BIO 374 -- Evolutionary Ecology
P.O. Box 805
Grinnell, IA 50112
Do not circulate without permission!
Note: these are the first two labs in an advanced (jr-sr.) course in evolutionary ecology. A third lab on functional morphology is completely student directed (i.e., they are posed with the problem of how describe the relationship between morphological variation and performance in a population of grasshoppers). The remainder of the labs are devoted to independent group projects.
Bio 374 -- Evolutionary Ecology
Fall 1998 -- J. Brown
Interactions between plants and their pollinators have been a focus of the study of how species interactions shape evolutionary change since Darwin's landmark studies of orchids and bees. Most studies have emphasized how pollinators influence plant characteristics, rather than the reverse, because phenotypic characters of plants can be easily measured and individual success during and after pollination can be easily followed. This laboratory exercise demonstrates the analytical techniques used in such studies, and will allow you to generate and test specific hypotheses about the roles of various plant traits in fitness.
As in all scientific investigations, the first step is to observe and describe the system before hypotheses may be generated and tested. I have chosen to concentrate on the interaction between a prairie plant, Solidago rigida, and its pollinator species. During the first week, we will characterize phenotypic variation in a single population of this species found in the restored prairie of the Krumm Nature Preserve. You will be responsible for analyzing these data before the second week, when we will try to determine what plant features are associated with greater attraction to pollinators. You will work in pairs and may work together on all aspects of the lab except the prose of your paper.
Since the population of S. rigida at Krumm consists of more individuals than we can measure, we will sample the population using random points along line transects. Each pair of students will sample individual plants along a transect laid through the population. Using the random number tables provided, find the distance along the transect to go before making your next measurement. At that spot, measure the individual closest to the transect point. Here are the phenotypes you should measure for each individual:
Express all (except 3 and 6) in centimeters.
There are spaces on the data sheets for any additional characters you wish to define. After you make your measurements on each individual, mark it with a flag and with an ID-number written on the flag with a Sharpie. Make measurements on as many individuals as possible -- the more you sample, the likelier you are to find significant results.
Sometime before the next lab meeting, you need to enter and analyze your data. The first step will be to familiarize yourself with the statistical program Minitab. Versions of Minitab are available for both Mac (v 10.5) and Windows (v 12). Either will be fine for what were are doing in this class and the directions for use are either identical or very similar. Please see me if you are having computer problems.
Entering your data: A new Minitab data worksheet should be open for you on the lower half of the screen. Enter the variable names at the top and then enter your data into the spreadsheet. If, for some reason, you are missing data for an individual, leave that cell blank. After entering your data, save the file!
Analyzing your data: At this point in the study we are interesting in describing population variation in two ways: (1) the extent of variation in individual phenotypic characters (univariate statistics), and (2) the correlation between various phenotypic characters (multivariate statistics).
1. Choose "Basic Statistics", then "Descriptive Statistics" from the Stat menu. Highlight, then Select all the variables, which should then appear in the variable box. Click OK to see output in the Session screen. These are the univariate statistics for your sample.
2. Choose "Basic Statistics", then "Correlation" from the Stat menu. Again, highlight and select all variables and click OK. You should see a matrix of correlation coefficients between all variables appear in the Session window.
3. While the Session menu is in front, highlight any text youd like to print and then choose "Print Selection" from the File menu (you can also save the output electronically, if you wish, by choosing "Save Window as . . ." from the File menu).
Study your statistical output. You should be able to get a sense of which plant phenotypes are particularly variable, and which phenotypes are strongly correlated with each other. Use the table of significant values for r to determine whether these traits are significantly correlated.
Plotting your data: It is often useful to represent your results in graphs. Find the two pairs of phenotypic measures with the highest positive or negative correlations and use Minitab to make scatter graphs that illustrate these correlations (Choose "Plot" from the Graph Menu. Bring in your graphs and be prepared to discuss your results informally by the next lab meeting.
Pollination Lab -- Week 2
Before coming to lab you should be prepared to discuss the results of your analysis of last weeks data on variation in characters of Solidago rigida. Study your results with the following questions in mind:
1. What plant characteristics are particularly variable?
2. What plant characteristics are highly correlated?
3. What plant characteristics do you hypothesize to be related functionally to success at attracting a pollinator?
This week we will search for correlations between plant traits and pollinator visitation. One approach to this is simply to look for associations between plant traits and rates of visitation. In this case, we are testing a causal link between plant variation and a pollinator response to this variation, so we use the statistical technique of regression (see Stats for Evolutionary Ecologists II). Our approach to this analysis is made more complex, however, by our recognition (as good ecologists) that plant traits are correlated with one another. Thus, we might find a significant regression between a single plant characteristic and fitness that was actually CAUSED by another highly correlated plant character. There are two ways that we might approach this problem:
1. If the correlated characters might be two different ways of measuring whatever the pollinator is reacting to, we might try regressing the pollination response with a composite of all these plant characteristics. Principle components analysis (PCA) can be used to generate a composite variable that explains much of the variation observed in multiple characters (see Stats for Evolutionary Ecologists II).
2. If we have no reason to believe that characters interact functionally to create a single "composite" character, we can use the technique of multiple regression (see Stats for Evolutionary Ecologists II) to estimate the contribution of each character individually to the pollinator response.
Think about which sets of plant characteristics might be appropriate to analyze in each of these fashions.
Finally, it is almost always preferable to manipulate phenotypic variation, rather than to rely on the variation present in the population. One reason for this is that if, for example, selection is favoring intermediate values of a character and character variation in normally distributed, there will be very few individuals in the tails of the phenotypic distribution to measure. Thus our power to discover how natural selection is acting can be increased by increasing the variance in the natural distribution of characters. Since we are working in a protected population, our options for manipulating plant characters is limited (e.g., we dont want to cut off flowers), but there are ways to increase variance in characters temporarily, without harming the population.
1. Return to your transect with a copy of your data sheet from last week, a digital stopwatch and a clipboard for each member of your group.
2. Go to all measured plants in your transect and quickly count again the number of open heads and measure again the inflorescence size, since these measures may have changed from last week..
3. Increase the variation in head number and inflorescence size by randomly choosing half of your plants and creating "super-inflorescences" by gently twist-tying each together with the nearest neighbor inflorescence.
4. Splitting up, watch each inflorescence from a short distance for 5 minutes. (If you remain as still as possible, you should not influence the pollinators behavior). Note the arrival and departure times of each pollinator on your sheet (use the lap function). If there is a pollinator there when you arrive, count it as having arrived at the beginning of your measuring period. Make notes of what sorts of pollinators/visitors you observe.
5. Circulate between plants for 5 minute observation periods as many times as time permits.
1. For each plant, add up the total number of visits, the total time spent on the plant over all observation periods, and the average time per visit. [NOTE: if you didnt watch each plant for the same total number of minutes, divide total number of visits and total time spent by the number of minutes watched for that plant.]
2. Add three new columns to your Minitab data sheet with the following titles: NVISIT, TOTTM, and TMPRVS and enter the data for each target plant. Make the appropriate changes in the other variables and save the data worksheet under a new name.
3. Regress each variable individually against each pollinator response [Choose Regression from the Stats menu, and select the response and predictor (plant) variable.]
The output should look something like this for each pair of variables:
The regression equation is
NVisits = 4.55 + 0.0040 Nheads
Predictor Coef Stdev t-ratio p
Constant 4.552 2.652 1.72 0.130
Nheads 0.00401 0.04297 0.09 0.928
s = 3.283 R-sq = 0.1% R-sq(adj) = 0.0%
Analysis of Variance
SOURCE DF SS MS F p
Regression 1 0.09 0.09 0.01 0.928
Error 7 75.46 10.78
Total 8 75.56
As you can see, the output tells you the regression equation and then provides an analysis of the estimates of the intercept (Constant) and regression coefficient or slope (here, Nheads, after the X variable). The p value indicates whether the estimate is significantly different from zero -- usually, values of p < 0.05 are considered significant (i.e., there is a less than 1 out of 20 chance that this difference could occur by chance alone). In this case, we are only really interested in whether the slope of the line is different from zero (the intercept doesnt matter). This is confirmed in the analysis of variance section.
4. Regress each of the pollinator responses against the following predictor variables in a multiple regression: Height of inflorescence, diameter of inflorescence and number of neighbors. Notice that you get p values for each of the predictor variables, indicating the significance of the partial regression coefficient for that variable. You may choose other combinations of character variables if you wish.
5. Choose a set of variables that you believe are different ways of measuring the same "composite" character of the plant and perform a principle components analysis:
Your output should look something like this:
Principal Component Analysis
Eigenanalysis of the Correlation Matrix
Eigenvalue 1.8869 0.9431 0.1700
Proportion 0.629 0.314 0.057
Cumulative 0.629 0.943 1.000
Variable PC1 PC2 PC3
DNIEGH -0.256 0.964 0.078
DIAINFL 0.677 0.236 -0.697
NHEADS 0.690 0.125 0.713
The eigenvalues are the variance along each principle component axis. The first section above indicates the actual and cumulative proportion of the variation explained by the three principle components; so in the above example, the PC1 explains 62.9% of the variation, PC2 31.4% etc. The second section gives the "loadings" of each variable on the PC1. The values of the coefficients gives you an idea of how important each character is in generating the PC score. In the above example, PC1 = -0.256[DNIEGH] + 0.677[DIAINFL] + 0.69[NHEADS]. [Note: You cant just plug in the raw scores for each of these variables to find the PC score. Values for each variable must first be "standardized," i.e., each value is subtracted from the mean and divided by the standard deviation. Thank goodness for computers . . . ]
You should have a column of PC scores for your analysis. Do a regression of the PC1 scores against each of the response variables.
6. Prepare your results in the form of a scientific paper (see handouts or Web page). You and your partner may submit a common Methods and Results section by Friday 9/11 at 5pm. I'll return by Monday morning. EACH of you should then write your own paper (adding Abstract, Introduction and Discussion sections), due at 5pm on 9/18. In discussing your results, you might consider the following questions:
Are our measures of pollinator visitation a good surrogate measure or predictor of fitness?
What other predictors of fitness might we consider (if we had more time)?
What do the alternative ways of analyzing the data tell you about selection on these plants?
Think about the biological meaning of all these analyses! Please come to talk to me if you wish to discuss your interpretations of this project.
Sexual selection in natural populations
Darwin described sexual selection as occurring due to "the advantages that certain individuals have over others of the same sex and species, in exclusive relation to reproduction." Studying sexual selection in natural populations entails the demonstration of phenotypic variation in heritable characters that is significantly associated with reproductive success either through competition with others of the same sex ("intrasexual" selection) or attraction to the opposite sex ("intersexual" selection). The study of sexual selection is important not only to demonstrate its ability to create intersexual differences; it has also been implicated in the processes that produce reproductive isolation during speciation.
One can easily measure the potential intensity of sexual selection in natural populations of organisms that have prolonged copulatory periods by comparing characteristics of individuals found mating with characteristics of solitary individuals. Local species of insects that exhibit these features include members of the coenagrionid damselfly genus Enallagma , commonly known as "bluets," and the acantharid beetle genus Chauliognathus, commonly known as the goldenrod soldier beetles. We will split into two teams to capture an adequate sample of solitary males and pairs of these species. We will then return to the lab to measure phenotypic differences between successful and unsuccessful males. In our analyses, we will again have to pay attention to potential correlations between characters in our interpretation of the mechanism of sexual selection.
Let us assume, for the moment only, that differences we may measure between successful and unsuccessful males are due to female choice, rather than male-male competition. One set of theories of sexual selection contends that females should make choice based on variation in certain traits because these traits are good indicators of the genetic quality of a male. For example, Hamilton and Zuk (1982) proposed that females choose male characters that are good indicators of low parasite load, since resistance to parasitism is a important and heritable determinant of fitness in many organisms. Several recent studies of vertebrate and invertebrate species have indicated that developmental stability may also be such a target of sexual selection. The most common measure of stability is fluctuating assymetry (FA), defined as "the random deviation from bilateral symmetry in a morphological trait for which differences between the right and left sides have a mean of zero and are normally distributed" (Watson and Thornhill, 1994). Under this model of sexual selection, males that are mated should be more symmetrical than those that do not obtain mates.
We will use a computer image analysis system to gather morphometric data on the individuals we collect. We will compare both absolute characters of males and their symmetry to explore how sexual selection may be acting in these populations. Finally, we will compare the morphology of mated pairs, to see whether there is any evidence of assortative mating ("like mating with like"), an important component of many models of speciation.
Hamilton, W.D. and Zuk, M. 1982. Heritable true fitness and bright birds: a role for parasites? Science 218: 384-7.
Watson, P.J. and Thornhill, R. 1994. Fluctuating assymetry and sexual selection. Trends in Ecology and Evolution 9:21-25.
Measuring sexual selection
NIH Image is an image analysis program that we will use to gather morphometric data on our insect samples. It is a free program (your tax dollars at work!), so you may do the measuring part of the exercise on any Macintosh computer (assuming it has a screen as big as the one the images are captured on).
I. Dissecting your organism
II. Scanning images into the computer using NIH Image
III. Measuring individual variation using NIH Image
IV. Analyzing the data
Looking for correlations:
Looking for sexual selection --
Looking for assortative mating --
A paper describing your results is due on September 30 by 5 PM. No rough drafts this time, but I will be happy to discuss your results with you in advance of the deadline.