Partitioning phenotypic variation: implications for morphometric analyses (Bryozoa)

Using clonal replicates of Electra pilosa (L.) colonies grown in controlled environmental conditions, it is possible to partition sources of morphological variation of skeletal hard parts into their Genotype, Environment and GXE interaction components. Preliminary studies showed these factors to account for only about 40% of variation among measured zooecial characteristics, resulting in a Residual of 60%. Here we accounted for additional, within colony sources of variation in an ANOVA models with a nested factorial design: Patches of zooecia within replicate colonics and Columns and Rows of zooecia within Patches. We analyzed data for five zooecial characteristics (opesia and zooecia width and length and spine number). The large Residual variance (among zooecia - within colonies) was reduced by two-thirds, to < 20% when sources of within colony variation were accounted for in ANOVA models; Significant environmental variation can be detected (Tank effect -2-3%), even in experiments designed to minimize it as a main factor. Patch effect within colonies was highly significant, and often was greater than Genotype, — 25-35%. Failure to account for Patch effects can bias conclusions regarding other factors. If zooecia are sampled randomly and widely across a colony and are pooled, Patch effect variance will be allocated to the Residual. However, if zooecia are all collected from a single, small patch on a colony, or as single small colony fragment (e.g. fosgil), the variance attributable to Patch is incorrectly and differentially allocated to Genotype, Environment, and GXE interaction. Row and Column effects within patches were small, - 5-10%, but significant. If Row and Column are omitted as factors in ANOVA, their variance does not affect the outcome of other factors and is allocated to the Residual. Morphometric data should not be collected from single patches or small fragments of bryozoans, because the within colony Patch effect will artifactually and significantly inflate the apparent among Colony factor, frequently used by palaeontologists and neontologists as a proxy for genetic differences.