The first two dedicated gamma-ray astronomical satellites, SAS 2 and COS B, yielded between them one identified extragalactic source of greater than 100 MeV gamma-rays, the radio source 3C273 [Bignami & Hermsen 1983]. The Compton Gamma-Ray Observatory (CGRO) was launched in 1991 April. Of the four detectors on board the Observatory, the Energetic Gamma-Ray Experiment Telescope (EGRET) is sensitive to the highest energy gamma-rays, those in the 20 MeV to 30 GeV range. To date, CGRO has discovered over 120 discrete sources of greater than 100 MeV gamma-ray emission.
Of these discrete sources the second EGRET Source Catalog, compiled from observations during phases I and II of the CGRO mission (from 1991 April to 1993 July), lists 40 high confidence identifications of strong, flat-spectrum extragalactic radio sources, and a further 11 marginal identifications [Thompson et al. 1995]. Identifications were classified as marginal if the candidate radio counterpart lies close to but outside the 95% uncertainty contour for the gamma-ray source position. Previously, the high confidence and marginal classifications were based solely on photon statistics [von Montigny et al. 1995a]. To avoid confusion, the following terminology will be used here: strong and weak describe the statistical significance of detection, and high-confidence and marginal describe the identifications. The rapid optical variability and large optical polarisation of many of the optical counterparts to the radio sources have resulted in their being classified as `blazars'.
The absolute gamma-ray luminosities for some EGRET sources are
exceedingly high ( 
10 
ergs s 
, approximately the
Eddington limit for a 
black hole) if isotropic
emission is assumed. However, there are reasons for concluding that
the emission is not isotropic. A number of the radio counterparts have
already been observed with VLBI.
These sources have generally exhibited apparent superluminal motions,
suggesting beamed emission from relativistic matter travelling close to
our line of sight. If the gamma-ray flux is
also beamed along our line of sight then the luminosities derived by
assuming isotropic emission will be overestimates.
The short time-scale variability of the gamma-ray emission from some of the stronger sources implies that the spatial extent of the gamma-ray emission region is on the order of 10 light days or less [von Montigny et al. 1995a].
From consideration of these two points it is generally postulated that the gamma-ray emission region lies within the base of the relativistic jet (c.f. references in von Montigny et al. 1995a).
The fact that many flat-spectrum radio sources have not been identified by EGRET and the assumption that
the gamma-ray emission is beamed has been used to infer that the width
of the gamma-ray beam is smaller than that of the radio beam, under a
``unified scheme'' where all bright, flat-spectrum radio
sources are gamma-ray sources. Adopting a beam size for the radio
emission of 14 
[Padovani & Urry 1992], Salamon & Stecker [1994] derive a
beam size for the gamma ray emission of 4 . However,
Dondi & Ghisellini [1995] argue that the radio and gamma-ray beams
may be collimated to the same extent and suggest that the EGRET
sources have been detected because they are currently in a high state of
gamma-ray emission, thus leaving undetected those sources currently in a low state of gamma-ray emission.
von Montigny et al. [1995b] have offered another alternative, which also relies on relativistic beaming, to explain why some radio sources have been detected in gamma-rays but others have not. von Montigny et al. [1995b] suggest that gamma-ray sources may have jets which remain straight from the scale of the gamma-ray emission region all the way to the kpc-scale and that we lie within the gamma-ray beaming cone as well as the radio beaming cone. On the other hand, jets which bend allow for the possibility that the gamma-ray emission is beamed away from our line of sight but that we still lie within the beaming cone of the radio emission. The implication of this suggestion is that the majority of gamma-ray loud sources have closely aligned small and large-scale jets whereas the majority of gamma-ray quiet sources have misaligned small and large-scale jets.
Indicators of relativistic beaming are therefore very relevant for models of
the gamma-ray emission. For example, the models proposed by Salamon &
Stecker [1994] and Dondi & Ghisellini [1995] make quite different
predictions about the distribution of line of sight angles of the
relativistic jet in EGRET sources. Salamon & Stecker predict that all
gamma-ray sources have their jets aligned within 4 to the
line of sight whereas Dondi & Ghisellini imply a broader distribution
of angles to the line of sight.
Such predictions may be tested by comparing the beaming characteristics of radio sources, such as superluminal motions and radio core brightness temperatures, over the full range of gamma-ray activity i.e. nondetections, weak detections, and strong detections, and examining any similarities or differences between these populations.
Thus, the three aims of the work described in this chapter are:
1] To increase the numbers of EGRET-identified radio sources well
studied with VLBI by targeting those in the Southern Hemisphere.
2] To begin to build a sample of strong, flat-spectrum radio sources which have not been identified by EGRET but are otherwise similar to the EGRET-identified sources, as a comparison sample.
3] To compare the relativistic beaming indicators derived for the sources presented here from 1] and 2], and also for observations from the literature.
To meet these aims, this chapter describes the first VLBI
observations of three EGRET-identified radio sources,
PKS 0208-512, 0521-365 and 0537-441, all high-confidence
identifications in the second EGRET catalog. Also described are the
first high-resolution VLBI observations of four radio sources
which have not been identified by EGRET, but show evidence for blazar
activity, PKS 0438-438, PKS 0637-752, PKS 1514-241, and PKS 1921-293. In 
4.2, a brief discussion of the beaming indicators which can be estimated from VLBI observations is given. In 4.3, the data reduction and analysis
methods specific to this work are outlined. The results for each individual source are
presented in 4.4, where estimates of the radio core brightness
temperatures and pc-scale to kpc-scale misalignment angles
are made. A discussion and comparison of the VLBI properties of the
sources considered in 4.4, as well as other radio sources from the literature, is given in 4.5.