This tension sits within ecotype-environment-gene-content. At Harvard Forest, treatment × horizon explained 41% of genus-level variance 1,2. In contrast, a species-level environmental comparison was null 1,2. Read side by side, the two results could seem to give different answers about how strongly environmental context shapes microbial variation. However, the two analyses use different units, environments and response variables 1,2. That caveat matters because the pairing could be misread as a contradiction, or treated as resolved, when the analyses may not measure the same thing.
Evidence Sides
Side A: environment explains substantial genus-level variance (Harvard Forest)
At Harvard Forest, treatment × horizon explained 41% of genus-level variance 1,2. Here “treatment” is the experimental manipulation and “horizon” is the soil layer sampled. The 41% applies to the combined treatment × horizon factor, not to treatment alone, and it describes variance at genus level 1,2.
Side B: species-level environmental comparison is null
The species-level environmental comparison was null 1,2. This is a null result for the comparison that was tested. It is not evidence of a negative effect, and it is not a measured effect size.
The concept page explicitly states that the two sides differ in unit of analysis, environment and response variable 1,2.
Possible Reconciliations
- Hypothesis 1: scale of biological unit. Environmental filtering may act mainly on which genera are present and how abundant they are, not on gene-content variation within species. If so, both results could hold at once without conflict.
- Hypothesis 2: response-variable mismatch. Community composition and within-species gene content are different quantities. A strong signal in one does not imply a signal in the other, so the two findings may not bear on the same claim.
- Hypothesis 3: environment-specific effects. The forest-soil warming setting may produce stronger structuring than the environments represented in the species-level comparison. On this view, the difference reflects context, not a general rule.
- Hypothesis 4: partitioning of the variance. The 41% figure belongs to a combined treatment × horizon factor. Some of the genus-level signal may come from horizon (soil layer) and not from treatment. The tension text does not report how much each factor contributes, so this cannot be checked from the evidence given here.
None of these hypotheses is established by the evidence supplied here, and none is preferred.
Resolving Work
- Same-site gene content: Use Harvard Forest metagenome-assembled genomes (MAGs, genomes reconstructed from community sequencing) or isolate genomes to test whether within-species gene content differs by treatment and horizon. This asks whether a species-level signal appears where the genus-level signal was measured; it is a related question, not a retest of the reported null.
- Taxonomic-rank sweep: Apply the same variance-partitioning method to the Harvard Forest data at phylum, family, genus and species ranks. This asks whether the explained variance declines as the taxonomic unit gets finer.
- Matched response variables: Re-run the species-level environmental comparison with responses that can be compared with the Harvard Forest genus-level response. This asks whether the null reflects the choice of response variable rather than environmental weakness.
- Factor decomposition: Report separate variance terms for treatment and for horizon alongside the combined term. This asks how much of the genus-level signal is attributable to the treatment itself.