Several speakers at this meeting presented their views on the important qualities of a good distance indicator. They noted the advantages of an indicator that is luminous, easily identified, and has easily measured properties (e.g. magnitudes). My list focuses more on the physical nature of the indicator rather than its technical properties. Thus, a good distance indicator:
This paper concentrates on the first of these. It is worth reviewing, though, how well the PNLF satisfies the remaining 5 criteria.
Metallicity effects were modeled by Dopita et al. (1992). Limited observational testing was performed by Ciardullo & Jacoby (1992). Most galaxies of interest have metallicities within a factor of 2 of solar abundances, and the predicted and observed errors in PNLF distances are smaller than 5% over this range.
The effects of population age on distance have been modeled by Dopita et al. (1992), Méndez et al. (1993), and Stanghellini (1995) and shown to be < 5% for galaxies having ages in the observed range between 3 and 11 Gyrs. Direct tests are complicated by our present inability to measure population ages accurately, but PNLF distances to young (LMC, SMC, M101) and old populations (M31's bulge, M81's bulge) where distances are known from Cepheids, fail to detect any age effect at all.
Internal extinction is less of a problem than intuitively suspected. In ellipticals, extinction is not an issue since the dust density is very low. In spirals, significant errors are expected if internal extinction in the galaxy is ignored. Observations, though, fail to reveal any measurable distance errors in the 7 late-type calibrators (Feldmeier, Ciardullo, & Jacoby 1997). Feldmeier, Ciardullo, & Jacoby (1997) modeled the effects of dust to understand this unexpected situation, and found that PN, which generally have scale heights well above the population I disk of a spiral, are either so extincted by heavy dust that they fall out of the PNLF sample, or they are so little affected that their magnitudes are not significantly diminished.
The excellent agreement between the PNLF and Cepheids (as well as SBF and with other methods to a lesser degree) demands that a physical basis for the PNLF must exist. Before all the comparisons were made, though, the theory had been described by Jacoby (1989), Dopita et al. (1992), and Méndez et al. (1993); recently, Stanghellini (1995) investigated the H PNLF. In short, it has proven easy to reproduce the constancy of the PNLF provided the population age is within the range of 3 to 11 Gyrs. If the progenitors are as young as 0.5 Gyrs, the PNLF brightens by ~ 0.3 mag. And, if all progenitors in a galaxy are much older than 11 Gyrs, they fail to produce observable PN at all.
The requirement that a distance indicator be testable against another method is fundamental to the concept of the scientific method. Because we never know the "right answer" in the distance scale business, we turn to intercomparisons between different methods assuming that 2 independent methods are very unlikely to repeatedly yield the same wrong answer. If a method is not testable, it relies solely on the validity of a model and scientists generally agree that models must be tested. By inference, an untestable indicator is equivalent to an untestable model. Fortunately, the PNLF can be tested against numerous methods (see Ciardullo, Jacoby, & Tonry (1993), Jacoby (1995), Feldmeier, Ciardullo, & Jacoby (1997)).
The second component of the last requirement is that multiple investigators must be able to derive the same answer using the same technique. This sounds simple enough, and again, is fundamental to the scientific method. Results from some methods, however, cannot be reproduced at a later time should a question arise about their validity. The most obvious of these methods is supernovae. A second observer cannot go back in time to observe a supernova in order to check on the observational accuracy of a prior observer's measurements, or to utilize a superior instrument. Thus, SN Ia fail to satisfy this requirement.
The PNLF technique satisfies the prescription for a good distance indicator on each count. Cross-testing with other methods (the last and most important criterion) shows that disagreements between the PNLF and other reliable methods (e.g. Cepheids) are smaller than 8%. Thus, systematic errors due to extinction, age, metallicity, or application of the method are not accumulating beyond the 8% level. In fact, when consideration is made for the error contribution from the Cepheid distances, the PNLF errors must be smaller than ~ 5%.