The assumption that a one-to-one mapping of words onto objects and vice versa is the basis for meaning and reference was made explicit in the work of the turn-of-the-century French linguist Ferdinand de Saussure. In his widely influential work on semiology (his term for the study of language), he argued that word meaning can be modelled by an element-by-element mapping between two planes of objects: from elements constituting the plane of signifiers (words) to elements on the plane of the signified (ideas, objects, events, etc., that word is refer to).
On this view, the mapping of vervet monkey alarm calls onto predators could be considered a signifier-signified relationship. But how accurate it does this model word-reference?
Although it is natural to imagine words as labels for objects, or mental images, or concepts, we can now see that such correspondences alone collapses a multileveled relationship into a single mapping relationship. It fails to distinguish between the rote understanding of words that my dog possesses and the semantic understanding of that in normal human speaker exhibits.
We also saw that the correspondence of words to reference is not enough to explain word meaning because the actual frequency of correlations between items on the two planes is extremely low. Instead, what I hope to show is that the relationship is the reverse of what we commonly imagined.
The correspondence between words and objects is a secondary relationship, subordinate to a web of associative relationships of a quite different sort, which even allows us reference to impossible things.
In order to be more specific about differences in referential form, philosophers and semoticians have often distinguished between different forms of referential relationships. Probably the most successful classification of representational relationships was provided by the American philosopher Charles Sanders Peirce.
As part of a larger scheme of semiotic relationships, he distinguished three categories of referential associations: icon, index, and symbol.
These terms were, of course, around before Peirce and have been used in different ways by others since.
Peirce confined the use of these terms to describing the nature of the formal relationship between the characteristics of the sign token and those of the physical object represented. As a first approximation these are as follows:
icons are mediated by a similarity between sign an object,
indices are mediated by some physical or temporal connection between sign an object,
and symbols are mediated by some formal or merely agreed-upon link irrespective of any physical characteristics of either sign or object.
These three forms of reference reflect the classic philosophical trichotomy of possible modes of associative relationship:
(b) contiguity or correlation, and
(c) law, causality, or convention.
The great philosophers of mind, such as John Locke, David Hume, Immanuel Kant, G.W.F. Hegel, and many others, had each in one way or another argued that these three modes of relationship described the fundamental forms by which ideas can come to be associated.
Peirce took these insights and rephrased the problem of mind in terms of communication (transmission of signs), organised by an underlying logic (or semiotic, as he called it) that is not fundamentally different for communication processes inside or outside of brains.
If so, it might be possible to investigate the logical thought processes by studying the sign production and interpretation processes in more overt communication.
To get a sense of this logic of signs, let's begin by considering a few examples.
When we say something is “iconic” of something else they usually mean that there is a resemblance that we notice. Landscapes, portraits and pictures of all kinds are iconic of what they depict.
When we say something is an “index” we mean that it is something causally linked to something else, or associated with it in space or time. A thermometer indicates the temperature of water, a weathervane indicates the direction of the wind, and a disagreeable odour might indicate the presence of a skunk. Most terms of animal communication have this quality.
Finally, when they say something is a “symbol”, we mean there is some social convention, tacit agreement, or explicit code which establishes the relationship that links one thing to another. A wedding ring symbolises the marital agreement; the typographical letter “e” symbolises a particular sound used in words (or sometimes, as in English, what should be done to other sounds); and taken together, the words of this sentence symbolise a particular idea or set of ideas.
No particular object or intrinsically icon is, indices, or symbols. They are interpreted to be so, depending on what is produced in response. In simple terms, the difference between iconic, index a call and symbolic relationships derive from regarding things either with respect to their form, the correlations with other things, all their involvement in systems of conventional relationships.
The diminishing competencies of the species corresponds with interpretations that are progressively less and less specific and progressively more and more concrete. But even at the bottom of this descent there is a possibility of a kind of minimalistic reference.
This demonstrates one of Peirce's most fundamental and original insights about the process of interpretation: the difference between different modes of reference can be understood in terms of levels of interpretation.
Attending to this hierarchical aspect of reference is essential for understanding the difference between the way words and animal calls are related. It's not just the case that we are able to interpret the same sign in different ways, but more important, these different interpretations can be arranged in a sort of ascending order that reflect a prior competence to identify a high-level associative relationships. In other words, reference itself is hierarchic in structure; more complex forms of reference are built up from simpler forms.
But there is more to this than just increasing complexity. This hierarchical structure is a clue to the relationships between these different modes of reference. Though I may fail to grasp the symbolic reference of the sign, I might still be able to interpret it as an index (as correlated with something else), and if I also fail to recognise any indexical correspondences, I may still be able to interpreted as an icon (recognise its resemblance to something else). Breakdown of referential competence leads to an ordered descent from symbolic to indexical to iconic, not just from complex icons, indices or symbols to simpler counterparts. Conversely, increasing sophistication of interpretive competence reverses the order of this breakdown of reference.
The hierarchic relationship between the three fundamental forms of reference – iconic, indexical, and symbolic:
Symbolic relationships are composed of indexical relationships between sets of indices and indexical relationships are composed of iconic relationships between sets of icons.
This suggests that kind of semiotic reductionism in whicg more complex forms of representation are analyzable to simpler forms. In fact, this is essentially what occurs as forms are interpreted. Higher-order forms are decomposed into (replaced or represented by) lower-order forms. Inversely, to construct higher representation, one must operate on lower order forms to replace them (represent them).
In Peirce’s terminology, each is an interpretive process, and the new signs substituted for the previous signs at a different level are “interpretants” of those prior signs.
We do not lose the indexical associations of words, despite the lack of correlation with physical reference, because the possibility of this link is maintained implicitly in the stable associations between words. It is by virtue of this sort of dual reference, to objects and to other words (or at least to other semantic alternatives), that the word conveys the information necessary to pick out the objects of reference. This duality of references captured in the classic distinction between sense and reference. Words point objects (reference) and words point to other words, but we use the sense to pick out the reference, not vice versa.
This referential relationship between the words - words systematically indicating other words – forms a system of higher-order relationships that allows words to be about indexical relationships, and not just indices in themselves.
But this is also why words need to be in context with other words, in phrases and sentences, in order to have any determinate reference. Their indexical power is distributed, so to speak, in the relationships between words. Symbolic reference derives from combinatorial possibilities and impossibilities, and we therefore depend on combinations both to discover it (during learning) and to make use of it (during communication).
Even without struggling with the philosophical subtleties of this relationship, we can immediately see the significance for learning. The learning problem associated with symbolic reference is a consequence of the fact that what determines the pairing between a symbol (like a word) and some object or event is not the probability of co-occurrence, but rather some complex function of the the relationship that the symbol has to other symbols. This is a separate but linked learning problem, and worse yet, it creates a third, higher-order unlearning problem. Learning is, at its base, a function on the probability of correlations between things, from the synaptic level to the behavioural level. Past relations tend to be predictive of future coalitions. This, as we have seen, is the basis for indexical reference. In order to comprehend the symbolic relationship, however, such indexical associations must be subordinated to relationships between different symbols. This is a troublesome shift of emphasis.
The construction of symbolic referential relationships from indexical relationships:
A collection of different indices are individually learned.
Second, systematic relationships between index tokens (Indexical stimuli) are recognised and learned as additional indices.
Third, a shift in mnemonic strategy to rely on relationships between tokens to pick out objects indirectly via relationships between objects. Individual indices can stand on their own in isolation, but symbols must be part of a closed group of transformations that links them in order to refer, otherwise they revert to indices.