Thursday, December 02, 2004

WWDD - I. Darwinian Method


If you happen to be one of the fretful minority who can do creative work,
never force an idea; you'll abort it if you do. Be patient and you'll give birth
to it when the time is ripe. Learn to wait. (Heinlein 1973)

What is Darwinian Method? Evolution is the central theme and unifying concept of all biology, as well as a useful mode of thinking in many other areas of human endeavor including economics, linguistics, pharmacology, engineering, medicine, and psychology. Charles Darwin's discovery of the Principle of Natural Selection (PNS) as the main mechanism of evolutionary change was, arguably, the most important discovery in the history of biology, and one of the most important events in the history of human civilization. It is unlikely that such a profound insight came to Darwin completely serendipitously. In order to form such a profound and novel theory he needed to employ a powerful scientific method and research heuristic.

Charles Darwin was not the first person to have a "gut feeling" that organisms have changed. He, like his predecessors, was a good naturalist and astute observer, but unlike the others, Darwin was not satisfied with "armchair" theorizing about the possible mechanisms. He knew that if he was to avoid the cruel fate of ridicule experienced by the writers of previous evolutionary tracts, he needed to put forward a rigorously tested theory. First, he had to be able to show proof that species change (microevolution), as well as that this change can be large enough to turn one species into a new one (macroevolution), no matter which taxonomic criterion or species definition might be used. Second, he had to provide a mechanism for evolutionary change which can explain every fact of biology known at the time. These explanations of facts of biology can all be divided into two categories: explanations of diversity (why are there so many species, why are they clustered in groups, and why they live where they do?) and explanations of adaptations (why organisms look, develop, function and behave the way they do?). Darwin's theory had to be capable of providing satisfactory answers to both categories of questions. However, the two kinds of explanations might require different methodological approaches, as well as different ways of presentation to the public. Was "Darwinian method" equally powerful in addressing these two categories?

The use of the phrase "Darwinian Method" is far from uniform. It is freely used to indicate a wide range of scientific methods. Most of the analysts applied the phrase to just one of the several approaches Darwin actually used. According to some, Darwin's methodology was inductive, according to others deductive, some claimed it was analytic, others synthetic. However, it was all of this, and more. In the following section, I will try to describe four aspects of Darwin's scientific method, as well as the mechanics of his scientific work: Feed-Forward Method, Integrative Approach, Comparative Method, and Consilience.



Feed-Forward Method:


Beware of the "Black swan fallacy. Deductive logic is tautological; there is
no way to get a new truth out of it, and it manipulates false statements as
readily as true ones. If you fail to remember this, it can trip you - with
perfect logic. The designers of the earliest computers called this the "Gigo
Law" i.e., "Garbage in, garbage out". Inductive logic is MUCH more difficult -
but can produce new truths. (Heinlein 1973)

Earlier in the history of biology, and certainly in Darwin's days, it was common practice for naturalists to observe (in the field or the lab), describe, make accurate drawings, and publish the observations without drawing any conclusions from the data. The thought was that a large number of people have to make a large number of observations before any patterns might start to emerge, and it was considered "bad taste" to theorize before such a wealth of information is accumulated. This thought did not prevent many of them from drawing what they expected to see, according to their preconceived notions, instead of drawing accurately.

Other naturalists of the time did test hypotheses, either in the field or in the laboratory, but here, the hypotheses were very narrow, and the reported data often inaccurate or "selected" to fit the preferred hypothesis. On the other hand, many books have been written by biologists of the era, proposing grand theories and sweeping generalizations based on a few selected observations. Needless to say, these theories were rarely if ever tested either by experiment or by further careful observations.

Darwin started out his naturalist career firmly within the tradition of the collectors and observers, trying hard to prevent any theory from biasing his observations. However, just like most of his colleagues of the time, he did read most of the theoretical literature. Like most of his colleagues, he was not able to dissociate the theory from practice, but unlike them, he insisted on a high degree of rigor, and he trusted the facts more then he believed in any particular theory. If observations did not fit the theory then there was something wrong with the theory, not the other way round. Before the voyage on the "Beagle" and in the very early part of the voyage, most of his work was purely observational. He collected specimens and gave them names. Soon enough he realized that his observations did not fit the prevailing theories of the time, mainly the belief that the species were immutable. He rejected all those theories and started on the quest for explanations of all the phenomena he encountered.

At first, his observations led him to ask a number of "smaller" questions about some particular phenomena. Further observations and thinking led him to propose hypotheses which he could test by further observations, as well as by experiments, as much as he was able to perform them while aboard a ship. By the time he came back home, he realized that all of his "small" hypotheses have a bearing on the one "big" question he was eager to answer. He started his "Transmutation" notebooks, read profusely, and started what was to become an enormous experimental program which he kept going until he died. This way of working, now almost taken for granted, brought a new level of stringency into the biology of the nineteenth century. A theory had to be tested, and had to fit the data, and not the other way round. The way Darwin worked is, in a way, how a modern biologist thinks and works today.

How does a modern biologist think and work? A good scientist looks at the nature with great curiosity. He (the "he/she"-phrase throughout the text is cumbersome, so, with no wish to appear sexist, I will use "he" for the sheer statistical fact that there are unfortunately still more men in science than women) has a keen eye and a quick mind. He notices things nobody noticed before. Or, he notices phenomena everybody takes for granted, but, as he realizes, nobody tried to explain before. So, he asks the questions about it: what, how and why? He tries to ask these questions with an open mind, but they will inevitably going to be led by his framework of thinking about the world which is informed by his knowledge and the current knowledge of the field. He might wish to do more observations first, in order to further refine his questions. Then, he will construct a story which has an explanatory potential of the phenomenon within his framework of understanding. He will, then, frame this story in the form of a testable hypothesis. So, he will go through phases of observing, theorizing and experimenting. The two main kinds of distinctions of scientific research are contrast between experiment and observation, and contrast between experimental and descriptive work (Brandon, 1998).

The difference between experiment and observation is in the manipulation of nature. In experiments we actively manipulate nature to produce certain conditions. Observations are passive - the conditions are already set by nature. If the phenomenon of interest is a variable which is dependent on other independent variables, then control of some or all of these variables is manipulation (experiment). The difference between experimental and descriptive work is the difference between testing hypotheses vs. measuring values of parameters. These two contrasts are not parallel, they are orthogonal. For instance, measurement of a parameter value may require manipulation. So, there are potentially four research methods.

Non-manipulative descriptive work can be exemplified by cataloguing flora and fauna of North Carolina, or the Human Genome Project. With manipulative descriptive work one manipulates one or more independent variables and measures the response variable without a specific hypothesis to test (just to get an answer to a straightforward question, or, just to see "what happens" in order to be able to design a hypothesis which can be subsequently tested). Non-manipulative tests of hypotheses include most of comparative work, including paleontological findings, as well as natural experiments (Diamond 1986). Natural experiments are cheap, but one has to be lucky to have nature perform one. For instance, a succession theory of island biogeography is naturally tested whenever a volcano or some such disaster completely wipes out life on an island. The scientist just needs to travel there, observe, count and measure for a few years. Of course, if one's question is about the general mechanism of all evolutionary change then all of life is a result of a natural experiment and pure observation can be used for a test of hypothesis, just as Darwin did. Manipulative tests of hypotheses are what is usually thought of as "experimenting".

Of course, these two dichotomies are not really dichotomies but graded continua. We can change one, more or all variables and have more or less precise controls. The same study can be a parameter measurement or a test of hypothesis depending on the current scientific importance of the hypothesis - at the time when the hypothesis is contested and "hot" the experiment will be a test of that hypothesis; later, when the hypothesis is generally accepted, it is not tested itself, but it is used in the same kind of experimental protocol for measurement of parameters. For instance, a set of data might have been used by Darwin writing the "Origin" to demonstrate that natural selection can work. In later works, he starts with the confidence that natural selection is the mechanism of evolution, and may have used the same set of data as illustrations or examples of other, narrower, principles, or as descriptions of phenomena which deserve further study.

So, why experiment? First, why manipulate? Because nature does not reliably and repeatably produce the conditions we need to observe in order to answer the question we pose.

Second, why test hypothesis instead of sticking to descriptive work? We test hypotheses because this is the most powerful method used to build theories that systematize and advance our knowledge of the world. In theory, one can keep observing, and perhaps counting and measuring, until every single case of the phenomenon has been described. That might take a bit too long if the hypothesis is very general - looking for fundamental laws of nature (looking at every atom in the Universe, every star, every blade of grass on Earth...). Even if possible, the complete description is still not an explanation. Even if every single observation is consistent with a world-view, this is still not the confirmation of that world-view. Hypotheses need to be generated and tested. A scientific test of hypothesis is an experiment.

There are three kinds of experiments: natural experiments, field experiments, and laboratory experiments (Diamond 1986). I described natural experiments above. Field experiments are labor intensive, often in difficult conditions, their results are less reliable due to uncontrolled variables, and the generality of results can be established only by further field experiments. Laboratory experiments allow for a stricter control of more variables, better repeatability, and thus stronger claims for generality of results. In biology, however, there is a caveat, as it is simply too easy to create phenomena in the laboratory that have no relevance to what is going on in nature (physics has no such problem - on one hand all data are relevant as they show what is and what is not possible; on the other hand relevance of data to bigger theories or to the human arena is often a matter of personal taste). A good research program is one in which data from three kinds of experiments (natural, field and lab), even if not performed by the same person, continuously inform each other.

In summary, a Feed-Forward method of scientific inquiry involves a continuous movement from observation, to question about the observation, to formation of a hypothesis which has a potential to explain the observation, to a model or an experiment testing the hypothesis, back to the hypothesis which needs to be modified in light of the experimental data, new hypothesis leading to new experiments, new experimantal data leading back to further observation, etc.

How does this Feed-Forward cycle start? It can be initiated at any point of the cycle. For instance, a mathematician can play on his computer and come up with a new model, which a biologist finds potentially useful for asking a particular question, so he picks it up and starts testing in nature. Or, a hypothesis might be born after a biologist reads a book or a paper and gets an idea. A naturalist might be walking in the great outdoors, or just tinkering in the laboratory, and happens to observe a curious phenomenon which starts him thinking. These days, most often a new graduate student is given an experiment to do. He performs it and, while analyzing the data and reading the relevant literature, realizes that a hypothesis may need modifying, or notices oddities in the data which need explaining. The cycle thus starts and the student becomes a scientist.

Integrative Approach:

Integrative approach is often understood to equal a "multi-level" approach - study of a physiological or behavioral function on all levels of organization including molecules, organelles, cells, tissues, organs, organ systems and whole organisms. However, this is too narrow definition of the term. Integrative method undoubtedly includes study at all levels, including even higher level of groups of individual organisms (e.g., breeding pair, population, species, community etc.), but it also includes a "multi-angle" approach - a study of different aspects of biology related to the main question of the study. In behavioral biology, this is commonly called "Tinbergian" method, (after Tinbergen 1963). Niko Tinbergen suggested that study of behavior, apart from pure observation, needs to integrate study of the mechanisms (physiology underlying the behavior), ontogeny of the behavior, the function (adaptive utility in relation to the environment), and evolution (phylogenetic history and evolutionary mechanism). The first two are often termed proximate ('how' questions) and the latter two ultimate ('why' questions) causes of behavior. Dewsbury (1992) recently suggested that it might be more fruitful to avoid this dichotomy and suggested an alternative scheme involving examination of the genesis (genetics, development, evolution), control (hormones, nervous system), and consequences (function, significance) of behavior.

Of course, there is no particular reason why such an approach needs to be restricted just to the study of behavior. Any morphological, developmental or physiological aspect of any organism can be studied from all of these angles, as well as on all levels of organization.

Did Darwin utilize integrative method? His main goal was to discover and characterize the mechanism of evolution, a much higher-level question than a study of one particular trait in a particular organism. However, in order to test his theory (natural selection), as well as numerous sub-theories (sexual selection, biogeography, evolution of mind etc.), he had to study a number of specific traits in order to show that, yes, they too can be explained by his theories. In order to convince his audience, he had to study a number of morphological, physiological and behavioral aspects of many organisms, and to show how they could have evolved by natural or sexual selection. He had to show that the trait is variable in nature, that it is heritable, that it aids in survival or reproduction of the organism, and that its form and mechanism reveals the path evolution took in forming that trait. Although physiology of his time was in its early pioneering days, and the study of cellular and molecular processes was still many decades away, Darwin's writings are full of examples of "multi-level" and "multi-angle" studies of particular traits in a variety of organisms. He started dissecting even before the voyage, had a microscope on board of the "Beagle" and conducted a number of physiological and behavioral experiments at his home in Downe after the voyage.

Comparative Method:

Earth is a home to a very large number of strikingly different life forms. All these organisms show a good fit to their immediate environments. These two observations, one about diversity, the other about adaptiveness, are what the science of biology is supposed to explain.

The two statements seem on the surface to contradict each other: if all organisms are perfectly adapted to their environments then all organisms found in the same environment should be identical, and there would be only as many forms of life as there are types of physical environment. The contradiction evaporates if one accepts that: a) adaptation needs not be perfect, just optimal in an engineering sense, b) there is more than one way to be optimally adapted to a local environment, c) environment also includes other organisms which actively change the environment, and d) adaptations have a history.

Diversity can then be explained by the process of evolution of adaptations. To put it very simply and in broad brush-strokes, heritable variation in a population is selected by the local environment leading, over a number of generations, to improved fit between the organisms and the environment. Organisms which find themselves at different places on Earth will have different variations and different environments, so they will evolve different adaptations. During that process they will become different from each other. As the local environments continuously change, all populations change, too. As the populations get more and more different from each other, and get adapted to more and more different environments, they will be less and less likely to ever meet, and if they accidentally do, they might not be able to recognize each other as potential mates (in case of sexually reproducing individuals, Paterson 1993). The lineage has split into two or more new lineages, and the process will get repeated again and again, leading to the appearance (and disappearance) of millions of species of organisms on this planet.

Current vs. historical adaptation. The oversimplified account of speciation (above) suggests that all of evolution is process of adaptation. If that was right, then every phenotypic (morphological, physiological and behavioral) trait would be an adaptation to the current conditions. Analysis of an organism could tell us everything about the environmental challenges imposed on that organism at the present time, or, at most, during very recent past period. However, traits without current utility, those that were perhaps useful in the past, would be undetectable. "Optimization ... erases history on the evolutionary time scale" (Wimsatt 1999). Without the presence of such useless traits - remnants of the past selective regimes - we would never be able to recognize that organisms have evolutionary histories. Fortunately for Darwin, and everybody since, history leaves traces in the body. Using such traces, we can analyze genealogical relationships between organisms, closeness of their relatedness, and the selective regimes of their common ancestors.

Why are non-adaptive traits retained for such long periods of time? If evolution is defined as change over time of relative frequencies of genes in a population, then a selective regime would eliminate "useless" genes or alleles and promote "useful" genes or alleles. The traits associated with the "useless" genes/alleles would disappear over time. But we know they do not. Perhaps this definition, although applicable to computer models of evolution, does not accurately describe evolutionary process in the organic realm. Natural selection alone cannot explain the persistence of type (Sterelny and Griffiths 1999, pp.224-234, 287-296). Instead, one might wish to consider Van Valen's definition: "evolution is control of development by ecology" (Van Valen 1973). The traits that were last useful hundreds of millions of years ago will persist if the process of their ontogenetic development is tightly coupled to the development of other, more recently useful traits. Alternatively, the disruption of the development of that particular trait would lead to the disruption of the integration of the whole organism. This phenomenon was termed "developmental canalization" by Waddington (1942), and somewhat explained by "generative entrenchment" of Wimsatt (1986, 1999). The change of gene frequencies in populations, although not sufficient as a definition or description of the evolutionary process, is a useful heuristic generalization. Gene sequences and patterns of gene expressions are genetic phenotypes which are the easiest to determine using current laboratory techniques. The direct measurement of gene (or expressed gene) frequencies is a technologically convenient "handle", or representation, of the evolutionary changes in the more complex system which is the organism. It serves as the entry point to the study of evolutionary change, as well as to the study of biological functions. One needs to keep in mind, however, that genes are just one of the aspects of the system, and that other (epigenetic) factors need to be taken into account as the genetic factors are elucidated. Also, an organism is a dynamic system which is constantly changing throughout its life cycle. The gene sequence is just a "snapshot" of one moment, and as such, does not contain sufficient information for the explanation of a higher-level trait or its evolutionary change.

The important point to be made here is that each organism is a mosaic of traits some of which hold current adaptive functions while others are adaptively neutral remnants of the past selective pressures. If two organisms are compared, their similarities can be attributed either to close genealogical relatedness or to similar ecological niches, while their differences are either due to phylogenetic distance or distinct current selective pressures. How is one to know which of the two - history or present function - is the cause for the presence of the trait in the organism?

Methodology of comparative research. In caricature, this is how the process often goes: One usually constructs phylogenetic trees by assuming that the similarities and differences in particular traits are correlated with genealogical distance and ignores the effects of adaptations. Then one takes the tree, maps the traits of interest onto the tree, and attributes the presence of "oddities" to adaptation. The choice of traits is crucial. The traits used for the construction of the tree should be the neutral traits, while the adaptive traits are used for the inference of selective pressures. But what is a neutral trait now, might not have been neutral in a more distant past. Even the most neutral aspect of phenotype - the nucleotide sequence in junk DNA - is not a universally reliable trait for the formation of phylogenetic trees, as that is the only phenotypic trait which is reversible (e.g., if T replaces G, then later G might replace T at the same position).

Let us now assume that a reasonably accurate phylogenetic tree has been built by combining DNA sequence, paleontological and comparative data. It has been drawn and published. How are the biologists going to study the biological functions of an organism? Are they going to use the tree at all? Depending on their concept of "function", or "adaptation", some will and some will not.

Pittendrigh (1958) suggested that the study of biological functions be called "teleonomy" It would ask the questions about functions of biological phenomena. First, it would assume that "some feature of the organism - morphological, physiological, or behavioral...serves some proximate end ....that the observer believes he can discern fully by direct observation and without reference to the history of the organism" (Pittendrigh 1958). However, this may be too optimistic. Errors can be (and have been) made when the functions are studied without the reference to the evolutionary history. Thus, evolutionary, or at least comparative data aid the understanding of the functional meaning of biological phenomena. Comparing a function in two related organisms aids understanding of that function in each of the two species (Pittendrigh 1958, Williams 1966). In a study of closely related organisms, similarities can be ascribed to phylogenetic inertia, and the differences to the recent selective environments (Brooks and McLellan 1991, Keeley 1999).

Did Darwin use comparative method? The answer is an obvious yes, as this is the most apparent aspect of his work. Every book of his (except perhaps the 'Earthworm' book, Darwin 1881) is a comparison of some aspect of biology of a long list of organisms. In the 'Origin' (Darwin 1859), the list includes everything alive, in the 'Domestication' (Darwin 1868) cultivated plants and animals, in 'Expressions' (Darwin 1872) and 'Descent' (Darwin 1871) many animals, in 'Movement of Plants' (Darwin 1880) many plants, in 'Orchids' (Darwin 1862) and 'Barnacles' (Darwin 1851) practically all members of the particular group. He used a large number of examples in order to find generalities, and to dissect the traits of shared history from the traits of current utility.


Consilience

Expertise in one field does not carry over into other fields. But experts
often think so. The narrower their field of knowledge the more likely they are
to think so. (Heinlein 1973)

Charles Darwin came up with a brand new general theory. It was so new that nobody before had any reason to perform any kinds of tests of that theory, and it was so big that it was impossible for a single person to quickly perform all the relevant observations and experiments. What Darwin faced was a great poverty of data. How was he going to persuade everybody about the correctness of his theory without a possibility of performance of a "critical test"? He did the second best thing. He showered his reading audience with thousands of reinterpretations of well-known observations and experiments, and effectively demonstrated that every little bit of knowledge about life on Earth he could think of is consistent with his theory and not with any competing theory. Without a short, elegant and definitive proof - he used brute force! He presented the theory as a hypothesis that still needs to be tested, but showed that it is worth testing as it offers the best explanation for all currently known phenomena in geology, biogeography, morphology, embryology, plant and animal breeding, physiology and behavior. If all known data, from many areas of inquiry, no matter how weak if looked at in isolation, all point toward the same conclusion, than the conclusion is very likely to be correct. That is consilience.

Darwin's heuristic

The difference between science and the fuzzy subjects is that science
requires reasoning, while those other subjects merely require scholarship.
(Heinlein 1973).


It is difficult to draw a clear line between a scientific "method" and the working "methodology" or "heuristic". In a way, the method one employs is also part of one's heuristic. However, some aspects of a scientist's work can be looked at as a personal "style" of work and need not be characterized as a scientific method.
For Darwin, it was of paramount importance to collect as many pieces of evidence as possible, as well as any facts which might challenge his theories. Thus, the most important aspect of his methodology is the way he searched for information. Without telephone, fax, photocopying machines, e-mail, and online databases, his search was, from present perspective, quite slow and tedious. He read papers in scientific journals, read books, read agricultural pamphlets, observed first-hand many phenomena during his voyage and later around his home in Downe, at the Kew Gardens and the London Zoo, and of course, performed a large number of experiments.
He is most famous for his personal queries, as he wrote letters to many people who were regarded as experts in their fields, most of whom he did not know personally, and asked them numerous questions and even suggested measurements and experiments (which some of them did for him). He asked very precise questions which should yield the most relevant information for the support of his theory, while never revealing to his correspondents the real reasons for asking them. At the same time, this rich correspondence prevented him from isolating himself at Downe. Over decades, the letters revealed to Darwin how people in a number of different fields thought about biology, and how they might react to his theory once he decided to publish it.
This combination of scientific methods and methodologies obviously worked for Darwin. At the outset, he was aware of all the competing hypotheses, but was agnostic about his own preferences. He realized that observations can be used as tests of those hypotheses. During the voyage of the Beagle, his observations led him to start preferring one hypothesis over the others, and subsequently he spent twenty years accumulating evidence and searching for an explanatory mechanism which would transform his speculative theory into a testable theory. He succeeded, so he must have done something right. However, what he discovered was a mechanism so general that it is applicable even outside the realm of biology. In the rest of this paper I will try to test if Darwin's method and heuristic is equally useful in the study of narrower questions in biology in his time, as well as today.




Style of presentation

The truth of a proposition has nothing to do with its credibility. And vice
versa. (Heinlein 1973)

Finally, something needs to be said about the way Darwin constructed his arguments and presented them to the public. Each of his books is one long argument for a particular theory. In each of the books, Darwin starts out with an outline of his thesis presented as a logical argument: If A, and if B, then C, where A and B are presented as very obvious, common-sensical facts and illustrated by several very familiar examples. The rest of the book is a long list of well-known facts, each shown to be explainable better by his theory than by any competing theory. He always had two opposite hypotheses running throughout a book. He presented his examples first as obvious proofs of the alternative theory, then showed how his theory fits the data even better. Also, in each book, he assumed that his previous books have already convinced the audience of his previous theories. For example, in "Variation under Domestication" he assumes that the readers are convinced of natural selection, in "Descent of Man" that pangenesis is an accepted truth, and in "Expressions" that sexual selection is a mainstream idea. Such confidence in his own ideas, use of large numbers of familiar examples, and presentation of theory in opposition to an alternative, made many a convert in his day, and is still as powerful today.

Go to Part 2:
http://sciencepolitics.blogspot.com/2004/12/wwdd-ii-darwin-on-time.html

posted by coturnix @ 1:20 AM | permalink | (0 comments) | Post a Comment | permalink