Teacher Notes

A. One step or many?

Outcomes 1. Pairs. Works well with a class tally on the board.

Expected observations

Of 50 hopeful monsters, most pairs will judge none, or one or two at a stretch, to be insect-like; most are small tangles, often radially symmetric. By breeding from the tree, most pairs reach something convincingly insect-like within a few dozen generations, typically through steady small improvements with occasional quick runs when a useful mutation appears.

Suggested answers

1. The source of variation was identical: random mutation. The difference is retention: cumulative selection keeps each small improvement and builds on it, so the probability of the end result is the product of many likely small steps rather than one astronomically unlikely leap.
2. The weasel works the same way: keeping the best partial match each generation turns a 1-in-10⁴⁰ event into a few dozen rounds. Dawkins himself stressed the weasel's limitation: it has a distant, fixed target, whereas natural selection has no target, only immediate reproductive success. The biomorph breeding with a goal shares that limitation (see I).
3. Evolution combines random variation with non-random selection. Selection is the opposite of chance, so "evolution is just chance" describes only half the process.

Watch for

• Students thinking mutations are directed towards what the breeder wants. They are not: the program mutates at random and the breeder only chooses.
• Students assuming complex forms need large mutations.

B. Selection or drift?

Outcome 2.

Expected observations

Drift wanders with no consistent direction. Complexity rises and falls: forms may balloon into large, many-twigged shapes or collapse into small ones. Resemblance to the starting tree is usually lost within a few dozen steps. A two-minute burst of breeding for height produces sustained, directional change by comparison.

Suggested answers

1. Change is any heritable alteration; adaptation is change that improves fit to a particular function or environment, and it requires selection. Drift produced plenty of change but no adaptation.
2. Genetic drift is random change in allele frequencies through sampling in a finite population (for example, which individuals happen to reproduce). A model would need a population of many biomorphs carrying different variants, parents drawn at random each generation, and a measure of variant frequencies over time, so that variants could be lost or fixed by chance. The program's Drift is closer to neutral mutation accumulation in a single lineage: a random walk through morphospace.

Watch for

• Equating drift with mutation. Make the distinction explicit, since the program's name invites the confusion.

C. From genes to bodies

Outcome 3. A gene-effect table makes a good short write-up.

Expected observations

• Genes 1–3 change the sideways spread of different kinds of branch, everywhere in the tree at once.
• Gene 4 changes the trunk and every vertical stem; gene 7 changes the downward-angled branches.
• Gene 9 changes the number of forkings, and each step doubles the twigs.
• Gene 10 stacks copies of the body (segments); gene 13 adds up-down or four-way symmetry.
• With gene 9 at 1, only the trunk is drawn, so genes 1–3 have no visible effect at all.
• A gradient on gene 4 makes each segment's stems longer (or shorter) than the last.

Suggested answers

1. Yes: every biomorph gene acts at every fork of its kind, so every gene is pleiotropic. This coordinates change (the whole body changes in step), but it also constrains it: a change that improves one part may spoil another. That favours mutations of small effect, a point made by R. A. Fisher.
2. Genes 1–3 are invisible when gene 9 is 1; gradients and segment distance do nothing in a one-segment biomorph; the scaling factor changes the effect of all the size genes.
3. Engineering requires knowing in advance which change produces which result. Nature has no engineer with foresight. Breeding needs only the ability to recognise a slightly better variant among those that turn up, which is exactly what differential survival and reproduction provide.

D. Evolving evolvability

Outcome 4. The longest investigation; pairs comparing Albums is the payoff.

Expected observations

In the nine-gene world students find trees, fans, insects, spiders and lattice-like shapes, all left–right symmetric, single-bodied and unsegmented. Once the new mutations are switched on, segmented "centipedes" and "plants", radially symmetric "stars", lopsided forms and tapering series appear, often within minutes.

Suggested answers

1. Segmentation changes the developmental rule rather than a particular shape: one mutation repeats an entire body plan, and gradients then let the repeats differ. It makes whole new regions of morphospace reachable in a few steps.
2. Candidate key innovations include segmentation in arthropods, annelids and vertebrates (somites); duplication of Hox genes; multicellularity; bilateral symmetry; jaws; and the tetrapod limb.
3. Dawkins (1989) suggested that lineages whose embryologies generate richer variation might come to dominate over long timescales: a kind of higher-level selection for evolvability. The idea is debated; it is a good one for students to argue about rather than accept.

E. Reading the fossil record

Outcome 5. Remind students to Initialize the record before breeding.

Expected observations

Change is gradual from fossil to fossil, with a visible turn where the goal changed, and occasional reversals. Fossils ten generations apart can look quite different, and their relationship is not always obvious.

Suggested answers

1. Sampling every tenth fossil makes gradual change look like a series of jumps. The record alone cannot tell gradual change from rapid change between samples.
2. Wherever the samples are far apart. "Missing links" are an artefact of incomplete sampling of a continuous process. Note also that even this complete record follows one lineage: the side branches (the unchosen offspring) left no fossils at all.

F. Descent from a common ancestor

Outcome 6. Pairs well with any phylogenetics teaching.

Expected observations

Siblings differ only slightly, since they are one round of mutation from their parent. The tips of two long chains differ much more, and more so the further apart they are in generations. Killing a child of the Adam removes that child and every one of its descendants.

Suggested answers

1. Each lineage accumulates its own independent mutations, so differences grow with the number of generations since the common ancestor, even with no selection. This is the logic behind molecular clocks.
2. Branching descent with modification necessarily produces groups within groups: the traditional hierarchy of classification and the basis of phylogenetic trees. A world of species that did not fall into nested groups, for example traits scattered at random across groups, would be hard to explain by common descent.
3. The extinction of an early lineage: everything descended from it is lost with it, a whole clade.

Watch for

• Detach is a bookkeeping tool for rearranging the display, not a biological event.

G. A walk through Biomorph Land

Outcome 1. Short and visual; good as a demonstration.

Expected observations

The pointer preview changes continuously, but intermediates are often not halfway in appearance: some features switch suddenly (symmetry genes and the depth gene change in whole steps) and others interact. Outside the triangle, features are exaggerated beyond any anchor, sometimes into large or bizarre forms.

Suggested answers

1. Every step changes one gene a little, yet in a space of many dimensions there are many directions to step in. Any distant point can be reached through a long series of small steps, and selection only ever has to choose among near neighbours.
2. Almost all points in the space are unremarkable, so a random landing (a hopeful monster) is almost never interesting. Stepwise selection follows a path of small improvements to the rare interesting regions.

H. The alphabet challenge

Outcomes 1 and 4. Popular as a self-directed challenge.

Expected observations

Mirror-symmetric letters (A, H, M, T, U, V, W, X, Y) can be approached; asymmetric ones need the Asymmetry gene and careful work, and some are frustratingly hard. Students often hit dead ends and have to back off, breeding away from the target for a while, before progress resumes.

Suggested answers

1. The branching rule produces mirror-symmetric shapes unless gene 12 says otherwise, so symmetric letters lie close to the starting point and asymmetric ones far away. In biology, the way development works makes some variants far more likely to arise than others: developmental constraint.
2. Natural selection has no distant target and no memory of a goal; it favours whatever reproduces best now. Breeding towards a chosen letter is goal-directed artificial selection.

I. Who is the watchmaker?

Outcome 7. Best as a whole-class discussion.

Suggested answers

1. Breeding biomorphs is artificial selection. Dawkins acknowledged this: the selecting is done by a human eye. Like Darwin with domestic pigeons, the program demonstrates the power of cumulative selection, and natural selection then supplies the selector.
2. In nature the "choice" is differential survival and reproduction in an environment: predators, competitors, climate, mates. A mechanical rule such as "the most twigs" corresponds to something like "plants that intercept the most light leave the most seeds". No one needs to see the outcome in advance.
3. The program leaves out populations of many individuals, sex and recombination, environments and ecology, competition and death, selection on reproductive success rather than appearance, changing conditions, and realistic timescales. It also has a human with foresight. Interestingly, Dawkins wrote an automatic target-seeking mode for the program, which v1.1 left switched off.