James, Intentionality and Analysis*
1. Introduction: Prototypes and Philosophical Analysis
Recent work in empirical psychology has produced what has been referred to as a ‘crisis’ in analytic philosophy as a number of apparent presuppositions of traditional ‘conceptual analysis’ have been undermined.  I’d like to argue today that the problems currently faced by practitioners of traditional analysis help make salient important aspects William James’ own methodology. While, it is common knowledge that James’s approach to philosophical problems was more influenced by empirical psychology than classical analysts such as Russell, there has been a tendency (certainly among his critics, but also among his friends) to treat James as making an interesting, if often unpersuasive, attempt at providing traditional analyses of our philosophically central concepts. However, the wisdom of hindsight suggests that James was actually trying to provide an analysis of a radically different sort. Furthermore, this alternative mode of analysis can plausibly be thought of as a prescient anticipation of what philosophical analysis could look like in a ‘post-analytic’ age.
The particular topic of James’ that I’d like to use as an example is his account of how our thoughts come to be about the world, an account which plays a central role in motivating his views on truth.  While James’ account of intentionality (or "knowing") has appeared to many to face obvious counterexamples, I’d like to argue here that this appearance stems from the fact that James’ approach to analyzing a phenomenon is very different from that which philosophers have found natural when providing such analyses. 
Philosophers and non-philosophers alike often understand phenomena in terms of the concepts by which they are designated, and think of these concepts as having sharp boundaries. Concepts were traditionally viewed as including all and only those objects that possessed all the properties characteristic of them. Every object either was, or was not, within the concept, and no members were any more central to it than any others. An adequate analysis of a phenomenon would thus involve providing necessary and sufficient conditions for the corresponding concept’s satisfaction. Consequently, it is typically assumed that one can criticize another’s analysis of a phenomenon by showing that the resulting conceptual criteria includes things that the term doesn’t apply to, or fails to include things to which it does. 
If one treats James as trying to provide such an analysis of intentionality (i.e., as trying to provide necessary and sufficient conditions for a thought’s being about an object), counterexamples to his account will be easy to find. However, rather than trying to provide necessary and sufficient conditions for every case of a thought’s being about an object, James focused his analysis on the core/prototypical/paradigmatic cases. James claims to be interested in the "originals and prototypes" of the connection between our ideas and the world, and that other types of intentional relations can be understood in terms of their relation to the prototypical cases (PR 99). This analysis of the core was then supplemented with additional remarks about how the less prototypical cases could be understood in terms of their relations to the paradigm.  If one analyzes a concept in this way, marginal cases do not count as ‘counterexamples’ simply because they lack properties that are prominent in the analysis of the prototypical ones. Their lack of these properties merely explains why they are marginal rather than prototypical.
His pursuing this kind of analysis doesn’t represent any lack of rigor on James’ part. Recent studies of human categorization provide strong evidence that many, if not most, of our categories can’t be understood in terms of sets of necessary and sufficient conditions.  There simply may be no set of properties that all and only the things that fall within a category possess. If this is the case, then any attempt to provide an analysis of our concepts that assumes otherwise will inevitably be unsuccessful.
The following example from Dyirbal (an aboriginal language of Australia) should make clear some of the limitations of the traditional model of concepts.  Dyirbal classifies every object in the world into one of four categories, bayi, balan, balam, and bala. The members of the four categories include the following:
1. Bayi: men, kangaroos, possums, bats, most snakes, most fishes, some birds, most insects, the moon, storms, rainbows, boomerangs, some spears, etc.
2. Balan: women, dogs, platypus, echidna, some snakes, some fishes, most birds, fireflies, scorpions, crickets, the hairy mary grub, anything connected with water or fire, sun and stars, shields, some spears, some trees, etc.
3. Balam: all edible fruit and plants that bear them, tubers, ferns, honey, cigarettes, wine, cake.
4. Bala: parts of the body, meat, bees, wend, yamsticks, some spears, most trees, grass, mud, stones, noises and language, etc.
Instead of being determined by possession of a fixed group of properties, class membership is Dyirbal has the following characteristics:
Centrality: Categories have certain central members. For Balan the central members are (human) females, water, fire, and fighting.
Chaining: Complex categories are structured by chains incorporating specific knowledge. Birds, for instance, fall into category 2 because they are believed to house the spirits of dead females. Women are also mythologically linked to the sun, which is linked to sunburn, which is linked to the hairy mary grub (which produces a painful rash that feels much like sunburn). It is by virtue of such a chain that the hairy mary grub is in the same category as women. 
No Common Properties: Categories on the whole need not be defined by common properties. This is most clearly the case with the fourth category, Bala: which seems to simply include everything not in the other classes.
Dyirbal classification is, admittedly, an unusually vivid case, but these basic phenomena of centrality, chaining, and lack of common properties are widespread and reflected in James’ account of intentionality. Perceptual cases provide the prototype, and other cases of intentionality are understood in terms of their relation to the prototype even if they lack properties that are crucial to the prototypical cases. 
2. James’s basic account 
James wanted to explain how one piece of "flat content … [with] no self-transcendency about it" came to be ‘about’ something else.  James’ method was to look at the clearest cases where we treat our thoughts as being about the world, and analyze what is going on in them.  James took the most basic cases of our thoughts being about the world are found in perception,  where (as he puts it) "[t]he external and the internal, the extended and the not extended fuse and make an indissoluble marriage" (ERE 265). 
James extends the paradigm of perceptual reference by arguing that one’s ideas can know objects outside of one’s perceptual field by leading one through a series of experiences that terminate in an actual percept of the object referred to. For instance, James’ "Memorial Hall" idea may just be a dim image in his mind, but if this image allows James to go to the Hall and recognize it, then "we may freely say that we had the terminal object ‘in mind’ from the outset, even altho at the outset nothing was there in us but a flat piece of substantive experience like any other, with no self-transcendency about it" (ERE 29). Our ideas about objects outside of our perceptual field need not share all of properties with the core cases of perception, but they bear a ‘chaining’ relation to them: they are ideas that lead one to the actual percept. Indeed, such cases are the ‘prototype’ of non-perceptual (conceptual) reference for James. 
James’ account of the intentionality of our concepts, in which our ideas are about objects because they lead us into perceptual contact with those objects, faces a number of problems if viewed as a conceptual analysis of the form: P’s idea x is about object O if and only if x leads P to come into perceptual contact with O. Among such problems are:
1. It suggests that one can’t have thoughts about objects until one has actually tracked them down.
2. It ignores the ‘social’ character of language and cognition, and suggests that I couldn’t have thoughts about things that I couldn’t recognize.
3. It would seem to make reference to unobservable entities impossible. 
However, all of these problems are manageable once we realize that James typically keeps his discussion close to the core of the concept, and that while these problem cases don’t fit the prototype, they can still be understood in terms of it.
3. Virtual Knowing and the potentially verified
For instance, an apparent problem with James’ account is that while it allows my idea of Memorial Hall to have always referred to the hall once it actually leads me to it, common sense suggests that our idea refers to the hall before this happens, or even if I never track the hall down. Indeed, a large and significant portion of my thoughts seem outside James’ initial extension of the prototype to non-perceptual cases. I can think about, say, the Eiffel Tower, but that idea may never reach the stage of ‘face-to-face’ verification.
James is aware of this, and he claims that in such cases we ‘virtually’ refer to the objects of our thoughts.
The key to this difficulty lies in the distinction between knowing as verified and completed, and the same knowing as in transit and on its way. To recur to the Memorial Hall example lately used, it is only when our idea of the hall has actually terminated in the percept that we know ‘for certain’ that from the beginning it was truly cognitive of that. Until established by the end of the process, its quality of knowing that, or indeed of knowing anything, could still be doubted; and yet the knowing really was there, as the result now shows. We were virtual knowers of the hall long before we were certified to have been its actual knowers, by the percept’s retroactive validating power. 
James goes on to claim that while "the immensely greater part of all our knowing never gets beyond this virtual stage.… to continue thinking unchallenged is, ninety-nine times out of one hundred, our practical substitute for knowing in the completed sense" (ERE 34). As long as this ‘virtual knowing’ could be cashed out whenever it needs to be, there would be no practical difference between a theory which says that we are only virtually referring in such cases and one that claims that we are actually referring in them. Since virtual knowing plays the same practical role as actualized knowing in most cases, it is justifiable to extend our concept of intentionality to those cases that remain virtual.
However, these ‘virtual’ cases are still understood in terms of their relation to the non-virtual prototype. It is precisely in virtue of their potential to become like the prototype that the ideas in question count as ‘knowing’ the objects that they do. Virtual knowing is still a type of knowing, and with it, James’ account of intentionality expands from the most central cases of verification to the broader notion of verifiability. The verification-processes are still, as James puts it, what truth and intentionality mean "essentially," but this is only to say that we couldn’t understand verifiability independently of verification. 
4. The social character of language
James is frequently criticized for missing out on the social character of language and cognition, and we can see how one might get such an impression from his account. Prototypical cases of intentionality such as perception are non-social for James. However, even if the analysis of the perceptual prototype is individualistic, the extended cases of non-perceptual reference allow more room for social contributions.
Such social contribution seems essential to our intuitive conception of what we can refer to. If left entirely to my own devices, the beliefs and recognitional capacities I have at my disposal would wildly underdetermine the referents of many of my terms. In the core cases of intentionality we are always able to, on our own, recognize what we were referring to. Consequently, if one takes all the characteristics of the core to be essential, it can seem that to refer to something, one must at least be able to locate or recognize it on one’s own. 
Nevertheless, James’ account can accommodate such cases in virtue of its stressing that knowing an object involves being led to it "through a context which the world provides."  My ability to locate, say, my computer is based largely on the fact that it is the only one sitting on my desk, rather than my knowledge of perceptual features that distinguish it from all other computers.  Our being embedded in particular contexts is thus essential to our ability to think about various objects. However, there is little reason to think that prototypical cases of identification can’t be extended by including one’s social context as well. After all, I couldn’t find Memorial Hall on my own, but given my social context, I would have no trouble locating Memorial Hall if I were placed in Cambridge. I would only need to consult a map of the city, or ask people around me until I found someone who was able to lead me to it.
I refer to what I do by many of my technical terms because, given my social context, the experts I rely on would lead me to a particular set of objects. If I were in a different social context, many of my ideas would be cognitive of different things. How a term is used in one’s social surrounding can thus affect what one’s own ideas are about.  I am treated as referring to elms by "elm" even if I cannot uniquely pick them out myself, because I can rely on others to do so for me. As long as somebody can be relied upon to know just what elms are, the system works. 
It should be clear why such cases are outside of the prototype. I can refer to Memorial Hall in virtue of my reliance on others who have been acquainted with Memorial Hall in the prototypical fashion. Indeed, it is a sign of how far such cases stray from the prototype that the claim that we are thinking about the purported objects of our thoughts become more and more tendentious as our ability to locate the referent becomes more and more dependent upon our social context.
This brings us to the case of ‘unobservables’, those items which we could not, perhaps even in principle, have perceptual contact with (quarks, another’s thoughts, etc.). Such cases obviously present at least prima facie problems for an account of intentionality that tries to explain it in terms of perceptual contact with the objects referred to.
However, if our ideas can lead us to the environment of these objects, James claims that there is a sense in which we can know them ‘virtually’ as well. Earlier cases of virtual knowing involve a thought’s being about something even though no perceptual contact is never actually made, and such cases bear an important relation to some cases of "knowing" where perceptual contact could not even possibly be made. After all, James characterizes virtual cases for observables as follows:
Just as we here assume Japan to exist without ever having been there, because it works to do so, everything we know conspiring with the belief, and nothing interfering, we assume that thing to be a clock. We use it as a clock, regulating the length of our lecture by it. The verification of the assumption here means its leading to no frustrations or contradiction. Verifiability of wheels and weights and pendulum … is as good as verification. For one truth-process completed there are a million in our lives that function in this state of nascency. They turn towards direct verification; lead us into the surroundings of the objects they envisage; and then, if everything turns out harmoniously, we are so sure that verification is possible that we omit it, and are usually justified by all that happens. (PR 99-100)
Non-actualized cases of virtual knowledge thus "lead to no frustration or contradiction" and to "the surroundings of the objects." Both of these characteristics could be shared by our thought about unobservables. Standard cases of virtual knowledge count as ‘knowing’ in virtue of having the potential to become like the prototype; reference to unobservables counts as ‘knowing’ in virtue of sharing the evidential characteristics of the virtual cases.  Once we admit the standard cases of virtual reference and allow that "to continue thinking unchallenged is, ninety-nine times out of one hundred, our practical substitute for knowing in the completed sense" (ERE 34), reference to unobservables can be admitted as well. This further extension of the prototype takes us quite far from the original, and thus it is not surprising that some have found our actually being able to refer in such cases problematic.
Both the classically conceptual and James’ prototype-centered analyses of categories can, in some sense, be understood as trying to get at what is ‘essential’ to a category/phenomenon, but each has a different conception of what being essential involves. In the former case, the essential properties must be shared by all instances of the kind. In the latter, the essential properties need not be possessed by everything which falls under the term, but the intelligibility of the category depends upon the essential cases in that the peripheral cases are understood as members of the category in virtue of their relation to the more basic ones. The essential properties explain the point of having the concept. The latter allows one to get at what is essential to a category without committing oneself to any ‘essentialism’ about it. James’ account thus itself provides a prototype of sorts for the type of analysis compatible with a more psychologically realistic picture of our concepts.
Our categorization judgments are subserved by a taxonomic scheme that generates categorization intuitions that are too variegated and diverse to be captured by simple and nondisjunctive definitions. If being an intuitive instance of X is simply a matter of having a cluster of properties that is sufficiently similar to some prototype representation, and if there are a number of different ways this can be done, some of which may vary over different contexts, then any crisp definition comprised of some subset of these properties and treating them as necessary and sufficient is never going to pas the test of intuition. It will always admit of intuitive counterexamples because the range of diversity sanctioned by our conceptual representation scheme will be much greater than that allowed by any tidy, straightforward definition…. On the prototype view, intuitive counterexamples will always be forthcoming for any definition specifying necessary features. (Ramsey p. 171.)
We are not to ask, "How is self-transcendence possible?" We are only to ask, "How comes it that common sense has assigned a number of cases in which it is assumed not only to be possible but actual? And what are the marks used by common sense to distinguish those cases from the rest?" In short, our inquiry is a chapter in descriptive psychology. (MT 14)