The analysis of plant root responses to nutrient concentration, soil volume and neighbour presence: Different statistical approaches reflect different underlying basic questions

To investigate the responses of plants to their below-ground neighbours independently of nutrient availability, experiments generally require a solitary treatment with one plant grown alone with one unit of nutrients, and a neighbour treatment with two plants grown together with two units of nutrients. This can either be done by doubling nutrient concentration (C) or by doubling soil volume (V) in the neighbour treatment as compared to the solitary treatment. Statistically analysing the same dataset from an experiment that grew plants in solitary or neighbour treatment with a series of V given a fixed amount of nutrients per plant (e.g. 1 g), Chen et al. (2015a) found significant neighbour effects when they controlled for V, while McNickle (2020) found the effects to be insignificant when he controlled for C. The discrepancy in the results of the two studies is caused by a difference in their analytical approaches. This includes (a) different choices of data transformation for the controlling factor, and (b) a mathematical deviation of model structures between V-based and C-based analyses, due to the different inversely proportional V-C relationships between solitary (Formula presented.) and neighbour (Formula presented.) treatments. Choices for either V or C as a controlling factor in the analyses for ‘neighbour effect’ are based on two different perspectives, focussing either on neighbour-induced nutrient depletion (like McNickle, 2020) or on identity recognition (like Chen et al., 2015a). We also raise concerns about the use of mesh-divided root interaction design and replacement series design in the studies of plant–plant root interactions. We propose to adjust the experimental designs and analytical methods based on the focal perspectives of neighbour effect.