6: 205-254). 

Of course, the use of neural networks to explain or model cognitive functions is much older, going back at least to the work of Rashevsky in the 1930s.  But that is not what I’m after.  I’m after the use of the term “connectionism”.

The term “connectionist” was used in psychology much earlier than the 1980s, but it used to mean something slightly different (though related).  It used to mean, more or less, the explanation of learning and memory in terms of changes in connections either between stimuli and responses or between neurons.  It was used in this way most prominently by Thorndike (The Fundamentals of Learning, 1932; Selected Writings from a Connectionist Psychology, 1949) but also, very importantly, by Hebb (The Organization of Behavior, 1949), among others.

The main link between these two literatures appears to be the work of Rosenblatt.  In 1958, while introducing his Perceptron (an early kind of artificial neural network), Rosenblatt says that his theory takes the “connectionist” position–clearly using the term in the sense in its early sense (“The Perceptron: A Probabilistic Model for Information Storage and Organization in the Brain,” Psychological Review 65: 386-408).  Rosenblatt’s work links the two literatures because his Perceptron is the main ancestor of many contemporary artificial neural networks, most of which are “connectionist” in the early sense of the term.

Question 1: does anyone know whether any of the other artificial neural network people from the 1950s (e.g., Farley and Clark, Rochester et al, Widrow and Hoff) ever used the term “connectionist” or “connectionism”?

Question 2: does anyone know whether anyone before Feldman and Ballard used the term “connectionist” or “connectionism” in the modern sense, i.e. to refer to the explanation or modeling of cognitive functions using artificial neural networks?  (For instance, do Minsky and Papert use either term in the 1969 edition of their book, Perceptrons?)

Back to Top