The entire southern sky (Declination, $\delta <$ 30 deg) has been observed using the Murchison Widefield Array (MWA), which provides radio imaging of $\sim$2-arcmin resolution at low frequencies (72-231 MHz). This is the GaLactic and Extragalactic All-sky MWA (GLEAM) Survey, and we have previously used a combination of visual inspection, cross-checks against the literature, and internal matching to identify the 'brightest' radio-sources ($S_{\mathrm{151MHz}} >$ 4 Jy) in the extragalactic catalogue (Galactic latitude, $|b| >$ 10 deg). We refer to these 1,863 sources as the GLEAM 4-Jy (G4Jy) Sample, and use radio images (of $\leq$ 45-arcsec resolution), and multi-wavelength information, to assess their morphology and identify the galaxy that is hosting the radio emission (where appropriate). Details of how to access all of the overlays used for this work are available at https://github.com/svw26/G4Jy. Alongside this we conduct further checks against the literature, which we document in this paper for individual sources. Whilst the vast majority of the G4Jy Sample are active galactic nuclei with powerful radio-jets, we highlight that it also contains a nebula, two nearby, star-forming galaxies, a cluster relic, and a cluster halo. There are also three extended sources for which we are unable to infer the mechanism that gives rise to the low-frequency emission. In the G4Jy catalogue we provide mid-infrared identifications for 86% of the sources, and flag the remainder as: having an uncertain identification (129 sources), having a faint/uncharacterised mid-infrared host (126 sources), or it being inappropriate to specify a host (2 sources). For the subset of 129 sources, there is ambiguity concerning candidate host-galaxies, and this includes four sources (B0424$-$728, B0703$-$451, 3C 198, and 3C 403.1) where we question the existing identification.

The Murchison Widefield Array (MWA) has observed the entire southern sky (Declination, $\delta <$ 30 deg) at low radio-frequencies, over the range 72-231 MHz. These observations constitute the GaLactic and Extragalactic All-sky MWA (GLEAM) Survey, and we use the extragalactic catalogue (Galactic latitude, $|b| >$ 10 deg) to define the GLEAM 4-Jy (G4Jy) Sample. This is a complete sample of the 'brightest' radio-sources ($S_{\mathrm{151MHz}} >$ 4 Jy), the majority of which are active galactic nuclei with powerful radio-jets. Crucially, low-frequency observations allow the selection of such sources in an orientation-independent way (i.e. minimising the bias caused by Doppler boosting, inherent in high-frequency surveys). We then use higher-resolution radio images, and information at other wavelengths, to morphologically classify the brightest components in GLEAM. We also conduct cross-checks against the literature, and perform internal matching, in order to improve sample completeness (which is estimated to be $>$ 95.5%). This results in a catalogue of 1,863 sources, making the G4Jy Sample over 10 times larger than that of the revised Third Cambridge Catalogue of Radio Sources (3CRR; $S_{\mathrm{178MHz}} >$ 10.9 Jy). Of these G4Jy sources, 78 are resolved by the MWA (Phase-I) synthesised beam ($\sim$2 arcmin at 200 MHz), and we label 67% of the sample as 'single', 26% as 'double', 4% as 'triple', and 3% as having 'complex' morphology at $\sim$1 GHz (45-arcsec resolution). Alongside this, our value-added catalogue provides mid-infrared source associations (subject to 6-arcsec resolution at 3.4 micron) for the radio emission, as identified through visual inspection and thorough checks against the literature. As such, the G4Jy Sample can be used as a reliable training set for cross-identification via machine-learning algorithms. [Abstract abridged for arXiv submission.]

Past studies of compact active galactic nuclei (AGNs), the dominant population at high radio frequencies, selected them using flat spectral index criteria. This biases the sample due to the steepening of AGN spectra at high radio frequencies. We improve upon this by selecting 3610 compact AGNs using their angular size information ($\sim$0.15 arcsec scale) from the Australia Telescope 20 GHz (AT20G) high-angular-resolution catalogue. We cross-match these against the Wide-field Infrared Survey Explorer All-WISE catalogue and present a catalogue with 3300 (91%) matches, 91 (3%) rejects and 219 (6%) nondetections that are excellent high redshift candidates. Of the matched compact AGNs, 92% exhibit QSO mid-infrared colours (W1-W2>0.5). Therefore, our sample of high frequency compact sources has a very high rate of identification with mid-infrared QSOs. We find counterparts for 88% of 387 compact steep-spectrum (CSS) sources in the AT20G survey, 82%$\pm$5% of which exhibit QSO mid-infrared colours and have moderate redshifts (median redshift = 0.82), while those dominated by host galaxy colours in mid-infrared have lower redshifts (median redshift = 0.13). The latter classified into late- and early-type galaxies using their mid-infrared colours shows a majority (68%$\pm$4%) have colours characteristic of late-type galaxies. Thus, we find that a larger fraction of these CSS sources are embedded in hosts with higher gas densities than average early-type galaxies. We compare mid-infrared colours of our AGNs against those reported for AGNs primarily selected using non-radio techniques. This shows that mid-infrared SED of high frequency selected compact radio AGN is comparatively less red, possibly due to contributions from their hosts.

Around 10% of bright low-frequency radio sources observed with the Murchison Widefield Array (MWA) show strong Interplanetary Scintillation (IPS) on timescales of a few seconds, implying that almost all their low-frequency radio emission comes from a compact component less than 0.5 arcsec in angular size. Most of these objects are compact steep-spectrum (CSS) or MHz-peaked spectrum (MPS) radio sources. We have used mid-infrared data from the Wide-field Infrared Survey Explorer (WISE) catalogue to search for the host galaxies of 65 strongly-scintillating MWA sources and compare their properties with those of the overall population of bright low-frequency radio sources. We identified WISE mid-infrared counterparts for 91% of the bright sources in a single 900 square degree MWA field, and found that the hosts of the strongly-scintillating sources were typically at least 1 mag fainter in the WISE W1 (3.4 micron) band than the hosts of weakly-scintillating MWA sources of similar radio flux density. This difference arises mainly because the strongly-scintillating sources are more distant. We estimate that strongly-scintillating MWA sources have a median redshift of z ~ 1.5, and that at least 30% of them are likely to lie at z > 2. The recently-developed wide-field IPS technique therefore has the potential to provide a powerful new tool for identifying high-redshift radio galaxies without the need for radio spectral-index selection.

We report on a search for low-frequency radio variability in 944 bright (> 4Jy at 154 MHz) unresolved, extragalactic radio sources monitored monthly for several years with the Murchison Widefield Array. In the majority of sources we find very low levels of variability with typical modulation indices < 5%. We detect 15 candidate low frequency variables that show significant long term variability (>2.8 years) with time-averaged modulation indices M = 3.1 - 7.1%. With 7/15 of these variable sources having peaked spectral energy distributions, and only 5.7% of the overall sample having peaked spectra, we find an increase in the prevalence of variability in this spectral class. We conclude that the variability seen in this survey is most probably a consequence of refractive interstellar scintillation and that these objects must have the majority of their flux density contained within angular diameters less than 50 milli-arcsec (which we support with multi-wavelength data). At 154 MHz we demonstrate that interstellar scintillation time-scales become long (~decades) and have low modulation indices, whilst synchrotron driven variability can only produce dynamic changes on time-scales of hundreds of years, with flux density changes less than one milli-jansky (without relativistic boosting). From this work we infer that the low frequency extra-galactic southern sky, as seen by SKA-Low, will be non-variable on time-scales shorter than one year.

Powerful radio-galaxies feature heavily in our understanding of galaxy evolution. However, when it comes to studying their properties as a function of redshift and/or environment, the most-detailed studies tend to be limited by small-number statistics. During Focus Meeting 3, on "Radio Galaxies: Resolving the AGN phenomenon", SVW presented a new sample of nearly 2,000 of the brightest radio-sources in the southern sky (Dec. $<$ 30 deg). These were observed at low radio-frequencies as part of the GaLactic and Extragalactic All-sky MWA (GLEAM) Survey, which is a continuum survey conducted using the Murchison Widefield Array (MWA). This instrument is the precursor telescope for the low-frequency component of the Square Kilometre Array, and allows us to select radio galaxies in an orientation-independent way (i.e. minimising the bias caused by Doppler boosting, inherent in high-frequency surveys). Being brighter than 4 Jy at 151 MHz, we refer to these objects as the GLEAM 4-Jy (G4Jy) Sample. The G4Jy catalogue is close to being finalised, with SVW describing how multi-wavelength data have been used to determine the morphology of the radio emission, and identify the host galaxy. In addition, the MWA's excellent spectral-coverage and sensitivity to extended/diffuse emission were highlighted. Both of these aspects are important for understanding the physical mechanisms that take place within active galaxies, and how they interact with their environment.

We describe the parameters of a low-frequency all-sky survey of compact radio sources using Interplanetary Scintillation (IPS), undertaken with the Murchison Widefield Array (MWA). While this survey gives important complementary information to low-resolution survey such as the MWA GLEAM survey, providing information on the subarsecond structure of every source, a survey of this kind has not been attempted in the era of low-frequency imaging arrays such as the MWA and LOFAR. Here we set out the capabilities of such a survey, describing the limitations imposed by the heliocentric observing geometry and by the instrument itself. We demonstrate the potential for IPS measurements at any point on the celestial sphere and we show that at 160MHz, reasonable results can be obtained within 30deg of the ecliptic (2{\pi} str: half the sky). We also suggest some observational strategies and describe the first such survey, the MWA Phase I IPS survey. Finally we analyse the potential of the recently-upgraded MWA and discuss the potential of the SKA-low to use IPS to probe sub-mJy flux density levels at sub-arcsecond angular resolution.

We use Murchison Widefield Array observations of interplanetary scintillation (IPS) to determine the source counts of point ($<$0.3 arcsecond extent) sources and of all sources with some subarcsecond structure, at 162 MHz. We have developed the methodology to derive these counts directly from the IPS observables, while taking into account changes in sensitivity across the survey area. The counts of sources with compact structure follow the behaviour of the dominant source population above $\sim$3 Jy but below this they show Euclidean behaviour. We compare our counts to those predicted by simulations and find a good agreement for our counts of sources with compact structure, but significant disagreement for point source counts. Using low radio frequency SEDs from the GLEAM survey, we classify point sources as Compact Steep-Spectrum (CSS), flat spectrum, or peaked. If we consider the CSS sources to be the more evolved counterparts of the peaked sources, the two categories combined comprise approximately 80% of the point source population. We calculate densities of potential calibrators brighter than 0.4 Jy at low frequencies and find 0.2 sources per square degrees for point sources, rising to 0.7 sources per square degree if sources with more complex arcsecond structure are included. We extrapolate to estimate 4.6 sources per square degrees at 0.04 Jy. We find that a peaked spectrum is an excellent predictor for compactness at low frequencies, increasing the number of good calibrators by a factor of three compared to the usual flat spectrum criterion.

We report the first astrophysical application of the technique of wide-field Interplanetary Scintillation (IPS) with the Murchison Widefield Array (MWA). This powerful technique allows us to identify and measure sub-arcsecond compact components in low-frequency radio sources across large areas of sky without the need for long-baseline interferometry or ionospheric calibration. We present the results of a five-minute observation of a 30x30 sq. deg MWA field at 162 MHz with 0.5 second time resolution. Of the 2550 continuum sources detected in this field, 302 (12 per cent) show rapid fluctuations caused by IPS. We find that at least 32% of bright low-frequency radio sources contain a sub-arcsec compact component that contributes over 40% of the total flux density. Perhaps surprisingly, peaked-spectrum radio sources are the dominant population among the strongly-scintillating, low-frequency sources in our sample. While gamma-ray AGN are generally compact, flat-spectrum radio sources at higher frequencies, the 162 MHz properties of many of the Fermi blazars in our field are consistent with a compact component embedded within more extended low-frequency emission. The detection of a known pulsar in our field shows that the wide-field IPS technique is at the threshold of sensitivity needed to detect new pulsars using image plane analysis, and scaling the current MWA sensitivity to that expected for SKA-low implies that large IPS-based pulsar searches will be feasible with SKA. Calibration strategies for the SKA require a better knowledge of the space density of compact sources at low radio frequencies, which IPS observations can now provide.

We present the first dedicated observations of Interplanetary Scintillation (IPS) with the Murchison Widefield Array (MWA). We have developed a synthesis imaging technique, tailored to the properties of modern "large-N" low-frequency radio telescopes. This allows us to image the variability on IPS timescales across 900 square degrees simultaneously. We show that for our observations, a sampling rate of just 2Hz is sufficient to resolve the IPS signature of most sources. We develop tests to ensure that IPS variability is separated from ionospheric or instrumental variability. We validate our results by comparison with existing catalogues of IPS sources, and near-contemporaneous observations by other IPS facilities. Using just five minutes of data, we produce catalogues at both 79MHz and 158MHz, each containing over 350 scintillating sources. At the field centre we detect approximately one scintillating source per square degree, with a minimum scintillating flux density at 158MHz of 110mJy, corresponding to a compact flux density of approximately 400mJy. Each of these sources is a known radio source, however only a minority were previously known to contain sub-arcsecond components. We discuss our findings and the prospects they hold for future astrophysical and heliospheric studies.

We present a source catalogue and first results from a deep, blind radio survey carried out at 20 GHz with the Australia Telescope Compact Array, with follow-up observations at 5.5, 9 and 18 GHz. The Australia Telescope 20 GHz (AT20G) deep pilot survey covers a total area of 5 deg^2 in the Chandra Deep Field South and in Stripe 82 of the Sloan Digital Sky Survey. We estimate the survey to be 90% complete above 2.5 mJy. Of the 85 sources detected, 55% have steep spectra (alpha_{1.4}^{20} < -0.5) and 45% have flat or inverted spectra (alpha_{1.4}^{20} >= -0.5). The steep-spectrum sources tend to have single power-law spectra between 1.4 and 18 GHz, while the spectral indices of the flat- or inverted-spectrum sources tend to steepen with frequency. Among the 18 inverted-spectrum (alpha_{1.4}^{20} >= 0.0) sources, 10 have clearly defined peaks in their spectra with alpha_{1.4}^{5.5} > 0.15 and alpha_{9}^{18} < -0.15. On a 3-yr timescale, at least 10 sources varied by more than 15% at 20 GHz, showing that variability is still common at the low flux densities probed by the AT20G-deep pilot survey. We find a strong and puzzling shift in the typical spectral index of the 15-20 GHz source population when combining data from the AT20G, Ninth Cambridge and Tenth Cambridge surveys: there is a shift towards a steeper-spectrum population when going from ~1 Jy to ~5 mJy, which is followed by a shift back towards a flatter-spectrum population below ~5 mJy. The 5-GHz source-count model by Jackson & Wall (1999), which only includes contributions from FRI and FRII sources, and star-forming galaxies, does not reproduce the observed flattening of the flat-spectrum counts below ~5 mJy. It is therefore possible that another population of sources is contributing to this effect.

We present polarisation data for 180 extragalactic sources extracted from the Australia Telescope 20 GHz (AT20G) survey catalog, and observed with the Australia Telescope Compact Array during a dedicated, high sensitivity run. For the sake of completeness we extracted the polarisation information for 7 extended sources from the 9-yr WMAP coadded maps at 23 GHz. The full sample of 187 sources constitutes a 99% complete sample of extragalactic sources brighter than S(20 GHz)=500mJy at the selection epoch with declination <-30 deg. The sample has a 91.4% detection rate in polarisation at 20GHz (94% if considering the sub-sample of point like sources). We have measurements also at 4.8 and 8.6 GHz within 1 month of the 20GHz observations for 172 sources to reconstruct the spectral properties of the sample in total intensity and in polarisation: 143 of them have a polarisation detection at all three frequencies. We find that there is no statistically significant evidence of a relationship either between the fraction of polarisation and frequency or between the fraction of polarisation and the total intensity flux density. This indicates that Faraday depolarisation is not very important above 4.8 GHz and that the magnetic field is not substantially more ordered in the regions dominating the emission at higher frequencies (up to 20 GHz). We estimate the distribution of the polarisation fraction and the polarised flux density source counts at 20GHz.

We present the high angular resolution catalogue for the Australia Telescope 20 GHz (AT20G) survey, using the high angular resolution 6-km antenna data at the baselines of ~ 4500 m of the Australia Telescope Compact Array (ATCA). We have used the data to produce the visibility catalogue that separates the compact Active Galactic Nuclei (AGNs) from the extended radio sources at the 0.15 arcsec angular scale, corresponding to the linear size scale of 1 kpc at redshifts higher than 0.7. We find the radio population at 20 GHz to be dominated by compact AGNs constituting 77% of the total sources in the AT20G. We introduce the visibility-spectra diagnostic plot, produced using the AT20G cross-matches with lower frequency radio surveys at 1 GHz (the NRAO-VLA Sky Survey (NVSS) and the Sydney University Molonglo Sky Survey (SUMSS)), that separates the 20 GHz population into distinct sub-populations of the compact AGNs, the compact steep-spectrum sources, the extended AGN-powered sources and extended flat-spectrum sources. The extended flat-spectrum sources include a local thermal emitting population of high latitude planetary nebulae and also gravitational lens and binary black hole candidates among the AGNs. We find a smooth transition in properties between the compact CSS sources and the AGN populations. The visibility catalogue, together with the main AT20G survey, provides an estimate of angular size scales for sources in the AT20G and an estimate of the flux arising from central cores of extended radio sources. The identification of the compact AGNs in the AT20G survey provides high quality calibrators for high frequency radio telescope arrays and VLBI observations.

We cross-matched the 6-year Swift/Burst Alert Telescope (BAT) survey of active galactic nuclei (AGN) with the AT20G radio survey of the southern sky, which is one of the largest high-frequency radio surveys available. With these data we investigated the possible correlation between the radio and the X-ray emission at the highest radio and X-ray frequencies. We found 37 AGN with a high probability of association (>80 per cent), among which 19 are local Seyfert galaxies (with median redshift z = 0.03) and 18 blazars. We found that \approx 20 per cent of the AGN detected in hard X-rays are also bright radio sources at 20 GHz, but the apparent correlation between the radio and hard X-ray luminosity is completely driven by the different median redshifts of the two subgroups of AGN. When we consider only the local Seyfert sample we find no evidence of a correlation between their 20 GHz and 15-55 keV power. Therefore it appears that at high frequencies the radio-X connection, which had been previously observed at lower frequencies, disappears. The disappearance of the radio-X correlation at high radio and X-ray frequencies could be tested through Very Long Baseline Interferometry and the use of the Nuclear Spectroscopic Telescope Array (NuSTAR) satellite.

We use high angular resolution data, measured from visibility of sources at the longest baseline of 4500 m of the Australia Telescope Compact Array (ATCA), for the Australia Telescope 20 GHz (AT20G) survey to obtain angular size information for > 94% of AT20G sources. We confirm the previous AT20G result that due to the high survey frequency of 20 GHz, the source population is strongly dominated by compact sources (79%). At 0.15 arcseconds angular resolution limit, we show a very strong correlation between the compact and extended sources with flat and steep-spectrum sources respectively. Thus, we provide a firm physical basis for the traditional spectral classification into flat and steep-spectrum sources to separate compact and extended sources. We find the cut-off of -0.46 to be optimum for spectral indices between 1 and 5 GHz and, hence, recommend the continued use of -0.5 for future studies. We study the effect of spectral curvature on redshift cut-off of compact AGNs using recently published redshift data. Using spectral indices at different frequencies, we correct for the redshift effect and produce restframe frequency spectra for compact sources for redshift up to ~5. We show that the flat spectra of most compact sources start to steepen at ~30 GHz. At higher frequencies, the spectra of both populations are steep so the use of spectral index does not separate the compact and extended source populations as well as in lower frequencies. We find that due to the spectral steepening, surveys of compact sources at higher frequencies (>5 GHz) will have redshift cut-off due to spectral curvature but at lower frequencies, the surveys are not significantly affected by spectral curvature, thus, the evidence for a strong redshift cut-off in AGNs found in lower frequency surveys is a real cut-off and not a result of K-correction.

Our current understanding of radio-loud AGN comes predominantly from studies at frequencies of 5 GHz and below. With the recent completion of the Australia Telescope 20 GHz (AT20G) survey, we can now gain insight into the high-frequency radio properties of AGN. This paper presents supplementary information on the AT20G sources in the form of optical counterparts and redshifts. Optical counterparts were identified using the SuperCOSMOS database and redshifts were found from either the 6dF Galaxy survey or the literature. We also report 144 new redshifts. For AT20G sources outside the Galactic plane, 78.5% have optical identifications and 30.9% have redshift information. The optical identification rate also increases with increasing flux density. Targets which had optical spectra available were examined to obtain a spectral classification. There appear to be two distinct AT20G populations; the high luminosity quasars that are generally associated with point-source optical counterparts and exhibit strong emission lines in the optical spectrum, and the lower luminosity radio galaxies that are generally associated with passive galaxies in both the optical images and spectroscopic properties. It is suggested that these different populations can be associated with different accretion modes (cold-mode or hot-mode). We find that the cold-mode sources have a steeper spectral index and produce more luminous radio lobes, but generally reside in smaller host galaxies than their hot-mode counterparts. This can be attributed to the fact that they are accreting material more efficiently. Lastly, we compare the AT20G survey with the S-cubed semi-empirical (S3-SEX) models and conclude that the S3-SEX models need refining to correctly model the compact cores of AGN. The AT20G survey provides the ideal sample to do this.

Cold massive cores are one of the earliest manifestations of high mass star formation. Following the detection of SiO emission from G333.125-0.562, a cold massive core, further investigations of the physics, chemistry and dynamics of this object has been carried out. Mopra and NANTEN2 molecular line profile observations, Australia Telescope Compact Array (ATCA) line and continuum emission maps, and Spitzer 24 and 70 \mum images were obtained. These new data further constrain the properties of this prime example of the very early stages of high mass star formation. A model for the source was constructed and compared directly with the molecular line data using a 3D molecular line transfer code - MOLLIE. The ATCA data reveal that G333.125-0.562 is composed of two sources. One of the sources is responsible for the previously detected molecular outflow and is detected in the Spitzer 24 and 70 \mum band data. Turbulent velocity widths are lower than other more active regions of G333 which reflects the younger evolutionary stage and/or lower mass of this core. The molecular line modelling requires abundances of the CO isotopes that strongly imply heavy depletion due to freeze-out of this species onto dust grains. The principal cloud is cold, moderately turbulent and possesses an outflow which indicates the presence of a central driving source. The secondary source could be an even less evolved object as no apparent associations with continuum emissions at (far-)infrared wavelengths.

The Australia Telescope 20 GHz (AT20G) survey is a blind survey of the whole Southern sky at 20 GHz with follow-up observations at 4.8, 8.6, and 20 GHz carried out with the Australia Telescope Compact Array (ATCA). In this paper we present an analysis of radio spectral properties in total intensity and polarisation, sizes, optical identifications, and redshifts of the sample of the 5808 extragalactic sources in the survey catalogue of confirmed sources over the whole Southern sky excluding the strip at Galactic latitude |b|<1.5deg. The sample has a flux density limit of 40 mJy. Completeness has been measured as a function of scan region and flux density. Averaging over the whole survey area the follow-up survey is 78% complete above 50mJy and 93% complete above 100mJy. 3332 sources with declination <-15deg have good quality almost simultaneous observations at 4.8, 8.6, and 20GHz. The spectral analysis shows that the sample is dominated by flat-spectrum sources. The fraction of flat-spectrum sources decreases from 81% for 20GHz flux densities S>500mJy, to 60% for S<100mJy. There is also a clear spectral steepening at higher frequencies with the median spectral index decreasing from -0.16 between 4.8 and 8.6GHz to -0.28 between 8.6 and 20GHz. Simultaneous observations in polarisation are available for all the sources at all the frequencies. 768 sources have a good quality detection of polarised flux density at 20GHz; 467 of them were also detected in polarisation at 4.8 and/or at 8.6GHz so that it has been possible to compare the spectral behaviour in total intensity and polarisation. We have found that the polarised fraction increases slightly with frequency and decreases with flux density. Cross matches and comparisons have been made with other catalogues at lower radio frequencies, and in the optical, X-ray and gamma-ray bands. Redshift estimates are available for 825 sources.

We present the full source catalogue from the Australia Telescope 20 GHz (AT20G) Survey. The AT20G is a blind radio survey carried out at 20 GHz with the Australia Telescope Compact Array (ATCA) from 2004 to 2008, and covers the whole sky south of declination 0 deg. The AT20G source catalogue presented here is an order of magnitude larger than any previous catalogue of high-frequency radio sources, and includes 5890 sources above a 20 GHz flux-density limit of 40 mJy. All AT20G sources have total intensity and polarisation measured at 20 GHz, and most sources south of declination -15 deg also have near-simultaneous flux-density measurements at 5 and 8 GHz. A total of 1559 sources were detected in polarised total intensity at one or more of the three frequencies. We detect a small but significant population of non-thermal sources that are either undetected or have only weak detections in low-frequency catalogues. We introduce the term Ultra-Inverted Spectrum (UIS) to describe these radio sources, which have a spectral index alpha(5, 20) > +0.7 and which constitute roughly 1.2 per cent of the AT20G sample. The 20 GHz flux densities measured for the strongest AT20G sources are in excellent agreement with the WMAP 5-year source catalogue of Wright et al. (2009), and we find that the WMAP source catalogue is close to complete for sources stronger than 1.5 Jy at 23 GHz.