Two alternatives to the stress-gradient hypothesis at the edge of life: the collapse of facilitation and the switch from facilitation to competition



New evidence demonstrates that facilitation plays a crucial role even at the edge of life in Maritime Antarctica. These findings are interpreted as support for the stress-gradient hypothesis (SGH) – a dominant theory in plant community ecology that predicts that the frequency of facilitation directly increases with stress. A recent development to this theory, however, proposed that facilitation often collapses at the extreme end of stress and physical disturbance gradients. In this paper, we clarify the current debate on the importance of plant interactions at the edge of life by illustrating the necessity of separating the two alternatives to the SGH, namely the collapse of facilitation, and the switch from facilitation to competition occurring in water-stressed ecosystems. These two different alternatives to the SGH are currently often amalgamated with each other, which has led to confusion in recent literature. We propose that the collapse of facilitation is generally due to a decrease in the effect of the nurse plant species, whilst the switch from facilitation to competition is driven by environmental conditions and strategy of the response species. A clear separation between those two alternatives is particularly crucial for predicting the role of plant–plant interactions in mediating species responses to global change.

Molina-Montenegro et al. (2013) recently reported the occurrence of strong facilitation in one of the most stressful environments on Earth; the moss and lichen-dominated communities of Maritime Antarctica. The lichen species Usnea antarctica hosted four of 11 lichens and five of 13 mosses in this community, and increased the survival of transplants of one of only two native vascular plants, Deschampsia antarctica, present in this ecosystem. In a commentary published in the same issue, Callaway (2013) states that the facilitation observed by Molina-Montenegro et al. (2013) was as strong as reported at the extreme end of alpine gradients in Callaway et al. (2002) and Butterfield et al. (2013), although the two latter studies were not specifically designed to assess interactions at the edge of alpine plant life. Callaway concluded that ‘facilitation remains an important organizing principle of life, even to the point where vascular plants begin to drop out of communities altogether’ consistent with the stress-gradient hypothesis (SGH: Bertness & Callaway 1994; Brooker & Callaghan 1998). This statement is also in agreement with the recent meta-analysis of He et al. (2013), showing that competition switches to facilitation with increasing environmental stress. However, like the Callaway et al. (2002) and Butterfield et al. (2013) studies, this meta-analysis did not analyse variations in interactions at the extreme end of stress gradients. Interestingly, Koyama & Tsuyuzaki (2013), in the same issue of JVS, found that facilitation collapsed in a post-mined Sphagnum peatland in extreme environmental conditions. Casanova-Katny et al. (2014) also identify important limitations to the Molina-Montenegro et al. (2013) study: the facilitation detected was artificial because the transplanted target species D. antarctica never naturally occurs within lichen cushions. The goal of this forum is not to stress other limitations of the Molina-Montenegro et al. study but to clarify an important conceptual misunderstanding in the literature regarding alternative hypotheses to the SGH at the edge of life by: (1) describing two alternatives to the SGH found in the literature, (2) proposing causal mechanisms for each alternative, and (3) highlighting important theoretical and applied implications of these theory refinements.

The two alternative hypotheses to the SGH at the edge of life

There is an increasing confusion in the literature between two fundamentally different patterns of variation in plant–plant interactions along environmental gradients, i.e. between the collapse of facilitation in extreme environmental conditions (Michalet et al. 2006) and the switch from facilitation to competition described in water-limited ecosystems (Tielbörger & Kadmon 2000; Maestre & Cortina 2004). A large body of recent studies have shown that facilitation does not increase linearly with increasing stress or disturbance, but conforms to a unimodal (i.e. hump-backed) curve with diminishing interaction frequencies at the extreme end of environmental gradients (Kitzberger et al. 2000; Brooker et al. 2006; Smit et al. 2007; Forey et al. 2010; Le Bagousse-Pinguet et al. 2012; Koyama & Tsuyuzaki 2013). This addition to the SGH was proposed conceptually by Michalet et al. (2006) with respect to species richness effects and was further developed by others (Xiao et al. 2009; Holmgren & Scheffer 2010; Malkinson & Tielbörger 2010; Le Bagousse-Pinguet et al. 2013a; Verwijmeren et al. 2013).

The second viable alternative hypothesis is that the SGH is not supported at extremes of water gradients due to increasing competition for water (Davis et al. 1998; Tielbörger & Kadmon 2000; Maestre & Cortina 2004; Armas & Pugnaire 2005). Numerous models of increasing competition with decreasing resource availability describe these patterns (Taylor et al. 1990; Davis et al. 1998) and serve as viable alternative hypotheses to the SGH at extreme end points of gradients – most likely for water. The meta-analysis of Maestre et al. (2005) similarly found no changes in interactions with increasing drought in water-stressed ecosystems, but may also be interpreted differently (Lortie & Callaway 2006; Maestre et al. 2006, 2009; Michalet 2006, 2007; Malkinson & Tielbörger 2010). Hence, this alternative hypothesis proposes that interactions switch not from positive to neutral but to competition due to limiting water resources.

These two alternatives have been gathered into a single model (Holmgren & Scheffer 2010; Malkinson & Tielbörger 2010) because the frequency of interactions is unimodal in both. Unfortunately, this has led to confusion between them (e.g. de Bello et al. 2011; Dvorský et al. 2013) and we should not be focussing on the peak of the hump-shaped curve but instead on the mechanism associated with generation of the tail.

The underlying mechanisms of the two alternatives to the SGH

We propose here a simple means to differentiate support for the two alternative hypotheses at the extreme ends of SGH studies. The collapse of positive interactions is typically driven by changes in the ‘nurse plant effect’, whereas the switch from facilitation to competition is typically driven by the ‘target species responses’ to a particular environmental limitation. To date, most cases of collapse have been shown when stress or disturbance decreases the facilitative effect of nurse plants, primarily on grazing gradients (Baraza et al. 2006; Brooker et al. 2006; Smit et al. 2007; Le Bagousse-Pinguet et al. 2012), but also on physical disturbance gradients (Forey et al. 2010; Maalouf et al. 2012a,b) and nutrient stress gradients (Le Bagousse-Pinguet et al. 2013b). Baraza et al. (2006) showed that very high levels of grazing decreased the refuge effect of nurse plants. Similarly, Forey et al. (2010) evidenced a collapse of the facilitative effect of the shrub Helichrysum stoechas for sand dune beneficiary species with decreasing shrub size due to increasing sand deposition in the Atlantic coastal dunes. Le Bagousse-Pinguet et al. (2013b) found, in the same system, that nutrient stress may also decrease shrub size leading to a similar collapse of facilitation. These studies support the alternative hypothesis that a collapse of facilitation is due to changes in the effect of a nurse plant.

In contrast, the switch from facilitation to competition is due to the response of beneficiary species to a decrease in a resource. It is now well recognized that resource vs non-resource factors can lead to different sets of interactions. Facilitation generally switches to competition in highly stressed conditions when species are interacting for a resource factor such as soil water (Michalet 2007; Maestre et al. 2009; Saccone et al. 2009). This alternative does not appear to apply to environmental gradients driven by non-resource factors, including temperature, salinity or aerial humidity, even at the extremes (Michalet 2007; Maestre et al. 2009; Holmgren & Scheffer 2010). Hence, changes in the nurse effect clearly differ from the response of beneficiary species to shifting resource availability. Importantly, the life-history strategies of both the nurse and beneficiary species involved in the interaction can explain the absence of facilitation (Michalet 2007; Maestre et al. 2009). Michalet (2007) proposed that competition is likely to be important in water-stressed systems when the life-history strategy includes a water-demanding tree or shrub interacting with a dominant grass, as shown by Davis et al. (1998) in the tall-grass prairie for oak species competing with exploitative grasses, and by Maestre & Cortina (2004) for the shrub Pistacia lentiscus competing with the grass Stipa tenacissima. Similarly, Armas & Pugnaire (2005) showed in southern semi-arid Spain that Stipa tenacissima had negative effects on the dwarf shrub Cistus clusii during periods of water shortage. In summary, this alternative hypothesis adequately explains decreases and even absence of facilitation when water limitations increase to the point of competition.

One may argue that these contrasting patterns of interactions may also be due to the method used to measure plant–plant interactions, i.e. the importance vs the intensity of interactions (Brooker et al. 2005). Interaction importance may always converge to zero at extreme stress, because any type of positive or negative interaction will be overwhelmed through the effects of stress, while interaction intensity may still be positive or shift to negative at extreme stress conditions. However, several authors have also evidenced a collapse of the intensity of interactions (e.g. Forey et al. 2010; Le Bagousse-Pinguet et al. 2013b).

Theoretical and applied implications

The two alternative hypotheses to explain absence of facilitation at extreme ends of gradients, where the SGH has generally applied, are important to decouple as they concern very different community capacities to maintain species and respond to climate change. Decoupling effect from response in facilitative studies may help in understanding changes in patterns of plant interactions at the edge of life, as described in the competition literature over 20 yr ago (Gaudet & Keddy 1988; Goldberg 1990), and since revisited (Liancourt et al. 2009; Violle et al. 2009). An indirect way to distinguish effect from response in facilitative studies is to measure variation in nurse traits along gradients, together with classical measurements of target species responses (Le Bagousse-Pinguet et al. 2013b; Schöb et al. 2013). Unfortunately, facilitative effects are difficult to measure in common gardens because of their dependence on adult nurse size and on the environmental stresses acting in natural conditions, and are thus less amenable to this approach common in competition studies (e.g. Novoplansky & Goldberg 2001). Nonetheless, there are numerous species of nurse plant and most can occur very widely on environmental gradients, thereby providing adequate opportunities to measure both nurse plant effects and variation in their traits and life-history strategy.

This is an important topic not just for theory development and conceptual refinement. Distinguishing effect from response is important to species extinctions, as climate change may influence each pathway independently and very differently (Michalet et al. 2014). Climate change can impact the nurse, the beneficiary (response) species, or both, and each set of impacts will influence the outcome of interactions very differently. Plant interactions may collapse or alternatively change from positive to negative. For example, Schöb et al. (2013) have shown that facilitation collapsed towards lower elevations in Mediterranean alpine communities with increasing drought stress because of the decreased facilitative effect of the drought-intolerant alpine cushion species Arenaria tetraquetra. The predicted increase in summer drought due to climate change in the Mediterranean will thus push the collapse of facilitation to higher elevations and accelerate species loss at the low altitudinal range limit of beneficiary species (Michalet et al. 2014). During the European 2003 heat-wave, however, Saccone et al. (2009) showed that adult trees from a temperate deciduous forest increased water competition for the seedlings of soil water-demanding tree species, but this is not likely to lead to species extinctions in the short term at least. Hence, climate change may influence the loss of positive interactions at the edge of life via different pathways, and one pathway will likely lead to higher extinction rates than the other.

Global change in general influences both stress and disturbance levels, and these two drivers often co-vary (Forey et al. 2010; Maalouf et al. 2012a). Environmental stress and disturbance may thus interact to accelerate the critical transitions in ecosystem states through changes in plant interactions (Kéfi et al. 2007; Verwijmeren et al. 2013), with significant implications for diversity and the maintenance/extinction of plant communities (Michalet et al. 2006; Kéfi et al. 2007; Maalouf et al. 2012a; Le Bagousse-Pinguet et al. 2013a; Verwijmeren et al. 2013). To conclude, global change will undoubtedly generate complex variation in plant–plant interactions resulting from changes in effect, response, or even both. To improve our predictions of the mediating role of plant–plant interactions for species responses to climate change, we need to clearly disentangle changes in effects from those of responses, and also begin to explore the potential for synergistic interactions.


J.P.M. was financially supported by the French ANR 09 - STRA - 09 O2LA and Y.L.B.P by the project Post-Doc USB ( CZ.1.07/2.3.00/30.0006) realized through EU Education for Competitiveness Operational Program. This latter project is funded by European Social Fund and Czech State Budget.