Abstraction and Organisation in signs and Sign Systems

William Edmondson

Cognitive Science Research Centre, School of Computer Science, The University of Birmingham, Edgbaston, B15 2TT, West Midlands, Britain.

email:w.h.edmondson@birmingham.ac.uk

Abstract

The call for participation in the Brighton Workshop on Iconic Communication was a challenge to commit to paper some more general thoughts on a topic in which my interest had been inspired, and subsequently focused, by many years research on sign language linguistics. I developed some material for a seminar presentation, and following acceptance of the abstract for the Brighton workshop the seminar material was revised and is presented here. The theme of the presentation was, and is here, analytical in its approach to the concepts of index, icon and symbol, and ultimately doubtful of the feasibility of developing a system of icons for human communication. An additional perspective here concerns the use of icons in interfaces.

1 Some questions

In the call for participation the following paragraph appeared.

Iconic communication is the attempt to build cross-language communication systems that completely avoid the use of words and rely solely upon pictorial symbols. The main interest is in general purpose person-to-person communication systems but there is also interest in purely iconic interfaces to existing software.

This gives rise to three concerns. Is 'iconic communication' possible? Can the problems with terminology be resolved? Is the main purpose (as stated) important?

I will address these questions in reverse order so as to lead up to the main question, and also because the last is relatively easily dismissed. Attempts to find some sort of pan-cultural, universal, language or communication system (e.g. Esperanto, Bliss Symbols) have often been accompanied by absurd, almost magical, claims. Examples from Dreyfuss' book (Dreyfuss, 1972) illustrate this well.

In the great overall evolutionary trending of humanity's gradual learning to produce ever more with ever less, it is implicit that the present discoveries of the electromagnetic behaviors of the brain and its local nerve system controls by mind will eventuate in telepathy's being graduated from society's assessment of it as mystical-magical phenomenon to an everyday communication facility.

Henry Dreyfuss' contribution to a new world technique of communication will catalyse a world preoccupation with its progressive evolution into a worldian language so powerfully generalized as to swiftly throw into obsolescence the almost fatally lethal trends of humanity's age-long entrapment in specializations and the limitations that specialization imposes upon human thinking. [From R. BuckminsterFuller's foreword to Dreyfuss, 1972]

Many of us dream of one world, devoid of geopolitical boundaries and futile bickering. Our astronauts, alone in outer space, must sense continually the importance of all men being brothers.

In an infinitesimal way, this book is an effort to help us bring that concept into fulfillment. Communication - people to people, nation to nation - is a vital ingredient to understanding .... new symbols are springing up daily. But as these symbols multiply, confusion, contradiction, and duplication become rampant. Hopefully, with this Source book as a start, standard symbols will some day be understood by all, regardless of language or culture.
[From the introduction to Dreyfuss, 19721

But the symbols [Bliss-Symbols] belong only to Aspect 1 of my work. The other 5 aspects are even more fascinating. Aspect 2: a simple symbolic logic which even children can learn to use in their daily problems, and which would help them later as husbands and wives to avoid unnecessary quarrels. Aspect 3: a simple semantics which would help even children to recognize (and avoid) those dangerous words by which demagogues and dictators in the homes and nations threaten the peace of mankind.

By trying to find appropriate symbols for mankind's most important meaning like ethics, evolution, life, liberty, religion, God, etc. etc. I made some important discoveries already acknowledged by scholars. Aspect 4: a universal natural ethics encompassing all religions. Aspect 5: the biochemical discovery that cells act ethically in all creatures. Aspect 6: the archeological discovery that old stone age man was not a killer of his fellowman. Cannibalism and war began with new stone age man by the introduction of dangerous words.
(From an introductory article in Dreyfuss, 1972, by Charles K. Bliss)

2 Terminology

We can now focus on our main purpose in this paper, namely the exploration of the feasibility of general purpose non-verbal communication systems. We begin by considering symbols and symbol systems from a terminological perspective. Two dimensions of description are useful when talking about signs and sign systems. These can be illustrated:

 

The terms pragmatics, semantics and syntax are widely used in semiotics. They can simply be defined as the relationship between signs and i) their users and the circumstances of use; ii) their referents; iii) each other. The intended implication is that these three aspects of signs are independent of one another, which would imply further that we should consider them to be the three axes of a descriptive framework. Instead, the suggestion here is that in effect they are one axis, as arranged, where comprehensiveness of characterisation of any system or language is indicated by location on the axis, with the origin to the left of 'pragmatics'.

The order chosen for the linearisation of the three terms is one which is often implicit in discussions of them, and of sign systems. The orthogonal dimension under semantics, above, is the probable underlying cause for this - indices are somehow very tied to the circumstances of use, and they are 'self-evidently' the simplest type of sign, semantically speaking. Icons are obviously about semantics - they are supposed to have 'self-evident' meanings. Symbols are obviously more abstract and difficult to discuss, from a semantic perspective, and derive their utility in part from the conventional or arbitrary patterns of their organisation. This sort of qualitative assessment of the three domains in semiotics surfaces also in the relative emphasis given to them - syntax is the most formalised and contentious component of linguistics; pragmatics is 'soft' and still looking for a formalism! A more subtle categorisation of signs is available (Sebeok, 1976) and this can be construed as an elaboration of the pragmatics, semantics, syntax axis:

 

Signals and symptoms can be discussed very briefly they are the inflexible components in 'stimulus (signal) - response (symptom)' situations. This end of the scale is obviously not the difficult end. Naming, on the other hand, is also readily appreciated for what it is: Sebeok's definition of a name is 'a sign which has an extensional class for its designatum'. His ordering, reflected above, locates name to the right of symbol by virtue of its arbitrariness, and its uninterpretability (you either know the name, or you don't). Its location on the scale is not of great moment; it serves, like signal and symptom, to draw out from the three main categories of sign some 'special cases' which might otherwise obscure the discussion. We can now turn our attention to these categories, not just with the presumption of relation in mind but, as implied above, also the need to examine that relation.

3 Index, Icon and Symbol

Sets of definitional statements are a useful prelude to further discussion. We will consider first the definitions - 3.1 - and then some problems not addressed in the definitional approach - 3.2.

3.1 Definitions
3.1.1 Index

An index represents its object by virtue of 'being really effected' by it, that is, by being in a dynamical or causal relation to it.
Peirce, summarised in Greenlee, 1973:70

A sign is said to be indexic insofar as its signifier is contiguous with its signified, or is a sample of it. Sebeok, 1976:131

Indices are 'those whose relation to their object consists in a correspondence in fact'. Peirce, quoted by Sebeok, 1976:128 

Examples of indices:

'Natural Signs' such as footprints (Peirce), weathervanes (Greenlee), or 'Polaris ...an index of the North celestial pole...' (Sebeok).

3.1.2 Icon

An icon 'represents its object by virtue of a character which it possesses regardless of whether or not the object exists'. Peirce, summarised in Greenlee, 1973:701 

A sign is said to be iconic when there is a topological similarity between a signifier and its denotata. Sebeok, 1976:1281

Icons are 'those whose relation to their objects is a mere community in some quality'. Peirce, quoted by Sebeok, 1976:128

Examples of icons:
Maps. Sub-classes: images, diagrams, metaphors.

3.1.3 Symbol

A sign without either similarity or contiguity, but only with a conventional link between its signifier and its denotata, and with an intensional class for its designatum, is called a symbol.
Sebeak, 1976:1341

'A Symbol is a sign which refers to the Object that it denotes by virtue of a law, usually an association of general ideas, which operates to cause the Symbol to be interpreted as referring to that Object.'
Peirce, quoted by Greenlee, 1973:93/

 

According to Aristotle 'a noun signifies this or that by convention. No sound is by nature a noun: It becomes one, becoming a symbol' 
Greenlee, 1973:941 Examples of symbols: Words, and letters of an alphabet.

 

3.2 Problems with the definitions
3.2.1 Problems with the notion of index

Figure 1. 

The illustration in Figure 1 is an example of an intended index. The drawing is of a control device, and the problem is one for the user - how is the control used? How does the user get from la to lb and what expectations might s/he have?

Electrical rotary controls usually involve less than one full turn. They usually involve clockwise motion for increased value, with anticlockwise motion decreasing the value, sometimes including an on/off switch at the point of minimum value (e.g. the volume control on some radios). Water controls, in contrast, mostly work 'the other way', but there is a newish variant to confuse the issue. The introduction of ceramic disc quarter-turn mechanisms for conventional use (other than in mixer taps, etc.) permits a simple tap handle to rotate in either direction, but only by a quarter turn, to turn on or off the flow. There is no physical force required, just rotation. More conventional taps incorporate the usual 'right hand thread' for the parts and require anticlockwise rotation of the handle to increase the flow. Clockwise motion decreases the flow, and in some cases force may be required to screw the handle down tight enough to stop the flow altogether.

The control device illustrated is for an electric water heater; the one control is used both for flow and heat. The apparently more obvious indexical functionality - shown in lb - is ambiguous because it is not clear whether the numbers refer to flow or (redundantly) to heat. More problematically, how does one turn it on or off?

The answer is relatively simple. Increased flow is at the expense of heat - the numbering redundantly repeats the colour-based information and flow rate is not indicated at all. The 'start' and 'stop' arrows indicate direction, not position - off is achieved by reducing flow, through maximum heat, by rotating the knob clockwise to the off position. It is noteworthy that the electrical intensity convention (clockwise to increase) is used simultaneously with the water flow convention (anticlockwise to increase). The numbering would thus be equally confusing if it ran the other way around. The designer's problem is clear - two opposing conventions are to be combined in one control. The user can be expected to be familiar with both, and is aware that both may be relevant to this particular control. Confusion is inevitable. Once the water is flowing, however, the indication for heat level is clear, and unambiguous, which is perhaps the safest compromise. There is no index for flow level.

Put more concisely, in terms of the discussion being developed, we can say that the user's problem is that the indexical functioning of the knob, in relation to the dial, is symbolically conditioned by the conventions of electrical and fluid controls. These conventions cannot both be unambiguously expressed in the single control, so the interpretability of the index must be learnt anew. The knob has to be 'read' as an electrical control or a fluid control, and thus is a symbol - in fact two symbols, because the control is ambiguous. In reality the quality of the symbols is not equivalent - the control is a clearer symbol of 'electrical control' than it is of 'fluid control', and this is safe but unfortunately unhelpful for getting the water to flow.

3.2.2 Problems with the notion of icon

Icons present several difficulties, and these all reflect definitional inadequacies of one form or another. The first and most obvious of these problems concerns the putative property of icons - iconicity. 'Topological similarity" is not an objectively definable property of a sign in relation to its referent. Iconicity is the perceived relationship between a sign and its referent; it is in the eye of the beholder. Greenlee (1973: 77) points out that conventionality is relevant - not conventionality in the sign itself, although this may also be a factor in some cases, but conventionality in the potential for its relation to the referent (and of course the same observation holds of the perception of the referent). In other words, convention guides the eye in relation to both sign and referent, to potentiate iconicity.

Sebeok has indicated other causes for concern with iconicity - symmetry and regression. The issue of symmetry arises when one considers chronological priority. A painting of a person is an icon. A reproduction of that painting is an icon of the first. A photo of the repro ....... and so forth. This is not controversial but does not surface in definitions. The regression problem is one of scope - it seems similar to the asymmetry problem but is actually different. The latter concerns just one sign-referent pairing (the set of photos of repros... is chosen to make the point in exaggerated fashion). The regression issue is that of long chains of reference. A child is an icon of its parents, and of their parents.... and of Adam and Eve. Sebeok's two concerns can probably be addressed by elaborating Greenlee's observation about convention potentiating iconicity. This work is not done here; it is complex because formal statement of the elaboration will require consideration of signs in their contexts, and then further elaboration must deal with the case of signs being taken in the context of other signs, as well as the notion of chronological priority. However, it is worth quoting Greenlee on this, if only to see how much elaboration remains to be done (op.cit.:7778).

In order for a sign to designate an object through a likeness or similarity, the sign must have been APPOINTED, as it were, to serve as a representative of the sort that it does. A convention or rule must have been established to the effect that the sign should be interpreted to signify in a certain way, which is through some respect of resemblance. A CONVENTION OR RULE OF INTERPRETATION MUST ESTABLISH THE GROUND OF THE REPRESENTATION (AND ACCORDINGLY SIGNIFICATION) OF THE ICON.

Note lastly that if iconicity is not perceived, despite all the potentiation, then the 'icon' is an arbitrary sign, a symbol.

3.2.3 Problems with the notion of symbol

As with icons, there are some definitional difficulties with symbols. Sebeok's definition points again to the use of convention, but by now this concept looks in danger of being overworked. It is not, it is misunderstood; the use of the term convention here relates to arbitrariness and agreement about it. A failed icon, of the sort mentioned above in 3.2.1, if its perpetuated use is agreed upon, is a symbol. Its meaning would have to be explained in terms of other signs (the intensional component in Sebeok's definition). But, how do we recognize arbitrariness, to whom is it recognisable, and is the whole point indeed that arbitrariness itself is recognised as an essential characteristic of a symbol - when it is recognised as a symbol?

Sebeok seems to have missed this point. His discussion of arbitrariness in non-human symbol use includes examples of/ail-wagging (dogs - friendship; cats - hostility; horses - flies) and tail holds. In rhesus monkeys a horizontally held tail indicates fear, but in baboons 'fear is conveyed by a vertical tail' (and in passing one could add that in cats the vertical tail indicates friendly greeting). However, in baboons the tail may also be vertical for other reasons (grooming; a balancing aid for infant baboons on the mother's back). My point here is that we can see the arbitrariness (the possibility that a behaviour could be otherwise) - both within our species and across species - but do the non-humans also see it, and if not is the behaviour really arbitrary (it could be genetically programmed)?

It could be argued that the above question is related to another - multiple meanings for symbols. A speaker's use of uncommon words, for example with Latin roots or perhaps of foreign origin, might be taken as a symbol for class, or race, or authority, or gender... This, however, is itself arbitrary and not at all like a baboon not responding to a rhesus monkey's tail orientations. It could be the case, of course, that we do not yet know enough about non-human behaviour to know that it is not arbitrary in the acknowledged sense above.

3.3 Coherence

The preceding discussion shows that despite the obvious validity of the different definitional perspectives, captured in the terms pragmatics, semantics and syntax, it is clear that indices, icons and symbols relate both to each other and to the semiotic categories - there seems to be a clear possibility of intermingling, if not a straightforward progression from index through icon to symbol.The single most important factor in this is probably the ubiquitous role played by convention. It is convention which makes plausible the notion that the semiotic categories are not underlyingly orthogonal. The interrelatedness of indices, icons and symbols is also evidenced in the discussion of sign systems.

4 Systems of signs
4.1 Systematicity and indices

Indices can be systematised - sets of dials and controls can be identified by virtue of shape and/or arrangement (even colour). They can be read in clusters also. Note, however, that typical groups of dial or knobs or levers, e.g. the engine throttle controls in an aircraft cockpit, have an iconic value and a symbolic value as well as an indexical value.As with other types of signs it is important to recognize that isolated instances can be misleading - they don't tell the full story.

4.2 Systematicity and icons

Interestingly, icons often appear more symbolic in isolation than they do when presented as part of their set. Here the influence of the systematised set upon its members is the opposite of the case with symbols, and in addition because the set confers iconicity on its members so those individual icons can become more abstract (symbolic). Dreyfuss (1972) provides some sets of symbols for animals (Figure 2) and grains (Figure 3) which illustrate this point well. Another example of this effect is seen in the set of signs used for electronic circuits (Figure 4).The icon for resistor has changed, over time, to become essentially abstract and symbolic; its iconic properties are read off the circuit context.This potential for change is important - we will return to it below.

CLICK ON IMAGE TO ENLARGE.

Figure 2: Illustrations of symbolsforgrains, and related concepts. Dreyfuss' Symbol Sourcebook) page 39.

and

Figure 3: illustrations of symbolsfor Livestock. Dreyfuss, Symbol Sourcebook,page 40 ~ 41.

CLICK ON IMAGE TO ENLARGE.

Figure 4.

4.3 Systematicity and symbols

An important aspect of the concept of arbitrariness is that it is a property which is incompletely demonstrated in isolated symbols - it is therefore confusable with randomness. A symbol system such as a spoken language demonstrates arbitrariness comprehensively. The class of spoken languages, for example, can be described as the set of arbitrary subsets of the set of possible vocal sounds.These sounds are arbitrarily organised into groups - syllables comprising one or more sounds. The individual sounds (in groups of one or more - but not necessarily syllables) are also arbitrarily assigned meanings - these are morphemes. Morphemes can be assembled (arbitrarily) into larger units, and etc. The systems provide structure - arbitrariness not randomness - and this cannot be divined from symbols in isolation.This patterning of arbitrariness is known as 'duality of patterning' or 'double articulation'; the sounds pattern arbitrarily and independently of the meanings which also pattern arbitrarily.

4.4 Systems as signs

There is another point about systems of signs which needs to be stressed - it too is overlooked in most discussions. Such systems - whether a written language or circuit diagrams - have very complex structures which can also be iconic or symbolic. This is not just a rephrasing of the interrelationship between a sign and its cohorts in a particular set, mentioned above. Rather, the problem is that arrangements of signs are themselves signs, with conventional aspects to their interpretation. This is too frequently taken as a defining property of language(s), but in fact it is a general property of sign systems, whether the signs are symbols or icons.

Consider again the circuit diagrams in Figure 4. Three circuits are illustrated. Such diagrams have three 'readings' - a static or structural reading, a dynamic reading, and more obscurely a chronological reading. These depend upon the reader (i.e. convention) for successful interpretation. Other examples - musical notation and programming languages, to identify just two - are not difficult to find. The concern here is that this sort of 'double articulation' (systems of systems) looks uncommonly like that usually invoked in relation to symbols (especially languages), and indeed, in musical notation we find the requisite arbitrariness for the characteristic to be clearly identified. In circuit diagrams the arrangements may be iconic, and very effectively so, even though the separate signs are themselves often symbolic. This need not concern us; the point really is that signs can be composed of signs, and that convention is required, as for the case of signs in isolation or in systems. Clearly this compositionality is not bounded - sign systems can be composed of signs each of which is the product of a sign system.

5 Change in signs and sign systems
5.1 Changes in individual signs

We begin our discussion by taking up an earlier point - the possibility of change in the form of a sign. Signs can change for reasons which are not significant from a semiotic viewpoint. Studies of writing, for example, show changes over time - cuneiform writing is well documented (see Figure 5.) It is not difficult to see the force of articulatory simplification at work, and this yields abstract versions of earlier signs. Note that this articulatory argument does not go beyond the individual sign. Another 'force for charge' should not be ignored, and also provides a reminder to be careful with the notion of iconicity. The early cuneiform sign for barley compares favourably with Dreyfuss' symbol (see Figure 3). However, the span of 5000 or more years which separates these symbols should urge caution rather that inspire wonder. How do we know what Sumerian barley looked like? Indeed, when we look at the range of cereal available today, and at the evidence we have concerning early versions of those grains, we can only wonder at the temerity of those who would assume from a single pictograph that a Sumerian would recognize as barley the grain now grown. In other words, icons may change because their referents may change.

 CLICK ON IMAGE TO ENLARGE.

Figure 5: Illustration of the development of cuneform characters,from Diringer (1968), vol 2, page 19.

5.2 Systematic change

Earlier it was noted that systems or sets of signs have properties or characteristics as systems which exert evolutionary pressure on the forms of the individual signs comprising the larger group. This may happen, in principle, in two ways. The system as a collection of arbitrary symbols may evolve and this will surface as changes in the individual symbols (see the Greek alphabet, Figure 6 - horizontal development). Alternatively, the system in use may evolve, and this likewise can surface as changes in the symbols (Figure 6 - vertical columns; also, standardisation, printing etc., for electronic circuits). The reasons for these drifts in form will vary (can we ever know them?). Aesthetic factors, writing materials, articulatory pressures may all contribute to the evolution of sign systems, just as for individual signs, but they will do so in a systematic way, as a migration to a set with sufficiently clear differences between all the signs (articulatory and/or perceptual). Likewise, migration can be towards sets which are themselves composed of organised sub-sets (speech sounds). Whilst some sets do not exhibit change (Morse code, semaphore) some others do, and occasionally the changes can be described and attributed to some cause. Individual signs in American Sign Language have changed and articulatory and perceptual factors are implicated, but it seems that the real change is that the organisation of the language, as a system, has evolved (see Klima and Bellugi, 1979).

  CLICK ON IMAGE TO ENLARGE.

Figure 6: Illustration of the development of the greek alphabet, from Diringer(1968),vol 2, pages 306 & 307

5.3 Rate of change 

The discussion thus far has considered change in signs and sign systems which has occurred over tens of years, or perhaps thousands. There is, in principle, no reason to suppose that more rapid change cannot occur, and as will be argued below every reason to suppose that potentially change is an option every time a sign is used. Signs systems which emerge from time to time undergo change in their early stages, or emerge in several different versions. Systems of notation for signed languages are one example, indices, icons and symbols on control panels are another. Special systems for special needs Blissymbolics and Makaton, for example - also undergo changes (cf. Kiernan et al., 1982:11).

6 Abstraction and organisation in signs and sign systems
6.1 Summary

We can summarise the account so far and draw some conclusions regarding the prospect for contrived systems and for human-computer interfaces. 

6.1.1 Convention

The notions of index, icon and symbol seem usefully distinct, although not unproblematic. Of particular note is the ubiquitous dependence on conventions of one sort or another - the three classes of sign may differ in arbitrariness or abstractness but conventionality is always involved.

6.1.2 Change

Indices, icons and symbols are not immutable. More fundamentally, they are objects of perception, not triggers and responses. Perception is an active, context sensitive business and this facilitates (even drives) change. Three points need noting.

6.1.2.1 Arbitrariness, especially when 'layered' in 'double articulation' (as in speech sounds) is probably incorporated into systems of signs to prevent (or slowup) drift in form or meaning, perhaps consequent upon 'reperception' or articulatory simplification.

6.1.2.2 Pressures for change, even in arbitrary systems, may reflect the need to maximise differences between signs, if not articulatory or aesthetic factors.

6.1.2.3 Immutable systems (Morse, semaphore) are contrived recastings of pre-existing systems.

6.1.3 Systems 

Systems of signs have properties independently of the individual signs (or components) which make up such systems (this is consequent upon the generalized notion of 'double articulation').

6.2 Abstraction versus organisation

The possibly unexpected conclusion here is that the situation regarding abstraction and organisation of signs and sign systems is one of conflict.

6.2.1 Abstraction

Abstraction is, in effect, perception - an unavoidable aspect of a human's response to objects (both signs and their referents). Every perception is potentially an abstraction, a re-representation as an index, icon or symbol. This is not true of signal, symptom, or name. This process of perception as abstraction has no time scale; every perception therefore has the potential to introduce a change.

6.2.2 Organisation

Arbitrariness is the most extreme form of organisation (note also that it is never random). Organisation is probably deployed to block, in so far as this is possible, the inevitable effects of perception. Organisation layers stability onto the unstable results of perception/ abstraction. The evidence is all around us - languages change, as do symbols. Convention, as part of the implementation of organisation, is required, but cannot prevent drift.

6.3 Origin of signs and systems 

The origin of individual signs and of sign systems must be seen in the light of the foregoing discussion. The tension between abstraction and organisation is that between creation and novelty, on the one hand, and congruence and uniformity on the other hand. Importantly, systematicity provides some conventionality for the process of creation and the tolerance of change. In some languages, for example, new signs can be coined relatively freely because the creative processes are themselves conventionalised. In sign languages there is productive use of morphological processes, but these are supported and constrained by the phonological organisation of those languages. Likewise, many spoken languages, e.g. German, have a productive, but constrained, morphology.

6.3.1 Invented sign systems

In this context, then, what are the prospects for an invented system of icons, for special or general communication? The answer must be negative, and for two reasons. Firstly, any invented system which is not a recasting of a pre-existing system is very unlikely to have hit upon a near stable configuration - in terms of the individual signs, in terms of the organisation of the system of which they are part, and in terms of conventions for managing the tension between abstraction and organisation. Pressures for change will thus outstrip organisational stability, if the system is used sufficiently. Secondly, the dependence on icons is likely to be just)fied by reference to notions of self-evident meaning, etc. This is quite false - icons are no less conventional than other symbols. For this reason, if no other, the notion of a pan-cultural system of communication is unlikely to work. One only has to look at a typical symbol for a building (e.g. in Dreyfuss, Blissymbolics, etc.) to see its origins in a particular culture; its value as a self-evident icon for a specific referent is culturally constrained.

6.4 Icons and interfaces

The situation regarding human-computer interfaces is much more hopeful. The use of icons, in sets, such as might be found in a word-processing package, for example, is unproblematic. Arbitrariness and set conferred iconicity (see 3.2.2 and 4.2 above) are readily encountered. Isolated icons can freely be created, on many systems, and these are probably just names. Where it seems likely that an analytical approach can bring dividends is that it will clarify the importance of convention and of sets or systems of icons.

It is also important to note, in the context of icons and interfaces, that the functional constraints on the circumstances of the use of such icons provides an important set of conventions which goes some way to systematising the signs concerned. Computers and applications which run on them have evolved to behave in certain restricted ways, and these behaviours underpin and conventionalise the icons used. It is a mistake to see the apparent ease of use of such signs (this needs investigation, in any case) as an indicator of the value of icons in the general sense. What must be acknowledged is the overall system and its limited functionality - it is the totality which is being used, interpreted, etc., not just the icons.

7 Conclusion

The single most helpful conclusion to be offered is the following. For most people the notion of an icon is that it has one or more, perhaps most, of the following properties: it can be considered in isolation, its origin is irrelevant, it is stable, it has an objectively

determined link with its referent, its meaning or referent is obvious, it dispenses with conventionality and arbitrariness, it is essentially different from symbols, especially linguistic symbols. I have attempted to show here that all these 'common sense' notions are misguided.

The Author

Dr. William Edmondson is a senior lecturer in the School of Computer Science at the University of Birmingham. He is a cognitive scientist with qualifications in physics, psychology, electrical engineering and linguistics.

Acknowledgments

William Edmondson's research is supported by a grant from Apricot Computers Ltd., a subsidiary of Mitsubishi Electric UK Ltd. The publisher would like to thank Random House UK Limited and McGrawHill, Inc., New York, for permission to reproduce the illustrations used in this paper.

References

Brophy,J.J. (1966) Basic Electronics for Scientists. New York: McGraw-Hill.

Camm, F. J. (1933) Newnes Wireless Constructor's Encyclopaedia. London: George Newnes Ltd.

Diringer, D. (1968) The Alphabet. (3rd Edition, in 2 volumes) London: Hutchinson.

Dreyfuss, H. (1 972) Symbol Source book. New York: McGraw-Hill. Paperback version published in 1984. New York: Van Nostrand Rheinhold Company.

Edmondson,W. H. (1977) Novel frequency-analysis system for a vibrotactile speech training aid for the deaf. Electronic Circuits and Systems, 1 (2), 57-63.

Greenlee, D. (1973) Peirce's Concept of Sign. The Hague: Mouton.

Kiernan, C., Reid, B. and Jones, L. (1982) Signs and Symbols:A review of literature and survey of the use of non-vocal communication systems. London: Heinemann Educational Books.

Klima, D. and Bellugi, U. (1979) The Signs of Language. Cambridge MA: Harvard University Press.

Sebeok,T.A. (1976.) Contributions to the Doctrine of Signs. Bloomington: Indiana University, with Lisse: The Peter de Ridder Press.