Metaphors, religious language and linguistic expressibility

Abstract

This paper examines different functions of metaphors in religious language. In order to do that it will be analyzed in which ways metaphorical language can be understood as irreducible. First, it will be argued that metaphors communicate more than just propositional contents. They also frame their targets with an imagistic perspective that cannot be reduced to a literal paraphrase. Furthermore, there are also cases where metaphors are used to fill gaps of what can be expressed with literal language. In order to clarify this function of catachresis the notion of de re metaphors will be introduced. With those metaphors we can convey contents that we cannot conceptualize independently from a certain context of utterance. Hence, with such metaphors we can reach beyond the limits of our conceptual repertoire which is a crucial function for religious language. Finally, the consistency and plausibility of the radical position that all assertions about God are irreducibly and necessarily metaphorical if they are supposed to be true will be discussed on the basis of the results of the former considerations about the irreducibility of metaphors.

Introduction

It is undeniable that metaphors are used very often in religious contexts. The following sentences are, for example, from the Hebrew Bible/Old Testament:¹

(1) The Lord is my rock, my fortress and my deliverer; my God is my rock, in whom I take refuge, my shield and the horn of my salvation, my stronghold (Ps 18, 2/3).

(2) But now, O Lord, You are our Father. We are the clay, and You our potter. And all of us are the work of Your hand (Jes 64, 7).

(3) For the Lord God is a sun and shield (Ps 84, 11).

(4) Hear us, shepherd of Israel, you who lead Joseph like a flock (Ps 80, 1).

It seems to be absurd to understand the expressions “rock”, “fortress”, “shield”, “horn of salvation”, “stronghold”, “father”, “potter”, “hand”, “sun”, “shield” or “shepherd” in these sentences literally. They have to be interpreted metaphorically. Due to its apparent capacity to transcend the limits of literal language, theologians, as well as philosophers of religion are traditionally very interested in the role of metaphor in religious language,² which will, in the following, be understood as language about God.³ The focus on religious language should neither be understood as implying that metaphors are working in a fundamentally different or special way in religious language nor that metaphors are not of great interest in other areas, too. Still, the arguments in this paper are developed in regards to language about God where metaphors traditionally play a crucial role because God is generally conceived of as being at the very limits or even beyond the human capabilities of conceptual understanding. Some authors such as Kenny (2004, 39–41), McFague (1982, 194), Jüngel (1974, 112) or Kaufman (1972, 95) even go so far as to claim that because of God’s radical difference to our mundane world, it can only be talked metaphorically about God. In order to precisely describe the prospects as well as the risks of metaphors in religious language, I will discuss in which ways metaphors are irreducible. First, I will argue that even if there might be literal language that expresses the same propositional contents as certain metaphorical utterances, it is convincing that such literal paraphrases cannot capture the imagistic perspective or “framing effect” the metaphorical utterances evoke. Secondly, the sense in which metaphors might also enable the expression of propositional contents that we cannot convey with literal paraphrases will be delineated. It will turn out that there is a subclass of metaphors that can be called heuristic de re metaphors which are indeed in principle irreducible to literal paraphrases. In the end, based on the results of the former considerations about irreducible metaphors, the consistency and plausibility of the claim that all assertions about God must necessarily and irreducibly be understood metaphorically if they are supposed to be true, i.e. the position sometimes called “panmetaphoricism”, will be discussed.

In this paper, a metaphor is taken to be a non-literal, figurative use of language where one kind of thing (the target) is characterized in terms of another (the source). Hence, the target is the subject matter which is talked about and the source is the subject matter which is exploited in order to describe, frame, or portray the target. In example (1) above the source is the subject matter conventionally related to the expression “rock” and the target is God. In (2) the target is the Lord, i.e. again God, and the source is a father. The target does not have to be mentioned by a definite description. In (4) it is introduced with only a pronoun (“you”), from which the context indicates that it refers to God who is portrayed as the shepherd of Israel. This characterization of metaphor is admittedly tentative. It is only introduced for some basic distinctions (target and source) as well as some basic clarifications of how metaphors are understood here. It is furthermore claimed that metaphorical interpretations are different from literal ones since the former are conceptualized as figurative interpretations. Roughly, literal interpretations should be understood as the interpretations that are conventionally associated with a certain expression. This does not mean that literal interpretations are only derived via a mechanistic application of conventionally associated meanings to certain expressions within sentences. In most cases, the conventionally connected meanings have to be contextually adapted to the specific situation in which they are applied. See for example the following cases:

(5) I already had breakfast.

(6) All bottles are empty.

With (5) it is normally communicated that the speaker had already had breakfast today and not on any day in his life. Equally an utterance of (6) is normally not used to express that all bottles in the world are empty but only those of a certain domain as for example the bottles of apple juice in the fridge. Although these interpretations are contextually adapted, we would not call them figurative interpretations. Authors such as Travis (1985), Recanati (2004), Stanley (2000), Bach (2001), and Anne Bezuidenhout (2002) have shown that contextual influence on what we would normally understand as literal interpretation is pervasive. Irrespective of the question whether the boundary between metaphorical and literal language is a categorical or a more continuous one⁴, there appear two major differences between literal and metaphorical language:

D1: Literal meanings of expressions can be learned apart from certain contexts. Metaphorical interpretations are, in contrast, not teachable apart from certain linguistic and extralinguistic contexts.

D2: There is an asymmetry between metaphorical and literal interpretations: Whereas metaphorical interpretations necessarily presuppose the knowledge of the literal meanings of the expressions used metaphorically, literal interpretations do not presuppose the knowledge of possible metaphorical interpretations.⁵

In cases where a former metaphorical interpretation does no longer differ in these respects from literal interpretations, the metaphor has become “dead” or lexicalized. Since the important characteristics of metaphorical interpretations are lost in such cases, it is suitable to say that dead metaphors are in fact no more metaphors but further instances of literal interpretations of expressions. The interpretation “a movie which is financially extraordinary successful” for the expression “blockbuster” is for example a literal interpretation, although it was the result of a metaphorical interpretation once.

The irreducibility of metaphors

The framing effect

To talk about the potential irreducibility of metaphors in a qualified way, we first have to elaborate on the notion of a literal paraphrase. Camp (2006, 2) argues convincingly that an adequate paraphrase does not have to reproduce all effects a metaphor is accompanied with. She excludes intended implicatures or other merely insinuated contents from the paraphrase. She also does not think that an adequate paraphrase must capture the non-propositional perspective or the seeing-as effects a metaphor produces. Instead, the paraphrase should state in literal terms the content expressed by the metaphor without demanding the same or similar interpretative effort as the metaphor itself.⁶ With this understanding of an adequate paraphrase as a background assumption we can highlight a first aspect in which metaphors are not reducible to literal language when we focus on the non-propositional, imagistic perspective metaphors evoke. As described, Camp claims that an adequate paraphrase does not have to reproduce these non-propositional effects. It will be argued here that a literal paraphrase construed in the way delineated above indeed cannot reproduce this non-propositional imagistic perspective.⁷ The evocation of such a perspective by metaphorical language, which Richard Moran called the “framing effect” (see Moran, 1989, 96), was most famously described and vindicated by Donald Davidson. The latter compares metaphors to jokes and pictures. It seems to be misguided for him to ask in one of these instances for meanings or propositional contents that are conveyed. Concerning pictures Davidson writes strikingly:

How many facts or propositions are conveyed by a photograph? None, an infinity, or one great unstatable fact? Bad question. A picture is not worth a thousand words, or any other number. Words are the wrong currency to exchange for a picture (Davidson, 1978, 47).

Following Wittgenstein, Davidson furthermore makes an illuminating distinction between “seeing that” and “seeing as” (see Davidson, 1978, 47). To grasp a truth-conditional content would be to see that something is the case. But this is entirely different from the non-propositional seeing something as something else. That can be illustrated with the following two metaphors:

(7) Tolstoy was a great moralizing infant.⁸

(8) Saul Kripke is an alchemist of ideas.

When interpreting the metaphor (7) we for example do not only infer some analogies between Tolstoy and infants. All propositions we might extract based on these comparisons such as “Tolstoy acts immature” or “Tolstoy cannot control his emotions properly” would be part of seeing that something is the case. But all these propositions, at least according to Davidson, do not sum up to the perspective and imagery this metaphor evokes. This non-propositional, imagistic aspect can be described as seeing Tolstoy as an infant. These considerations can be transferred to example (8). To fully grasp the metaphor, it is not enough to understand that the speaker wants to communicate for example that the philosopher Saul Kripke can produce valuable philosophical insights from the combination of ordinary ideas. The recipient normally additionally also imagines Kripke to be working in a middle age laboratory dealing with different substances in order to produce something valuable such as gold. This perspective cannot be true or false and is, at least according to Davidson, non-propositional in nature. Through this perspective, the literal meaning of an expression remains active also after the metaphorical interpretation is successfully derived. The metaphorical characteristic of an interpretation is in part produced by a productive tension between the literal meaning of an expression and its new metaphorical interpretation.

To further illustrate that metaphors convey more than just propositional contents, consider the following example of a mixed metaphor from Tirrell (1989, 25):

(9) If you see a student with a spark of imagination, water it.

It is imaginable that somebody wants to use this utterance to communicate that the addressee should support students who have some imagination. This interpretation can be derived from the circumstance that watering a plant is helping the plant to grow. Hence, “watering” students can metaphorically mean to sponsor them in their careers. Although such a metaphorical interpretation of “to water” seems to be perfectly fine in general, (9) is infelicitous if “to water” is interpreted this way. The reason for this infelicity lies in the fact that the perspectives which are produced by the metaphorical interpretations of “spark” and “water” are in a contradictory tension with each other. If you water something, sparks are normally put out. But this tension does not appear at the level of the propositional content expressed with this mixed metaphor. It is on the level of the imagistic perspectives which are evoked by the literal meanings of the expressions that are interpreted metaphorically.⁹

This irreducible perspective of metaphors transgresses into the realm of religious language. The following are possible literal paraphrases of the metaphorical interpretations of (1) and (2):

P 1: I lead my life from a deep trust in God’s love and forgiveness which gives me hope and safety also in difficult situations of my life. Nothing can stop the love and unconditional acceptance that God has for me as I am.

P 2: God is the ultimate reason that brought our universe and, thereby, also every human being into existence. As contingent beings we are dependent on God, the ens necessarium.

In both cases P 1 and P 2, it is imaginable that for example contemporary speakers of the respective lines in the psalms have such propositional contents in mind. These contents can, obviously, be true or false. But even in the case that these literal paraphrases are apt, they miss a central element of the metaphors in the psalms. They do not reproduce the perspective or imagistic aspect that the metaphors evoked. When we describe God as in P 1, we do not see God as a rock, a fortress or a stronghold. And the description of P 2 does not conceptualize God as a potter or a father. It is, among others, this imagistic aspect of letting us see something as something else that makes metaphors interesting and attractive devices. It is also not hard to understand, why this framing effect is especially interesting for religious language. It can make very abstract reflections about God accessible to our sensory and visual experiences, at least to some extent.

Paraphrase and irreducible propositional contents

As Stanley Cavell famously noted, metaphors are always pregnant with meaning. This can, according to Cavell, be seen by the fact that “and so on” can always be added to possible paraphrases of metaphors (see Cavell, 1969, 79). Donald Davidson infers from this circumstance, among other things, that there is no determinate content that metaphors convey except their often blatantly false literal interpretation (see Davidson, 1978, 31 f.) One might conclude that metaphors are only or mainly used to evoke certain perspectives on a circumstance or an entity but no truth-conditional contents. Davidson and, following him, Rorty (1987), Reimer (2001), White (1996) as well as Matthew Stone and Ernie Lepore (2010) argue that the whole enterprise to look for propositional contents that are communicated by metaphors is misguided. Although it is undoubtedly true that many metaphors are not mainly used to express propositional contents, it is difficult to defend the claim that they never also express propositional contents. We can, for example, successfully make potentially true or false statements with metaphors (10) and successfully incorporate them into conditionals (11) or subordinate clauses (12).

(10) The arc of the moral universe is long, but it bends toward justice. (Martin Luther King)

(11) If music is the food of love, then play on. (William Shakespeare)

(12) Because Jack is a Gorilla, John is afraid of him.

The result of the metaphorical interpretation of (10) can intuitively be true or false. Circumstances where somebody continues playing music because she thinks that the antecedent of (11) is true are also easily imaginable. Finally, the causal subclause makes a meaningful and plausible contribution to the truth-conditional content expressed with (12). All this is very difficult to explain if we do not ascribe propositional contents to metaphors at all.¹⁰ Still, paraphrases of metaphors often seem to fall short of capturing the whole richness of metaphorical interpretations. But when we use the description of an adequate paraphrase by Camp that was delineated above, it does not have to capture all these additional effects metaphors produce. Furthermore, often one has to take the context into account in which metaphors are uttered. The contextual information, also about the presuppositions the different participates of a conversation share or pretend to share, normally disqualifies several interpretations. These two factors significantly reduce the indeterminacy of literal paraphrases of metaphors in most of the cases.

Even if all these points are accepted, there seem to be some metaphors that resist a literal paraphrase. It would be implausible to infer that such metaphors never express propositional contents. At least in some cases, the most convincing interpretation would be to ascribe truth-conditional contents to the respective metaphors while accepting that these contents cannot be captured by a literal paraphrase, even in the moderate way described above. Elisabeth Camp gives the following metaphor as an example for such a case:

(13) When he finally walked out the door, I was left standing on the top of an icy mountain crag, with nothing around me but thin cold air, bare white cliffs and a blindingly clear blue sky. (Camp, 2006, 11)

According to Camp, we can, with appropriate context information, infer via analogical reasoning the content the speaker intends to convey with (13). The speaker ascribes to herself a certain feeling for that she has no words but that can be captured only via analogies between the state of standing on the top of an icy mountain and the moment she felt when her partner left her. The recipient has to use imaginative skills in order to successfully capture the property that the speaker ascribes to herself with this metaphor (see Camp, 2006, 11). The mutual understanding of the speaker and the recipient does, in this case, not necessarily presuppose the availability of a literal paraphrase of the content conveyed with (13). One can generalize from this example that there are metaphors that express contents that cannot be adequately paraphrased in purely literal terms. Therefore, one might conclude that these are instances of irreducible metaphors.

William Alston would agree with the existence of such irreducible metaphors in the sense that we can, at the moment, offer no literal paraphrase of them although they convey a certain propositional content. Still, he claims that this does not mean that we cannot in principle give a literal paraphrase of their contents. He argues that in cases where we have some property P in mind, which is part of the truth-conditional content we want to express with a metaphorical statement, we also must have a concept C of the property P. What is cognitively accessible is, according to Alston, also conceptually accessible. If we, in turn, have a concept C for P, it is then always possible to introduce an expression E which is associated with the concept C that refers to the property P. Additionally, if we introduce an expression E that conventionally stands for C, we are able to literally paraphrase metaphors that express the property P. In short, if we have cognitive access to a certain property via metaphorical language, we can, according to Alston, form a concept that captures that property, which can via stipulation be related to a certain expression. If there is some truth-conditional content we can express with metaphorical language, this content can, therefore, always in principle become part of the expressive power of literal language (see Alston, 1989, 27–30). John Searle presented a very similar argument. He defends a “Principle of Expressibility”:

[…] for any meaning X and any speaker S whenever S means (intends to convey, wishes to communicate in an utterance, etc.) X then it is possible that there is some expression E such that E is an exact expression of or formulation of X. […] (Searle, 1969, 20).

Searle claims that for any meaning X we can think of and that we intend to convey to another person, we may introduce an expression E into the respective linguistic system. It is important to note that Searle, as Alston, does not say that we can in fact express everything that can be meant with every language. In his formulation is a modal operator. It is just always a possibility that a certain expression will be introduced for every thought. Searle claims that every upper bound a language reaches concerning thoughts it can express is a contingent, not a necessary fact (see Searle, 1969, 20). Hence, according to Searle all meanings that can be part of propositional contents we have cognitive access to, can in principle be conventionally associated with a certain expression E. Therefore, Searle claims that everything that can be thought of can also be expressed within a linguistic system. And since the expression E Searle talks about is described as an “exact expression of or formulation of” the respective meaning X, E can be understood as a conventionally established expression for X which is nothing else than a literal expression for X. This meaning X can in turn be furthermore described as referring to a certain property or sets of properties P.

There are different arguments that can be put forward against the expressive optimism proposed by Alston and Searle. It can, for example, be argued that there might be entirely private thoughts which cannot be communicated to others because of their idiosyncrasy. It is, in this case, not possible to express these thoughts with literal language at all. It seems pointless to introduce a term for a thought that in principle nobody else can share. Both Searle and Alston saw this possible objection against their optimism towards linguistic expressivity. Searle simply accepts the possibility of such expressions with meanings nobody but the speaker can understand. The denial of this possibility would rely, according to Searle, on the controversial claim that there can be no private language. The most famous critic of the possibility of a private language is Ludwig Wittgenstein (see Wittgenstein, PU, § 243 ff.). Searle, in contrast, remains agnostic about this possibility and, therefore, leaves open the question if there can be linguistic expressions with entirely private meanings (see Searle, 1969, 20).

Alston argues that one can in religious contexts expect people who claim to have had special, private or mystical experiences with God. These people might argue that they cannot communicate these private experiences if the recipients did not have the very same kind of experience with God. And even if somebody else had an experience with God, it does not seem possible to compare these experiences and look if they are really the same. Alston sees Wittgenstein’s argument against the possibility of a private language as an easy way out of this problem (see Alston, 1989, 29). He does not explicate this line of reasoning further. An imaginable argument would be that the impossibility of private meanings in a language would undermine the claim that these private experiences have any impact on the expressivity of language because such experiences can in principle never become part of a language. Alston does not go down this route but writes that he is not convinced by Wittgenstein’s argument against a private language. Alston’s strategy to deal with the privacy objection is that he takes it for granted that the factors available to disseminate socially shared meanings in religious communities are also available for establishing new meanings (see Alston, 1989, 29). Indeed, it seems to be the case that the need for justification lies on the side of the critics who deny that one can introduce every possible meaning into a linguistic system. It is not clear why special mystical experience should be in principle banned from being socially shared via our linguistic capacities.

A further possible objection against the expressive optimism of Searle and Alston is that there are properties we cannot conceptualize but which are nonetheless expressible with metaphorical language. This objection will be delineated in the next section.

The limits of expressibility and de re metaphors

A person who makes the aforementioned objection would have to explain how we can successfully refer to properties and express them with metaphorical assertions while we cannot adequately conceptualize these properties. A theory about this was developed by Josef Stern with his notion of de re metaphors.¹¹ He explicates what he means by such metaphors in reliance on Tyler Burge’s usage of de re belief (see Burge, 1977). Burge explained the notion of de re belief in relation to de dicto beliefs that have fully conceptualized contents as their objects. Hence, all elements of the respective propositional content are characterized by elements from the conceptual repertoire of the believer in the case of de dicto beliefs. De re beliefs, in contrast, are according to Burge beliefs about certain objects themselves, while there is no conceptual representation of the objects involved in the propositional contents which are believed. The de re believers relate to the respective objects in a non-conceptual and contextual way. Burge clarifies this point by claiming that he does not want to suggest that there are no concepts involved in the acquisition of de re attitudes towards the respective objects. He just wants to stress that the concepts do not enter the propositional content that is believed as concepts of the respective object. They do not denote the object (Burge, 1977, 345–347). The difference between those types of beliefs can be illustrated with sentences (14) and (15):

(14) The mayor of Helsinki likes snow.

(15) This book [pointing towards a book] is very old.

If a person, say Alfred, utters (14) while he has no particular person in mind, when he thinks of the mayor of Helsinki, he has a belief de dicto. He does not even need to know who the mayor of Helsinki is, in order to successfully formulate this belief. The reference of the singular term is determined solely by the definite description “The mayor of Helsinki”. When Alfred utters (15) in contrast, the reference of the demonstrative construction “This book” is not determined solely by the meanings of the expressions involved. Additionally, there is a pointing involved which enables the speaker and the audience to determine the reference of this complex demonstrative. Now, the pointing gesture is not a conceptual representation of the object but a contextual means to fix the reference. Alfred might not be in a position to conceptually represent the book he is pointing at. He might, for example, not be able to tell exactly at which stack the book lies in the library or which identification number it has, etc. Without the non-conceptual means of the pointing gesture, the speaker would not be able to individuate which book he is referring to.¹²

Stern applies these considerations of Burge to general terms (i.e. predicates). According to Stern, there are, in analogy to singular terms, two ways how they can refer to properties: they can do it de dicto via a concept or they can do it de re. In the latter case, they refer to what Stern calls a “bare property”. Part of the propositional content is, according to Stern, in these cases not a conceptualized representation of the property but the property “itself”. Similar to Burge, Stern would not deny that concepts play a crucial role in individuating the respective properties in these de re cases. He does not claim that we have immediate contact with the properties. The epistemic relation is just different from de dicto cases in the sense that the concepts and devices we use to refer to a certain property are not a conceptual representation of this very property that can be used independently of the respective extralinguistic context of utterance (see Stern, 2000, 188).¹³ Now, de re metaphors are, according to Stern, exactly such devices: they enable us to express properties for which we do not possess a context-independent conceptual representation. Hence, we cannot offer context-independent conditions that have to be fulfilled in order to ascribe the respective property to an entity. Instead, these properties can only be demonstratively “pointed” at relative to a network of associations within a certain context of utterance. Stern claims that de re metaphors function as predicate demonstratives (see Stern, 2008, 17 f.). Such de re metaphors are, according to Stern, devices that fulfil the function of catachresis (i.e. they fill lexical gaps in our language). These gaps are not only due to a lack of words but also due to a lack of understanding: we cannot refer to the respective novel properties because we have no representation of them in our conceptual repertoire (see Stern, 2000, 189). De re metaphors, therefore, enable us to reach beyond the limits of our conceptual capacities and thereby also help to extend the expressivity of our languages.

To illustrate de re metaphors, Stern cites an example from Ted Cohen (see Cohen, 1990), who tells about his father that he sometimes expressed a certain and specific deficiency of beers with the following remark:

(16) It tastes green.

According to Cohen, this was not an entirely private subjective experience. His father was able to identify this property in several beers and could also induce others (unfortunately, not Ted Cohen) to recognize this very property. However, his father was not able to refer to this specific property of beers without the metaphorical usage of “green”. Hence, the property he was picking out with this metaphor was identified based on relations between associations he had with green things and the taste of some kinds of beer. There was no lexicalized meaning such as “flat”, “stale” or “bitter” which could substitute the metaphorically derived meaning of “green” in these contexts. Metaphors which are not de re would, in contrast, express properties for which we have lexicalized concepts. Hence, these properties can also be expressed context-independently without the activation of specific metaphorical networks in certain contexts. These metaphors could be called “de dicto metaphors”.¹⁴ It should be noted that the distinction between de re and de dicto metaphors is an epistemic one. Hence, it might be the case that the very same metaphor is for one person de re and for another de dicto. The status of the metaphor is tied to the conceptual repertoire of the person who is using or interpreting the respective metaphor. Since the conceptual repertoire of a person may evolve, a metaphor that is de re for a certain person can become de dicto over time because new concepts were developed that can fill the former gaps of literal expressibility.¹⁵

With the terminology of Stern in hand¹⁶, the metaphor (13) can be understood as a de re metaphor. The property referred to is activated relative to a contextual network of associations and implications, while the speaker does not possess a context-independent mean to describe these properties. Instead, the metaphorical interpretation of (13) depends on specific association one has of icy mountain tops and the ending of relationships. Therefore, we have compelling arguments against the assumption of Alston that cognitive accessibility in forms of propositional content expressed always presupposes conceptual accessibility. According to Alston, a speaker “cannot have the property in mind without having a concept of that property” (Alston, 1989, 28). Hence, for Alston cognitive access to a property P implies that one in fact has a concept C for P. Alston concedes that a speaker or recipient might lack an expression E in her linguistic system that is conventionally related to the concept C that in turn refers to the property P. Now, if there are de re metaphors, we can have cognitive access to a property P without having a concept C for P. In such a case, we do not just lack an expression E for the concept C but we also lack the concept C itself that context-independently refers to the property P.

One might argue that even if there are de re metaphors, we can simply move a step further than Alston did and not only introduce an expression E but also a new concept C*¹⁷ for the property we are metaphorically referring to.¹⁸ One might, therefore, claim that in cases we have socially shareable cognitive access to a certain property, it is always in principle possible to introduce a concept C* that stands for this very property. This new concept can and perhaps also must also be associated with an expression E.¹⁹ This expression would then be a term that can be used literally to express the property P via the concept C*. The possibility to introduce a new concept C* is plausible for cases such as (13) or (16) because there is a certain specific property that is metaphorically referred to. It is easily imaginable that the expression “green” can receive a new meaning that refers to a certain taste of beer. The respective meaning of the taste of beer might then just become part of the conventional meaning of the polysemous expression “green”. The usage of “green” with this new meaning would then be a case of a dead metaphor. As was argued above this is not a case of a genuine metaphorical interpretation anymore but just a further literal interpretation of the respective expression. Similarly, the feeling the speaker described with (13) may be shared by others and after a certain time, a new concept and a corresponding expression might be introduced to refer to the property expressed by (13) in the context of utterance described above.

Nonetheless, it should be noted that this possibility does not make the metaphorical language in general theoretically eliminable. Even if it is in principle possible to introduce literal expressions for the properties expressed metaphorically, the metaphors are in many instances necessary to get a first hold onto the properties which are not fully conceptualized yet. At the time we use the metaphor for the first time, a literal paraphrase may not be available (see Camp, 2006, 16). This leads to a further step we might take: There can be phenomena or properties we can touch or glimpse at with de re metaphors but which we cannot conceptualize. Hence, we cannot introduce a new concept C* for these properties. Therefore, we are in turn also not able to introduce an expression that conventionally stands for these phenomena or properties in these cases. According to Camp, such scenarios can for example occur when we metaphorically speak about sub-personal cognitive processes that are mainly identified by their causal roles. We often cannot fix these processes without considering the complex causal network they are embedded in. Here, metaphors are helpful devices to get a first grab onto a property for which we have no fixed conceptualization. Camp presents the example where the process of cognitive memory retrieval is conceptualized as the opening of a computer file.²⁰ Through the analogical structure between the cognitive process of memory retrieval in our brain and the opening of computer files, we can, according to Camp, establish a link to a certain causal property we have no direct access to. Camp concedes that such an analogical equation might introduce referential vagueness since there might be many cognitive processes that are, in some respect, similar to the opening of a computer file. Still, she does not think that this undermines the possibility that this metaphor is useful and can be productive for research. First, the possible analogies one might infer from this metaphor are restricted by other assumptions we have about neural processing and the field of investigation. Second, when we enter a new research territory, it is okay to deal with referential vagueness. Especially, through their degree of indeterminacy, metaphors can establish a first connection or first access to certain phenomena that might later be conceptualized more precisely. Our conceptual repertoire often reaches limits when confronted with new phenomena we are not directly acquainted with. Metaphors and analogical patterns of thought are great tools to get a first hold of such phenomena (see Camp, 2006, 17–20).

In sum, metaphorical language can, in several respects, express propositional contents for which we do not have adequate literal paraphrases. De re-metaphors are means of referring to properties, although we have no concepts available to denote these properties independently of the utterance's context. Some of these de re metaphors, which shall in the following be called “heuristic de re metaphors”, are special because we cannot even introduce conceptualizations for the properties expressed by these metaphors. They help us to get a hold within a new research territory while we are not in the position to introduce concepts for the properties we are reaching out to. These functions are of high value for religious language where the central object is at the limits of our epistemological capabilities. With metaphorical language, we can grasp or at least point to properties of a divine reality that transgresses our conceptual repertoire. Janet M. Soskice famously tried to account for this special way metaphorical language can be applied to God when she argues that metaphors are not “directly descriptive” but still “reality depicting” (see Soskice, 1985, 148). We do, according to Soskice, not “define” God with metaphorical language but just “refer or point to” God (Ibid. 140 f.). At another occasion she writes that metaphorical language “denominates” rather than “describes” God (Ibid. 154). With the distinctions made so far in hand, we can clarify these considerations of Soskice: In theological speech heuristic de re metaphors are used in many cases to depict a reality we cannot fully conceptualize and understand.

Panmetaphoricism

So far, we have discussed several respects in which metaphors can be understood as irreducible. In the remainder, it will be described which consequences the results of this analysis have for the position that is sometimes called “panmetaphoricism”.²¹ According to this position there are not only some metaphors in religious language which are in some respects irreducible. Adherents of panmetaphoricism argue that language about God can only be true when it is understood metaphorically. Hence, it follows that all language about God must be interpreted as necessarily and irreducibly metaphorical when it is supposed to be true. Panmetaphoricists do not deny the possibility that we can make literal statements about God. They just claim that such statements must all be classified as false simply because they are literal statements. In order to avoid confusion, it should be clarified that this position should not be misunderstood as the claim that every expression in an assertion about God must be interpreted metaphorically. Obviously, in most cases we use metaphorical language, we just interpret some expressions of utterances metaphorically. In sentence (8) it is for example only the expression “an alchemist” which is interpreted metaphorically but neither “Saul Kripke” nor “is”, nor “of ideas”.²² It is assumed in panmetaphoricism that all predications about God must always, necessarily and irreducibly be understood metaphorically when they are supposed to be part of true assertions about God. The expression “God” which is a singular term, as well as for example connectives, conjunctions or operators do, for adherents of panmetaphoricism, not have to be understood metaphorically to be part of true assertions about God.

Reasons to hold panmetaphoricism are usually God’s supposed transcendence and complete “otherness” when compared to the entities we normally apply our expression to. Anthony Kenny for example claims that we always have to leave the language game where a word is “at home” when we apply it to God (see Kenny, 2004, 39–41). Alston (1989) famously attacked the consistency of panmetaphoricism. As was already delineated above, he argued that a necessary condition for cognitive access to a certain property P is that we have a concept C for the property P. Hence, if we are able to integrate a property P into a metaphorical assertion that is supposed to be true about God, then we always need to have a concept for P, which in turn enables us to introduce an expression E that is conventionally related to C and thereby literally expresses P. This undermines panmetaphoricism since the metaphorical language about God can, according to Alston, not be understood as in principle irreducible to a literal paraphrase.

Josef Stern criticized Alston’s objection against panmetaphoricism. He claims that with de re metaphors we can refer to properties which we cannot fully conceptualize. Hence, as was already described above, cognitive access to properties we can refer to metaphorically does, according to Stern, not presuppose conceptual access to the respective property. In fact, in the case of de re metaphors, we do not lack words but concepts and full understanding. This might then, according to Stern, enable us to be consistent panmetaphoricists because all predications about God could be made with irreducible de re metaphors (see Stern, 2000, 193 f.).

Although Sterns argument is convincing against the position Alston proposes, it can be put under pressure when we allow the introduction of new concepts. As we saw above, it is without further problems imaginable that the metaphors (13) and (16) become conventionalized over time if they are used to communicate socially accessible contents. Then it might be argued against the consistency of a de re panmetaphoricism that the irreducibility of de re metaphors is just a contingent fact. We can always introduce concepts for the properties referred to by the metaphors. Therefore, one might argue that panmetaphoricism, also in the de re version, is not a consistent position since the assertions about God are not and cannot be necessarily and irreducibly metaphorical when supposed to be true. It is always possible that we establish concepts for the respective properties that can then be conventionally related to certain expressions.

But Stern’s argument can be repaired if we introduce a distinction within the set of de re metaphors. When Stern explicates the reasons for the plausibility of panmetaphoricism, he argues that we cannot fully understand the nature of the divine properties and can, therefore, only “point” at them through the context-sensitive mechanisms of metaphor (see Stern, 2000, 194). This description fits well with the heuristic metaphors introduced by Camp. As was delineated above, she presents as an example the case when the cognitive process of memory retrieval is described as the opening of a computer file. Such a metaphor can be understood as de re because we have no established concept for the property we are trying to capture. At the same time, we are also not in the position to introduce a concept for the property since the metaphor is a first access to the respective property, which we are not fully able to identify. Hence, we might specify the position of Stern in the sense that for panmetaphoricism to be consistent, the assertive speech about God cannot consist in de re metaphors such as (13) or (16), but must be completely made up of heuristic de re metaphors as they were described above.²³

Therefore, Alston’s argument that panmetaphoricism is from a linguistic point of view not consistent, is not convincing. Still, this does neither imply the plausibility nor the attractiveness of panmetaphoricism. First, it is for linguistic reasons difficult to offer a convincing justification for panmetaphoricism. Such justifications would have to involve claims about the nature of God from which we can derive that we can only make true assertions about God when we use metaphors. If we can only hint at God’s nature with heuristic de re metaphors, there is great danger that the respective justifications are all highly arbitrary. They would all leave room for several interpretations. And since there are no literal statements we could truthfully make about God, it is difficult to see how we can establish constraints on the possible interpretations of the heuristic de re metaphors about God that are supposed to justify panmetaphoricism.²⁴

This is closely related and also directly leads to the next and perhaps most important problem for panmetaphoricism: Even if we accept panmetaphoricism, it is not clear which criteria we have to evaluate which metaphorical statements about God are more apt than others. It is not clear how panmetaphoricists can offer substantial reasons why not all metaphors to some degree adequately point to certain properties of God. To illustrate this point, consider cases where we judge metaphors to be more apt than others. If we say that Charles is more aptly described as a sheep than as a shark, we justify such a claim with further statements about Charles. We could for example say that Charles is very friendly and helps other people a lot, perhaps sometimes even too much. He, for example, has problems with saying “No” and is, therefore, an easy victim for people who want things to be done by him. It does not seem apt to metaphorically characterize Charles as a shark, since sharks are commonly associated with aggression, danger, and violence. Jonathan can on the other hand not aptly be described as a sheep, since he is violent and aggressive. If you meet him, you are always in danger that he snaps, when you say or do something wrong, while his attacks can either be verbally or physically. In both cases, the reasons for and against the aptness of a certain metaphor were given by literal statements that describe Charles and Jonathan. The same holds even more in cases of heuristic de re metaphors. The metaphorical usage Camp described when a sub-personal process is conceived of as the opening of a computer file only enables us to get a certain glimpse of the property we are looking for because there are important constraints in place. We know what sub-personal cognitive processes are and we know a lot about our biology and neurological processing specifically. Hence, the metaphor only succeeds if these constraints enable to get hold onto some useful analogies. If there are no such constraints in the form of predicates that are at least assumed to truthfully apply to the target, i.e. the entity talked about metaphorically, there is not much hope that we have any chance to delimit the potentially endless chain of associations and analogies the metaphorical interpretation evokes. All metaphors about God then just come out as equally apt. Daniel Howard-Snyder infers from this situation that panmetaphoricists cannot say why one claim about God is true and another false. Therefore, they also cannot substantially reason about God and cannot express what is good about God. Hence, they have, according to Howard-Snyder, also no reason to believe that it is desirable to be in a proper relation to God. He concludes that panmetaphoricism, therefore, leads to “acute religious anaemia – or worse” (Howard-Snyder, 2017, 47). Even if one does not go as far as Howard-Snyder, it should be clear that panmetaphoricists have great trouble in explaining, why some metaphors are more apt than others when applied to God. The usage of heuristic de re metaphors does not help here but only makes the problem more severe. Such metaphors develop their explanatory value against a background of assumptions about the target. If these assumptions are also all made up of heuristic de re metaphors, it is not clear where the possible interpretations should get their constraints from. Neither for theologians nor religious believers this seems to be an attractive consequence.

Still, it has to be admitted that these arguments put forward against panmetaphoricism are certainly not refuting the position. They are just considerations that put pressure on panmetaphoricism. In contrast to Alston, it is not claimed here that we can only on the basis of linguistic considerations about the relation between metaphors and literal language infer the inconsistency of panmetaphoricism. The plausibility of panmetaphoricism also depends on its alternatives. Thomas Aquinas famously argued similar to Kenny and other panmetaphoricists that our cognitive capacities are deficient when it comes to the description of God. For Aquinas it is our modus significandi that demands the composition of subject and predicate and is, therefore, not suited to adequately describe God. Because of God’s ontological simplicity, we can, according to Aquinas, with the necessarily compositional structure of our language and thinking not grasp God’s nature as it really is. Still, Aquinas does not claim that we can only speak about God metaphorically. Predicates of perfections such as “is wise” or “is good” are for Aquinas used properly (proprie) about God and not only in a metaphorical way (ST I,13,1–3). Still, the expressions for predicates are for Aquinas not used univocally, i.e. with the same meaning as they are used in mundane contexts but with analogical, i.e. similar meanings when applied to God (see for example ST I,13,5). Other authors such as Swinburne (2016, 66–77) or Wolfhart Pannenberg (2011, 181–201) developed their own theories of analogical language that crucially departs from the one of Aquinas. Alston (2005) proposes a theory based on “partial univocity” where a univocal core between the mundane usage of certain predicates and their meanings when applied to God exists in order to avoid problems he sees in the approach of Aquinas. The plausibility of these and other models of analogical language cannot be discussed in detail here. They share the interest to go beyond metaphorical interpretations concerning at least some predicates when they are applied to God. The point made in this section is that metaphorical interpretations need constraints to be helpful devices for communication. And if there are no constraints in the form of assumed literal truths about a subject matter, the interpretations of the metaphors about that subject matter are in great danger of becoming arbitrary.²⁵

Conclusions

With the help of de re metaphors it was possible to protect panmetaphoricism against Alston’s charge of inconsistency. Nonetheless, panmetaphoricism is confronted with the problem of formulating constraints for the different possible metaphorical interpretations of predications about God. It is not clear that there are enough resources to significantly reduce the indeterminacy of metaphorical interpretations without a theory of analogical or partially univocal theological language. Still, it is highly plausible that metaphors belong to the most promising and important devices to approach a radically distant reality such as the God of the monotheistic religions. First, metaphors produce an imagistic perspective that enables a framing of the target of the metaphor. This can in many cases function as a way to relate the descriptions of a divine reality to our sensible experiences. Second, with the de re metaphors we are able to express propositional contents we can, at least at a certain moment, not express with literal language. Especially heuristic de re metaphors can be seen as important tools to reach out to areas we do not have good conceptual access to. It is, therefore, reasonable to assume that the progress in theological modelling will heavily and perhaps necessarily depend on the usage of metaphors.