Community composition differed between disturbed and undisturbed plots. Disturbed plots had higher species richness, consistent with previous results in old field systems [e.g 15]
. Nonmetric multidimensional scaling confirmed that the two habitat types differed in their species composition. This difference was exaggerated in the analysis incorporating the phylogenetic node structure of the plots. This means that disturbed plots are quite phylogenetically distinct from undisturbed plots, with each type of plot represented by different terminal clades. This can be seen visually in and .
Recently disturbed plots displayed larger variation in phylogenetic ordination scores than undisturbed plots. This is likely because most species and clades that are dominant in undisturbed habitats are also found to a lesser degree in disturbed plots. Several factors are likely responsible for this, including the fact that many are perennial plants which can resprout from rhizomes after the plots were plowed. In addition there is probably a large propagule pressure from old field dominants entering into the disturbed plots. On the other hand, very few species or clades characteristic of disturbed habitats were also found in undisturbed habitats. This may be because most are annual weeds which cannot easily establish under the conditions of high density and shading in the undisturbed plots, because they are adapted to open habitats.
Overall, I found that when all plots were considered together, they were on average significantly clustered phylogenetically, when compared to a null model. This was true both for PSV, and PSE when compared to a null model of species assembly. It was also true when pairwise co-occurrence values were correlated with phylogenetic distance. This is consistent with many other studies of phylogenetic structure in plant communities, most of which have found phylogenetic clustering when structure was found [reviewed by 41]
. However, simulation studies have found that these tests can be liberal under several circumstances 
, including cases of spatial autocorrelation due to limited dispersal, and phylogenetic structure in the experiment-wide abundances. Abundance Phylogenetic Deviation index (APD 
) measures the clustering of species abundance on the phylogeny by comparing the mean phylogenetic distance of species in the experiment, with the abundance-weighted mean phylogenetic distance. Positive values suggest clustering of abundances, whereas negative values suggest overdispersion of abundances. In this study the APD value for undisturbed plots was 0.08 and for disturbed plots it was 0.10 – slight clustering. This could have made the overall test liberal, however, it is the comparison among habitat types which is the important result. Since the APD values are close, there is little reason to suspect that the test for the disturbed sites is more liberal than that for the undisturbed sites, and so there should not be a higher probability of finding significant results in the disturbed plots.
Despite no difference between disturbed and undisturbed habitat types in phylogenetic diversity indices, there was a difference in how they deviated from their expected phylogenetic diversity, based on null models of community assembly (). Disturbed plots were significantly more ‘clustered’ than expected under a null model of community assembly, whereas undisturbed plots did not deviate from random expectation. This could only have come about if the ‘expected’ phylogenetic structure of each habitat differed.
In this study, I found that the available species pool for disturbed and undisturbed habitat differed, and that the phylogenetic diversity of each pool also differs. The average phylogenetic diversity expected under a null model of community assembly is reflective of the underlying phylogenetic structure of the species pool. Disturbed plots had a higher expected phylogenetic diversity, suggesting that at a regional scale disturbed areas contain lineages that are less related than in undisturbed areas. This is contrary to my expectation, as it is usually thought that disturbance should select closely related species. One possible explanation is that the undisturbed communities actually constitute a harsher environment for species due to their high level of competition, and that there may be a suite of traits that make species suited to this environment which are phylogenetically conserved. On the other hand, since I essentially only have a single sample for each habitat's pool of species, this difference may be due to chance alone. It is impossible to assess the generality of this pattern because most studies which have found higher clustering in disturbed habitats have failed to distinguish between regional species pool differences and more local plot level differences 
The results at the plot-level within each habitat type were different. Disturbed plots were more phylogenetically clustered than expected by chance, so that individual plots had, on average, lower phylogenetic diversity than their regional habitat pool. In undisturbed habitat, phylogenetic diversity in individual plots did not differ significantly from the regional habitat pool. A weaker competitive environment in recently disturbed plots could lead to more phylogenetic clustering than expected in several ways. If the environmental tolerances of species are phylogenetically conserved, then differences in the environment could act as a filter, and closely related species will be more likely to coexist. A heavy disturbance such as plowing could create a harsh or unique environment that selects for species that can tolerate these conditions 
Stripping away dominant vegetation could lay bare environmental variation which was masked before. This could happen if the competitive stresses of the environment are strong enough that they become more important than anything else, and so in a sense, homogenize the environment. If so, such a competitive environment will act as an initial filter, reducing the species pool, but thereafter species are distributed randomly with respect to phylogeny. This could also happen if traits relating to competition are less phylogenetically conserved than those involved with dealing with abiotic stresses which may be more prevalent in recently disturbed environments.
Another possible explanation is that both environments are experiencing forces that promote phylogenetic clustering, but in the undisturbed environment there are also strong counteracting forces promoting phylogenetic overdispersion. This would happen if ecological traits related to niche partitioning were phylogenetically conserved, and that the stronger competition in undisturbed plots led to stronger niche differentiation. Opposing processes that counteract each other's effect on phylogenetic structure has been demonstrated before. One study found that when environmental factors were statistically removed from sunfish communities, phylogenetic overdispersion was revealed 
. An argument against this possibility is that phylogenetic conservation of niches has been difficult to demonstrate in plants. For example a study of meadow communities showed that phylogenetic distance was not correlated with niche seperation along several axes of soil conditions 
. Another study came to the conclusion that the intensity of competition between plant species pairs was only weakly correlated with phylogenetic distance in a meta-analysis of pot experiments, and only for certain taxa 
Though competition and environmental filtering are often thought of as dominant forces in the structuring of communities, there are other factors that could come into play in this system and others. Predation, or herbivory in this case, could be involved in the structuring of communities. Herbivory could promote phylogenetic overdispersion if herbivores fed on more than one species, and those species tended to be closely related, through the action of ‘apparent competition’ 
, in a manner analogous to resource competition. Theory has shown that apparent competition can act very similarly to resource competition 
, and so limiting similarity may act here as well, only in this case the similarity is in shared predators rather than shared resources. This effect can also be thought of as a phylogenetic extension of the Janzen-Connell hypothesis 
as described in a recent review of community phylogenetics 
. Such “Janzen-Connell” effects could be stronger in the undisturbed plots. Though it is unlikely that herbivore pressure differs greatly between the habitats (given their close spatial proximity), herbivore effects would likely be gradual and only result in significant difference in community structure over many years – years which the undisturbed plots have experienced and which recently disturbed plots have not.
Predation could also act as a filter. Though smaller specialized herbivores such as insects may not vary between the habitats, deer herbivory may. Deer are common on the property on which I conducted my surveys and they may impose pressure on old field plant communities. The increased exposure of disturbed fields makes plants more apparent, and so deer could be a stronger force in recently disturbed plots. If traits that lead to deer-resistance are phylogenetically conserved, deer could act as a filter leading to phylogenetic clustering in areas where deer are more common.
Many of the effects discussed above vary in the timescale over which they act. Most of the filters promoting clustering will act immediately, whereas those thought to promote overdispersion will act gradually. It may be useful to define filters as density-independent effects on fitness, whereas competitive effects (including both resource and ‘apparent’) are density-dependant, in that they become stronger in high densities. It may be then that disturbance exposes plant communities to environmental filters which leads to greater than expected phylogenetic clustering at low densities, after which communities gradually return to their ‘expected’ level of phylogenetic structure, through the action of weak dispersion promoting forces, such as limiting similarity and Janzen-Connell effects, which become important as densities increase. It is particularly interesting to note however, that in this system this process leads to no difference in the absolute phylogenetic diversity of the different habitats, due to differences in their species pool and therefore differences in the expected phylogenetic structure of each habitat.
This could have implications for conservation. It is becoming clear that phylogenetic diversity has consequences for ecosystem functioning 
. If so, reductions in phylogenetic diversity could have negative effects that may be independent of the effects of species richness. Indeed, one study found that urban areas (assumed to be more disturbed) actually had higher species diversity of plants but that phylogenetic diversity was lower 
. This likely reduces the positive aspects of increased species richness. Consistent with many other studies 
, I demonstrated the ability of disturbance to decrease the phylogenetic diversity of an area, however, I also show that whether this leads to an absolute difference in phylogenetic diversity between disturbed and undisturbed habitats is dependant on the pool of available species in each habitat. In this case, disturbed areas had the potential for higher phylogenetic diversity than undisturbed areas, but clustering at the plot level led to statistically indistinguishable values for phylogenetic diversity in each habitat.
Factors that influence ecological succession may act in a biased manner with respect to phylogeny, because of a correspondence between phylogeny and ecological similarity. Therefore, phylogenetic information should be useful in understanding these forces. In this study I found phylogenetic information could be used to get a fuller picture of compositional changes in plant communities. In particular, disturbed plant communities were more phylogenetically clustered than expected by chance, suggesting the action of environmental filters on phylogenetically conserved traits. Importantly, this led to no difference in phylogenetic diversity between disturbed and undisturbed plots, because the underlying species pool for disturbed plots had a higher phylogenetic diversity. This suggests that processes that structure communities can have different effects on phylogenetic diversity at different scales, from the regional to the plot level. This necessitates the careful choice of null models when comparing phylogenetic diversity indices amongst habitats. Analyzing differences in phylogenetic structure and composition at different scales can lead to useful insights into habitat differences in community composition.