Biology Reference
In-Depth Information
Species are not independent
Statistical testing of hypotheses requires that data points are independent. Imagine we
wanted to compare body size in men and women, and that we measured several males
from the Smith family and several females from the Jones family. Although, at a population
level, men tend to be larger than women, we might not obtain this result in our study.
Maybe members of the Jones family are particularly tall, or maybe they are wealthier and
have more food. In this case, we would obtain a spurious result that is driven by differences
between families rather than between the sexes. Put another way, the data points within
each sex are not independent, and this can increase the likelihood of incorrect conclusions.
Exactly the same problem can arise when comparing between species (Clutton-Brock &
Harvey, 1977). Closely related species tend to be similar because they share traits by
common descent rather than through independent evolution. To give an extreme example,
the Australian mammals are mostly marsupials that carry infants in pouches, whereas
British mammals are all placental, using a placenta to nourishing their young in the
mother's uterus. If we compared British and Australian mammals, then we would find that
having a pouch would correlate with any environmental variable that differed between
Britain and Australia, such as mean temperature, rainfall, proportion of land that is desert
and so on. However, to infer that any of these variables has selected for this difference would
be foolish. The difference is much more easily explained by the historical fact that the
placental mammals evolved after Australia became isolated from the other continents.
The problem of non-independence does not rely on such extreme historical patterns
or a lack of evolutionary flexibility in a trait (Ridley, 1989). For example, considering
the patterns across primates, the gibbons (
Hylobates
spp.) are all monogamous, eat fruit
and hold territories. Consequently, every time you add a new species of gibbon to a
comparative study, you increase the extent to which these traits are correlated across
species. Whilst this could plausibly be because natural selection links these traits
(e.g. large territories are required to obtain enough fruit, and this spreads individuals
out, favouring monogamy), it could also be explained by an almost infinite number of
possibilities (e.g. gibbons are fruit eating specialists, but it is something else that they all
do that favours monogamy).
Species may be
similar through
common descent …
… which may bias
comparative
analyses
Phylogenies
To solve the problem of species non-independence, it is necessary to take phylogenies
into account. A phylogeny is a tree which shows the evolutionary relationships among
species (Fig. 2.8a; ignore the details for the moment). Initially, morphological traits
were used to construct these trees, but nowadays they are usually based on similarity
in DNA sequences in nuclear or mitochondrial genes; the more similar two species are,
the more recently they must have shared a common ancestor. If the mutation rate is
known (by calibration from fossils or geological events of known dates), then the
magnitude of the difference in DNA sequence becomes a 'molecular clock' which
estimates the time since two species last shared a common ancestor. Branch lengths in
the tree can then indicate the time that has elapsed since divergence. Various statistical
methods are used for reconstructing the most likely phylogeny, given the DNA
sequences of extant species.
