I think this new paper <http://arxiv.org/abs/1212.5519> deserves some discussion here. It states that visual observations of eclipsing binary minima are biased and are therefore useless.
I am out of state right now, and just caught this on a friend's computer. Wow! I read the short abstract, but not the full length paper. I would say I have to disagree strongly with their deduction. Perhaps the particular people who observed this one star they cite were biased, who knows. But to say that all the visual observations are biased and no good is, in my opinion, not a strong conclusion to come to.
I hope others will chip in on this.
Chris Stephan SET
Robert Clyde Observatory
I agree with some points and not with others.
BS Vul was an EB-type star with a very short period (0.47 d.)
All the bad things that could affect visual estimates are strengthened when observing such a star.
Stars like this, with continuous variability are very difficult to measure visually because the eye won't detect the small variations happening all the time, specially around maxima times.
And the eclipses of an EB are wide, which makes it more difficult to determine.
It is different with EAs where the changes happen usually rapidly or in a steeper way and where you can see the star at a constant maximum most of the time.
I think that the most important point here is that visual observations of eclipsers are mostly useful when there are no other observations available. It makes no sense to compare them with photometric observations and the accuracy will always be lower.
And I wouldn't observe EBs but only EAs.
For the EAs I observe that have an eclipse caught in ~1990 by Hipparcos and maybe 1 or 2 from ASAS, a visual point in eclipse could be useful to determine a rather accurate period.
BUT I wouldn't try to study period changes based on those observations.
I think we need to focus on what is useful knowing what we can't do. Determining a period by catching a new eclipse doesn't even need a complete light curve, because our goal is different. We need to combine our data with other data to derive elements that in the future will be useful to make photometric observations to improve those parameters.
It makes no sense to fight the new technologies, we need to focus on what we can do.
I derived elements for a dozen of eclipsers, mostly long period ones, that had no period using visual observations and combining them with survey data (V353 Hya was an example where also Chris observed eclipses of that star visually). So it is not okay to make general statements about the usefulness or not of visual observations, observers just need to know what they can't do and what is still useful.
...merely those attempting to capture a particular time of minimum, which they say are biased in response to the predictions. They essentially echo what Sebastian says above:
It is evident from our findings that the heavily subjective character of visual
observations of short period EBs disqualifies them as a source of unbiased
information apt for fine EB period analysis. On the other hand we could
use 200 visual estimates of brightness made by S. Piotrowski in two time
intervals 1935–1939 (155 estimates) and 1945–1946 (45 estimates) (data in
Szafraniec, 1962), because these estimates have a different character than
later AAVSO and BBSAG visual observations. The distribution of these old
visual estimates shows that they were obtained in ‘monitoring mode’ without
the primary aim to obtain minima timings. Furthermore they passed
our careful analysis (p. 9).
Edit: Attached a copy of the paper for your convenience.
Visual observers can certainly be biased by the known predicted time of minimum (in rare instances, observers have "seen" an EB dim and brighten when it later turned out that the prediction of minimum was erroneous - clearly a case of anticipation bias (more commonly, a visual observer will "force" the beginning of a minimum when the star is running late; then, when the star refuses to continue dimming, the observer logs a flat interval in the light curve and picks up the real dimming to complete a record of the minimum).
That's why visual observations must be handled with caution. I would not rely on an isolated visual ToM to reach substantial conclusions about a period change. Visual data is always more useful when it is supported by additional visual data - three or more ToM that confirm the same thing.
But a mere glance at the many well-populated O-C diagrams in the AAVSO publications shows both the (sometimes) large scatter, isolated erroneous timings, and most important, the consistency of the vast majority of timings in tracing a star's varying period (and the often close agreement of visual and PEP times of minima). This data shows that visual EB timings are useful when used with due caution.
Slow stars with shallow minima are always the most difficult and the most prone to observational bias. An experienced analyst knows how to assess the data and how far to go in drawing conclusions from photographic, visual and PEP data.
The arXiv paper 1212.5519 looks like a shortened re-pub of a paper by the same authors (in different order) recently published in AJ. (Zhu et al, in AJ 144:37)
I've just spent the morning re-graphing, at enlarged scale, an O-C diagram of their data, that they listed in Table 4 of the AJ paper. I was hoping to see that the problem was the author's use of a parabolic fit, and that by forcing a parabola they pre-selected against honouring the visual data. However, looking at my graph, I am not willing to try to make that argument. The alternative that I was thinking of, a piecewise linear ephemeris that fits all the visual (through its larger scatter) and the CCD data, is just too much of a stretch. It requires a very abrupt period change just at the time that the observations switched from visual to CCD, and first a period decrease and then an increase to fit the unambigous trend of the CCD data. That's too much of a coincidence, even without delving in to the physics (which would be a bigger task).
There's an irony here: the authors were able to more fully examine the AAVSO data because as I understand it, Gerry Samolyk has in recent years been going back and putting the visual photometry in to the AID (i.e. the observations used to determine the ToMs that were published in the EB monographs). Also, when he switched to CCD observing, his data agrees perfectly with other CCD observers... but breaks from his own visual data.
All that being said, I think the authors are a bit extreme in how they word things.
I fully agree that the authors are too extreme in their wording.
Unfortunately I don't have access to the AJ paper, but I think they put too much weight on the correctness of their parabolic fit. Only rarely such a perfect fit exists.
Also I don't understand why they absolve BRNO observers from having any bias, simply because they used predictions rounded to the nearest half hour. Were the AAVSO and BBSAG predictions any different? Furthermore, the spread on the visual timings is about half an hour and deviates on average less than half an hour from the parabolic fit, how can that then make much of a difference? In addition the way these predictions were used by Gerry Samolyk and Hermann Peter is not clear to me: were they used at the telescope or just to plan the night's targets? In the former case they still require Gerry to have "followed" an almost perfect light curve every time.
It is clear that there is little value in visual timings when contemporaneous CCD observations are available, but dismissing visual data based on only one example is not a good practice.