International Society for History, Philosophy, and Social Studies of Biology

twitter 2015
     facebook 2015


TUESDAY, JULY 7  /  09:00 - 10:30  /  DS-M240
Organized session / standard talks
Heterogeneity and the problem of ecological prediction

Eric Desjardins (University of Western Ontario, Canada)

“The fundamental aim of many empirical studies is to predict” (Spirtes et al 2000). A common view in philosophy of science is that predictions are the ultimate measure of a theory’s success or, at the very least, that they are inextricably linked to successful explanations in a particular field (Salmon 1989). In the field of ecology, however, this relationship does not seem to hold. Even though ecologists in many subfields provide explanations which greatly increase our understanding of particular phenomena, predicting future instances of these phenomena remains largely elusive. This is particularly problematic in some subfields of ecology that deal with practical issues such as bioinvasion, conservation, restoration, epidemics, etc. The purpose of this symposium is to explore the issue of prediction in ecology from an interdisciplinary perspective, in order to uncover some of the underlying theoretical and methodological difficulties in making precise predictions. The first paper, by Antoine Dussault, will revisit the Clements/Gleason controversy about the formation of ecological communities and thus provide an historical viewpoint on early debates that shaped subsequent views about the law-like vs. random nature of ecological succession. In the second paper, Katie Marshall will approach the issue from the standpoint of an insect ecologist, and give an account of the practical constraints faced by ecologists when making predictions about an invader’s response to environmental pressures, in the light of climate change. The third paper will provide a philosophical diagnosis of the problem of prediction in ecology. Alkistis Elliott-Graves will argue that some ecological phenomena are characterized by extreme causal heterogeneity, resulting in limits on the generalizability of results and a tradeoff between generality and predictive accuracy.

Rethinking the Clements-Gleason debate

Antoine C. Dussault (Université de Montréal, Canada)

Discussions of F. Clements and H. Gleason in ecology are often presented in terms of polarities. Clements’ Climax theory of species succession for example is commonly conceived as law-like and supporting of the idea that ecologists should be able to consistently predict the evolution of ecological communities after perturbation. On the contrary, Gleason’s theory of species assembly is commonly presented as random and therefore not supporting of the idea that such predictions should be possible. However, recent historical and philosophical investigations have suggested that this polarization between these two authors is unrepresentative. On the one hand, Eliot (2011; 2007) and Hagen (1992; 1988) have demonstrated that Clements does not conceive of ecological succession as exceptionless. On the other hand, Nicolson and McIntosh (2002), McIntosh (1998; 1995; 1975) and Nicolson (1990), have shown that Gleason does not endorse a purely random picture of the ecological world. However, these conclusions are so far apart from the received view that they leave the reader wondering whether there was actually something at stake between Clements, Gleason and their followers, or why the triumph of the Gleasonian view in the 1950s was perceived by many ecologists as a radical shift in the discipline (Barbour 1996). The goal of my presentation is to propose a more subtle and accurate picture of the Clements/Gleason controversy, which explains the perception of a disciplinary shift, while not falling into the simplified readings criticized by Eliot and others. My suggestion, building on some observations by Allen, Mitman and Hoekstra (1993), will be that Clements’s commitment to neo-lamarckianism allowed him to reconcile mechanistic explanations for ecological succession and the origin of repeatable species association; and that Gleason and later Gleasonians were led to reject such mode of reconciliation by their commitment to neo-Darwinism and its genetic view of inheritance.

Why it’s so hard to model the response of the ‘average’ insect to climate change

Katie Marshall (University of British Columbia, Canada); Brent Sinclair (University of Western Ontario, Canada)

Understanding the effects of climate change on organisms is a critical part of preparing for the future. Drawing on lab and field-acquired data, biologists produce models of the both the present and predicted future distribution and abundance of organisms. The models can either be “top down”, correlating current distributions to abiotic conditions and the extrapolating to future conditions, or “bottom up”, attempting to model current distributions based on lab-measured physiological tolerance. Modelling animal distributions is generally based on three key assumptions: 1) that the limits of organismal tolerance are fixed, 2) that means, extremes, and variability of environmental factors have no additive physiological effects and therefore can be treated equally in models, and that 3) the environment that organisms experience is what is being measured by dataloggers. While some biological systems do not violate these assumptions, many others do. Using the economically and ecologically important example of insect overwintering biology, we demonstrate how all three of these assumptions are violated in this system. Overwintering insects have extremely plastic cold tolerance on both short and long time scales, variability of low temperature stress has additional effects on survival, energy use, and reproductive output, and snow cover can have a large effect on thermal environment leading to downstream potential fitness effects. Violating these assumptions leads to very differing model outcomes at both the individual and population level. We will discuss the relative merits of each of these approaches, framed as ways of understanding the emergent property of population distribution, and outline what empirical work and assumptions would be necessary to increase the accuracy of these models.

Causal heterogeneity constrains predictive power in ecology

Alkistis Elliott-Graves (University of Western Ontario, Canada)

Ecologists study systems which are complex and are characterized by heterogeneous phenomena. In many cases, such as ecological succession and bioinvasions, part of the heterogeneity is causal as each particular instance of a phenomenon can be caused by a unique combination of factors. I argue that this has some important implications. First, it limits the generalizability of explanations and results. For example, the invasion of the north american lakes by the aquatic cattail Typha angustifolia was caused by the displacement of the native Bolboschoenus fluviatilis, fueled by allelopathy (the exudation of toxins from the roots of a plant). Even though this is generally considered to be a sufficient explanation of this particular invasion, the results cannot be generalized to other invasions, given that most other plants and animals are not allelopathic and toxins diffuse differently in other environments. Second, generality comes at the expense of causal detail. That is, a description or explanation of a phenomenon becomes general if we abstract away causal particulars. In the case of the aquatic cattail, the invasion could be explained in a more general way by abstracting allelopathy and conceptualizing it as an instance of interspecific competition. However, a significant subset of ecological predictions must be precise and localized in order to be useful. For instance, it is not sufficient to predict that allelopathic species tend to become invasive; the key is to determine when and where the next invasion is likely to occur. This leads to the third implication, namely that in these heterogeneous systems, the very causal factors which must be abstracted in order to generalize from a single case are the same factors which are necessary for making precise predictions. Thus, the key to successful predictions is to conceptualize each phenomenon with the right level of generality.