Hello Ian,

I have investigated further details of the pt and multiplicity distribution 
of the charged tracks in the CMS pseudorapidity acceptance region.
You suggested me to separate the samples into pt<0.6 GeV and pt>0.6 GeV:
I have followed your suggestion and I have also quoted the effect of the
variation in the parton distribution functions.
As you know, using the most recent pdfs (ex. cteq5L), I expect to be able to 
explore the tails of the charged multiplicity distribution for the 
"large pt-hat" (jet) events. More time (and students) are needed to finalise
such studies.
Here I quote the effects just on the averages (charged multiplicities in
the overall sample and in the two pt regions).
First of all let me quote the Averages for a reference sample (tuned CTEQ4L), single pp interaction:
<n_ch>=39.2
<pt>=0.55 GeV
<n_ch>(pt<0.6)=27.1
<n_ch>(pt>0.6)=11.5 
--> Ratio = <n_ch>(pt<0.6)/<n_ch>(pt>0.6) = 2.36

Effect: Fragmentation
Tested: Delphi Fragmentation Tuning (CERN PPE 96-120) vs
Rel Difference on occupancy -4.6% (overall), -8.2% (pt<0.6), +6.2% (pt>0.6)
Rel Difference on Ratio -13.6%

Effect: PDF
TESTED: GRV94L, CTEQ2L (as example of OLD PDF) vs CTEQ4L (Reference)
Rel Difference on occupancy (GRV94L) +0.2% (overall), -0.8% (pt<0.6), +3.6% (pt>0.6)
Rel Difference on Ratio (GRV94L) -4.4%
Rel Difference on occupancy (CTEQ2L) -2.0% (overall), -0.2% (pt<0.6), -11.8% (pt>0.6)
Rel Difference on Ratio (CTEQ2L) +13.1%

Effect: Multiplicity tuning (UA5,CDF data) (CERN 2000-004)
Tested: Max (and Min) values for the multiple interaction ptmin vs Reference
Rel Difference on occupancy -7.3% (+4.3%) (overall), -6.1% (+4.0%) (pt<0.6), -7.6% (+3.4%) (pt>0.6)
Rel Difference on Ratio Negligible (inside mc stat errors).

Effect: Multiple interaction model
Tested Model 1 vs Model 3 (Reference)
Rel Difference of 6.4% uniform in all the pt regions

It is now up to you to interpretate such numbers in terms of rel. uncertainties. As mentioned
in my previous mail, only the RMS of the multiplicity distributions would change between low and
high luminosity.
I suggest to adopt my scheme in order to identify the systematic "source": Don't mention 
the pt spectrum, I would rather refere to the fragmentation tuning, indeed.
For what concern the PDF it is quite hard to find an agreement. CTEQ2L is certainly ruled out.
GRV94 is certainly "too close" to CTEQ4L (same data are used for the fits).
For the multiplicity tuning I would suggest a conservative approach (just quote the highest numbers).
I hope this does help.

cheers.
	Paolo