fwb12453-sup-0002

advertisement
1
Appendix S1. Coherence between pairing success measured in the laboratory and observed pairing
2
frequency in the field
3
Under laboratory conditions, amphipods originating from MOTUs genetically diverging by
4
approximately 3% formed an amplexus in 85% of trials. This means that, when amphipods from these
5
MOTUs meet, they have a 0.85 probability of pairing up. Their probability of meeting up depends on
6
their respective frequency in the population they originate from. Consider p(A) as being the
7
frequency of the MOTU A in the population and p(B) as the frequency of the MOTU B. Note that
8
these frequencies do not need to sum up to one, as other MOTUs may also be present in the
9
population. The probability for an A individual of meeting a B individual is Pm = 2 × p(A) × p(B). The
10
sex ratio is assumed to be even (which is the case in the populations we sampled) so Pm does not
11
depend on whether a male A encounters a female B or the opposite. Probability of pairing between
12
individuals A and B is thus Pm = 2 × p(A) × p(B) under random pairing and Pnon-random = 2 × p(A) × p(B) ×
13
0.85 when individuals follow the same rules for pairing as in our laboratory experiment. Determining
14
whether individuals do or do not discriminate between potential partners involves comparing the
15
frequencies of observed MOTU A/MOTU B pairs to the odd of finding such pairs if individuals were
16
pairing without discrimination. This is done by computing an odds ratio as a measure of effect size of
17
the difference between observed and expected frequencies for the considered pairing combination.
18
Considering that we sample MOTU A/MOTU B pairs from a natural population where individuals do
19
follow the rules described in our experiment, how many pairs would we need to sample to detect
20
that individuals from MOTU A and MOTU B only mate 85% of the time? Figure S2 represents the
21
odds ratio of the difference between observed frequency of pairing and the frequency of pairing
22
expected under no discrimination between A and B individuals, as a function of the number of pairs
23
sampled. For this example, p(A) = p(B) = 0.5. Expected frequency of MOTU A/MOTU B pairs is thus Pm
24
= 2 × 0.5² = 0.5 while the observed frequency should be Pnon-random = 2 × 0.5² × 0.85 = 0.425. In order
25
to be confident about the statistical significance of the difference between observed and expected
26
pairing frequencies (i.e. when the confidence interval of the odd ratio does not cross 1), a minimum
1
27
of 400 amphipod pairs should be sampled (Fig. S2). Biasing the frequency of one MOTU at the
28
expense of the other MOTU or decreasing the frequencies of both MOTUs only increases the number
29
of pairs needed to detect a statistically significant effect (result not shown). In the populations we
30
studied, it is possible that amphipods from MOTU genetically diverging by approximately 3% paired
31
up with a 0.85 probability. However, as we sampled about 100 couples per population, it is unlikely
32
that we could have detected a statistically significant difference between observed and expected
33
pairing frequencies between these MOTUs. This may explain the discrepancy between our field
34
observations and laboratory results.
35
Fig. S2. Odds ratio of the difference between observed frequency of pairing and the frequency of
36
pairing expected when individuals originating from genetically divergent MOTUs (~ 3%) pair up
37
without discrimination, as a function of the number of sampled pairs. Red lines represent the 95%
38
confidence interval for odd ratios. p(A) = p(B) = 0.5
39
2
Download