Äîêóìåíò âçÿò èç êýøà ïîèñêîâîé ìàøèíû. Àäðåñ îðèãèíàëüíîãî äîêóìåíòà : http://www.mso.anu.edu.au/~bessell/Beauty.ps Äàòà èçìåíåíèÿ: Wed Jan 28 05:20:25 2004 Äàòà èíäåêñèðîâàíèÿ: Mon Oct 1 21:29:13 2012 Êîäèðîâêà: Ïîèñêîâûå ñëîâà: reflection nebula

Beauty and Astrophysics
Michael S. Bessell
Research School of Astronomy & Astrophysics, Institute of Advanced Studies, Australian
National University, Private Bag, Weston Creek PO, ACT 2611, Australia
bessell@mso.anu.edu.au, http://www.mso.anu.edu.au/~bessell
Abstract
Spectacular colour images have been made by combining CCD images in 3 different pass­
bands using Adobe Photoshop. These beautiful images highlight a variety of astrophysical
phenomena and should be a valuable resource for science education and public awareness of
science. The wide field images were obtained at SSO by mounting a Hasselblad or Nikkor
telephoto lens in front of a 2Kx2K CCD. Options of more than 30 degrees or 6 degrees square
coverage are produced in a single exposure in this way. Narrow band or broad band filters
were placed between lens and CCD enabling deep, linear images in a variety of passbands to
be obtained. We have mapped the LMC and SMC and are mapping the Galactic Plane for
comparison with the Molonglo Radio Survey. Higher resolution images have also been made
with the 40 inch telescope of galaxies and star forming regions in the Milky Way.
Keywords: techniques: image processing --- surveys ---miscellaneous
1 Introduction
1.1 Foreword
I chose this title to remind the jaded astronomers amongst us of the major reason why we chose
astronomy as a career. It also explains why there are so many dedicated amateur astronomers and
why the general public readily supports astronomy. Astronomy, combining beauty with mystery
and wonder, remains a formidable combination of attractions for the minds of students and others
with a restless intellect.
To a few theoreticians, E=mc2 represents the highest form of beauty, but most of us are visual
people stimulated more by pictures than equations. I well remember as a student attending Bart
Bok's public lectures in Hobart and participating in his evocation of the wonders of Astronomy
through his black and white slides of the star­forming regions in Norma, Carina and the Magellanic
Clouds. This was for me, I thought.
In the current world many of the best students are turning to economics and law and many
students are more attracted to astrology and creationism than to mainstream science. In addition
within some of the liberal arts disciplines there has been an increasing attack on the scientific
method and the absolutism of science. It is therefore essential for the 'hard scientists', especially
the physical scientists, to speak out in support of the scientific method of determining the truth
in the physical world and to challenge the pessimistic outlook of the Jeremiahs in the community
with the optimism from scientific discoveries both on Earth and elsewhere.
1

Part of this fight­back is not only to attract some of the best students back into mainstream
science but to educate all students and the general public about the importance of the scientific
method and the importance of scientific research. A public better educated on these matters
is more likely to result in governments that make well informed decisions and which will fund
research and education adequately for the betterment of future generations.
School teachers, university lecturers and the media all have an important part to play in this
and to assist them it is essential to provide a wide range of visual stimuli that can attract attention,
maintain interest and affect the emotions. The coloured images from Mt Wilson, NASA, the AAO
and more recently the HST are excellent resources but in this day of visual overload it is important
to provide many new images of high impact that can be used for many different purposes and to
provide these on a continuing basis.
1.2 Stimulus and Opportunities
In this presentation I would like to show some of the results from imaging projects that I have
been involved with at SSO using modest optics but with a large science grade CCD. I hope that
this will encourage some of you to experiment with your own images and to utilize all these images
that have been produced in your lectures and presentations.
The processing into colour images is the most time consuming and in many ways the most
difficult step, albeit the most rewarding. The most exciting moment is the instant three different
black and white images combine to produce colour. There are invariably surprises, as, unlike
earthly scenes of mountains, streams, skies and people that we see every day, we often have no
idea of what colours or distribution of colours will appear in an astronomical image. I believe that
maximum impact of an image requires the skills of a graphic designer/artist and an astronomer.
There are few people born with these skills but it raises the possibility of collaborations. I can see
many possible projects involving astronomers and creative artists and more specifically graphic
design students and even a major project for producing an interactive CDROM of the survey
images.
Finally I hope that it will encourage amateur astronomers and teachers from schools with access
to the new low and medium cost CCD systems to take on the challenging but very rewarding and
creative astronomical colour imaging projects.
1.3 The Many Uses of Colour
There are many ways in which the use of colour enhances the information in an astronomical image.
Although normal broad­band blue­green­red images approximating the eye's response can be used
to illustrate many topics in astrophysics, it is the use of narrow­band images that provide, more
specific astrophysical information, particular that involving the gas and dust in the interstellar
medium. It is also the transformation of images made in passbands in the UV, IR or radio into
optical colours that enhances our insights into astrophysical properties or highlights the unusual
object. Such a technique has wide spread applications one of which is to illustrate the photometric
redshifts of distant galaxies. We also nowadays often make use of false­colour images to record
intensity differences in an image made with a single passband. Colour is therefore the best way to
present most data but its use has been inhibited in the past by the expense of colour reproduction
in the main science journals. However, with more and more electronic journals being produced,
the provision of coloured JPG files should not be an added expense. David Malin (eg. Malin 1992)
has pioneered the use of colour in astronomical imaging and you are refered to the AAO images
page http://www.aao.gov.au/images.html for his suite of images derived from photographic plates
2

Table 1. Camera Focal Lengths and Fields
Name Focal length F no. Aperture Scale Field
Hasselblad 80mm 2.8 60mm 60 30
Nikon 400 4.5 90 12 6.8
40 inch 8128 8 1016 0.6 0.3
and film, images that inspired the CCD work reported here.
1.4 The Wide­Field Imaging Project
I had been experimenting at the 16 inch telescope with the 2.3m quartz optics reimaging camera
and a 1K CCD to provide UBVI and Hff images of the Magellanic Clouds. But because the
scale was a little low and access to the imager camera was uncertain, John Hart suggested I try
out a commercial Nikon telephoto lens bought many years earlier for an unsuccessful image tube
project. The results were excellent. So, when Anne Green and Lawrence Cram made a request for
Hff images for comparison with their Molonglo Radio Surveys of the galactic plane (Green et al.
1999; http://www.astrop.physics.usyd.edu.au/MGPS/) we decided that this was the appropriate
setup. Funds were provided from Sydney University and the IAS for a collaborative WFI project
and a computer and data drive were purchased. A mechanical lens mount, filter holders and CCD
mounting brackets were made and the Nikon imager providing a 6.8 degree field on a 2K CCD
was installed on the 16 inch telescope mount in place of the existing telescope tube. Dome flat
field screens and lamps were also provided. For the first year we used a CCD controller, computer
and 2K CCD borrowed from the other SSO and MSO telescopes when they were unscheduled but
recently have obtained a dedicated 2K CCD and controller for the project. More details of the
Hff survey can be found on http://www.mso.anu.edu.au/~buxton/halpha.html
We have also borrowed a Hasselblad lens from another ANU Department and made a new
mounting so that we can image over 30 degrees of sky with a single exposure. We have also used
the 40 inch telescope to obtain images at higher resolution of smaller objects, especially galaxies
outside the local group.
In Table 1 are given the mountings, cameras, image scales and field sizes.
1.5 Acknowledgements
Major contributors to the project have been Michelle Buxton (MSSSO) and Bob Watson (Uni­
versity of Tasmania). They did many of the initial observations and Bob developed all the IRAF
scripts for the data processing. Other students from the Universities of Sydney, NSW and Wol­
longong have contributed to the observing together with a team of amateurs from Coonabarabran
organised by John Shobbrook who also wrote the observing manual. In 1999 Paul Price took
over the Hff survey as an honours project and Newcastle amateur Ken Hargreaves has done much
observing. Ralph Sutherland (MSSSO) has been indispensable for his instructions on the use of
Adobe Photoshop and his inspired image processing. The workshop staffs at SSO and MSO have
been very supportive of the project both in its construction and operation. Hwankyung Sung
provided many of the 40 inch images from his work on young star forming regions.
3

2 Atlas of Images
2.1 Preamble
Some examples of images from all these instruments will be shown. The images do not represent
limiting observations but rather are illustrative of the information content and the astrophysical
problems that they can address. The images are presented in a somewhat arbitrary progression
from external galaxies to our galaxy, the Milky Way, then move along the galactic plane from
Centaurus to Monocerotis commencing with the widest field views to the narrow views. The aim
is to integrate what we see in our galaxy and the Magellanic Clouds with what we see in more
distant galaxies.
The images of the external spiral galaxies show clearly the spatial distribution of the star
forming regions, the giant HII regions (like red beads in a necklace) and the dust lanes lying along
the arms and reaching right into the centre of the galaxies. The central bulges made up of older
stars are also very obvious. When we look at the wide field images of the Milky Way we can
recognise similar HII regions and dust lanes and we can then zoom in on these to reveal bright
blue stars and shimmering curtains of ionised hydrogen. We can see the results of extensive stellar
winds and the remnants of exploding stars with their delicate twisted shock fronts. By providing
deep images of the interstellar region we can see the complicated interaction between hot stars
and gas and marvel at the beauty of these interactions.
The image scale (12 arcsec/pixel) with the Nikon lens is good for most large scale structure but
is unable to resolve the fine collimation of the Herbig­Haro jets associated with pre­main sequence
stars or the disks of distant planetary nebulae, both of which are a couple of arcsec across. These
are seen extremely well on the films of the new AAO Schmidt Hff survey.
2.2 Filters Used
Many of the images combine two narrow band filters, Hff (red) and [OIII] (green) with a short
exposure B (blue). These bands are actually red (6563 š A), green (5007 š A) and blue (ú4300 š A).
The combination of HII (13.6 eV) and [OIII] (35.1 eV) with their significantly different ionization
potentials traces the ionization gradients. It highlights the hottest stars and the most energetic
shocks. A few images will be shown that feature more subtle ionization differences using [SII],
Hff, and [NII], all 'red' lines but shown as RGB in order of wavelength. Two images will be also
be shown with out­of­order colours that result in a more striking images.
2.3 List of Images
The images are available for viewing from http://www.mso.anu.edu.au/ bessell/images/. Thumb­
nails of the images are displayed three per line for 21 lines. Table 2 lists the images and identifies
them by line and position. It also gives the filters comprising the BGR of the colour image, the
camera used (see Table 1), the observers, and a brief comment about the image or series of im­
ages. Detailed descriptions of many of the astronomical objects can be found with David Malin's
images at http://www.aao.gov.au/images.html. The shorthand O3, S2, N2 are given for [OIII],
[NII] and [SII] respectively. The code for the observers is msb: Bessell; b.w: Buxton & Watson;
sung: Hwankyung Sung; murph: Murphy; b.dc.k: Bessell, Da Costa & Keller; suth: Sutherland;
s.b: Sutherland & Bally.
4

Table 2. List of Images
No. Name Filters Camera Obs. Notes
1a Antenna galaxies BVHff 40 inch msb Pair of colliding galaxies. Note knots of star forming regions
1b Centaurus A BO3Hff 40 inch b.w Radio galaxy. Note green [OIII] jet top left
1c M 95 BVHff 40 inch msb Theta galaxy. Note the low surface brightness outer arms
2a NGC 1365 BVHff 40 inch msb Famous barred spiral. Note the dust lane leading into the core
2b NGC 2997 BVHff 40 inch msb Grand Spiral. Note the HII regions in the spiral arms
2c NGC 6744 BVHff 40 inch msb A different type of barred spiral
3a M 31 BVHff Nikon msb Andromeda galaxy. Note the dust lanes and HII regions
3b Magellanic Clouds BVHff Hasselb msb The closest galaxies to the Milky Way
3c SMC BVHff Nikon msb The SMC and wing + globular clusters 47 Tuc and NGC 362
4a SMC zoom BO3Hff Nikon msb The gas­rich SMC. The HII regions dominate in this image
4b LMC BO3Hff Nikon msb HII regions and the old stellar population in the bar
4c LMC zoom BO3Hff Nikon msb 30 Doradus and shells of HII regions dominate this image.
5a 30 Doradus BVR 40 inch sung Three beautiful views of 30 Doradus. Stars dominate here
5b 30 Doradus BVHff 40 inch sung In this image stars and the gas share the stage
5c 30 Doradus BO3Hff 40 inch sung Here the gas dominates. It appears 3 dimensional
6a Crux and pointers BVHff Hasselb msb The archetypal view of the southern skies
6b Scorpio BVHff Hasselb msb The huge constellation of Scorpio extends from the plane
6c Sag­Oph BVHff Hasselb msb View looking toward the centre of the Milky Way
7a Jewel Box BVHff 40 inch b.w A young cluster of massive stars near the Southern Cross
7b NGC 2004 164VHff HST b.dc.k A similar cluster in the LMC with many times more stars
7c NGC 2004 zoom 164VHff HST b.dc.k These are HST images. Note the blue main sequence stars
8a NGC 2100 zoom 164VHff HST b.dc.k Be stars (pink), red supergiants (orange);
8b NGC 330 zoom 164VHff HST b.dc.k A supergiants (white)
8c NGC 346 BO3Hff 40 inch msb The largest cluster of young massive stars in the SMC
9a Eta Carina BO3Hff Nikon msb The beautiful j Car nebula. Note the lace work
9b Eta Carina BO3Hff 40 inch msb A closer view of the nebula and young clusters
9c Eta Carina BVHff 40 inch sung Note the strange variable j Car
10a Trumpler 24 BVHff 40 inch sung A young cluster and HII region
10b Bochum 14 BVHff 40 inch sung An HII region surrounding a Wolf Rayet star
10c Bochum 10 BVHff 40 inch sung Another young cluster and HII region
11a Lagoon & Trifid BO3Hff Nikon msb The well known Lagoon and Trifid nebula
11b Trifid BO3Hff 40 inch msb A blue reflection nebula, HII region and dust lanes
11c Lagoon zoom BO3Hff Nikon msb Color change shows the ionization gradient.
12a Lagoon inner BO3Hff 40 inch msb Yellow near the hottest stars. Red shows lower ionization
12b Lagoon inner BVHff 40 inch sung Beautiful dust lanes and starkly etched globules
12a Lagoon inner zoom BO3Hff 40 inch msb Colours rotated to produce moody image `a Turner'
13a Eagle Nebula BO3Hff 40 inch b.w Star forming region. HST image available.
13b Eagle Nebula zoom BO3Hff 40 inch b.w Rotated field
13c CG4 Nebula BVR 40 inch msb Interesting reflection nebula with dust clouds
14a NGC 3603 BVHff 40 inch sung Young cluster
14b NGC 6188 BO3Hff Nikon b.w Another interesting obscured nebulosity
14c Pismis 24 BVHff 40 inch sung Young cluster
15a IRAS 10.60.100¯ IRAS IRAS From Sutherland & Bally
15b Orion­Eridanus BO3Hff Hasselb suth Wide field view mosaic
15c Orion BO3Hff Hasselb msb Showing Barnard's Loop around the Orion constellation
16a – Orionis BO3Hff Nikon msb The 'bubble' surrounding the O star – Orionis
16b Orion BO3Hff Nikon msb The region between the Belt and Sword of Orion.
16c Horsehead Nebula BO3Hff 40 inch sung The famous dark cloud in Orion
17a The Orion Nebula BO3Hff 40 inch msb Different views of the great nebula of Orion
17b The Orion Nebula BVHff 40 inch msb Different filters highlight small ionization differences
18a The Orion Nebula S2HffN2 KPNO 1m b.s The seeing is much better in these next images
18b The Orion Nebula S2HffN2 NTT 3.5m b.s Note the detail in this excellent ground based image
18c The Orion Nebula S2HffN2 HST b.s Note the stellar tails all pointing away from the trapezium
19a NGC 2264 N BVHff 40 inch sung Part of young cluster in Monocerotis
19b NGC 2264 W BVHff 40 inch sung Note the beautiful reflection nebula and dust clouds
19c Rosette Nebula BO3Hff Nikon murph Strong ionization produces these unusual colouring
20a Vela­Puppis BO3Hff Hasselb msb This wide field view shows the comet Hale­Bopp lower right
20b Vela BO3Hff Nikon msb Note the [OIII] and Hff filaments of the SN remnant
20c Vela zoom BO3Hff Nikon msb Note the details of the twisted [OIII] filaments
21a Crab Nebula BO3Hff 40 inch b.w Note the green [OIII], red Hff and blue synchrotron light
21b Helix Nebula BO3Hff 40 inch b.w Planetary nebula. The blue and green colours have been reversed
21c Comet Hale­Bopp V Nikon msb Note the two globular clusters. The colors represent intensities
5

3 Discussion
3.1 Are the Colours Real?
The colours that are seen, are 'real' colours, in distinction to the 'false' colour images that one
often sees these days. The blue, green and red colours are assigned to those actual observed
colours in most cases, in the few cases where they are not they are at least assigned in the order
of increasing wavelength. However, because the images utilize at least one narrow band filter
centered on an emission line from ionized hydrogen, we 'see' the gas brighter than the star light
by a factor that is the ratio of the width of the narrow band filter to the width of the broad
band filter that represents the eye's colour band. That is, we increase the brightness of the hot
interstellar gas in comparison to the brightness of the stars. We draw attention to the star forming
regions by the effect that the hot stars have on the gas. We do not alter the relative colours of
stars but we make them fainter in comparison to the brightness of the gas. But we do alter the
colour of the gas depending on which emission line we choose to highlight.
For example, in the three images of 30 Doradus (line 5) we see very graphically the effect of
changing the combination of filters used. Replacing the narrow green filter with a broad green
filter weakens the green light from the gas compared to the star light. Finally replacing the narrow
red filter with a broad red filter weakens the red light from the gas. The images shift from being
star dominated to gas dominated and the colour of the gas changes while the colour of the stars
stays the same.
In most of the images that are shown, by combining narrow band green ([OIII]) and narrow
band red (Hff) light, we produce a yellow colour when the [OIII] line is similar in strength to the
Hff line; we see this for example in the Lagoon and Trifid Nebulae (line 11) where the ionization
is produced by radiation from imbedded hot stars. Colour gradients between red and yellow show
the varying strength of these lines as the radiation temperature of the gas changes.
However, in supernova remnants, like the Crab (line 21) and Vela (line 20) some of the ioniza­
tion results from high energy shocks and in these cases the [OIII] lines can be strong while Hff
is very weak. In that case the gas will appear green. We also see regions in SN remnants where
there are a series of ionization fronts, some green while others are yellow and red illustrating the
complex ionization structure and interaction between collisions and radiation. The use of [SII],
Hff and [NII] filters as seen in the series of Orion images from KPNO, NTT and HST (line 18)
shows more subtle ionization differences.
3.2 What is reality in an image?
We distort 'reality' in many of our images in several ways. CCDs are linear detectors that faithfully
record the relative brightnesses. However, our eyes are logarithmic detectors so is it more 'real' to
process the CCD data to be logarithmic? We do this anyway in our data processing as explained
below. But we need to do other distortions as well. The range of intensities that the eye can
see, or the monitor or hard copy can provide, is limited. There is a great advantage in further
compressing the intensity scale so as to keep detail in the bright areas of an image as well as in the
fainter part. Photographic plates and prints did this a little by their non­linearity at the bright
end but we can achieve it much better digitally. For example, in the images of galaxies (lines 1,
2 and 3) the nuclear regions and bulges of these galaxies tend to be much, much brighter than
the outskirts. We have tried as hard as possible to retain the information in those bright regions
whilst enhancing the brightness and contrast in the outer spiral arm regions. The nuclear ring in
NGC 2997 is just visible and the dust lanes with their imbedded star forming regions are seen in
6

these barred spirals to reach into the central regions. Such details are not obvious in previously
published images showing the outer regions.
Another way we distort reality or rather we enhance reality is by emphasizing the hot interstel­
lar regions. Normal broad­band optical images of galaxies and galactic fields mainly show stars.
Within a narrow dynamic range these stars are coloured although the brighter ones are invariably
white because the images are saturated. However, we know that in most galaxies, in particular in
spiral galaxies and disk galaxies like the Milky Way and the Magellanic Clouds the gas content
is very high and in fact often dominates the mass distribution in the disk. Is it 'real' then to
show a picture that mostly ignores a major physical component? By enhancing the faint light of
ionized hydrogen as we have done in our images we are bringing to attention an important mass
component. For example, in the wide field view of the Milky Way (line 6) we begin to see that
the galactic plane is bathed in a sea of ionized gas and all the bright HII regions are probably
connected. In the SMC and LMC (line 3,4) we see galaxies that are dominated by gas, which
is what they are. So these images bring a new perspective and in many ways are more 'real'
than previous 'reality'. We still cannot see the neutral hydrogen but by combining the Parkes
multibeam survey http://www.atnf.csiro.au/research/multibeam/multibeam.html with these im­
ages, using another colour, we should be able to encompass that as well. So the difficulty and
challenge is to present a beautiful image whilst retaining and emphasizing the important physical
information, the integrity, of the object.
3.3 Image Processing
Finally, some specifics about the observing procedure and image processing. We aim to take three
images through each filter with the telescope slightly offset between each exposure in the set of
three. The three images through each filter are bias subtracted and flat fielded then superimposed
and combined by median filtering to produce a single grey image without cosmic rays and bad
columns. The combined images through different filters are registered onto a reference image
using the programs GEOMAP and GEOTRAN within IRAF. The 16 bit FITS files are then
imported into PHOTOSHOP using a program written by Ralph Sutherland. This program, which
generates 8 bit PHOTOSHOP data, samples the image and suggests upper and lower levels for the
conversion. It also offers linear or logarithmic scaling. Most of the images shown here have had
logarithmic scaling for all three colours (channels). (Sometimes special effects can be produced
by combining logarithmic scaling in one or more colours with linear scaling in another.) We often
need to reimport images with different levels chosen, especially near the sky to ensure that faint
detail is not lost and to maximise the 256 levels in PHOTOSHOP by not wasting them.
In PHOTOSHOP the images are copied into the R, G or B channel of a new image and then
manipulated by adjusting levels, linearity and contrast. The manipulation of the levels and curves
are where the artistic endeavour is required; however, one also needs to address what story is
being told in order to determine how best to present the data. This is where the astronomy is
needed together with the preconceived ideas of what a coloured astronomical picture should look
like. Generally, we feel that the background sky should be dark with no dominant colour; that
the bulk of the stars should be white or their colours soft shades of blue­white, white, yellow or
orange. Normally, blue stars come out blue­white, most stars white and red giants are generally
deep yellow. The glowing interstellar region then produces its own unexpected colours. An image
can of course be produced without any a priori astronomical knowledge and this would be an
interesting exercise for an artist, but with one exception I will not present any such images here.
The HST images of the Magellanic cloud clusters (line 7,8) show more subtle colours. Because
the blue colour is the PHOTOSHOP image represents the far ultraviolet colour of the star. The
7

brightness of the blue image of the hot stars is much brighter than normal optical images. As a
consequence, these stars come out very blue. But if those hot stars have strong Hff emission (the
Be stars), they appear pink in the colour image; you can see the high proportion of Be stars in
these Cloud clusters. In addition, the great drop in UV light (compared to blue light) between an
A­F star and a KM star makes the AF stars white in the colour images and the KM stars orange.
We have made no attempt to produce real intensities in our manipulations. The intensity
mappings are however, monotonic although not linear. The manipulations are done for their
impact without undermining the astrophysics. Where exact quantitative data is required this is
best done with the individual linear monochromatic images in IRAF. We have generated good
magnitudes or good stellar free Hff images by subtracting or dividing the individual images in
IRAF. Michael Murphy (UNSW) has also identified many probable planetary nebulae and Be
stars in the Magellanic Clouds by doing photometry on Hff­[OIII] and [OIII]­continuum images
(Murphy & Bessell 1999).
3.4 Future Work
Paul Price (Price 1999) is doing an honours project completing the Hff survey and comparing
it with specific portions of the the Molonglo Radio Survey (Green et al. 1999) and the MSX
IR Galactic Plane Survey (Cohen 1999) http://gibbs1.plh.af.mil/. We are also completing the
wide­field Hasselblad survey of the whole galactic plane visible from SSO and exploring with
a commercial organisation the possibility of making an interactive CDROM using all the CCD
images we have processed. However, generating colour images is very time consuming so it is
unlikely that much more can be done without additional resources.
Acknowledgements
I would like to thank David Malin for suggestions that led to considerable improvements in the
paper and for inspiring the pursuit of beauty in astronomical colour imaging.
References
Cohen, M. 1999, private communication
Green, A.J., Cram, L.E., Large, M.I. & Ye, T. 1999, ApJS, in press
Malin, D.F. 1992, QJRAS, 33, 321
Murphy, M.E. & Bessell, M.S. 1999, MNRAS, in press
Price, P.A. 1999, private communication
8