Sorry, QO = Quantum Optics.

Click on the "View as Tree" options and the Google Groups will show you
the parent-child relations between posts in the tree form, with new
messages shown in bold typeface. That reply was in the middle of a back
& forth sequence between E. Stefanovich and myself which the Tree View
will show you.

The sub-poissonian photo counts refer to (alleged) additional
exclusivity (anticorrelation) one is supposed to observe when detecting
a "single photon"  via multiple detectors e.g. after the "single
photon" goes through a 50:50 beam splitter. The classical model for the
"split photon" detection by two detectors in a beam splitter experiment
is a pair of Poissonian processes, one for each detector i.e. the
number of photo-electrons k ejected on each detector is given via
Poissonian distribution:

  p(k,n) = (n^k / k!) exp(-n)    ....  (1)

where n = n(t) = <k> is the average k, which is proportional to
incident light intensity. The two counts k measured (as amplified
photocurrents) by two detectors in each try (or sampling window) are
independent of each other. They depend only on n(t).

There is a folkore in pedagogical and popular QM & QO literature which
asserts that the probability p(k,n) on one detector will drop below the
Poissonian value, the RHS in eq. (1), as soon as the other detector
"triggers" (the "trigger" is defined by the 'sensitivity settings' of
the detector's electronics, as a photocurrent above some arbitrarily
selected threshold  i.e. it is photo-curent value coarsened down to 1
bit precision). It is claimed by the proponents of this "collapse"
folklore that this drop in the p(k,n) is due to the the state collapse
-- allegedly, as soon as the detector 1 has "measured" (an ill-defined
term) the photon, the composite (EM field) state collapses and the EM
field state at the detector 2 is supposed to vanish instantaneously and
across any distance.

This "collapse" folklore (which is based on an operational
misinterpretation of the von Neumann's QM state collapse postulate for
composite systems), has failed to show up in the experiments, despite
numerous attempts to demonstrate it since 1950s. At that time (1956)
Hanburry Brown and Twiss demonstrated that the two detections (e.g.
after the beam splitter) behave exactly as the classical Maxwell ED
would predict -- given the incident light intensity I(t)=const*n(t) on
the two detectors in a given sampling window, the triggers on one
detector are completely independent on whether the other detectors has
triggered or not in that sampling window. The only dependence between
the two trigger rates (or photocurrents) is due _solely_ to the common
value of n(t), which is what classical EM theory predicts.

The 1956 HBT experiment generated great deal controversy when the
collapse folklore proponents promptly performed their own experiment
contradicting HBT (which they declared to be an experimental error) and
confirming their theoretical prediction. Shortly thereafter it turned
out the collapse experiment and its theoretical justification were
bogus and the HBT were correct (as shown by Purcell in 1956). Check an
interesting interview with Hanbury Brown on this story (rarely
mentioned nowdays):

  http://www.nla.gov.au/events/oralhist/bhathal.html

In the early 1960s, Quantum Opticians (the main promoters of the
"collapse" folklore) recovered from the HBT embarrassment when Glauber
developed modern QO terminology which redefined the terms "counts" and
"correlations" so that they refer to the adjusted counts (the
adjustments are called "standard QO subtractions") obtained by
discarding the sampling results "we are not interested in". The
Glauber's "counts" do show the "collapse", but since G-counts are not
actual counts of anything, the G-collapse is not the collapse (as the
term would be understood by physicists outside of QO).

Unfortunately, the fake experiments of this kind (demonstrating the
"collapse") are being done every few years to this day. The flaws of
the latest "success" are usually not acknowledged until the next
"successful" experiment. In the related ongoing thread in the
sci.physics.research newsgroup, I gave a more specific numeric example
of the "sub-poissonian photo-counts" with some references:

http://groups.google.com/group/sci.physics.research/browse_thread/thr...

There is also a more in depth discussion of Glauber's QO, kicked off by
the taking apart of the most recent (AJP 2004) "demonstration" of the
"collapse" (which was a blatant magic trick, where the cheating was
done by tweaking the coincidence window timings to fall well outside of
the component-ready timing specs, yielding the loss of almost all of
the double triggers), in the PhysicsForum threads:

  http://www.physicsforums.com/showthread.php?t=71297
  http://www.physicsforums.com/showthread.php?t=44964

with cited reference summarized at end of the thread:

 http://www.physicsforums.com/showpost.php?p=544829&postcount=122