Under the assumption of a void-filled Universe we investigate if the characteristic scale of voids can be determined from existing surveys. We use the Voronoi tessellation to create mock surveys and study the properties of the first zero-crossing of the two-point correlation function for various survey geometries. Our main conclusion is that the available data sets should be able to discriminate between 5000 \kms and 12000 \kms voids, if one of these scales actually characterizes the distribution of galaxies.

Using a long wavelength iteration scheme to solve Einstein's equations near the Big-Bang singularity of a universe driven by a massive scalar field, we find how big initial quasi-isotropic inhomogeneities can be before they can prevent inflation to set in.

We re-examine the systematic errors in the determination of the primordial helium abundance, $Y_{\rm P}$. We find that the systematics are significantly larger than the statistical errors. The uncertainty in (the determination of) $Y_{\rm P}$, is thus, larger than is currently claimed. Furthermore, most of the systematics lead to underestimate of $Y_{\rm P}$. The new upper bound allows cosmological models with no non-baryonic dark matter in which $\Omega_{baryons} = \Omega_{BBN} = \Omega_{dyn}$.

We study the propagator of a non-relativistic, non-interacting particle in any non-relativistic ``time-machine'' spacetime of the type shown in Fig.~1: an external, flat spacetime in which two spatial regions, $V_-$ at time $t_-$ and $V_+$ at time $t_+$, are connected by two temporal wormholes, one leading from the past side of $V_-$ to t the future side of $V_+$ and the other from the past side of $V_+$ to the future side of $V_-$. We express the propagator explicitly in terms of those for ordinary, flat spacetime and for the two wormholes; and from that expression we show that the propagator satisfies completeness and unitarity in the initial and final ``chronal regions'' (regions without closed timelike curves) and its propagation from the initial region to the final region is unitary. However, within the time machine it satisfies neither completeness nor unitarity. We also give an alternative proof of initial-region-to-final-region unitarity based on a conserved current and Gauss's theorem. This proof can be carried over without change to most any non-relativistic time-machine spacetime; it is the non-relativistic version of a theorem by Friedman, Papastamatiou and Simon, which says that for a free scalar field, quantum mechanical unitarity follows from the fact that the classical evolution preserves the Klein-Gordon inner product.

We suggest to use the common appearance of voids with a scale of $5000 \kms$ in the galaxy distribution to estimate the power spectrum on this scale. We use a simple model for a gravitational formation of voids and we compare the results with the matter fluctuations as constrained by the CMBR observations of COBE. We find that a power spectrum $P(k)\propto k^n$ with $n \approx 1.25$ is compatible both with COBE and with gravitational growth of large voids in an $\Omega=1$ universe. A Harrison-Zel'dovich spectrum, $n=1$, normalized to produce the observed CMBR fluctuations, does not have enough power for gravitational growth of voids with a diameter of $5000 \kms$ in an $\Omega=1$ Universe. Such a spectrum would be compatible if (i) void diameters are smaller ($3500 \kms$), the voids are shallower or if the voids are rare (we assume that the universe is void filled); (ii) $\Omega < 1$ i.e. the Universe is open; (iii) galaxies do not trace matter on very large scale or if (iv) the voids do not grow gravitationally.