An observational study of the clustering of faint galaxies in the magnitude range 19.5< mj <22.0 is undertaken. High quality plate material from the United Kingdom Schmidt Telescope, and automated machine measures from the COSMOS facility at the Royal Observatory Edinburgh constitute the primary data. Particular attention is paid to quantifying the amplitude and nature of random, and systematic errors associated with the plate material and the measuring process. The details of calibration procedures and properties of the COSMOS scanning spot are found to be particularly important for a full understanding of the measurement process. New information concerning the success of automated image classification techniques is obtained, through the use of very deep plate material from the Anglo- Australian Telescope to "calibrate" the wide field plate material used in the investigation. Automated image classification using COSMOS parameters is found to be considerably less successful than previously claimed. Consideration of the methods used to estimate the covariance function from galaxy samples shows that some discrepancies between earlier work on the Zwicky and Lick catalogues, and more recent, deep machine measured samples are due to different estimation procedures, and not the data themselves. Data from five high galactic latitude fields in the south, subtending a total area of more than 100 square degrees, allows precise limits to be placed on the amplitude of spurious galaxy clustering due to small scale extinction fluctuations within the galaxy. Upper limits to the amplitude of extinction fluctuations are derived from the form of the stellar covariance function and the cross -correlation between the stellar, and deep galaxy samples. The observations are compared to a number of extinction models, and it is concluded that mean values of absorbtion at 2 the poles of A <0.1 magnitudes are favoured.