Shepherd on Meaning, Reference, and Perception

The Meaning of Natural-Kind Terms

Shepherd’s account of mental representation is intimately tied to her account of the causal relation, both because objects’ causal powers are the primary objects of our mental representations, and because the mind itself and its representative powers are part of the causal structure of the world. Shepherd articulates and defends her account of the causal relation in An Essay upon the Relation of Cause and Effect (ERCE). Here is Shepherd’s definition of ‘cause’.

A Cause, therefore, is such an action of an object, as shall enable it, in conjunction with another, to form a new nature, capable of exhibiting qualities varying from those of either the objects unconjoined. This is really to be a producer of new being.—This is a generation, or creation, of qualities. (ERCE 63)

As Shepherd sees it, cause and effect is the relation whereby two objects combine to bring into existence some third object. The two objects combined are the causes. The new object is the effect.

An Effect is the produced quality exhibited to the senses, as the essential property of natures [causes] so conjoined. (ERCE 63)

Just as causes are those objects that combine to create a new object, effects are complementarily those objects that result from that combination. Notice that in both passages Shepherd refers to the creation of new natures and qualities. It is important to note that, for Shepherd, an object necessarily has all and only the qualities that it does.

But an object is nothing else (in relation to us,) than a mass of peculiar qualities; and when observations inform us, that any known mass is produced by similar circumstances, on various occasions; such mass or object must necessarily contain all its qualities, and be equal to exhibit all its effects in hitherto untried events. (ERCE 53–54)

‘An object must necessarily contain all its qualities’, so the production of any new quality (an effect) amounts to the production of a new object.¹ Effects are these new objects; causes are the objects that combine to create them. Notice the final clause in the quoted sentence: an object must contain all its qualities, and ‘be equal to exhibit all its effects’. The qualities of an object are themselves results of the causal combinations that produce them, and in turn, serve as causes of other such objects. The qualities of an object are specifically causal properties. Therefore, objects are essentially capable of combining to form new objects and are equally essentially the products of previous combinations (EPEU 281–82).²

As Shepherd sees it, these causal powers are represented by our use of words for these objects. The meaning of natural-kind terms includes information about the causal powers that we take objects of that kind to have.³

Therefore fire, in order to have a right to the sign of the word fire, for an expression of its attributes, in order to be a ‘like cause,’ must of necessity burn as much as it must be red, otherwise the red object were not fire; and could not have been produced by those causes that elicit that element. (ERCE 54)

An object is correctly denoted by ‘fire’ only insofar as the object has all the causal powers implicitly attributed to that kind of object. For example, an object is a fire only if it is caused by the combination of, say, a spark and some kindling, and is in its turn the cause of burning, and sensations of heat and certain colors. Again,

It becomes therefore part of the definition of fire to burn certain bodies, to melt others; of bread to nourish the human body; of snow to be cold, and white; and these qualities they must have, in order to compose that entire enumeration of qualities, for which appropriate names have been formed, and to the exhibition of which similar and efficient causes have been in action. (ERCE 55)

The word ‘fire’ contains as part of its definition that it is applicable only to those objects that burn some bodies, melts others, etc. Similarly, what we mean by the word ‘bread’ includes that which comes from wheat and nourishes the body; what we mean by ‘snow’ is that which is cold and white and falls from the sky, etc. Of course, in defining ‘fire’ as that which arises from the combination of a spark and some kindling, we likewise implicitly give partial definitions of ‘spark’ and ‘kindling’ as those objects that combine to create fire. And, of course, fire can be used to cook bread or to melt snow, and so both the causal powers of all these objects, and the definitions of the words that designate them, are inter-related. The full definition of ‘fire’ will make use of the words ‘spark’, ‘kindling’, ‘bread’, and ‘snow’, and vice versa, so each of these words is inter-defined. That is, Shepherd appears to be committed to a version of meaning holism: the meaning of any word depends on its place in a system of interrelated meanings.⁴

Of course, thus far, all of these are claims about the meaning of the term ‘fire’, and it likely occurs to a contemporary reader to wonder how Shepherd thinks of their referents. That is, while we might mean by ‘fire’ that which is caused by the combination of spark and kindling, and that which causes burning, etc., whatever meets such a description in our world, might well have different causal properties in a different possible world. Fire, that which burns in our world, might not have had that power in another world. Meanings serve to fix a referent for a term in our world, but once that referent is fixed, that term refers to that referent in all possible worlds in which it exists, even ones in which it has different causal powers. So, isn’t Shepherd simply mistaken to claim that, ‘fire […] must of necessity burn’? Isn’t this to run together a claim about the meaning of the term ‘fire’ with a claim about the object to which that claim refers?

To answer this question, we will need to make use of the full apparatus of Shepherd’s account of mental representation, specifically her account of perception, to which we will soon turn. In the meantime, we can notice now just how tricky an answer to this question is going to be. Recall that we saw a moment ago that an object necessarily has the causal powers that it does. So, whereas we might think of the term ‘snow’, for example, as meaning the cold flaky white stuff that falls from the sky, but referring to, say, the specific crystalline structure of what happens to meet that description in our world, these two aspects of the term ‘snow’ cannot be distinguished so easily for Shepherd. For Shepherd, the meaning of ‘fire’ will consist of specifically causal descriptions, and since an object necessarily has the causal powers that it does, what meets that description in our world, say ‘that which burns’, has those same causal powers in all possible worlds in which it exists. This is precisely because being the cause of burning is part of the very essence of that object.⁵ Thus, if the only way we fix reference is via a specification of causal powers, and an object necessarily has the causal powers that it does, then meaning and reference cannot come apart. So, if we specify the referent of ‘snow’ via a specification of those causal powers, then it does follow that snow, qua that object in this world that causes sensations of cold, necessarily causes sensations of cold in all possible worlds because what is referred to in this way is necessarily something with that causal power.⁶ More on this later.⁷

For the moment, the important takeaway is again that Shepherd appears to be committed to a version of meaning holism: the meaning of any word depends on its place in a system of interrelated meanings. Meaning holism is a familiar position in contemporary philosophy of language (e.g., Brandom 1994). However, Shepherd’s version might strike one as particularly implausible. For example, it might appear to imply that we can know the causal powers of worldly objects a priori, by doing nothing more than reflecting on the meanings of our terms.⁸ Or, to frame what amounts to the same concern slightly differently, that all true propositions about the causal powers of objects will turn out to be true by definition.⁹ Those are serious worries of which Shepherd appears to be aware. Her response appeals to the possibility of discovering that objects in the world may not be properly designated by the words we use for them.

It follows then from the definitions given in the preceding section, and the reasonings on which they are formed, that were a body in all other respects resembling snow, to have the taste of salt and feeling of fire, it would be an extraordinary phenomenon, no doubt; and one which might for ought we know take place, but it would not be snow; and such a body could not fall from the clouds but by new causes efficient to its formation;—it would therefore, be entirely a different object, and would require a new name; and the phenomenon could offer no ground for the conclusion, that reason does not afford an argument, for the expectation of similar effects from similar causes. (ERCE 69)

Shepherd holds that if something white fell from the sky, but caused in us sensations of saltiness and burning, then the object would not be snow or be properly be named ‘snow’. This object would, ‘require a new name’. Again, objects of the natural kind snow necessarily have the causal powers of snow, so any object with different causal powers, would of necessity be a different kind of object. According to Shepherd, while it is true by definition that for something to be properly called ‘snow’ it must feel cold, before touching whatever is falling from the sky to feel it, we cannot yet know that it will feel cold, and so cannot yet know that it is worthy of the name ‘snow’. What we might call our conceptual scheme represents that certain kinds of objects bear certain causal connections to one another, but it might be that the kinds of objects it represents are not the kinds of objects with which we meet in experience. The empirical aspect of our experience is not to be found in the meanings of our object terms, but rather in whether those object terms name anything in the world.¹⁰

Again, a contemporary reader might interject here that it is only a contingent fact about snow that it is cold, et al., and that we come to know this fact about it only a posteriori. Notice, though, that formulating this thesis requires that we have a way to refer to snow independently of its causal powers. If, for example, I could refer to snow as merely ‘that stuff over there’ then it would be an open question whether that stuff over there feels cold. What I will suggest, however, is that for Shepherd the only way we have to fix the referents of our terms is via a description of their causal powers. ‘That thing over there’ is always shorthand for ‘that thing over there with such-and-such causal powers, including causing this perception in me’.¹¹ And again, since an object necessarily has the causal powers that it does, in representing that object over there as being of a certain kind, we attribute to it certain causal powers which causal powers it would have in all possible worlds in which it exists. So, all such attributions will be necessarily true. The only question possible, then, about such an object is whether anything meeting that description in fact exists in our world.

But how do we determine whether our object terms name anything in the world? That is, how do we test our conceptual scheme as a whole in the crucible of experience? In order to answer this question we must look beyond Shepherd’s account of definitions to her account of perception: the mental faculty by which definitions are applied to objects of experience.

Shepherd’s Account of Perception

Shepherd’s meaning holism raises three distinct but related questions:

1. How do we represent that the world contains objects of the kinds represented by our definitions of natural-kind terms? For example, how does one represent not just that fire burns, but also that the world contains fire?

2. How do we represent not just the natural kind fire, but also that this particular object here is fire?

3. Once we represent that the world contains fire, and that this object here is fire, how do we go about answering the question of whether this particular object here is, in fact, fire?

Shepherd’s distinctive account of our representation of causation answers these questions. But in order to see how it does so, we must first examine Shepherd’s account of perception.

Shepherd stipulates at the outset of EPEU precisely how she will use certain terms for mental items, and how her usage differs from those of her predecessors (most prominently, Berkeley, Hume, and Reid). Firstly, she uses ‘sensation’ as the most generic term for any mental item: ‘sensation […] is a generic term, comprehending every consciousness whatever’ (EPEU 6). As she notes, this makes her use of ‘sensation’ much like Hume’s use of ‘perception’.

All the perceptions of human mind resolve themselves into two distinct kinds, which I shall call impressions and ideas. (T 1.1.1.1; SBN 1)¹²

What Shepherd calls ‘sensations’ and Hume calls ‘perceptions’ include the most generic constituents of the human mind.¹³ Under this genus fall a variety of species, such as, ‘thought, notion, idea, feeling, and perception’ (EPEU 7). We need not concern ourselves with the differentia of these species, since Shepherd’s concern, and ours, is with perception, which she defines as, ‘a “consciousness of sensation,” a sensation TAKEN NOTICE OF BY THE MIND’ (EPEU 9). At first blush this definition seems to indicate that first we have some sensation and then the mind takes notice of that sensation. So, perception is, or depends on, a kind of meta-cognition, a mental state of awareness of another, independent mental state—an existing sensation—that precedes the awareness of it. Shepherd’s later account of perception in EPEU, however, complicates this relatively simple picture.

Now I repeat there is one sense in which it may be said that objects are perceived immediately, as existing outwardly, by the senses. It is this; the conscious powers of the understanding, and the senses, are blended together in man; we are analysing them, but in nature they are united as intimately as are the prismatic colors in one uniform mass of light. (EPEU 67)

Here Shepherd is clear that while sensations are the most generic kind of mental entity, they are not temporally, metaphysically, or epistemologically, the most fundamental kind of mental entity. That status belongs to perception. While it is possible to analyze perception as composed of the understanding on the one hand, and the senses on the other, these faculties are united in a single uniform faculty of perception. It is not the case that first we have some sensation, and then through a numerically distinct mental act, a perception that is a kind of awareness of that sensation. Rather, our first encounters with the world are perceptions, which can be analyzed into something like their nominal parts: those parts that result from the exercise of the senses and those parts that result from the exercise of the understanding. While we might be able to separate out the contributions of the senses, doing so is the result of an analysis of the more fundamental faculty of perception.¹⁴

How does Shepherd’s account of perception as consciousness of sensation fit with her later account of perception as a combination of the understanding and the senses? Perhaps the simplest answer is that the senses provide the sensation at issue (in this case a sensible quality) and the understanding provides the consciousness of that sensible quality.¹⁵ Notice that in the later passage, Shepherd refers to ‘the conscious powers of the understanding.’ One might wonder what the understanding does to make the sensible quality conscious. Shepherd answers that question later in the same passage:

This being the case, they [the senses and understanding] are acting in concert when any object affects the senses. Therefore the understanding knowing the simplicity of mental sensation, it follows, that the varieties of the causes, (which create varieties in the effects,) are instantly perceived and detected, and that immediately with the conscious use of the senses; whilst also the mind as immediately mixes that idea of which the understanding is aware; viz. that these varieties, as complex objects, continue to exist unperceived and independant, when unnoticed by the senses. (EPEU 67–68)

What the understanding ‘adds’ to sensible qualities in a perception to make them conscious is knowledge of (or at least hypothesis about) the causes of those sensations: ‘the varieties of the causes […] are instantly perceived and detected.’ For one to become conscious of a sensation is to think of that sensation as being caused by some external object.¹⁶ Shepherd repeats this claim later in EPEU, where she implicitly distinguishes being conscious of a sensation with merely contemplating one:

The perception of external, continually existing, independant objects, is an affair of the understanding; it is a mental vision; the result of some notions previously in the mind, being mixed with each sensation as it arises, and thus enabling it to refer the sensations to certain reasonable causes, without resting merely in the contemplation of the sensations themselves ; by which it comes to pass, that names stand for these compound mixtures ; and that the organs of sense are the instruments which immediately detect the presence of those things which are external to, and independant both of the organs of sense and the mind. (EPEU 168–69)

Taking sensible qualities to be caused by external objects ‘is an affair of the understanding.’ Shepherd adds here that taking sensible qualities to be caused by external objects constitutes a ‘mental vision’ that is a combination of sensible qualities and understanding, ‘enabling it to refer the sensations to certain reasonable causes.’¹⁷ A sensible quality may be merely contemplated. But in perception a sensible quality becomes part of a mental vision that refers to an object insofar as that perception itself represents the sensible quality that is its nominal part as an effect of that object on the mind.¹⁸ This understanding of sensible qualities as caused by objects is what conscious awareness of sensations is over and above mere contemplation. It is because the understanding is responsible for causal reasoning that it is implicated in perception along with the senses.

We can now return to the question of whether conscious awareness of a sensible quality as being caused by some external object is something added to an otherwise non-conscious sensible quality, or whether it is from the perception as a whole that we derive the notions of the deliverances of the senses and those of the understanding respectively. Notice Shepherd’s repeated use of the idiom of immediacy and instantaneousness in the passages above. In the first passage, the causes of perceptions are ‘instantly perceived and detected,’ and this occurs, ‘immediately with the conscious use of the senses,’ and, ‘the mind immediately mixes that idea of which the understanding is aware.’ In the second passage, the notion of the cause of a sensible quality is, ‘mixed with each sensation as it arises,’ and perceptions, ‘immediately detect the presence of those things which are external.’ Shepherd’s insistence on immediacy supports reading perception as a primary unity, of which the senses and understanding are mere theoretical abstractions. Our most fundamental kind of sensation is a perception: an awareness-of-sensible-qualities-as-caused. The hyphenation here indicates that such awareness is not a conjunction or congery of otherwise disparate sensations, for example, first a sensible quality, then a numerically distinct awareness of it as caused, but rather as a single sensation that is a unity of what can, but need not, be analyzed into an aspect that is a sensible quality and an aspect that is an awareness of this sensible quality. We immediately perceive external objects via a perception, and only upon subsequent philosophical analysis do we abstract from this original perception to the notions of the contributions of the senses and understanding.

Shepherd presents her account of our mental faculties in contrast to Berkeley, Hume, and Reid, whom she argues are wrong about the nature of perception. She begins by noting the three components of perception that we have discussed: the object that causes the perception, the contribution of the senses, and the consciousness of sensible qualities as caused by the object that is contributed by the understanding.

The union of the three following things are required to form the proximate cause for that great effect, the formation and combination of those aggregates of sensible qualities usually called objects; namely, first, the unknown, unnamed circumstances in nature, which are unperceived by the senses; secondly, the organs of sense, whose qualities mix with these; and thirdly, the living, conscious powers necessary to sensation in general. (EPEU 71–72)

Shepherd declares that her predecessors’ mistake is to take each of these components as numerically distinct, rather than merely nominally separable aspects of the single fundamental faculty of perception.

And they all, as I think, err in this, in considering them [‘what Berkeley calls “ideas,” and “sensations in the mind”’] as first formed, and then contemplated, and taken notice of afterwards. Whereas, the sensible qualities of things are only formed by being taken notice of. (EPEU 72)

Sensible qualities are formed by being taken notice of. And the taking notice of a sensible quality is a perception. And a perception is a combination of sensible qualities and reasoning about their cause. So, it is not that sensible qualities come first, and then a hypothesis about their cause is added to them in perception, but rather that perception comes first—a representation of a sensible-quality-as-caused-by-some-object—which perception is only subsequently treated as involving ‘distinct’ contributions of the other faculties.

I return to the philosophical consequences of the priority of perception in Shepherd’s refutation of Berkeley, Hume, and Reid in the final section. Now we can return to the questions posed at the beginning of this section. Recall Shepherd’s holistic account of natural kind terms. Each kind term includes in its definition the causal relations that the objects to which it refers bear to other kinds of object. ‘Fire’ designates that which is caused by the combination of spark and kindling, and in turn causes burning and sensations of heat, and ‘fire’ means what it does in virtue of the role it occupies in a system of such definitions. At the end of the previous section, we wondered how a holistic conceptual scheme gains purchase on the world of objects it represents, especially in light of the fact that it appears that claims about the causal structure of that world would be a priori rather than a posteriori.

What we are now in a position to see is that it is via the perception of fire that a conceptual scheme containing ‘fire’ gains this purchase on the world. Because perception consists of a combination of a sensible quality—say, heat—and reference to that sensible quality’s cause—fire—a conceptual scheme containing ‘fire,’ deployed in such a perception, purports to pick out a case of fire in the world. In perceiving the sensible quality of heat as being caused by fire, the conceptual scheme acquires representational purport: it is not merely a conceptual ‘frictionless spinning in the void’ (McDowell 1994: 67). The conceptual scheme is directed at the external world as the causal explanation of our experience of that world. Because perception is our most fundamental cognitive faculty, while we can abstract the contribution of the understanding to perception to form a notion of a set of reciprocally-defined terms that do not bear a relation to the world, such a notion is derivative of one that connects immediately to the world it represents. Our concept of fire is first acquired in perception: in the nominally compound, but actually unified, act of explaining the existence of a sensible quality of heat in our mind via an appeal to its external cause. Consider Shepherd’s claim about the meaning of the word ‘table’.

Now the truth is, that no real table is formed, no image of a table is formed, unless the whole united mass of the unknown objects in nature exterior to, and independent of the instruments of sense, (not yet worthy of the name of ‘TABLE,’) unite with the mechanical action of these, and by their means with the sentient principle, in order to create in such a union that object which alone can properly be termed ‘TABLE.’ (EPEU 148–49)

Contrast the view that Shepherd articulates here with that of someone like Hume. For Hume, the word ‘table’ serves as a prompt to recall certain sensible qualities arranged in a certain manner, for example, color points in a table-shaped array. By contrast, for Shepherd, the meaning of the word ‘table’ specifies a table’s causal qualities, including that a table is the kind of object that causes the kind of array of sensible qualities on which Hume focuses. The contributions of the senses alone are not sufficient for explicating the meaning of ‘table’, but neither, presumably are the contributions of the understanding. Representing a table requires more than merely thinking of the causal relations in which tables stand to other objects in the abstract. It requires perceiving a table: thinking of the table as that which causes my sensible qualities of it. Shepherd continues,

Yet after experience, the OUTWARD OBJECTS, the CONTINUALLY EXISTING PARTS of the whole causes necessary to the creation of a table, must be named by the name by which the whole is named; for there is no other name whereby they can be called, nor any other ideas by which the memory of them can be introduced into the mind, save by the appearance of ‘the faint images of those sensible qualities’ which their presences originally created. (EPEU 148–49)

There is no application of the word ‘table’ other than to the outward objects that are the causes of the sensible qualities, ‘which their presences originally created.’ Neither the sensible qualities themselves nor the entirely general definitions of ‘table’ are alone sufficient to account for the meaning of the word ‘table’. The meaning of that word is constituted both by its place in the conceptual scheme of which it is a part and by its reference to the unknown outward objects that are the causes of our sensible qualities.

The passage above also brings out another important feature of Shepherd’s view. Notice that Shepherd emphasizes that the name ‘table’ names the continually existing parts, ‘of the whole causes necessary to the creation of a table,’ and insists that, ‘there is no other name whereby they can be called, nor any other ideas by which the memory of them can be introduced to the mind.’ Here Shepherd is insisting that the meaning of ‘table’ includes not just this or that contingent feature of tables, but rather all of the causes necessary to its creation. That is, the name table describes the entire causal role that we take tables to occupy, and Shepherd holds that it is only in virtue of this holistic definition that we can name tables at all.¹⁹ This point is of importance to us because it resolves a question that we had considered earlier. Can’t we refer to a table as something like, ‘whatever object caused my current sensible quality’, and leave open the question of whether it has the causal properties of a table? Aren’t questions of what causal properties belong to tables answered a posteriori? We can now see why Shepherd’s answer to this question is ‘no’. Shepherd holds that the only way to fix a referent of a term is via the description that constitutes its definition;²⁰ she also holds that these definitions describe a holistic set of causal powers. So, what is described by ‘table’ is something with all the causal powers of tables, and since an object necessarily has the causal powers that it does, what is described by this holistic definition must have the causal powers that it does in any possible world in which it exists. The relevant a posteriori question is not whether tables have such-and-such causal properties, but rather whether anything described by the definition of ‘table’ exists.

Another way to put this latter question, is to ask whether ‘table’ designates some real object, and here is Shepherd on what it means to be real:

they are real which fulfil the definitions for which their names were first formed. The being true to expectations formed of their qualities, is the very criterion of reality. (EPEU 188)²¹

What it means for an object to be real is to fulfil the definition for which its name is formed. Therefore, a definition cannot misdescribe its object, if that object exists. A definition can fail to correspond to any object, but insofar as its object is real, its definition must fit it. This is precisely because the only way we have to fix the referents of names is via a holistic description of their causal powers: ‘there is no other name whereby they can be called.’

While it is true that ‘fire’ has as part of its definition that it is caused by the combination of flint and steel, it also contains as part of its definition that it is the unknown outward cause of my sensible quality of heat. What this earns Shepherd is the answer to at least one of the questions posed at the beginning of this section. Our conceptual scheme purports to be about the world insofar as its constitutive definitions include in their scope that certain kinds of objects are the reasonable causes of our sensations. It also suggests an answer to another of those questions: in virtue of what does one represent not just fire in general, but also that this particular object is fire? Since part of the definition of ‘fire’ is that which causes not just some sensation or other, but also this sensation of which I am currently aware in perception, it is by deploying the concept ‘fire’ in perception that natural kind terms can be used to refer to particular objects of that kind.

I answer as before, that by reason the mind judges that the causes of those sensations in particular, which come under the definition of external objects, must needs be out of, and distinct from the mind, or the cause of sensation in general; for the notion of outward existence does not suit the definition given to inward existence: Inward existence is the capacity for sensation in general; outward existence is the exciting cause for some sensation in particular. (EPEU 40)

Note that Shepherd includes being the cause of particular sensations as part of the definition of external objects. It is not the understanding alone that provides complete definitions of objects, but more fundamentally perception: the unity of understanding and the senses. It is in representing our sensations as caused (perception) that we first hypothesize that there are objects of certain kinds, with certain causal powers. It is thus part of the definition of fire that it is of a kind that includes objects with the causal powers appropriate for causing this very sensation. As we have seen, a sensation of fire-as-the-cause-of-this-sensation is precisely what a perception of fire is. Natural kind terms represent particular objects of those kinds because they first come to be defined in the context of perception, in which objects of particular kinds are hypothesized to be the causes of particular sensations.

A question remains: how does one learn whether this particular object here is, in fact, fire? If ‘fire’ means, in part, whatever object it is that causes this sensation of which I am aware in perception, then does it follow that whatever is the cause of this sensation is thereby fire? Can’t we be wrong about the causes of our sensations? Shepherd’s lengthy treatment of dreaming in EPEU indicates that we can. How is this possible?

If ‘being the cause of my sensation of heat’ were the sole definition of fire, then we could not be wrong about the causes of our sensations. But it is not the sole definition of fire. ‘Fire’ refers to that which causes my current sensation of heat, and is caused by the combination of spark and kindling, and causes burning, etc. ‘Fire’ is applicable only to objects meeting all of the parts of its definition. If what is causing my current sensation of heat is not fire, but, say, radiation, then what follows is that the concept ‘fire’ has an empty extension, so to speak: it is not applicable to anything.²² Of course, something very similar to fire might still exist in the world: something that has all the causal powers of fire other than causing this particular sensation of heat. And we might use the sign ‘fire’ to designate that thing as well, but in doing so, we are strictly speaking redefining that term. Thus, while any true claims about the causes and effects of fire are true by definition, this leaves open the possibility that no such thing exists. Our conceptual scheme as a whole purports to represent the causal relations between objects in the world as a whole, but it might not do so. It might be that we discover that we are in need of some entirely different conceptual scheme.

Thus do we find Shepherd describing the process whereby we engage in hypothetical reasoning about the causal powers of objects in relation to our sensations.

the unknown causes of all our perceptions, are as the unknown quantities in algebra, which yet may be measured, valued, reasoned on by their signs; and the signs of these outward objects are the sensations they can create; and they may always be spoken of, and compared together, as though they did truly exist, in these forms in which they appear to the mind. (EPEU 47–48)

We use the signs of objects, sensations, in order to form a hypothetical picture, ‘the forms in which they appear to the mind’, of how objects relate to each other in the world, ‘as though they did truly exist’. These forms together constitute the conceptual scheme of interrelated definitions that I have been describing. The question that we then confront about such a conceptual scheme is whether the objects that it portrays ‘truly exist,’ as it portrays them, ‘in these forms’. Or, as we earlier saw Shepherd put it, whether the objects described by such a conceptual scheme are ‘real,’ that is, whether the objects represented fulfill their definitions.

One way to put my thesis, then, is that for Shepherd, we bring hypotheses about causal powers to perception in order to test them in the crucible of experience, rather than, say, deriving these hypotheses from experience itself. Shepherd does, however, sometimes write in a way that seems to imply the opposite order. For example, in arguing that it is reason rather than custom that affords us a notion of cause and effect, she reconstructs the reasoning at work in perception as follows.

Here is a new quality, which appears to my senses: But it could not arise of itself; nor could any surrounding objects, but one (or more) affect it; therefore that one, (or more) have occasioned it, for there is nothing else to make a difference; and a difference could not ‘begin of itself’. (ERCE 43–44)

The point that Shepherd is making here is that, contra Hume, repeated experiences of co-occurrence are not necessary for us to form a hypothesis about relations between causes and effects because the causal maxim affords us a principle that allows us to determine causes from a single instance. What is important to us just now, though, is her description of how this takes place. That description suggests a picture of causal reasoning according to which we first encounter some new object or quality, the causal powers of which we have no representation. Then we engage in some experimental reasoning about its hypothetical cause, and finally conclude that such-and-such an object (of an unknown kind) is the cause of that new quality. That is, it suggests that we do not come to experience with a set of definitions that represent a hypothetical world of causes, and test it against the actual world. Rather we observe the world first, and then use reason to deduce its causal structure.

However, this appearance is belied a short while later when Shepherd reflects on this very same process.

If then an existence now in being, conjoined with any other, forms thereby a new nature, capable of exhibiting new qualities, these new qualities must enter in to the definition of the objects; they become part of their natures; and when by careful experiment, or judicious observation, no new prevening circumstances are supposed to make an alteration in the conjunction of the same bodies, the new qualities, that are named effects, are expected without a doubt to arise upon every such conjunction; because, they as much belong to this newly combined nature, as the original qualities did to each separate nature, before their conjunction. (ERCE 47)

Suppose we have two objects that we know to have certain causal powers. The definitions of those objects will include those causal powers. So, for example, we might represent that fire causes sensations of heat, and snow sensations of cold. Now we combine those objects and discover that they have previously unknown causal powers: the fire melts the snow. Notice that what Shepherd describes as happening in this case is that, ‘these new qualities must enter into the definition of the objects’ and that this new effect is subsequently, ‘expected without a doubt to arise upon every such conjunction.’ What this suggests, again, is that rather than learn, of fire, that it melts snow, as Shepherd understands it, the primary role of empirical observation is to allow us to test whether the definition we already have of fire is a good one (accurate and complete). When we discover that it is not, because it does not include melting snow among the powers of fire, we must alter that definition. We change what we mean by ‘fire’ so that it includes melting snow, and we thereby come to a definition of fire (and snow) that comes closer to representing something that ‘truly exists’ or ‘fulfills its definition’.

What of the earlier passage, then, where Shepherd describes us as first encountering a new quality, then searching for its cause, and only subsequently hypothesizing the nature of that cause? Well, I would suggest that that portrayal is an abstraction of the nominal parts of what I have described previously as a more fundamental unity. It is a just-so story. We can abstract from perception the different contributions of the senses and reason, and so portray these as sequential contributions, but really they are united. Because the causal maxim is an a priori principle of reason, we already, so to speak, come to experience expecting every event to have a cause. And because perception itself is suffused with reason, to even so much as perceive an object, is in the same sense, already to hypothesize it as the cause of the very sensation that is our perception of it.

How, then, do we know what the essence or essential properties, of a natural kind are? And what role does experience play in discovering these? Shepherd’s answer is that we use ‘good sense […] observation enables it [the mind] to judge, when an object is presented, what causes have been used in its formation’ (ERCE 102–3). The answer that I discern in Shepherd is that we guess, and then test those guesses against experience to see if what we have hypothesized, the ‘reasonable causes’, are real, or truly exist. To use Shepherd’s own analysis of the situation, one finds oneself with some sensation. Armed with the causal maxim, one knows that this sensation cannot have caused itself, so one hypothesizes that it was caused by some object in the local environment. Of course, since that object too must have a cause, one forms a further hypothesis about its cause, etc. Now, while I have, following Shepherd, described that process sequentially, because the hypothesized causes are all part of the definition of the (internal and external) objects at hand, that is potentially very misleading. Rather, in first thinking of a sensation at all, I think of it as a sensation-caused-by-some-external-object. And in thinking of that external object, I think of it as an object-caused-by-some-other-object-and-causing-my-sensation-of-it. Etc. That is, to think any of these objects individually, I think form a holistic set of definitions that represent them as causally related to each other, and indirectly to all other objects. (Recall again that Shepherd includes in the definition of bread such distal causes as the creator’s intention for us to benefit from the fruits of our labor.) Such a set of definitions, however, might not be applicable to the world; its objects might not be ‘real,’ or ‘fulfill their definitions,’ or ‘truly exist’. So, I observe the world, and the appearances of causes and effects in it, to determine whether this set of definitions will do, or whether I am in need of a new one. In a sense, that is the primary job of perception: to represent a particular sensation as having some particular cause, thus allowing us to put our entire holistic set of definitions into the crucible of experience.

As such, perception is our most fundamental representative faculty. In perception, we represent a sensation-as-caused-by-some-external-existence. We can analyze such representations into contributions of the senses, sensible qualities, and contributions of reason, becoming conscious of sensible qualities as caused, etc. The understanding’s contribution consists in part of a system of definitions that represents the world, including our minds, as having a certain causal structure. It is in virtue of this system’s including the very sensible qualities to which it is applied in perception that the system as a whole gains representational purport: that it represents not only kinds, but also objects as individuals of those kinds, and that it can either successfully or unsuccessfully be tested in the crucible of experience.

Meaning Revisited

With the substance of this interpretation now in hand, it is worth noting that others interpret Shepherd’s accounts of meaning and reference, and by extension perception, differently. For example, Lolordo (2019) confronts the worries about Shepherd’s claim that it is part of the meaning of ‘snow’ that it causes a feeling of cold, and offers a reply that is different from the one offered above.

Shepherd might think it’s part of the concept of snow that it’s cold, she does not in general think that truths about the causal powers of objects are analytic. (Lolordo 2019: 10n24)

Shepherd does not use the Kantian distinction between the analytic and the synthetic, but if she did, she would say that truths like fire consumes its fuel are synthetic, a posteriori necessities. (Lolordo 2019: 11n25)

One can understand wanting to avoid attributing to Shepherd the view that ‘snow causes a feeling of cold’ is analytic, but it is not clear how Lolordo proposes to do so, especially in light of her admission that, ‘it’s part of the concept of snow that it’s cold’. That sounds like the very definition of analyticity that Kant offers: that the predicate is contained in the concept of the subject (A6–7).²³ This suggests that these are analytic truths after all, although it remains an open question whether the system of definitions of which they are a part applies, as a whole, to the world. Either way, Lolordo’s description of how we learn what she casts as synthetic a posteriori necessities is precisely what we have already described.

How can I tell, for instance, whether the glass on the kitchen counter contains water or sulfuric acid? One option is just to examine it more closely. If the liquid is syrupy, it’s not water. A second option is to experiment with it, to see how it behaves in different circumstances. If a slice of lemon in the glass appears unaffected, it’s not sulfuric acid. (Lolordo 2019: 10)

Lolordo’s question is: how we can tell whether the glass on the kitchen counter contains water or sulfuric acid? I would reframe that question: how can we tell whether what is in the glass meets the definition of ‘water’ or ‘sulfuric acid’. Either way, the procedure for answering these questions is the same: perform experiments to determine the causes and effects of what is in the glass. My disagreement with Lolordo concerns how to interpret Shepherd’s claims about the meaning of terms like ‘definition’.

Boyle also endorses a view like Lolordo’s, interpreting Shepherd as anticipating Kripke’s necessary a posteriori propositions.²⁴ Boyle carefully distinguishes what she calls nominal or ‘absolute’ (ERCE 157) definitions from real definitions. A nominal definition is merely a name for certain enumerated sensible qualities. For example, we might use the name ‘gold’ to apply to a collection of sensible qualities such as yellow, shiny, metallic taste, etc. Real definitions, by contrast, aim at representing the causal powers of their object. Gold is that which causes those sensible qualities in us and also dissolves in aqua regia, etc. Boyle interprets Shepherd as holding that while we can know the nominal definitions of objects a priori, we have to discover their real definitions, their causal powers, a posteriori. Boyle also notes Shepherd’s explicit endorsement of Locke’s use of ‘definition’ in just these ways.

Kripke formulates his account of the meanings of names and natural kind terms against the background of Russell’s account of names as definite descriptions (Kripke 1980). Kripke objects to Russell in part on the grounds that we could discover that arbitrarily many (or even all) of our beliefs about, say, Plato are wrong. For that to be possible, the name ‘Plato’ would have to designate Plato independently of those beliefs. So, Kripke concludes that the meanings of names and natural-kinds terms are constituted entirely by their rigidly-designated referents, and not by their associated senses. ‘Water’ refers to that kind of thing over there in the lake (which, it turns out, is H₂O), regardless of what we might believe about it. Thus, we have to learn a posteriori that water is H₂O, even though when we do so, what we come to learn is necessarily true.

Consider the question of whether Kripke is right that a person could be wrong about arbitrarily many of her beliefs about water while successfully referring to water. If Lolordo and Boyle are right, then Shepherd holds that ‘water’ refers directly to the essence of some object, its real definition. While we can determine what the causal powers of that object are via experiments, we might turn out to have arbitrarily many false beliefs about those causal powers (especially before our experiments). I hold that Shepherd could not endorse this scenario. The key to seeing why is to see how Shepherd’s holistic account of the meaning of natural kind terms interacts with a causal theory of reference. In order for meaning and reference to come apart, one would have to be able to fix the reference of ‘water’ using some feature of water that it might not have in some other possible world. As Shepherd sees it, though, since the essence of any object is necessarily connected to its causal powers, and the holistic meanings of our terms purport to represent the causal powers of the objects that they describe, if those terms describe anything at all, they describe that object as having the causal powers that it has in all possible worlds. We can’t be wrong about whether snow is cold because ‘snow’ means that which is cold. So, if ‘snow’ describes anything in the actual world, it describes something with the causal power to produce sensations of cold. Since snow necessarily has its causal powers, if snow exists in any other possible world, it will likewise have the power to cause sensations of cold there as well. Anything that is like snow, but is not cold is not worthy of the name ‘snow’. The sense of ‘snow’ is part of the meaning of that natural-kind term, and since reference is fixed via the application of such natural kind terms to objects in perception, ‘snow’ also refers to an object only insofar as that object necessarily has the causal powers attributed to it. Nothing can be snow that does not fit that definition.

I agree with Lolordo and Boyle that the referent of natural kind terms plays an important role in Shephard’s account. Just as, for Kripke, we use ‘water’ to designate that kind of thing over there in the lake, part of the meaning of ‘snow’, for Shepherd, is that which is the cause of my present perception. However, because that aspect of its meaning is essentially constituted by its place in a holistic system of such definitions, whether what causes my sensation of cold is snow will also depend on whether it has the other causal powers appropriate to an object of that name. The a posteriori question that needs to be answered by experience and experiment is not, most fundamentally, whether my beliefs about snow are true, but rather, as we saw Shepherd put it earlier, whether snow is real, whether anything fulfils the definition of ‘snow’ (EPEU 188). Shepherd’s claim that perception consists of contributions from both understanding and the senses is best understood by interpreting her as committed to the claim that the meaning of a natural kind term is its holistic definition, including its role in being perceived, and its referent is that object which satisfies that definition in this, and all other possible worlds.

Shepherd’s Use of her Account of Perception

The present interpretation of Shepherd has profound implications for understanding her criticisms of her philosophical predecessors. For example, as Landy (2020b) points out, if we conceive of perception as the original fundamental faculty, then Hume’s opening gambit in the Treatise—his claim that our minds consist of just impressions and ideas, or as Shepherd would put it, sensible qualities and contemplations of sensible qualities—is a nonstarter. If impressions and ideas are themselves theoretical abstractions from the more fundamental faculty of perception, then Hume owes us an argument showing that we must artificially limit our philosophical resources to just these two aspects of it. Landy illuminates the problem by highlighting the different treatments of the meaning of the world ‘table’ that Hume and Shepherd offer. Here is Hume:

The table in front of me is alone sufficient by its view to give me the idea of extension. This idea, then, is borrow’d from, and represents some impression, which this moment appears to the senses. But my senses convey to me only the impressions of colour’d points, dispos’d in a certain manner. If the eye is sensible of any thing farther, I desire it may be pointed out to me. But if it be impossible to show any thing farther, we may conclude with certainty, that the idea of extension is nothing but a copy of these colour’d points, and of the manner of their appearance. (T 1.2.3.4; SBN 34, emphasis added)

Our complex idea of a spatial complex comes to represent the spatial complex that it does by being a collection of simple ideas of colored points arranged in a way that exactly resembles the arrangements of the spatial complex being represented.²⁵ We represent the relation that some simple impressions stand in relative to one another by arranging simple ideas of each of these impressions into the same relation. We represent a as being next to b by placing an idea of a next to an idea of b in phenomenological space. The idea of a spatial complex is nothing more than a spatial complex of ideas. Contrast that with Shepherd’s treatment of the same representation:

Now the truth is, that no real table is formed, no image of a table is formed, unless the whole united mass of the unknown objects in nature exterior to, and independant of the instruments of sense, (not yet worthy of the name of ‘TABLE,’) unite with the mechanical action of these, and by their means with the sentient principle, in order to create such an union that object which alone can properly be termed ‘TABLE.’ Yet after experience, the outward OBJECTS, the CONTINUALLY EXISTING PARTS of the whole causes necessary to the creation of a table, must be named by the name by which the whole is named; for there is no other name whereby they can be called, nor any other ideas by which the memory of them can be introduced into the mind, save by the appearance of ‘the faint images of those sensible qualities’ which their presence originally created. (EPEU 148–49)

Shepherd’s account of the representation of a table begins by drawing a distinction between what we see—a table—and what we see of what we see, say, the facing surface of the top of the table. While ‘the faint images of [the] sensible qualities’ of the table in some sense represent just the sensible qualities that correspond to the seen parts of the table—say a rectangular patch of brown—what our idea of the table represents, is precisely that such sensible qualities are the result of our perspectival encounter with a table, which comprises many other possible and actual sensible qualities united by the idea of the table as their common cause. ‘Table’ is the name of a complete object, not just the name of certain of the sensible qualities that such an object excites in the mind of someone perceiving it.²⁶

Hume constructs the idea of a table from just sensible qualities. From Shepherd’s point of view, this is a deliberately impoverished approach, which Shepherd rejects for the very reason that Hume himself embraces. Shepherd and Hume agree that one cannot construct the notion of the table as a continually existing locus of causal powers from sensible qualities alone. Hume concludes that we have no idea of such a table. Shepherd concludes that we must have representative resources over and above those afforded by the table’s sensible qualities alone.

Shepherd makes a similar series of points arguing against first Berkeley and then Reid.²⁷

Berkeley never affixed the names of objects to any thing, but the combined sensible qualities which the organs of sense helped to form; omitting the idea of their constant ability, to return upon the sense when called for, and of outwardness being equally a regular attendant upon their appearance, and a capacity in nature necessary to their existence in relation to us, and to our own in relation to them; which circumstances are included in their names. (EPEU 68)

While one might be able to offer a nominal definition of ‘table’ as applying to a mere collection of sensible qualities, the real definition of ‘table’ includes in it a conception of the table as that which causes those sensible qualities. The definition of ‘table’ applies to the object of perception, not just the contribution of the senses, and so any philosophical account that begins with the latter has already made a series of implicit, illicit theoretical moves to arrive at that starting point.

His method of incomplete definition, and naming only the combined sensible qualities the effects of things, when all men name them as united with the perceptions of the understanding, and the observations of experience, is the reason why his philosophy seems at once plausible, contradictory, and unanswerable. (EPEU 69)

Berkeley’s philosophical system substitutes nominal definitions for real ones. Shepherd calls for an argument that the contributions of the senses to that perception can be separated from it despite perception’s being our most fundamental representative mental faculty.

Shepherd aims to shift the burden of proof. Hume and Berkeley begin by limiting our legitimate cognitive faculties to sensible qualities (and certain of their behaviors) and draw radical conclusions from this starting point. Shepherd’s account of perception calls into question that beginning. If our thinking begins with perception, which is a combination of the senses and reason, then by what right do Hume and Berkeley limit the scope of our representative powers to a merely nominal aspect of that faculty?

Reid, it turns out, makes a similar mistake, as Shepherd sees it. He accepts Berkeley’s and Hume’s restrictions on our representational powers, denies their idealisms, and ends up with the world as itself consisting of sensible qualities, rather than the cause of sensible qualities in our minds.

Whilst Dr. Reid, when he asserted, that the primary qualities are conceived by clear ideas of them as they exist when unperceived, and unlike any sensation they yield, was not aware that he explained these conceptions of unperceived qualities, by other qualities which still require the senses, in order to their formation; and therefore such as could only exist in a sentient being. Thus he explained ‘hardness,’ as ‘a firm cohesion of parts;’ ‘figure’ as ‘the relation of parts to each other;’—‘visible figure,’ as ‘the relation of parts in respect to the eye;’ ‘sound’ by ‘the vibrations of the air,’ &c. &c.—as though these things, after being perceived, could be planted as they appear to the inward sense and consciousness of the soul, outwardly again, as independent modes of existence, and objects of contemplation; as though the very system he is arguing against does not suppose cohesion, parts, vibrations, figure, &c. &c. &c. to be perceptions, which are inward; because all perception is conscious, and all consciousness is inward and sentient; thus assuming as his premises the very idea which is in question; and which premises involve the difficulty his argument is raised to answer. (EPEU 70–71)

Hume and Berkeley confine our representational powers to awareness of sensible qualities (and their dynamics) and conclude that all that we can represent using those powers are further sensible qualities. As Shepherd sees it, Reid accepts the mistaken premise about sensible qualities, but recognizes that we can represent external existence nonetheless, and therefore concludes that external existence must itself consist of sensible qualities. But his assumption that we can represent external objects begs the question against Hume and Berkeley. Reid needs to reject some premise of their argument in order to undermine their conclusions. This is precisely what Shepherd takes herself to have done.²⁸ She rejects the theory of mental representation on which their idealisms are founded. Our most fundamental representational faculty is perception, which is more than merely the deliverances of the senses, and so our most fundamental representational faculty already includes in it the power to represent that-which-causes-sensible-qualities.

Competing Interests

The author has no competing interests to declare.