"... In PAGE 7: ... After the end of game, we sort and choose top 15 strategies in the population as the test strategies. Table3 shows the opponent strategies which are selected from the initial population, when the number of players is 8 and history length 1. Tables 4 and Table 5 show the results of the evolved strategies playing against best random strate- gies and themselves, respectively.... ..."

"... In PAGE 11: ... An alternative to land area that can be used to control for the size of a country is population. The results when population is used are reported in Table4 ; I use the average population between 1400 and 1800.12 The results are robust to this alternative control for country size.... ..."

"... In PAGE 4: ... We studied the performance of the genetic algorithm on the data set used above, varying the population sizes. The results are shown in part in Table2 . The table focuses on tness function values only, as this is an appropriate measure of the quality of the optimiza- tion process itself.... In PAGE 6: ... Therefore, the traditional measure of convergence is not applicable. Indeed, the gures in Table2 demonstrate that the homogeneity metric can vary signi cantly over the course of the run. Most of the conventional wisdom regarding popula- tion size selection derives from the problems with di- versity and convergence.... ..."

"... In PAGE 13: ... Table8 present slightly better F1 categorisation performance results for the case of the language background-speci c cohort compared with the gender- speci c cohort results. Again, there is a general improvement, though not dra- matic, in performance as the the number of e-mails in each language cohort class increases.... ..."

"... In PAGE 12: ....1. Scenario 1 - The estimation of metapopulation persistence in the study areas The results of simulations of the behaviour of popula- tions of G. leadbeateri in the old growth patch struc- tures within the Ada and Steavenson Forest Block are presented in Table2 . There were significant differ- ences in the extinction probabilities between the two study areas, irrespective of whether fires were, or were not, included in the simulations (Table 2).... In PAGE 12: ... leadbeateri in the old growth patch struc- tures within the Ada and Steavenson Forest Block are presented in Table 2. There were significant differ- ences in the extinction probabilities between the two study areas, irrespective of whether fires were, or were not, included in the simulations ( Table2 ). In all cases, the values for the extinction probabilities were significantly greater in the Ada Forest Block than those derived from simulations of the Steavenson Forest Block (Table 2).... In PAGE 12: ... There were significant differ- ences in the extinction probabilities between the two study areas, irrespective of whether fires were, or were not, included in the simulations (Table 2). In all cases, the values for the extinction probabilities were significantly greater in the Ada Forest Block than those derived from simulations of the Steavenson Forest Block ( Table2 ). Not surprisingly, we recorded higher values for the predicted probability of extinc- tion when wildfires were more frequent or more extensive.... In PAGE 13: ...lock were occupied by G. leadbeuteri. The number assigned to each patch corresponds to that given in Figure 2. This value is the increase above the value generated for the predicted prob- ability of extinction after 150 years (= P[E],,,) when all old growth patches were included in the patch structure (P[E],,, = 34%; see Table2 ). A minimum of 300 simulations were com- pleted in each case where a given patch was deleted from the patch structure.... In PAGE 13: ... Patch no. Patch size Identity of Proportion Decline (%) (ha) other patches of time in viability to which patch is connected I 3 2 4 3 11 4 5 5 2 6 2 7 8 8 14 9 5 10 8 11 12 12 16 13 16 14 16 15 6 16 7 17 3 18 2 19 2 20 1 21 2 22 2 23 3 24 2 25 2 26 I 27 1 2 1 4 3, 5 4,6 5 8 7,9 8, 10 9 12 1 I, 13, 16 12, 19, 20 15 14, 16, 17 12, 15 15 19 13, 18,20 13, 19, 21 20 - - - - - - occupied from deletion* 11 - 15 53 8 22 - - 6 5 - - 32 - 48 11 25 - 28 12 66 24 82 61 82 26 74 29 51 26 57 23 16 - 11 - 13 - - 8 9 4 - 7 3 3 2 8 - - - - - - *These values correspond to the increase in P[EIl5, relative to the scenario with all old growth patches were included in the patch structure (P[E],,, = 34%; see Table2 ). The % of the via- bility that was lost was calculated from the equation ... In PAGE 14: ... The analyses were completed without the impacts of wildfires and the patch deletions were limited only to stands of old growth forest. Identifying number Area (ha) of Decline of of patches deleted system viability due to patch deletion (%)* Deletion of areas by patch size classes All patches I 3 ha 28 All patches I6 ha 48 All patches I 9 ha 71 94 All patches 5 15 ha I08 All patches I I2 ha Deletion of ensembles of patches 1 amp;2 7 3,4 amp;5 18 7, 8,9 amp; 10 35 12 amp; 13 32 12 amp; 14 32 13 amp; 14 32 12, 13 amp; 14 48 8 amp; 12, 13, 14 62 11-21 (inclusive) 73 22,23,24,25 amp; 26 10 - II 18 61 67 20 73 86 53 91 100 91 Denotes the increase in P[E] relative to the scenario with all old growth patches were included in the patch structure (P[E],,,, = 34%; see Table2 ). For example, if the value for P[E],,,, fol- lowing the deletion of a patch is 70% then the loss of remaining viability is (70-34)/( 100-34) = 55%.... In PAGE 15: ... Column 4 con- tains information on the change in the probability of metapopu- lation extinction in response to the deletion of a given patch. This value is the increase above the value generated for the pre- dicted probability of extinction after 150 years (= P[E],,,,) when there were no deletions and all old growth patches were includ- ed in the patch structure P[E],50 = 48%; see Table2 ). A dash in column 4 signifies that there was no significant change (i.... In PAGE 18: ...1. Scenario 1 - The estimation of metapopulation persistence in the study areas Our analyses revealed large differences in the proba- bility of extinction between the two wood production forest blocks that were modelled ( Table2 ). This result was expected as there are major differences in the amount and spatial configuration of old growth patches in the two areas (compare Figs.... In PAGE 18: ... values for the predicted probability of extinction of G. leadbeateri in the Ada Forest Block over the next 150-300 years were substantial, even in the absence of the impacts of wildfires ( Table2 ). These findings highlight a concern for the long-term prospects for the survival of the species in many areas of its present distribution.... ..."

"... In PAGE 10: ... If different populations exposed to dif- ferent environmental stimuli and ecological challenges were to be compared, the repertoires of the agents in the population would be even more different. Table1 shows the interpopulation category variance cvH11032, a metric used to show how well categories compare across populations. It is the average of the category variance com- puted between all agents of two different populations P and PH11032.... In PAGE 10: ... It is the average of the category variance com- puted between all agents of two different populations P and PH11032. The number of agents in the populations P and PH11032 are n and m, respectively, which are assumed to be equal for all populations being compared: (12) Table1 shows that the category sets of agents within and across populations are quite dissimilar (an intuitive grasp can be obtained by comparing the values in this table with other category variance tables in the following sections). If the categories of agents are similar between two popula- tions, cvH11032 decreases.... In PAGE 44: ... Actual agents do not show the close agreement that language could potentially support. As an illustration of this, Table1 compares the average within-language variance to the variance in mean foci across languages for red, green, blue, or yellow terms for the WCS respondents, based on an analysis by Webster and Kay (in press). (These are calculated from the raw distances in the Munsell palette for the Hue and Value dimensions separately.... In PAGE 44: ... (These are calculated from the raw distances in the Munsell palette for the Hue and Value dimensions separately.) For each language, terms corresponding to the En- glish terms were determined by finding the focal choices for con- Commentary/Steels amp; Belpaeme: Coordinating perceptually grounded categories through language 512 BEHAVIORAL AND BRAIN SCIENCES (2005) 28:4 Table1 (Webster amp; Kay). Average variance in individual foci within a WCS language compared to the variance of mean foci between languages, computed for the hue or lightness of red, green, blue, or yellow.... In PAGE 45: ... 2002). For example, Webster and Kay an- alyzed whether differences between the average foci for the WCS languages were larger than predicted by random sampling across languages; and as Table1 shows, differences were significant for all terms. Nevertheless, as indicated earlier, the within language variances are much higher than the variances in mean foci be- tween languages.... ..."

"... In PAGE 14: ... While in this paper, the members do not keep history of their previous state, future work will investigate use of previous state in appropriateness evaluation. Table3 shows the function used by members to evaluate their appropriateness for status change. For the self-evaluation of appropriateness in the simulations presented in this paper, members use the midpoint (thM) between the upper threshold (thU) and the lower threshold (thL) as the target group size for local groups.... ..."