Item 3a - International Seed Testing Association

S UPPORTING
D ATA AND E VIDENCE
F OR T HE P ROPOSED R ULES
C HANGES 2004 –
ITEM 3A
FOR INFORMATION AND CONSIDERATION AT THE
ORDINARY MEETING 2004

SUPPORTING DATA AND EVIDENCE FOR THE
PROPOSED RULES CHANGES 2004 - ITEM 3A
PAGE 2 OF 22
SUPPORTING DATA AND EVIDENCE FOR PROPOSED RULES CHANGES - ITEM 3A:
CHANGES TO THE UNIFORM BLOWING METHOD FOR POA PRATENSIS VARIETIES
PURITY ANALYSIS OF SMALL SEEDED VARIETIES OF POA PRATENSIS
TABLE OF CONTENTS
0. Summary 2
1. Introduction 3
2. Preliminary work before 1998 4
3. Working group 1998-2002 6
4. Working group 2003-2004 8
5. Conclusions 13
6. Proposed Rules Change 2004 13
7. Literature 15
8. Acknowledgements 15
9. Annex 1: Statistical analyses blower speeds in
purity assessments of Poa pratensis 16

SUPPORTING DATA AND EVIDENCE FOR THE
PROPOSED RULES CHANGES 2004 - ITEM 3A
PAGE 3 OF 22
0. Summary
In 1994 the ISTA Purity Committee received information that seed companies had problems
with the purity test of small seeded varieties of Poa pratensis. Internal work in seed
companies in the early 1990s had revealed that in the purity test of small seeded varieties of
Poa pratensis many seeds with caryopsis were blown into the light fraction, and classified as
inert matter according to the ISTA Rules.
Work carried out by the ISTA Working Group on blowing in the late 1990s indicated that the
percentage of pure seed in the light fraction was 65% (average of 10 small seeded varieties).
If the blower setting was lowered to 0.90, the light fraction still contained more than 40% of
pure seeds.
The subject was discussed at the ISTA Extra Ordinary Meetings in 2002 and in 2003. The
discussion in 2003 revealed that several items needed to be evaluated further, and it was
decided to start a new series of experiments. Laboratories from Denmark, the Netherlands
and the USA were involved. This report summarises the results from the previous years and
present the results received from Denmark and the Netherlands in 2003. USA and Canada
(Representing ISTA and AOSA) participated in planning of the new experiment and in
discussion of the results.
When accepting the factor 0.82 for Poa pratensis in 1992 the results of the blowing by this
factor were compared with the hand method. For that reason it was decided that the main
principle for accepting a new blowing method in the ISTA Rules should be the comparison of
the new blowing factor with the hand method (based on PSD41 in the ISTA Rules).
If the amount of particles classified wrongly is the same for both the heavy and the light
fractions, it is assumed that the result of the hand method will be the same as the result
obtained by means of the blower. The equivalence in number of misplaced seeds in the
heavy and the light fraction is at present used when calibrating the seed blowers with the
ISTA calibration samples.
In the examinations in 2003 participated 5 seed testing laboratories. In total 32 samples of 10
small seeded varieties were blown at the blower settings 1.00, 0.90 (only 22 samples), 0.82,
0.70, and 0.60.
The results of the tests revealed that:
• It was possible to obtain similar results when identical samples are blown with
different blower settings at four seed laboratories.
• Identical results were obtained when seed samples of five varieties with different
1000 seed weight were re-analysed at the same seed laboratory.
• Due to variable results between samples of the same variety and between varieties it
is concluded that the pure live seed (PLS) is not suited for selection of the blower
setting of Poa pratensis.
• The equivalence of g misplaced seeds in the heavy and in the light fraction provided
uniform results for all small seeded varieties tested (1000 seed weight from 0.22-0.34
g).
• The factor 0.82 was the optimal blower setting, and it represents the best possible
compromise between the increase in % pure seed and the decrease in %
germination, when the blower settings are reduced from 1.00 to 0.60.
• The factor 0.82 has already been in use for Poa trivialis at the ISTA seed laboratories
for more than 10 years.
• It is, accordingly, proposed that the factor 0.82 should be included in the ISTA Rules
for blowing of all varieties of Poa pratensis with a 1000 seed weight

SUPPORTING DATA AND EVIDENCE FOR THE
PROPOSED RULES CHANGES 2004 - ITEM 3A
PAGE 4 OF 22
1. Introduction
Breeding of Poa pratensis has diverted in two directions. One direction is aiming at high
yielding varieties for forage purposes. Another direction is aiming at varieties for amenity
purposes, leading to finer leafed types that grow slower.
It appears that leaf size and plant growth are correlated with seed size.
As a consequence forage types are getting heavier seeds, and amenity types are moving
towards smaller and lighter seeds.
In the 1980s new small seeded varieties of Poa pratensis have entered the markets in
Europe. Presently seed companies are already testing varieties having a 1000 seed weight
as low as 0.19g!
The low seed yield of these varieties is compensated by a premium for good genetic quality.
In the cleaning plant it appeared that the purities requested in the market could only be
reached when heavy seed losses during cleaning were accepted.
Internal work in seed companies in the early 1990s revealed that of these low 1000 seed
weight varieties many seeds with caryopsis (‘pure seed’) are blown into the light fraction, and
classified as inert matter according to the ISTA Rules. Seed yields of these varieties in terms
of kilograms per hectare are low as well.
As a consequence of this, good varieties have not reached the market, because on top of an
already low yield of raw material too many valuable seeds had to be cleaned out in order to
reach the purity level as assessed by the present purity test.
In 1994 the ISTA Purity Committee received information that seed companies had problems
with the purity test of small seeded varieties of Poa pratensis. Evidence was presented
indicating that when samples of such varieties were blown on blowers calibrated with the
ISTA calibration sample, a considerable amount of pure seeds were blown into the light
fraction and accordingly considered as inert matter.
In 1996 a questionnaire regarding problems with purity analysis of small seeded varieties of
Poa pratensis was sent out to Danish and Dutch seed companies. The replies were that they
had problems with in total 15 small seeded varieties of Poa pratensis.
The received information lead to some activity late 1996, when Hans Arne Jensen (chairman
of the ISTA Purity committee), Danish Plant Directorate sent a draft working plan to
R. Minnaar of NAK in the Netherlands on how to identify the correct blower setting for such
varieties. A student did some preliminary work, but this was never published.
Due to change of personnel, not much happened during the following year or two. After the
ISTA Congress in 1998 in South Africa, when the results of the preliminary work was
presented by Harry Nijënstein during the meeting of the ISTA Purity Committee, a working
group started some activities in the field of this subject. Leadership of the working group was
taken over by Anny van Pijlen of NAK.
Anny van Pijlen sent an enquiry to all members of the purity working group: Ken Allison
(Canada), Sharon Davidson (USA), Ola Kristian Dille (Norway), Hans Arne Jensen
(Denmark), Maria Rosaria Mannino (France), Harry Nijënstein (the Netherlands), Arnold
Larsen (USA), Elzbieta Malluzynska (Poland).
The question asked was: “Do you have any problems with purity analysis of small seeded
varieties of Poa pratensis? And what actions did you take?”

SUPPORTING DATA AND EVIDENCE FOR THE
PROPOSED RULES CHANGES 2004 - ITEM 3A
PAGE 5 OF 22
Table 1.1:
The results of the enquiry of 1998.
Country
Answer
The Netherlands
Some private companies have problems
Norway
No problems
Denmark
Some private companies have problems
Poland
No problems
France
No problems
USA
No problems
The reasons for not having problems in some countries are probably twofold: In some
countries little or no production of Poa pratensis takes place (Norway, France, Poland). In the
USA seed companies did not produce many small seeded varieties at that time.
However, because it was quite clear that seeds were classified wrongly in some important
growing areas for Poa pratensis, it was decided to continue the work.
The results of the working group were available in 2001. However, due to communicational
problems, these data were never presented to the purity committee and ISTA meetings in
time for change of the ISTA Rules.
The problems with purity tests of small seed varieties and proposals for change of the ISTA
Rules were discussed at ISTA Extraordinary Meetings in 2002 and in 2003.
During the ISTA EOM in Zürich, Switzerland the latest results of the working group were
presented. As it became obvious that the proposed factor 0.90 did not solve the problem and
number of new questions were asked and it was realized that harmonisation with the AOSA
Rules should be examined, a new series of experiments were set up.
This paper wraps up activities and results from: Before1998, from 1998-2002, and from
2003-2004.
2. Preliminary work before 1998
In the 1940s some research on climax blowing point has been carried out. Leggett (1941)
suggested ‘that considerations should be given to a redefinition of pure seed according to a
standard synthetic sample prepared to conform to the “climax” blowing point.
However, there is no proof that these discussions have lead to adoption of this principle for
acceptance of new methods for the ISTA Rules. On the contrary: the paper of Jensen &
Bülow-Olsen (1992) indicates that for adopting the Poa trivialis blower settings in the ISTA
Rules, the results of the blower were compared with those of the hand method.

SUPPORTING DATA AND EVIDENCE FOR THE
PROPOSED RULES CHANGES 2004 - ITEM 3A
PAGE 6 OF 22
The entire purity chapter of the ISTA Rules is based on pure seed definitions. Purity
determinations of other Poa species like Poa secunda/ampla, Poa annua, Poa compressa,
Poa nemoralis, Poa palustris are based on Pure Seed Definition (PSD) 41:
“ Spikelet, with lemma and palea enclosing a caryopsis, with or without awn, plus
attached sterile floret.
Floret, with lemma and palea enclosing a caryopsis, with or without awn.
Caryopsis.
Piece of caryopsis larger than one-half the original size. “
In the present experiments PSD41 is used for classifying seeds.
In the early 1990s, when seed companies encountered problems in reaching the requested
purities for small seeded Poa pratensis, these companies did several experiments in order to
find the cause for these problems. This paragraph summarises the results of these
experiments.
Analyses of small seeded varieties of Poa pratensis revealed that a considerable percentage
of pure seeds (according to PSD 41) were blown into the light fraction (see Table 2.1).
Table 2.1:
Experiment carried out by Cebeco Seeds in 1994.
Germination percentage of pure seeds (PSD 41) from light fraction. Results are averages of
two seed lots per variety, blown at 2*0.5g.
Variety Weight of 1000
seeds
Purity % Pure seed in
light fraction
Limousine 0.23 87.9 12.5 51
Ampellia 0.29 86.9 9.2 21
Julia 0.29 86.6 7.9 15
Enprima 0.33 84.8 8.5 17
Delft 0.35 90.8 6.9 10
Miracle 0.38 88.8 0.9 12
LSD (p

SUPPORTING DATA AND EVIDENCE FOR THE
PROPOSED RULES CHANGES 2004 - ITEM 3A
PAGE 7 OF 22
Table 2.2
Average seed weights, purity and germination percentages of seeds from the light fraction of
eight varieties. Experiments by Mommersteeg in 1995, 2*0.5g.
Variety
Number of
samples
Weight of
1000 seeds
Purity
Germination
of total light
fraction
Variation in
germ. light
fraction
Germination
heavy
fraction
Monopoly 19 0.26 91.3 9.9 05 – 20 92.4
Parade 10 0.33 91.3 3.4 01 – 07 89.8
Cynthia 14 0.33 93.5 18.2 08 – 30 92.6
Asset 11 0.36 90.8 10.1 02 – 23 88.8
Geronimo 25 0.37 92.5 16.2 03 – 36 94.0
Saskia 13 0.37 94.1 5.9 01 – 16 86.2
Minstrel 10 0.39 92.5 2.3 00 – 06 85.2
Cocktail 12 0.43 91.7 1.3 01 - 04 83.1
In a third experiment the relation between purity and germination of Poa pratensis varieties
was investigated (Table 2.3).
Although the purity is increased by 3.2% on average at a lower blower setting, germination
percentage is hardly influenced.
The percentage PLS increases from 79% to 81% when the blower setting is increased.
Table 2.3
Relation between purity and germination percentages of 6 varieties, blown at 4 different
blower settings at 2*0.5g. Results are averages of two samples per variety.
Blower setting figures 2.9 3.0 3.1 3.2
Variety
Weight
of 1000
seeds
Purity Germ. Purity Germ. Purity Germ Purity Germ
Limousine 0.23 94.2 86 91.6 87 88.6 89 88.3 87
Ampellia 0.29 89.6 88 88.2 88 86.9 89 85.8 90
Julia 0.29 91.6 85 90.5 86 89.5 86 88.8 86
Enprima 0.33 91.0 88 89.8 88 88.9 88 88.3 88
Delft 0.35 92.9 87 91.6 86 90.7 89 89.9 91
Miracle 0.38 97.4 90 96.9 87 96.6 87 96.2 87
average 92.8 87 91.4 87 90.2 88 89.6 88
% PLS 81 80 79 79
3. Working group 1998 – 2002
The ISTA working group for blowing consisted of members from the following countries:
Canada, USA, Norway, Denmark, France, the Netherlands and Poland.
The results in the following tables are the averages of work carried out by five laboratories:
three ISTA accredited laboratories in three countries (UK, Denmark, the Netherlands), and
two company laboratories from the Netherlands.

SUPPORTING DATA AND EVIDENCE FOR THE
PROPOSED RULES CHANGES 2004 - ITEM 3A
PAGE 8 OF 22
Table 3.1.
Percentage of pure seed blown into the light fraction and percentages of empty seeds blown
in the heavy fraction. Per variety on sample was subdivided in five sub-samples, and send to
five laboratories: three ISTA accredited laboratories in three countries (UK, Denmark, the
Netherlands), and two company laboratories from the Netherlands.
Each lab blew the samples at 1*0.5g (submitted sample).
Weight
of
Blower setting
0.90
factor
0.90
Blower setting
1.00
factor
1.00
Variety 1000 Pure in light Empty in Pure in light Empty in heavy
seeds
heavy
Unique NL 0.20 22.1 12.2 36.5 14.4
Unique USA 0.26 34.5 0.1 72.8 0.0
Barblue 0.26 57.9 5.3 63.2 6.8
Barmax 0.26 83.3 4.0 92.5 8.1
Limousine 0.26 27.5 4.2 42.5 6.1
Washington 0.28 55.4 0.2 86.6 0.2
Monopoly 0.29 16.7 0.1 84.8 0.0
Julia 0.31 51.4 0.0 68.5 0.0
Midnight 0.33 69.5 0.1 74.3 0.2
Viva 0.47 0.0 0.6 29.0 0.6
average 0.29 41.8 2.7 65.1 3.6
At normal blower settings (1.00) the light fraction contained on average 65.1% of pure seeds.
At a blower setting of 0.90 of the normal setting for Poa pratensis, the light fraction still
contains 41.8% of pure seeds (Table 3.1.). In the average figures also the heavy variety
number 2 (1000 seed weight=0.47g) is included. Averages for pure seeds blown into the light
fraction of small seeded varieties only would even be higher.
Although a blower setting of 0.90 improves the situation considerably, still far too many good
seeds are blown into the light fraction and valued as inert matter.
Table 3.2.
Percentage germination of pure seeds after blowing at two different settings. Per variety on
sample was subdivided in five sub-samples, and send to five laboratories: three ISTA
accredited laboratories in three countries (UK, Denmark, the Netherlands), and two company
laboratories from the Netherlands.
Each lab blew the samples at 1*0.5g (submitted sample).
Weight of blower setting 0.90 blower setting 1.00
Variety 1000 seeds. % germination % germination
Unique NL 0.20 66 69
Unique USA 0.26 83 82
Barblue 0.26 69 78
Barmax 0.26 83 84
Limousine 0.26 87 89
Washington 0.28 90 92
Monopoly 0.29 95 93
Julia 0.31 89 89
Midnight 0.33 88 78
Viva 0.47 92 92
average 0.29 84 85

SUPPORTING DATA AND EVIDENCE FOR THE
PROPOSED RULES CHANGES 2004 - ITEM 3A
PAGE 9 OF 22
In another experiment the effect of blower settings on the germination level was examined.
On average hardly any difference exist between the two blower settings. Varieties number 5
and 4 deviate. The reason for this is not known (Table 3.2.).
In the enquiry of 1996, 15 varieties of Poa pratensis in Europe with a weight of 1000 seeds
lower than 0.35g were reported to provide problems with the purity tests. In 2003 this number
in Europe had increased to 57 (Table 3.3.), and at present more than half of the varieties in
Europe belongs to the group of small seeded varieties.
Apart from the number of small seeded varieties, also the weight of 1000 seeds of newly
breed varieties appears to decrease.
In Europe the weight of 1000 seeds in new varieties tested now are as low as 0.19g.
Table 3.3.
Number of normal (=>0.35g)and of small seeded (=0.35g
weight

SUPPORTING DATA AND EVIDENCE FOR THE
PROPOSED RULES CHANGES 2004 - ITEM 3A
PAGE 10 OF 22
The aim of the new experiment was to elucidate:
• Can similar results be obtained when testing seeds from same seed sample in
different seed laboratories?
• Can similar results be obtained when re-testing seed samples of different varieties in
the same seed laboratory?
• Which blower settings provide both optimal and uniform results when blowing small
seeded varieties of Poa pratensis?
• Can all small seeded varieties be blown at the same blowing point or is it necessary
to have f. inst. one blower setting for varieties with a weight of 1000 seeds between
0.34 and 0.30g and another blower setting for varieties with weight of 1000 seeds

SUPPORTING DATA AND EVIDENCE FOR THE
PROPOSED RULES CHANGES 2004 - ITEM 3A
PAGE 11 OF 22
The experiment comprised in total 32 samples distributed on 10 varieties of Poa pratensis.
Table 4.2
Key figures extracted from table 4.1
Variety
Weight of
1000 seeds
Number
of
samples
Number
of seed
lots
Laboratory
Number
Harmony 0.23 4 2 4
Mardonna 0.24 4 2 4
Unique 0.24-0.25 2 1 1,3
Julia 0.26 1 1 1
Platini 0.26-0.28 4 2 4
Sobra 0.27-0.30 4 2 4
Eprima 0.32-0.34 4 4 1,2,3,5
Cynthia 1 1 5
Saskia 1 1 5
Balin 0.31-0.34 4 2 4
Evora 0.45-0.48 3 1 1,2,3
The overall averages from Table 4.1 show that the percent of pure seed increase from 94.6
to 99.8 when the blower setting is changed from 1.00 to 0.60. The increase in percent pure
seed is specific obvious when the blowing factor is changed from 1.00 to 0.82 (Table 4.3).
For germination on average 9% lower germination is obtained when the blower setting is
changed from 1.00 to 0.60. As Poa pratensis in general have a low germination, the blowing
factors 0.60 and 0.70, which give the lowest germination, should be avoided (Table 4.3).
Table 4.3:
Averages for the heavy fraction: Percentage germination and percentage PLS.
In the right part of the table, all samples of which the blower setting of 0.90 was not tested,
were excluded from calculation of the averages.
See Table 4.2. for information on varieties, number of samples etc.
Pure
live
seeds
Excluding
0,90
blower
setting
Including 0,90 Germination (PLS)
Heavy Germination
blower setting % %
%
% %
0.60 99.8 79 79 0.60 99.9 80 80
0.70 99.4 80 80 0.70 99.7 81 80
0.82 98.2 83 81 0.82 98.8 82 81
0.90 97.7 84 82 0.90 97.7 84 82
1.00 94.6 88 83 1.00 95.5 86 83
Pure
live
seeds
(PLS)
%
The calculated content of pure live seeds (PLS) shows a considerable variation between
samples of same variety as well as between varieties (see Table 4.1). The statistical analysis
revealed that the differences were significant (1% level). This factor is, accordingly, not
suited for establishment of the optimal blowing factor for Poa pratensis.

SUPPORTING DATA AND EVIDENCE FOR THE
PROPOSED RULES CHANGES 2004 - ITEM 3A
PAGE 12 OF 22
The uniformity of testing samples from the same variety was examined for the variety
Enprima. Identical samples were analyzed at four different seed laboratories (gram pure
seeds in light fractions and gram empty seeds in heavy fractions was only recorded at three
laboratories) For this variety the results at blower speed 0.82 showed no significant
differences among laboratories for g light seed in the heavy fraction and g heavy seed in the
light fraction, except for % heavy seeds, where the amount of was slightly less for laboratory
3 than for laboratory 2 and 5. The result at laboratory 1 may seem high, but this is most
probably because this laboratory received seeds from a different sample.
The blower speed 0.82 had on the average the best balance between g pure seeds in light
fraction and g empty seeds in heavy fraction. This was also the case for 7 out of the 11
varieties. For the remaining 4 varieties the absolute difference for the 0.82 blower speed was
only significant larger than the blower speed with best balance for one variety, Julia (where it
was significant at the 4% level).
The variance between repeated analyses of the same sample are for most variables smaller
for data from blower speed 0.82 than for all data. This indicates that the variability between
repeated analyses is lower for this blower speed than for the others. For % heavy seeds the
standard deviation is *0.017=0.13 which show that the probability of exceeding the ISTA
tolerances will be very low.
The ability to re-test seed samples of five different varieties with different seed weight was
examined by one seed laboratory (laboratory no. 4). The results of the re-test for blower
setting 0,82 are presented in table 4.4. In the table is the differences are the differences
between the analysis results compared with the ISTA Tolerances.
Table 4.4:
Differences between two purity tests on same submitted sample of five varieties compared
with the ISTA Tolerances, Table 3.1. Blower setting 0,82. Laboratory 4.
Variety Pair % Heavy seed
1. test 2. test average difference ISTA Tolerance
Harmony I 99.6 99.6 99.6 0.0 0.5
II 99.2 99.2 99.2 0.0 0.7
Maradonna I 97.7 97.9 97.8 0.2 1.1
II 99.1 99.1 99.1 0.0 0.7
Platini I 99.0 99.2 99.1 0.2 0.7
II 99.4 99.4 99.4 0.0 0.6
Sobra I 99.0 99.0 99.0 0.0 0.8
II 98.1 98.1 98.1 0.0 1.0
Balin I 98.9 99.2 99.1 0.3 0.7
II 98.2 98.6 98.4 0.4 1.0
It appears from table 4.4, that the reproducibility within laboratory 4 is good when blowing
small seeded varieties at blower setting 0,82. The differences between pair of purity analyses
were all within the ISTA Tolerances.
If looking at the blowing level where the grams of pure seed in the light fraction are equal to
or nearly equal to the grams of empty seeds in the heavy fraction, a consistent picture
appears (see table 4.1). All varieties up to Enprima show an optimal blowing point of 0.82.
For Balin the blowing point can either be 0.82 (2 samples) or 0.90 (2 samples).

SUPPORTING DATA AND EVIDENCE FOR THE
PROPOSED RULES CHANGES 2004 - ITEM 3A
PAGE 13 OF 22
The variety Evora has a weight of 1000 seeds >0.35 and shall, accordingly, be blown at a
blower setting 1.00.
Returning to the hand method is not an option. The blowing method is both faster and gives
more uniform results than the hand method (Nelson, 1970; Porter & Leggatt, 1942; Leggatt,
1941)
The present ISTA Poa pratensis calibration samples are based on a mixture of commonly
grown varieties with a weight of 1000 seeds ranging from 0.31 to 0.40 g (average 0.35 g).
Care was taken that the new calibration samples gave the same purity results as those
produced in the Netherlands 10-15 years ago.
The differences in seed weight between Poa pratensis and Poa trivialis did for many years
prevent the ISTA stations from using the blowing method for Poa trivialis. AOSA was the
pioneer in solving the problem and introduced the multiplication factor 0.82, when blowing
Poa trivialis on a blower, calibrated with a Poa pratensis calibration sample. Within the ISTA
Purity Committee a study was initiated, and this study confirmed the AOSA factor 0.82
(Jensen & Bülow-Olsen, 1992). The factor of 0.82 for Poa trivialis was introduced into the
ISTA Rules in 1992.
If we look at the 1000 seed weight’s of the small seeded varieties tested in these
experiments, they appear to be on the same level as the weight of 1000 seeds for Poa
trivialis.
Data received from the Netherlands and Denmark give the following weight of 1000 seeds for
varieties of Poa trivialis: Sabre: 0.24 (two lots), Solo: 0.21 (six lots), and Dasas: 0.24 (ten
lots).
Therefore, it is not a surprise that the suggested optimal blower setting 0.82 for small seeded
Poa pratensis varieties is the same as the one already in use for Poa trivialis.
Choosing the already existing blower setting for Poa trivialis has the additional advantage
that no extra blowing points have to be added. This makes it easier for all involved to
implement the blowing factor also for small seed varieties of Poa pratensis.

SUPPORTING DATA AND EVIDENCE FOR THE
PROPOSED RULES CHANGES 2004 - ITEM 3A
PAGE 14 OF 22
5. Conclusions
Based on previous results and the results from this study, it can be concluded that:
• It was possible to obtain similar results when identical samples are blown with
different blower settings at four seed laboratories.
• Identical results were obtained when seed samples of five varieties with different
1000 seed weight were re-analysed at the same seed laboratory.
• Due to variable results between samples of the same variety and between varieties it
is obvious that Pure Live Seed (PLS) is not suited for selection of the blower setting
of Poa pratensis.
• The equivalence of g misplaced seeds in the heavy and in the light fraction provided
uniform results for all small seeded varieties tested (1000 seed weight from
0.22-0.34 g).
• The factor 0.82 was the optimal blower setting for all small seeded varieties tested,
and it represents the best possible compromise between the increase in % pure seed
and the decrease in % germination when the blower settings are reduced from 1.00
to 0.60.
• The factor 0.82 has already been in use for Poa trivialis at the ISTA seed laboratories
for more than 10 years.
• It is, accordingly, proposed that the factor 0.82 should be included in the ISTA Rules
for blowing of all varieties of Poa pratensis with a 1000 seed weight

SUPPORTING DATA AND EVIDENCE FOR THE
PROPOSED RULES CHANGES 2004 - ITEM 3A
PAGE 15 OF 22
List of varieties of Poa pratensis with an average weight of 1000 seeds

SUPPORTING DATA AND EVIDENCE FOR THE
PROPOSED RULES CHANGES 2004 - ITEM 3A
PAGE 16 OF 22
7. Literature
Brown … , Porter … (1935). An improved method of testing seeds of Kentucky Bluegrass
(Poa pratensis L.). Proceedings AOSA 27: 44-49.
Jensen, H.A. & Bülow-Olsen, A. (1992) Comparison of purity analysis results from Poa
trivialis L. samples tested according to the ISTA and the AOSA method. Seed Science &
Technology 20: 655-661.
Leggatt, C.W. (1941) The “Climax” blowing point in the testing of grass seeds for
percentages of pure live seed. Contr.: No. 657. Division of Botany and Plant Pathology,
Science Service, D. of A. Ottawa, Canada.
Nelson, B. (1970) Uniform Blowing Procedure for Canada and Rough Bluegrass
Subcommittee. Proceedings of the Association of Official Seed Analysts. 60: 24-25.
Porter, R.H., Leggatt CW (1942) A new concept of pure seed as applied to seed technology.
Scientific Agriculture 23(2): 80-103.
8. Acknowledgements
The working group acknowledges the analytical work performed by the participating seed
laboratories, the valuable comments received during preparation of the manuscript and the
statistical analysis carried out by Dr. Kristian Kristensen, Biometry Research Unit, Danish
Institute of Agricultural Sciences.

SUPPORTING DATA AND EVIDENCE FOR THE
PROPOSED RULES CHANGES 2004 - ITEM 3A
PAGE 17 OF 22
Annex 1:
Statistical analyses blower speeds in purity assessments of Poa pratensis
Kristian Kristensen
Biometry Research Unit, Department of Animal Breeding and Genetics
Danish Institute of Agricultural Sciences
November 2003
Method
Each variable were first analysed in an analysis that took into account the effects of variety, laboratory and
blower speed as well as all two-way interactions and the three-way interaction. The residual effects were
separated into 1) within sample variation and 2) between sample variation. Mathematically the model may be
written as:
Y = µ + α + β + γ + ( αβ ) + ( αγ ) + ( βγ ) + ( αβγ ) + D + E
vlbsr v l b vl vb lb vlb vs vlbsr
where
Y is the observed value for variety v at laboratory l with blower speed b for replicate r of sample s
D
E
vlbsr
vs
vlbsr
is the random effect of sample s in variety v
is the random effect within sample
D and E are asumed to be independently normally distributed with mean zero and vaiances σ
2
vs vlbsr D
and σ , respectively
2
E
The greeck letters are the fixed effects.
Next all data with blower speed 0.82 were analysed in a model that took into account the effects of variety and
laboratory as well as the two-way interaction. The residual effects were separated into 1) within sample variation
and 2) between sample variation. Mathematically the model may be written as:
Y = µ + α + β + ( αβ)
+ D + E
vlsr v l vl vs vlsr
where
Y is the observed value for variety v at laboratory l with blower speed 0.82 for replicate r of
vlsr
sample s
D is the random effect of sample s in variety v
E
D
vs
vlsr
vs
σ
is the random effect within sample
and E
vlsr
2 2
D E
are asumed to be independently normally distributed with mean zero and vaiances
and σ , respectively
The greeck letters are the fixed effects.
The set of varieties analysed differed from laboratory to laboratory in such a way that the table variety by
laboratory were disconnected. Therefore, in order to estimate the effects of the factors (variety, blower speed and
laboratory) it was necessary to separate the data into two groups: 1) data from laboratory 4 and 2) data from
laboratory 1, 2, 3 and 5. In addition it was necessary to delete the blower speed 0.90 in the second group
(laboratory 1, 2, 3 and 5). In the estimation also all interactions with laboratory were excluded (they were nonsignificant
or almost non-significant for all variables (see table 1 and 4)).

SUPPORTING DATA AND EVIDENCE FOR THE
PROPOSED RULES CHANGES 2004 - ITEM 3A
PAGE 18 OF 22
Results
Model check
The model was checked for the underlying assumptions using graphical methods. The graphics were constructed
to check whether the variability dependent on the response, whether the residuals looked normally distributed
and whether deviating observations were present. This check revealed that rather large discrepancies were
present for the two replicates of Mardona for blower speed 1 at laboratory 4. However, the discrepancies were
judged to be of a size that could have happened by change, so the observations were kept in the analyses.
All data
The analysis showed that there were a significant interaction between blower speed and variety for all characters
except for % pure life seeds. The blower speed had a significant effect on all variables. For only one variable, g
empty seeds in heavy fraction, significant differences between laboratories were found. See table 1. The last
column, Balance, is the difference between g pure seeds in light fractions and g empty seeds in heavy fractions.
The variance components are shown in the bottom part of table 1. The residual variance is the within samples
variation. The sample component is the additional variation when the results comes from different samples. They
were for most variables of similarly size - only for the variables balance and g pure seed in light fraction was the
component for between sample considerable less that the component for within sample. For those two variables
most of the variance were caused by the within sample variability. For the other variables the variance between
two observed values was approximately doubled if the observed values come from different samples compared
to observed values from the same sample.
The effect of Blower speed is shown in table 2. The effect of blower speed was of the same type for both groups
of laboratories although the effects were lager for laboratory 1, 2, 3 and 5 than for laboratory 4. For the variable
g empty seeds in heavy fraction the magnitude of the figures were considerable larger for laboratory 1, 2, 3 and 5
than for laboratory 4. However, this was mainly caused by the variety Julia (see table 7).
The blower speed 0.82 had on the average the best balance between g pure seeds in light fraction and g empty
seeds in heavy fraction (table 2). This was also the case for 7 out of the 11 varieties (table 7). For the remaining 4
varieties the absolute difference for the 0.82 blower speed was only significant larger than the blower speed with
best balance for one variety, Julia (where it was significant at the 4% level).
Table 1 Significance of fixed effects and estimated variance components for each variable
using all data
Effect name % heavy Germination % pure live g pure in light g empty in Balance, g
seeds percent seeds frac. heavy frac.
Significance of fixed effects *
Variety ** ** * *** **
Laboratory ***
Blowers. *** *** *** *** *** ***
V×L *
V×B *** ** (*) *** *** ***
L×B
V×L×B
Estimated random effects
Sample 0.647 5.698 8.515 0.000016 0.000003 0.000008
Residual 0.607 9.001 7.995 0.000045 0.000005 0.000058
* ) One, two of three asterisks denote that the effect is significant at the 5%, 1% or 0.1% level,
respectively

SUPPORTING DATA AND EVIDENCE FOR THE
PROPOSED RULES CHANGES 2004 - ITEM 3A
PAGE 19 OF 22
Table 2 Estimated effects of Blower speed
Blower speed % heavy seeds Germinatio % pure live mg pure in mg empty in Balance, mg
n percent seeds light frac. heavy frac.
Laboratory 4
0.60 100.0 79.3 79.3 0.0 7.0 -7.0
0.70 99.8 80.4 80.2 0.1 6.1 -6.0
0.82 98.9 81.0 80.2 1.7 2.0 -0.3
0.90 97.8 82.7 81.0 7.5 0.3 7.2
1.00 95.5 85.6 81.9 28.7 0.0 28.7
Laboratory 1, 2, 3 and 5
0.60 99.4 77.7 77.3 0.0 25.2 -25.5
0.70 98.7 80.0 79.0 0.4 22.4 -22.1
0.82 96.9 85.1 82.5 5.8 14.7 -9.0
1.00 91.7 90.0 82.6 35.2 11.1 23.9
For the variable g empty seeds in heavy fraction there was a significant effect of laboratory. This was mainly
caused by laboratory 1 that had a high value and laboratory 4 that had a low value (table 3). However, it should
be noted that the difference between laboratory 4 and the other three couldn't be separated from the effect of the
groups of varieties analysed at the two laboratories.
Table 3 Effect of laboratory for the variable mg empty seeds in heavy fraction
Laboratory Lab 4 Lab 1,2,3,5
1 25
3 13
4 3
5 17
Data at blower speed 0.82
When analysing data where seeds were cleaned at the blower speed 0.82 significant effects of variety were found
for most variables. For the same data significant effects of laboratory were found for 3 variables (see table 4). I
addition a specific test for differences between laboratories for the variety Enprima was added (Enprima was the
only variety analysed by all samples). Significant differences between laboratories were only found for % heavy
seeds and were mainly caused by the relative low value at laboratory 3 (note: laboratory 1 received a different
sample).
The variance components (bottom part of figure 4) are for most variables smaller for data from blower speed
0.82 than for all data. This indicates the variability between repeated analyses are lower for this blower speed
than for the others. For % heavy seeds the standard deviation is √0.017=0.13 which show that a difference larger
than 0.6 between two samples will be very rare.
Table 4 Significance of fixed effects and estimated variance components for each variable
using only data from blower speed 0.82.
Effect name % heavy Germination % pure live g pure in light g empty in Balance, g
seeds percent seeds frac. heavy frac.
Significance of fixed effects *
Variety ** ** ** *** ***
Laboratory *** *** **
V×L *
Laboratories *
for Enprima
Estimated random effects
Sample 0.309 1.100 1.827 0.000002 0.000003 0.000005
Residual 0.017 11.450 11.256 0.000003 0.000001 0.000038
* ) One, two of three asterisks denote that the effect is significant at the 5%, 1% or 0.1% level, respectively

SUPPORTING DATA AND EVIDENCE FOR THE
PROPOSED RULES CHANGES 2004 - ITEM 3A
PAGE 20 OF 22
Table 5 Estimated effects of varieties for blower speed 0.82
Blower speed
% heavy
seeds
Germination
percent
% pure live
seeds
mg pure in
light frac.
mg empty in
heavy frac.
Balance, mg
Laboratory 4
Balin 98.7 81.5 80.5 0.9 6.2 -5.3
Harmony 99.4 86.5 86.0 0.2 0.3 -0.1
Mardona 98.5 62.5 61.5 2.3 1.5 0.8
Platini 99.2 84.5 83.9 4.0 0.6 3.4
Sobra 98.5 90.3 88.9 0.9 1.3 -0.4
Laboratory 1, 2, 3 and 5
Cynthia 95.5 87.0 83.1 5.5 2.3 3.1
Enprima 97.7 88.0 85.9 2.0 3.1 -1.2
Evora 98.0 93.3 91.4 -1.7 4.1 -5.8
Julia 95.2 84.5 80.4 13.8 66.3 -53.0
Saskia 98.7 86.0 85.0 1.7 3.2 -1.6
Unique 95.9 69.2 66.3 14.3 8.3 6.0
The estimated effects for each variety are shown in table 5. For germination and % pure live in seeds the
varieties Mardona and Unique had significant low values. For the three variables mg pure seeds in light fraction,
mg empty in heavy fraction and balance Julia had values that deviated from most other varieties. The variety
Unique had a very high value for the variable mg pure in light fraction.
The estimated effects for each laboratory are shown in table 6. For the variable % heavy seeds laboratory 4 and 1
had higher values than the other, bur only laboratory 1 and 3 differed significantly (laboratory 4 can not be
compared statistically with the other laboratories). For the variables mg empty seeds in heavy fraction and
balance the values at laboratory 4 are much closer to zero, than for the other 3 laboratories. However, the
estimates for laboratory 4 cannot be compared statistically with the values for the other laboratories.
Table 6 Estimated effects of laboratories for blower speed 0.82
Blower speed
% heavy
seeds
Germination
percent
% pure live
seeds
mg pure in
light frac.
mg empty in
heavy frac.
Balance,
mg
Laboratory 4
4 98.9 81.1 80.2 1.7 2.0 -0.3
Laboratory 1, 2, 3 and 5
1 97.4 81.2 79.2 8.4 21.7 13.2
2 96.7 82.0 79.3
3 96.2 85.8 82.6 5.0 9.3 -4.3
5 96.9 89.7 86.9 4.2 12.7 -8.5

SUPPORTING DATA AND EVIDENCE FOR THE
PROPOSED RULES CHANGES 2004 - ITEM 3A
PAGE 21 OF 22
Table 7 Estimated blower by variety effects for two groups of laboratories
Blower
Laboratory 4 Laboratory 1, 2, 3 and 5
speed Balin Harmony Mardona Platini Sobra Cynthia Enprima Evora Julia Saskia Unique
% heavy seeds
0.60 100 100 100 100 100 99.4 99.8 99.1 98.7 99.9 99.6
0.70 99.6 99.8 99.9 99.9 99.7 98.1 98.9 98.8 98.1 99.5 98.7
0.82 98.7 99.4 98.5 99.3 98.5 95.5 97.6 98.0 95.7 98.7 95.9
0.90 97.2 99.1 96.1 98.8 97.8 . . . . . .
1.00 95.1 98.0 89.8 97.9 96.6 88.5 95.2 96.3 84.7 97.0 88.5
Germination
0.60 78.3 86.0 62.2 83.2 87.0 80.9 83.7 87.4 73.6 85.9 54.7
0.70 80.8 86.3 62.5 84.3 88.3 78.9 84.5 87.4 80.6 90.9 57.7
0.82 81.5 86.5 62.5 84.5 90.2 89.9 88.7 92.4 82.6 88.9 68.2
0.90 84.8 87.0 62.2 86.0 93.5 . . . . . .
1.00 87.0 88.8 73.0 86.5 92.8 90.9 94.5 93.7 89.6 90.9 80.2
% pure live seeds
0.60 78.2 86.0 62.3 83.3 87.0 80.4 83.6 86.6 72.8 85.8 54.6
0.70 80.4 86.1 62.4 84.2 88.0 77.3 83.6 86.3 79.2 90.4 57.0
0.82 80.5 86.0 61.5 83.9 88.9 85.7 86.7 90.5 79.2 87.7 65.4
0.90 82.4 86.2 60.0 84.9 91.5 . . . . . .
1.00 82.7 87.0 65.6 84.7 89.6 80.2 90.0 90.2 76.2 88.1 71.1
mg pure seeds in light fraction
0.60 0.1 -0.0 0.0 -0.0 0.1 2.1 0.0 -3.0 -2.7 2.0 0.6
0.70 0.1 0.1 0.2 -0.0 0.1 2.0 0.2 -2.5 0.6 2.0 0.4
0.82 1.0 0.3 2.3 4.0 0.9 5.8 2.0 -2.2 13.3 2.0 14.1
0.90 5.5 3.0 18.5 6.0 4.7 . . . . . .
1.00 20.7 13.0 79.1 14.8 16.0 39.6 13.5 5.6 98.1 9.7 44.6
mg empty seeds in heavy fraction
0.60 12.5 4.1 5.8 3.2 9.4 6.1 16.2 15.7 91.1 4.9 17.1
0.70 11.1 1.9 8.1 2.5 6.7 6.6 9.4 10.4 96.8 3.9 7.4
0.82 6.3 0.3 1.5 0.6 1.3 1.9 3.0 4.9 67.1 2.8 8.5
0.90 1.1 0.0 0.2 0.1 0.2 . . . . . .
1.00 0.0 -0.0 -0.0 0.0 0.0 1.5 1.0 -4.8 64.0 1.4 3.7
Balance, mg
0.60 -12 -4.2 -5.7 -3.2 -9.4 -3.7 -16 -19 -94 -2.6 -17
0.70 -11 -1.8 -7.9 -2.5 -6.6 -4.3 -9.4 -13 -97 -1.6 -7.2
0.82 -5.3 -0.1 0.8 3.4 -0.4 4.2 -1.2 -7.5 -54 -0.5 5.4
0.90 4.4 2.9 18.3 5.9 4.5 . . . . . .
1.00 20.7 13.0 79.1 14.8 16.0 38.4 12.3 10.0 33.7 8.6 40.7