I. The Theater of Representation*
One of the important developments in science studies has been the increased awareness that scientific inquiries are social and ideological constructions. Donna Haraway's explorations of primatology, Shapin and Schaffer's investigations into the sociology of Boyle's laboratory, and Bruno Latour's study of "black boxes" in science are only a few of the seminal analyses that have challenged accounts of how science is done (1). So extensive and successful have these critiques been that it now seems the aspect of science most in need of explanation is its power to arrive at apparently ahistorical and transcultural generalizations. Given that science is socially constructed, how can we explain, as Michel Serres puts it, that "entropy increases in a closed system, regardless of the latitude and whatever the ruling class." (2)

A clue can be found in a curious lacuna that occurs when this question is discussed within the philosophy of science. There the debate has been constructed as a division between the realists and the anti-realists. Both sides grant that there is something called observables, and that these observables have an instrumental efficacy in the world. You tighten a loose battery cable, and the car starts where it would not before. The difference comes in whether or not the observables relate to entities that exist in reality as such. The realists say there really is an electrical current that flows, while the anti-realists want to weaken or deny this claim. The lacuna occurs in the anthropomorphic grounding that underlies the idea of observables. Without being explicit about it, both sides mean observable from a human perspective. This assumption has important implications.

Consider a frog's visual cortex. Studies indicate that objects at rest elict little or no neural response in a frog's brain. (3) Maximum response is elicted by small objects in rapid, erratic motion--say, a fly buzzing by. Large objects evoke a qualitatively different response than small ones. This arrangement makes sense from a frog's perspective, because it allows the frog to identity prey from non-prey, and prey from predators that want to eat it. Now imagine that a frog is presented with Newton's laws of motion. The first law, you recall, says that an object at rest remains so unless acted upon by a force. Encoded into the formulation is theassumption that the object stays the same; the new element is the force. This presupposition, so obvious from a human point of view, would be almost unthinkable from a frog's perspective, since for the frog moving objects are processed in an entirely different way than stationary ones. Newton's first law further states, as a corollary, that an object moving in a straight line continues to move so unless compelled to change by forces acting upon it. The proposition would certainly not follow as a corollary for the frog, for variation of motion rather than continuation counts in his perceptual scheme. Moreover, it ignores the size of the object, which from a frog's point of view is crucial to how information about movement is processed.

My point is not that humans know what frogs cannot fathom. The scientists who did the frog research put it well: their work "shows that the [frog's] eye speaks to the brain in a language already highly organized and interpreted instead of transmitting some more or less accurate copy of the distribution of light upon the receptors" (Lettvin et al., 1950). This and other studies conclusively demonstrate that there can be no perception without a perceiver.(4) Our so-called observables are permeated at every level by assumptions located specifically in how humans process information from their environments.

Observing with instruments rather than unaided human perception does not rescue us from our anthropomorphism, for the instruments we design and build are just those that would be conceptualized by someone with our sensory equipment. Instruments extend and refine human perceptions, but they do not escape the assumptions encoded within the human sensorium. Add the profound influence of acculturation upon cognitive processing, and it becomes clear that observables really mean observations made by humans located at specific times and places and living in specific cultures. In short, we are always already within the theater of representation. Everything we perceive, think, or do is always already a representation, not reality as such.

Yet representation may be too passive a concept to account for the complexities involved. Research by Walter Freeman and Christine Skarda on the olfactory bulb of rabbits indicates that perceptual processing is context-dependent as well as species-specific.(5) Rabbits continually sniff; these sniffs take in molecules of odorants that fall on the cilia of receptor cells in the nose, which in turn are connected to mitral cells in the olfactory bulb of the cortex. When the odors are neutral, oscillatory bursts of neural activity appear that can be reliably identified as characteristic of a given animal. When the animal sniffs an odor that he has been conditioned to recognize as significant, a different pattern appears. Then the burst is amplified in a cascading effect that brings together selectively co-activated neurons in a nerve cell assembly. This amplification happens very fast, within milliseconds. At certain critical thresholds, further changes take place that affect the entire global area of the olfactory bulb. The data demonstrate that perception is not a passive response to stimuli but an active process of self-organization that depends on prior learning and specific contexts. "Perception begins within the organism with internally generated neural activity," Skarda writes (p. 52). "What happens within the brain is about interaction" (p. 53). Although the data vary with individual animals and between species, additional experiments on the visual cortex of the monkey and the somatosensory cortex of a human subject indicate that the active, self-organizing nature of perception applies in these cases as well.(6) On this basis, Skarda and Freeman have argued that neuroscience should give up the concept of representation (which Skarda calls "representationalism"), because it encourages the fallacy that perception passively mirrors the external world. Representation in this sense happens only when an observer enters the scene. It is the experimenter's viewpoint, Skarda writes, which "requires that conclusions be drawn about what the observed activity patterns represent to the subject" (p. 57). From this vantage,our anthropomorphism has not only led us to universalize our species-specific perspective into a vision of an autonomously existing reality but also to falsify the nature of our own perceptual processing.

The point is telling. I am not willing, however, to relinquish a term as central to literary discourse as representation. I want to introduce another way of formulating it that will make representation a dynamic process rather than a static mirroring. Suppose we think about the reality "out there" as an unmediated flux. The term emphasizes that it does not exist in any of the usual conceptual terms we might construct (such as reality, the universe, the world, etc.) until it is processed by an observer. It interacts with and comes into consciousness through self-organizing, transformative processes that include sensory and cognitive components. These processes I will call the cusp. On one side of the cusp is the flux, inherently unknowable and unreachable by any sentient being. On the other side are the constructed concepts that for us comprise the world. Thinking only about the outside of the cusp leads to the impression that we can access reality directly and formulate its workings through abstract laws that are universally true. Thinking only about the inside leads to solipsism and radical subjectivism. The hardest thing in the world is to ride the cusp, to keep in the foreground of consciousness both the active transformations through which we experience the world and the flux that interacts with and helps to shape those transformations. For as soon as the thought forms, we become aware of the paradox: what we imagine is not the cusp itself, but the representation of it that is in our conceptual realm.

The reflexive mirroring that enfolds cusp into concept shows how we can be trapped within the prison house of language. This inherent reflexivity was part of what Derrida had in mind when he famously proclaimed "There is no outside to the text."(7) As long as positive assertions are made, there is indeed no way out of the reflexive loop, no way to conceptualize the cusp without always already falling short of what the conceptualization attempts to represent. Negation, however, is a more complex and ambiguous function. In negation, possibilities for articulation exist that can elude the reflexive mirroring that would encapsulate us within textuality and nothing but textuality. This elusive negativity authorizes a position that grants the full weight of the constructivist argument but draws back from saying anything goes.

Such a position is necessary if science is to retain its distinctive characteristic as an inquiry into the nature of the physical world, while also rightfully being recognized as an arena of social discourse and cultural practice. Central to it are contexts, consistency, and constraints. Their interaction allows the cusp to be posited and its relation to elusive negativity explored.

II. Riding the Cusp: What We Remember, What We Forget
This afternoon Hunter and I went for a walk. Hunter is a handsome, medium-sized dog, half beagle and half hound. Hunting rabbits is bred into his genes, and there are a lot of rabbits where we live. It is not uncommon for a rabbit to run across the road in front of us. He sees it, I see that he sees it, he sees that I see he sees it. Having lived with Hunter for over ten years, I know that I have about two seconds to convince him to remain at heel rather than run after the rabbit. I also know that the outcome will depend in part on how authoritative my voice is, how close the rabbit, how intense the scent and how bad his arthritis. Most of the time I succeed in convincing him not to run; occasionally I fail. In either case, complex communications take place between us about an external reality that we both perceive and that affects our actions. How does this happen?

No doubt Hunter processes the world in a very different way than I do, from the limited color range he experiences to the vastly richer role scent plays in his universe. Despite these differences, we are able to communicate because we share a context that remains largely consistent from day to day. I do not perceive the world as he does, but my perception of his perception stays relatively constant. I know the kinds of things that excite his attention and what his probable responses will be, just as he knows mine. When the rabbit runs across our path, we each react within our different sensory realms to a stimulus that catalyzes our responses, which are also conditioned by past experiences with the world and each other. This consistency allows for the shorthand "Hunter sees the rabbit," although on reflection I am aware that "rabbit" is an anthropomorphic concept that Hunter does not share with me in anything like the same sense another human being could. The unmediated flux impinges on him, impinges on me; I see the rabbit and Hunter's response in my way, he sees the rabbit and my response in his. We both know that we are responding to an event we hold in common, as well as to a context that includes memories of similar events we have shared.

The temptation to forget the complexities of this account and abstract to the shorthand is very strong. From such abstraction comes the belief that nature operates according to laws that are universally and impartially true. What is the harm in moving tothe abstraction? The implications become clear when we look at what it leaves out of account. Gone from view are the species-specific position and processing of the observer; the context that conditions observation, even before conscious thought forms; and the dynamic, interactive nature of the encounter. In such a pared-down account, it is easy to believe that reality is static and directly accessible, chance and unpredictability are aberrations, and interaction is nothing more than an additive combination of individual factors, each of which can be articulated and analyzed separate from the others.

This is, of course, the world of classical physics. It continues to have a vigorous existence in popular culture as well as in the presuppositions of many practicing scientists. When the TV camera, accompanied by Carl Sagen's voice-over, zooms through the galaxy to explore the latest advances in cosmology, these presuppositions are visually and verbally encoded into an implied viewpoint that seems to be unfettered by limitations of context and free from any particular mode of sensory processing. As a representation, this simulacrum figures representation itself as an inert mirroring of a timeless, objective reality.

Perhaps its most pernicious aspect is the implicit denial of itself as a representation. The denial is all the more troubling because of the ideological implications encoded within it. Among those who have explored these implications are Evelyn Fox Keller, who points out the relation between an "objective" attitude, the masculine orientation of science, and the construction of the world as an object for domination and control; Ilya Prigogine and Isabelle Stenger, who relate the appeal of a timeless realm to a fear of emotional involvement and death; Nancy Cartwright, who demonstrates that the idea of scientific "laws" always derives from the act of analysis and never intrinsically from the situation itself; and Michel Serres, who reminds us that deviations from idealized, abstract forms are not exceptions but the noise that constitutes the world.(8) These critiques can be seen as acts of recovery, attempts to excavate from an abstracted shorthand the complexities that unite subject and object in a dynamic, interactive, on-going process of perception and social construction.

A model of representation that declines the leap to abstraction figures itself as species-specific, culturally determined, and context-dependent. Emphasizing instrumental efficacy rather than precision, it assumes local interactions rather than positive correspondences that hold universally. It engages in a rhetoric of "good enough," indexing its conclusions to the context in which implied judgments about adequacy are made. Yet it also recognizes that within the domains specified by these parameters, enough consistencies obtain in the processing and in the flux to make recognition reliable and relatively stable.

Since the claim for consistency separates this position from strict social construction, it is worth exploring more fully. Central to this claim is the idea of constraints. By ruling out some possibilities--by negating articulations--constraints enable scientific inquiry to tell us something about reality and not only about ourselves. Consider how conceptions of gravity have changed over the last three hundred years. In the Newtonian paradigm, gravity is conceived very differently than in the general theory of relativity. For Newton, gravity resulted from the mutual attraction between masses; for Einstein, from the curvature of space. One might imagine still other kinds of explanations, for example a Native American belief that objects fall to earth because the spirit of Mother Earth calls out to kindred spirits in other bodies. No matter how gravity is conceived, no viable model could predict that when someone steps off a cliff on earth, she will remain spontaneously suspended in mid-air. This possibility is ruled out by the nature of physical reality. Although the constraints that lead to this result are interpreted differently in different paradigms, they operate universally to eliminate certain configurations from the range of possible answers. Gravity, like any other concept, is always and inevitably a representation. Yet within the representations we construct, some are ruled out by constraints, others are not.

The power of constraints to enable these distinctions depends upon a certain invariability in their operation. For example, the present limit on silicon technology is a function of how fast electrons move through the semiconductor. One could argue that "electron" is a social construction, as are "semiconductor" and "silicon." Nevertheless, there is an unavoidable limit inherent in this constraint, and it will manifest itself in whatever representation is used, provided it is relevant to the representational construct. Suppose that the first atomic theories had developed using the concept of waves rather than particles. Then we would probably talk not about electrons and semiconductors, but indices of resistance and patterns of refraction. There would still be a limit, however, on how fast messages could be conveyed using silicon materials. If both sets of representations were available, one could demonstrate that the limit expressed through one representation is isomorphic with the limit expressed in the other.

Note that I am not saying constraints tell us what reality is. This they cannot do. But they can tell us which representations are consistent with reality, and which are not. By enabling this distinction, constraints play an extremely significant role in scientific research, especially when the representations presented for disconfirmation are constrained so strongly that only one is possible. The art of scientific experimentation consists largely of arranging situations so the relevant constraints operate in this fashion. No doubt there are always other representations, unknown and perhaps for us unimaginable, that are also consistent with reality. The representations we present for falsification are limited by what we can imagine, which is to say, by the prevailing modes of representation within our culture, history, and species. But within this range, constraints can operate to select some as consistent with reality, others as not. We cannot see reality in its positivity. We can only feel it through isomorphic constraints operating upon competing local representations.

The term I propose for the position I have been urging is constrained constructivism. The positive identities of our concepts derive from representation, which gives them form and content. Constraints delineate ranges of possibility within which representations are viable. Constrained constructivism points to the interplay between representation and constraints. Neither cut free from reality not existing independent of human perception, the world as constrained constructivism sees it is the result of active and complex engagements between reality and human beings. Constrained constructivism invites--indeed cries out for--cultural readings of science, since the representations presented for disconfirmation have everything to do with prevailing cultural and disciplinary assumptions. At the same time, not all representations will be viable. It is possible to distinguish between them on the basis of what is really there.

Are constraints not themselves representations? If so, how is the claim for their invariability justified? With thesequestions, the distance between articulation and cusp threatens to collapse, cutting off the connections that interactively put us in touch with the unmediated flux. To answer them and elaborate the dynamic figure of representation, I return to the crucial difference between congruence and consistency. Congruence implies one-to-one correspondence. In Euclidean geometry, one can test for congruence by putting one triangle on top of another and seeing whether they match. If the area and shape of one exactly fits the other, congruence is achieved; any deviation indicates that they are not congruent. Congruency thus falls within the binary logic of true/false. Consistency, by contrast, cannot adequately be accounted for in a two-valued logic. In addition to true and false, two other positions--let us call them not-true and not-false--are necessary. The introduction of these two values reveals an important asymmetry between affirmation and negation. From this asymmetry emerges a sense of the relation between language and representation that steps outside the reductive dichotomies of the realist/anti-realist debate.

III. The Semiotic Square and Elusive Negativity
Mapping the four positions mentioned above onto a semiotic square will make explicit the multiple connections and disjunctions that constitute their interactions. A. J. Greimas introduced the semiotic square as a way to represent the possibilities for signification in any semiotic system.(9) These possibilities, although very rich, are not infinite. They are created through the interaction of what Greimas called "semiotic constraints"--deep structures that enable meaning to emerge by restricting articulations to certain axes of signification. Ronald Schleifer has interpreted and expanded on Greimas's construction of the semiotic square, and the discussion that follows is indebted to his work as well as to Greimas. (10 )

If we grant that we are always already within the theater of representation, it follows that no unambiguous or necessary connection can be forged between reality and our representations. Whatever the unmediated flux is, it remains unknowable by the finite subject. Representations arise in response to such historically specific factors as prevailing disciplinary paradigms and cultural assumptions, as well as such species-specific factors as the human sensorium and neurophysiology. Observations are culturally conditioned and anthropomorphically determined. We can never know how our representations coincide with the flux, for we can never achieve a standpoint outside them. Consequently, the true position cannot be occupied because we cannot verify congruence.

The false position, however, can be occupied. Within therange of representations available at a given time we can ask,"Is this representation consistent with the aspects of reality under interrogation?" If the answer is affirmative, we still know only our representations, not the flux itself. But if it is negative, we know that the representation does not adequately account for our interaction with the flux in a way that is meaningful to us in that context. The asymmetry revealed by this analysis should not be confused with Popper's doctrine of falsification. (11) Understanding that theories could not be verified, Popper nevertheless maintained congruence as a conceptual possibility. The problem for him was that congruence was empirically based and so always liable to exceptions that might appear in the future. In the scheme articulated here, future exceptions do not play a privileged role in explaining why congruence cannot be achieved. Even if by some fiat we could be sure that no future exceptions would exist, the most we could say is that a model is consistent with reality as it is experienced by someone with our sensory equipment and previous contextual experience. Congruence cannot be achieved because it implies perception without a perceiver.

The four positions are mapped onto a modified semiotic square as shown below.

exclusion
(inconsistent) False --------------- True (unoccupied)

overlap
(unknown) Not-True ---------------- Not-False (consistent)

The horizontal relation between the two top positions, false and true, is constructed through a contrary relation that makes them mutually exclusive alternatives. What is true cannot be false, and what is false cannot be true. The bottom two positions, not-true and not-false, are in a more complex relation. Not-false, designated as the more restrictive, is occupied by models found to be consistent with the flux as it is interactively experienced. Not-true is occupied by models which have been imperfectly tested or not tested at all; these I call unknown. Between the negated categories of not-false and not-true, two kinds of oppositions are in play. One is a polarity between negation and affirmation (false/true), the other between indefinite and definite (unknown/consistent). This ambiguity folds together the ability to negate with the ability to specify. In doing so, it opens an escape hatch from the prison house of language.

The entanglement of negation with specificity can be explored through the linguistic concepts of modality and marking. Traditionally defined, a modality is a statement containing a predicate that is affirmed or denied by other qualifications. The modern definition expands a modality into any statement about another statement. Non-modal articulations appear as mere statements of fact. In this sense they are unmarked, allowing for a reading that does not take the speaker's position into account. In general unmarked terms arethose which have been naturalized by cultural assumptions and so rendered transparent. "Man" is an unmarked noun, "woman" a marked one; "as old as" is an unmarked phrase, "as young as" a marked one. In modality the marking is accomplished by the qualifying phrase that calls attention to the statement's swerve from facticity. Affirmation and negation are non-modal; denial and assertion are modal. When the President's press secretary says, "The rumor is false [or true]," he has negated [or affirmed] it. When he says "I say that the rumor is false [or true]," he has denied [or asserted] it. Denial implies negation while subtly differing from it, just as assertion implies affirmation without exactly being affirmation.

As their compound form signals, not-true and not-false are markedterms. Realism tends to elide the differences indicated by these markings, assimilating not-false into true and not-true into false.When a scientific textbook states "All the matter in the universe was once contracted to a very small area," the difference between the model and the reality tends to disappear, as do the position and processing of the observer for whom the statement makes sense. Far from eliding markings, the semiotic square displays them along the vertical axis. Expanding the binary dichotomy of realism to the quadrangle of semiotics, this distance-as-difference reminds us that articulations emerge from particular people speaking at specific times and places, with all of the species-specific processing and culturally-conditioned expectations that implies. The vertical axis thus separates as well as implicates, as shown in the schematic below.

exclusion
(inconsistent) False --------------- True (unoccupied)

implication/separation overlap

(unknown) Not-true -------------- Not-false (consistent)

Beyond the marking that not-true and not-false share is the additional negativity inhering in not-true. Located at the lower left corner of the square, it occupies the space that on a Cartesian grid represents the negative of both axes. The negative of a negative, it is the position most resistant to assimilation into the transparencies of non-modal statements. Fredric Jameson calls it "the place of novelty and of paradoxical emergence," noting that it is "the most critical position and the one that remains open or empty for the longest time." (12)

The implications of its excess negativity can be unpacked by again referring to modality. It is possible to negate a modality, creating as it were a double marking. The press secretary may say "I cannot say that the rumor is false [or true]," in which case the status of the rumor remains indeterminate. This situation corresponds to a residue within the not-true position that cannot be articulated--models that we cannot conceive because they are alien to our mode of processing the world. Not coincidentally, it also points to the reason why we cannot say a model is congruent with reality. Because we can never achieve a viewpoint outside our viewpoint, "unknown" overlaps with and implies "unknowable."

Schleifer has argued that this kind of ambiguous negation is characteristic of scientific theories and art forms that elude either/or categorization, particularly quantum mechanics and literary modernism. (13) Shoshana Felman has called it "radical negativity," which "belongs neither to negation, nor to opposition, nor to correction . . . --it belongs precisely to scandal." (14) Calling this scandal the "outside of the alternative" because it emerges from a "negativity that is neither negative or positive" (p. 141-2), she suggests that it opens the way to reconceive referentiality (p. 76-77). In my terms, it allows the question of reference to be re-introduced without giving up the insights won by the new sociology of science when it bracketed reference.

The relation of constraints to representation can now be articulated more precisely. When constraints become representations, they necessarily assume a positive cognitive content that moves them from the cusp into the theater. When I say "The total entropy of a closed system never decreases," I am expressing a representation of a constraint. Representations of this kind operate along the diagonal that connects inconsistent and consistent models. At the cusp, the interactions expressed by these representations have no positive content. The inability of language to specify these interactions as such is itself expressed by the elusive negativity that exists within the not-true position. The diagonal connecting true and not-true reveals their common concern with the limits of representation. At the positive ("true") end of the diagonal, the limits imply that we cannot speak the truth. At the negative ("not-true") end, they paradoxically perform the positive function of gesturing toward that which cannot be spoken. Elusive negativity, precisely because of its doubly negative position, opens onto the flux that cannot be represented initself.

The complete semiotic square can now be given.

exclusion (inconsistent) False ------------ True (unoccupied)

implication/separation

(unknown) Not-True ------------ Not-False (consistent)

It is no accident that the semiotic constraints generating the semiotic square bring the not-true position into view. Language structures how we conceptualize any representation, including mathematical and scientific ones. But language is not all there is. Elusive negativity reveals a synergy between physical and semiotic constraints that brings language in touch with the world. Physical constraints, by their consistency, allude to a reality beyond themselves that they cannot speak; semiotic constraints, by generating excess negativity, encode this allusion into language. There is a correspondence between language and our world, but it is not the mysterious harmony Einstein posited when he said that the mystery of the universe is that it is understandable. Neither is it the self-reflexivity of a world created through language and nothing but language. Our interactions with the flux are always richer and more ambiguous than language can represent. Elusive negativity, acknowledging this gap, gestures toward this richness and so provides a place within semiotic systems to signify the unspeakable--to signify the cusp.

IV. Making Connections: The Language of Metaphorics
To posit a model for scientific inquiry is to presupposeor evoke a correlative view of language. A realistic model calls for and is reinforced by the assumption that language is a transparent medium transmitting ideas directly from one mind to another; a positivist model produces and is produced by attempts to formalize language into theory and observation components; a social constructivist model is associated with a non-referential view of language that sees discourse operating through relations of sameness and difference. These correspondences are not accidental. They must obtain in any coherent account of scientific inquiry, for inquiry is constituted as such only when it enters the social arena of discourse. Like other representations of scientific inquiry, constrained constructivism corresponds to a particular view of language. The view of language correlative with it can be found within the emerging field of metaphorics. The difference between a representation consistent with reality and one that depicts reality is the difference between a metaphor and a description. Constrained constructivism thus implies that all theories are metaphoric, just as all language is. Metaphorics, defined as the systematic study of metaphoric networks as constitutive of meaning production, presents a view of scientific inquiry that enriches and implies the figure of representation presented here.

Since Max Black's influential analysis of metaphor, it has become customary to emphasize the power of metaphor to create new understanding. (15) According to this argument, metaphors not only express similarities between disparate concepts; they also set up complex currents of interaction that change how the terms brought into relation are understood. (16) A similar argument is adopted by Lakoff and Johnson in Metaphors We Live By. (17) Like Black, Lakoff and Johnson are concerned with systems of associated commonplaces that infuse into each other when two terms are brought into metaphoric interplay. Their emphasis falls on ordinary metaphors which, precisely because they do not surprise, reveal presuppositions deeply embedded within the culture.

In Arbib and Hesse's The Construction of Reality, metaphorics is explicitly connected with scientific inquiry. (18) They argue that perception takes place through schema which operate through relational similarities and differences. The category "dog" has as its reference not some Platonic idea that captures the essence of dog, but a network of individual perceptions that form a group, albeit one fuzzy at the edges. In their account, the tension between similarity and difference characteristic of metaphor, farfrom being a special subset of language usage, is fundamental to how language works. The "loose bagginess" of the metaphoric relation allows for constantly changing configurations within metaphoric networks; these changes in turn correlate in a systematic fashion with shifts in paradigms. "Scientific revolutions," Arbib and Hesse write, "are, in fact, metaphoric revolutions, and theoretical explanation should be seen as metaphoric redescription of the domain of phenomena" (p. 156).

In James J. Bono's account, metaphorics allows cultural presuppositions to be articulated together with scientific discourse systems. (19) Bono argues that metaphor functions "as both the site and means for exchanges among not only words or phrases, but also theories, frameworks, and most significantly, discourses" (p. 73). He envisions interactive, synchronic networks of metaphors that span disciplinary boundaries, in which traces of metaphors inherited diachronically from disciplinary traditions interfere and intersect with other metaphoric systems within the culture. Meaning production in this account can never be contained within a scientific field alone. Rather, it depends upon and emerges from resonances and interferences between inter- and extra-scientific networks of metaphors that engage one another at highly specific sites.

Constrained constructivism matches these views of scientificlanguage with an interactive, dynamic, locally situated model of representation. Recognizing that scientific theories operate within the theater of representation, it emphasizes that meaning production is socially and linguistically constructed. The elusive negativity that is a consequence of taking consistency rather than congruence as a standard for correctness reveals ambiguities intrinsic to any account of scientific models. These ambiguities ensure fluidity in language, thus reinforcing the claim that scientific revolutions are effected through metaphoric redescription. Finally, the transformative nature of interactions at the cusp makes the model context-dependent as well as species-specific, encouraging the idea that specific exchanges take place at local sites. Constrained constructivism thus presents a figure of representation that itself can be a metaphor for the inquiries of metaphorics.

V. Situated Knowledge: No Outside But A Boundary
Constrained constructivism puts limits on Derrida's aphorism that there is no outside to the text. Although there may be no outside that we can know, there is a boundary. The consequences that flow from positing a boundary or cusp rescue scientific inquiry from solipsism and radical subjectivism. At the same time, constrained constructivism acknowledges that we cannot have direct, unmediated access to reality. There is much to be said on why this acknowledgment is felt as an intolerable limitation by some realists. In "Situated Knowledges: The Science Question in Feminism and the Privilege of Partial Perspective," Donna Haraway alludes to the ideology embedded within an omniscient viewpoint when she calls t a "god trick." (20) Objectivity is associated with a view from everywhere, and hence from nowhere--a view with no limitations and hence no connections to humans located at specific places and times. (21) That it is a power trip is undeniable. That this power has frequently been misused is also undeniable. The illusion that one can achieve an omniscient vantage point, and the coercive practices associated with this illusion, have been so thoroughly deconstructed that they do not need further comment here. The liberatory spirit with which the critiques of objectivity were undertaken has been realized in the valuable contributions they have made to our understanding of how ideology and scientific objectivity mutually reinforce each other.

But in the process, objectivity of any kind has gotten a bad name. I think this is a mistake, for the possibility of distinguishing a theory consistent with reality from one that is not can also be liberating. If there is no way to tell whether the claim that blacks and women have inferior brains is a less accurate account of reality than the claim that they do not, we have lost a valuable asset in the fight for liberation. George Levine eloquently made this point when he argued for the need to break out of coterie politics and strive for a faithful account of reality. (22) Donna Haraway also recognizes this possibility when she calls for a paradoxical, non-innocent stance that will recognize limited objectivity at the same time that it continues to deconstruct all claims to omniscient knowledge. The problem she wrestles with is underscored by Levine as the central issue of the contemporary sociology of knowledge: "how to have simultaneously an account of radical historical contingency for all knowledge claims and knowing subjects, a critical practice for recognizing our own 'semiotic technologies' for making meanings, and a no-nonsense commitment to faithful accounts of a 'real' world, one that can be partially shared and that is friendly to earthwide projects of finite freedom, adequate material abundance, modest meaning in suffering, and limited happiness" ("Situated Knowledges" p. 579).

Haraway's solution is to emphasize that every perspective is partial, all knowledges situated. She tackles the difficult task she sets herself by continuing the vision metaphor but insisting that it is partial and contingent rather than full and unlimited. I am fully in sympathy with her project, and I think that she has articulated the central problem that a feminist sociology of knowledge faces. I am concerned, however, that the idea of partial vision can be easily misconstrued. It can be taken to suggest that part of our vision sees things as they really are, while only part is obscured. Whatever our vision is, this is not the case; we see things whole, not in parts. An alternative approach is to follow the lead of Merleau-Ponty when he suggests that situatedness, far from being a barrier to knowledge, enables it. (23) Given that we are not God, we can only come in touch with the universe through particular sets of sensory apparatus located within specific cultures and times. Constrained constructivism has this double edge: while it implies relativism, it also indicates an active construction of a reality that is meaningful to us through the dynamic interplay between us and the world. Renouncing omniscience and coercive power, it gains connectedness and human meaning.

Endnotes *
In writing this essay, I have benefited from conversations and correspondence with F. C. McGrath, Ronald Schleifer, Walter Freeman, Evelyn Keller, and James Bono. George Levine and Gillian Beer gave helpful encouragement and guidance.

1. Donna Haraway, "Animal Sociology and a Natural Economyof the Body Politic, I and II," Signs 4 (1978): 21-60; Steven Shapin and Simon Schaffer, Leviathan and the Air Pump: Hobbes,Boyle, and the Experimental Life (Princeton: Princeton Univ, Press,1985); Bruno Latour, Science in Action: How to Follow ScientistsEngineers through Society (Milton Keynes: Open Univ. Press, 1987).

2. Michel Serres, Hermes: Literature, Science, Philosophy,ed. Josué V. Harari and David F. Bell (Baltimore: Johns HopkinsUniv. Press, 1982) 106.

3. "What the Frog's Eye Tells the Frog's Brain," J. Y. Lettvin,H. R. Maturana, W. S. McCulloch, and W. H. Pitts, Proceedings of the Institute for Radio Engineers, 47 (1959): 1940-51.

4. For a summary of visual mechanisms in different species, see Models of the Visual Cortex, eds. David Rose and Vernon G. Dobson (New York: John Wiley and Sons, 1985).

5. Christine A. Skarda, "Understanding Perception: Self-Organizing Neural Dynamics," La Nuova Critica 9-10 (1989): 49-60. See also Walter Freeman and Christine Skarda, "Mind/Body Science: Neuroscience on Philosophy of Mind," John Searle and His Critics,eds. E. LePore and R. van Gulick (London: Blackwell, 1988);and "Representations: Who Needs Them?" Proceedings 3rd Conference on the Neurobiology of Learning and Memory (forthcoming).

6. Walter Freeman, private communication.

7. Jacques Derrida, Dissemination, trans. Barbara Johnson Chicago: Chicago Univ. Press, 1981).

8. Evelyn Fox Keller, Reflections on Gender and Science (New Haven: Yale Univ. Press, 1985); Ilya Prigogine and Isabelle Stengers, Order Out of Chaos: Man's New Dialogue with Nature (New York: Bantam, 1984); Nancy Cartwright, How the Laws of Physics Lie (New York: Oxford Univ. Press, 1983); and Michel Serres, Hermes (1982).

9. A. J. Greimas, "The Interaction of Semiotic Constraints," On Meaning: Selected Writings in Semiotic Theory, trans. Paul J. Perron and Frank H. Collins (Minneapolis: Univ. of Minnesota Press, 1987) 48-62.

10. Ronald Schleifer, A. J. Greimas and the Nature of Meaning:Linguistics, Semiotics and Discourse Theory (London: Croom Helm,1987) 22-55.

11. Karl L. Popper, Conjectures and Refutations: The Growthof Scientific Knowledge, 2nd ed. (New York: Basic Books, 1965).

12. Fredric Jameson, "Foreword," On Meaning (xvi).

13. Ronald Schleifer, "Analogy and Example: Heisenberg, Negation, and the Language of Quantum Mechanics,". ms. See also Ronald Schleifer, Rhetoric and Death: The Language of Modernism and Postmodern Discourse Theory (Champaign: Univ. of Illinois Press, 1990).

14. Shoshana Felman, The Literary Speech Act: Don Juan with J. L. Austin, or Seduction in Two Languages, trans. Catherine Porter (Ithaca: Cornell Univ. Press, 1983) 141-2.

15. Max Black, Models and Metaphors (Ithaca: Cornell Univ.Press, 1962). See also "More About Metaphor," Metaphor and Thought, ed. Andrew Ortony (Cambridge: Cambridge Univ. Press, 1979) 19-43.

16. Paul Ricoeur emphasizes the torque that metaphors put on terms in Interpretation Theory: Discourse and the Surplus of Meaning (Fort Worth: Texas Christian Univ. Press, 1976).

17. George Lakoff and Mark Johnson, Metaphors We Live By (Chicago: Univ. of Chicago Press, 1980).

18. Michael A. Arbib and Mary B. Hesse, The Construction of Reality (Cambridge: Cambridge Univ. Press, 1986) 147-70.

19. James J. Bono, "Science, Discourse, and Literature: The Role/Rule of Metaphor in Science," Literature and Science: Theory and Practice, ed. Stuart Peterfreund (Boston: Northeastern Univ. Press, 1990) 59-89.

20. Donna Haraway, "Situated Knowledges: The Science Question in Feminism as a Site of Discourse on the Privilege of Partial Perspective," Feminist Studies 14 (1988): 575-99.

21. For a different (and more realist) position on how subjectivity and objectivity can be integrated, see Thomas Nagel, The View from Nowhere (New York: Oxford University Press, 1986).

22. George Levine, Plenary Address at the Society for Literature and Science Society Conference, September 1988.

23. Merleau-Ponty, The Phenomenology of Perception (London: Routledge and Kegan Paul, Ltd., 1962).