1989 - extragalactic reality- the case of gravitational lensing

Philosophy of Science Association

Extragalactic Reality: The Case of Gravitational LensingAuthor(s): Ian HackingSource: Philosophy of Science, Vol. 56, No. 4 (Dec., 1989), pp. 555-581Published by: The University of Chicago Press on behalf of the Philosophy of ScienceAssociationStable URL: http://www.jstor.org/stable/187781Accessed: 11/05/2010 09:33

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available athttp://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unlessyou have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and youmay use content in the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained athttp://www.jstor.org/action/showPublisher?publisherCode=ucpress.

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printedpage of such transmission.

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range ofcontent in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new formsof scholarship. For more information about JSTOR, please contact [email protected].

The University of Chicago Press and Philosophy of Science Association are collaborating with JSTOR todigitize, preserve and extend access to Philosophy of Science.

http://www.jstor.org

http://www.jstor.org/stable/187781?origin=JSTOR-pdf

http://www.jstor.org/page/info/about/policies/terms.jsp

http://www.jstor.org/action/showPublisher?publisherCode=ucpress

Philosophy of Science

December, 1989

EXTRAGALACTIC REALITY: THE CASE OF GRAVITATIONAL LENSING*

IAN HACKINGt

Institute for the History and Philosophy of Science and Technology University of Toronto

My Representing and Intervening (1983) concludes with what it calls an experimental argument for scientific realism about entities. The argument is evidently inapplicable to extragalactic astrophysics, but leaves open the possibility that there might be other grounds for scientific realism in that domain. Here I argue for antirealism in astrophysics, although not for any particular kind of antirealism. The argument is conducted by a detailed examination of some current research. It parallels the last chapter of (1983). Both represent the meth- odological opinion that abstract or semantic realism/antirealism debates are empty, and typically lead to confused or wrong conclusions because they pay so little attention to the details of a science.

1. Introduction. My Representing and Intervening (1983) concluded with an experimental argument for scientific realism about entities that cannot in any literal sense be observed. It was based on the fact that we regularly use such entities to investigate other parts of nature, and that we reliably build apparatus to take advantage of some of the causal properties of such entities. Evidently the argument cannot apply to unobservable entities that are postulated to exist outside our galaxy. There might, however, be some other compelling argument about scientific realism for extragalactic entities. In this paper I defend a modest astrophysical antirealism. The ar-

*Received June 1987; revised January 1988. tThis research was done at the Institute for Advanced Study, Princeton, where the author

was supported by an Isaac Walton Killam Research Fellowship (Canada Council), supplemented by funds from the Henry Luce Foundation (I.A.S.). The paper was presented at the "Newton and Scientific Realism" conference, Van Leer Institute, Jerusalem, April 27-30, 1987.

Philosophy of Science, 56 (1989) pp. 555-581. Copyright e 1989 by the Philosophy of Science Association.

555

556 IAN HACKING

gument is conducted by a quite detailed examination of one current body of research. It concerns a newly detected kind of entity, the gravitational lens. Gravitational lensing is a relativistic phenomenon, although it seems first to have been suggested by a Newtonian in 1919. The first strong contender for a gravitational lens system was reported in 1979. Since then the field has been very active. Black holes and superconducting cosmic strings are, of course, even less amenable in principle to "observation" and "experiment" (no matter how in the future we extend our use of those terms) than are gravitational lenses. But lensing has several advantages for philosophical scrutiny of today's real-life observational and theoretical cosmology. First, the fundamental physical idea appears to be very simple. Secondly, there is widespread conviction that we now know the lo- cation of a number of lens systems, so that individual examples of these entities are not merely postulated, but actually given geographical coor- dinates. Fascinating though strings and holes may be, the unfanciful gravitational lenses provide a more serious test of an antirealism that professes something more than merely a foundation in semantics and the philosophy of language, an antirealism, in short, tied to present scientific practice.

2. Gravity Lenses, What. I shall proceed as follows. In sections 3-7 I describe a philosophical problem about scientific realism in extragalactic astrophysics. In sections 8-17 I present the gravitational lens effect and its history from 1919 to the end of 1986. Then in section 18 I return to the philosophical concerns, armed with this new and I hope rich example. Readers unfamiliar with the effect will want, however, to know at once what a gravitational lens is. At first the answer is easy. Imagine a light source (radio source or whatever) at a very great distance from us. Be- tween this source S and the observer 0 is a massive object L. Suppose that L is very slightly off the line of sight OS. Then radiation from S grazing L on either side will be deflected by the mass of L so that 0 will observe S from two directions, seeing two images of S as in Figure 1. If L were exactly on the line of sight OS, then S could be imaged as some sort of luminous circle around L. Not only would one see two (or more images) or a halo, but also there will be magnification. That is, the flux, or rate, of light (or radio or x-ray) energy from S entering a cross-section of the observer's "telescope" will be substantially greater than would be the case if L were not in the way and S were observed directly.

Under these circumstances SLO constitutes a gravitational lens system and L is the gravitational lens. A glance at Figure 1, combined with the predicted magnification, explains the metaphor of lensing. I should warn at once that things will not prove so simple as this introduction. In Figure 1 we are really thinking of a point mass L. When correctly treated as a sphere, there must always be an odd number of images. Yet almost all,

THE CASE OF GRAVITATIONAL LENSING 557

Image of S S Image of S

L

0

Figure 1. Simplest model of a gravitational lens system. The lens is L, slightly off the line of sight of a source S and observer 0. Light from S is deflected as it passes L, creating two images and substantial amplification. This model is not even approximately sound unless L can be treated as a point mass.

558 IAN HACKING

or perhaps all, of known lenses appear to have an even number of images. Theoretical prediction for more complex lens systems is terrible: for clumpy universes, for example, or for objects with singularities such as cosmic strings.

Everything in Figure 1 could happen in our galaxy. First thoughts about gravitational lenses, including those of Einstein, did tend to think in galactic terms. But present candidates for lens systems are extragalactic. There are two good reasons for this. The lenses themselves are galaxies, which are massive enough to serve as lenses, and plentiful enough so that with non-negligible probability they will be on the line of sight to a distant source. Secondly, the present known sources are quasi-stellar objects, which, at the immense distances required, are bright enough to be observed.

Incidentally I shall use the name "quasi-stellar object" or QSO through- out (except when quoting) rather than the "quasar" of popular and text- book writing. The latter name is reifying for it suggests that there is one natural kind of thing-that for example there are (a) quasars and (b) pul- sars, two natural kinds. QSOs are incredibly bright distant blue star-like objects originally noticed in radar sources, and first dubbed "quasi-stellar radio sources" or QSRS, hence "quasars". But we now know that they (they?) need not have detectable radio emission; at any rate, in many it is very faint. They are objects but not primarily radio sources. They might not be a natural kind at all, but several completely different kinds of object. Conceivably they might not be anything: within a year of the discovery of QSOs, there began a long sequence of papers by the Bar- nothys (for example, J. M. Barnothy 1965; J. M. Barnothy and M. F. Bamothy 1968, 1983) urging that QSOs are "mirages" produced by none other than the gravitational lens effect! That's probably a mistake (de Silva 1970; Sanitt 1971).

In this paper I wish entirely to evade questions about QSOs. But it happens that all our probable lens systems use a QSO as a source. Hence I should say how many QSOs have been detected. Hewitt and Burbidge (1987) list 3594. At the beginning of their catalog they note that the question, "how many QSOs?" is not a precise one. QSOs are "star-like" and have a "very large redshift". Some definition. These cataloguers intro- duce what they call precise "operational" (that is, nontheoretical) criteria to obtain their 3594 QSOs. For comparison, the category of QSO was first used in 1963. A 1977 catalog lists 637; one for 1980 has 1549; and in 1985, 2835 were compiled.

We can in a fairly natural sense "observe" QSOs-even if we don't know what they are, and even if "quasi-stellar object" turns out to denote several fundamentally different kinds of object. In the case of lenses, at most we can see multiple images and deduce that they are images of the


same object, and we can look for a massive object in the right place that would do the lensing. We may be able to observe an object A, but we can't observe that this object is a gravitational lens. One can only infer that from observations, where by "observation" I hope it is by now self- evident that I don't mean just looking at the night sky with the naked eye or even looking into a telescope. Certainly we cannot interfere with a gravitational lens system (except by choosing to look at its effects). More- over, the remark with which I conclude section 15 shows that even in the case of the handful of lens systems known at present, the lens itself may be primarily dark matter, by definition matter that cannot be observed. Finally, as we see in section 17, our whole vision of the distant reaches of space may be extraordinarily misleading thanks to the action of undetected gravitational lensing objects. These objects may be "in principle" detectable-many are, after all, just stars. But they have not been detected. Many cannot be detected with present instruments because they are too dim. It is entirely possible that a system of minilenses can be studied only as a statistical ensemble whose individual elements will not be detected at all.

In short we have no need for philosophical quibbling about what is observable and what observation is. Wherever one draws the line, some body of lenses oversteps it.

3. Astronomy and the Experimental Method. Only one commonplace philosophical thesis about science is inculcated in grammar school: the scientific method is the experimental method. The Baconian adage has, however, one intrinsic problem. Two of our earliest sciences were geometry and astronomy. Take geometry first. In the Babylonian era it may have been an empirical science, but we are taught that after Thales it became a matter not of experiment but of deduction. Doesn't that create a problem for identifying scientific method and experimental method? No, because the very identification of science and experiment-one that arose only in the seventeenth century-by now takes for granted that geometry is not a science in the relevant sense of the term. On the other hand, if we do, in a post-Riemannian world, understand the propositions of geometry to be about our actual space, and we deny that geometry is a matter of convention, then geometry so understood is subject to experimental enquiry and hence no counterexample. The case of astronomy has no such swift resolution. True, we are already able to experiment on the moon and the planets which are close, but we cannot experiment on or with the sun. Galactic experimentation is science fiction, while extragalactic experimentation is a bad joke. Yet the study of the sun, the galaxy and the universe is assuredly science. Hence the scientific method cannot be the experimental method.

560 IAN HACKING

This is a fine poser for the alert child to put to the schoolteacher, but no one worries much about it. One element of the experimental method is observation, is it not? In astronomy we are largely restricted to observation, so our science is in certain ways truncated. But that is a fact of life, a fact about our size and powers, as opposed to the size and powers of the sun, the galaxy or the universe.

A school teacher as alert as the pupil may even respond with a historical message. In the beginning of Western science there was (aside from geometry, which went its own way) only observation. Only the solar system was obliging enough to suggest to us firm quantitative regularities that gave us the concept of a law of nature. Only in astronomy founded upon observation were we able to make strikingly precise predictions of phenomena and ephemera. Then in the seventeenth century a long tradition of merely empirical tinkering with nature closer to home was transformed into that experimental method whereby we create phenomena about which we can frame laws as exact as those hitherto known only for astronomy.

Astronomy, according to this reply of the schoolteacher, is not a counterexample to the identity of scientific method and experimental method. It just happens that we are able to study the nature of the macrocosmos without interfering with it, simply by observing it. Indeed were this not the case, we would probably never have attained to our conception of law of nature at all. The full experimental method of intervention in nature is needed for the microcosmos. Doubtless we would learn more about our galaxy were we able to explore it and manipulate it. We certainly will learn more about our solar system by exploring it, and experimenting on bits of it just as we now experiment on rocks from the moon. But it happens that in astronomy we can get by without the full experimental method engineered into being in the seventeenth century.

4. Astrophysics and Scientific Realism about Entities. The nebulous worry of section 3 is more serious for me. I take a somewhat Duhemian view of theories, holding them to be possible representations that are not, and perhaps could not be, literally and exactly true. But I am realistic about certain unobservable entities. Electrons do exist, I argue, even though we may be unable to give true descriptions of them over and above a purely phenomenological level. Naturally, such a point of view cannot be combined with a Russellian theory of denotation in terms of descriptions. It rides easily, however, with the sort of theory of reference advanced by Putnam.

I did not say in my (1983) that an experimental argument is the only viable argument for scientific realism about unobserved entities. I said only that it is the most compelling one, and perhaps the only compelling


one. On my last page I concluded with a deliberately flippant remark about black holes:

I must confess a certain scepticism, about, say, black holes. I suspect there might be another representation of the universe, equally consistent with phenomena, in which black holes are precluded. I inherit from Leibniz a certain distaste for occult powers. Recall how he in- veighed against Newtonian gravity as occult. It took two centuries to show he was right. Newton's aether was also excellently occult. It taught us lots. Maxwell did his electromagnetic waves in aether and Hertz confirmed the aether by demonstrating the existence of radio waves. Michelson figured out a way to interact with the aether. He thought his experiment confirmed Stokes's aether drag theory, but in the end it was one of many things that made aether give up the ghost. The sceptic like myself has a slender induction. Long-lived theoretical entities, which don't end up being manipulated, commonly turn out to have been wonderful mistakes. (Hacking 1983)

1 was not speaking about arbitrary cosmological objects. A black hole is as theoretical an entity as could be. Moreover, it is in principle unobservable. (Well, maybe there are problems about cosmic censorship, but let that be.) At best we can interpret various phenomena as being due to the existence of black holes.

The putative counterexample to the Baconian identity "scientific method - experimental method" was, "there's no experimentation in astronomy". The school teacher's reply was: "There's observation, not experimentation, and there is an instructive story to tell about that". But now I have switched from astronomy to extragalactic astrophysics. Here we have entities, postulated by splendid theories, and which cannot be observed. That's a poser even for the school teacher. But it is worse for me. Do I really mean that, as part of my experimental philosophy of science, there can never be evidence for the existence of black holes? At the very least I need to articulate my position!

5. Why Use Gravitational Lensing as an Example? Because I want to take a case that is as unfavorable to my ideas as possible. Black holes may in principle be unobservable but we cannot with any confidence point to any region of the sky and say, there's one there, even if we suspect that in fact indefinitely many are there. But we really can point to a few regions of the sky, and say with some confidence, there is a gravitational lens system there. Yet I am very disinclined to say that we can observe the lens system. I do not propose now to go into the semantics of that open-textured verb, "observe", but I would continue along lines set forth in chapters 10 and 11 of my book. The more valuable, if more demand-

562 IAN HACKING

ing, exercise is to consider scientific realism about entities in the context of quite specific instances of a kind of entity that we cannot observe, and with which we cannot interfere.

6. Gravitational Lenses Not Idle Curiosities. Postulation of black holes and cosmic strings arises from profound theorizing. Gravitational lensing looks like a trivial corollary of more than one theory about gravitation. But gravitational lensing "provides a potentially powerful tool for inves- tigating two of the outstanding problems in astronomy: the distribution of mass in galaxies and clusters of galaxies, and the values of three fundamental cosmological parameters", namely, the Hubble constant (rate of expansion of the universe), the rate of change of expansion rate with time, and the cosmological constant, which defines the mass-energy density of the vacuum (Lawrence et al. 1984).

These are grand hopes, although not everyone is sanguine that they will be fulfilled. The idea of using gravitational lensing to determine the Hubble constant goes back to Refsdal (1964b, 1966). His model, which supposes a lens to be point mass, is ideal for determining the Hubble constant, but it turns out to be unrealistic. As we shall see, the geometry of more realistic models may be so messy that details internal to the lens system may entirely dominate subsidiary effects due to the rate of expansion of the universe (Falco et al. 1985). There is another idea afoot, that lenses may enable one to investigate the character of the "missing" mas-s of the universe (Gott 1981). Finally, as I note in section 17, gravitational lenses may be populating the universe with artifacts. Whatever the upshot, lenses have captured the imagination of the astrophysical community.

7. A Shift in Philosophical Emphasis from Kinds of Entities to In- dividuals. Scientific realism about unobservable entities in physics has typically concerned itself with natural kinds: are there compelling arguments for thinking that electrons exist? Neutrinos? Quarks? We are concerned with sorts, not particulars. Even when we do have a way of uniquely designating a subatomic particle-the ax-partical that caused that click- we have no interest in that very individual.

There are of course counterexamples. Since 1981 the Penning trap has enabled one to "trap" individual atoms and see how they behave. Yet even here we do not care about this atom, but use it solely as a specimen of the type. In the celebrated events first observed on Earth on 23 Feb- ruary 1987, several neutrino detectors around the world showed events that we assert indicate the passage of some neutrinos that left the explo- sion we call Supernova 1987A or Supernova Sheldon some 160,000 years ago. Moreover we are speaking of a small number of events, 5 here, 11


there. Nor are all the neutrinos identical, so in a sense we do discuss particular, individual neutrinos whose history is conjectured for 160,000 years. Yet even here the interest lies only incidentally in the individuals that have been observed. We study them to answer questions about the type: what (for example) is the mass of the neutrino, if indeed it has a mass at all?

It is otherwise with gravitational lenses. The first probable gravitational lens system to be detected involved 0957 + 561 A, B. A and B might be twin quasi-stellar objects, but it looks as if A and B are identical, two images of the same object created by a lens system. Shades of Frege's example of Hesperus and Phosphorus! Are A and B identical? Does that theoretical object, a gravitational lens system, exist here, with the consequence that A = B? It should be instructive to have a real-life example in which we pass from traditional concern with kinds of unobservable entity, where we can never interestingly single out an individual, to a case of a specific inferred but unobservable individual entity with its own idiosyncratic properties. What reasons can we have for thinking that right here at 0957 + 561 we have an instance of a conjectured kind of entity, a gravitational lens?

8. Newton: Hello and Goodbye. A "Newtonian", Sir Oliver Lodge, foresaw the possibility of gravitational lensing (Lodge 1919). It was in the context of the first news of the famous Eddington-sponsored expe- dition to test the general theory of relativity. Lodge notes "that a gravitational force acting obliquely on light would probably be unable to alter speed, but through the co-operation of its transverse and longitudinal components, it might be expected to produce an extra dose of deflection-assuming light to be subject to gravitation, as Newton surmised" (my emphasis). He does not favor the lens metaphor:

[I]t is not permissible to say that the solar gravitational field acts like a lens, for it has no focal length. If the sun were backed by a nebula or any luminous area, the light grazing the rim all round would be brought to a focus at a place seventeen times the distance of Neptune, while light from any larger circle would focus still further off in pro- portion to the area of the circle. So from a uniformly luminous area there would result a focal line of constant brightness. (Lodge 1919, p. 354)

I shall not develop Lodge's reasoning, which is a rearguard action, no doubt. "Dynamics have served us so well in the past that it must be still legitimate to try, wherever possible, to apply well-established principles to new phenomena". Eddington had more to say about such quasi-lenses the next year (Eddington 1920, p. 133). Searches for such lens effects

564 IAN HACKING

among stars began immediately at the Yerkes Observatory (Zwicky 1937b, n. 3), but fizzled out.

9. Einstein: Hello and Almost Goodbye. "Some time ago, R. W. Mandl paid me a visit and asked me to publish the results of a little calculation, which I had made at his request" (Einstein 1936). Thus was Einstein induced to write about the "lens-like action of a star by the deviation of light in the gravitational field". He considers the light from a star S trav- ersing the gravitational field of another star L whose radius is r. Let (x be the angle of the deflection of a ray of light from S grazing L at distance r from the center of L. (According to theory, a- = const./r, where const. = 4 x gravitational constant X mass of L.) Then an observer at distance d from L, and exactly on the extension of the line SL will not see a point- like star L, but L surrounded by a halo of angular radius or/d. If the line of sight is slightly off the extension of SL, the source S is imaged as two point-like sources separated by about the same angle. The lens- like effect of L increases the apparent brightness of S by roughly oxd/x where x is the distance at L of the line of sight from SL, and is small compared to ioxrd.

Einstein noted that when S is directly behind L, one would not rec- ognize S, for L would merely look brighter than it "really" is. Likewise, x must be so small that the two images would not be distinguished by "the resolving power of our instruments". "Therefore, there is no great chance of observing this phenomenon, even if the dazzling effect of the much nearer star L is neglected".

What struck Einstein as interesting about this (undetectable) effect is that the increase in brightness of the source S does not decrease with increasing distance d of the observer, but actually increases as \/d. Other cosmologists expressed deep regret that the "most perfect tests of general relativity furnished by the gravitational lens effect are unavailable" (Rus- sell 1937). Einstein had no regrets. One speculates that he was delighted by this phenomenon of a magnification that increases as \/d. Here is a fundamental phenomenon that is known but unobservable.

Einstein's comparative indifference to experimental investigation of the general theory may be further illustrated by the fact that he did not make another calculation. R. W. Mandl put the question of lensing not only to Einstein but also to V. K. Zworykin, who in turn aroused F. Zwicky's interest (Zwicky 1937a). Zwicky very easily showed that although there may not be much hope of visible lensing in our galaxy, extragalactic nebulae are much more promising. Thinking in terms of nebulae of magnitude greater than 21.5, and using the results of the best exposures with the 100-inch telescope, he showed that one should observe the halo of a


source S once in every 400 nebulae; allowing for magnified brightness due to lensing, which would allow one to observe distant nebulae not otherwise observable, one might (assuming uniform distribution of nebulae) expect that as many as 1 percent of all nebulae were gravitational lenses. The chief application foreseen by Zwicky was the determination of the masses of lensing nebulae. Incidentally, I think we owe the name "gravitational lens" to Zwicky; he uses it in quotation marks in (1937a), and without them in (1937b).

If Zwicky had been simply and straightforwardly right, gravitational lenses should have been known in abundance reasonably soon after 1945. They weren't.

I titled this section, "Einstein: Hello and Almost Goodbye". He thought that the universe is full of gravitational lensing by stars, but that we would almost certainly never detect a single instance in the form of split images or luminous circles. This thought was long dismissed as either wrong or irrelevant, although sheer failure to detect lenses was a bit problematic. Einstein's discarded thoughts have a habit of returning in new guise: see section 17 below, where my references are chiefly to articles published in 1986 through January 1987.

10. Detecting Lenses. Refsdal (1964a) brought new life to lensing with a very clear review analysis of the gravitational lens effect. His opening paragraph ends with a sentence of interest to the philosopher of experiment: "Due to progress in experimental technique we find, contrary to Einstein, that the [gravitational lens] effect may be of practical interest" (my emphasis). He computes the probability of occurrence of a detectable gravitational lens and finds it substantial. Thinking of the passage of one distant object in front of another very distant object, he writes that "It seems safe to conclude that passages observable from the Earth occur rather frequently. The problem is to find where and when the passages take place" (p. 306).

I have already drawn attention to the most dramatic proposal about gravitational lensing, namely the Barnothys' idea that the quasi-stellar objects are mirages (imaged by Seyfert galaxies). Less iconoclastic con- jectures also failed. One was reported in 1974, a pair of quasars, 1548 + 115 A, B (Gott and Gunn 1974). The prima facie evidence lay in the similarity of the spectrum of two nearby objects A and B. There were problems in providing an adequate detailed lens model for this couple (Bourassa and Kantowski 1976). The first lens system to have stood the tests of time was that of Walsh et al. (1979). Some titles of successive reports on this and subsequent images are instructive. I shall number the lenses in order of publication, although it is probable that at least one of these will not survive as a lens-like object system at all. Note how we

566 IAN HACKING

pass from titles ending in "?' to "probable" to "discovery".

Lens I: "0957 + 561 A, B: Twin quasistellar objects or gravitational lens?" (Walsh et al. 1979). "The double quasar Q0957 + 561 A, B: A gravitational lens image formed by a galaxy at z = 0.39" (Young, Gunn et al. 1980). [z is a redshift.] "Q0957 + 561: Detailed models of the gravitational lens effect" (Young, Gunn et al. 1981). "Q0957 + 561: Effects of random stars on the gravitational lens" (Young 1981).

Lens II: "The triple QSO PG1 115 + 08: Another probable gravitational lens" (Weymann et al. 1980). "The triple quasar QI15 + 080 A, B, C: A quintuple gravitational lens?" (Young, Deverill et al. 1981).

Lens III: "Discovery of a third gravitational lens" (Weedman et al. 1982) [2345 + 007 A, B].

Lens IV: "Discovery of a new gravitational lens system" (Lawrence et al. 1984) [A triple radio source MG2016 + 112].

Lens V: "Discovery of a new gravitational lens" (Djorgovski and Spinrad 1984) [QSO 1635 + 267 A, B].

Lens VI: "2237 + 0.35: A new and unusual gravitational lens" (Huchra et al. 1985).

There have been other suggested lenses (Paczyn'ski and Gorski 1981) which have not survived further scrutiny. But the sequence of titles displays a growing confidence in the candidates for gravitational lenses. In "Ob- servations of gravitational lenses" Walsh (1983) provides a survey of results to date (Lenses I-III), wryly noting that in only one case (Lens I) is there a candidate for the massive object that is doing the lensing, "so that in only one case, 0957 + 561, has the lens been observed (notwith- standing the title of this paper)". These words seem contrary to my sug- gestion that one does not "observe" lenses, for Walsh is saying that we have done so. But what we have observed is a very large number of galaxies, most of which are, under some model, consistent with the observations and the hypothesis of some gravitational lens effect or other.

11. Why Do You Think It is a Gravitational Lens? Lens system VI probably consists of a nearby galaxy almost directly in front of a QSO, but in the preceding cases there were two or more apparent objects. It was conjectured that these are images of a single object. An advantage of gravity lensing as an example for a general audience is that the reasoning behind the conjecture is easy to follow. I take Lens I as typical, at least from the point of view of scientific method (Walsh et al. 1979).

The quasi-stellar objects of Lens I were identified by the Jodrell Bank


radio survey. They were optically identified by people from Jodrell Bank, using the Kitt Peak facilities in Arizona, March 29, 30, and April 1, 1979. The arguments for identifying A and B as images of a single quasi- stellar object are as follows. First, A and B are close, 5.7 arc seconds apart. Secondly, they have the same magnitude (17) and almost identical redshifts of about 1.40. What is fundamental, however, is that they have "almost identical spectra" (this becomes simply "identical" in later lit- erature). In particular, emission and absorption lines of particular elements match strikingly well. In the blue part of the spectrum, A is brighter than B by about 1/3 magnitude, while in the redder parts they are about equal.

The observers then reason as follows.

(a) The great similarity in spectra of the two very close QSOs with the same redshift "seems to constitute overwhelming evidence that the two are physically associated, regardless of the nature of their redshifts, and we do not think that a useful a posteriori statistical test of this assertion can be carried out". They then assume that the redshift is in fact cosmological (that is, due to universal expansion and hence a measure of distance, and not, for example, a gravitational redshift). More recent work suggests that similarity of spectra is less compelling than at first thought. But there is more.

(b) Noting that (a) suggests two images of the same object, the team first examines "the more conventional explanation involving two distinct QSOs". This requires a large number of coincidences. Either the emission spectra are the same by coincidence, or else the two QSOs had the same initial conditions and the same subsequent history. Moreover the near identity of absorption lines is more mysterious still "regardless of the mechanism invoked to explain absorption". Absorption lines in QSO spectra are ill understood, and a whole bunch of mechanisms have been suggested. But on any plausible model at present on the cards, it is very unlikely that two objects so close would have the same absorption lines.

(c) Then the lens hypothesis is considered. The relevant factors are: angular separation of the images, their shapes and sizes, and their amplification. Everything fits such a model. In particular, the fact that one image ought to be brighter than the other in an equivalent cosmological setting was already known, although fully demonstrated only in Schneider (1984). If we use the magnification of A compared to B to estimate the luminosity of the source, the postulated source has luminosity typical for a QSO.

(d) (Half a sentence, in passing: There is no strong C IV emission line of the sort to be expected from a strongly magnified Seyfert galaxy nu- cleus. The implication is that we need not trouble ourselves with the Bar-

568 IAN HACKING

nothy hypothesis mention above in section 2.) (e) There's a missing link. In (c) a source QSO was proposed to ex-

plain the phenomena. But where is the lens L? Only a probabilistic calculation is offered. If we have a lens system, then there should be a galaxy of such and such a mass, a galaxy from which A + B is very distant, and which lies close to the line of sight to A + B. A calculation on the relative frequency of massive elliptical galaxies of the right size cannot yield any precise number but, "while such coincidences [alignment, mass, etc.] must be very rare, it is not out of the question that we should have one example in the -1,000 QSOs known".

12. Inference to the Best Explanation. This looks like a classic example of what philosophers, following Harman (1965) have come to call inference to the best explanation. Philosophers divide on the merits of such a mode of inference as a ground for scientific realism, with Salmon (1984) in favor of it and van Fraassen (1980) against. But Walsh et al. are not arguing about scientific realism. They are comparing the merits of two rival hypotheses (with a half-sentence dismissal of a third rival, the Barnothy hypothesis). Contrary to Harman's original scheme of inference to the best explanation, we have no prior probabilities of the two hypotheses, A ? B vs. A = B. We have only the following: if the former is true we have some unquantifiable but extremely unlikely coincidence. If the latter is true, a conjuncture of only modest improbability has been observed. I shall return to the question of probabilities in section 16.

Note that unlike Salmon's use of inference to the best explanation in arguments for scientific realism, we have two seemingly quite clear physical or rather astrophysical hypotheses. It is not like comparing a precise hypothesis against "sheer cosmic coincidence". The two-quasi-stellar-objects hypothesis is deemed to demand an extremely improbable and mean- ingless set of coincidences. The hypothesis of one double-imaged source is deemed to demand a not very uncommon arrangement in the heavens, an arrangement that explains the observed phenomena.

Inference to the best explanation is often traced back to C. S. Peirce and what he called the method of hypothesis (and also retroduction, ab- duction). In maturity Peirce did not think that the method of hypothesis gave ground for belief, but only for further investigation, testing, and induction. Unknowingly writing in the spirit of Peirce, Walsh et al. note that "further observations would shed light on the gravitational lens hypothesis". They suggest, for example, that if the conjectured source has variable flux, we would see similar curves for A and B but with a time lag due to the different path lengths. Time lags are a matter of greatest interest, for people thought that one ought to be able to determine the Hubble constant from the time lag and hence estimate the rate of expan-


sion of the universe. Naturally more commonplace investigations are also in order, namely the radio spectrum, and for some theorists, this has been the decisive factor.

13. Observational Confirmation of Lens I. After sixty years of off- and-on predictions of gravitational lensing, a real live candidate is quite exciting. A spate of checks followed. Somewhat confusing studies of radio spectra were immediate (Greenfield et al. 1980; Pooley et al. 1979; Porcas et al. 1979; Roberts et al. 1979). So were initial reports of the possible lensing object itself (Adams and Boronson 1979). A more ex- tensive search of the field of the QSO for a lensing object used a series of deep red pictures and revealed a very large cluster of galaxies (Young, Gunn et al. 1980). Young's group used a standard galaxy model and noted that a very large number of configurations of the galaxy would produce phenomena like those discovered in 1979-1980. "What does emerge clearly is that the galaxy-cluster combination should make multiple images, and that there are several plausible ways to reproduce the observations. We regard the case for some of these to be very good." All the models involve some sort of galaxy-cluster, not some thing that can be represented by a point mass gravitational lens. This enables one to understand the confusing results of radio maps of A + B, for the imaging will be complex, and there is the possibility that the imaging galaxy is itself a radio source.

This investigation was continued (Young, Gunn et al. 1981). A catalog of 146 galaxies and 32 stars around A + B went as far as one could go, namely magnitude 24.5. The position of the prime lensing galaxy was determined more accurately by Stockton (1980) and the infra-red structure by Soifer et al. (1980). For a thorough survey, see Walsh (1983).

14. A Happy Ending Good for Traditional Philosophy of Science? I seem to have described an unusually felicitous example of scientific method. There is an overall conjecture of the sort that Popper would have called metaphysical: gravitational lenses exist. For sixty years, from 1919 to 1979, we lacked any instance of a gravitational lens. The conjecture re- mained unfalsifiable and therefore "metaphysical". Then in the spring of 1979 it was proposed that 0957 + 561 A, B is a gravitational lens. This was argued by an inference to the best explanation, that is, it provided ground for further inquiry. The conjecture entailed that there is a lens to be found. It also implied that there is variability in the ratio of flux between A and B. Intensive study corroborated these predictions. The existence of a gravitational lens was confirmed and the hunt was on for more.

570 IAN HACKING

Unfortunately, as we all know, things are never that simple in the fore- front of a scientific inquiry.

15. Doubts: The Odd or Even Number of Images. I have, for practical reasons, chiefly addressed observational practice and issues. The best theoretical review essay is that of Peacock (1983). In the next few sections of this paper I shall address a very few of the problems that arise when theoretical modeling clashes with observation. One concerns the number of images.

If, as in Figure 1, the lensing object is treated as a point mass, it readily follows that a source will have two images that merge to one as alignment of source, lens and observer increases. Unfortunately computer simula- tions of rays in a lensing galaxy show that the point mass model won't do (Bourassa and Kantowski 1976). We need to be more realistic and regard the galaxy as a distribution mass.

It was then observed that for a transparent spherical mass distribution, one would always obtain an odd number of images (Dyer and Roeder 1980). "Transparent" here means that almost all rays from the source pass through the lens. Thus a solid object (a star) or a black hole is excluded. This result readily generalizes to any detectable arbitrary mass distribution in a transparent galaxy (Burke 1981). Here "detectable" means that the lens is not so massive that it is within its own event horizon. The argument is an elementary topological one. The conditions for a ray from S to be deflected by L and perceived at 0 are that the difference between the radial vectors SL1 and L1O should be zero. This occurs only for saddle points (which give "inverted" images), and peaks (maxima or minima), and there must always be one more of the latter than the former.

Unfortunately the lenses I-V show an even number of images. Young, Gunn et al. (1980, 1981) went to work on Lens I, trying to answer their question, "Where then is the third image?" (1981, p. 727). In one of the models proposed in their (1980) they had shown that a five-image model was possible, but could not define a two-image model. No one is satisfied with any of the candidates for a third image, such as proposed in Gor- enstein et al. (1983).

Here then we have a characteristic "anomaly". What is going on? Here are a few suggestions (Blandford and Narayan 1986). (1) The missing images are too faint to be detected. (2) The lensing potentials may be far greater than those of ordinary galaxies, so that the two observed images are relatively close together, while other images are separated from these by a great angular distance, hence not deemed to be candidates for multiple imaging. (3) The galaxies may not be entirely transparent, with some rays being absorbed by passing through the center of the galaxy, or actually being impeded by stars in the center of the galaxy. (4) The compact cores of galaxy potentials deamplify the missing images to the point of


undetectability. (5) Our lens may be a being impeded by stars in the center of the galaxy. (6) Our lens may be a singularity, for example, a black hole or a cosmic string. With respect to this final case, Blandford and Narayan show that "a point singularity (for example, a black hole) can sometimes be observationally distinguished from one with a long singularity (for example, a cosmic string) using the parities of the observed images". At present, none of these six possibilities, which all break down into subpossibilities, is compelling.

On the basis of the remark by Blandford and Narayan just quoted, I thought to title this paper "How to Tell a Black Hole from a Cosmic String Using a Gravitational Lens". Unfortunately at present that would be to write science fiction. From 1980 it has been suggested that some multiple QSOs might be the consequence of lensing by cosmic strings. An excellent theoretical exposition is Hogan and Narayan (1984). Pac- zynski (1986a, b) suggests several QSO images as candidates. Blandford, Phinney and Narayan (1987) shoot down the most promising of these candidates (1146 + 111 B, C).

The odd-number anomaly is a readily understood conflict between observation and theory. There are others of a more technical sort. One is that despite the plausible success of finding the actual lensing object in the case of Lens I, a good lens is hard to find. Remember, what we have is a few images, and we look for an object in the region that will serve as a lens. There are theoretical constraints on the objects. The image separation of the present proposed lenses suggests high velocity disper- sion galaxies, but in general deep searches in the vicinity of the images produce no such objects. Hence one is led to think that dark matter may be primarily responsible for lensing.

If that were the case, then indeed most of our present proposed lenses are inferred, not observed!

16. Doubts: The Scarcity of Lenses. Statistical considerations enter our story in several ways. There are tests of significance of the hypothesis that A, B is a pair of images, against the "null hypothesis" of a pair of similar QSOs. There is statistical treatment of lensing discussed in section 17 below. Then there is the fact that lenses are too infrequent. With the notable exception of Einstein, everyone from Zwicky on to the present has thought that gravitational lenses ought to be quite easy to find, and they certainly are not. This casts doubt at least upon our specific modeling. Hacking (forthcoming) discusses all these probabilistic arguments, and provides further information on the unexpected rarity of lens systems.

17. Microlensing: Are Little Lenses Flies in the Astronomical Soup? Gravitational lenses have been getting bigger and bigger: Lodge's stars, Zwicky's nebulae, and then galaxies. So it is fitting that I conclude

572 IAN HACKING

this story of lensing by going back, if not to stars, to very compact objects. These enter as "minilenses" or "microlenses".

I shall merely suggest some of the ramifications of this idea. First, our problems with big lenses may in part be caused by their structure. It may be that the stars in the lensing galaxy, when close to a ray which reaches us, have a lensing effect upon that ray. This was in fact one of the first suggestions made in connection with our old friend Lens I. The title of Chang and Refsdal (1979) is self explanatory: "Flux variations of QSO 0957 + 561 A, B: Image Splitting by Stars Near the Light Path". The authors argue that a single star could split a QSO image into two or four images with an angular separation of 10-5 arc seconds. The star is a point mass, so the even number is all right, and such splitting, coming in even numbers, will not affect the odd-number theorem.

More perturbing, in both senses of the word, is the possibility that minilenses are thoroughly distorting our view of distant objects. Einstein noted that even a small lens would be expected to have substantial amplification effects. In particular, it might magnify an object, even though it was impossible to resolve the several images of the object being produced by the lens. Thus an object lensed in this way would appear closer or brighter than it really is. Moreover a star in passage close to an object would cause an apparent variation in this object.

An example is a class of objects only fairly recently investigated, the BL Lacentae, originally not distinguished from quasars. The Hewitt and Burbidge (1987) quasar catalog is supplemented by a list of 87 BL Lac objects. These are extragalactic highly variable polarized sources with significant emission all the way from the radio spectrum to X-rays. Note the "highly variable" in this definition. Is the great variability an artifact of gravitational lensing? This is proposed by Ostriker and Vietri (1985), who suggest that at least some BL Lac objects are QSOs, admittedly odd ones ("optically violent variable quasars"). They present a model that is consistent with this interpretation for a substantial number of the better- studied BL Lacentae. (But see Schneider 1986 for a modification of this calculation.)

The chasm of uncertainty is yawning before us. In general there is no way that we are going to see the stars that are doing the conjectured lensing. Their magnitude will be very much less than the QSO, BL Lac or whatever it is, and characteristically beyond the range of detection, even with Space Telescope.

Moreover, what about standard QSOs? We are very interested in the evolution of quasars, or what appears to be the evolution of quasars. As always, when no one has much of an idea of what is going on, we resort to statistics, attempting to discern some regularities in the evolution of quasars as an ensemble. But suppose most of this evolution is the con-


sequence of passages of undetected stars, with consequent amplification changes without discernible splitting of images? (Turner 1980; Peacock 1982). In section 16 I was lamenting the fact that there are not enough lenses. Suddenly we are battling a seething mass of undetected and perhaps undetectable minilenses.

Take another oddity. QSOs apparently like to hang out around galaxies. Is there some physical reason for this? Or perhaps here is another illusion. Consider two QSOs of a magnitude at present not detectable on its own. One is near a galaxy. It may be magnified by the galaxy, or rather by a star or successively by several stars in the galaxy (for we are considering an example in which no split images are discernible). So we may observe this QSO and not the other, hence explaining "quasar-galaxy association" (Schneider 1986). We shall doubtless hear a number of other phenomena "explained away" in this manner (for example, Schneider and Weiss 1987). And of course there will regularly be questions about alternative causes of image distortion, for example, "Image Distortion Near Galaxies: Dwarfs or Lensing?" (Tyson 1985).

I just said that a QSO might be successively lensed by several stars. First thought on this topic confined itself to one compact object, but stars are plentiful as blackberries. So the next question to address is necessarily a statistical one. Given certain assumptions about the distribution in depth of stars, what in general are the gravitational microlensing effects that are to be expected? (Paczyn'ski 1986c).

Paczyn'ski does not strictly answer this question. What he does is to provide a great many new models of gravitational microlensing. I next turn to the concept of modeling and hence to the philosophy of scientific realism. Before doing so, we should recall Einstein and section 9 above. He held that the universe was full of gravitational lensing which we would never detect in the form of luminous halos or split images. He does have a habit of having the last word.

18. Models. Philosophers say precious little about models. Cartwright (1983) has perhaps the most extended discussion, and, following her, there is a brief discussion in my (1983, chap. 12). The latter uses "theories, models and phenomena" as one possible framework in which to think about natural science. It notes a simple reason for not regarding models as true-of-the-world, namely, that it is absolutely standard practice to use a number of different models in solving the same class of problems, even though the models are formally inconsistent with each other, if taken to be making literal assertions about the world (p. 216). In short, one must be anti-realist about models, which is not to say that "anything goes".

I suspect that there is no branch of natural science in which modeling

574 IAN HACKING

is more endemic than astrophysics, nor one in which at every level modeling is more central. We do have what in a rather strict sense of the word are properly called theories. We have general relativity. We have imported almost all of the theories of particle physics, in the way made celebrated by Weinberg (1972). Now of course the word "theory" is as open-textured as you will, and I could cite a goodly number of tokens of the words "the theory of the gravitational lens effect", for example. There is no grave impropriety in speaking of Friedmann's or Lemaitre's or Rob- ertson's or Walker's theory of the universe, but it is standard practice, and one reflecting good sense, to speak of Friedmann-Lemaitre models or Robertson-Walker models. Such is the practice of Eddington (1931), Weinberg (1972) and Peebles (1980), works that rightly set the standards in these matters. I follow their usage, as most writers do, when a theory/ model distinction could matter.

The topic of gravitational lensing is typical, in respect of models, of much astrophysics. We have a gamut of models of increasing level of specificity, starting with a model of the universe, and ending with a model of 0957 + 561 A, B. Better: we have on the shelf a number of models of the universe, and at the level of Lens I we have many local models. This talk of a lot of models should trigger any anti-realist instincts har- bored in the reader's soul.

Before proceeding to a particular lens, let us consider a stage of modeling intermediate between everything and one thing. A recent paper will serve: "Self-Consistent Probabilities for Gravitational Lensing in Inhom- ogeneous Universes" (Ehlers and Schneider 1986). This is not concerned with individual lenses, that is, not with single, isolated distributions of mass that do the lensing. It is rather in the tradition that considers the statistical distribution of lenses, alluded to in section 17. Far from being peculiar to the question of microlenses, it was discussed long before there was any observational point in discussing giant or tiny lenses (Dyer and Roeder 1972, 1973; Press and Gunn 1973). There is a problem about the distribution of lenses at large, succinctly stated by Ehlers and Schneider:

gravitational lensing takes place only in a universe in which matter is clumped. However, no solution to the equations of General Rel- ativity is known which is appropriate to describe such a clumpy universe. (1986, p. 57, col. 1)

Dyer and Roeder dealt with the difficulty in a way typical of applied physics. They invented a differential equation for the diameter of a bundle of light rays, using certain (by then) rather routine procedures connected with the Friedmann-Lemaitre model. Here are Ehlers and Schneider again, with their emphasis:


Although this description of light progagation in a clumpy universe has not been derived by a perturbation approach or otherwise from General Relativity, but is based in a number of more or less plausible ad hoc assumptions, these "Dyer-Roeder distances" are usually applied to the gravitational lens equation, due to the lack of any more founded theory. (1986, p. 57, col. 2)

Dyer and Roeder did what physicists who are not Diracs do: they pre- tended that on a large scale a clumpy universe can be described by a Friedmann-Lemaitre metric, which agrees almost nowhere with any clumpy universe. (Almost nowhere means probability 0 on any reasonable to- pology.)

Now Ehlers and Schneider are critical of Dyer-Roeder distances. As they see it, the Dyer-Roeder formulae are taken blindly, plugged into a gravitational lens equation, and then consequences are cranked out. But in effect the consequences combined with the original Dyer and Roeder assumptions are inconsistent. Ehlers and Schneider therefore make the next move of the working physicist. Take the assumptions of Dyer and Roeder and the gravitational lens equation, and find solutions that are consistent with both.

At any rate, the validity of the Dyer-Roeder equation for the prop- agation of light beams through a clumpy universe has not been established within General Relativity. The best one can achieve with this model is a self-consistent description of ray bundles. (1986, p. 59, col. 1)

The authors construct just such a set of formulae, and conclude that "the self-consistent equation shows that hitherto the effects of gravitational lenses have been underestimated" (1986, p. 61, col. 1). This assertion is part of a program of which we have seen instances in section 17, of studying the statistical distribution of lenses in the universe, with a view to understanding how they might be influencing our observations. But the revised equations for the light bundle in a lensing situation should apply not only to statistical problems but also to specific lenses.

Now let's move down to something more specific, our familiar Lens I. In sections 11 and 13 I traced the following story. First 0957 + 561 A, B was proposed as a pair of images in a lens system, on the basis of the extreme coincidence of emission and absorption lines. Then Young et al. (1980, 1981) studied this pair in depth. One project was to identify the lensing object. Although their conclusions have not gone unchal- lenged, they did produce a plausible set of objects for the lens. Now why do they think that this set, picked out of the swarm of objects, is the immediate part of space? Is it because this "galaxy cluster combination

576 IAN HACKING

should make multiple images, and . . . there are several plausible ways to reproduce the observations. We regard the case for some of these to be very good". That is the way it has to be, even in the most favored case of Lens I where we actually have a candidate for the lensing object. There are a number of models, mutually inconsistent with each other, and which collectively justify the assertion.

In constructing the family of possible models for the lenses proposed at present, one does not in general even use the differential equation for ray bundles derived by Dyer and Roeder, or by Ehlers and Schneider. One uses plain old geometrical optics, with a paragraph explaining why this, although false in space, makes things simple and is not too misleading.

Even in the present paper one will find other examples of multiple models, for example, the models that account for flux variation and time delay observations of Lens I. Thus we have model upon model upon model. More precisely: we have a Friedmann-Lemaitre model for the universe. We have a model of ray bundles, a differential equation con- structed along Friedmann-Lemaitre lines for a universe which is clumpy, that is, not F-L, so that the equation just isn't true. Finally we have fam- ilies of models for a particular lens system, models that are mutually inconsistent and which are derived by optics even more patently false than that of Dyer and Roeder, or Ehlers and Schneider.

When I say "false" here, I am not in any way intending to impugn the physicists. I am speaking from the standpoint of the scientific realism/ anti-realism debate among philosophers. In astrophysics we have only models, and all too many of them, at every possible layer of investigation. There are no propositions of detail, at the level of the science I have been describing, which are more than models. These models are not literally true. Nor are they "converging on the truth", for all we will ever have are more models. That is why I am an anti-realist about astrophysics. I stand by the concluding paragraph of Representing and Intervening, quoted in section 4 above.

19. "To Save the Phenomena". This famous expression, a translation from the Greek, has a literal meaning: to reconcile observed phenomena and a theory that they contradict. The reference was astronomy. The phenomena did not conform to motions calculated from celestial models. Phenomena and models were regularly reconciled by adding epicycles to the models.

During the seventeenth and eighteenth centuries the expression was to some extent adapted to other sciences, but with a change in verb. Making a pun on the Latin salve, one got "to solve the phenomena". Here "solve" retained the ancient sense of solving a problem in geometry (that is, mak-


ing a construction). To solve the phenomena was to construct an empir- ically adequate theory. The solving and even the saving of phenomena then dropped out of common usage. There were revived in this century by philosophers, with Pierre Duhem's book To Save the Phenomena and a chapter of that title in van Fraassen's (1980). The history of astronomy was very present to Duhem but not to van Fraassen. The latter holds that all science aims only at saving the phenomena, at empirical adequacy, at ability to derive observable phenomena. I don't imagine that science "aims" at anything at all, but if we do use the metaphor, then it seems to me that most natural science aims more at manipulating and interfering with the world in order to understand it. Hence saving the phenomena seems an entirely subsidiary aspect of scientific activity. There is one, and perhaps only one, branch of science where the tag "to save the phenomena" has a central place: astronomy and astrophysics.

Surprise, surprise! Is it an accident that the catch phrase has a place in the family of disciplines for which it was coined, and nowhere else? No. The technology of astronomy and astrophysics has changed radically since ancient times, but its method remains exactly the same. Observe the heavenly bodies. Construct models of the (macro)cosmos. Try to bring observations and models into line. In contrast: the methods of the natural sciences have undergone a profound transformation, chiefly in the seventeenth century. Or one might say: the natural sciences came into being then and thereafter, while astronomy is not a natural science at all. (The organization of many university faculties and departments reflects this.) The transition to the natural sciences was precisely the transition to the experimental method, to interference in nature, to the creation of new phenomena, as already discussed in section 3 above.

Astronomy did experience a fundamental technological change at the time of Galileo. The advent of the telescope is a change in kind of observation, compared to which all subsequent changes, aside from inter- planetary exploration, are changes of degree. Yet it left unaffected the method of astronomy: observation and modeling. The data and the analysis of data have undergone massive technological change. We observe phenomena different in kind from any imagined by Galileo: radio emis- sions and absorption lines, for example. No science is in such a state of flux as today's astrophysics, not even molecular biology. (Reviewing M. Bartusiak's Thursday's Universe in the New York Times, December 14, 1986, L. A. Marschall rightly calls the book authoritative; "But read it with dispatch, for the frontiers of astrophysics are a region . . . in which one has to move quickly to stay in the same place".)

Even a reader of the present essay will be aware, in a very modest way, of the rapid changes now occurring in astrophysics. But I persist in saying that the method of the science is the same as that of astronomy

578 IAN HACKING

in hellenistic times. Model, observe, and remodel in such a way as to save the phenomena. I have twice quoted Young et al. (1980) saying that they have a family of models, any one of which enables them to "reproduce the observations" What is reproducing the observations but saving the phenomena?

Van Fraassen is fundamentally in error when he holds that all science is a matter of empirical adequacy and saving the phenomena. He holds this erroneous view because, like almost all philosophers, he is totally theory-oriented, and thereby blind to experiment. Natural (experimental) science is a matter not of saving phenomena but of creating phenomena in the sense of my (1983, chap. 13). But in astrophysics we cannot create phenomena, we can only save them.

We believe in the reality of many entities postulated by theory because we can construct devices that use those entities in order to interfere in other aspects of nature, and to investigate the inner constitution of matter (my 1983, chap. 16). People, it has been said, are tool-making animals. When we use entities as tools, as instruments of inquiry, we are entitled to regard them as real. But we cannot do that with the objects of astrophysics. Astrophysics is almost the only human domain where we have profound, intricate knowledge, and in which we can be no more than what van Fraassen calls constructive empiricists.

Eddington (1920) used a few lines from Milton as his epigraph. Be- cause my anti-realism about astrophysics hinges on a doctrine of modeling, I may use them as my conclusion:

the rest From Man or Angel the great Architect Did wisely to conceal, and not divulge His secrets to be scann'd by them who ought Rather admire; or if they list to try Conjecture, he his Fabric of the Heav'ns Hath left to thir disputes, perhaps to move His laughter at thir quaint Opinions wide Hereafter, when they come to model Heav'n And calculate the Starrs, how they will weild The mightie frame, how build, unbuild, contrive To save appeerances, . . .

Paradise Lost VIII, 71-82.

REFERENCES

Adams, M. T., and Boronson, M. A. (1979), "Direct Imaging of the Twin QSOs 0957 + 561 A and B: The Gravitational Lens Interpretation", Nature 282: 183-185.


Barnothy, J. M. (1965), "Quasars and the Gravitational Image Intensifier", Astronomical Journal 70: 666.

Barnothy, J. M., and Barnothy, M. F. (1968), "Galaxies as Gravitational Lenses", Science 162: 348-352.

. (1983), "Quasars in a FIB Universe", in J. P. Swings (ed.), Quasars and Grav- itational Lenses: 24th Liege International Astrophysical Colloquium. Liege: Univ- ersite de Liege, pp. 139-142.

Blandford, R. D., and Narayan, R. (1986), "Fermat's Principle, Caustics, and the Clas- sification of Gravitational Lens Images", Astrophysical Journal 310: 568-582.

Blandford, R. D.; Phinney, E. S.; and Narayan, R. (1987), "1146 + 111 B, C: A Giant Gravitational Lens?" Astrophysical Journal 313: 28-36.

Bourassa, R. R., and Kantowski, R. (1976), "Multiple Image Probabilities for a Sphe- roidal Gravitational Lens", Astrophysical Journal 205: 674-687.

Burke, W. L. (1981), "Multiple Gravitational Imaging by Distributed Masses", Astro- physical Journal 244: LI.

Cartwright, N. (1983), How the Laws of Physics Lie. Oxford: Oxford University Press. Chang, K., and Refsdal, S. (1979), "Flux Variations of QSO 0957 + 561 A, B and Image

Splitting by Stars Near the Light Path", Nature 282: 561-564. De Silva, L. N. K. (1970), "Quasi-Stellar Objects and Gravitational Lenses", Nature 228:

1180-1181. Djorgovski, S., and Spinard, H. (1984), "Discovery of a New Gravitational Lens", As-

trophysical Journal 282: LI-L4. Dyer, C. C., and Roeder, R. C. (1972), "The Distance-Redshift Relation for Universes

with No Intergalactic Medium", Astrophysical Journal 174: LI 15-LI 17. . (1973), "Distance-Redshift Relations for Universes with Some Intergalactic Me-

dium", Astrophysical Journal 180: L3 1-L34. . (1980), "Possible Multiple Imaging by Spherical Galaxies", Astrophysical Jour-

nal 238: L67-L70. Eddington, A. S. (1920), Space, Time and Gravitation. Cambridge: Cambridge University

Press. . (1931), "The Expansion of the Universe", Monthly Notices of the Royal Astro-

nomical Society 91: 412-416. Ehlers, J., and Schneider, P. (1986), "Self-Consistent Probabilities for Gravitational Len-

sing in Inhomogeneous Universes", Astronomy and Astrophysics 140: 119-124. Einstein, A. (1936), "Lens-Like Action of a Star by the Deviation of Light for the Grav-

itational Field", Science 84: 506-507. Falco, E. E.; Gorenstein, M. V; and Shapiro, I. I. (1985), "On Model-Dependent Bounds

on H0 from Gravitational Images: Application to Q0957 + 561 A, B", Astrophysical Journal 289: LI-L4.

Gorenstein, M. V.; Shapiro, I. I.; Cohen, N. L.; Corey, B. E.; Falco, E. E.; and Marcaide, J. M. (1983), "Detection of a Compact Radio Source Near the Center of a Gravi- tational Lens: Quasar Image or Galactic Core", Science 219: 54-56.

Gott, J. R. (1981), "Are Heavy Halos Made of Low Mass Stars? A Gravitational Lens Test", Astrophysical Journal 243: 140-146.

Gott, J. R., and Gunn, J. E. (1974), "The Double Quasar 1548 + 115 A, B as a Grav- itational Lens", Astrophysical Journal 190: L105-L108.

Greenfield, P. E.; Roberts, D. H.; and Burke, B. F. (1980), "The Double Quasar 0957; + 561: Examination of the Gravitational Lens Hypothesis Using the Very Large Ar- ray", Science 208: 495-497.

Gunn, J. E. (1981), "Gravitational Lenses in Astrophysics", Annals of theNew YorkAcad- emy of Sciences 375: 287-296.

Hacking, I. (1983), Representing and Intervening. Cambridge: Cambridge University Press. . (1989), "Astronomical Improbability", in J. E. Fenstad, I. T. Frolov, and R.

Hilpinen (eds.), Logic, Methodology and Philosophy of Science VIII. Proceedings of the Eighth International Congress, Moscow, USSR, August 1987. Amsterdam: North- Holland.

580 IAN HACKING

Harman, G. (1965), "The Inference to the Best Explanation", Philosophical Review 74: 88-95.

Hewitt, A., and Burbidge, G. (1987), "A New Optical Catalog of Quasi-Stellar Objects", Astrophysical Journal Supplement Series 63: 1-246.

Hogan, C., and Narayan, R. (1984), "Gravitational Lensing by Cosmic Strings", Monthly Notices of the Royal Astronomical Society 211: 575-591.

Huchra, J.; Gorenstein, M.; Kent, S.; Shapiro, I; Smith, G.; Horine, E.; and Perley, R. (1985), "A New and Unusual Gravitational Lens", Astronomical Journal 90: 691- 696.

Lawrence, C. R.; Schneider, D. P.; Schmidt, M.; Bennett, C. L.; Hewitt, J. N.; Burke, B. F.; Turner, E. L.; and Gunn, J. E. (1984), "Discovery of a New Gravitational Lens System", Science 223: 46-49.

Lodge, 0. S. (1919), "Gravitation and Light", Nature 104: 354. Ostriker, J. P., and Vietri, M. (1985), "Are Some BL Lac Objects Artefacts of Gravi-

tational Lensing?" Nature 318: 446-448. Paczynski, B. (1986a), "Will Cosmic Strings Be Discovered Using the Space Telescope?"

Nature 319: 567-568. (1986b), "Is There a Black Hole in the Sky?" Nature 321: 419-420. (1986c), "Gravitational Microlensing at Large Optical Depth", Astrophysical Journal

301: 503-516. Paczynski, B., and Gorski, K. (1981), "Another Possible Case of a Gravitational Lens",

Astrophysical Journal 248: L101-L104. Peacock, J. A. (1982), "Gravitational Lenses and Cosmological Evolution", Monthly No-

tices of the Royal Astronomical Society 199: 987-1006. . (1983), "Review of a Gravitational-Lens Theory", in J. P. Swings (ed.), Quasars

and Gravitational Lenses: 24th Liege International Astrophysical Colloquium. Liege: Universit6 de Liege, pp. 86-104.

Peebles, J. (1980), The Large Scale Structure of the Universe. Princeton: Princeton Uni- versity Press.

Pooley, G. G.; Browne, I. W. A.; Daintree, E. J.; Moore, P. K.; Noble, R. G.; and Walsh, D. (1979), "Radio Studies of the Double QSO, 0957 + 561 A, B", Nature 280: 461-464.

Porcas, R. W.; Booth, R. S.; Browne, I. W. A.; Walsh, D.; and Wilkinson, P. N. (1979), "VLBI Observations of the Double QSO, 0957 + 561 A, B", Nature 282: 385-386.

Press, W. H., and Gunn, J. E. (1973), "Method for Detecting a Cosmological Density of Condensed Objects", Astrophysical Journal 185: 397-412.

Refsdal, S. (1964a), "The Gravitational Lens Effect", Monthly Notices of the Royal As- tronomical Society 128: 295-306.

. (1964b), "On the Possibility of Determining Hubble's Parameter and the Masses of Galaxies from the Gravitational Lens Effect", Monthly Notices of the Royal As- tronomical Society 128: 307-310.

. (1966), "On the Possibility of Testing Cosmological Theories from the Gravi- tational Lens Effect", Monthly Notices of the Royal Astronomical Society 132: 101- 111.

Roberts, D. H.; Greenfield, P. E.; and Burke, B. F. (1979), "The Double Quasar 0957 + 561: A Radio Study at 6-Centimetres Wavelength", Science 205: 894-896.

Russell, H. N. (1937), "A Relativistic Eclipse", Scientific American (February) 156: 76- 77.

Salmon, W. (1984), Scientific Explanation and the Causal Structure of the World. Prince- ton: Princeton University Press.

Sanitt, N. (1971), "Quasi-Stellar Objects and Gravitational Lenses", Nature 234: 199- 203.

Schneider, P. (1984), "The Amplification Caused by Gravitational Bending of Light", Astronomy and Astrophysics 140: 119-124.

. (1985), "A New Formulation of Gravitational Lens Theory, Time-Delay, and Fermat's Principle", Astronomy and Astrophysics 143: 413-420.


(1986), "Statistical Gravitational Lensing and Quasar-Galaxy Association", As- trophysical Journal 300: L31-L34.

Schneider, P., and Weiss, A. (1987), "A Gravitational Lens Origin for AGN-Variability? Consequences of Micro-Lensing", Astronomy and Astrophysics 171: 49-65.

Shapere, D. (1982), "The Concept of Observation in Science and Philosophy", Philosophy of Science 49: 485-525.

Soifer, B. T.; Neugebauer, G.; Matthews, K.; Becklin, E. E.; Wynn-Williams, C. G.; and Capps, R. (1980), "IR Observations of the Double Quasar 0957 + 561 A, B and the Intervening Galaxy", Nature 285: 91-93.

Stockton, A. (1980), "The Lens Galaxy of the Twin QSO 0957 + 561", Astrophysical Journal 242: L141-L144.

Turner, E. L. (1980), "The Effect of Undetected Gravitational Lenses on Statistical Mea- sure of Quasi Evolution", Astrophysical Journal 242: L135-L139.

Turner, E. L.; Ostriker, J. P.; and Gott, J. R. (1984), "The Statistics of Gravitational Lenses: The Distribution of Image Angular Separations and Lens Redshifts", Astro- physical Journal 284: 1-22.

Tyson, J. A. (1985), "Image Distortion Near Galaxies: Dwarfs or Lensing?" Nature 316: 799-800.

Van Fraasen, B. C. (1980), The Scientific Image. Oxford: Oxford University Press. Walsh, D. (1983), "Observations of Gravitational Lenses", in J. P. Swings (ed.), Quasars

and Gravitational Lenses: 24th Liege International Astrophysical Colloquium. Liege: Universit6 de Liege, pp. 106-124.

Walsh, D.; Carswell, R. F.; and Weymann, R. J. (1979), "0957 + 561 A, B: Twin Quasi- Stellar Objects or Gravitational Lens? Nature 279: 381384.

Weedman, D. W.; Weymann, R. J.; Green, R. F.; and Heckman, T. M. (1982), "Dis- covery of a Third Gravitational Lens", Astrophysical Journal 255: L5-L9.

Weinburg, S. (1972), Gravitational and Cosmology. New York: John Wiley & Sons. Weymann, R. J.; Latham, D.; Angel, J. R. P.; Green, R. F.; Liebert, J. W.; Turnshek,

D. A.; Turnshek, D. E.; and Tyson, J. A. (1980), "The Triple QSO PG1115 + 08: Another Probable Gravitation Lens", Nature 285: 641-643.

Young, P.; Deverill, R. S.; Gunn, J. E.; Westphal, J. A.; and Kristian, J. (1981), "The Triple Quasar QII15 + 080 A, B, C: A Quintuple Gravitational Lens Image?" As- trophysical Journal 244: 723-735.

Young, P.; Gunn, J. E.; Kristian, J.; Oke, J. B; and Westphal, J. A. (1980), "The Double Quasar Q0957 + 561 A, B : A Gravitational Lens Image Formed by a Galaxy at Z=0.39", Astrophysical Journal 241: 507-520.

. (1981), "Q0957 + 561: Detailed Models of the Gravitational Lens Effect", As- trophysical Journal 244: 736-755.

Zwicky, F. (1937a), "Nebulae as Gravitational Lenses" Physical Review 51: 290. . (1937b), "On the Probability of Detecting Nebulae Which Act as Gravitational

Lenses", Physical Review 51: 679.

1989 - extragalactic reality- the case of gravitational lensing

Documents

conditions of use

noncommercial use

jstor archive

scientific realism

jstor transmission

youmay use content

compelling argument

experimental argument