Title: Florida Entomologist
ALL VOLUMES CITATION DOWNLOADS THUMBNAILS PAGE IMAGE ZOOMABLE
Full Citation
STANDARD VIEW MARC VIEW
Permanent Link: http://ufdc.ufl.edu/UF00098813/00274
 Material Information
Title: Florida Entomologist
Physical Description: Serial
Creator: Florida Entomological Society
Publisher: Florida Entomological Society
Place of Publication: Winter Haven, Fla.
Publication Date: 1938
Copyright Date: 1917
 Subjects
Subject: Florida Entomological Society
Entomology -- Periodicals
Insects -- Florida
Insects -- Florida -- Periodicals
Insects -- Periodicals
 Notes
General Note: Eigenfactor: Florida Entomologist: http://www.bioone.org/doi/full/10.1653/024.092.0401
 Record Information
Bibliographic ID: UF00098813
Volume ID: VID00274
Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: Open Access

Downloads

This item has the following downloads:

PDF ( 1 MBs ) ( PDF )

PDF ( PDF )

PDF ( PDF )


Full Text





bhe


Florida Entomologist
Official Organ of the Florida Entomological Society
VOL. XXI OCTOBER, 1938 NO. 3

ADVANTAGES OF STATISTICAL METHODS FOR THE
PRACTICAL ENTOMOLOGIST
By L. B. REED, Junior Entomologist
Bureau of Entomology and Plant Quarantine
United States Department of Agriculture
During the past few years statistics has joined other branches
of science in the rapid march of progress. Fifteen years ago
most research men and some of the leading statisticians rec-
ognized that the statistical methods available were not suitable
for practical research problems. The elaborate methods of that
day were based upon the theory of infinitely large samples and
were not sufficiently accurate for practical data obtained from
limited observations. Those who attempted to apply statistics
to practical problems usually became so involved in complicated
mathematics that they either lost sight of the practical aspects
of their problem or gave up statistics as hopeless. One authority
said of the traditional machinery of statistical processes: "Not
only does it take a cannon to shoot a sparrow but it misses the
sparrow!" The author of this statement, Professor R. A. Fisher,
recognized the trouble and successfully introduced a remedy.
He developed a system by which data from small experiments
could be analyzed and judged on their own merits by means of
ordinary arithmetic. His application of the analysis of variance
method to small experiments revolutionized this phase of statis-
tics and furnished to the agricultural science worker a practical
tool which enables him to evaluate properly the results of his
undertakings.
The object of this paper is to point out the value of this
relatively new Analysis of Variance method in ordinary ento-
mological experiments. It is unnecessary to tell a group of
entomologists that individual insects of the same species will
not always react to changes in environment in exactly the same
manner. Their actions will vary according to sex, age, health,
weather conditions, and other factors. For example, a recent









THE FLORIDA ENTOMOLOGIST


investigation determined that female house flies are about twice
as hard to kill as male flies of the same age and breeding, also
that young flies of either sex are harder to kill than older ones.
The results of our simplest experiments vary, depending upon
numerous Conditions that we cannot hope to keep constant. Even
if it were possible to keep these conditions constant, our experi-
ments would be so limited in their application to practical prob-
lems that they would lose much of their significance. Thus for
ordinary experiments we must accept the fact that the results
will be variable and must make allowance for that variation,
which tends to cause incorrect conclusions.
In some simple experiments where wide differences are ex-
hibited between treatments, and the discrepancy is small, correct
evaluation may be made by simple examination without statis-
tical analysis. However, unless recognized methods of analysis
are used, there is danger that different investigators will draw
different conclusions from the same simple set of data. The
method of analysis of variance aids the investigator in making
precise evaluations, and it inspires confidence in his results
because others are informed of the degree of precision with
which his experiments were conducted.
The analysis of variance places numerical values on the vari-
ation arising from the several sources. These numerical values
are known as "variances". Thus, the analysis of variance is
just what the name implies. If we analyze a sample of fer-
tilizer, we weigh the sample, then separate and weigh the dif-
ferent plant foods or active ingredients; we then subtract the
weight of the active ingredients from the total weight and we
have left the weight of the inert ingredients. In analyzing
data, we calculate the total variance, then we separate and
measure the variances due to treatments or other controlled
factors. We subtract those variances that are under our con-
trol from the total and have left the error variance which we
could not control. By comparing the treatment variance with
the error variance, we can tell immediately whether the experi-
mental results are reliable or not. Or by use of the error vari-
ance and a table of odds, one can estimate the odds in favor
of any one treatment in comparison with any other treatment.
Each experiment is judged on its own merits, and the degree
of confidence placed in it is automatically adjusted according to
the number of replicates made and the error encountered.









VOL. XXI-No. 3 35

However, one should not get the idea that the statistical
method is a mill through which data may be passed and the
final interpretations produced mechanically without use of per-
sonal judgment. This would be as foolish as expecting an
unskilled laborer to build a fine house just because he had the
best available carpenter tools. Even the best carpenter needs
all of his senses, proper building materials, and a well-planned
design. Nevertheless, a carpenter can use his tools without
being able to manufacture them. The statistical method is
merely a tool and does not interpret data, neither does it do
our thinking. Yet we can make practical use of this tool with-
out understanding all of the complicated theory upon which
its operation is based.
Statistical methods have done more for us than make it
practical to measure the reliability of results. They have pro-
vided methods of testing technique and have made possible the
design of more efficient and more comprehensive experiments.
In this connection it should be emphasized that the form of
analysis appropriate for any data depends upon the design of
the experiment producing them. If the requirements for
replication and randomization are not fulfilled, the standard
methods of analysis are not at all applicable. At least two
replicates must be provided in any experiment and it is prefer-
able to have four or more. The test plots or test insects should
be assigned to the treatments by some method of deliberate
randomization. Any deviation from absolute randomization
invalidates the analyses unless the effects of such departures
are measured and due adjustments made. A third requirement
is that the experiment contain checks or controls in order to
avoid the possibility of unseen accidents causing apparently
positive results. Thus, if a single insecticide is tested, its
effect must be compared either with the effects of a well-known
insecticide, or with the effects of some suitable control; other-
wise there is no certainty whether the results are due to the
insecticide in question or due to an accident.
The simplest experiment that employs the best principles
of experimentation consists of pairing the test plots or insects
so as to get the members of a pair as nearly alike as possible.
Then one random member of each pair is exposed to the condi-
tion to be tested, while the other member is either observed
as an undisturbed check or is exposed to some alternative. In
such an experiment the variation between pairs is not error









THE FLORIDA ENTOMOLOGIST


because it affects the two individuals of any one pair alike.
Therefore, a separate item is provided in the analysis of vari-
ance for pairing and the variance due to pairing is thereby
excluded from the estimate of error. This method is very
effective in reducing experimental error and employs in sim-
plified form the chief principles embodied in the more complex
designs. However, it is limited to the testing of only two treat-
ments.
In field experiments with three or more treatments, the
"randomized block" arrangement of plots is popular. This con-
sists of dividing the experimental area into as many equal sized
blocks as there are to be replicates, with an attempt to get the
blocks as uniform within as possible and to place the major
differences between the blocks. Each block is divided into as
many plots as there are treatments to compare. The treat-
ments are then assigned at random to the plots with the restric-
tion that each treatment appear once and only once in each
block. In this arrangement the differences between blocks have
similar effects upon all treatments and do not tend to cause
erroneous results. Thus, a portion of the usual experimental
error loses its effect and ceases to be error. Skill in the arrange-
ment of blocks to conform with the variations in soil, plants,
and insects results in greater precision. The randomized-block
design was developed for field experiments but the principle of
grouping exemplified by it is being used with equal success in
laboratory experiments.
When from four to seven treatments are to be tested and
it is convenient to use the same number of replicates, the "latin
square" design is appropriate and will as a rule be more efficient
than randomized blocks. If more than six or seven treatments
are to be tested, the latin square becomes too cumbersome and
is not recommended. A four by four latin square requires 16
plots arranged in a rectangle with 4 plots on each side. The four
experimental treatments are assigned at random to the plots
except that each treatment must occur once and only once in
each row of plots in two directions across the field. This design
can be employed in laboratory experiments but it is chiefly
used in small-plot field experiments. The double grouping of
plots featured by it provides for a double reduction of error. It
also insures a representative distribution of treatments through-
out the experimental area and thereby prevents the chance place-
ment of like-treated plots side by side as is sometimes done in









VOL. XXI-No. 3


randomized blocks. The latin square is especially appropriate
in the situation so often encountered in which a reduction of
error is greatly needed but little is known of its source or dis-
tribution. In other words, the rigidness of the latin square
makes it safe and more foolproof for general use. However,
this quality also limits its adaptability to special conditions,
and situations may arise in which it is less efficient than the
more flexible randomized blocks. The latin square is preferred
for the arrangement of field plots unless situations do arise
that definitely indicate some other arrangement.
Often it is desirable to test in the same experiment several
control schedules and several different insecticides and to deter-
mine what influence the insecticides have upon the schedules.
If there are four schedules and four insecticides, a total of 16
combinations will be possible and unless all 16 are tried there
is no certainty just which would be best. The factoriall design"
is appropriate for experiments of this type and can be used with
or without randomized blocks. It is especially efficient in labor-
atory experiments where the natural distribution of plants, soils,
and insects does not limit the number of treatments that can
be effectively handled at one time. For example, at Sanford,
Fla., last fall all combinations of 3 poisons, 3 rates of application,
2 diluents, and 2 dosages were tested simultaneously, there being
in all 36 treatments. The large blocks ordinarily required to
conduct such an experiment in the field are a disadvantage. How-
ever, if circumstances make it sufficiently important to do so,
there are ways of handling a large number of treatments with
small blocks.
In investigations of insects affecting winter vegetables we
have found the randomized blocks, the latin square, and the
factorial design all to be useful and easy to put into operation.
Any one who confines himself to the simpler experiments can
with little outside aid learn to analyze results obtained from
these experiments. Simple instructions are given by Fisher
and Wishart (1930) or by Paterson (1933), while tables for
making tests for significance are given and explained by Snedecor
(1934). The bulk of the calculations consists of adding numbers
from a table of squares and can be done by any careful assistant
that can operate an adding machine. The chief requirement,
then, is the use of common sense in designing the experiments
and in making interpretations.










THE FLORIDA ENTOMOLOGIST


REFERENCES
BARLOW, PETER. 1930.
Barlow's Tables of Squares, Cubes, Square Roots, Cube Roots, and
Reciprocals of All Integer Numbers Up to 10,000. Third edition
(L. J. Comrie, Ed.) E. and F. N. Spon. London.
FISHER, R. A. 1936.
Statistical Methods for Research Workers. Sixth edition, revised and
enlarged. Oliver & Boyd. London.
FISHER, R. A. 1937.
The Design of Experiments. Second edition. Oliver & Boyd. London.
FISHER, R. A., and J. WISHART. 1930.
The Arrangement of Field Experiments and the Statistical Reduction
of the Results. Imperial Bureau of Soil Science: Technical Com-
munication No. 10. Published by His Majesty's Stationery Office.
London.
PATERSON, D. D. 1933.
Experimentation and Applied Statistics for the Practical Agriculturist.
Tropical Agriculture, Vol. X, Nos. 10, 11, and 12.
SNEDECOR, GEORGE W. 1934.
Calculation and Interpretation of Analysis of Variance and Covariance.
Collegiate Press, Inc. Ames, Iowa.
SNEDECOR, GEORGE W. 1938.
Statistical Methods Applied to Experiments in Agriculture and Biology.
Revised. Collegiate Press, Inc. Ames, Iowa.
TIPPETT, L. H. C. 1927.
Random Sampling Numbers. Tracts for Computers, No. XV. Cam-
bridge University Press. London.

FOR SALE
Index to Florida Entomologist, Vols. I to XIX inclusive ....$ .75
Volumes I to XX Florida Entomologist .-.....-.........--....---...... 21.75
Back Numbers of the Florida Entomologist (single No.) .... .35
Series of papers on systematic subjects such as Blatchley's
papers on Coleoptera of Florida, Professor J. R. Watson's papers
on Thrips and other systematic papers can be furnished.
Orders should be sent to J. W. Wilson, Business Manager,
Belle Glade, Florida.

A THIRD OF A CENTURY OF EXPERIENCE

W. W. OTHERS
Consulting Entomologist
457 Boone Street, Orlando, Fla.
Advisory Work Confined to Citrus
Citrus Literature Bought and Sold Without Profit
REPORTS AND APPRAISALS OF CITRUS PROPERTY









Uhe
FLORIDA ENTOMOLOGIST
Official Organ of the Florida Entomological Society
Gainesville, Florida

VOL. XXI OCTOBER, 1938 No. 3

J. R. WATSON, Gainesville --..- ----................--...................... Editor
E. W. BERGER, Gainesville--...................................Associate Editor
J. W. WILSON, Belle Glade..........----..--.......Business Manager
Issued once every three months. Free to all members of the
Society.
Subscription price to non-members is $1.00 per year in ad-
vance; 35 cents per copy.

LABORATORY STUDIES OF POISONED BAITS FOR THE
CONTROL OF THE SOUTHERN ARMYWORM
By C. B. WISECUP and L. B. REED
U. S. Department of Agriculture
Bureau of Entomology and Plant Quarantine
The southern armyworm, formerly called the semitropical
armyworm, Prodenia eridania (Cram.), is at times one of the
most destructive pests on truck crops in Florida. It feeds on
a great variety of host plants, notably celery, cabbage, tomato,
sweetpotato, Irish potato, pepper, beet, carrot, and cowpea. In
the younger stages the larvae have habits similar to those of
the true armyworms, as indicated by the common name, but
as they near maturity they take on the characteristics of cut-
worms, such as a tendency to feed alone, to hide during the day,
and to cease migrating any great distance.
Quite often this insect appears in large numbers on vege-
table crops just prior to harvest, causing a definite loss unless
controlled by artificial means. To date the only successful
control measures have been applications of arsenical or fluorine
dusts and sprays, which may be responsible for dangerous resi-
dues on the marketed product. For the past several years
studies at the Sanford, Fla., laboratory of the Bureau of Ento-
mology and Plant Quarantine have been directed towards the
development of control measures that would employ materials
least objectionable from the residue standpoint. As a part of
this program, poisoned baits were investigated because (1)
arsenical baits are less dangerous from a residue viewpoint than
are arsenical sprays or dusts, (2) they are readily fed upon









THE FLORIDA ENTOMOLOGIST


in cages by the larger larvae, and (3) they have not been
entirely satisfactory in practical use under field conditions.
OBJECT
The hypothesis upon which this investigation was based
was the belief that poisoned baits would be effective in the
control of the southern armyworm under field conditions, pro-
vided they were made more attractive than the natural food
by combining sufficient quantities of the proper poisons, foods,
and attrahents. The ultimate objective, of course, was to find
this ideal combination and to prove its effectiveness. The two
experiments discussed herein have necessarily been limited to
the testing of only a few of the many possibilities. The purpose
of this report is not merely to make known the results of these
tests but also to describe and illustrate an experimental method
that is widely applicable to entomological problems.
The mortality caused by a bait is affected by several dif-
ferent factors, chief of which are the availability of normal
food, the kind of poison, the quantity or rate of poison, the
kind of food used as bulk in the bait, and the kind of attrahent.
A poison may be more effective with certain foods or attrahents
than with others; likewise, an attrahent may be more effective
in certain combinations than in others. Thus the objective
was not only to test the influence of each factor upon the mor-
tality but also to test the influence of each factor upon the
effectiveness of each of the other factors. The first experiment
was devoted to the study of the mortality as affected by the
availability of unpoisoned normal food and the kind and rate
of poison in the bait.
METHOD
The usual method of testing different kinds of poisons is
to change them while keeping all other conditions as nearly the
same as possible. The rates at which poisons are used are
usually tested in the same manner. Such a method has its
advantages but it will not satisfy the objectives of this investi-
gation because it does not provide for a study of the relation-
ship between poisons and rates. This relationship cannot be
studied unless every combination of poison and rate be tried
simultaneously. In other words, rates and poisons must both
be varied at the same time and in the same experiment. It was
found that if this were done in an orderly manner according to
a definite scheme, and not haphazardly, approximately twice as









VOL. XXI-No. 3


much information could be obtained per unit of effort. This
is due to the fact that the total data obtained can be utilized
in studying poisons and can again be utilized in studying rates.
This principle of experimentation has been described by Fisher1
as the factoriall design", evidently because several different
factors are tested simultaneously and according to a definite
design. The factorial design was used in each of the experi-
ments considered in this report. This type of experiment is
not new except for the fact that until within the last few years
research men have seldom recognized and taken full advantage
of it, evidently owing to the difficulty in interpreting the com-
plex results. Fisher has largely solved this difficulty by his
method of analysis of variance.
The effectiveness of each treatment was determined by cag-
ing five larvae, which were half grown or larger, singly in salve
tins each of which contained 1/20 gram of bait. After 18 to
20 hours' exposure to the bait each larva that remained alive
was transferred to a fresh salve tin containing only sweetpotato
leaves, as these were known to be a favored natural food. After
being confined with unpoisoned food for a period of four days
the larvae were discarded and records made of the mortality.
A group of five larvae constituted a sampling unit. All treat-
ments were tested simultaneously. In order to insure repre-
sentative results, the tests were replicated'on several different
dates with freshly-mixed baits and with new broods of larvae.
Usually it was necessary to employ several rearing jars to
supply sufficient larvae of the same size for a test. In pre-
liminary experiments this was found to be a frequent cause of
bias, evidently due to unpredictable differences in larval vigor
or resistance in the different jars. In order to insure an equal
distribution of these and similar variations and thereby to
eliminate bias, the following method was adopted for assigning
the individual larvae to the treatments: A group of larvae of
uniform size equal in number to the number of treatments tested
was selected from a single jar. These were assigned, in a hap-
hazard manner, one to each treatment. This process was
repeated until the required five larvae per treatment were
obtained.

'Fisher, R. A. 1935. The Design of Experiments. Oliver and Boyd.
London. Chapter 6.









THE FLORIDA ENTOMOLOGIST


As an aid in the interpretation of the resulting data, Fisher's
method of analysis of variance was used as described by Fisher
and Wishart.2.
Experiment I. Poisons and Rates. The first experiment
was devoted to a study of the mortality as affected by the
availability of unpoisoned normal food and the kind and rate
of poison in the bait.
The four materials chosen for this study were paris green,
synthetic cryolite, phenothiazine (thiodiphenylamine), and pow-
dered cube. Paris green and cryolite were known to be toxic
to this insect when applied to its natural food, but were objec-
tionable because of possible dangerous residues. For this reason
the other two materials were tested with the idea that a con-
centration or rate of one might be found that would be effective.
Three rates for each material were used, viz., 1, 21/2, and
7 pounds per 100 pounds of bait, incorporated in the following
standard bait mixture:
Bran .................... ..... .. ........... 50 pounds
Cottonseed meal ....................................... 50 pounds
Molasses ........................................ 1 gallon
Water ....................... ................ As needed to moisten
Bait alone was supplied for food in one-half of the treat-
ments. In the other half an abundance of sweetpotato leaves
was included with the bait. The mortality in the absence of
natural food was expected to show the toxicity of the materials
as compared with starvation checks, while the mortality in the
presence of natural food should show the degree of attraction
of the various baits. Thus there were the following three
factors in this study: (1) Poisons, with four alternatives (paris
green, synthetic cryolite, phenothiazine, and cube); (2) rates,
with three alternatives (1, 21/2, and 7 pounds per 100 pounds
of bait); and (3) food, with two alternatives (either present
or absent).
Six replications of the tests were conducted on different
dates, in which a total of 30 larvae were exposed to each of
the 28 treatments. The mortality data have been summarized
in Table 1. An analysis of variance of the data from all treat-
ments showed that the standard error of a treatment total was
2.19 and that a difference of 6.2 was required between treat-
ment totals to show significance at odds of 19 to 1.
'Fisher, R. A., and J. Wishart. The arrangement of field experiments
and the statistical reduction of the results. Imperial Bureau of Soil Science,
Technical Communication No. 10. 1930.









VOL. XXI-No. 3 43

TABLE 1.-MORTALITY OF SOUTHERN ARMYWORM LARVAE IN TESTS WITH
POISONED BAITS.

Material and Concentration Number of Dead Larvae
Bait Alone | Bait and Food

Control'- ......................--- -...----....-..-- ... 0 0
Control --.....~~........... ............................ 0 0
Paris green, 1 pound per 100 ....-..-.....-........ 23 8
Paris green, 21 pounds per 100 ........----........... 22 12
Paris green, 7 pounds per 100 .....--- ............ 28 21
Cryolite, 1 pound per 100 ............................ 26 12
Cryolite, 21/2 pounds per 100 .......................... 28 22
Cryolite, 7 pounds per 100 ........................... 29 29
Phenothiazine, 1 pound per 100 ................ 3 7
Phenothiazine, 21/2 pounds per 100 ................ 15 12
Phenothiazine, 7 pounds per 100 ................ 23 14
Cube, 1 pound per 100 ..............--- .......---- 0 0
Cube, 21/ pounds per 100 ....................-........ 1 0
Cube, 7 pounds per 100 ............................... 6 0
A difference of 6.2 is required between any two figures, irrespective of
position, to show significance at odds of 19 to 1.
'There were two control treatments without any bait at all and two with unpoisoned bait.

A more detailed analysis was later necessary for the precise
study of the interrelationships of the three treatment factors.
For this purpose the data from the six cube treatments and the
four control treatments were omitted in order to avoid the large
number of zero values which would tend to bias the estimates
of error and interaction variances. The complete analysis of
variance of the 18 remaining treatments is given in Table 2
because it summarizes so precisely the effects and the inter-
relationships between the various things studied. For example,
the significant variance for between tests not only shows that
the data vary from test to test more than can be explained
by chance but also tells how much more. This is worthwhile
information because it indicates that the treatments studied
were compared under widely different sets of conditions and
are therefore more representative and more widely applicable
than if they had been obtained under uniform conditions.
The highly significant variances for poisons, rates, and foods
prove that some of the alternatives of each were more effective
than others. The examination of the total mortalities indicates
that cryolite with 146 dead worms out of a possible 180 was
more effective than paris green with 114, and that both were
much more effective than phenothiazine with 74. Likewise the










44 THE FLORIDA ENTOMOLOGIST

7-pound rate with 144 dead larvae was more effective than the
21/2-pound rate with 111, and both were more effective than the
1-pound rate with 79. When unpoisoned host food was absent,
197 larvae died out of a total of 270, whereas in the presence
of such food only 137 died.

TABLE 2.--ANALYSIS OF VARIANCE FOR STUDY OF POISONS, RATES, AND
PRESENCE OR ABSENCE OF FOOD.


Source of Variation


SDegrees
[Freedom


Total (Between sampling units) .............. 107
(5 larvae)
1. Between treatments ................. 17
2. Between tests ..............-............. 5
3. Remainder (error) ........................... 85

1. Between treatments .......... ........- 17
(a) Between poisons ..................... 2
(b) Between rates ...........-.......-... 2
(c) Between foods .....--.-................. .1
(d) Remainder (interaction) ............ 12
(1) Poisons and rates ............-..-. 4
(2) Poisons and food ................ 2
(3) Rates and food ................... 2
(4) Poisons, rates, and food ..-... 4
**Indicates a highly significant variance when compared with
*Indicates a significant variance.


Sum of IVariance
Squares I

305.074

195.074 11.47**
17.296 3.46*
92.704 1.091

195.074 11.47**
72.296 36.15**
58.685 29.34**


33.333
30.760
4.760
8.001
1.167
16.832


33.33**
2.56**
1.19
4.00*
.58
4.21*


the error variance.


The lack of a significant interaction3 for rates times poisons
indicated that the increase in rates affected all poisons alike
and that the apparent differences in this respect were due to
chance. Likewise the interaction variance for rates times foods
was too small to be significant, indicating that the addition of
sweetpotato leaves attracted the larvae away from one rate of
poison as much as from any other rate. On the other hand,
there was a significant interaction variance for poisons times
foods, proving that the sweetpotato leaves attracted the larvae
away from some poisons more than from others. Or, to express
it another way, some of the poisons acted as if they might have
been repellent, while others did not. This can be studied in more
detail by examining the following mortalities:


Mortality
without food
Cryolite ................................... 83
Paris green ......-......................... 73
Phenothiazine ........................ 41


Mortality
with food
63
41
33


"The reversible interrelationship between two factors is usually ex-
pressed as the interaction of one times the other.









VOL. XXI-No. 3


Since a difference of 13 is required for significance, the data
show that although there was as a rule a decrease of mortality
in the experiment as the result of the presence of sweetpotato
leaves, this decrease did not prove to be significant when pheno-
thiazine was used. Likewise the data show that although paris
green was as a rule more effective than phenothiazine, this was
not demonstrated to be true in the presence of sweetpotato
leaves. This is interpreted as probably due to the low mortality
obtained with phenothiazine as a whole rather than to a lack
of repellency. This might not have been the case if stronger
concentrations had been used, as indicated in the results ob-
tained with the highest concentration of phenothiazine shown
in Table 1.
The variance for the triple interaction was significant, show-
ing that there was an interrelationship between the various
factors due to something apart from their own actions. This
is demonstrated in Table 1 by a general tendency for the pres-
ence of sweetpotato leaves to have less effect upon the mortality
of the lower rates of phenothiazine and the higher rates of
cryolite and paris green.
Experiment II. Constituents of Baits. While the first part
of this study showed that phenothiazine was inferior to both
cryolite and paris green, yet in the absence of natural food its
strongest concentration was quite toxic. It therefore seemed
desirable to study the effect of this material in connection with
various bait ingredients to determine if some combination would
not be sufficiently attractive to cause the insects to feed upon
the bait rather than the natural food. Both paris green and
phenothiazine were used at the low rate of 21/2 pounds per 100
pounds of bait to see if mortalities could not be obtained which
were as good as those secured with cryolite in the first part of
this study.
Each factor selected for Experiment II had two alterna-
tives, viz.:
(1) Poisons; either paris green or phenothiazine, at the rate
of 21/2 pounds per 100 pounds of bait.
(2) Bulk; either corn meal or cottonseed meal, at the rate
of 50 pounds per 100 pounds of bait.
(3) Syrup; present at the rate of 2 gallons per 100 pounds
of bait, or absent.
(4) Lemons; present at the rate of 4 pounds per 100 pounds
of bait, or absent.










THE FLORIDA ENTOMOLOGIST


All baits contained 50 pounds of bran as a basic material,
to which all possible combinations of the other ingredients were
added, resulting in 16 treatments. These were tested only in
the presence of abundant natural food, as the main object was
to determine those combinations capable of overcoming the
attraction of the natural food.
Eight time replicates of each test were conducted with
different broods and fresh baits, the results of which are
presented in Table 3. No treatment resulted in a really satis-
factory number of dead larvae, showing that no combination
was sufficiently attractive to overcome the attraction exerted
by the presence of natural food, since the greatest mortality
(18 larvae) was only 45 per cent of the 40 larvae used for each
treatment.
TABLE 3.-MORTALITY FROM COMBINATIONS OF INGREDIENTS.

Poison Bulk Ingredients Number of
SLarvae Dead
Paris green Corn meal Syrup Lemon 10
Phenothiazine 5
Paris green Cottonseed meal "13
Phenothiazine 5
Paris green Corn meal No syrup 13
Phenothiazine 2
Paris green Cottonseed meal ." 5
Phenothiazine 5
Paris green Corn meal Syrup No lemon 18
Phenothiazine 6
Paris green Cottonseed meal 14
Phenothiazine 7
Paris green Corn meal No syrup 7
Phenothiazine 3
Paris green Cottonseed meal 9
Phenothiazine 3

Difference required for significance, odds of 19 to 1 ..-..-......-........- 7

As this part of the study had a fourth factor, the analysis
of variance (Table 4) is more complex than that for Experi-
ment I. It contains three triple interactions and one quadruple
interaction.
It is quite probable, as indicated by the data, that paris
green was better than the phenothiazine. The summary total
for the mortality arranged by factors is presented in Table 5.
This shows that for the experiment as a whole there were no
significant differences between cottonseed meal or corn meal as
a bulk material for these baits, nor between the presence or










VOL. XXI-No. 3 47

TABLE 4.-ANALYSIS OF VARIANCE FOR STUDY OF BAIT CONSTITUENTS.

Source of Variation Degrees Sum of Variance
Freedom Squares
Total (between sampling units) .................... 127 122.93
I. Between treatments ...........-......-- .... 15 39.805 2.654**
II. Between tests .............................----...... 7 10.242 1.463
III. Remainder (error) ........................... 105 72.883 .694
I. Between treatments .............-......... 15 39.805 2.654**
a. Bulk ......... .... ....- ..... .................. 1 .071 .071
b. Poisons ... ........................ ......... 1 21.946 21.946**
c. Syrup .................. .... ....- .......... 1 7.508 7.508**
d. Lemons .. ......... ............ ......... ....- 1 .633 .633
Remainder (interaction) ..............- 11 9.647 .877
Interactions:
Bulk X poisons ........ ---..... - ...... 1 .945 .945
Bulk X syrup .-..........-............ .. ........ 1 .070 .070
Bulk X lemons ...-............................ 1 .008 .008
Poisons X lemons ................................. 1 .195 .195
Poisons X syrup ...... ..........- ....... ..... 1 .945 .945
Syrup X lemons .................. ...... ...-- ...| 1 1.758 1.758
Bulk X poisons X syrup ........ .......... 1 .383 .383
Bulk X poisons X lemon .......-....-..... 1 .195 .195
Bulk X syrup X lemon ...........-......--.. 1 1.320 1.320
Poisons X syrup X lemon ...........-....-- 1 .383 .383
Bulk X poisons X syrup X lemon ....... 1 3.445 3.445*

**Indicates a highly significant variance when compared with the error variance.
*Indicates a significant variance.

absence of lemon. However, the difference between the total
mortality for all combinations containing paris green as the
toxic agent and for those containing phenothiazine was highly
significant. The presence of syrup resulted in a definite increase
in mortality over the absence of syrup.
(To be continued)



Printing for All Purposes
Carefully Executed
Delivered on Time

Pepper Printing Company
Gainesville, Florida

We recommend the goods advertised in The Florida Ento-
mologist. Please mention Entomologist when you write our
advertisers.




University of Florida Home Page
© 2004 - 2010 University of Florida George A. Smathers Libraries.
All rights reserved.

Acceptable Use, Copyright, and Disclaimer Statement
Last updated October 10, 2010 - - mvs