Our aim in this thesis is to use the clustering of QSOs to investigate large-scale structure and cosmology. We are particularly concerned with estimating the cosmological parameters which govern the evolution of structure in the Universe.
We first investigate how QSOs trace the distribution of `normal'
galaxies by measuring the correlation between
a sample of 
QSOs and faint, 
galaxies. At 
we find that the cross-correlation amplitude is marginally negative.
This low signal clearly rules out models in which QSOs inhabit rich
environments. The environments of QSOs are more similar to those of
average galaxies. The slight negative correlation can be
explained by gravitational lensing, but this has no effect on our
conclusions concerning QSO environments.
We determine the clustering properties of a combined sample of >1500
QSOs including the LBQS and Durham/AAT QSO surveys. This data set has
a clustering amplitude 
for
at
. On 
scales the limit
on detected signals in 
is 
. A model of clustering evolution
which includes the effect of bias was used to compare QSO clustering
to the clustering of low redshift galaxies and Seyfert galaxies. If
Seyferts and QSOs are similarly clustered, then the data prefer a low
or high bias for QSOs and galaxies.
In contrast, comparisons to the CMB measurements of COBE assuming a CDM-type
power spectrum suggest low bias.
This might be taken as evidence for low 
, but the data is still
consistent with 
and 
.
We consider the possibility that nearby galaxy clusters can
gravitationally lense background QSOs. We apply the lensing hypothesis
to the result of Boyle et al., (1988) and find that cluster masses
required are too large. A small
dust component could retrieve the lensing model and allow more
reasonable mass estimates for clusters from this method.
The requirement for a new, deep, wide-field, QSO survey is clear. We
discuss the construction of the candidate catalogue for the 2dF QSO
Redshift Survey, which will contain 
QSOs. We calibrate the photographic plates used for the candidate
catalogue and assess the sources of errors and incompleteness. From
preliminary spectroscopic observations we conclude
that the completeness of the 2dF catalogue is 
,
compared with an estimated completeness of 
. We propose to
substantially increase the catalogue completeness (to 
), by
the introduction of UKST r plates into our candidate catalogue.