Notes on icons, indices and symbols with personal interpretation, based on Deacon, Terrence (1997) `The Symbolic Species; The Coevolution of Language and the Brain' (Norton). For some relevant excerpts from The Symbolic Species, see the associated page.

Robert Port, Oct, 1999
www.cs.indiana.edu/~port/teach/q500/deacon

ICON.

Defined by a distinct physical pattern, that is, a `type' or category. Tulips strongly resemble each other physically and comprise a distinct category of physical object. Multiple productions of the word `tulip' (/tulIp/) bear direct resemblance to each other. But these are two icons - a visual one and a phonetic one. The visual Tulip can be generalized across dimensions like color, size, angle of view, etc. And the phonetic `tulip' can be generalized across speaker voices, voice pitch, etc. Icons are simply `perceptual categories'.

INDEX.

Often one icon (eg, the word `fire') correlates with another icon or category or set of them (eg, the appearance and smell of burning material). If the correlation is strong, then, eg, the smell of smoke becomes an Index for fire. The smell is a `sign' of fire in most contexts. Notice that an index is always based on at least two icons - one at each end, so X is an index for Y. Roughly, indices are `learned associations'.

Examples:

1. In animal expts, a bell can be an index (or sign) for food presentation if there is a temporal and spatial correlation between them. Similarly, `a patch of light' across the floor of my parking garage is my personal index for `empty parking space ahead'.

2. Some household pets (dogs, cats) can acquire a (small) set of English word icons and use them as indices for matters of interest: `dinnertime', `take a walk', etc. (Still, there is a good chance that the actual sign here is NOT the English word, but something in the tone of voice coupled with looking at the pet directly, etc. But SOME pets probably learn SOME English words.)

3. Bees `know' an indexical relation between features of a dance within the hive and the direction and amount of a food source. This is indexical, but apparently innately acquired.

Note 1. Indices are intrinsically independent of each other; Hierarchical relationships are completely outside this picture. There may be multiple icons correlated with `fire' (visible smoke, smokey smell, certain colors and lights, fleeing animals, etc), but the fact that an animal may eat banana, apple, bread, etc, does NOT imply the existence of a category `the things-that-I-eat'. If an animal is hungry, of course, it will seek ANY of the the things that are food for it. But that does not imply the existence of a grouping of indices into a category or class that might itself have a name or index (as opposed to an index for any of the specific foods).

Note 2: An index depends on correlation (either learned or genetically acquired) for creation and maintenance of the index relation. Thus, an index can not normally be used for counterfactuals. So, bees cannot mislead other bees and dogs cannot talk about (or think about) topics named by words they know.

SYMBOL

So far, this discussion is based on Charles Sanders Peirce (sic, not Pierce, the more common spelling). But Deacon has his own story to tell about real symbols.

1. The architypal symbol is a content word (eg, noun, verb, adjective) in any human language. (Im not so sure what to say about the status of function words like `in, this, a, because, etc'.)

2. Just as an Index is constructed from two (or more) Icons, a Symbol is constructed from two or more Indices (says Deacon).

3. So symbols get their meaning not just from a relationship between a phonological pattern (eg, /tsher/) and sensory icons of chairs, but ALSO from various kinds of associations with other words (eg, table, floor, desk, lamp, stand, sit, dinner, eat, wood, etc). As noted by Saussure a century ago, there are two axes here: the syntagmatic axis, of relationships within strings of words as they are spoken together, and the paradigmatic axis, the relations of words to other words not normally occurring in the same stream of speech.

4. This independence of the phonetic icon of a word from the direct relation to its semantic icons implies:

a) the statistical correlation between sound icon and sensory icon need not be maintained. Indeed,

b) no semantic icon need ever have existed (eg, unicorn or dragon. We can learn a word just by hearing it and using it with other words)

c) we can use words counterfactually (in the absence of the icons for its meaning). Disconnection from meaning is no problem. So we can lie, mislead, hypothesize, fantasize, idealize, etc.

5. But this dependence of a word on other words (of symbols on other symbols) implies:

a) all words tend to have a large set of `associates' - some of the other words that the target word is `defined with respect to'. So, `Pharoah' is associated with `Egypt, king, first millenium BC, Nile, Ra, etc'. These associates seem to be what Tversky's contrast model works with.

b) Words have a tendency frequently to organize themselves into various organization groups, including hierarchic sets (eg, basic level categories, superordinates, subordinates, and soldier < squad < platoon, etc.). Other sets include:

a) I, you, he/she/it; we, y'all, they ;; present, past, future

b) king, queen, prince, princess; mommy, daddy, gramma, grampa; big-small, True-False, inside-outside

c) eat (using knife, fork, plate) food (eg, bread, meat, fruit) at (table, restaurant, birthday party)

From this perspective, then, frames seem so plausible to us because they are built from words which are already defined (in part) with respect to each other. It seems possible that frames are not the ``underlying organization of human knowledge,'' but just reflect the way that words tend to organize themselves as we acquire competence with our vocabulary. They may be just the epiphenomena of a rich (human) vocabulary.

c) Obviously, these word-word organizations reflect in part the structure of the world (hill-valley, mother-father) but also significantly reflect specific human cultures and their history (bowl-spoon, House-Senate, body-soul).

6. PHYSIOLOGICAL SPECIALIZATION. Deacon maintains that:

a) symbols exist only in human brains, because

b) all these word-word and word-meaning relationships require specialized neural architecture based on pathways from prefrontal cortex to many other brain regions. Since no other species have expanded prefrontal cortex and its recurrent pathways, symbols are extremely difficult for them to acquire in any form, though they are natural and unavoidable in humans.

Notes by Robert Port, Oct, 1999