Seminar by G.A. Wright

1
Seminar by G.A. Wright Stat 601 Spring 2002
2
(No Transcript)
3
While flying slowly in a patch of flowers, a bee
may encounter an inflorescence every 0.14 s
(Chittka et al., 1999)
4
How do bees recognize floral perfumes among
different flowers?
What characteristics of a floral perfume do they
remember?
5

• Characteristics of floral perfumes
• often, made up of many (100 ) odor compounds
• some compounds are present at high
  concentrations, others are present at low
  concentrations
• (sometimes several orders of magnitude
  difference)

6
Robertson et al., 1993, Phytochemical Analysis
7

• How do floral perfumes vary among individual
  flowers?
• temporal variation diurnally and developmentally
• inter-plant variation individuals, varieties,
  species, and families

8
Variation in major scent components of snapdragon
varieties
N. Dudareva and N. Gorenstein, Purdue Univ.
9
Three parameters of a floral perfume that may
affect the learning and memory of honeybees
1) Types of compounds present
2) Variation in the intensity of the components
3) Intensity of each component relative to the
intensity of the perfume
10
Methods

• Two types of 3-component mixtures
• Similar compounds
• hexanol, heptanol, and octanol
• Dissimilar compounds
• hexanol, geraniol, and octanone

• Two concentrations
• Low 0.0002 M
• High 2.0 M

A 3-component mixture where one odor
concentration is fixed and the others are allowed
to vary randomly at low or high conc. produces 22
4 possible mixture combinations.
11
Three Experiments
1) Constancy of a single odor component
2) Average concentration of each component versus
variation in individual components
3) Variability of all components versus mixture
osmolality
12
Experiment I
Constant odor at the low concentration
Two concentrations used to make odor mixtures
low and high
13
Constant odor at the high concentration
Two concentrations used to make odor mixtures
low and high
14
Using either the similar odors or the dissimilar
odors, each component of the mixture was
systematically held constant Bees were trained
over 16 trials with either - constant odor at
low or - constant odor at high eg. dissimilar
mixture hexanol constant odor Then, they
were tested with each odor component of the
mixture at either - low concentration or -
high concentration eg. dissimilar mixture
tested with hexanol, geraniol, and octanone

15
Proboscis extension by honeybees during
associative conditioning
Trial 1
Trial 2
Trial 3
Odor
Sucrose
Proboscis extension
16
test concentration
constant odor
mixture type
low
low
high
Similar
high
low
high
low
low
high
Dissimilar
high
low
high
17
Used PROC LOGISTIC in SAS for analysis of
data Variables entered in the analysis 1)
Level of the constant odor (coded 0,1) 2) Level
of the test components (coded 0, 1) 3) Identity
of the test components (coded 0, 1) 4) Response
variable 0 no response, 1 response The
analysis was separated by mixture type (similar
and dissimilar)
18
Experiment I
Similar
SAS Output for logistic regression
Parameter DF Estimate SE
Chi-Square Pr gt ChiSq Exp(Est) Intercept
1 0.4055 0.1757 5.3266 0.0210
1.500 mixlev 1 2.0959 0.3729
31.5902 lt.0001 8.133 tstlev 1
-1.0655 0.2527 17.7822 lt.0001
0.345 mixlevtstlev 1 -2.5753 0.4618
31.0968 lt.0001 0.076
19
Experiment I
Dissimilar
SAS Output for logistic regression
Parameter DF Estimate Error
Chi-Square Pr gt ChiSq Exp(Est) Intercept
1 1.0442 0.2625 15.8182 lt.0001
2.841 mixlev 1 1.1816 0.4654
6.4461 0.0111 3.259 tstlev 1
-0.6369 0.3558 3.2051 0.0734
0.529 tstodre 1 0.4446 0.3280
1.8373 0.1753 1.560 mixlevtstodre 1
1.0923 0.4252 6.5980 0.0102
2.981 mixlevtstlev 1 -1.8121 0.5079
12.7287 0.0004 0.163 tstlevtstodre 1
-1.3244 0.4249 9.7167 0.0018
0.266
20
Conclusions of Experiment I
Similar odors Sensory system adaptive gain
control If training (constant) odor is high and
test odor is low, the response to all odors
decreases, and visa versa Dissimilar
odors Gain Control same as for similar
odors Constant odor preferred If training
(constant) odor the same as the test odorant,
then the response to constant odor
increases Interaction between test odorant
identity and odorant intensity Suggestion of an
interaction between variation and intensity
21
Experiment II Average concentration of each
component vs. variation in individual
components
Using either the similar odors or the dissimilar
odors, Bees were trained over 16 trials with
either - a mixture with all odorants at a
constant middle (0.02 M) - or only one odor at
a constant middle (0.02 M), and the others at
either low or high (thus, average middle) Then,
they were tested with each odor component of the
mixture at the low concentration
22
Used PROC LOGISTIC in SAS for analysis of
data Variables entered in the analysis 1)
Experiment type (coded 0,1) 2) Identity of the
test components (coded 0, 1) 3) Response
variable 0 no response, 1 response The
analysis was separated by mixture type (similar
and dissimilar)
23
Experiment II
Similar
Dissimilar
SAS Output for logistic regression
Parameter DF Estimate SE Chi-Square
Pr gt ChiSq Exp(Est) Intercept 1 1.3863
0.5590 6.1497 0.0131 4.000
tstodre 1 -0.7672 0.6499 1.3936
0.2378 0.464 exp 1 -1.7540
0.7075 6.1465 0.0132 0.173
tstodreexp 1 0.6723 0.8404 0.6400
0.4237 1.959
Similar
Parameter DF Estimate SE
Chi-Square Pr gt ChiSq Exp(Est) Intercept 1
-0.6190 0.4688 1.7433 0.1867
0.538 exp 1 1.1045 0.6494
2.8928 0.0890 3.018 tstodre 1
0.9213 0.5675 2.6352 0.1045
2.512 exptstodre 1 -1.6944 0.7882
4.6218 0.0316 0.184
Dissimilar
24
Conclusions of Experiment II
When tested with the low concentration
components Similar odors If the training
odorants are at a constant concentration, the
response to the test odorant increases Dissimilar
odors Constant odor preferred If one of the
odorants is constant in the mixture, the response
to the constant odorant increases Suggestion of
an interaction between variation and intensity
and mixture type
25
Experiment III Variability of all components
versus mixture osmolality
Using either the similar odors or the dissimilar
odors, Bees were trained over 16 trials with
either - a mixture with all odorants at a
constant (0.7 M), producing a mixture with
osmolality 2.1 M - a mixture with all
odorants at varying concentrations producing a
mixture with osmolality 2.0 M - a mixture
with all odorants at varying concentrations
producing a mixture with osmolality 0.03 M
Then, they were tested with each odor
component of the mixture at the low
concentration
26
Used PROC LOGISTIC in SAS for analysis of
data Variables entered in the analysis 1)
Variability (high or low) (coded 0,1) 2)
Mixture osmolality (coded 0, 1) 3) Response
variable 0 no response, 1 response The
analysis was separated by mixture type (similar
and dissimilar)
27
Experiment III
Similar
Dissimilar
SAS Output for logistic regression
Parameter DF Estimate SE Chi-Square
Pr gt ChiSq Exp(Est) Intercept 1 1.5755
0.3950 15.9125 lt.0001 4.833 mixlev
1 1.3802 0.2205 39.1896
lt.0001 3.976 cv 1 -2.3977
0.3469 47.7793 lt.0001 0.091
mixtype 1 -0.6612 0.2089 10.0177
0.0016 0.516
28
Conclusions of Experiment III

• Similar and Dissimilar odors
• The magnitude of the response to the low
  concentration components is a measurable function
  of
• 1) Variation in the concentration of the
  components
• 2) Osmolality of the mixture

29
Conclusions

• Types of compounds present affect generalization
  to constant
• components

2) Variation in the intensity of the components
increases generalization to the components
3) The intensity of the perfume produces an
adaptive gain control which affects the
ability of bees to detect low level components
30
Photo courtesy of NOVA
Acknowledgements Thanks to Brian Smith, Amanda
Mosier, Beth Skinner, Cindy Ford, Joe Latshaw,
Sue Cobey, Natalia Dudareva for the snapdragons
and volatiles data. Funded by NIH.