1. Testing Alternative Models of Social Evolution Vickie L. Backus, Middlebury College Joan M. Herbers, Ohio State University

2. Disney View of Social Insects Single queen Workers who are all sisters (r=0.75) Every individual working for the “good” of the nest Workers sacrifice own reproduction in order to help mother because this increases her own inclusive fitness.

3. Real world view of social insects Number of queens can vary from 0–100’s Workers may not be related to each other Queens and workers may disagree on what is “best” for each other and the nest.

4. Potential conflicts in ant nests How many resources to invest in each offspring (packaging decisions) Invest in growth (workers) or reproduction (male and female alates) Invest in males or female alates (sexual allocation)

5. Leptothorax longispinosus <5 mm in length Lives in pre–formed cavities in northern temperate forest Nests usually <40 workers Multiple queens in nests Seasonal polydomy

6. Location of Studies VT site-Mallett’s Bay State Park NY site-Edmund Niles Huyck Preserve

7. Average Female Alate Weight NY82NY83NY87NY88NY89NY90VT82VT83VT86VT87VT88VT89VT90 Data Set Mean female weight (ug)

8. How do nests decide how to package new ants? Herbers 1990 American Naturalist 136(2):178–208 Backus 1993 Oecologia 95:283-289 Both papers used path analysis but each paper took a slightly different approach to constructing the paths

9. Herbers 1990 Model A Path model where alate weight (G F and G M ) are determined directly by nest demographics (Q=Queen number; W=Worker number).

10. Herbers 1990 Model B Alate weight (G F and G M ) are determined after nests have decided how much energy to allocate to each sex. F= total biomass of females M=total biomass of males

11. Methods used by Herbers 1990 Decompose the paths using the “rules” presented in Li 1975 Standardize the variables (Q, W, G F, G M, F and M) with mean=0 and s 2 =1.0 Calculate multiple- regression coefficients for the standardized variables=path coefficients

12. Methods of Herbers 1990 Test internal consistency by comparing observed correlations of internal variables with that expected under the model

13. Testing internal consistency Expected correlation between GF and GM = qq’ + ww’ +qw’R QW + q’wr QW

14. Results of Herbers 1990 analysis Neither model showed consistent strong fit with the data but the fit was in the same direction for model A

15. Enhanced path model for packaging decisions (Backus 1993) Nest demographics still the heart of the model Addition of maternal effects (mean queen mass) as an independent variable Addition of worker packaging as a dependent variable

16. One other issue… Path analysis produces very large data sets E.G. 9 path coefficents for every data set analyzed Backus 1993 analyzed 6 data sets; Herbers 1990 another 7

17. Presentations in print can accommodate large data sets Large data table from Backus 1993 showing the path coefficients for the 6 data sets for each of 3 dependent variables.

18. Graphical presentation of data Both magnitude and direction of coefficients shown

19. Packaging for female alates Herbers 1990; Backus 1993 Data setEffect of Queen #Effect of Worker #Effect of Mean Queen Mass NY82+--NA NY83-+NA NY87+--NA NY88- - -+ + +-- NY89-+- NY90NA VT82- - - -NA VT83+- - NA VT86+-NA VT87++ NA VT88- -+ VT89NA VT90NA

20. What did we find out about packaging decisions? Able to detect some detailed trends in packaging. Example of packaging of males from Backus 1993

21. Two strong conclusions emerged: Significant queen worker conflict to packaging occurred. The proposed models don’t explain the whole story

22. Evidence of queen worker conflict in the path analysis Data for female alates as an example. Direction of coefficients different for queens and workers in 7/10 data sets

23. Explanatory power of the models was low. Data SetFemale mass Male massWorker mass NY88404.011.6 NY8919.430.515.1 NY90NA9.418.3 VT888.825.039.0 VT89NA5.12.2 VT90NA2.26.3

24. Why so little variation explained? Many unexplained variables Requires a deep understanding of the biology of the system Clearest use of the technique is when you have two equally realistic models and want to determine which is correct

25. Queen–worker conflict in Hymenoptera Haplio-diploid sex determination system is critical Q-females, sons or workers r=0.5 W-sisters (females or new workers) r=0.75 W-brothers r=0.25

26. Prediction for sexual allocation decisions: Queens and workers have different optima for sexual allocation Queens prefer equal investment in new female alates and males. Workers prefer biased investment with most biased being 1:3 males:females Trivers and Hare 1976

27. 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 NY81NY82NY83NY87VT82VT83VT84VT85VT86VT87 Data set Allocation to males (0=all female) Allocation to Males

28. Herbers 1990 Am. Nat. 136(2):178–208. Leptothorax longispinosus Queen number and worker number as independent variables Final dependent variable is proportional allocation to males.

31. Major conclusions from Herbers 1990 Evidence for significant queen-worker conflict Variation in the effect of nest demographics is important Nest demographics doesn’t explain everything.

32. What about growth? Reproduction isn’t the only allocation decision that nests need to make.

33. Backus (1995 American Naturalist 145(5):775–796) Added growth to the model. Tested alternative models for how nests may be making allocation decisions.

36. Comparing the two models Model A: proximate level factors important Model B: ultimate level (selection) important

37. Which model fits the data? Internal consistency tested. Compared the inferred correlation between GB and RB for both models with the observed correlations.

38. Best fit differed for the two sites. Data SetInferred by Proximate ObservedInferred by Ultimate NY88.6037.3768.2340 NY89.0466-.0595.0227 NY90.0491.0875.1068 VT88.0752.0731.0320 VT89.1035.1512.0938 VT90.1326.0209.1348

39. Further conclusions from the analysis of the two models: Natural selection is acting differently on the two populations Queen-worker conflict does occur over the decision to grow or reproduce. Nest demography is important

40. What has path analysis allowed us to conclude? Conflict occurs over all types of allocation decisions made ant nests Different populations have different levels of conflict Different populations may apply different rules to allocation decisions Nest demography is important but not the only factor to consider in allocation decisions.

41. Acknowledgments JMH’s work funded by National Academy of Sciences, E. N. Huyck Preserve, National Science Foundation VLB’s work funded by E. N. Huyck Preserve and Sigma Xi. Presentation of this talk funded in part by Middlebury College. Samuel Scheiner for inviting me to present this work.

42. And Especially…