Fotografia y Conocimiento

SCOTTWALDEN

Photography and Knowledge

Photography has been associated with the ac-quisition of knowledge since its inception.Elizabeth Eastlake and Charles Baudelaire, twoof the earliest commentators on the medium, sawit as the business of photography to give evi-dence of facts and allowed that the new technol-ogy should be the secretary and clerk of whoeverneeds an absolute factual exactitude in his profes-sion. Later, modernist thinkers such as RudolfArnheim and Siegfried Kracauer talked of theauthenticity of the medium and praised it forbeing uniquely equipped to record and revealphysical reality. Contemporary thinkers such asPatrick Maynard and Barbara Savedoff devotecrucial chapters of their respective books to dis-cussions of the detective function or the docu-mentary authority of the medium.1

I am highly sympathetic with the idea that pho-tographs typically offer an epistemic advantagerelative to other kinds of representations. But I ad-mit to being uneasy with the metaphors that thesethinkers offer. Photographs are not, literally, sec-retaries or clerks, nor do they have authority overus. Yes, photographs are often authentic in somesense, and they can certainly record or reveal ordetect physical reality, but the same can often besaid for other kinds of representations, such aswords or nonphotographic pictures. It would begood to know in nonmetaphorical terms whereinthe epistemic advantage of the medium lies.

Surprisingly, this is a difficult task. Perhaps evenmore surprisingly, it is one that has received care-ful attention only very recently, and it is far frombeing completed.Here I critically examine two im-portant papers and a follow-up commentary coau-thored by Jonathan Cohen and Aaron Meskin,explaining in what respects I think their analy-sis succeeds and in what respects it fails.2 I then

offer my own analysis, showing how it avoids theproblems that render Cohen and Meskins analy-sis inadequate.

i. photographs as spatially agnosticinformants

Cohen and Meskins analysis relies on a notionof information according to which information iscarried when there is an objective, probabilistic,counterfactual-supporting link between two inde-pendent events.3 An event e1 carries informationabout another event e2 if and only if

(I) the probability that e2 has propertyQ is muchhigher given that e1 has property P than itwould be if e1 had had property R

and

(II) if e2 had not been Q, then (all things beingequal) e1 would not have been P.

For example, the retinas of my eyes carry infor-mation about the greenness of the leaves on thetree in front of me because (i) the probability ofthose leaves being green conditional on the rel-evant patches of my retinas being irradiated bygreen light is much higher than the probability ofthe leaves being green conditional on the relevantpatches being irradiated by nongreen light, and(ii) if the color of the leaves had been different(if, for example, I was looking at the tree in theautumn), then the relevant patches on my retinaswould have been irradiated by nongreen light.

When information construed in this way per-tains to the visible properties of objects, Cohen

c 2012 The American Society for Aesthetics

140 The Media of Photography

and Meskin call it v-information, and in themanner just sketched, our visual system is de-signed to utilize v-information to help determinethe character of the world around us. But in thecourse of its normal operation, our visual sys-tem also utilizes what Cohen and Meskin call e-information, which is information regarding thespatial location of objects relative to our bodies.4

If I am looking at a tree, my visual system not onlyprocesses v-information in ways that ultimatelyenable me to form beliefs about the color of itsleaves, but as well it processes e-information inways that ultimately enable me to form beliefsabout whether the tree is in front of me, aboveme,below me, and so on. In this way, ordinary visionexploits both v-information and e-information tohelp guide us through the world.

However, creating an informational system thatprovides e-information is frequently quite diffi-cult, as the arrangements required to maintainthe truth of the relevant probabilities and coun-terfactuals often require physical proximity, andsuch proximity can be difficult or costly to main-tain. E-informational systems are, to use Cohenand Meskins term, demanding sources of infor-mation.5 Furthermore, often we want to learnabout visible properties of objects, but we do notcare about where those objects are positioned inrelation to our bodies. The upshot of these nor-

mative pressures is that we frequently seek infor-mational systems that furnish v-information with-out at the same time attempting to furnish costlye-information. We seek what Cohen and Meskinrefer to as spatially agnostic informants.6

As it turns out, spatially agnostic informants arenot hard to come by, for pictures can, and some-times do, satisfy the two conditions for informa-tion transfer with respect to the visible propertiesof objects they depict, but not with respect to thelocation of those objects in relation to the viewer.

Consider the class of pictures Cohen and Me-skin refer to as veridical landscape paintings, anexample of which is Rackstraw Downess 110thand Broadway, 19781980 (see Figure 1).7

To keep matters simple, focus on a detail suchas the color of the lettering on the Sloans signacross the intersection (in the original, which isin color, the lettering on the Sloans sign is red).Given Downess realist aims and plein air tech-niques, it is very likely true that (i) the probabilityof those letters being red conditional on the rel-evant patches of the picture being red is muchhigher than the probability of the letters being redconditional on the relevant patches not being red,and (ii) if the color of the letters had been dif-ferent, then the relevant patches on the picturewould have been some color other than red. Thecanvas thus carries v-information about the color

Figure 1. Rackstraw Downes, 110th and Broadway (19781980). 23 39 in. Oil on canvas. Original in color. c RackstrawDownes, Courtesy Betty Cuningham Gallery, New York.

Walden Photography and Knowledge 141

of the lettering on the sign and, in all likelihood,carries v-information about many other featuresof the intersection at the time that the paintingwas made. And yet there is no similar correlationbetween features of the canvas and its location inrelation to the objects depicted. If the painting ismoved farther from or closer to the Sloans store,the color patches and shapes on the canvas willremain the same. Thus, the canvas does not offere-information to its viewers even though it doesoffer v-information; it is a spatially agnostic infor-mant.

Photographic pictures are likewise often spa-tially agnostic informants. A Polaroid snapshottaken from the same vantage point as the one oc-cupied by Downes as he created his canvas wouldexhibit color patches and shapes that satisfy (I)and (II) with regard to visible properties of thescene, but not with regard to spatial-relationalproperties holding between the snapshot and theobjects depicted in it. One could carry the snap-shot to the other side of the planet and not alterthe shapes and colors on its surface in any way.

It is thus not their status as spatially agnosticinformants that epistemically sets photographicpictures apart from veridical landscape paintings.Instead, according to Cohen and Meskin, it is thecognitive status of typical viewers of the two kindsof images that does so. More specifically, it is cer-tain beliefs that viewers bring with them to the actof viewing:

By and large, viewers believe that the type of pho-tographs is one whose members carry v-information.And by and large, viewers believe that the categoriesto which they assign veridical paintings are ones whosemembers may fail to carry v-information.

Furthermore,

subjects who come into visual contact with a photograph(under ordinary viewing conditions) typically categorizethat object as a photograph; in contrast, subjects whocome into visual contact with veridical . . .paintings donot typically categorize them as veridical . . .paintings,but rather as paintings.8

Here Cohen and Meskin distinguish betweenthree types of pictures. The first, photographs,typically supply v-information and are believedtypically to supply v-information. The second,veridical paintings (such as the Downes canvas),likewise typically supply v-information and are

believed typically to supply v-information. Thethird, paintings (such as Poussins depictions ofan imagined Arcadia), frequently fail to supply v-information and are believed frequently to fail tosupply v-information.

At the level of types,CohenandMeskins analy-sis thus does not epistemically distinguish betweenphotographs and veridical paintings. But at thelevel of tokens it does. Because token veridicalpaintings exhibit the same kinds of physical char-acteristics as token paintings, viewers frequentlyform false beliefs about them, categorizing themas paintings when they are in fact veridical paint-ings. Viewers therefore often fail to recognize thev-informational richness of token veridical paint-ings and are thus unable to exploit it. Their situ-ation is analogous to that of Othello who, whenconfronted with a rich source of (propositional)information in Desdemona, fails to recognize heras belonging to the informationally rich type towhich she in fact belongs, and so is unable to ex-ploit the epistemic virtue that she offers.

With photographs, however, matters are dif-ferent. The physical characteristics of token pho-tographs (small size, glossy surface, and so on) ren-der them easily recognizable as tokens of the typetowhich they belong, and so viewers believe tokenphotographs to be v-informationally rich, and arethus able to exploit this richness. Photographs areakin to the legendary Delphic oracle in that bothare rich sources of information and easily recog-nizable as such. According to Cohen and Meskin,it is this that accounts for the epistemic advantagethat photographs have over veridical nonphoto-graphic pictures. As they put it in a later discus-sion, [y]ou will only take as evidence [regardingfacts in which you have an interest] a [picture]that does carry information if it is recognizable toyou as an instance of a type whose instances youbelieve carry information.9

My first concernwithCohen andMeskins anal-ysis centers on its psychological plausibility. Tosee the problem, note first of all that Cohen andMeskin emphasize that the notion of informationthey invoke is utterly non-doxastic. That is, wedo not understand the provision of informationin terms of a capacity to provide true beliefs nor,in fact, do we construe information in any cogni-tive terms at all.10 It may be that the correlationscharacterized by (I) and (II) function in the ear-liest stages of visual processing in ways that ulti-mately lead to the formation of perceptual beliefs,


but such correlations are nonetheless far removedfromanyof the familiar doxastic occurrences stud-ied at the level of psychology. Information, in thesense invoked by Cohen and Meskin, is utterlynonpropositional and, hence, is not the sort ofthing that can be true or false. It is, and is only,a pattern of correlations exhibited by some statesand events and not others, a pattern characterizedby highly technical propositions expressed by (I)and (II).

Although I have no objection to such a con-strual of information, I worry that it mixes poorlywith the belief ascriptions that are essential toCohen and Meskins analysis. The viewers of pic-tures that Cohen and Meskin have in mind areordinary people looking at snapshots in photo al-bums or paintings on gallery walls, people wholack the highly technical concepts involved inCohen andMeskins notion of information. Giventhat it is a plausible constraint on correct propo-sitional attitude attribution that the individualsto whom those attitudes are ascribed possess theconcepts involved in the embedded contents, itis unlikely that viewers believe that the typeof photographs is one whose members carry v-information or that they believe that the cat-egories to which they assign veridical paintingsare ones whose members may fail to carry v-information.

In order to understand my second concern withCohen and Meskins analysis, a preliminary dis-tinction is required between two aspects of inves-tigation into why we value objects or processes ingeneral. The first, ontic, aspect involves develop-ing an understanding of the objects or processesthat persons value, but doing so apart from a con-sideration of the cognitive status of those persons.The second, cognitive, aspect involves developingan understanding of the beliefs that those personsdevelop about the ontic matters, and how thosebeliefs interact with desires to yield the value as-signed to the objects and processes.

To take a simple example, it is a fact about mostpeople that they value gold more than they doiron. Part of the story about why this is so has todo with facts about gold and iron themselves, factsthat we can study in isolation from a considerationof anyones cognitive status. Gold, we learn in de-veloping our chemistry, has an atomic structurethat renders it chemically very stable, that rendersit physically malleable, and that gives it a certainbright and reflective color. Iron, we learn, has an

atomic structure that renders it highly reactive,brittle, and that gives it a dark and unreflectivecolor. But an understanding of these ontic mat-ters alone does not yield an understanding of whyit is that gold is valued over iron. We must extendour investigation into the cognitive domain, exam-ining the beliefs that people have about gold andiron and the various desires that they have thatinteract with these beliefs. People, we learn in de-veloping our folk psychology, believe that gold isnoncorrosive, that it can be worked, and that it isa pleasing color. Furthermore, people believe thatiron rusts, that it will crack if formed in intricateways, and that it is dull in appearance. And wenote as well that people desire metals that do notcorrode, that can be worked, and that are pleasingto the eye. These cognitive facts, in conjunctionwith the ontic facts, begin to yield a complete un-derstanding of why it is that people value goldmore than they do iron.

One might reasonably expect that the sameconsiderations apply to our understanding of whypeople typically assign greater epistemic value tophotographs than they do to other types of pic-tures. If so, on the ontic side, we have to investigatetypical photographic technologies and the varioushigh-level generalizations about photographs inrelation to the objects they depict that are sub-tended by the proper functioning of these tech-nologies. And we have to investigate how thesetechnologies and supported high-level generaliza-tions differ from those associated with other typesof pictures.On the cognitive side,wehave to inves-tigate the beliefs that people typically have about,if not the technologies and formative processes as-sociated with the various kinds of pictures (sincepeople often lack these), then at least the high-level generalizations associated with each kind ofpicture. And, finally, we have to understand howthese beliefs interact with commonly held desires.

Returning to the Cohen and Meskin proposal,we have seen that they contribute substantially tothe cognitive aspect of this investigation. They as-cribe to viewers (although implausibly in my view;see above) beliefs concerning the v-informationalcarrying capacity of types of pictures and beliefsregarding to which of these types token picturesbelong. But when it comes to the required onticside, their analysis is inadequate.

To see why, consider once again the gold-ironexample, and note that characteristically thereis a connection between the beliefs ascribed on


the cognitive side of such an analysis and theoccurrences described on the ontic side. Morespecifically, people are frequently caused to havethe beliefs that they have because the world is asthose beliefs represent it to be. In the example,people are caused to have the belief that gold ismalleable by the fact that gold is malleable. Suchcausation in turn subtends explanation insofar aswe can explain the fact that people believe thatgold is malleable by pointing to the fact that goldis malleable. Extending this, we can begin to ex-plain the fact that people value gold more thanthey value iron by pointing to this ontic-groundedbelief, the ontic-grounded belief that iron is notmalleable, and thewidely helddesire formalleablemetals. Call cases involving this pattern of expla-nation standard cases.

There is, of course, a large class of cases thatdeviate from the standard cases insofar as there isno ontic basis for the beliefs in the ascribed cog-nitive structures. Consider, for example, forkedsticks that are used as divining rods. Some peo-ple value forked sticks more than they do sticksshaped in other ways because those people desireto locate underground water and possess the be-lief that forked sticks indicate the presence of un-derground water along with the belief that sticksshaped in other ways do not. Here there is noontic basis for the former beliefs for the simplereason that forked sticks are as inefficacious atindicating the presence of underground water asare sticks shaped in other ways. The former beliefsare false. Call such cases error-theoretic, as they in-volve widespread and systematic error on the partof a significant group of people.

Returning to the Cohen and Meskin proposal,their text suggests that it does not belong tothe error-theoretic type. With regard to the v-informational richness (or lack thereof) of thetypes photograph, veridical painting, and paint-ing, they allow that some of the relevant back-ground beliefs are false in some cases but addthat the relevant background beliefs are true inmany cases.11 Assuming that explanations of whypeople value items are either standard or error-theoretic, we can thus conclude that the Cohenand Meskin proposal is standard. There must,therefore, be a discoverable and explicable on-tic base grounding and explaining the beliefs thatviewers have regarding the v-informational con-tent of the types photograph, veridical painting,and painting. And it is precisely in this respect

that the Cohen and Meskin proposal is inade-quate. For they offer no account of why tokensof the type painting are v-informationally impov-erished relative to tokens of the types photographand veridical painting. The beliefs that viewers arealleged to have about such matters are thereforeleft unsupported.12 In the gold-iron example, themalleable character of gold and the brittle charac-ter of iron cause the widespread beliefs that goldis malleable and iron brittle, and such causationcan thus subtend the corresponding explanationsof the presence of those beliefs. But there is noanalogous explanation available in the Cohen andMeskin proposal.

Such an omission is surprising, and I findI can account for it only by speculating thatCohen and Meskin believe that they do have anontic basis to offer, and that it can be found inthe spatially agnostic character of photographs.As noted above, Cohen andMeskin point out thatsuch character renders photographs undemand-ing sources of v-information and that this makesphotographs epistemically valuable because(i) frequently, we are not interested in obtaininge-information, but we are interested in obtainingv-information, and (ii) information systems thatfail to carry e-information are often thereby ableto carry v-information about a wider range of sub-jects aboutwhichwedesire v-information. If this iscorrect, it is an ontic characteristic of photographsand thus one that can cause, and therefore ex-plain, widespread beliefs about an epistemic fea-ture of photographs. The problem is, however,that this is not the ontic base that Cohen andMeskin require to complete their analysis. WhatCohen and Meskin in fact require is an ontic dif-ference between the type photograph and the typepainting that renders the former type typicallyv-informationally rich relative to the latter type.Spatial agnosticism does not do this, however, forthe simple reason that, except in very unusual cir-cumstances, all types of pictures, if they v-inform,do so in spatially agnostic ways. Spatial agnosti-cism factors out of the current analysis, an on-tic common denominator among the three typesof representation under consideration and, I sus-pect, amere vestige of Cohen andMeskins earlierdisagreement with Kendall Walton regarding thetransparency of photographs.13

If I amcorrect thatCohenandMeskins analysissuffers from this ontic omission, then perhaps theycan easily patch things by offering the required


ontic component in a future discussion. How-everand perhaps even more worrisomely thanthe apparent ontic omission itselfI find in theirtexts an explicit rejection of what I take tobe the most promising characterization of therequired ontic base. I have in mind a character-ization that has appeared in the literature on pho-tography almost since the invention of the tech-nology. It has to do with pictorial etiology. Manyhave pointed out that photographic pictures, asopposed to handmade pictures such as veridicalpaintings and paintings, have a mechanistic eti-ology, one that importantly excludes the menta-tion of the photographer from the formative pro-cess.14 Very roughly, and subject to qualificationsthat will be discussed below, when a photogra-pher presses the shutter, the process that mapsfeatures of the scene onto features of the resul-tant picture bypasses her mentation so that fea-tures of the scene are rendered in the picture re-gardless of whether or not she noticed them. Incontrast, when a painter creates an image, fea-tures of the scene must register in his cognition toat least some extent in order for them to be ren-dered on the canvas.15 Likewise, it can be arguedthat themechanistic character of the photographicprocess places any desires the photographermighthave about what will be depicted in the pictureoutside of the formative process, thereby limitingher control in this regard. But this is not so in thecase of the painter, who can add features at will.In information-theoretic terms, handmade imagessuch as paintings are much more susceptible toequivocation and noise than are photographs andthus as a type are more v-informationally im-poverished than photographs as a type.16 Cohenand Meskin could use this alleged ontic featurein their analysis. However, they explicitly rejectthis mechanistic-etiological approach, citing withapproval the currently popular view that the men-tation of the picture maker is as much involved inthe formation of a photograph as it is in the for-mation of a veridical painting or a painting.17 I amthus at a loss to see how they might proceed.

ii. the objectivity-based epistemic advantage

i. Ontic Component. My own ontic story takesas its foundation the mechanistic character of thephotographic process just discussed. It is no secretthat this position has received substantial criticism

over the past few decades.18 Indeed, with a fewexceptions, most today accept as orthodoxy thatthe production of photographs is shot throughwith intentional conditions on the selection oflenses, the development process, and so forth,and that therefore no purchase can be gainedalong these lines on an ontic distinction betweenphotographic and handmade images.19

However, I have argued elsewhere that al-though critics of the traditional mechanistic viewcorrectly note that the mentation of the photogra-pher typically has a substantial effect on the char-acter of the photographs produced, they nonethe-less fail to distinguish between two importantlydistinct ways in which mentation can be involvedin the formation of pictures. Mentation is primar-ily involved in the formation of a picture whenit is a component of the sequence of events thatsubtends the mapping of features of the depictedscene onto features of the picture, whereasmenta-tion is secondarily involved when it has a bearingon the character of the resultant picture, but is notdirectly involved in the subtending process.20

To take a simple example, the photographer de-cides in which direction to point her Land camera,and her decision in this regard will have a sub-stantial effect on the character of the photographproduced. But once she has made this decisionand pressed the shutter button, her mentation isexcluded from the optical-chemical process thatsubtends the mapping of features of the sceneonto features of the resultant picture. Her deci-sion is secondarily involved in the formation ofthe picture, but not primarily involved. Contrastthis with the case of the painter, whose mentationis involved secondarily in the decision aboutwhichscene to paint, and then again primarily insofar asfeatures of that scene must register in his mindin order for them to be depicted in the resultantpicture.

Because the term objectivity is generally as-sociated with processes that importantly excludementation (a process that objectively measuresstudent performance, for example, is one that by-passes the teachers thoughts on the matter), wemay refer to picture-forming processes that lackprimary mental state involvement as objective. Itis in this sense that there is merit to the traditionalview that photography is an especially objectivemedium.

There are, of course, many objections that canbe raised at this point, several of which merit


serious consideration: the use of Land camerasis the exception and not the norm; the primary-secondary distinction is ad hoc; the use of digitalimaging techniques enables photographers easilyto involve their mentation primarily in the forma-tion of their pictures, and so on. As I have offeredreplies to such objections elsewhere, I do not doso here.21 Instead, I simply assume that the tra-ditional photographic process is objective in thisway, and I explain how such objectivity lays downthe ontic base that is tracked by beliefs of viewersof pictures.

Two further elements are required in orderadequately to characterize this ontic base. Thefirst is what I refer to as a proto-belief .22 Proto-beliefs are perceptually or cognitively inducedintentional states that are candidates for becom-ing full-fledged beliefs but that differ from full-fledged beliefs insofar as they lack the functionalroles that the latter exhibit. They are states withcontents that can be true or false, but that we donot act upon in the course of attempting to ful-fill our desires. They are what in folk psychologywe call appearances.23 In perception, the vast ma-jority of proto-beliefs quickly become perceptualbeliefs, as matters are generally as they appearto be. But there are occasions on which the con-tents of proto-beliefs are so much in conflict withthe contents of background beliefs that the for-mer are rejected. Familiar examples include thesituation in which a stick half immersed in waterappears bent, or the roadway ahead on a sunnyday appears to be covered in pools of water.

The second required element is that of a high-level regularity, an idea that emerges from reflec-tion on our interactions with technologies whosemechanisms we do not understand. Imagine Fred,who knows nothing about how automobiles workin terms of the detailed operation of the engine,transmission, and so on, but who does know howto operate an automobile in the rough-and-readysense of being able to press it into service in thecourse of his daily life. The contents of Freds be-liefs pertain not to the detailed mechanisms, butrather to high-level regularities that are subtendedby these mechanisms. Fred carries with him back-ground beliefs with contents such as turning thekey starts the car and placing the shifter in theD position enables the car tomove forward. Fredis no isolated exception. Even those who are rela-tively sophisticated in terms of their understand-ing of technologies that subtend high-level reg-ularities in one domain must rely on knowledge

of those high-level regularities in other domains.People who understand automotive technologymight be ignorant of the details of lithographic,audio, or aeronautical technology, for example,and so must lapse into an understanding based onthe regularities subtended by those technologiesin order to function in relation to them.

Bringing proto-beliefs and high-level regulari-ties together with objectivity, my central ontic as-sertions are that

O1: proto-beliefs induced in the minds of viewers ofobjectively formed pictures are frequently true

and

O2: proto-beliefs induced in the minds of viewers ofnonobjectively formed pictures are frequently false.

Furthermore, because photographic pictures aretypically formed objectively and handmade pic-tures are not, it follows from the correctness of O1and O2 that proto-beliefs induced by looking atphotographs are much more frequently true thanproto-beliefs induced by looking at handmade pic-tures.

What reasons does the skeptical reader have foracceptingO1 andO2? There are several, some ob-vious, others less so. Here are two: (a) Painters orsketchers can easily fail to notice visible featuresin the scene that they are depicting and, becausetheir mentation is primarily involved in the for-mation of their pictures, such failures will resultin the omission of features in the pictures thatotherwise would have induced true proto-beliefsin the minds of viewers. Similar failures to noticedetails on the part of photographers will not like-wise result in the omission of features on the pho-tographs, precisely because their mentation is notprimarily involved in the formation of the pho-tographs. Thus, viewers of the photographs willnot likewise fail to have true proto-beliefs inducedin their minds.

(b) Painters or sketchers, their mentation pri-marily involved in the formative process, can eas-ily add features to their pictures that have no ana-logues in the depicted scene and that thus typ-ically induce false proto-beliefs. Photographersmay wish to do the same, but have a much hardertime doing so, as their mentation is only secon-darily involved in the formative process. And thusthe pictures they produce rarely induce such falseproto-beliefs.


What about veridical paintings, such as theDownes canvas discussed above? It might be ob-jected that their existence falsifies O2, as such pic-tures by definition are ones that frequently inducetrue proto-beliefs.24 However, while it is true thata prevalence of veridical handmades would fal-sify O2, it is also true that such pictures are rarelyencountered. Photorealist canvases are not onlyvery difficult to create, they are as well not thesort of picture that most painters and sketcherswant to create, even if they could. As the truthof O2 can survive exceptions, such rarely encoun-tered examples are not a threat.

It might also be objected that O1 and O2 arenot fully ontic assertions, since the high-level reg-ularities they express hold between picture-typescharacterized in etiological terms on the one handand the truth values of contents of certain kindsof mental states on the other. O1 and O2 thusstraddle the mentalextra-mental divide and soarguably mix poorly with my earlier characteriza-tion of the ontic side of the analysis, which mightbe construed as pertaining exclusively to extra-mental occurrences. But the essence of the onticside of the analysis is not its extra-mental char-acter, although typically it does deal with extra-mental occurrences. Instead, the essence of theontic component is that it investigates those occur-rences that are represented by beliefs embeddedin the cognitive component.And, as iswell known,human beings have the capacity for beliefs abouttheir beliefs and, specifically, the capacity to eval-uate their beliefs for truth or falsity. As I hope toshow below, typical viewers of pictures have be-liefs that represent the regularities expressed byO1 and O2, thus rendering them bona fide as-pects of the ontic component, their partial mentalcharacter notwithstanding.

ii. Cognitive Component. With the ontic baselaid down, the cognitive component can easily bestated. My central cognitive assertions are that

C1: viewers believe that proto-beliefs induced in theirminds by looking at objectively formed pictures are fre-quently true

and

C2: viewers believe that proto-beliefs induced in theirminds by looking at nonobjectively formed pictures arefrequently false.

It will immediately be objected that C1 andC2 ascribe to viewers contents involving technicalconcepts such as that of a proto-belief and an obje-ctively formed picture, and that they are thereforenomore psychologically plausible thanCohen andMeskins ascriptions involving concepts such as v-information.However, the difference between theascriptions is that the concepts involved in C1 andC2 can easily be replacedby (less precise) folk psy-chological analogues. Couched in the language offolk psychology, C1 and C2 can be restated as

C1FP: people believe that the appearances that resultfrom looking at photographs are frequently true

and

C2FP: people believe that the appearances that resultfrom looking at handmade pictures are frequently false,

both of which are plausible ascriptions. Starting ata very young age, people encounter myriad pho-tographs and handmade pictures and the thingsdepicted in those pictures. Young Johnny seessnapshots of his mother in the family photo albumalongside his art-class efforts to depict her, andcompares the appearances each of these engen-ders in his mind with the appearances engenderedwhen he looks at his mother directly. He quicklyrealizes that there is no contest between the twothe appearances engendered by his art-class ef-forts are disappointingly false. Such experiencescontinue through life, as he repeatedly encoun-ters instances of both types of picture and learnson the basis of this that the appearances engen-dered by looking at photographs are much moreoften true than those engendered by looking athandmade pictures. He comes, in other words, tohave the beliefs ascribed by C1FP and C2FP.25

It might nonetheless be objected that even thisinvolves attributing too sophisticated a psychol-ogy to viewers of pictures. Do ordinary folk re-ally think about the truth values of their thoughts,looking for correlations between those truth val-ues and the etiologies of those thoughts? The casecan be made that the answer to such a rhetoricallyasked question is yes. Placing pictures to one side,it is a truism that in our daily lives we encounterboth people who frequently utter true sentencesand those who (whether through deliberate at-tempts to deceive or through careless assertionsregarding matters about which they are ignorant)


frequently utter false ones. In the face of this, wemake it a practice to search for correlations be-tween the (eventually revealed) truth values ofbeliefs that we have formed on the basis of suchutterances on the one hand and the identities of in-dividuals whomade the utterances on the other. Ifcertain individuals are thereby revealed to be un-reliable, we discount their assertions in the future.I conjecture that our practice with regard to pho-tographic and handmade pictures piggybacks onthis perhaps innately endowed capacity, and thatit should be no more surprising that we engagein such complex inferences in the case of picturesthan it is in the case of learning about the worldvia assertions made by other persons.

Returning to the more precise formulations,and assuming C1 and C2 are correct ascriptions, itremains to be determined how they interact withother intentional states so as to generate the epis-temic value typically associated with photographsbut not handmade pictures. I suggest that when aviewer encounters a photograph, she quickly andon the basis of the characteristic features of pho-tographs discussed above forms the belief that

F: this picture is objectively formed,

and that F, in conjunction with C1, leads her to in-fer (typically unconsciously) that the proto-beliefsengendered by looking at the picture are likelytrue, whereupon she elevates them to the status offull-fledged beliefs. And when a viewer encoun-ters a handmade picture, he forms the belief that

NF: this picture is nonobjectively formed,

and NF, in conjunction with C2, blocks the proto-belief from being elevated to the status of a belief.

Note that this inferential pattern in the case ofan encounter with a photograph constitutes war-rant for the beliefs formed about visible featuresof the scene depicted in the photograph. The factthatO1 obtainsmeans that viewers have strong in-ductive support for C1. And, because objectivelyformed pictures are typically identifiable as such,viewers also have confidence in the truth of F.26

Given that the inference on the basis of C1 and Fis demonstrative, viewers are not only compelledto form true beliefs about visible features of thescene depicted, but they are compelled as well tobelieve that those beliefs are true. They have war-ranted, true beliefs; they have knowledge.

And, finally, as has been recognized at leastsince Plato, we have desires, not only for true be-liefs, but as well for reasons to believe that ourtrue beliefs are true. These desires, working inconjunction with the meta-beliefs regarding thetruth of the first-order perceptual beliefs, yield atleast someof the epistemic valuewe associatewithphotographs.

iii. photographic epistemic advantage and thelarger picture

Barbara Savedoff argues convincingly thatMargaret Bourke-Whites famous Bread LineDuring the Louisville Flood, Kentucky (1937) suc-ceeds as a photograph inpart becauseof the ironyof circumstance that it conveys, an irony involv-ing the juxtaposition of rural black poverty andsuburban white affluence.27 Savedoff points outthat establishing this irony depends on the docu-mentary authority of the photographic mediuminsofar as it establishes that the breadline was in-deed in proximity of the billboard. Had the im-age had some other sort of etiologyhad it beena digital construction, a pencil-on-paper sketch,or some similar type of image that lacked such au-thorityit would run the risk of generating merelyan irony of idea, a less-stinging irony arisingfrom the imagination of the image maker, ratherthan from the actual circumstances of 1937. Butwhat exactly is documentary authority? We nowhave an answer. Bourke-Whites image is an ar-rangement of lines and tones on a surface thatinduces in the minds of typical viewers a proto-belief with the content that there was a breadlinein proximity of a billboard. But similar arrange-ments of lines and tones with different etiologieswould likewise induce a proto-belief with this con-tent: a digital construction or a pencil-on-papersketch would do so as well. In the case of thephotograph-induced proto-belief, however, view-ers who believe that they are looking at a photo-graph have reason to believe that this proto-beliefis true. This reason comes in the form of an infer-ence, with C1FP in the role of the major premiseand the belief that a photograph is being viewedin the role of the minor. And to say that they havereason to believe that their proto-belief is true isto say that it is likely that the billboard was infact in proximity of the breadline, an arrangementthat subtends the irony of circumstance. By wayof contrast, the proto-belief induced by a sketch


is typically accompanied by the belief that the im-age is a sketch, with the result that C2FP is takenas the major premise and the inference drawn isthat the proto-belief is likely false, an arrange-ment that subtends merely an irony of idea. Inthis way, we have a fuller understanding of whythis photographand, I suspect, many other pho-tographs from the modernist canonsucceed asphotographs.

But much remains to be investigated. If digi-tal imaging techniques make it easy to underminethe objectivity-based epistemic advantage, willthe difference between photographic and hand-made images dissipate, or will institutional factorslimit the extent to which this takes place? If pho-tographs have an advantage in terms of yieldingknowledge of visible features, does this advan-tage extend to knowledge of nonvisible featuressuch as the ethical knowledge investigated bySusan Sontag?28 If the epistemic advantage canhelp us better to understand the success of pho-tographs from themodernist canon,will it likewiseenable us better to understand the success of pho-tographs from the postmodernist canon? Investi-gations along these lines, and others, are reservedfor another occasion.29

SCOTTWALDEN

Department of PhilosophyNassau Community CollegeGarden City, New York 11530

internet: [email protected]

1. Elizabeth Eastlake, Photography, in The LondonQuarterly Review, April 1857, reprinted in Vicki Goldberg,Photography in Print (University of New Mexico Press,1988), pp. 8899, at p. 97; Charles Baudelaire, Salon of1859, trans. Jonathan Mayne, reprinted in Photography inPrint, pp. 123126, at p. 125; RudolfArnheim, The TwoAu-thenticities of the PhotographicMedia,The Journal of Aes-thetics and Art Criticism 51 (1993): 537540; Siegfried Kra-cauer, Theory of Film (Princeton University Press, 1997),p. 28; Patrick Maynard, The Engine of Visualization (Cor-nell University Press, 1997), chap. 5, Seeing Machines,pp. 117148; Barbara Savedoff, Transforming Images: HowPhotography Complicates the Picture (Cornell UniversityPress, 2000), chap. 5, Transforming Media: Painting, Pho-tography, and Digital Imagery, pp. 185210.

2. Jonathan Cohen and Aaron Meskin, On the Epis-temic Value of Photographs, The Journal of Aestheticsand Art Criticism 62 (2004): 197210; Photographs as Evi-dence, in Scott Walden, ed., Photography and Philosophy:Essays on the Pencil of Nature (Oxford: Wiley-Blackwell,

2008), pp. 7090; and Photography and Its Epistemic Val-ues: Reply to Cavedon-Taylor, The Journal of Aestheticsand Art Criticism 67 (2009): 235237.

3. Photographs as Evidence, p. 72.4. The terms are introduced in Photographs as Evi-

dence, p. 74.5. The distinction between demanding and undemand-

ing sources of information is firstmade in On theEpistemicValue of Photographs, p. 204.

6. On the Epistemic Value of Photographs, p. 204.7. Veridical paintings are first mentioned in On the

Epistemic Value of Photographs, p. 205. Veridical land-scape paintings are first mentioned in Photographs as Evi-dence, p. 76.

8. Photographs as Evidence, p. 76.9. Photography and Its Epistemic Values: Reply to

Cavedon-Taylor, p. 236.10. Photographs as Evidence, p. 82.11. Photographs as Evidence, p. 77.12. Dan Cavedon-Taylor senses this omis-

sion in his important critique of the Cohen andMeskin proposal. He notes that [t]he existence ofobjective, viewer-independent differences [that is, what Iam calling ontic differences] between the two media [thatis, photographs and paintings] could arguably explain whyviewers have the background beliefs about photographsthat they do (The Epistemic Status of Photographs andPaintings: A Response to Cohen and Meskin, The Journalof Aesthetics and Art Criticism 67 [2009]: 230235 at p. 234).Absent such objective, viewer-independent differences,we are left with no explanation of why viewers have thebeliefs that Cohen and Meskin ascribe.

13. Cohen and Meskins criticism of Kendall Waltonsclaim that we see through photographs can be found in Onthe Epistemic Value of Photographs, especially Sections Iand II. Waltons original statement of his position occurs inhis Transparent Pictures: On the Nature of PhotographicRealism, reprinted inPhotography and Philosophy: Essayson the Pencil of Nature, pp. 1449.

14. Eastlake refers to the camera as an unreasoningmachine (Photography, p. 97). Stanley Cavell describesthe automatism of photography as removing the humanagent from the task of reproduction (The World Viewed[Harvard University Press, 1971], p. 23). The clearest de-scription of the mechanical character of photography isfound in Waltons Transparent Pictures: On the Natureof Photographic Realism, Section 5.

15. Dominic Lopes has argued that painters and sketch-ers need not, and typically do not, form beliefs about vis-ible features of the scene before them in the course ofcreating their pictures. But he does allow that the scenemust register in their minds to some degree. See his Under-standing Pictures (OxfordUniversity Press, 1996), especiallypp. 183184.

16. For a discussion of the relevant construals of equiv-ocation and noise, see Fred I. Dretske, Knowledge and theFlow of Information (MIT Press, 1981), chap. 1.

17. Cohen and Meskin reject the mechanistic (or auto-matic) view in On the Epistemic Value of Photographs,p. 209, and again more recently in Photography and ItsEpistemic Values: Reply to Cavedon-Taylor, p. 237.

18. See especially the highly influential criticism of themechanical model of photography offered by Joel Snyder


and Neil Walsh Allen in their Photography, Vision, andRepresentation, Critical Inquiry 2 (1975): 143169.

19. Cohen and Meskin, Photography and Its EpistemicValues: Reply to Cavedon-Taylor, p. 237. Themost notableexample of an exception is Walton. See Transparent Pic-tures: On the Nature of Photographic Realism, Section 5.

20. See Section IV of my Objectivity in Photography,The British Journal of Aesthetics 45 (2005): 258272.

21. See Walden, Objectivity in Photography, espe-cially Section IV, and my Truth in Photography, Pho-tography and Philosophy: Essays on the Pencil of Nature,pp. 108110.

22. Walden, Truth in Photography, pp. 9697. Proto-beliefs are perhaps equivalent to the outputs of informa-tionally encapsulated modules as discussed by Jerry Fodor,The Modularity of Mind (MIT Press, 1983). An intriguingline of investigation would explore the intersection betweenthese and the class of propositional attitudes that KendallWalton and Patrick Maynard refer to as states of imagin-ing. See Kendall Walton,Mimesis as Make-Believe: On theFoundations of the Representational Arts (Harvard Univer-sity Press, 1993) andMaynardsThe Engine of Visualization.Tamar Gendlers notion of alief is also relevant, althoughproto-beliefs have more conventional, propositional con-tents. See her Alief and Belief, The Journal of Philosophy,105 (2008): 634663, especially p. 635 n4.

23. While in perception there is almost always a phe-nomenology accompanying proto-beliefs, it is the under-lying propositional core of the proto-beliefs to which thisanalysis appeals, a core that will typically be much less in-trospectively salient to the person who has the proto-beliefthan is its phenomenal character. My claim is thus not thatphenomenological appearances somehow match the ob-served reality, whatever thismightmean. Thus, the criticismsof the visual model offered Snyder and Allen in Photog-raphy, Vision, and Representation, pp. 143169, would notapply.

24. Perhaps here I should register my discomfort withthe term veridical painting. Veridicality is associated withtruth and truth with propositionspropositions that somemight assume are expressed by the pictures themselves, in

the same way that propositions are expressed by declar-ative sentences. But I have no better term to offer and, aslong as veridical paintings (or, better, veridical handmades)are understood to be by definition nonphotographic imagesthat tend to induce true proto-beliefs, no confusion shouldresult.

25. This characterization of the cognitive side of the anal-ysis replaces my earlier attempt to do so in terms analo-gous to Grices characterization of the cognition associatedwith implicature (see PaulGrice, Logic andConversation,reprinted in his Studies in the Ways of Words [HarvardUniversity Press, 1989], pp. 2240, and my Objectivityin Photography, pp. 271272). In that earlier work, I as-sumed that viewers have beliefs with contents pertainingdirectly to the objectivity of the photographic process. Ipapered over the psychological implausibility of this bysuggesting that viewers did indeed have such beliefs butthat they did not realize that they had such beliefs in thesame way that according to Grice typical speakers have be-liefs pertaining to conversational maxims but do not realizethat they do. I was never entirely satisfied with this ap-proach, however, as the concepts involved in beliefs aboutconversational maxims are much less complex than con-cepts having to do with objectivity. It was Jody Azzouniwho in conversation pointed out to me that the contentsof the ascribed beliefs could pertain, not to the objec-tive character of the process that subtends high-level reg-ularities (or, as he calls them, gross regularities; see hisKnowledge and Reference in Empirical Science [London:Routledge, 2000]), but rather to the high-level regularitiesthemselves.

26. Although the advent of digital imaging techniquesplaces this in jeopardy. For more on this concern, see myTruth in Photography, pp. 108110.

27. Barbara Savedoff, Transforming Images, pp. 202209.

28. Susan Sontag, In Platos Cave, in her On Photog-raphy (New York: Picador, 1973), pp. 324.

29. I am indebted to Jody Azzouni for his extremelyhelpful suggestions and to an anonymous referee for thisjournal.

Documents

Fotografia y Conocimiento