Comet disintegration proceeds both through sublimation and discrete splitting events. The cross-sectional area of material ejected by a comet may, within days, become many times greater than that of the Earth, making encounters with such debris much more likely than collisions with the nucleus itself. The hierarchic fragmentation and sublimation of a large comet in a short period orbit may yield many hundreds of such short-lived clusters. We model this evolution with a view to assessing the probability of an encounter which might have significant terrestrial effects, through atmospheric dusting or multiple impacts. Such an encounter may have contributed to the large animal extinctions and sudden climatic cooling of 12,900 years ago, and the near-simultaneous collapse of civilisations around 2350 BC.

I find evidence for clustering in age of well-dated impact craters over the last 500 Myr. At least nine impact episodes are identified, with durations whose upper limits are set by the dating accuracy of the craters. Their amplitudes and frequency are inconsistent with an origin in asteroid breakups or Oort cloud disturbances, but are consistent with the arrival and disintegration in near-Earth orbits of rare, giant comets, mainly in transit from the Centaur population into the Jupiter family and Encke regions. About 1 in 10 Centaurs in Chiron-like orbits enter Earth-crossing epochs, usually repeatedly, each such epoch being generally of a few thousand years duration. On time-scales of geological interest, debris from their breakup may increase the mass of the near-Earth interplanetary environment by two or three orders of magnitude, yielding repeated episodes of bombardment and stratospheric dusting. I find a strong correlation between these bombardment episodes and major biostratigraphic and geological boundaries, and propose that episodes of extinction are most effectively driven by prolonged encounters with meteoroid streams during bombardment episodes. Possible mechanisms are discussed.

Intersection with the debris of a large (50-100 km) short-period comet during the Upper Palaeolithic provides a satisfactory explanation for the catastrophe of celestial origin which has been postulated to have occurred around 12900 BP, and which presaged a return to ice age conditions of duration ~1300 years. The Taurid Complex appears to be the debris of this erstwhile comet; it includes at least 19 of the brightest near-Earth objects. Sub-kilometre bodies in meteor streams may present the greatest regional impact hazard on timescales of human concern.

We calculate the expected flux profile of comets into the planetary system from the Oort cloud arising from Galactic tides and encounters with molecular clouds. We find that both periodic and sporadic bombardment episodes, with amplitudes an order of magnitude above background, occur on characteristic timescales ~25-35 Myr. Bombardment episodes occurring preferentially during spiral arm crossings may be responsible both for mass extinctions of life and the transfer of viable microorganisms from the bombarded Earth into the disturbing nebulae. Good agreement is found between the theoretical expectations and the age distribution of large, well-dated terrestrial impact craters of the past 250 million years. A weak periodicity of ~36 Myr in the cratering record is consistent with the Sun's recent passage through the Galactic plane, and implies a central plane density ~0.15 M_Sun pc^(-3). This leaves little room for a significant dark matter component in the disc.

It is shown that all of the 32 point X-ray sources which lie within about 10' of the centre of nearby galaxies, and which have so far been optically identified are high redshift objects - AGN or QSOs. Thus the surface density of these QSOs p similar or equal to 0.1 per square arc minute. Some of them were originally discovered as X-ray sources and classified as ultraluminous X-ray sources (ULXs), nearly all of which lie near the centers of active galaxies. We demonstrate that this concentration around galactic nuclei is of high statistical significance: the probabiliy that p that they are accidental lies in the range one in a thousand to one in ten thousand, and apparently this excess cannot be accounted for by microlensing.

It is shown that there are at least 21 QSOs within 1 degree of the nearby active spiral galaxy NGC3079. Many of them are bright (mag<18) so that the surface density of those closer than 15 arc minutes to the galaxy centre is close to 100 times the average in the field. The probability that this is an accidental configuration is shown to be less or equal to one in a million. Discovery selection effects and microlensing fail by a large factor to explain the phenomenon, suggesting that the QSOs may lie in the same physical space as NGC3079. However, two of them make up the apparently lensed pair 0957+561A, B whose lensing galaxy lies at z=0.355. This problem is discussed in the concluding section.

Two new samples of QSOs have been constructed from recent surveys to test the hypothesis that the redshift distribution of bright QSOs is periodic in $\log(1+z)$. The first of these comprises 57 different redshifts among all known close pairs or multiple QSOs, with image separations $\leq$ 10\arcsec, and the second consists of 39 QSOs selected through their X-ray emission and their proximity to bright comparatively nearby active galaxies. The redshift distributions of the samples are found to exhibit distinct peaks with a periodic separation of $\sim 0.089 $ in $\log(1+z)$ identical to that claimed in earlier samples but now extended out to higher redshift peaks $z = 2.63, 3.45$ and 4.47, predicted by the formula but never seen before. The periodicity is also seen in a third sample, the 78 QSOs of the 3C and 3CR catalogues. It is present in these three datasets at an overall significance level $10^{-5}$ - $10^{-6}$, and appears not to be explicable by spectroscopic or similar selection effects. Possible interpretations are briefly discussed.