Updated Sept 4, 2000 (R. Port), Linguistics L103, Fall,
2000
People who study signs and communication differentiate three kinds of signs: an ICON from an INDEX from a SYMBOL. This distinction is very important and derives from philosopher C. S. Peirce in the late 19th century. This page is an attempt to sharpen the difference between these three which are described in your text (CELL, pp. 1-4). The critical issue is to appreciate what a symbol is. This is the key to understanding language and how it differs from any animal communication systems.
First, we must note that a sign is a stimulus pattern that has a meaning. The difference between the various kinds of sign has to do with how the meaning happens to be attached to (or associated with) the pattern.
The icon is the simplest since it is a pattern that physically resembles what it `stands for'.
Of course sometimes there could be a dispute about what `physical resemblance' means and how similar it must be. And just because we humans can recognize a little picture of something doesn't mean that any other animal could. (Do you think your cat could recognize a picture of a can of catfood and interpret it as ``Time to eat''? Not likely.) So physical resemblance is by no means a simple concept. But how and why an image or a sound has some particular semantic content for us humans is fairly easy to understand.
An `index' is defined by some sensory feature, A, (something directly visible, audible, smellable, etc) that correlates with and thus implies or `points to' B, something of interest to an animal. All animals exploit various kinds of indexical signs in dealing with the world. The more intelligent animals are good at learning and exploiting more sophisticated indices (thus a cat will use and learn many more indexical signs than a frog, a fish or an ant -- which tend to be restricted to ones acquired innately).
Thus,
Note that all of these above depend on a certain statistical regularity of part A (the signal pattern) with part B (the behaviorally relevant state). The exploitation of this regularity requires first, detecting property A (which is not necessarily simple) and either learning (or innately knowing) its correlation with the B. In that case the animal will use A as an index for B.
Note that for humans, some indices can be artificial and manmade (rather than environmentally natural or innate to particular species):
Notice that the correlation need not be perfect. It isn't always warmer closer to the sea surface, dark clouds in the west don't always mean the rain is coming this way, and even a stoplight can be broken sometimes. This doesn't detract from the usefulness of these signs as a way for an animal to guide its life in a confusing and only partly predictable world.
Words are said to be indexical when they directly point to their meaning - without depending on any relationship to other words. Thus, words like here, there, I, me, you, this, etc. For all of these there is an implied pointing gesture. (Remember in Latin, index really meant the index finger.)
Words as Symbols.
Now, what about a noun word in a human language? Let's say English `KITTY'? Isn't this just a kind of arbitrary index? Isn't KITTY just an index for the presence of a cat (just for English speakers of course)? In support, one might note that a small child and its mother would be likely to say KITTY in the presence of a cat (so there should be some correlation between the cat and the word KITTY). The sounds [kIDi] correlate partially with the presence of cat (so A predicts B). Doesn't that show that this is just an indexical sign like those above? Unfortunately no - even if its true that most early words for children are learned indexically (that is, by pointing to what they refer to). In general, however, it is very rare for the utterance of a word to correlate with the thing it refers to. Sometimes such a correlation exists, of course, buta word in any language is vastly more complex and sophisticated even for language-learning infants. Notice that:
These word-word relationships (sometimes called word-associates) are critical for anchoring the meaning of a word without requiring any correlation in space and time between the signal (the sound of the word) and its meaning. Indices do not require any such set of relationships to work as signs. In summary, symbols like most words in a human language are (a) easily removable from their context, and (b) are closely associated with large sets of other words.
Notice that humans easily learn words for things we have never experienced. Children who grow up in the tropics learn to correctly use words like SNOW and ICE without ever seeing snow or ice. It is not a big problem for them because they have heard descriptions of them in terms of words they do know, like COLD, WHITE, CLEAR, HARD, SOFT, FLUFFY, WATER, MELT, FALLING, SLIPPERY, etc. From these descriptions, they get a pretty good idea what snow and ice are like - enough to read and produce the words appropriately.
This is the enormous power of human symbols: When you have learned a basic vocabulary (based in part on indexical relationships), you can use it to bootstrap to many other new concepts and words. And given the possibility of cultural transmission from generation to generation, human knowledge and understanding become cumulative and have grown at a very rapid rate (relative to the creation and transmission of innate knowledge).
IMPORTANT CLAIM:
Apparently no living nonhuman animals are able to use word-like symbols.
There are, however, some (disputed) claims that a few individual animals (mostly higher primates like monkeys, chimpanzees and gorillas) have been trained by humans to use a small (< 50) inventory of symbol-like units using hand signs or small physical tokens. If this claim is true, it implies a huge divide between humans and nonhuman animals. It means that no animal communication systems can be understood as just `simple versions of human languages'. This claim is daring and provocative, but probably true. [Of course, if one believes that humans are derived from nonhuman animals, then somehow our ancestors must have passed through stages that were intermediate between index-based communication systems (like dogs, monkeys, bees, whales, etc) and modern-human symbolic language even though we have very little direct evidence about how this evolution took place.]
Nonword Symbols.
Words (especially nouns, verbs and adjectives) are the architype for symbols. But the most common use of the term symbol in everyday, nontechnical language is for signs that are not words: eg, a flag or totem animal as the symbol of a country (bald eagle for USA, bear for Russia, etc), a cross for Christianity, star of David for Judaism, swastika for Nazism, a particular type font for a specific product (eg, Coca-Cola, Indiana University, etc).
It seems that a similar set of associations to other words exist for such symbols. Thus, the US FLAG (that is, the graphic pattern in red, white and blue, not the English word FLAG) gets its meaning partly from its association to words and concepts like: HOMELAND, WASHINGTON, BALD EAGLE, PATRIOTISM, MOM, DAD, APPLE PIE, PRIDE, HEROISM, DEMOCRACY, `OH SAY CAN YOU SEE...', `I PLEDGE ALLEGIANCE...', SACRIFICE, etc etc.
Mathematical and logical symbols also get their meaning from their relation to other symbols. Thus pi is defined as the ratio of the circumference of a circle to its diameter: pi = c/d.
So, nonword symbols are much like words but often lack a phonetic form.
The term sign is often used for all three of these: icons, indices and symbols. All have a signal aspect, some physical pattern (eg, a sound or visible shape) and a meaning (some semantic content that is implied or `brought to mind') by the signal. But they differ in that icons have a physical resemblance between the signal and the meaning and an index has a correlation in space and time with its meaning. But a symbol is an arbitrary pattern (usually a sound pattern in a language) that gets its meaning primarily from its mental association with other symbols and only secondarily from its correlation with environmentally relevant properties.