The analogy between depiction and description is usually argued to undermine rather than support the platitude that depiction is mediated by resemblance. From the claim that depiction and description are both kinds of symbol system Nelson Goodman, for example, draws the conclusion that “Almost any picture may represent almost anything; that is, given picture and object there is usually a system of representation, a plan of correlation, under which the picture represents that object” (Goodman, 1968, 38). Because they are both kinds of symbol systems, depiction and description, according to Goodman, are supposed to be equally arbitrary.
Defining depiction as a kind of symbol system is interesting because it develops an even closer analogy between depictive and descriptive representation. The moral that can be drawn from this chapter is that it’s possible to maintain such a close analogy between depiction and description, while still maintaining the platitude that whereas words are connected to what they represent merely by arbitrary convention, pictures are connected to what they represent by resemblance. The insight behind defining depiction as a kind of symbol system can be accepted, while the consequence that depictive representation is arbitrary is rejected.
This chapter and the next also address the objection that depiction should be analysed in analogy with sentence meaning, rather than speaker meaning. This chapter argues that there is an exact analogy between the analysis of depictive symbol systems and the analysis of speaker meaning, except whereas the latter is arbitrary and mediated by convention, the former is non-arbitrary and mediated by resemblance. Nevertheless, it doesn’t follows that depiction should be analysed in analogy with sentence meaning and not speaker meaning, because some depictions don’t belong to depictive symbol systems, and depict only in virtue of their perpetrators’ intentions.
The first section describes the tension between defining depiction as a kind of symbol system and the platitude that depiction is mediated by resemblance. The second section describes an analogous tension between the definition of a language and the platitude that description is mediated by convention. The third section describes an analysis of conventional language which resolves this tension. And the fourth section argues for an analogous analysis of depictive symbol systems which reconciles the analogy between depiction and description with the platitude that whereas description is mediated by convention, depiction is mediated by resemblance.
4.1 Goodman’s definition of symbol systems
A symbol system is a set of characters correlated with a set of extensions (Goodman, 1968, 143). In the symbol system of Arabic numerals, for example, the characters are the numerals ‘1’, ‘2’, ‘3’, ‘4’, … and the extensions are the numbers one, two, three, four, … and so on. The symbol system correlates ‘1’ with one, ‘2’ with two, ‘3’ with three, ‘4’ with four, … and so on. Likewise, in the symbol system of traffic lights, for example, the characters are red, orange and green and the extensions are stop, slow and go. The symbol system correlates red with stop, orange with slow and green with go. Depiction, according to Goodman, is a kind of symbol system.
A correlation between a set of characters and a set of extensions is simply a function from the characters to the extensions. A function is a mapping between two sets: it delivers a member of the second set for every member of the first set. The function of doubling, for example, takes every member of the set of natural numbers to a member of the set of even numbers: it takes one to two, two to four, three to six, four to eight, ... and so on. Every many-one mapping between two sets, no matter how arbitrary, is a function. There’s a function, for example, which takes every letter of the alphabet to the number of the day of the month on which it was first written.
The advantage of defining symbol systems as functions from characters to extensions is that any symbol system can be fully specified by such a function. The symbol system of traffic lights, for example, is fully specified by the function that takes green to go, orange to slow and red to stop. Similarly the symbol system of Arabic numerals is fully specified by the function that takes ‘1’ to one, ‘2’ to two, ‘3’ to three, ‘4’ to four, ... and so on. The same point applies to depictive symbol systems. The symbol system of chess diagrams, for example, is completely specified by the function which takes each chess diagram to the position it represents.
Since a function is any mapping, no matter how arbitrary, between two sets, the definition of symbol systems as functions from characters to extensions shows that any depiction may represent anything in some symbol system or other. There is a function from portraits to people, for example, that takes the Mona Lisa to Socrates, and so it follows that there is a symbol system in which the Mona Lisa represents Socrates. Just as words represent other things in other languages, pictures depict other things in other symbol systems, so what a depiction represents appears to depend not on what it resembles but only on its extension relative to a symbol system.
So if depiction is a kind of symbol system, then any depiction, just like any word, may represent any thing in some symbol system. Goodman draws the conclusion that depictive representation is like descriptive representation in being arbitrary. As he writes, “Descriptions are distinguished from depictions not through being more arbitrary … for what describes in some symbol systems may depict in others. Resemblance disappears as a criterion of representation…” (Goodman, 1968, 230-231). So if, as Goodman claims, depiction is a kind of symbol system, then depictive representation appears to be both arbitrary and unmediated by resemblance.
But even if it follows from the definition of symbol systems that any picture may represent anything, it does not follow that any picture may depict anything, since the alternative symbol systems relative to which characters possess other extensions may lack the syntactic and semantic properties required for being depictive. So although, for example, there is a symbol system in which the Mona Lisa represents Socrates instead of Lisa, that symbol system may not be a depictive one, in which case the fact that there is a symbol system in which the Mona Lisa represents Socrates would not show that the Mona Lisa’s depiction of Socrates is arbitrary.
Furthermore, there seems to be little obstacle in principle to combining Goodman’s thesis that depiction is a kind of symbol system with the thesis that depictions resemble what they represent, or any other thesis which maintains that the relationship between depictions and what they represent is non-arbitrary. As Dominic Lopes writes, “The claim that pictures are symbols in this [Goodman’s] sense is not incompatible with perceptual explanations of depiction. Nothing in the symbol model rules out pictures being correlated with, and standing for, their subjects because they resemble them…” (Lopes, 1996, 57).
The following analysis, for example, might be taken to reconcile defining depiction as a kind of symbol system with the platitude that depiction is mediated by resemblance:
(13) |
A symbol system is depictive if and only if every character in that symbol system resembles its extension. |
So the symbol system of maps, for example, is supposed by this analysis to be depictive because every map resembles the region it represents in that symbol system. If this analysis is right, defining depiction as a kind of symbol system establishes neither that depiction is arbitrary nor that it’s unmediated by resemblance.
But the following example shows that this analysis cannot be right. Often, letters of the alphabet are used to represent themselves, so that ‘a’ represents ‘a’, ‘b’ represents ‘b’, ‘c’ represents ‘c’, and so on. Since resemblance is reflexive, every letter in this symbol system resembles and represents itself, but it is intuitively not the case that every letter in the symbol system depicts itself, or that the symbol system described is depictive. The letters’ resemblance to themselves is incidental to their representation of themselves: even if, for example, capital letters were used to represent lower case letters, the way in which the letters represent each other would be the same.
So reconciling defining depiction as a kind of symbol system with the platitude that depiction is mediated by resemblance is not straightforward. The following sections exploit an analogy between symbol systems and languages which shows how to resolve this problem. The role of resemblance in depictive symbol systems, I will argue, is analogous to the role of precedent in conventional language. This suggests an analysis of depictive symbol systems which supports a strong analogy between depiction and description and the platitude that whereas description is mediated by convention, depiction is mediated by resemblance.
4.2 Formal definition of languages
Just as a symbol system is a function from characters to extensions, a language is a function from sentences to truth-conditions (Lewis, 1975, 163). So ‘snow is white’ means in English that snow is white, for example, because English is a function from sentences to truth-conditions that takes the sentence ‘snow is white’ to the condition of snow’s being white. Similarly, ‘grass is green’ in English means grass is green because ‘grass is green’ is taken by the function to the condition of grass’ being green. This section describes the tension between this definition of a language and the platitude that description is mediated by convention.
The main advantage of defining languages as functions from sentences to truth-conditions, as for defining symbol systems as functions from characters to extensions, is that every language is fully specified by such a function. English, for example, is fully specified by the function that takes ‘snow is white’ to the condition of snow’s being white, ‘grass is green’ to the condition of grass’ being green, ‘it’s raining’ to the condition of its being raining, ... and so on. Once the relevant function from sentences to truth-conditions is specified, there’s no further question about what the sentences of the language specified by that function mean.
Since truth-conditions are the extensions of sentences and sentences are the characters of languages, this definition of languages is just a special case of the definition of symbol systems (Lopes, 1996, 59). So defining depiction as a kind of symbol system is closely analogous to defining languages as functions from sentences to truth-conditions. Although I’ll argue below that depiction should not be defined as a kind of symbol system, I’ll also argue that the strong analogy behind this definition is correct: depictive symbol systems are exactly analogous to conventional languages.
Just as there’s a tension between the definition of depiction as a kind of symbol system and the platitude that depiction is mediated by resemblance, there’s a tension between the definition of languages as functions from sentences to truth-conditions and the platitude that description is mediated by convention. The reason is that a function from sentences to truth-conditions necessarily takes the sentences it does to the truth-conditions it does: just as the function of doubling necessarily takes two to four, the function from English sentences to their truth-conditions, for example, necessarily takes ‘snow is white’ to the condition of snow’s being white.
If sentences necessarily mean what they do in a language, then it appears convention could have no role in linking them to their meaning. If ‘snow is white’, for example, necessarily means in English that snow is white, then no convention is needed to determine that it does. If languages are functions from sentences to truth-conditions, having a convention that ‘snow is white’ means that snow is white in English is like having a convention to the effect that falling objects must accelerate at approximately ten metres per second per second: clearly, the acceleration of falling objects is independent of any convention which may purport to govern it.
But to conclude on these grounds that language is not mediated by convention would be obviously incorrect. This is because, while it is a matter of necessity rather than convention that ‘snow is white’ in English means that snow is white, it is contingent that English, or the language given by the function from English sentences to their truth-conditions, is a language of this island, and this is a fact which is plausibly mediated by convention. It’s mistaken to infer from the fact that sentences necessarily mean what they do in a language that language is not mediated by convention.
In general, while the meaning of a sentence in a language is a matter of necessity, it’s a contingent matter whether a language or function from sentences to truth-conditions is used by a population. So although it’s a matter of necessity, for example, that ‘snow is white’ means snow is white in English, it’s nonetheless an arbitrary matter that a language in which ‘snow is white’ means that snow is white is spoken on this island, rather than a different language in which it means that grass is green. Convention doesn’t mediate the meaning of a sentence in a language, but it does mediate which languages amongst many a population uses (Lewis, 1975, 166-167).
The moral of this point is to distinguish between the study of languages in use and in the abstract. The study of languages in the abstract focuses on the study of languages as abstract mathematical objects such as functions from sentences to truth-conditions and investigates further questions about the structure of those objects. The study of language in use focuses, in contrast, on how those abstract mathematical objects are connected with concrete speech and interpretation. Convention is part of this latter study, because the role of convention is to determine which languages in the abstract are used in speech and interpretation (Lewis, 1969, 160-165; 1975).
Likewise, one should distinguish between symbol systems in the abstract and symbol systems in use. The fact that any depiction may depict anything in some symbol system or another is a mathematical fact about the nature of symbol systems in the abstract. Nothing follows from this about the nature of symbol systems in use, since it doesn’t follow that the choice between which symbol systems in the abstract are adopted for use is an arbitrary one. So the platitude that depiction is mediated by resemblance should not be interpreted as a thesis about symbol systems in the abstract, but as a thesis about symbol systems in use.
So the platitude that whereas description is mediated by convention, depiction is mediated by resemblances suggests that resemblance is responsible for determining which symbol systems in the abstract are symbol systems in use. If this suggestion is right, then the difference between depictive and other symbol systems is not in the abstract mathematical syntactic and semantic properties of symbol systems, but in the way that depictive symbol systems are connected to the populations which use them. The following sections attempt to confirm this suggestion by providing an analysis of depictive symbol systems analogous to the analysis of conventional language.
4.3 Lewis’ analysis of convention
This section describes the analysis of conventional language; the next adapts it to the analysis of depictive symbol systems. A convention is a rationally self-perpetuating regularity in behaviour. Driving on the left, for example, is a convention in Australia because there is a regularity of driving on the left in Australia and because the existence of this regularity provides Australian drivers with a reason for continuing to drive on the left. Driving on the right is a convention in Europe because there is a regularity of driving on the right in Europe, and the existence of this regularity gives European drivers a rational reason for continuing to drive on the right.
To have a convention of driving on the left it must be, firstly, that there is a regularity of driving on the left and, secondly, that drivers are aware that there is a regularity of driving on the left and, thirdly, that drivers have a reason to drive on the left on condition that the others do. If any of these conditions failed then the regularity of driving on the left would not be rationally self-perpetuating: drivers would not continue driving on the left in the first case because it would not be better to do so, in the second case because they would not know that it was better to do so and in the third case because others driving on the left would not give them any reason to do so.
This suggests convention may be analysed as follows (Lewis, 1969, 58):
(14) |
A regularity in the behaviour of a group is a convention if and only if: |
a. |
everyone conforms to that regularity |
b. |
everyone expects everyone else to conform to that regularity |
c. |
and everyone has reason to conform on condition that everyone else conforms. |
Driving on the left is a convention in Australia, for example, because everybody drives on the left, everybody expects everybody else to drive on the left, and everybody has reason to drive on the left if everybody else drives on the left.
Three amendments. First, imagine that everyone drives on the left because they expect others to and because everybody has reason to drive on the same side as others. However, nobody believes that others drive on the left for these reasons: rather everyone believes that others drive on the left merely out of habit or because driving on the left is more scenic. The regularity of driving on the left is not a convention for avoiding collision in this population, because members of the population would continue driving on the left even though they believe others may not care about collision (Lewis, 1969, 59).
To avoid this problem the analysis should be altered to reflect that everyone must be aware the conditions of the analysis are fulfilled. According to this amendment:
(15) |
A regularity in the behaviour of a group is a convention if and only if: |
a. |
everyone conforms to that regularity |
b. |
everyone has reason to conform on condition that everyone else conforms |
c. |
and everyone believes (a)-(c). |
So, for example, driving on the left is a convention in Australia because everybody drives on the left, everybody prefers to drive on the left given that everybody else drives on the left and because everybody believes that everybody else conforms to the regularity for these three reasons.
Second, the condition that everyone has reason to conform on condition that others do is supposed to capture the cooperativeness of convention, since it’s because of their common interests that everyone has reason to conform to the same regularity in conventional activities. Driving on the left in Australia is a convention, for example, because everybody has a common interest in driving on the same side, and this common interest gives them a reason to conform to the regularity of driving on the left if others do. In general, all convention arises from common interests (Lewis, 1969, 69).
But the conditions of the analysis may be met even when common interests are absent. Imagine an office in which everyone hopes to be promoted by working harder than their colleagues. A regularity of hard work obtains in this office, everybody in the office has reason to conform to this regularity of hard work on condition that everyone else does, and everybody in the office is aware of these three facts. But hard work is not a convention of the office, because there’s no common interest in everybody working hard. In contrast, since everyone in the office is interested in promotion, their interests are best served by everyone else slacking off.
To avoid this problem the analysis should be amended to reflect that everybody has a common interest in conformity to the regularity. According to this amendment (Lewis, 1975, 165):
(16) |
A regularity in the behaviour of a group is a convention if and only if: |
a. |
everyone conforms to that regularity |
b. |
everyone has reason to conform on condition that everyone else conforms |
c. |
everyone prefers everyone to conform, on condition that most do |
d. |
and everyone believes (a)-(d). |
So driving on the left, for example, is a convention since everyone prefers everyone to drive on the left on condition that most drive on the left, since driving on the left on condition that most do reduces crashes, and everyone prefers not to crash.
Third, the condition that everybody has reason to conform because others do is supposed to capture the arbitrariness of convention, since everyone is supposed to conform to the regularity for no other reason than that others do. The convention of driving on the left, for example, is supposed to be arbitrary because nobody has much reason to drive on the left except the fact that others drive on the left. The regularity of driving on the left is no better than the regularity of driving on the right, which everyone would happily adopt were it adopted by others.
But this condition is not sufficient to capture arbitrariness. Suppose, for example, that it’s a regularity in our behaviour to meet for coffee once a week, on Fridays. We only like to drink coffee together, so that I have reason to go to the café on Friday only if you go to the café on Friday and both of us prefer to go if the other does. And suppose we’re both aware of all this. Nevertheless, our regularity of meeting at the café on Fridays may not be chosen arbitrarily, because on Friday the café offers a heavy discount. Our regularity of meeting for coffee on Fridays meets the conditions of the analysis, but it is not a convention.
To avoid this problem the analysis has to be amended to reflect that the option chosen is a matter of indifference. According to this amendment (Lewis, 1969, 76):
(17) |
A regularity in the behaviour of a group is a convention if and only if: |
a. |
everyone conforms to that regularity |
b. |
everyone has reason to conform on condition that everyone else conforms |
c. |
there is an alternative regularity which everyone would have had reason to conform to if others had conformed to it |
d. |
everyone prefers everyone to conform, on condition that most do |
e. |
and everyone believes (a)-(e). |
So, for example, driving on the left would not be a convention if it weren’t for the existence of the option of driving on the right, which everyone else would have reason to do if others did.
Since conventions are regularities in behaviour, the analysis of convention doesn’t apply directly to functions from sentences to truth-conditions, but only to a regularity in behaviour concerning those functions. David Lewis (1969, 177) suggests that the relevant regularity is truthfulness in a language, which consists in sometimes uttering sentences of the language, while trying not to utter sentences of the language which are false in that language. Being truthful in English, for example, consists in sometimes uttering sentences such as ‘snow is white’, while trying to abstain from uttering sentences such as ‘the moon is blue cheese’ if the moon is not blue cheese.
This suggests that a language is used if and only if there’s a convention of truthfulness in that language. Together with the analysis of convention, this leads to following analysis of conventional language (Lewis, 1969, 177; 1975, 167-168):
(18) |
A group has a convention of using a language if and only if there is a regularity in the group such that: |
a. |
everyone is truthful in that language |
b. |
everyone has reason to be truthful in that language on condition that everyone else is truthful in that language |
c. |
there is an alternative regularity of truthfulness in another language which everyone would have reason to conform to if others did |
d. |
everyone prefers everyone to conform to a regularity of truthfulness in that language on condition that most do |
e. |
and everyone believes (a)-(e). |
English, for example, is a language spoken by English speakers since English speakers are truthful in English, English speakers expect all other English speakers to be truthful in English, and all English speakers prefer to be truthful in English given that everybody else is.
4.4 Analysis of depictive symbol systems
It’s a platitude that whereas words are connected to what they represent merely by arbitrary conventions, pictures are connected to what they represent by resemblance. That suggests that the role of resemblance in depictive representation is analogous to the role of convention in descriptive representation, so that instead of simply stating that depictive symbol systems are those whose characters resemble their extensions, depictive symbol systems may be defined by substituting symbol systems for languages and resemblance for arbitrariness in the analysis of conventional language.
Arbitrariness is ensured in the analysis of convention by the second condition, which ensures that the population conforms to the regularity for no other reason than that other members of the population conform to it, and the third condition, which ensures that there are other regularities which all members would have preferred to conform to had others done so. So to adapt the analysis of the conventions governing linguistic representation in order to provide an analysis of depictive symbol systems, these two conditions ought to be altered.
The second condition of the analysis captures the arbitrariness of linguistic conventions by specifying that which language the members of a population have reason to use depends on which language other members of that population choose to use, rather than any feature of the language independent of the choices of others (Lewis, 1969, 70). To provide an analysis of depictive symbol systems this condition should be amended to include resemblance, because the resemblance of a symbol system’s characters to their extensions in that symbol system provides an additional reason for its use.
The third condition of the analysis further ensures the arbitrariness of linguistic conventions by specifying that there must be an alternative to the regularity members of the population actually conform to which would serve their purposes just as well. Since the relationship between depictions and what they represent is not arbitrary, this condition becomes unnecessary and should be dropped: although there may be alternative regularities which would serve just as well, there need not be such alternatives in order for the use of a symbol system to count as depictive.
This suggests the following analysis of depictive symbol systems:
(19) |
A symbol system is depictive if and only if there is a regularity of truthfulness in that symbol system such that: |
a. |
everyone is truthful in that symbol system |
b. |
everyone has reason to be truthful in that symbol system, since its characters resemble their extensions |
c. |
everyone prefers everyone to conform to a regularity of truthfulness in that symbol system on condition that most do |
d. |
everyone believes (a)-(d). |
So, for example, the symbol system of maps is depictive because everybody uses them, everybody expects everybody to use them, everybody has reason to do this because maps resemble what they represent, rather than simply because everybody else uses them, and everybody is aware that these conditions obtain.
By specifying the role of resemblance in depictive symbol systems, this analysis avoids counterexamples to simply defining depictive symbol systems as those whose characters resemble their extensions. The symbol system of using letters to represent themselves, for example, is not counted as depictive because although all the characters in that symbol system resemble their extensions, it is not for this reason that this symbol system is preferred over others as a method of representing the letters: the alternative of using capital to represent lower case letters, for example, would serve equally well.
Furthermore, the analysis shows that it does not follow from depiction being a kind of symbol system that the relationship between depictions and what they represent is merely arbitrary. While it is true that there are always other symbol systems in which the same pictures would have different extensions, which symbol system is selected for use in communication is not arbitrary but depends on the resemblance between the characters and extensions of that system just as, while it is true that sentences have their meaning in English necessarily, linguistic meaning is not a matter of necessity since which language is spoken depends on arbitrary conventions.
Whereas the analysis of depiction in the previous two chapters paralleled the analysis of speaker meaning in terms of intention, the analysis of depictive symbol systems in this chapter parallels the analysis of sentence meaning in terms of convention, and so accommodates the possibility of depiction without intention. Just as a person uttering a sentence whose meaning in a language they do not know might accidentally utter a sentence which means something they don’t intend, a person who doesn’t know that the top of a map, for example, represents north may accidentally depict the orientation of a nation in a way which they don’t intend.
In chapter two, I said that using convention to accommodate unintentional depiction risks failing to distinguish between depiction and onomatopoeia. The word ‘woof’, for example, resembles but does not depict a dog’s bark. But just as sentences now have their sentence meaning because of conventions arising from what they were used to speaker mean in the past, onomatopoeic words now have their meaning because of conventions arising from what they were used to depict in the past. The word ‘woof’, for example, represents a dog’s bark due to a convention, which arose because ‘woof’ was used to depict a dog’s bark in the past.
Though the analysis of depictive symbol systems in this chapter parallels the analysis of sentence meaning in terms of convention, it can still distinguish between depiction and onomatopoeia. Although there are symbol systems in use in which ‘woof’, for example, represents a dog’s bark, these symbol systems are not depictive but merely conventional, since the reason for being truthful in those symbol systems is no longer because the characters in the symbol system resemble what they represent, but is now merely because others are truthful in that symbol system. In other words, use of those symbol systems is no longer mediated by resemblance, but merely by convention.
Two clarifications. First, the definition of truthfulness in a language cannot be straightforwardly applied to symbol systems, since symbol systems are functions from characters to extensions rather than truth-conditions and extensions need not be true or false. One cannot be truthful in the symbol system of Arabic numerals, for example, since the extensions of the numerals in the system are numbers, which exist or not rather than obtaining or being true or false. Similarly, if the extensions of chess diagrams are chess positions, one cannot be truthful in the symbol system of chess diagrams, because chess positions cannot be true or false.
The solution is to observe that depiction is of states of affairs as well as objects. The Mona Lisa, for example, does not merely depict Lisa but also the state of affairs of Lisa smiling. Similarly, depictions resemble states of affairs as well as objects: the Mona Lisa does not merely resemble Lisa (the object), but the Mona Lisa’s having certain properties also resembles Lisa’s smiling (the state of affairs). So the difficulty can be overcome by applying the analysis first towards the depiction of states of affairs and then stipulating that something depicts an object if and only if it depicts a state of affairs of that object’s having a property (see section 8.4).
Second, both the analysis of conventional language and of depictive symbol systems have to be altered to accommodate non-assertoric utterances. The sentences ‘the door is closed’ and ‘shut the door’, for example, have the same truth-condition. But their meanings in English are different, since the first is an indicative and the second is an imperative. The solution is to redefine languages as functions from sentences to ordered pairs of moods and truth-conditions. English, for example, is a function taking ‘the door is shut’ to the ordered pair of the indicative mood and the door’s being shut and ‘shut the door’ to the ordered pair of the imperative mood and the door’s being shut.
As a result of this revision, the definition of truthfulness in a language also has to be revised. Truthfulness, according to the revision, is sometimes uttering sentences of the language in the indicative mood while abstaining from uttering indicative sentences which are false in the language and trying to make true in the language sentences of the language addressed to one in the imperative mood, if those sentences are uttered by the appropriate agent – such as someone in authority, someone one can trust or whom one wants to please (Lewis, 1969, 184). Analogous revisions are required to accommodate other moods, depending on the usual force of the mood.
The analysis of depictive symbol systems requires less revision, since the definition of symbol systems need not be altered to include moods. This is because, while some pictures are intended to produce beliefs and others actions, this difference is not marked in the syntactic structure of the pictures in the way that the intended force of the utterance of a sentence is marked using the mood of the sentence. Differences in mood are syntactic differences between sentences used as prima facie markers of the intended force of a sentence, but there are not usually corresponding syntactic differences between depictions.
There is no difference corresponding to mood, since there is no syntactic difference at all, between, for example, the picture of a Lego castle on the front of the Lego box which tells you how the Lego will look when it is built and the picture of the Lego castle contained in the Lego instructions which tells you where to put the final bricks to complete the castle. Although the picture on the front of the box – like ‘the door is shut’ – is designed to induce belief and the picture in the instructions – like ‘shut the door’ – is designed to induce action, this difference in force is not marked by any syntactic difference in mood between the pictures.
Nevertheless, the fact that some depictions are used to produce action rather than belief does require a revision of the definition of truthfulness used in the analysis of depictive symbol systems. The definition of truthfulness says that someone is truthful in a symbol system if and only if they sometimes perpetrate characters in that symbol system while abstaining from perpetrating characters in the symbol system when the state of affairs they represent does not obtain. This definition only accommodated depictions intended to induce belief: Lego instructions which tell you where to put the bricks, for example, are not perpetrated only when the bricks are already arranged.
So the definition of truthfulness in a symbol system has to be revised. Truthfulness in a symbol system, according to the revision, is sometimes perpetrating characters of the symbol system while trying to abstain from perpetrating characters of the symbol system which represent states of affairs which don’t obtain or else trying to bring about the state of affairs the character represents, if the character is perpetrated by an appropriate agent (further problems involving truthfulness are addressed in the next chapter). So, for example, a picture of Lego may be perpetrated both if the Lego is arranged in a certain way or if it is to be arranged in that way.
There are two kinds of philosophy of language, which address two different kinds of question. The first kind is descriptive: it answers questions about what kinds of syntax and semantics our languages have. Answers to these descriptive questions include, for example, the definition of a language as a function from sentences to truth-conditions. The second kind is foundational: it answers questions about what makes it the case that our languages have the syntax and semantics that they do. Answers to these foundational questions include, for example, the thesis that language is mediated by convention (Lewis, 1969, 204; Stalnaker, 1984, 32-35; 1997, 166-168).
Correspondingly, there are two kinds of question about the nature of depiction: descriptive questions about the syntax and semantics of pictures and foundational questions about what makes it the case that depictions have the syntax and semantics they do. The descriptive questions include, for example, the question of whether the semantics of depiction is compositional or the content of a picture is a state of affairs. The foundational questions include, for example, the questions of whether depiction is mediated by resemblance, and of how what a depiction represents is dependent on its perpetrators’ intentions.
According to the platitude that depiction is mediated by resemblance the difference between depiction and other kinds of representation is a foundational issue about what makes it the case that depictions represent what they do. An important alternative is that the difference between depiction and other kinds of representation is a descriptive issue concerning the descriptive syntactic and semantic properties of symbol systems. As Goodman, for example, writes: “... whether a denoting symbol is representational [depictive] depends not upon whether it resembles what it denotes but upon its own relationships to other symbols in a given system” (Goodman, 1968, 226).
Which symbol systems in the abstract are symbol systems in use is not a descriptive question about the syntactic and semantic properties that symbol systems actually use, but a foundational question about what makes it the case that we use certain symbol systems rather than others. English, for example, is a symbol system in use rather than merely a symbol system in the abstract because we have a convention of speaking and writing in English. So even if depiction is to be defined as a kind of symbol system, this definition must still address foundational questions about what makes it the case that our symbol systems have the syntactic and semantic properties that they do.
In terms of the distinction drawn at the beginning of this section, questions about the definition of depiction are not descriptive questions about the syntax and semantics of pictures, but foundational questions about how depiction is mediated. As Kulvicki writes “… one appealing direction in which to proceed is to claim that facts about us … contribute to determining which representational systems are pictorial … After all, a merely structural analysis of representational systems must fail to take into account how and by whom such representational systems are used …” (Kulvicki, 2006, 42). If a symbol system is depictive, it is so because of the way in which we use it.