The brightness and relative compactness of reflection nebulae have led to determinations of dust scattering properties with a finer wavelength sampling and smaller uncertainties than in either dark clouds or the DGL (see Fig. 1). Due to the usual presence of strong emissions (ERE and non-equilibrium emission), reflection nebulae have only provided scattering information in the ultraviolet and blue-optical. While dark clouds are faint, their relative compactness and usual lack of strong emissions has made them valuable as probes of dust scattering properties in the red-optical and near-infrared (Fig. 2). The one drawback to both reflection nebulae and dark cloud studies is that they probe higher density media than diffuse interstellar medium. This leads to questions to the general applicability of reflection nebulae and dark clouds results.
The faintness and large extent of the DGL makes it the most challenging object to both observe and model. The importance of measuring dust scattering properties in the diffuse interstellar medium and determinations of the extragalactic background has resulted in considerable effort being expended on the study of the DGL. Determinations of DGL dust scattering properties in the blue-optical and near-ultraviolet have been fairly uncontroversial, while such determinations in the far-ultraviolet have been quite controversial. This controversy is highlighted by the back-to-back reviews on the ultraviolet background in which it was argued that the majority of the ultraviolet background is from dust scattered light (Bowyer 1991) or extragalactic light (Henry 1991). The difficulties in both observing and modeling the DGL in the far-ultraviolet has led to studies in the literature with conflicting results, either low a and g values or high a and g values. This is the area which was most affected by the fourth criteria (not superseded by a more recent study). A number of initial analyses giving low a and g values were superseded by subsequent studies using more sophisticated models which found higher a and g values. As a result, the scatter in the plots of a and g for the DGL (Fig. 3) is now within the quoted uncertainties. The values of a and g derived from DGL measurements are similar to those derived from reflection nebulae and dark clouds.
By examining Figs. 1 - 3, it can be seen that the a and g values derived for each object type are internally consistent. For example, the measurements of a and g from different reflection nebulae all are consistent with a single wavelength dependence. When examing these figures, it is important to remember that they include results from broad-band filters along with those derived from spectroscopy. Comparing the results for all three object types in Fig. 4, it can be seen that the same may be true in general and not just internal to an object type. The only area where the data from different object types may disagree is in the far-ultraviolet region for a measurements only. The uncertainties are large enough to make a strong statement either way difficult. The high level of agreement between object types implies that their different strengths and weaknesses as well as model assumptions are not severely biasing our view of dust scattering properties. This result goes a long way in answering the concerns raised by Mathis, Whitney, & Wood (2002) about possible biases in reflection nebulae measurements of the dust albedo caused by ignoring the known clumpiness of dust.
In fact, it is not unreasonable to say that the current data (see Fig. 4) indicates the wavelength dependence of the albedo is ~ 0.6 in the near-infrared/optical with a dip to ~ 0.4 for 2175 Å bump, a rise to ~ 0.8 around 1500 Å, and a drop to ~ 0.3 by 1000 Å. The wavelength dependence of g is simpler with a monotonic rise from 0.6 to 0.8 from 10000 to 1000 Å. The uncertainties do allow for significant real variation in the a and g values around these qualitative averages of at least 0.1.
The study of dust scattering properties has shed light on the nature of the 2175 Å bump and the far-ultraviolet rise features of extinction curves. Early work on the 2175 Å bump indicated that it was likely an absorption feature with no scattered component (eg., Lillie & Witt 1976). Subsequently, evidence for a scattered component in the 2175 Å bump was found in two reflection nebulae, CED 201 and IC 435, by Witt, Bohlin, & Stecher (1986). This results was found to definitely be spurious for IC 435 when much more sensitive observations were taken and analyzed by Calzetti et al. (1995). By inference, the observations of CED 201 are also likely to be spurious. As a result, all evidence currently supports a 2175 Å extinction bump which is only due to absorption. Similarly, examining the results of reflection nebulae (see Fig. 1) gives good evidence that a significant portion of the far-ultraviolet rise in extinction curves (1700 Å and shorter wavelengths) is due to absorption.