Radio Source Survey: Disputes, 1948–1957

Abstract

The decade of the 1950s became one of the most eventful in the history of not only radio, but of all astronomy, as a result of the first significant surveys of radio sources. During this decade it was realised that the radio sources included an isotropic population of extragalactic sources which could be observed in the radio at greater distances than the deepest optical observations. Radio astronomy suddenly became a part of extragalactic astronomy, and most significantly, a completely new and unexpected probe of cosmology. The story of the development and interpretation of these surveys is a complex one, involving technically challenging instrumental issues, one controversy involving the different results obtained by the observational groups located in Sydney and in Cambridge, and another controversy between the cosmologists and the observers. The situation was further exacerbated by the intense emotional interactions between some of the key personalities involved.

Perhaps the greatest discontinuity [in my career] was with the identification of Cygnus A. That showed that [in 1953] we were in the cosmology game … To me that was the point where one said, "Well now, if other things like Cygnus exist, here is something which we can likely see, even with our little instrument, as far away as the [Mt Palomar] 200-inch can see. This is something much more interesting than it might have been—much more interesting than [the radio sources] being galactic objects …

–Martin Ryle (Interview with W. Sullivan, 1976, cited in Sullivan, 1990, 2009)

Introduction

The decade of the 1950s became one of the most eventful in the history of not only radio, but of all astronomy, as a result of the first significant surveys of radio sources. During this decade it was realised that the radio sources included an isotropic population of extragalactic sources which could be observed in the radio at greater distances than the deepest optical observations. Radio astronomy suddenly became a part of extragalactic astronomy, and most significantly, a completely new and unexpected probe of cosmology. The story of the development and interpretation of these surveys is a complex one, involving technically challenging instrumental issues, one controversy involving the different results obtained by the observational groups located in Sydney and in Cambridge, and another controversy between the cosmologists and the observers. The situation was further exacerbated by the intense emotional interactions between some of the key personalities involved.

The facts of the story are now fairly well known, with substantial reviews and interpretation in Edge and Mulkay (1976), Bertotti, Balbinot, Bergia, and Messina (1990), Kragh (1996), and Sullivan (2009). Here we summarise these events and present the controversy from the Sydney perspective, which is less well known. Conflicts between the Sydney and Cambridge groups had already started over failed collaboration attempts, especially in relation to the early optical identifications proposed for some radio sources.¹ New, more serious, conflicts were now arising due to major discrepancies between the early survey catalogues and disagreements about how to interpret the results, and these differences are explored more thoroughly in this theme chapter.

Exploring the disagreements and controversies about the surveys of the 1950s offers many enticing opportunities for gaining insight into how the field of radio astronomy developed in this period. We can also delve into how and why some scientists missed the key insights that might have delivered swifter progress, and how and why others reversed their opinions. This was a period of fast-changing conceptions about what radio observations could accomplish, the nature of the phenomena observed and the implications of these observations for not only astronomy, but also cosmology. These are the details that we have attempted to elucidate in this chapter.

Early Radio Source Surveys

In 1948 Pawsey suggested a radio continuum survey of the southern sky to complement the solar work.² The Cambridge group under Martin Ryle also started a survey in 1948. The initial objective of these surveys and those that followed throughout the 1950s was to extend the short list of discrete radio sources then known in order to understand what they were. It was naturally assumed that the brightest radio sources in the sky would be associated with the brightest optical stars, thus making a catalogue to identify and classify individual radio sources was the obvious next step. Initially there were no obvious identifications with bright stars apart from the sun and the Milky Way. The groups needed to measure more accurate positions so the radio sources could be identified with known astronomical objects. This initial work started at Sydney and Cambridge at about the same time. As the 1950s progressed, additional objectives of understanding the distribution of the population of radio sources became a topic of intense interest. How were they distributed across the sky—isotropic or concentrated along the Milky Way? How were they distributed in intensity—were there more faint sources farther away? By 1950 the radio astronomy research group at Jodrell Bank³ also began to conduct surveys using their large transit dish.⁴

The main radio source surveys conducted in the 1950s and early 1960s are summarised in Table 35.1. We discuss some of the more critical results here; more technical details can be found in the ESM_35.1.pdf, Radio Source Surveys.

Table 35.1 Radio source surveys

For their survey, the Cambridge group built a two-element interferometer called the “Long Michelson” (see ESM_35.1.pdf, Radio Source Surveys). Fifty radio sources were detected and a “Preliminary survey of the radio stars in the Northern Hemisphere” was published by Ryle et al. (1950), later to be known as the 1C (First Cambridge) survey. Based on the lack of any correspondence with the brightest galaxies, they concluded that these sources were “radio stars” in our galaxy although the positions were still too poor to make any associations with specific stars. There were too many possibilities.

In Sydney, Bernard Mills used three antennas configured as two phase-switched 101 MHz interferometers with baselines of 60 and 270 m at a site at Badgery’s Creek, southwest of Sydney (Fig 35.1a).⁵ This was first introduced in Chap. 21 and more details are provided in ESM_35.1.pdf, Radio Source Surveys.

Fig. 35.1

(a) One of the three elements used for the 101 MHz all sky survey made by Mills in 1950–1951. Mills stands next to one of the aerials (See also Chap. 21, Fig. 21.1). The primary beam was 14 by 24 deg. The survey was conducted in 10 deg. steps, i.e. the aerial was fixed at a certain elevation (0 to 90 deg) and the sky drifted past. The declination ranged covered was +50 to −90 degrees. Credit: CSIRO Radio Astronomy Image Archive B2594-4. (b) Shows the different behaviour of sources near ① or away ② from the galactic plane. Credit: Fig. 7, “The Distribution of the Discrete Sources of Cosmic Radio Radiation”, Mills, B. Y. (1952b). Australian Journal of Scientific Research A Physical Sciences 5: 266

The survey yielded 77 sources (Mills, 1952b) and an important result was that Mills divided the radio sources he found into two classes: Class I sources were concentrated along the galactic plane with galactic latitudes <12 deg. tending to be the more intense sources. On the other hand the Class II sources were displaced from the galactic plane and showed a different log N- log S behaviour (Fig. 35.1b).

They used the famous log N- log S description of the counts of the number of radio sources of different intensity, a powerful diagnostic for the way the radio sources are distributed in space. N is the number of sources per unit area with a flux density S or greater. This presentation is often characterised by a slope, α, where N ∝ S^α or Log (N) ∝ α Log (S). In a 3D Euclidian Universe, the volume of space and hence the number of sources, N, increases with the cube of the distance and the flux density of the sources decreases as the square of the distance so α = −3/2 = −1.5. However, if the sources are confined to a finite volume there will be fewer fainter sources at larger distance so α > −1.5.⁶ In the Mills survey the galactic plane sources of Type I showed a slope (α) of −0.75 while the Class II sources indicated a slope of −1.5. Class I sources were characteristic of a population of Milky Way objects in a finite depth disk as observed from the sun. Mills pointed out that the Class II sources could arise either from a nearby local population (e.g. stars) in the galaxy or from extragalactic objects.

At the time many astronomers, and in particular the Cambridge group, held the view that the discrete radio sources were stars in our galaxy because if these brightest radio sources in the sky were outside our galaxy they would have to be extraordinarily luminous (see Chap. 16). Therefore, Mills’s data, one of the first suggestions of an extragalactic population of radio sources which were not just a few abnormal sources such as Virgo A and Centaurus A (identified by Bolton, Stanley, & Slee, 1949, as external galaxies) was potentially revolutionary. Mills was cautious, but he wrote to Bowen, who was in the UK to give a presentation to the RAS in London at the time: “I can say, but not for publication, that the evidence favours the extragalactic hypothesis …”.

Mills needed to be cautious since, already in 1951, there were discrepancies between the first survey results, as everyone involved was keenly aware. The Ryle et al. (1950) 1C survey had 50 sources fewer than detected by Mills; they claimed that it showed no evidence for the two classes. We note that Sullivan (2009, p. 365) repeated the analysis of the source counts from the 1C list and obtained results very similar to Mills: “I find a log N- log S slope of 2.1 with a demarcation similar to Mills’s (1952a) for sources at low and high galactic latitude. The 1C sample also shows an excess of strong sources near the galactic plane (six of seven of the strongest sources are at latitudes ≤12°).” Mills wrote to Bowen that he thought a major problem in the northern galactic plane for Ryle and Smith was the effect of sidelobes from the two strongest sources in the north which were both near the galactic plane, Cassiopeia A and Cygnus A.⁷ The former was circumpolar (i.e. always above the horizon as observed from Cambridge). Mills wrote: “Since Ryle has to deal with two powerful sources right on the galactic plane, I think ours is more likely to be correct (it may not be a good idea, however, to claim any superiority, particularly if Ryle should be present).” Ryle did attend the 11 May 1951 presentation by Bowen at the RAS in London.⁸

In Sydney Bolton, Stanley, and Slee (1954) published the results of the final survey made with the Dover Heights sea-cliff interferometer (see ESM_35.1.pdf, Radio Source Surveys). The 104 discrete sources detected were compared with other radio source catalogues made up to that time. As already noted by others there was good agreement for the brighter sources but significant disagreements for many of the fainter sources. Bolton et al. stated that some disagreements were to be expected since, in addition to the differing responses to extended sources, there was the issue of confusion. For cases where the source density was high enough to have two sources in the primary beam, both interferometers and single dishes will give incorrect [but different] results. They emphasised the value of multiple surveys with different types of antenna and noted that some source confusion was removed with the sea-cliff interferometer due to the sharp edge of the earth’s shadow. Remarkably Bolton, Stanley, and Slee (1954) concluded their paper with a most significant result.

… There seem to be too many faint sources compared with an isotropic distribution of objects … A plausible explanation is that the Sun (if these sources are galactic) or the Galaxy (if these sources are extragalactic) is in a local region of low source density and that somewhere towards the limit of the survey we reach a region of much higher density … however, there is not much point in speculating too far on this result as it could also be produced by a large dispersion in absolute magnitude [intrinsic luminosities] amongst the sources of the survey.

This is the first published evidence for an evolving population of radio sources, but unfortunately, Bolton and his colleagues did not realise that their concern about the effect of a large dispersion in luminosity was misguided. This was understandable. Fainter sources could either be further away, or intrinsically less luminous; one would naturally think that these two effects could not be disentangled. But this argument for an evolving population of sources was far stronger than Bolton, Stanley, and Slee (1954) realised. The next year Ryle and Scheuer (1955) showed that while a large dispersion in intrinsic luminosity could decrease the slope of the source counts, it can never increase the slope! Unfortunately, Bolton was not in a position to use this information. Due to the decision to fund the Mills Cross telescope rather than a new instrument at Dover Heights (see Chap. 22), Bolton had left radio astronomy in Aug 1953 to work on cloud physics (rain making). He had no further involvement in radio astronomy until after he moved to Caltech to set up the Owens Valley Radio Observatory in 1955. But, as we will see at the end of this chapter, Bolton eventually had the last word on the surveys.

Sullivan (1990) has presented a striking image that showed that for the four main surveys up to 1953 (see Table 35.1) few optical identifications had been made and the surveys “profoundly disagree with one other” (Sullivan, 1990, p. 318 in Modern Cosmology in Retrospect). Some of the disagreement related to an issue which was to continue to be controversial for many years: that of the nature, as well as the position, of many of the radio sources. Were these (sun-like) stars or other kinds of objects, and were they inside, or beyond, our galaxy? In this period, Ryle was quite committed to the assumption that the radio sources were stars in the galaxy, which would have extremely small angular size.

All three radio astronomy groups were strongly influenced by their understanding of the instruments they were using, which were very different. Many of the discrete sources (especially in the galactic plane) were later found to be extended, so the responses were quite different for single dishes—as were used particularly at Jodrell Bank—which detected all the radio emission from an extended source, and interferometers which are insensitive to extended structure.

In his thorough analysis of the underlying reasons for the survey controversies of 1948–1953, Sullivan (2009, p. 361) discussed the complexity of disputes centred on instrumentation. He pointed out that Ryle’s use of long interferometer baselines was well motivated, since it was the result of his conviction that the radio sources were stars which have very small angular size. His success in using an interferometer with its phase switch (invented at Cambridge, see Chap. 37) to remove the strong diffuse galactic emission also made this the optimum instrumental design for a survey of point sources. Ryle saw this as a major advantage over the single dish, which would be confused by structure in the diffuse galactic emission. Sullivan (2009, p. 361) has described the importance of using an interferometer to eliminate confusion caused by small scale variations in a letter that Ryle wrote to Tandberg-Hanseen on 10 February 1950:

As Ryle explained at the time⁹:

It is very easy [with large dishes] to obtain maxima in the received flux which appear to be due to a source of small angular diameter, but which in reality are due to the angular variations of the general structure of the Galaxy. For this reason, we have always used interference systems of considerable resolving power to discriminate between “point” sources (i.e., sources having a diameter of less than 5–10 minutes of arc) and the general background radiation from the Galaxy.

This is a nice example of how a concept that turned out to be mostly incorrect—the notion of radio “stars”—can be central to instrument design.¹⁰ As Sullivan points out, in this case it was an example of an effect discussed by Pickering (1981) and Galison (1987) where a researcher shaped his instrument to obtain the anticipated result. While this is often necessary for success, there was also the obvious danger of prolonged misleading research results. So it is interesting to see how these preconceived ideas affected the eventual outcome. As we will see by the end of this chapter, Ryle was in one sense correct—the interferometers and arrays were far superior to the single dishes for surveys of the extragalactic discrete source population. But this was not obvious at all in the early 1950s, and the conflicting results from the single dishes were a huge distraction that greatly confused the discussions at the time.

After their first survey, the Cambridge group began planning for a new enlarged instrument, designed in 1950–1951 and constructed in 1952. (See Fig. 35.2 and ESM_35.1.pdf, Radio Source Surveys, for a description of the 2C aerial.) But since Ryle only expected point sources which would have the same amplitude on any interferometer baseline, he only used a single baseline. This produced a disastrously incorrect survey. Multiple point sources within his beam, or even within the sidelobes of his beam, were incorrectly catalogued as single sources in the wrong position if the interferometer fringes added, or were missed altogether if the fringes cancelled.

Fig. 35.2

Comparison of the Cambridge 2C array (top) and the Mills Cross (bottom). See ESM 35.1.pdf, Radio Source Surveys, for more details. Diagrams based on sketches by Martin Ryle. Courtesy of and copyright: Cavendish Laboratory, University of Cambridge, all rights reserved

Disagreements about instruments in this period were worsened by Ryle’s personality and difficulty in collaborating. Ryle’s disdain for other instruments made it harder, not easier, for a collective scientific assessment of the affordances and limitations of different instruments and of the consequent issues of confusion and potential identifications of sources with astronomical objects. For example, Hanbury Brown reflected later on the first Jodrell Bank survey (see ESM_35.1.pdf, Radio Source Surveys) which he and Cyril Hazard had conducted with a 218-foot transit parabolic dish (Hanbury Brown & Hazard, 1953b) and compared with the Cambridge 1C interferometer survey of Ryle et al. (1950)¹¹:

A comparison of the two surveys showed that of the 13 most intense sources detected with our pencil-beam (10 of which lay within 5° of the galactic plane) only 4 appeared in the [Cambridge 1C] survey. I shall never forget the strenuous arguments we had with those dedicated interferometrophiles in our efforts to convince them that our sources in the galactic plane were real and not side-lobes of our pencil-beam. It was part of the conventional wisdom at Jodrell Bank that Cambridge had only three standard reactions to our work: (1) "it is wrong", (2) "we have done it before", or (3) "it is irrelevant". Indeed, as we later showed, at least 6 of the 10 sources in the galactic plane had angular sizes exceeding 1° and were either partially or totally resolved, and therefore largely undetected, by their interferometer.¹²

The issue of the influence of possible extended extragalactic sources took another decade to settle. Although these were real, and were missing in the Cambridge survey, it was a small effect that had not influenced the conclusions from Cambridge. However, the fact that some of the strongest sources in the southern hemisphere were very extended (eg Centaurus A) led Mills astray. He assumed there were similar sources in his survey, and he overcorrected for this effect. So, while Pickering and Galison were correct in noting a potential bias from instrument design at Cambridge, this bias did not affect the final outcome. Even though Ryle started with strong pre-conceptions and a personality that hampered the capacity to acknowledge error, his own evidence eventually convinced him to dramatically change his view.

Sullivan (2009) ends his comprehensive coverage of the early history of radio astronomy at the end of 1953 (though reference is made to some post-1953 events in his discussions of the repercussions of some of the pre-1953 radio astronomy history). Technology was developing at a rapid rate and an entirely new field of astronomy was being opened. Completely new and unexpected classes of radio sources were being found (or at least, tentatively suggested) all over the sky; the radio source surveys became the focus of future investigations, especially for the groups in the UK and Australia who dominated this field.

The 2C survey and Extragalactic Radio Sources: Radio Sources as a Population

During 1953, plans for the next generation of survey instruments were being developed in Sydney and in Cambridge. These included the conceptual design of the Mills Cross (as a survey telescope) in Australia, while Smith at Cambridge emphasised the advantages of the planned 2C antenna, at 3.7 m. These instruments could not and did not resolve the issue of confusion, i.e. the probability of multiple sources simultaneously in the telescope beam, and discussions continued about how to interpret their findings.

Thus perhaps the greatest impact of the 2C survey was its effect on Martin Ryle, who would suddenly reverse his position that radio “point sources” were new radio-emitting forms of stars that occurred within our galaxy. Ryle’s reversal of position—as he well understood—had huge cosmological implications.

In Sydney, this extragalactic population had been long recognised following the identification Bolton et al. (1949) of Virgo A (M87) and Centaurus A (NGC 5128)¹³ and Mills’s 1951 division of the discrete radio sources into two classes: Class I (galactic) and Class II (possibly extragalactic), as we discussed above. Those cosmologists—Fred Hoyle (1915–2001), Thomas Gold (1920–2004) and Hermann Bondi (1919–2005)—and astronomers—Baade and Minkowski (see Chap. 16)—who were aware of radio astronomical investigation in this period were also very willing to consider alternative explanations and extra-galactic distances for radio sources.

But as is well known, before 1953 Ryle had inflexibly rejected speculation that most sources were extragalactic. Why was this? The strongest source in the northern sky had no bright nearby galaxy identification; from Ryle’s perspective, if the emission was from a very distant extragalactic source, it would be so powerful as to stretch credibility. He had dismissed the Bolton et al. (1949) identifications of Centaurus A as “not certain” and the case of Virgo A identified with the galaxy M87 as “less interesting”, just a normal galaxy like the nearby Andromeda galaxy, which had been detected as a very weak radio source at Jodrell Bank. Extragalactic emission was also not compatible with his view that the radio emission from the Milky Way was the sum of all the radio stars in the galaxy (Boyd, 1951).

Hoyle’s autobiography (Hoyle, 1994) includes a colorful description of an exchange between Ryle and Gold following talks by Gold and Ryle at the Massey meeting (Boyd, 1951) in April 1951.

By 1951, about half a dozen radio sources had been definitively related to astronomical objects, not one of which had turned out to be a star. There had been identifications at Jodrell Bank with weakly emitting nearby galaxies, and there had been the two cases identified by John Bolton, the Crab Nebula and the galaxy NGC 5128. So Gold said that perhaps the other possibility for explaining the isotropic distribution of radio sources—namely, that most radio sources were very distant—should be taken seriously. He said nothing more than that, and he expressed it temperately. It was in these circumstances that Ryle began an attack that was to persist for almost two decades … he [Ryle] began, “What theoreticians have failed to understand … ” (with the word theoreticians implying some inferior and detestable species) …

I do not think it unreasonable to say that Ryle’s motivation in developing a program of counting radio sources, a program that was to occupy a major fraction of his group over the next ten years, was to exact revenge for his humiliation over the radio-star affair. This was to be done by knocking out the new form of cosmology with which Gold, Bondi, and I were associated.

Ryle could hold onto his misconception that the radio sources were mostly galactic stars, even though Minkowski had identified Cygnus A with a distant galaxy using the Cambridge position,¹⁴ by assuming Cygnus A was an abnormality, an outlier which, while interesting, could still be ignored. But in late 1953 he suddenly saw how all the data could make sense, if this “outlier” was typical for the population.

Sullivan (1990, p. 321) has analysed Ryle’s notes from the early 1950s to find out when this dramatic change of view occurred. Before 1953 Ryle was an adamant supporter of the galactic radio star model. But in October 1953 he presented a talk to the Cavendish Physical Society:

Now it so happens that there is a large isotropic component of radiation which cannot be explained in terms of emission from galactic sources. If indeed it is extragalactic, it offers the possibility of being able to distinguish between some of the cosmological theories.

Whether the observations will ever be sufficiently accurate one cannot say, but it is nice to think that the cosmologists may one day [lose their] complete freedom of choice of the conditions beyond the optical limit.¹⁵

The first direct archival evidence related to the new (2C) survey found by Sullivan is from 30 Jul 1954 when Ryle made a note: “Study of extragalactic sources—cosmological application—little hope of much identification … Main object will be to see further—even if area of sky has to be restricted …”.

Sullivan goes on to document Ryle’s later recollection of this period:

Perhaps the greatest discontinuity [in my career] was with the identification of Cygnus A. That showed that we were in the cosmology game … To me that was the point where one said, "Well now, if other things like Cygnus exist, here is something which we can likely see, even with our little instrument, as far away as the 200-inch can see. This is something much more interesting than it might have been—much more interesting than [the radio sources] being galactic objects, much more interesting than M87's [fairly normal galaxies]" … (Interview with W. Sullivan, 1976, cited in Sullivan, 1990, 2009)

It was the interpretation by Shakeshaft et al. (1955) of the 2C catalogue of radio sources paper, called “The Spatial Distribution and the Nature of Radio Stars” by Ryle and Scheuer (1955), that caused a commotion. In the logN-logS plot in Fig. 35.3, N is the number of sources per steradian (I) with intensity greater than (I) in units of 10 Jy. Ryle and Scheuer pointed out that the steep slope (<−1.5 which is the expected slope for an isotopic distribution in a static Euclidean universe) could only be explained if the sources were extragalactic, of similar luminosity as Cygnus A, and “of much greater number density at larger distances than nearby” (giving more faint sources than expected and hence the steep negative slope at faint flux density levels). (What seems surprising to us now is that they were not concerned by the very rapid flattening of the counts for faint sources. This effect would only be possible in a finite universe or if there was a mistake in measuring amplitudes for faint sources; it should have alerted the group to instrumental error.)

Fig. 35.3

2C survey results from Ryle and Scheuer (1955) showing the log of the number of sources versus the log of the source intensity. This is referred to as the log N- log S source counts. Credit: Fig. 1, “The Spatial Distribution and the Nature of Radio Stars”, Ryle, M., and Scheuer, P.A.G. (1955). Proceedings of the Royal Society of London Series A 230, pp. 448–462

Scheuer (Scheuer & Bertotti, 1990, p. 331) included a brief historical timeline. He asserted that Ryle was already convinced in late 1953 that the numbers “were not thinning out with distance, as they should, sooner or later, if they were local galactic stars … At this point—it must have been in the winter of 1953-54 or the early spring of 1954—Martin Ryle’s attitude on ‘radio stars’ changed almost overnight. They had to be extragalactic.” And the population of sources had to increase with distance. This was evidence that contradicted the Steady-state theory recently developed by Hoyle, Gold and Bondi, which predicted a constant density in the universe. This effect could not be explained by a local distribution of galactic stars unless the sun were located in the centre of a spherical hole in the nearby universe.

Ryle’s Halley Lecture, Oxford, 6 May 1955

Ryle (1955) presented his new view on the nature of the radio sources in his Halley Lecture. He described a series of well-focussed steps leading to the need for a survey of radio sources to understand their distribution in the universe. He made no reference to similar, earlier suggestions of just such an extragalactic population made by the Dutch, and by the Australians who had made this suggestion 4 years earlier. It had also been proposed by his adversarial theoreticians Gold and Hoyle, all of whom he had previously refuted.

He gave no hint that he had just made a complete reversal of his previous galactic radio star interpretation and instead he presented his new world view as if it were all the logical consequence of the Cambridge work. This was a very Cambridge-centric version of the history. Nor should we necessarily find this surprising. Memory, we now know, does not withdraw information about the past from storage. Rather, memory is an active reconstructive process that generates and validates a particular conception of the self; indeed, research shows that when we describe our memories differently to different audiences, the memory itself (and not just the message) alters in what is known as the “audience tuning effect”. Ryle’s version of history can be understood as an example of this effect.¹⁶ Ryle became very assertive, even dogmatic, about his new views; interestingly, Mills, who had already made this case long before, remained characteristically conservative.

In this lecture, Ryle proceeded to clearly describe the nature of the source counts which would have N(s) ∝ s^-1.5 in a static Euclidean Universe. He noted that the shape was not altered even if the sources all had different luminosities.¹⁷ The results from the first catalogue (2C with 1936 sources, Shakeshaft et al., 1955) required an increase in apparent spatial density or luminosity of “radio stars” with distance. Note that Ryle continued to use the term “radio star” to describe both galactic and extra-galactic sources; clearly the terminology no longer implied that the objects were “stars”, but by keeping the old terminology the Cambridge group had disguised their previous misconceptions. Most other radio astronomers of this era had ceased using the terminology “radio stars”.

Ryle proposed that many of these radio sources could have luminosities comparable to Cygnus A, but be located more distantly in the universe. They would likely not be identifiable even with the 200-inch telescope. This point was very important—and with this step forward he had leapfrogged all the other groups. It explained why most radio sources could not be identified with galaxies—a problem which many observers of the survey controversy took to imply a poor-quality radio catalogue and a lack of progress.

A reason why this step may have been more obvious to the Cambridge group is the asymmetry between the Northern and Southern Hemispheres. In the South there were several relatively nearby bright radio galaxies: NGC 5128 (Centaurus A), NGC 1316 (Fornax A) and M87 (Virgo A) while the brightest source in the Northern Hemisphere is the far more distant Cygnus A. It had been implicitly assumed that the brightest sources would be associated with the brightest and closest galaxies, but this was not the case.

Ryle had been explicit in his conclusion of the Halley Lecture: “This is a most remarkable and important result, but if we accept the conclusion that most of the radio stars are external to the galaxy, and this conclusion seems hard to avoid, then there seems no way in which the observations can be explained in terms of a Steady-state theory.”

As noted by Sullivan (1990, p. 23), the Steady-state group was quick to respond.

This Steady-state theory had been invented in the late 1940s at Cambridge by Bondi, Gold and Hoyle and provided an attractive and testable alternative to the various older “big bang” models developed in the decade following the introduction of general relativity. Ryle’s sweeping disproof of Steady-state theory naturally caught the imagination and attention of the general public and the scientific community. Here was no less than a major theory of the universe being overthrown.

Within a week, the Steady-state proponents had a chance to respond at the Royal Astronomical Society meeting the following Friday (13 May 1955). An image of the 2C survey and a newspaper billboard from London on 13 May 1955 are also shown in Fig. 35.4.

Fig. 35.4

2C survey (Shakeshaft et al., 1955) and P(d) analysis (Scheuer, 1957) with a newspaper billboard from London on 13 May 1955. Credits: Fig. 3, “A survey of radio sources between declinations −38° and+ 83°”, Shakeshaft, J. R., Ryle, M., Baldwin, J. E., Elsmore, B., & Thomson, J. H. (1955). Memoirs of the Royal Astronomical Society, 67, 106; “A statistical method for analysing observations of faint radio stars”, Scheuer, P. A. G. (1957). Proceedings of the Cambridge Philosophical Society 53, pp. 764–773

Figs. 35.5 and 35.6

Radio source counts. Left: Mills Cross compared to Ryle and Scheuer. Right: additional 1030 Mills Cross sources. Credit: Fig. 1 & 2 from “Preliminary Statistics of Discrete Sources obtained with the ‘Mills Cross’”, Pawsey, J. L. (1957). IAU Symposium No 4 Radio Astronomy, ed. van der Hulst, H.C. pp. 228

Shakeshaft and Ryle spoke on behalf of the radio astronomers, while Gold and Bondi responded from the Steady-state camp. Gold praised the quality of the survey but did not miss the chance to revive the earlier controversy with Ryle about whether the radio sources were extragalactic, as well as questioning the current cosmological evidence. Gold wrote:

I have been greatly impressed by this magnificent survey, which has exceeded all expectations of a few years ago. I am also glad to see that there is now agreement that many of these sources are likely to be extragalactic, as I suggested here and elsewhere, with much opposition, four years ago. Mr Ryle then considered that such a suggestion must be based on a misunderstanding of the evidence.¹⁸

If the interpretation which has now been adopted is correct, then it is true that radio observations may make a direct contribution to cosmology. The great distance of the identified Cygnus source is an indication that some other sources may also be far, and some even further than any optically recognizable objects. But on present evidence it is very rash to regard the great majority of weak sources as extremely distant. Yet this is implied in attributing cosmological significance to the curve of the number-intensity relation. A wide spread in the intrinsic intensity of sources would imply a dilution of far with near in the count of the weak sources.¹⁹

At Jodrell Bank, Hanbury Brown and Hazard had detected faint radio emission from nearby “normal” galaxies using the 218-ft fixed reflector. These were faint and close by, but the prevailing view was that fainter radio sources were simply more distant. At the same RAS meeting, Hanbury Brown asked the very good question about the role of the large variation in intrinsic luminosity of different sources based on the cosmological interpretation. Ryle in his Halley lecture had argued that the luminosity of the sources does not matter because sources of different luminosity will all have the same source count slope of −1.5 so the combination of sources of different luminosity will still have a slope of −1.5. However, this argument is only correct if the distribution of intensities is a power law. This would be the case for a Euclidian universe with no boundary, but it does not apply in a real universe at large distances where the curvature of space and time dilation break the power law assumption. Ryle and Scheuer (1955) presented a more sophisticated version of this argument and showed that the luminosity function could not increase the slope, however, a reduction was possible.

In fact, we can now see that the effect of the shape and width of the luminosity function would continue to be either ignored, or blamed for unexpected effects, until it was fully included in the models by Longair (1966) and explained by Von Hoerner (1973). They both showed that it was still the case that either the luminosity or the number of sources had to increase in the distant universe. Von Hoerner (1973) also demonstrated that for some luminosity functions a population of sources need not have an inverse relation between average flux density and distance, i.e. the weaker sources in a survey are not necessarily more distant as was often naively assumed.

Gold continued: “It is a fortunate fact that the Steady-state theory of cosmology is very definite in its observational implications. If that theory is correct, then it must be expected that any error in an observation or its interpretation will lead to discord.”

It is interesting to note that Gold, and some other cosmologists, used the predictive power of the Steady-state theory to assert that the observations were flawed, rather than contemplate abandoning a theory when predictions are not confirmed.²⁰ Proponents of other theories with less clear predictions would not engage in such contentious arguments.

Sullivan (1990, p. 325) reports that in 1976, Ryle told him that he had been unprepared for the intensity of the opposition from the Steady-state adherents. As Ryle recalled to Sullivan:

[The intense controversy] was a considerable shock, because of course the trouble with cosmology up till then was that it had been a playground for mathematicians—"Is space curved this way or that way?"—and all these things. It was nothing very much to do with the real world and observations had never, and apparently would never, make any effect on it. It was a game mathematicians could play, safe from all possible attack.

But the development of the Steady-state model was an important break-through. Here was something that made specific predictions in wide range of not-necessarily-thought-of possible measurements. It said that the universe was in a state that could remain the same through time as well as space … And as soon as you know you can detect sources at redshifts large enough for things to happen on other cosmologies, then you can detect a difference … It was remarkable what an absolute storm it provoked. Well of course, it wasn’t helped by the fact that the press got hold of the story.

The Australian group were taken by surprise as Ryle promoted his new evolutionary paradigm based on the 2C Survey statistics. From May to July 1955, leading up to the conflicts during the August IAU conference at Jodrell Bank, Ryle and Pawsey corresponded about discrepancies in their results and their interpretation; Pawsey in a characteristically cautious, but open-minded style.²¹

On 14 June 1955 Pawsey wrote Ryle after having read The Times newspaper from London:

I recently saw in a cutting from The Times that you have completed all, or at any rate a major part, of the job of analysis of your 81 MHz source records and had described your results at the Halley Lecture [at Oxford on 6 May 1955]. It must have been a tough job with this large number of sources.

The Times article then went on to say that conclusions of major importance followed from the analysis, but I was unable to gather any essential points from the article. So, we in Australia are left highly intrigued by things which are now common gossip in England, and I wondered if you could let us in on the secret. Perhaps you have a spare copy of the lecture …

I shall look forward to seeing you at Jodrell Bank [IAU Symposium on Radio Astronomy] and Dublin [IAU General Assembly].

On 18 June 1955, Ryle answered an earlier letter of 24 May 1955 from Pawsey [which has not been located in the NAA archive]:

Many thanks for your letter of 24 May enclosing a preprint of the catalogue of sources [likely a list of the new Mills Cross sources] …

As soon as I have some copies of the figures, I shall be sending you a manuscript of a paper which has gone to the Royal [Publications of the Royal Society].

On 27 June 1955, Ryle wrote a handwritten letter to Pawsey, now responding to the letter of 14 June:

There was, of course, certainly no intention of keeping Australia out of the picture! In fact, we have intended a more or less simultaneous “release” of the new data coincident with the Halley Lecture. Before that, I am afraid we had been somewhat cagey because of the large number of hungry cosmologists we have prowling around [probably Hoyle, Gold and Bondi— the proponents of the Steady-state theory of the Universe] —we wanted to be able to collect our thoughts a little on the next stage before they pounced!²²

Rather than send you the simplified account given in in the Halley lecture (which you might have punched holes in!), I thought it best to send you the real paper—the latter is not yet common gossip in England.

When you have had a chance to read this, we should very much appreciate your comments— both on what you think of the arguments and on the observational side. I imagine that Mills’s survey is now in a position where we could make a similar analysis—he has in fact probably already done so and it will be most interesting to see what his results show …

We look forward to seeing you and Paul Wild [at Jodrell Bank and the IAU].

In mid-May to mid-June 1955, Pawsey continued his conversations with Ryle, sending some early results from the Mills Cross. Independently, Fred Hoyle wrote Mills asking if his survey was complete at a level that could provide a check on the 2C survey statistics.

Pawsey replied to Ryle once more on 6 July 1955 before his arrival in London on 21 August.

Thank you for your two letters and for the copy of your paper, which arrived today.

The content of the paper is revolutionary. If the data are correct, I think your explanation the most plausible (i.e. the best) and the implications in cosmogony²³ immense.

I immediately checked the current state of the 80 MHz Mills Cross results. The present position is this. We have not yet begun a systematic survey but have a lot of observations of areas chosen on a peculiar basis. In these areas discrete sources have been noted as the records became available. There are 500 or 600 on the list and when these are plotted, log I versus log N they fall on a slightly irregular line of slope of −1.5 which flattens off (sources getting small) [weaker?] at an intensity of about 5 [Jansky ]. Thus, the observations appear not to agree with yours, but we must investigate the situation much more carefully before there is any certainty.

Another count of a small number of sources away from the Milky Way showed a slope steeper than −1.5 but at an intensity less than where you found this.²⁴

This is all mysterious. We will do our best to clarify the issue as soon as possible, I hope before I come over, and shall let you know what eventuates. In the meantime, I don’t need to do more than say our results constitute a case for careful investigation. In this context it will be most helpful to have the full paper which you sent me so that we can compare the really crucial points. Thank you very much for sending it.

The next stage was to occur when Pawsey was at Jodrell Bank in late August 1955. As additional checks were made in Sydney and a larger sample of the southern radio sky was available, the statistical disagreement between the two surveys had increased as the papers from the conference were presented.

1955 IAU Symposium no 4 on Radio Astronomy

See Chap. 26 and NRAO ONLINE.51 for details on the organisation of this first IAU sponsored symposium in the new field of radio astronomy.

This 3-day symposium was held on 25–27 August 1955 at Jodrell Bank. The second day of the programme included a discussion of the statistics and optical identifications of discrete radio sources.²⁵

After some introductory papers on radio source identifications, and a series of papers on the emission mechanism (see Chaps. 26 and 34), there was a short presentation by John Shakeshaft describing the “Cambridge Survey of Radio Sources”. The main fireworks of the morning followed with Ryle’s presentation “The Spatial Distribution of Radio Stars”. This was followed by Pawsey, who presented the competing point of view, on behalf of Mills, from Sydney, “Preliminary Statistics of Discrete Sources Obtained with the ‘Mills Cross’”. Ryle’s presentation was similar to his Halley Lecture²⁶ and also included the P(D) curve (see Figs. 35.3 and 35.4). From the data shown, Ryle asserted again that the steepening of the log N- log S at faint intensities was a real effect, with the spatial density or the luminosity of the sources showing a progressive increase with distance.

Pawsey began his presentation with a stirring declaration: “The statistics of the discrete sources observed in Cambridge and the interpretation given by Ryle and his colleagues constitute one of the most interesting items of recent astronomy. It is therefore of great importance to check the observational data and this can be done [by using the new Mills Cross at 85 MHz]”.

Pawsey explained to the audience that he had received, some months previously, a pre-publication copy of the 2C survey. At that time, the Mills survey had detected 550 sources over a solid angle of about one steradian at 85 MHz; 180 of the sources were well displaced from the galactic plane. In Fig. 35.5, we show his comparison with the Ryle and Scheuer distributions; in Fig. 35.6 we show the additional sources detected (total 1030 sources) just before the Jodrell Bank Symposium; the three Sydney curves showed no deviation from the −3/2 law “which we can be sure is significant”. Pawsey wrote:

There is thus a substantial disagreement between the Cambridge and the preliminary Sydney results, and it seems best to withhold judgement on the more interesting interpretation put forward by Ryle and Scheuer until the Sydney observations are complete. At that stage quite definite conclusions should be reached because the pencil-beam technique used is substantially free from confusion

During the discussion there were heated exchanges between Ryle and both Gold and Bondi, the latter two suggesting that confusion might be a problem with the 2C survey. Gold stated after Pawsey’s talk:

Another way in which a steepened curve could be brought about is by the erroneous judgment of intensity of some of the faint sources. When there are several sources in the beam, it might frequently occur that one is recorded of greater than correct intensity. This would produce an increase in the number in one range of the curve at the expense of a proportionally much smaller decrease in a higher section of the curve.²⁷ An interpretation of that sort would imply that the Cambridge survey is much more liable to such an error, and already at a higher intensity than the Australian one.

Bok (1955, p. 21) reviewing the Jodrell Bank Symposium on Radio Astronomy, was cautious as he summarised his opinions about the conflicts. He was worried about the need for a detailed comparison of the two surveys and the failure of the Cambridge group to secure optical identifications: “the Australian observers were somewhat luckier, with Pawsey reporting 10 normal galaxies, one pair of apparently colliding galaxies, one additional supernova remnant and several [HII regions] … as having marked radio sources at their optical positions.” Bok concluded:

There was considerable discussion—especially later in Dublin—of the possible cosmological consequences of the surprisingly large numbers of faint sources in the Cavendish statistics, but in view of the rather different Australian results and the lack of positive identifications with optical objects, the time does not seem ripe for such speculations. We are still very far from a real understanding of the nature of the faint sources. Ryle considers that they are mostly galaxies (possibly like the faint colliding pair 200 million light-years²⁸ from the sun that is responsible for the very strong Cygnus A source), and practically all of them beyond the reach of even the 200-inch telescope.

In retrospect we can see that contrary to Bok’s, and indeed everyone’s, assumptions, identifications with optically known objects did not help the early radio astronomers understand their sources better—instead, these identifications led Mills astray! With typical position errors of fractions of degrees in 1955, the only secure identifications were with nearby galaxies. By chance this meant that a few Southern sources were able to be identified with nearby galaxies. Unlike Ryle, Mills assumed that most radio sources were relatively nearby, so he expected a Euclidian source count with index −1.5, and interpreted this as no evolution.

But we can see now that Ryle’s assumption that the dominant radio population had luminosities similar to Cygnus A and were at large redshifts was correct; hence comments about the lack of identifications, like those from Bok, were misleading. Lots of identifications with bright optical galaxies would not be expected, and Bok’s concluding comment that “… practically all of them beyond the reach of even the 200-inch telescope” was correct and explains the low identification rate since faint galaxies could not be identified with the poor positions available.²⁹ We can now see that Ryle was fortunate in his choice of Cygnus A as the prototype since this was the correct assumption about the average distance of the radio source population. With this large distance even a source count index of −1.5 required evolution. This assumption of a large average distance and not the bogus steep slope of the 2C source counts, was the key to Cambridge case for evolution.

Pawsey Correspondence with Southworth, Oort and Appleton

Pawsey and Ryle were similar insofar as both were strongly influenced in their interpretation of the data by their knowledge of the reliability of their own instruments (and presumably by human factors such as loyalty to, and defense of, their own group). Thus Pawsey clearly backed Mills and Mills’s results in his letters to other astronomers at the time. For example, on 26 November 1955, George Southworth (see Chap. 11 and NRAO ONLINE.20) had written to Pawsey. Pawsey’s reply included an update on the Mills Cross survey that had begun in 1954:

Our major current program is a survey of the sky with the 85 MHz Mills Cross. [see ESM_35.1.pdf, Radio Source Surveys] … It is working excellently. Incidentally, we are engaged in a hot controversy with Ryle on results. He deduces thoroughly interesting cosmological ideas from results from the recent Cambridge survey with an interferometer, for example, that most of the observed faint radio stars are beyond the limits of the 200-inch telescope. We think his interpretation interesting, but his observations phoney.

A letter from Pawsey to Oort was sent a few weeks later (16 December 1955) with a similar message:

I think the principal gossip from here is that the Cambridge-Sydney controversy over the statistics of discrete sources is reflected in similar discrepancies in comparisons between individual sources. So, the Sydney guess as to the cause of the discrepancies is still inadequate resolution for the Cambridge observations.

On 22 February 1956, Pawsey wrote Appleton³⁰:

Incidentally, as you may know, we are in complete disagreement with Ryle as to the discrete sources. Ryle found from his statistics many more faint sources than would have been expected in a Euclidean universe with a uniform (average) density of radio sources. He drew very pretty cosmological inferences about the days when the universe was young. The disagreement between Mills’s and Ryle’s statistics is paralleled in a sample comparison of individual sources.

On 19 September 1956, Appleton asked Pawsey for advice as he prepared to give the Reith Lecture later in 1956.³¹ Appleton wanted to tell the story of the 2C survey, writing:

As a matter of mere sentiment, I would like the radio people to be right and the quite arrogant theorists wrong. But that merely indicates the need to steel oneself to the discipline of science.

Can you, then, please give me the Australian view, with advice as to what can be said and can’t. You have different parts of the universe to examine, but Ryle says he gets the same result in all directions. (Appleton’s emphasis)

Pawsey replied on 28 September 1956 with a decisive letter.³² He noted that Appleton was dealing with an “intensely controversial question and I should like to suggest that you move with great caution.” Pawsey pointed out that the results of the previous year from Cambridge had major implications but only if the data were correct. However, “the disagreement [of the two surveys] is appalling. One or other of these surveys is completely haywire.” Pawsey then pointed out that Ryle’s data was in error due to confusion with only two beamwidths per source compared to tens of beamwidths per source in the Mills survey from Sydney.³³ The Mills survey showed no excess of sources at the low flux density level over that expected from a “Euclidian universe with a constant space density”. Pawsey showed all the reasons for remaining cautious.

1955–1957: Increased Controversy Between Sydney and Cambridge

After the heated debates of 1955, the Sydney group (Mills, Slee and Hill) continued their southern sky survey, counting sources down to 7 Jansky; 368 sources were detected over an area of about one steradian. The region of declination of +10 to −20 deg. and RA range 0 h to 8 h was chosen since it overlapped with the 2C survey preliminary catalogue sent by Ryle to Mills “for the purpose of checking”.

Publishing a paper in 1957 detailing these results, Mills et al. were even more vehement than at the Jodrell Bank 1955 conference. The source lists were discrepant: “Simple inspection of the maps reveals that the two catalogues are almost completely discordant. The conclusion follows that instrumental effects play a decisive part in determining the positions and intensities of sources in at least one of the surveys.”

The authors have carried out an analysis of the two surveys, concluding that the major limitation was caused by instrumental errors in the 2C survey. The claimed culprit was an effect of the low resolution of the 2C interferometer.

There are two important factors which tend to increase the apparent number of sources with flux densities just above the survey limit … Confusion or blending effects in which sources below the limit cause a random variation in the output … Large chance excursions are then counted as single sources. [Also] the effect of observational selection in the presence of noise; the rapid increase in numbers with decreasing flux density provides many more sources …

The Mills’s and colleagues’ logN-logS curve was almost identical to the one shown in the 1955 Jodrell Bank conference proceedings. The message was clear:

We have shown that in the sample area, which is included in the recent Cambridge catalogue of radio sources, there is a striking disagreement between the two catalogues. Reasons are advanced for supposing that the Cambridge survey is very seriously affected by instrumental effects which have a trivial influence on the Sydney results. We therefore conclude that discrepancies, in the main, reflect errors in the Cambridge catalogue, and accordingly deductions of cosmological interest derived from its analysis are without foundation.

An analysis of our results shows that there is no clear evidence for any effect of cosmological importance in the source counts …

Scientific American Interchange Sept 1956 Between Ryle and Mills

In 1956, Ryle wrote an article, “Radio Galaxies”, for Scientific American in what was probably the first time the scientifically minded public were made aware of this remarkable new window upon the Universe. His focus was, of course, on the great distance of radio galaxies “beyond the range of the 200-inch [Palomar] telescope!”. He presented an account of the 2C survey including the claim that the slope of the log N- log S was very steep, providing evidence for an expanding universe. Ryle noted:

The Cambridge conclusion about the distribution of radio stars in space far beyond our galaxy has been questioned by workers in Australia. A survey with the Mill-cross radio telescope has failed to show a marked excess of faint sources such as was found by the Cambridge group. The Australian survey, however, has not yet covered a large area of the sky and it does indicate that radio stars are not distributed uniformly with distance.³⁴

The article concluded with a preference for an evolutionary cosmology:

If these surveys verify that the density of radio sources in space does indeed increase with distance, they should help to make possible a decision between the evolutionary and Steady-state theories of the universe … Thus, our present conclusions from the radio work at Cambridge support the evolutionary view.

This article—perhaps not surprisingly—made the Sydney-Cambridge disagreements both public and (hence) more acrimonious. Mills, though cautious in interpreting data in general, was as inflexible as Ryle about the validity of his own survey. He also had a strong value for exactitude in undertaking surveys, and this was more important to him than cosmological speculation. The group in Sydney at CSIRO were offended by having what they considered to be superior results publicly dismissed. Mills wrote a two-column rejoinder. Pawsey and Bart Bok, who was still at Harvard but was preparing to move to Australia to the Mt. Stromlo Observatory in January 1957, helped to arrange its publication. Bok wrote to Pawsey on 23 October 1956³⁵ that he had just been in New York at the Scientific American offices and had met the editor, Dennis Flanagan.³⁶ Flanagan was “honoured and delighted to be able to print the Mills letter. They promised to put it in a good conspicuous spot, so that the reader couldn’t possibly miss it.” Mills’s rejoinder was published in the letter section in December 1956 (page 8); Ryle followed with three plus columns and a complex figure on page 10.

Mills’s letter—rather like Gold’s earlier critique of Ryle at the 1955 IAU symposium—accurately questioned the accuracy of Ryle’s observations but then tried to justify his own results with an erroneous assumption. Beginning with a precise description of the large discrepancies between the two catalogues (“it is obvious that at least one of the catalogues is hopelessly wrong”), he explained how “difficulties might be expected to arise when two or more radio sources are sufficiently close together for a radio telescope to respond to each simultaneously” and noted that this will occur frequently with the Cambridge instrument. After this persuasive argument that it was almost certainly the 2C catalogue that was in error (which was true), he destroyed his own case with a circular argument that the Mills catalogue must be right because it was consistent with a static Euclidean Universe!

Ryle’s rejoinder in the Scientific American began mildly:

In a subject which is advancing as rapidly as radio astronomy it is natural that there should be differences of opinion from time to time; in some cases these differences may be finally resolved only by more detailed observations. It is possible that in the present case a definite conclusion may have to await the completion of the new survey of greater resolving power which is now in progress at Cambridge on a frequency of 160 MHz.³⁷

Ryle stated that preliminary results from the 3C survey agreed fairly well with the 2C sources. But then he made the point that was key to the whole dispute: “Errors in the positions of individual sources have no effect on the estimated number of sources falling within a given intensity range.”

In other words, Ryle was saying that it didn’t matter if the sources in the 2C catalogue were in the wrong place because incorrect positions did not affect the slope of the source counts—a statistical result, and one that requires evolution. Ryle was able to make this argument because of Scheuer’s P(D) analysis (see next section), something that would have been too difficult to explain in the Scientific American. Mills had been making a much more straightforward assertion that if the catalogues don’t agree, you can’t trust the results; Ryle was changing the question from a focus on understanding individual sources primarily from their position, to what can be learned from the distribution of amplitudes of a population of sources.³⁸

Ryle also asserted that the flatter curve from Sydney was influenced by a contamination of flatter spectrum galactic sources; subsequent data from Sydney showed that this was not the case.

Ryle’s concluding remarks: “We look forward to seeing further details of Mills’s survey, but in the meantime, we feel that our conclusions concerning the spatial distribution of sources of small angular size are substantially correct.”

In later reflections, Mills told the author (WMG) that he now regretted that he had entered the Scientific American exchange which had further inflamed the dispute and had resolved nothing.³⁹

Probability of a Deflection—P(D) Analysis

While Ryle consistently used the erroneous steep slope of the 2C catalogue to argue for evolution, his young Cambridge collaborator, Peter Scheuer (see Chap. 39), had realised in 1954 that an analysis of the population based on the statistics of the observed interferometer deflections would be more effective than using the sources listed in the 2C catalogue. Scheuer was well aware that confusion from multiple sources was resulting in serious errors in the 2C catalogue; but he also realised that his statistical analysis of the raw interferometer deflections could provide information on the population which allowed for the effects of overlapping sources and eliminated the subjective element when identifying individual sources.

Ryle was distrustful of theory and still preferred to use direct counts of sources in the catalogue, but Scheuer’s P(D) statistical analysis supported his case for an evolving population. When reflecting on the controversy over source counts, Scheuer (1990) provided a succinct description of his invention and utilisation of this “Probability of a Deflection” P(D) method of analysis and also his assessment of the view of the Cambridge group at the time:

One models the probability distribution P(D) of deflection amplitudes D on the interferometer records to be expected for sources with a given N(>S) - S relation sprinkled in random positions over the sky and compares this directly with the histogram of observed interferometer amplitudes. This eliminates the subjective element in extracting individual sources from the records … On the one hand, the P(D) analysis confirmed that the observations required a log N(>S)-log S relation with a slope steeper than −1.5, and that was very important in the long controversy that followed. On the other hand, it indicated a slope much less steep than the slope of −3 that came from the [2C catalogue] source counts. The second conclusion was almost as unwelcome to the rest of the Cambridge group as the first was reassuring [our emphasis].

The statistical P(D) approach was underlying Ryle’s comment in his Scientific American letter when he argued that a determination of the statistics of the source intensities was more important than having the correct positions of the sources in a catalogue. However, the P(D) analysis was always inconsistent with the very steep slope initially claimed by Ryle; but this detail was omitted in most presentations Ryle made in this era. While Ryle included use of the P(D) analysis in the 1955 paper interpreting the 2C survey results, he had no enthusiasm for the more theoretical approach and preferred to base his argument on the properties of the individual sources in the catalogue. As we have indicated, the positions of many of these sources were wrong.

In retrospect we can now see that if Ryle had only admitted that his original 2C catalogue and the resulting steep source counts was seriously flawed, much of the acrimony in the ensuing controversy between Sydney and Cambridge would have been avoided. The case against the Steady-state cosmology would have still been valid.

Due to undertaking military service, Scheuer did not publish his P(D) methods until 1957, at the time of the first publication of the preliminary Mills Cross catalogue of the first 383 sources. Therefore, Mills and Slee were only aware of the assertions based on use of the P(D) analysis from the 1955 Ryle and Scheuer publication (above) but not the methodology used. Two footnotes in Mills and Slee (1957) dealt with the statistical analysis method:

Ryle and Scheuer (1955) give curves which they have derived from this probability distribution with various types of source distribution, but no flux density scales are appended, and no details of the calculations are given, so that we are unable to check their correctness.

Note added in Proof: We have just received from PAG Scheuer a copy of his paper (1957 in the Proceedings of the Cambridge Philosophical Society) giving the theoretical derivations on which the probability distributions curves are based; however, we have not been able to compare the curves directly with our own results because of the lack of essential numerical data. [No details of the nature of the missing data were given.⁴⁰]

Scheuer’s publication of the P(D) method did not enable Mills to improve the analysis of his own survey. We now know that the process would have been ineffective since it only works for surveys which are close to, or at, the confusion limit. The 2C survey was close to the confusion limit in contrast to the Mills Cross survey. Mills himself summarises the reasons for his continued support for a non-evolving (Euclidean) interpretation of his source counts in his review paper in Annual Review of Astronomy and Astrophysics (2006). Mills gives a positive view of the P(D) analysis process and goes on to explain why his interpretation of the source counts was flawed:

Although the Cambridge 2C catalogue was largely useless, an important result had been obtained from an analysis of the statistics of the interferometer output (from Scheuer, 1957). This showed that the observed radio emission could not have originated in a population of unresolved discrete radio sources randomly distributed throughout a nonevolving universe. My view expressed at the time was that many of the stronger sources would have been resolved by the interferometer [this turned out to not be the case], producing smaller output deflections, and it seemed likely that the distribution of the sources was not random, so that nothing could be said directly about evolution.

We now know that the Cambridge assumptions, both about the large distances and the minor effects of extended sources, were correct. Thus Mills’s remaining reservations about the use of P(D) were unfounded. This was perhaps more a result of good luck than an incisive understanding of the radio population.

Pawsey’s Matthew Flinders Lecture

Pawsey was aware of the importance of the bigger picture, while Mills remained focused on defending the accuracy of his catalogue. In 1957, Pawsey was asked to present the prestigious first Matthew Flinders Lecture at the Australian Academy of Science⁴¹ which was a tribute to his standing in the scientific community. In the talk, he reviewed all of radio astronomy at that time, and his summary picked up on Ryle’s interest in the possibilities for radio research, as we can see from the following comment on the radio continuum surveys and cosmology:

Radio studies of “discrete sources” [not radio stars!] have been reported which appeared to show an ‘edge of the universe’ effect but we now consider the observations to be at fault … In the field of cosmology, then, radio astronomy is in a tantalising position. There is every reason to suppose that some galaxies which are far beyond the current optical limits are visible to radio telescopes. But so far we have been unable to make use of this potential source of information because of lack of detail in radio observations. This position has been aggravated by the disagreement between Cambridge and Sydney observations after the raising of high hopes by the Cambridge work. But there is every reason to suppose that improved radio telescopes and more critical analysis will take the effective radio horizon out to distances where the recession velocity approaches the velocity of light and decisive results on “world models” can be obtained.⁴²

Although Pawsey remained a strong critic of the 2C catalogue and sceptical of Ryle’s cosmological inferences, he was certainly not blind to the far reaching possibilities raised by Cambridge. By 1957 these ideas were influencing his planning for instruments and research directions, particularly in relation to the question of investment in a large dish “Giant Radio Telescope” (see Chap. 27). His analysis of the performance of radio telescopes for observations of extremely distant objects is included in ESM_35.2.pdf, Surveys with arrays.

AAS Symposium on “Radio Sources Outside Our Galaxy”

The recognition of the increasingly important role being played by the radio astronomers captured the attention of the optical astronomers and cosmologists in the US. After World War II the US had fallen behind Australia and the UK in observational radio astronomy and that was now starting to change. The group at Harvard had detected the 21 cm radio emission line from neutral hydrogen, the Naval Research Laboratory was operating a 60-foot dish and going to higher radio frequencies than other groups. The US had also recognised the extra-ordinary opportunities to embark on radio astronomy research as “big science” with the genesis of the National Radio Astronomy Observatory (Kellermann, Bouton, & Brandt, 2020).

In this environment the US astronomy community held a symposium at the Urbana, Illinois, meeting of the American Astronomical Society on 20 Aug 1957. A remarkable collection of papers, which are not at all well-known today, were included in this small symposium volume published by the Astronomical Society of the Pacific in 1958. This symposium heralded the beginning of the era of rapprochement between the observational groups, although the acrimonious public debate was to continue, culminating the infamous Paris IAU symposium the following year, which we discuss in the next chapter.

In Illinois, Pawsey presented the status of the “Sydney Investigations and Very Distant Radio Sources” based on the report discussed in the last section. Remarkably, this paper also included one of the earliest numerical simulations of the effects of radio source confusion using realistic beam models. The simulations were carried out by Mullaly and T. Pearcey (unpublished) using Australia’s CSIRAC digital electronic computer (McCann & Thorne, 2000), one of the few astronomical projects for which CSIRAC was used. For the simulation they generated an artificial sky with a random distribution of point sources in a static Euclidean universe. Fig. 1 from Pawsey (1958) is shown below (Fig. 35.7). It clearly demonstrates the effects of confusion which can cause sources to be lost or have the wrong amplitude and position. Below about 10 or 20 beam areas per source was a “dangerous level” where peaks could disagree seriously with the real sources. Pawsey compared this to the Sydney and Cambridge (2C) surveys and concluded that the Sydney survey just avoided this limit (beams per source), but the Cambridge survey was well below it, possibly by a factor of 10.

Fig. 35.7

Examples of the trace resulting from the passage of the ideal beam shown over randomly distributed point sources. The dotted line indicates a level below which serious blending effects occur. Credit: Fig. 1, “Sydney Investigations and Very Distant Radio Sources”, Pawsey, J. L. (1958). Publications of the Astronomical Society of the Pacific 70, pp. 133–140. All rights reserved

Hewish, from the Cavendish Laboratory in Cambridge reported on “The Distribution of Radio Stars” based on the new 3C catalog. Hewish would have been privy to discussions (eg by Scheuer) of the problems with the 2C survey, and now he admitted the serious confusion errors in that catalogue. Minkowski, from the Mount Wilson and Palomar Observatories, discussed “The Problem of the Identification of Extragalactic Radio Sources” and McVittie, a cosmologist from the University of Illinois, discussed the cosmological implications.

By the end of the meeting it was clear that the major disagreements between the surveys had been resolved. It was clear, and accepted by Hewish, that the 2C Survey was limited by source confusion but Hewish now had the much superior 3C Survey results at hand; the evidence for evolution had persisted but at a much more moderate level. This harmony may have only been possible due to the fact that two surrogates for the strong personalities were involved. Hewish represented Ryle and Pawsey represented Mills.

Following this meeting Pawsey wrote Bowen a letter from the US (received in Sydney 28 August 1957)⁴³:

The third [Cambridge] survey, called 3C, agrees much with the Mills survey. In fact, there are about 30 per cent coincidences … According to Mills and me [Pawsey], confusion due to background sources becomes very serious at the level corresponding to 10 or 20 beam areas per recorded source for a pencil beam, and probably about the same for an interferometer. I therefore conclude that the 3C survey probably contains about 100 sources per steradian which are reasonable, the rest are probably phoney … You will be pleased to know that Hewish admits in public that the original 2C survey was over-interpreted.