In the beginning of the previous chapter, I offered two theses that credit-giving views of science endorse: (CG1) there is a world independent of human inquiry, and (CG2) knowledge of that world is such by virtue of accurately describing it.  We can pick out a parallel pair of theses for extreme social constructivist views to endorse: (SC1) our only access to reality is through social processes, and (SC2) reality is constituted by the social processes that allow us “access” to it.  This is as neat a comparison as I would allow myself, and complicating issues turn up right away.  Most apparently, the order is reversed in the second pair.  The epistemological claim comes first and the ontological one second.  I did this because there is a dependence relation between the two.  According to the way social constructivist arguments typically go, we should believe (SC2) because we recognize (SC1).  By way of comparison, (CG2) is dependent on (CG1) in that we could not know an indepent world if one did not exist.  This is a simple and straightforward dependence that causes the credit-giver no problems.
 The dependence relation between (SC1) and (SC2) is, however, a problematic one.  It consists of a unjustified move from a claim about what we know to a claim about what there is.  Even if we do not have any access at all to a reality separate from our social interests, that does not necessarily mean that there is no reality separate from our social interests.  A move from an epistemological claim to a ontological one is not warranted.  Especially in this particular case, quite the opposite is true.  If (SC1) is true, and we take the fact that our access to reality is unavoidably social to mean that we cannot perceive what reality is truly like, then we can make no claim about the character of that reality.  (SC2) is a claim about the character of reality, namely that no independent reality exists, and since we have agreed that we can’t know about that, (SC2) cannot be consistently asserted.  A probable response from social constructivists like Latour and Woolgar is that (SC2) is the kind of move characteristic of the idealism of the modern period.  The social constructions of scientists, as they are all we have epistemically, are constituative of reality.  This move depends on the idea that we should build ontology on epistemology.  I suggest that it only sounds plausible given a starting point of Cartesian-demon-type skepticism which I will not fully argue against.  I offer a partial respone (I happen to think it is good enough) in the following section, but avoid engaging the global skeptic fully, as that would bot be within the scope of this project.
 On this idealist rendering of (SC2), it does seem to follow from (SC1)--if we allow an unusual use of the term ‘reality.’  To me, this use of the term is so unusual that it almost exactly reverses the accepted, intuitivie meaning of the term, using it to apply to our social (not determined by correspondence with states of affairs in the world) construction as opposed to an independent world.  ‘Reality’ is conventionally used to refer to an independent world, while this use specifically picks out something that is not an independent world.  I suggest that this misuse of ‘reality’ makes (SC1) and the idealistically rendered (SC2) entirely counterintuitive.  Latour and Woolgar would probably cheerfully agree.  But the traditional concept that reality is something “out-there,” prior to and independent of our inquiry, should not be abandoned so quickly.  Even on the coherent--but counterintuitive--idealist rendering, we can still ask whether (SC1) and (SC2) are true or not.  I contend that they are not.  The way I propose we decide is by checking the relevant state of affairs in the world.  This may be unfair to social constructivists like Latour and Woolgar, who say that we cannot do so and that any account will survive or fail or be modified based on its power to convince, but let me just say that--for me and many others--an account’s power to convince is greatly increased by how well it matches with the state of affairs it putatively describes.

The observable world
Ordinary, mid-size objects like tables, chairs, dogs, cats, cows, and automobiles can be confidently described as independent of human inquiry.  Of course, our understanding of these objects is social to a certain degree.  I wouldn’t call a table a ‘table’ if I had not been brought up speaking English, and I wouldn’t have a certain set of ideas about what a table is used for, where it comes from, and so on if I had not been brought up in (or exposed to) a society which uses tables in the way ours does.  In short, I might not know an object with legs and a flat, horizontal surface as a ‘table’ or know what such an object is used for.  But it is also obvious that there are features of tables which do not depend on social factors like language and upbringing.  Being able to stub one’s toe one a table leg comes to mind.  Let me stick to the example of being able to see the features of things like tables.  The surface of a table will reflect light which can be sensed by the human eye.  This process will tend to bring about a belief about the presence, location, and other features of the table in the human observer.  These beliefs are caused and justified by causal interaction with the table, just as the pain in one’s toe is caused by interaction with the table leg.  Interactions like these can be performed by anyone with the relevant sensory apparatus, regardless of social factors.  We can devise two tests to effectively prove that beliefs formed about the table are not socially constructed.  (1) Have people from significantly different social backgrounds interact with the table in the same way--e.g., stub their toes, identify its location, size, color, etc.  (2) Have people in the (as close as possible) same social situation interact with many different tables and see if their interactions differ in correspondence to the differences in the tables--here we might use trick tables that do not feel as hard as they look or other oddities as a check for the consistent application of shared social schema for table-interaction.  The results, we can be sure, would indicate that much of the way people interact with tables is not determined by social factors.  We perceive consistency in the features of ordinary objects like tables because their physical behavior is independent of our epistemic interaction with them.  By this I mean to allow for the fact that we can, of course, interact with ordinary objects.  Stubbing one’s toe on a table leg, for instance, might jar the table.  This is a causal interaction between physical bodies and is not the result of a social construction.  But the kind of interaction indicated by (SC1) and (SC2)--something like the idea that my socially-induced concepts about tables are the reason that they can be moved by kicking them--should be ruled out.
 Richard Boyd puts the point nicely.  He says that realists must affirm the “no noncausal contribution thesis,” while constructivists will deny it.  The thesis is that “human social practices make no noncausal contribution to the causal structures of the phenomena scientists study.”   Of course humans can interact causally with objects--they can build houses, eat plants, and stub their toes--and social practices can even be the source of causal interaction, as someone who lives in a carpentered society like ours is more likely to build a house, thereby interacting with the materials she uses in a certain way because of that social factor.  But social practices do not interact with the world in a noncausal way--an example of which would be that table legs are actually hard enough to hurt people’s toes because of socially-transmitted fears about hurting toes focused on tables.  This would be the kind of reality-constituting social activity that Latour and Woolgar attribute to their endocrinologists.  Scientific realists must be committed to the idea that this does not happen.
 That an independent, real world of ordinary objects exists is surely the best explanation for the consistency of our perception of ordinary objects.  As I mentioned before, I do not want to get caught up in the argument against global skepticism.  I think that the strength and cross-cultural consistency of the way we experience the observable world is argument enough against the skeptic.  To deny an independent, real world would necessitate positing something else that performs the same function; that is, something else that acts just like an independent real world.  There is a lack of economy to such a model and there is, I think, intuitive inclination to say, “If it acts just like an independent, real world, what else could it be?”  Social constructivists, of course, may be willing to say that something that walks like a duck and quacks like a duck is still a social construction, but this means ignoring the power of a theory committed to reality to explain the consistency of the behavior of the observable world.

From observable to unobservable
I have mentioned that I think constructivists have a problem with the observable world.  Bloor admits that we can be right about the observable world, but he thinks scientific theories that invoke unobservable entities or phenomena can never be shown to be true because observable evidence will never fully determine the correctness of any theory.  This commits him to positing a cut-off point where objects cease to be observable.  Latour and Woolgar and Knorr-Cetina either follow the same strategy or, as I suggested before, don’t even believe that we can know about an independent, observable world.  If they follow the latter strategy, they are wrong, as I argued in the previous section.  Let us charitably assume that all the social constructivists agree about the independence of the observable world and so are committed to a cut-off point between observable and unobservable.
 Bloor does not mention where he thinks such a cut-off might be.  He might not need to, if the idea is plausible.  Is it?  Grover Maxwell has argued that there is no possible principled cut-off between observable and unobservable entities, only a continuum of harder-and-harder to observe.   He refers to the logical empiricist’s idea that looking through a microscope does not mean looking at physical things.  He says, “if this analysis is strictly adhered to, we cannot observe physical things through opera glasses or even through ordinary spectacles, and one begins to wonder about the status of what we see through an ordinary windowpane.”   Maxwell argues that there is a continuum from looking through a vacuum to looking through reading glasses, up to looking through a high-powered microscope.  Any line drawn between observable and unobservable will be arbitrary and depend on contextual factors like the resolution of available microscopes.  How can we justify the idea that things on one side of our arbitrary line have physical thinghood and those on the other side do not?  If we take any two consecutive stages in the continuum, can we really say that things observed at one stage are “slightly less real” than those at the previous?  Maxwell’s arguments are to the effect that, given the continuity between observability and unobservability, and the fact that a continuity between existence and non-existence sounds nonsensical, the view that unobservables are significantly, ontologically different in themselves from observables is implausible.  His article occupies a place within a debate between realism and logical empiricism, with its view that unobservables are “theoretical entities,” “convenient fictions,” or something of the kind.  Maxwell argues against the theoretical/observational dichotomy.  But the realist claims he makes can be aimed at social constructivism as well, since it includes the ideas that (at least) unobservables do not really exist, but are constructed.   Instead of the theoretical/observational dichotomy, we have a socially constructed/observational dichotomy to fight.
 Does Maxwell’s argument destroy the distinction between observables and unobservables?  Bas van Fraassen has countered Maxwell by claiming that, although there may be a continuum, there are clear cases of unobservables which would surely occupy a place close to the end of the continuum.  Such entities--the electron comes to mind--might still be able to take their place in Bloor’s theory as the unobservables about which we can never have true theories.  Van Fraassen argues that the observability of bodies seen through a telescope comes from the fact that we could, in principle, go and look (at the moons of Jupiter, for example).   We cannot, in actuality or in principle, go and look at an electron.  Quantum uncertainty aside, a being that could see electrons would simply not be human.  There is a genuine limit to what we can observe.  But there is more to Maxwell’s continuum argument than just stating it and exclaiming that there is no principled difference between observables and unobservables.

The continuum argument applied to microscopy, despite the fact that it does not prove that there are no unobservables in principle, is important as a model or outline of the type of argument that is needed to demystify unobservables.  What is important about it is what it indicates about the relation, in principle, between successive stages of magnification.  Take a hypothetical first stage of magnification, which will be very weak.  We can look through the microscope and see magnified bodies.  But an observer with very good vision can look closely at the bodies with the naked eye in order to compare what she sees with the microscope image.  This is what I call an observable-to-observable check.  It validates the first stage of magnification, putting it at a similar level of reliability with ordinary observation.  We can then use the first stage to validate a second stage, and so on.
 Of course each successive stage cannot be exactly compared.  There are elements of the magnified image which are simply not available at a previous stage, otherwise magnification would be pointless.  But we can at least see a mapping of the images on to each other that allows comparison.  If a circular body is looked at under a microscope, the magnified image reveals a larger circular object--and of course this holds for as great an amount of detail as can be perceived in the unmagnified image.  Ian Hacking offers another argument which lends support to the significance of the observable-to-observable check which he calls the argument of the grid.   Tiny, numbered metal grids are used to identify and compare the arrangement of microscopic bodies.  These grids are manufactured by drawing a large grid with numbered squares in ink, photographically reducing it, and depositing metal on the resulting micrograph.  If we then look at the tiny grid through a microscope, we see exactly the same grid and numbers as were drawn in the large version.  We know the microscope image is accurate because we know what the grid ought to look like, and that is what we see.
 This kind of continuum argument, based on showing the reliability of moving successively farther and farther away from ordinary observability, allows us to have confidence in the similarity of the microscopic world to the observable world.  It works for telescopy in a similar way.  On earth, we can look at something far away through a telescope and then go check with the naked eye.  In space, we have “gone and checked” as far as the moon--in terms of naked-eye vision.  We could, in principle, go farther.   The continuum argument in microscopy is not limited to optical microscopes.  Using something like the stage-by-stage check I rehearsed above, we can compare the representations of something as complex as an electron microscope with the representations of methods we have a great  deal of confidence in, like optical microscopes.  At a low resolution, electron microscopes provide images that can be compared with those of optical microscopes, which gives us confidence in the accuracy of this fundamentally different process.
 I also think this argument of degrees of observability helps us to infer the observability--in principle--of organisms in the past.  The physical impossibility of time travel obviously prevents a straightforward “go and check” strategy, but the time that passes in a normal human lifetime allows us to learn a considerable amount.  We can, for example, observe a cow easily enough.  Say the animal dies and the body is left to decompose.  We then know that the resultant skeleton came from the observed cow.  The observability of an organism which lived in the past and left a skeleton can be inferred by way of analogy with our late cow.  Fossilization happens within five to ten years, and so can be observed by humans as well, giving us reason to believe that fossilized skeletons of animals never seen by humans would have been as observable as cows, had anyone been there to look.  The other part of the solution is being able to explain the process by which unobservables become observable.  Explainability lends great support to the lack of a principled difference--although it may not sound persuasive to a social constructivist, who will doubt the explanation as much as the observation.  What is important to my use of the continuum argument is the possibility of an observable-to-observable check, which is what the “go and check” strategy with terrestrial telescopy and the analogy argument for the observability-in-principle of past organisms are.
 How does the continuum argument factor in my appraisal of the difference between observables and unobservables and what does that mean for the reality of the unobservable world?  The power of microscopes to produce accurate images of unobservable entities in effect makes those entities observable.  The observable-to-observable check gives us confidence in the process.  The comparability of successive stages gives us confidence in the magnitude to which we use the process.  The fact that unobservable entities under magnification behave like observables viewed with the naked eye combined with the very idea of a continuum of degrees of observability, which allows no cut-off point, indicates that the entities are not significantly different enough in themselves to justify a principled difference between observable and unobservable ones.  That is, it’s not necessarily significant to the nature of microscopic entities that we cannot normally see them with the naked eye.  That is a limitation of human beings’ sensory apparatus which we can escape through the use of instruments like microscopes.  To argue that unobservables are ontologically different from observables based solely on the fact that we cannot observe them makes no sense.  Are the limmitations of the sensory powers of human beings really the determining factor of the reality or the natrure of the existence of all entities in the universe?  This seems, to me, impossible.  A tree falling in the forest does make a sound, even if there is no one there to hear it.  It makes a sound because sound is a vibration of air molecules that does not depend on the presence of human ears for its occurence.  But this point leads us to what is so great about human ears.  They sense vibrations in air molecules by amplifying and converting vibrations upon the ear drum, allowing a causal interaction with the world that gives humans the ability to form correct beliefs about states of affairs in the world.  I will return to the issue of causal interaction later.
 Let me first restate just what the idea of a continuum between observable and unobservable does for a credit-giving view.  The indepence and reality of the observable world is unproblematic.  Indeed, barring some kind of global skepticism--which I have elected not to argue against--it is very difficult to deny.  The continuum argument shows that the ontological status of the observable world is contagious.  That we know it to be real means, if we think a little about the difficulty of finding a good place to put the border between observable and unobservable, it becomes clear that a radical change in ontological status between the two is implausible.  The change from “independent of human inquiry” to “dependent on human agonistic activity (or any other social process)” is just such a radical change.
 Take as an example a small insect with compound eyes.  Imagine that the compound eyes are just large enough to be visible as tiny dots.  The individual eyelets cannot be observed with the naked eye.  It makes no sense to say that the observable eyes of the insect are real and independent while the eyelets are not.  The unobservable eyelets, as constituent parts of the observable eyes, must surely have an independent reality also.  This is how the ontological status of the observable world is contagious.  But there is also an epistemological point to be made here.  Let’s say we wish to count the number of eyelets in our insect’s compound eyes.  We have established that the eyelets have an independent reality and so they must have a determinate number.  We can count the larger, observable eyes to learn how many of them there are.  Our observation of the observable world is thereby established as a source of knowledge about that world.  The continuum argument, with regard to observation, shows that a stark contrast between observation of observable objects and (previously) unobservable objects is untenable.  We can learn about observables through observation, and given the continuum between the two, we know that we can learn about ubobservables as well.  All we need is confidence in the reliability of the process.  The observable-to-observable check provides this, as do arguments presented in the following section.

Technology, explainability, and the argument from coincidence
The continuum argument will only get us as far as the highest resolution of instruments which work like microscopes and telescopes, with their gradually increasing degrees of magnification.  I have also suggested something like it can be used to infer the observability of organisms in the past.  This “time-travel” form is limited to inference about the results of processes which can be observed in a human lifetime.  This leaves a lot to be desired.  Van Fraassen’s clear cases of unobservables still stand.  You can’t turn a microscope up high enough to look at an electron.  However, the continuum argument is bolstered by others.
 Two related arguments are those from technology and explainability.  The technology argument is based on the manipulative success of the technology involved in scientific instruments.  This is the idea that technology, the development and use of which is at least somewhat independent of scientific theory, has a certain degree of recognizable power.  The power of technology independent of theory is evident in cases where the technology is developed and used before the theoretical explanation comes along.   Of course, when a theoretical explanation does accompany powerful technology, it strenthens the case for the accuracy of the instruments even more.  This is the explainability argument.  It is based on the idea that being able to explain the way an instrument works with the relevant scientific theory is another reason to believe in the accuracy of a representation yielded by the instrument.
 Technology and explainability connect in a special way.  That the manipulative power of instruments and methods can be explained by scientific theories is reason to believe that the theories are approximately true.  The move toward greater complexity and explanatory power over the course of the history of science has been accompanied by more and more powerful technology.  We have a body of scientific theories which have enabled the progress of technology, can explain the workings of the technology, and also fit to a great degree with an enormous body of empirical evidence.  It would be a miracle indeed if those scientific theories did not refer to the workings of the world in an approximately accurate and steadily improving way.
 Perhaps the most powerful of the arguments allied with the continuum and explainability arguments is one which Ian Hacking calls the argument from coincidence.  It is similar to my thought experiment which shows the independence of the observable world.  The idea is that different instruments, the technology of which depends on different parts of science, reveal the same phenomena.  Hacking cites the example of an electron microscope showing constellations of dense bodies on red blood platelets.   The experimenter suspected that the dense bodies were artifacts of the electron microscope equipment, and checked the same sample with a fluorescent microscope.  The light microscope revealed the same bodies in the same constellations, verifying their independence from the electron microscopy method.  Examples of observing the same phenomena with different instrumentation abound.  They prove that each individual method is not arbitrary or constructed.  The similarity of results means that, for social constructivists to prove that an eperimental result is constructed, they would have to prove that every other method that might test the same phenomena would also not only be constructed, but be constructed in a way that would yeild the same results.  This indicates a picture of science as something more than isolated practices which seem to lend themselves to being viewed as construced.  If different, relatively isolated parts of science indicate the same result, we are lead to a picture of science as one big, extremely coherent social construction, one which acts a lot like an activity that finds out about an independent, real world.  As Hacking puts it, “it would be a preposterous coincidence if two totally different kinds of physical systems were to produce exactly the same arrangements of dots on micrographs.”
 To get back to observability, the argument from coincidence is another good reason to believe in the accuracy of the representations of scientific instruments.  We can compare images across different methods, which gives us faith in the methods and in the idea that the entities we perceive with instruments act like entities we can observe without instruments.  Our increased confidence in the accuracy of instruments gives us reason to believe that the representations they produce allow us to observe the phenomena under study.  That the phenomena can be consistently observed in different ways is reason to believe in their independence from any of the methods of observation.  This is a general point, not one based on the resolution of microscopes.  The argument from coincidence is, in itself, a reason to believe in the reality of the observed phenomena, because of what it indicates about the possibility of the accuracy of our observations.  The power and reliability of instruments make the unobservable observable.  The observability of previously-unobservable phenomena in different and accurate experimental set-ups gives us confidence in the independence and reality of the phenomena in the same way as our observation of ordinary objects gives us confidence in their independence.

I have tried to keep this discussion about observability general, but the fact is that each instrument and each experimental set-up deserves its own justification.  I have sketched such a justification for microscopy, with an implication that it should serve as a model for others.  However, the continuum strategy in microscopy is not promising for the justification of a method like the use of bubble chambers in high energy physics.  But I have offered a combination of related arguments which should indicate how to defend such a case.  Showing the manipulative power of the equipment involved in HEP is likely to be very powerful.  Hacking argues that using an electron as a tool in an experiment commits the experimenter to believing in its reality.  I contend that using a particle accelerator commits the experimenter to believing in its manipulative power.  Although the technology argument with its emphasis on manipulative power may only warrant something like instrumentalism, the explainability argument, along with the well-known strategy of inference to the best explanation, warrants belief in the reality of and the theories about the phenomena observed with reliable instruments.
 The continuum between observable and unobservable casts doubt on our ability to draw a meaningful line--according to the criterion of observability--which separates things which exist, or things we can know and everything else.   Neither the ontological distinction upon which the empiricist depends nor the epistemological distinction upon which the constructivist depends can reasonably be based on assessments of observability.  The consistency of the results of different scientific experiments and the consistency between those results and the the relevant theories along with the consistency of scientific theories with a body of technology which has very real manipulative power is evidence for the fact that those theories approximately accurately describe the phenomena to which they refer.  There is not enough space in the present project to elaborate a full defence of this point, which would need to include answers to concerns about underdetermination,  the pessimistic meta-induction,  and other complex issues.  I think these concerns can be answered.  But the current project is an answer to social constructivism, which contends that science proceeds according to social factors, rather than input from an independent world.  All I really need to show is that science is not entirely controlled by social factors and that the social forces that are present are not necessarily damaging to its cognitive status.  The next two sections are concerned with these objectives.

Causal interaction
I have alluded to the idea that causal interaction is what’s so great about human perception.  We see because our eyes sense light which is reflected or emitted by nearby objects.  The way the light rays are organized tells us about the object from which they come.  Sense perception can serve as evidence because it involves interacting with the world in a way that allows the state of affairs in the world to affect the character of the relevant belief.  The use of instruments and systematic experimentation to learn things we cannot with our unaided senses and the limited empirical data gleaned in an ordinary individual’s life are what’s so great about science.  Of course the important part about experimentation is just the same as in unaided sense perception; that the state of affairs in the world affect the character of the beliefs formed.  This is what social constructivists argue doesn’t happen.  Latour and Woolgar contend that social processes control what becomes scientific fact, and that “the effect of reality” is produced as a result of the solidification of a fact.  For them, there is no reality “out there” for beliefs to be formed in accord with.  I have argued that the observable world is a ready example of reality “out there.”  Observable objects can be consistently perceived across different sense modalities, individuals, and cultures.  The behavior of observable objects is consistent in many important respects and general laws predicting and explaining their behavior can be formulated.  This forces the constructivist to retreat to the unobservable world, as Bloor does.  But using observability as a criterion for a distinction between real and not-real has also turned out to be problematic.  Noting the similarity of the unobservable world to the observable leads us to recognizing that it exhibits a similar reality and independence from human inquiry.  The question remains whether our inquiry allows us access to that world.
 The extreme constructivist thesis about scientific inquiry is that there is nothing about that inquiry which affords us access to the workings of nature.  This is a very strong claim, and the empirical research done in microsociological case studies simply does not support it.  Latour and Woolgar’s third chapter, “The Construction of a Fact,” mentions many assays, tests, and stages of experiment, and never shows that any of them is inconclusive or suspect in any way.  Indeed, without getting into the relevant endocrinology--which they don’t--there is no way to evaluate whether the experiments described interact with the world in a way conducive to producing a true account.  Surely some experiments allow scientists to interact with nature in a very obvious way.  In genetics, a scientist may simply count the number of offspring with a certain trait that are produced by parent organisms.  Ordinary sense perception may be an unproblematic method for such an experiment.  Constructivism has some difficulty showing that such measurements of observables are entirely socially constructed, because of the ease and consistency of interaction with the observable world.  Furthermore, just as the arguments of the above sections show the independence of unobservables by connecting their behavior and measurement to that of observables, we can assume that measurements of unobservables tend to afford genuine access to the relevant state of affairs because we know that measurements of observables do.
 The complete argument that a given experimental set-up involves sufficient causal interaction with nature will depend on the instruments used and the technology and theory behind them.  Inasmuch as some science does not use instruments and is based on measurements of observables, it is impossible to occupy the position that all of science is socially constructed.  I hope the trouble with the observable/unobservable distinction casts doubt on the possibility that all experiments which use instruments yield socially constructed results as well.  This point is easy to make in the field of microscopy.  But I must admit that a complete defense of the stance that a bubble chamber allows causal interaction with subatomic particles is not something I’m capable of.   I do not move from recognizing that to arguing that it is not something that can be done--which is what Latour and Woolgar do when they depend on an assumption of strangeness.  I think that the physics and technology in a bubble chamber can be explained in a way which should make us comfortable with the idea that experiments in high energy physics do allow scientists to interact with and observe particles.  And that is exactly what is so great about particle accelerators and bubble chambers.  This is a fact that probably seems so obvious to anyone who uses a particle accelerator that it’s not worth mentioning.  But philosophers tend to doubt obvious facts like it.  The recognition that instruments allow causal interaction in a way analogous to simple sense perception is just what we need to deny the extreme constructivist claim that there is nothing about science which is not socially constructed.  To prove the extreme opposing claim that there is nothing about science that is socially constructed would take more than an argument about the plausibility of the accuracy of measurements of unobservables.  But I won’t attempt to do that.  In fact, I don’t think it is possible.  There are social aspects of science.  There are social factors that are good for the cognitive success of science.  This is the topic of my next section.

Social factors in science
I recognize, along with the social constructivists, that science has an important social element.  I do not, however, agree with them that this means that science is entirely social or that it does not do what it claims to do.   I think that science’s superior epistemic status in our culture is warranted and that the socialness of science, jarring though it may be to holders of the traditional view of science, does not take away from that status.  It seems that there are two immediate strategies for the response of a holder of a credit-giving view of science to the socialness of the activity.  The first, and probably most obvious, is to invoke a kind of god-of-the-gaps argument.  This is the idea that the social factors that we have admitted are present in science only really affect elements of scientific practice that are irrelevant to the cognitive output of the activity.  This is basically no different from the traditional defense against accusations of socialness.   Here I stand with the social constructivists and say that this position is not tenable.  But a credit-giving view need not depend on it.  Another strategy to present itself is the idea that the social aspects of science are neither damaging nor irrelevant to its cognitive output, but are positive forces, helping scientists find out about the world.
 What does it take for a social process to be good for science?  The general criterion is, of course, the same as that for any process: epistemic reliability.  Reliable processes--that is, ones which tend to lead to true beliefs--are exactly what is wanted in an activity concerned with discovering truth.  The reliability of scientific instruments is what I tried to establish in my arguments concerning the observable/unobservable distinction.  This is just the kind of thing that needs to be shown about social processes.
 An analogy or two will be of help here.  One set of social processes often pointed to in studies of science consists of those which influence the motivation of scientists and the reward processes that exist in the practice of science.  Scientists are apparently often motivated by things other than a pure, noble pursuit of truth.  They are motivated by the desire for credit, fame, or even money.  This does not, however, mean that they do not attain truth.  David Papineau offers a game analogy.   He says that few poker players are motivated by the desire to conform to the rules of the game; they are much more likely to be motivated by the desire to win money.  But they are still playing the game, following the rules, and when they win to earn money, they still win.  As Helen Longino quotes David Hull, “knowledge may be a by-product of the activities that obtain the primary results (e.g., credit) desired by scientists; but, because of the rules by which the game is played, it is still a product.”   When a scientist comes up with a good theory in order to get famous, she still comes up with a good theory.  An even stronger case for the usefulness of social processes than showing that ulterior motives don’t hurt is to show that elements of scientific method may be significantly social and still be good for science.  Think of shooting free throws in basketball.  Children playing in school or at the park probably learn how to shoot free throws entirely through social processes; they are likely to be concerned with doing it in a way that looks good as much as anything else--probably a large motivating factor will be the desire not to be made fun of by the other children.  But their techniques can be evaluated for whether they tend to make the ball go into the goal or not.  In this case, what “looks good” is likely to correspond to what children see professional basketball players doing.  And what professional players do is likely to be good at making the ball go through the hoop.  So a method that is learned socially and reinforced socially helps with the unsocial process of getting the ball through the hoop.   Indeed, to return to the issue of motivation, the rules of the scientific community are such that--even assuming a scientist’s motivation is solely to get famous--the best way to do that is to have original ideas, careful methods, and repeatable results.  These are exactly the kinds of things we need to be present to make sure what’s being done is good science.

Just what kind of social processes can the credit-giver admit are present in science without having to worry about the epistemic status of science’s cognitive output?  We should find two basic categories.  The first consists of factors which facilitate evidential processes, and the second consists of factors which can be more directly characterized as evidential processes.  In the first category, we find things like fund allocation and specialization of labor.  Laboratory science could not exist without generous grants from universities, governments, foundations, etc.  Having enough money is necessary to building a particle accelerator, and having a particle accelerator is necessary to creating evidential links between limited human knowers and subatomic particles.  Similarly, as mentioned above, a particle accelerator could not work without a team of many people, each assigned to different, specialized tasks.  The presence of obviously social factors like these and their importance to science is easily acknowledged by the traditional view of science.  That they facilitate evidential processes is the obvious explanation for their place in the practice of science, and they are not threatening to a credit-giving view.  Scientists’ motivations should also be placed in this category.  Scientists may be motivated entirely be the desire to earn credit and respect in the relevant community, but because of the way the standards are set up in that community, the best way to do that is usually by doing good science.  This is not to say that these factors necessarily always lead to the best scientific results, but that they tend to.  The reliability of factors like these is based on the idea of encouraging individuals or groups who show promise.  That is, achieving one significant result will lead to more funding and resources, which is likely to lead to more significant results.  This is a fine explanation for why scientists are motivated by the desire for credit; it will become a resource which they can invest in future research.
 The second category of social factors in science consists of those which act more directly as evidential processes.  These are the processes which foil the traditional strategy of using the social/evidential distinction as the deciding factor in what should be counted as good science.  They are processes which are correctly identified as social but are used in an evidential way in scientific practice.  As in the first category, this often comes down to the idea that individuals or groups who have produced good results in the past will produce good results again.  To put it another way, credit is earned.  As an example, take the prestige of a particular laboratory group.  This prestige increases the confidence of others reading a new paper by the group that the findings the paper reports are accurate.  The prestige is a social factor, but it is used by members of the community in the evidential process of deciding about the status of a particular finding.  Thus the prestige, which is earned through the past production of good results, becomes evidence for the accuracy of the newer result.  This implies that some people can be better at scientific inquiry than others, which has a clear enough parallel with the idea that some people are better at ordinary sense perception than others (certain people have, for instance, better vision than others).  Just as people can allow their perception of ordinary objects to be corrected by others with better vision, so does the scientific community use discussion and criticism of results to decide upon accurate representations.
 Social constructivists suggest that the socialness of science is something deeper than the easily-recognizable patterns of prestige alluded to above.  They argue that the most central and productive elements of scientific practice cannot be considered purely evidential (that is, unsocial).  But the idea that even the most central aspects of science are social need not be a threat to a credit-giving view.  Helen Longino has argued that the objectivity of science is dependent upon understanding scientific inquiry as fundamentally social and that two central elements to the production of scientific knowledge--observation and justificatory reasoning--are significantly (and unproblematically) social.   Her discussions are not simply statements of these claims but explanations of the way the socialness recognized bolsters the cognitive status of science.  Here the socialness of the processes involved is admitted and used to defend the objectivity of science.  Longino’s arguments tend to involve a theme along the lines of “many knowers are better than one.”  That is, observation is social in that it is not simply the sense perception of an individual, but an activity that is regular and intersubjectively invariant.  These are important aspects of the repeatability of experiments--that anyone else in the same situation, or in contact with a relevantly similar experimental set up would perceive the same results.  When experimenters check their results with colleagues and with the larger scientific community, it is a social process, but one which helps guarantee the objectivity of science.

Although Latour and Woolgar’s destruction of the social/evidential distinction seems to be unwarranted, we have seen that the distinction is not very helpful in thinking about science.  Constructivists, despite their denials, usually destroy the social/evidential dichotomy in favor of the social.  That is, they say that we cannot show that the two sides of the dichotomy are sufficiently different, and--in light of the traditional dependence on the unsocialness of science for its cognitive status--we must conclude that science is merely a social construction.  I have discussed the problems with this approach above.  A holder of a credit-giving view sufficiently informed by the useful insights of social constructivism might be tempted to break down the social/evidential distinction in favor of the evidential instead.  I don’t think this is the right way to think about the issue.  We can tell the difference between social and evidential categories of processes, and we can tell when a process fits into both categories.  But the spirit of the suggestion is right.
 Let’s keep in mind that the subject of the discussion is the epistemic status of science.  If we really stick to the point, what matters is not how social any given process is, but how evidential it is.  I have offered examples of social processes which can still be evaluated for their effectiveness in the relevant domain--basketball or describing the world.  That is what should be focused on in either case: the effectiveness in the relevant domain.  If a process is good at creating accurate descriptions of the world, why should we care how social it is?  This is, rather than a breakdown of the social/evidential distinction in favor of either side, more like a setting aside of the distinction and the recognition that what is important to thinking about science is evaluating the epistemic effectiveness of the processes involved.  It should be noted that social constructivists like Latour and Woolgar and Knorr-Cetina only attempt to evaluate the socialness of elements of scientific practice, not their epistemic effectiveness.  Considering the debunking tone of their projects and the fact that their conclusions are aimed at the cognitive status of science, this is clearly the wrong focus.  Discovering and describing social aspects of science is, of course, still an interesting pursuit.  My point is that studies of science which do not take epistemic evaluation as part of the data-gathering process should make epistemic evaluations in their conclusions.
 What should really be done in studies of science is evaluations of elements of scientific practice for how well they afford human beings contact with the workings of the world.  To do this they must be causally responsive to elements of the world and reliable in terms of yielding accurate representations.  If these conditions are satisfied, then science is not just a social construction (or set of many constructions), but rather an activity with some real connection to an independent world.  I have not attempted to pin down nature of that connection in this project.  I hope to have shown the superiority of a credit-giving view over any strong form of social constructivism.  A credit-giving view, in order to be a serious contender, however, should be informed by the socialness of science and should not make the same mistakes the traditional view does--mistakes which social constructivism calls attention to.  My umbrella category of “credit-giving views” leaves room for versions of both empiricism and scientific realism, and picking out the best one of these is a natural follow-up to this project.  My treatment of the relationship between unobservables and observables will give the reader some clue as to which I favor.  I think that empiricism is an ultimately unsatisfying way to give science credit, and that realism is the most promising view of science.  I have avoided articulating a defence of realism over empiricism or other credit-giving views because that is not within the scope of the present project.  The intent of this project is to diffuse the apparent threat to realism from social constructivism.  I have done so by discussing the superiority of credit-giving views in general--of which my own scientific realism is one.  The (purposely vague) picture of science that should come out of this project is one of science as an activity that is social in important ways.  However, the socialness of science is not, prima facie, threatening to cognitive status.  Claims that science is “just” a social construction can be refuted by showing that the practice of science allows the state of affairs in the world to be at least one determining factor to the character of scientific theories.  The remaining, uncontested socialness of science does not necessarily make even “socially constructed” science have no connection with states of affairs in the world.  Social factors, too, can be reliable in terms of leading scientists to accurate representations.