Theory‎ > ‎

Symbolic & Subsymbolic

Wilma Bucci and Sean Murphy

Symbols are entities that refer to other entities and have the capacity to be combined in rule-governed ways, so that an infinite array of meaningful units can be generated from a finite set of elements.  Symbols may be images or words.  According to the classical symbolic approach within cognitive psychology, intelligent beings are symbol systems operating on representations that have the format of symbolic codes.  Until recently symbolic architectures of the mind have been accepted as the dominant approach in cognitive science and artificial intelligence.  This classical symbolic architecture follows the general design of the von Neuman computer and includes some version of processing units such as buffer memories, short-term memory, long-term memory and control structures which oversee the operation and integration of these processing units (Bucci, 1997). 

More recently, models based on contrasting architectures have been posited which have been termed subsymbolic, connectionist, or parallel distributed processing (PDP) models.  These models emphasize representations and processes in which the elements are not discrete, where organization is not categorical and processing occurs simultaneously in multiple parallel channels. Subsymbolic architectures address the essentially infinite array of rapid and complex computations, often carried on outside of awareness, often without explicit metrics, dimensions, or units, in most of the common acts of everyday life -- in entering a line of traffic, taking down a heavy book from a high shelf, or picking up a piece of paper that has fallen behind the desk:

Hundreds of times each day we reach for things. We nearly never think about these acts of reaching.  And yet, each time, a large number of different considerations appear to jointly determine exactly how we will reach for the object. The position of the object, our posture at the time, what else we may also be holding, the size, shape, and anticipated weight of the object, any obstacles that may be in the way--all of these factors jointly determine the exact method we will use for reaching and grasping.  (McClelland, Rumelhart, & Hinton, 1989, p. 4)

Along with the dominance of symbolic architectures, verbal learning and verbal behavior models also dominated the study of human cognition from the time of Watson (1913) until the cognitive revolution of the 1960's.  While verbal and symbolic aspects of mental functioning are important domains of study there are also important areas of mental functioning in the non-verbal and subsymbolic domains that need to be understood.  Since the cognitive revolution, and more recently through technological improvements and imaging techniques, a great deal of progress has been made in understanding non-verbal and subsymbolic processing.  Yet, little work has been done to explain how all four of these domains interact to produce a unified experience of mental life.

The matrix below outlines the interactions of these domains as areas that need to be understood in a comprehensive theory of mental functioning.  Bucci's Multiple Code Theory is one attempt to grapple with these diverse domains.



 Non-Verbal Subsymbolic

Verbal Subsymbolic


Non-Verbal Symbolic
Verbal Symbolic

The verbal symbolic domain largely concerns those aspects of mental functions that pertain to words and language as discrete units of meaning that are strung together to form increasingly complex representations.

The non-verbal symbolic domain concerns images and representations of objects and events that do not have the form of verbal language.  In addition to visual images, these include images and perceptions in other senses. 

The non-verbal subsymbolic addresses many of the experiences noted by McClelland, Rumelhart, & Hinton above, and also pertains to experiences like intuition or the emotional communication between individuals.

The verbal subsymbolic is the domain that has been least recognized and studied.  This concerns the non-symbolic aspects of language such as prosody, meter, rhythm and the phonemic qualities of language.  At least as early as Köhler, there has been mounting evidence that the sounds of words often have some inherent relationship to the objects they represent;  and so words may not be properly understood as purely abstract symbols that are assigned to things in an arbitrary way. 


Consider the images above, originally developed by Köhler (1929; 1947).  If you show these two images to people and say, "one is bouba and the other is kiki, try to guess which is which" 95% will pick the image on the left as "kiki" and the image on the right as "bouba" even though they have never seen these pictures before.  This is only one such demonstration.  A subfield of linguistics known as 'sound symbolism' explores this domain as its primary area of focus.  The neurologist V.S. Ramachandran uses the same example above as an illustration of his theories of the neurological and evolutionary underpinnings of language. (2001, p.19)


Bucci, W. (1997). Psychoanalysis and Cognitive Science: A multiple code theory. NY: Guilford Press.

Köhler, W. (1929). Gestalt Psychology. NY: Liveright.

Köhler, W. (1947). Gestalt Psychology (2nd. Ed.). NY: Liveright.

McClelland, J.L., Rumelhart, D.E., & Hinton, G.E. (1989). The appeal of parallel distributed processing. In D. E. Rumelhart, J. L. McClelland, & the PDP Research Group, Parallel distributed processing: Explorations in the microstructure of cognition (Vol. 1, pp. 3-44). Cambridge: MIT Press.

Ramachandran, VS & Hubbard, EM (2001). Synaesthesia--a window into perception, thought and language. Journal of Consciousness Studies, 8, 3-34.

Watson, J.B. (1913). Psychology as the behaviorist views it. Psychological Review, 20. 158-177.