The Future of Everything: The Science of Prediction

by David Orrell

  According to Galton, traits such as ability radiate down the family tree like a kind of Newtonian force, diffusing with distance. In the Law of Ancestral Heredity, parents contribute one-half of inherited traits, grandparents one-quarter, great-grandparents oneeighth, and so on, decreasing by a factor of two with each generation. The sum of all contributions from one’s predecessors is ½ + ¼ + + + . . . which adds in the limit to one.6 Of course, eminence isn’t the only thing that gets passed down. Galton also argued that the children of criminals were more likely to themselves show criminal tendencies: like the scorpion in Aesop’s fable, they have it in their nature.

  To prove his theory, Galton needed a quantitative way to measure people’s traits and characteristics. This exposed a problem with the metric: it is hard to measure what goes on in people’s minds. He therefore decided to concentrate on external, physical properties, and he set about measuring the dimensions first of English schoolboys and then of the population at large. During the International Health Exhibition in 1884, he set up an anthropometric laboratory and collected statistics such as height, weight, and strength on thousands of people (who seemed to enjoy being measured, and paid three pence each for the privilege).7

  WHATߣS NORMAL?

  Since Galton could not demonstrate using biology that traits were inherited, he instead tried to prove the point with statistics. He performed his analysis using two techniques, correlation and the bell curve, which are still the basis for modern statistical prediction in areas from demographics to the stock market.

  FIGURE 5.1. Histogram of hypothetical height data, based on Galton’s report of the adjusted mid-height of parents.8 The height of each column indicates the number of samples in that bracket. The distribution can be approximated with a bell curve of mean 68.3 and standard deviation 1.65.

  First proposed by Abraham de Moivre in 1718, the bell curve, also known as the normal distribution, is a method used to approximate the spread of randomly varying quantities. Laplace used it in 1783 to study measurement errors, and today IQ scores are based on it. Its shape is specified by two numbers—the mean, or average, and the standard deviation, which is a measure of its width. Sixtyeight percent of the results are within one standard deviation of the mean, and 95 percent within two standard deviations. Figure 5.1 shows a histogram of 928 measurements of height (the data, though hypothetical, is similar to that used by Galton). The height of each column indicates the number of samples in that bracket; most are clustered close to the mean, then the numbers tail off in a smooth way. Also shown is a bell curve that fits the data quite well.

  There’s something almost magical in the way that apparently random data turn out to correspond to a mathematical function. The name “normal distribution,” however, is somewhat misleading. Just as most numbers are not rational, so most statistical samples are not normal. The normal distribution owes its popularity to its mathematical properties: in the same way that a linear system is the sum of its parts, the sum of many bell curves, or other distributions, is a bell curve. If the heights of men and women each follow a bell curve, then a couple’s adjusted mid-height, which is the average of the parents’ heights after adjusting for the difference between men and women, will do the same. As we’ll see in the next chapter, many quantities that have been modelled using a normal distribution have turned out not to be so normal after all.

  Relationships between different sets of data, such as the heights of parents and children, were quantified by Galton using the concept of correlation. Suppose that one set of data Y is plotted against another set of data X. If the two have the same degree of noise or scatter, and a straight line is drawn which best interpolates the data, then the correlation is defined as the slope of the line (given by the rise over run). This is 0 if the line is horizontal, 1 if the angle is 45 degrees, and –1 if the angle is –45 degrees. A 1 or a –1 means that the data is perfectly correlated. Figure 5.2 (see page 180) shows a scatter plot of children’s heights against the adjusted mid-height of parents, based again on Galton’s data. Each point represents a particular child/parent data pair. The straight line, determined here by a technique known as linear regression, indicates a positive correlation of 0.63 between the height of parents and that of their offspring: tall parents tend to have tall children.9

  FIGURE 5.2. Scatter plot of children’s heights versus the adjusted mid-height of their parents. The slope of the line is 0.63, indicating a positive correlation.

  Galton used correlations to argue that traits were passed on from generation to generation, a reasonable assumption to make for height. But it is usually impossible to discern cause and effect from correlation alone, or to know whether both data sets have been influenced by a third factor. The data in figure 5.2 appear to show that tall parents cause tall children, and therefore that height is inherited. However, if the measurements came from poor and rich families—so some were very badly nourished and others well fed—then it could be that height is more a function of diet (nurture rather than nature).10 Experimenters usually try to screen out such effects (as Galton did, to an extent) by carefully controlling the data.

  Since the angle of the line in figure5.2 is less than 45 degrees (the correlation is less than 1), it means that while tall parents will have taller-than-usual offspring, those children will on average be shorter than the parents—a phenomenon that Galton called regression to the mean. Galton saw this as a kind of negative feedback mechanism that prevents traits from growing without limit. Without regression to the mean, members of successive generations of tall families would grow taller and taller, while short families would shrink to a dot. Regression to the mean implies that variance within a population is bounded, and it’s the theory behind a number of slogans still popular in places like Wall Street, such as “What goes up must come down.”

  If height could be inherited, then so in principle could complex traits like longevity, intelligence, or criminality. In Galton’s composite photographs of criminals, he used methods he had “frequently employed with maps and meteorological traces” to pick out correlations in features.11 A more successful crime-fighting technique was fingerprinting as a means of identification. He developed a classification scheme, which remains in use today, for the intricate whorls and vortices, themselves reminiscent of storm systems on weather maps.12 Human identity, it seemed, could be mapped and predicted just like the atmosphere.

  In the nineteenth century, the heyday of mechanistic science, all of this led to a horrible kind of deterministic logic: if human traits were in the blood, it followed that they could be improved by selective breeding. As Galton wrote, “No one, I think, can doubt . . . that, if talented men were mated with talented women, of the same mental and physical characters as themselves, generation after generation, we might produce a highly-bred human race, with no more tendency to revert to meaner ancestral types than is shown by our long-established breeds of race-horses and fox-hounds.”13 He introduced the word “eugenics” to describe “questions bearing on what is termed in Greek, eugenes, namely, good in stock, hereditarily endowed with noble qualities.”14

  The eugenic scheme was perhaps best expressed in Galton’s unpublished novel, Kantsaywhere, completed shortly before his death. It described a futuristic society whose inhabitants were required to pass a series of exams for health and intelligence. Those who did well were rewarded and encouraged to procreate, while those who were mediocre could have only a limited number of children. The punishment for failure was enforced celibacy. Kepler’s parents—a mercenary and an accused witch—might have flunked.

  Galton perhaps got the idea from Plato’s Republic. This too described a Utopian society, where members of the elite Guardian class were to be bred like “dogs for hunting.”15 Their children would be taken to a nursing area, where wet nurses would “provide for their nurture,”16 while “the offspring of the inferior, or of the better when they chance to be deformed, will be put away in some mysterious, unknown place, as they should be,”17 so t
hat the Guardians would be kept pure. (As always with Plato, it is hard to tell how literally this should be taken.) More recently, eugenics as a tool for improving society has been supported by people from the Nazis to Winston Churchill, and was part of government policy in places such as Alberta into the 1970s.18 Communists, on the other hand, championed nurture over nature and believed that traits could be moulded by the state. As Lenin boldly put it: “Man can be corrected. Man can be made what we want him to be.”19 Prediction has always been not just about foretelling the future, but about controlling it. And scientific predictions sometimes say as much about politics and sociology as they do about the system under study.

  SMOOTH OR WRINKLY

  We now think of ancestral qualities as being passed on through the genes. The idea that traits were transmitted from one generation to the other in discrete form—a kind of quantum theory of heredity— was first proposed by Gregor Mendel. The son of a farmer, Mendel was a keen gardener and a trained scientist (like Galton, he also published in the area of meteorology). In a series of careful experiments, carried out on the grounds of his Augustinian monastery, he tracked thousands of pea plants for several years, pollinating them by hand to control how they mated.20 He found that certain traits, such as colour and shape, were passed on in a discrete fashion. If a plant whose seeds had a smooth surface was bred with one whose seeds were wrinkly, the resulting seeds would be either smooth or wrinkly, rather than a mix of the two. He proposed that, for any particular trait, each parent plant had two “factors” and would contribute one to the next generation.

  Factors could be either dominant or recessive. Suppose that the factor for smooth seeds is S and the factor for wrinkly is W. A child plant inherits one factor from each parent, as shown in table 5.1 (on page 178). The factor S is dominant, so a seed that inherits at least one S will be smooth; a seed is wrinkly only if it inherits two Ws. If the factors S and W are equally distributed in the parents, 75 percent of peas should be smooth and 25 percent wrinkly. Mendel’s statistics for his 19,959 plants yielded 14,949 smooth (74.9 percent) and 5,010 wrinkly (25.1 percent), which was Pythagorean in its geometric precision (to the point that statisticians accused him of “correcting” the data). Heredity was all in the factors—or the genes, as they became known at the start of the twentieth century.

  PARENT l PARENT 2 CHILD

  S S S

  S w S

  w s s

  w w w

  TABLE 5.1. Each parent contributes one of its factors to the child. S denotes smooth, W is wrinkly. Because S is dominant, three of the four possibilities result in a smooth seed.

  Mendel’s theory was completely ignored until well after his death; it was not consistent with the views of Galton, who saw inheritance as a kind of blending, and it didn’t appear to fit with human experience.21 The traits of most children do not perfectly copy those of one parent or the other, but are usually a mix of the two. However, the gene theory was given a boost in the early twentieth century by experiments on Drosophila—those pesky flies that hang around the fruit bowl. The geneticist Thomas Hunt Morgan, at Columbia University, found that the flies inherited traits such as eye colour in the same way that peas inherited their texture.22 Later, Hermann Müller showed that the flies often suffered severe genetic mutations as a result of X-ray damage to their chromosomes (so named because they absorbed coloured dyes). This implied that the chromosomes were the location of the genetic material—and also, as Müller pointed out, that X-rays were more dangerous than they seemed.

  The history of chromosome research is another example of how scientific theories—and basic shape-recognition and counting skills—can be affected by sociological forces. In 1923, Müller‘s eminent colleague Theophilus Painter announced that, by means of complicated microscopic observations, he had counted the number of human chromosomes, and there were twenty-four pairs.

  Other scientists repeated his observations and came up with the same number. Some thirty years later, new methods allowed cells to be placed onto microscope slides, giving scientists a better look at the chromosomes. It soon became obvious that there were in fact only twenty-three of the little fellows. Painter had got it wrong, but his influence was so great that many scientists preferred to believe his count over the actual evidence. Textbooks from the time carried photographs showing twenty-three pairs of chromosomes, and yet the caption would say there were twenty-four.23

  Mendelian inheritance became widely accepted when it was found that the great majority of traits, including those studied by Galton, are the result of a combination of genetic and environmental factors, which is why they tend to vary over a continuous spectrum. The traits that Mendel had chosen to study were therefore the exception, rather than the rule (as Hermann Bondi wrote of Newton, “his genius selected an area where such perfection of solution was possible”). The mathematician R. A. Fisher, who was Galton Professor of Eugenics at University College London, also showed that while individual genes could vary in a discrete fashion at the individual level, the total “gene pool” would vary in the continuous manner assumed by Darwin’s theory of evolution.

  THE CENTRAL DOGMA OF THE SELFISH GENE

  By 1944, the genetic material in the chromosomes had been narrowed down to a “master molecule” by the name of deoxyribonucleic acid, or DNA. The key ingredients were four bases: adenine, cytosine, guanine (first found in guano), and thymine; these were denoted by A, C, G, and T. The quantum physicist Erwin Schrödinger suggested that the letters must spell out some kind of code.24 In 1953, the American biologist James Watson and the English physicist Francis Crick jointly published the structure of the molecule.25 Their double-helix model of DNA, confirmed by the X-ray diffraction results of Rosalind Franklin, won them the Nobel Prize; Franklin died at the young age of thirty-seven from ovarian cancer, perhaps caused by mutations in her own cells owing to Xray exposure.26 The unravelling of the structure of DNA appeared to solve a number of riddles almost at one stroke. It provided an explanation of how genetic information is stored and inherited; how it replicates when new cells are formed; and how it evolves. We learned how the information is translated into proteins, and potentially, how we can predict our future health.

  The double helix consists of two extremely long strings of the four bases. The strings are complementary, so a C on one always bonds to a G on the other and an A always bonds to a T. DNA is not a single molecule but is divided in humans into the twenty-three pairs of chromosomes. One of each pair contains genetic information from the mother, one from the father. Every cell in our bodies therefore contains two versions of the DNA story, with the exception of sperm or egg cells. In these, the genetic information is compressed into a single strand containing only a single copy; when egg and sperm merge at conception, the fertilized egg again has two copies. We therefore inherit a mix of genetic information from each parent, as Mendel described.

  When a cell divides to form a new copy, the two strands of DNA are unzipped and a complement to each strand is produced by adding the missing bases. The result is two copies of the original. The copying process is tightly controlled and most mistakes quickly repaired, so each human cell normally has only a few errors. New DNA errors are continuously introduced by external effects such as ultraviolet radiation or carcinogenic chemicals, however; sunlight and cigarettes both go right to the DNA. It’s a molecule that is always patching itself up. Despite this constant grooming, mistakes happen. A base is left out or a new one put in. If such point mutations happen in egg or sperm cells, they can be passed on to the next generation. Sometimes, this can result in serious genetic diseases; however, random mutations are also one of the drivers of evolution, since occasionally they provide a beneficial feature (like these fins on my back).

  While DNA seemed to be the “master molecule” of heredity, it had long been known to biologists that the workhorses of the cell, which gave it structure and carried out the metabolic processes that sustain it, were strings of amino acids known as proteins.
The problem of how the information in DNA is translated into proteins was cracked in the 1960s by a group of scientists that included Crick and a team at the Pasteur Institute in Paris. They showed that each three “letters” of DNA, such as the string ACT, represents one of the twenty different amino acids. Mendel’s factor—or gene—was a segment of DNA, consisting of perhaps thousands of letters, that codes for an entire protein.27

  FIGURE 5.3. The central dogma. Genes encoded in DNA are transcribed to form molecules of RNA , which are translated to form proteins, which carry out the work of the organism.

  Proteins are not produced directly from DNA. In an intermediate step, a molecule of RNA—a complement to DNA (like a photographic negative), but containing only the information relevant to the particular protein—is transcribed. The RNA is then translated into a string of amino acids, which must be folded into shape to form the actual protein. It is the shape of a protein that in large part determines its behaviour (like how it bonds with other molecules). At the molecular level, form and function are the same. The transcription of RNA—and its translation into proteins—is another tightly controlled process. Proteins that misfold or are flawed in any way are targeted for destruction by the cell—though different proteins may be folded from the same amino acid string.

  Francis Crick dubbed the DNA-RNA-protein model the “central dogma” of molecular biology—a Newton’s law for life. This highly stable, linear, one-way information flow has since been revealed to be considerably more complicated. Barbara McClintock showed that genes are themselves dynamic and can jump from chromosome to chromosome (or even species to species). The RNAs produced were also found to go through substantial splicing and modification before being translated. DNA is like a first draft that is reshaped by a very proactive editor. Evolution is therefore influenced by non-genetic factors, as well as random mutations.28 However, the central dogma was extremely successful in explaining the basic mechanics of inheritance—so successful that biologists like Richard Dawkins began to see life forms as little more than glorified delivery systems for their “selfish genes.”29