John Carmichael wrote :
>>In proofreading the new fifth edition of my "Sundial Owner's Manual",
when
>>discussing sundials, I think that I mistakenly used the words,
"precise" and
>>"accurate", interchangeably, as if they meant the same thing.

This is a confusing area.  I am not a metrologist, though I am involved
in measurement.  I would suggest that accuracy is a more generally
useful term.

The "small divisions" concept is perhaps more often related to
resolution than accuracy.  A (clockwork) stopwatch has a typical
resolution of 1/5 or 1/10 second (some are faster) and can discriminate
between values on this basis.  However its accuracy - and the accuracy
of a measurement made with it - can be independent of its resolution as
it might gain say 1 minute per hour or 1.6%.  The two errors will
respectively dominate at short or long interval measurements.  But with
a sundial, which is analogue, the observer will interpolate between
values and the more divisions are present the less chance of introducing
errors by poor interpolation.

Another concept is reproducibility - which with a stopwatch might depend
on the person pressing the button as much as on the constancy of the
mechanism.  If you measure something twice (difficult with time!) do you
get the same value?  For a sundial, would several simultaneous observers
agree on its reading (perhaps this is a red herring)?

In my own instinctive view of these things:

A noon mark may be
  precise - giving the time of noon to a few seconds (e.g. St Sulpice) 
  accurate - agreeing with the general definition of local noon
  imprecise - fuzzy shadow or solar image
  inaccurate - wrongly aligned
However it is definitely no good at all for telling the time at 4 p.m.!

It is not necessary to have a lot of fine divisions to be precise -
dictionary definitions are not always helpful!  However the divisions
that do exist have to be good ones.  A length end standard (slip gauge)
is very precise indeed - and, one hopes, also accurate - but has no
divisions at all.  Even more so is the prototype kilogramme - a lump of
metal near Paris which defines the unit for the rest of the world and is
therefore accurate by agreement.  Similarly a capacity measure such as a
volumetric flask (or pint glass) can be precise as well as accurate but
is not subdivided.  If the length standard were significantly short of
its nominal value then using it would make for a precise but inaccurate
measurement.  

So a dial whose gnomon is well shaped and throws the best possible
shadow onto an hour line - agreeing with the line all the way along it -
may be precise, even if it has only hour lines.  If it is wrongly set up
it would however be inaccurate.  If it were badly delineated it would
certainly be inaccurate but might I think still be precise - this is the
difficult case to argue.

A heliochronometer that is well made should be precise.  However if
incorrectly set up it would be inaccurate; you would get an excellent
measure, to the minute, of the wrong thing!

Finally, you can make an accurate measurement using an inaccurate
instrument of known (and reproducible) error at the measurement point.
Combining the instrument and knowledge of its error produces a better,
imaginary, instrument.  In one sense, applying the equation of time to a
good sundial (which even if it is at least precise normally provides an
only inaccurate measure of mean time) is doing just this to get an
accurate measure of mean time.

Sorry this has got rather long.  I don't know whether this is helpful or
the reverse ... let alone accurate or precise ;-)

Regards
Andrew James

Reply via email to