The high abundance of 4He allows accurate measurements in many locations. However, 4He is also produced by stars, and since such high accuracy is required, the primordial abundance is best measured in locations with the least amounts of stellar production. High accuracy is desired, since D/H predicts Yp to within 0.0014 (Yp / Yp = 0.006, 95% confidence), which is well beyond the typical accuracy of astronomical abundance determinations. In the local ISM, the amount of 4He from stars is about Y = 0.01 - 0.04; much less than Yp , but ten times the desired accuracy for Yp .
Helium has been seen in the intergalactic medium, where Carbon abundances are < 0.01 solar, and possibly zero in much of the volume. Strong absorption is seen from the He II Ly line at 304 Å in the redshifted spectra of quasars , however it is difficult to obtain an abundance from these measurements, because nearly all He is He III which is unobservable, and we do not know the ratio He II/He to within an order of magnitude. However, the strength of the He II absorption does mean that there is abundant He in the intergalactic gas , which has very low metal abundances, which is consistent with BBN, and probably not with a stellar origin for the 4He..
The best estimates of the primordial abundance of He are from ionized gas surrounding hot young stars (H II regions) in small galaxies. The two galaxies with the lowest abundances have 1/55 and 1/43 of the solar abundance. The 4He and H abundances come from the strengths of the emission lines which are excited by photons from near by hot stars.
Values for Yp from these extragalactic H II regions have been reported with small errors for more than 25 years, e.g.:
These values are lower than the value now predicted by low quasar D/H and they appear incompatible, because of the small errors. However Skillman et al.  argued that errors could be much larger than quoted, allowing Yp < 0.252, and Pagel  (and private communication 1994) agreed this was possible.
The measurement of Yp involves three steps. Emission line flux ratios must be measured to high accuracy, which requires good detector linearity and flux calibration, and corrections for reddening and stellar He I absorption. These fluxes must be converted to an abundance, which requires correction for collisional ionization and neutral He. Correction for unseen neutral He depends on the spectral senegy distribution adopted for the ionizing radiation and might change Yp by 1 - 2 percent. Then the primordial abundance must be deduced from the Y values in different galaxies.
Izotov, Thuan & Lipovetsky ,  have been pursuing a major observational program to improve the determination of Yp . They have found many more low metallicity galaxies and have been reporting consistently higher Yp values, most recently in their clear and persuasive paper :
The four main reasons why these values are higher are as follows, in order of importance , , , (Skillman, and Thuan personal communication 1998).
1. When stellar He I absorption lines underlying He emission lines are not recognized, the derived Yp is too low. This is a important for IZw18  which has the lowest metallicity and hence great weight in the derivation of Yp , and perhaps for many other galaxies.
2. The emission line fluxes must be corrected for collisional excitation from the metastable level. At low abundances, which correlate with high temperatures, these corrections can be several percent. The amount of correction depends on the density. There are no robust ways to measure these densities, and differing methods, used by different groups, give systematically different results. Izotov and Thuan  solve for the He II density, while Olive, Skillman and Steigman  use an electron density from the S II lines.
3. Izotov & Thuan  have spectra which show weaker lines, and they use the five brightest He lines, while Olive et al.  usually use only HeI 6678.
4. Izotov & Thuan ,  correct for fluorescent enhancement, which increases the Y values from for a few galaxies.
For these reasons Izotov & Thuan  obtain higher Y values for individual galaxies which have also been observed by Olive, Skillman & Steigman , and Izotov & Thuan  find a shallower slope for the regression to zero metal abundance (see  Fig 6). And most importantly, using higher quality Keck telescope spectra, they obtain high Yp = 0.2452 ± 0.0015 (random errors), from the two galaxies with the lowest metal abundances .
These measurement difficulties, combined with the recent improvements, lead most to conclude that the Yp is in accord with the SBBN. The Izotov & Thuan  values are very close to the low D/H predictions, while the lower Yp quoted by Olive , 0.238 ± 0.002 ± 0.005, is also consistent when the systematic error is used.
It is clear that the systematic errors associated with the Yp estimates have often been underestimated in the past, and we propose that this is still the case, since two methods of analyzing the same Helium line fluxes give results which differ by more than the quoted systematic errors. While the Izotov & Thuan  method has advantages, we do not know why the method used by Olive, Skillman & Steigman  should give incorrect answers. Hence the systematic error should be larger than the differences in the results: 0.007 using the most recent values, or 0.011 using earlier results.