Although the first near-infrared observations of Galactic Cepheids were made nearly 30 years ago (Wisnewski & Johnson 1968) their applicability to the distance scale was not appreciated until quite recently. In the first of a series of papers, McGonegal et al. (1982) unambiguously demonstrated that once periods were adopted from optical data, near-infrared observations of Cepheids provided a number of distinct advantages.
It was, of course, anticipated that by going to the infrared concerns about total and/or differential reddening would be significantly reduced. For instance, a blue extinction of AB = 0.32 mag (typical for Cepheids in the LMC, for instance) would translate to a total correction of AK = 0.05 mag at 2.2 µm; the full correction in the near infrared being comparable to the uncertainty alone associated with most optical extinctions!
But it was also immediately clear from the outset (see for example Figure 5) that the infrared had other advantages directly applicable to the establishment of the Cepheid distance scale. First of all, the decrease in the observed width of the PL relation was dramatic. Even single observations of the Magellanic Cloud Cepheids, (uncorrected to mean light) produced a PL relation with remarkably small scatter (± 0.2 mag). This narrowing of the width is due to two effects: a decreased sensitivity to differential reddening, but more significantly, due to a much decreased sensitivity of the infrared surface brightness to the temperature width of the instability strip. For exactly the same reason, the amplitudes of individual Cepheids (shown in Figure 1, as plotted earlier by Wisnewski & Johnson 1968) also decrease with the wavelength of the observations. Thus, Cepheids observed at long wavelengths and at random points in their cycle are (1) closer to their time-averaged mean magnitudes than the equivalent observation in the blue, and (2) the mean magnitudes themselves are in fact coming from a narrower projection of the instability strip into the infrared plane than the equivalent blue PL relation. From B to K, a typical Cepheid amplitude drops from 1.0 mag to 0.4 mag, while the width of the PL relation decreases from 1.2 mag to 0.5 mag. As a result, for distance determinations even single, random-phase observations of known Cepheids, when made in the near-infrared, are comparable in accuracy to complete time-averaged magnitudes (derived from a dozen or more observations) in the blue.
The periods for many extragalactic Cepheids had already been determined and published from photographic studies made at blue wavelengths in many long-term studies by Hubble, Baade, Sandage, Swope, Payne-Gaposchkin, Gaposchkin and others. As a result, in only a matter of nights, it was possible to reobserve the entire galaxy sample in the near infrared and thereby provide new accurate distances, almost unaffected by absorption effects.
However, until recently, with the advent of near infrared arrays, there have been limitations in the application of the IR to the Cepheid distance scale, even for nearby Local Group galaxies. Single-channel infrared detectors, available at the outset, were aperture devices which had to be ``chopped'', ``on'' and ``off'' the source in order to continuously monitor the intense and fluctuating terrestrial sky contribution to the signal. For Galactic Cepheids, and even for those in the Magellanic Clouds, a typical aperture of 5 arcsec could be placed over a star and chopped to a nearby reference region a few arcseconds away with relatively small uncertainty. However, for Cepheids at larger distances, the near-infrared observations of Cepheids in M31 (Welch et al. 1986) and M33 (Madore et al. 1985) show much more noise than could be attributed to photon statistics alone. The most likely source of error is crowding and confusion; that is, contamination in one of the two comparison fields where ``skies'' were being measured, or contamination in the object aperture itself. Although no systematic error is expected from this contamination, the random errors are appreciable. New observations are being obtained with near infrared InSb and HgCdTe arrays and two-dimensional image analysis is allowing more accurate infrared magnitudes to be measured in M31 and M33.