1 Links between soil ecology, general ecology, modelling and evolutionary approaches

There are low percentages of cross-citations between soil ecology journals and journals specialized in modelling and evolutionary approaches. Moreover, soil ecology journals publish low percentages of articles based on modelling and evolutionary approaches and, conversely, journals specialized in evolutionary and modelling approaches publish low percentages of soil ecology-related articles. Even inside the journals non-specialized in soil ecology, articles related to soil ecology use less often models and evolutionary approaches than the other articles published in the same journals. Although this does not replace a much wider bibliographic study comparing different fields of ecology (soil ecology, marine ecology, bird ecology … etc), this suggests that soil ecology uses modelling and evolutionary thinking more parsimoniously than other fields of ecology. Besides, soil ecology is not well represented in generalist ecology journals. Finally, SBB (and probably other soil ecology journals) is little cited by or does not cite much generalist ecology journals, theoretical ecology journals and evolutionary ecology journals, while cross-citation could be a way to make up for the lack of modelling and evolutionary orientated articles published in soil ecology journals. Reciprocally, we could have expected generalist journals to cite more often specialized journals such as soil ecology journals as a source of patterns to be interpreted and data to test their general theories. All these results confirm the relative independence between soil ecology and general ecology and the five hypotheses presented in the introduction.

It must be marked that the trend described using journal articles is confirmed by the examination of text books on soil ecology that hardly mention explicitly evolutionary issues and model results [14], [15]. Meanwhile, at least some text books on plant ecology [16], [17] and insect ecology [18], [19] or aquatic ecosystems [20], [21] largely refer to the advances made using models and evolutionary thinking.

In the following discussions, we try to interpret these results and more particularly the relative independence between soil ecology and general ecology, and the parsimonious use of modelling and evolutionary interpretations made by soil ecology. We first propose two explanations that could explain some of our results and show that they are not valid or do not explain the whole bibliographic pattern we have described. A first explanation would be that soil ecologists have had to face more technical problems than other ecologists (section 2). That would have impeded them to develop their field as fast as in other fields of ecology and would have slowed down efforts of modelling and evolutionary questioning. We argue that this explanation contains some truth but is not sufficient to explain the whole bibliographic pattern we have described. A second explanation would be that generalist journals, with higher impact factors, necessarily publish higher percentages of model- or evolutionary-based studies because modelling and evolutionary approaches lead to more general results (section 3). We show that this explanation does not hold. We then propose that the scarcity of modelling and evolutionary approaches in soil ecology is an important proximal cause of the relative independence of soil ecology from general ecology (section 4). In this section we thus detail the reasons while soil ecology would highly benefit from more conceptualisation and evolutionary thinking. We then try to propose some historical reasons for the whole bibliographic pattern we have described and particularly for the scarcity of evolutionary and modelling approaches in soil ecology (section 5). We finally conclude by some remarks on the role of the publication system on the development of soil ecology and suggest that this system might hamper interdisciplinary thinking and the building of links between general theories and specific empirical studies (section 6).

2 Technical difficulties, so what?

A first explanation for the relative independence between soil ecology and general ecology would be that soil ecology is intrinsically difficult to study due to the following points: (1) Soil is a black box. It is more difficult to manipulate and observe soilo rganisms without disturbing their environment than above-ground organisms. (2) Soil is a very complex environment in which it is difficult to disentangle biotic and non-biotic interactions. (3) These interactions involve a solid phase, an aquatic phase, a gaseous phase and complex exchanges between these phases. (4) Soils are extremely heterogeneous at all special scales [from the micron to the kilometre, 22]. (5) Soil processes depend directly on a huge variety of organisms, whose size is typically low and whose taxonomy and diversity are poorly known in comparison to aboveground organisms. (6) Soil processes involve a huge variety of organic molecules and chemical reorganizations that are poorly understood. Possibly, soil ecologists have, for the moment, focussed on solving their technical problems and developing methods to investigate soil processes. This could contribute to explain that they are centred on their own discipline and less open to general ecology. For example, the development of investigation methods has involved the development of molecular techniques in soil microbiology.

The technical difficulties could thus partially explain why soil ecology makes a parsimonious use of models and evolutionary rationales and why soil ecology has developed partially independently of general ecology. However, these difficulties have not impeded the production of a huge empirical corpus and technical problems are progressively overcome by new methods based, for example, on molecular biology [23], or stable isotope marking [24]. Besides, a huge number of soil ecology articles use already validated methods and standard protocols. This suggests that technical difficulties are not the only reason why soil ecologists parsimoniously use models and evolutionary interpretations.

3 Generalist journals should not publish more model and evolutionary oriented articles than specialized journals

To explain that generalist journals publish more studies based on modelling than soil ecology journals, it could be simply argued that, by definition, generalist journals with high IF publish studies addressed to a wider readership than specialized journals and that modelling help obtaining general arguments that are likely to attract a wide readership. This is probably true (see below) but does not explain the whole bibliographic pattern. First, we have also found that, inside journals not-specialized in soil ecology, soil ecology articles are less based on modelling and evolutionary approaches than the articles of other field of ecology. Second, this explanation assimilates implicitly models to theoretical models that indeed aim at testing and building general theories. In fact, they are many kinds of models that can be classified according to the degree of generality, realism and precision of the predictions [25]. To be general a model must take into account few fundamental mechanisms but remains a non-fully realistic idealization and cannot give quantitatively precise predictions because it has not been fitted to any particular system. Conversely, such predictions require taking into account more mechanisms and more particularities of the modelled ecological systems so that models giving precise predictions cannot be general [25]. Recent analyses of Levin's classical article support the existence of tradeoffs between the different type of models he has described [26], [27].

Clearly, building general theoretical models is useful but it is also imperative to build models focussing on particular systems. We thus suggest that journals specialized in soil ecology should publish more studies based on models that apply to soil systems. These studies should encompass theoretical models applying to soil systems in general (to build general theories on soil ecology, e.g. a general model showing the implications for decomposition of the existence of two pools of organic matter with a different degree of recalcitrance) or models built to answer specific questions on particular soil systems (a model that predict the decomposition rate of soil organic matter in a given site as a function of climatic variations). Meanwhile, generalist journals could publish a theoretical model studying the effect of decomposition rate on primary production if its conclusion can be applied to a wide class of systems (for example both on soil and aquatic systems). We want to emphasize that it is only by applying different modelling approaches to study the same issue that the robustness [25] and relevance of modelling results can be assessed. In particular, while general and theoretical models are efficient to suggest broad theories to be tested, more realistic and precise models are more efficient to compare model outputs to empirical observations.

The low percentage of articles linked to evolutionary issues published in soil ecology journals can be interpreted in the same way as the low percentage of articles using models published in these journals: evolutionary interpretations help to reach general conclusions that could be applied to many systems. The parallel drawn here between the way models and evolutionary interpretations can help building general theories is confirmed by the high percentages of models published in journals specialized in evolution and the high percentages of evolution-related studies published in journals specialized in modelling (with the exception of Ecological Modelling). Indeed, some evolutionary mechanisms and issues are very general and results about them should be published in generalist journals (especially when they are studied using theoretical models). However, other evolutionary mechanisms and issues only concern particular ecological systems. Once again, soil ecology journals should publish articles involving evolutionary rationales that are particular to soil systems or studying how general evolutionary questions translate into soil ecology. For example, generalist journals should publish general results on the way dispersal ability and other life-history traits coevolve while soil ecology journals should publish studies on the consequences of the limited dispersal ability of many soil organisms on the evolution of their life-history.

4 Proximal explanation of the relative independent development of soil ecology and general ecology

An explanation would be that soil ecologists tend to present their results in such a way that they are poorly linked to general theories of ecology, i.e. they rarely interpret their results in the lights of widely accepted ecological theories such as food-web, competition, coexistence or evolutionary theories. Their articles would thus be more difficult to publish in generalist journals and would be mostly cited by other soil ecologists (as confirmed by our analysis on SBB), which would at last result in lower numbers of citations and specialized journal with low IF. According to this explanation the lack of modelling and evolutionary interpretation would be one of the causes of the weak connection between general ecology and soil ecology. We detail below how soil ecology would benefit from an effort of conceptualisation and evolutionary thinking.

Modelling is an integral part of natural sciences in the sense that conceptual verbal models are already models. Such models are necessary to sort out hypotheses and make clear statements about our mental representation of nature and the links between observed patterns and mechanisms [28], [29]. Then, mathematical models are necessary when the consequences of hypothesized processes cannot be predicted verbally which arises as soon as the studied systems become complex (more compartments, more interactions, retroactions, space is explicitly taken into account). Moreover, without prejudging the existence of general laws in ecology [30], [31], conceptual verbal models and results of mathematical models sum up our ecological knowledge. The model results that are widely accepted at a given time often constitute a paradigm in the Kuhnian acceptation [32]. Taken together, models, either mathematical or verbal, are necessary both to predict the consequences of newly discovered processes and to provide new hypotheses to be tested empirically.

In this way, soil ecologists should more often take advantage of modelling results generated in other fields of ecology to build new hypotheses suggesting in turn new experiments. Conversely, they should benefit from constructing more often their own models to explain their empirical results. An effort of conceptualisation should help soil ecology to build its own paradigm and thus to become more visible and independent. It should also help soil ecology to build more bridges with other fields of ecology because models are easier to compare than experimental data gathered on very different systems. These two trends, independence and integration, are both necessary to foster progress in a given scientific field [33], [34].

An effort of conceptualisation in soil ecology will also lead to the development of new theories on aspects of ecology often overlooked by general ecology. For example, relations between soil organisms often involve ecosystem engineering activities [35], exchanges of signal molecules [see for example 36], and relations mediated through the recycling of nutrients [37], [38]. Such relations have so far been poorly taken into account by classical food web models [39]. Their consequences on population dynamics, community stability and evolution are poorly understood. Moreover, many soil processes are based on the activity of microbes. Linking the population dynamics of microbes, their plasticity and their capacity to evolve to soil and whole ecosystem properties is a new field of investigation that requires new models linking population dynamics, functional ecology and evolution [40]–[42]. We of course do not wish to suggest that there has not been any effort of conceptualisation in soil ecology. For example, Clarholm's work on the microbial loop [43], without the help of any mathematical model, has a long lasting influence on soil ecology researches. Setälä's work on decomposer biodiversity [44] is also based on a strong conceptual base and many other good examples can be found [see for example in 45]. However, we think that these efforts of conceptualisation should be reinforced and become a habit of all soil ecologists.

Although describing the proximate mechanisms involved in the dynamics of populations and ecosystems is a challenging task, it is primordial to take into account the fact that organisms are the result of a long evolution process. The biological traits that determine the nature of interactions between organisms have also evolved. As models, evolutionary arguments can interplay bidirectionally with empirical results, helping to interpret empirical results and suggesting new experiments. Thus, besides studying evolution for its own sake, studying the evolution of soil organisms should be scientifically beneficial for three reasons. (i) Interpretation of empirical results must be consistent with evolutionary knowledge. For example, when new ecological interactions or new biological traits are pointed out, their ecological significance must take into account the fact that these traits have evolved and thus that they should benefit to their owners or that their evolution is linked to a constraint. It is relevant to identify the nature of this benefit or constraint. (ii) Conversely, evolutionary theories and new evolutionary models applied to soil issues are likely to lead to predictions on the way biological organisms have evolved and thus on the biological traits of present soil organisms and the type of interactions that link them. Such predictions can be used as guides to design new experiments. (iii) Finally, it is more and more recognized that evolution is often quicker than formerly believed [46], and thus that some temporal patterns observed on the human time scale could be due to evolutionary processes. This should particularly be the case for soil processes because they depend on short-lived organisms with a high potential for rapid adaptation such as bacteria or protozoa [6], [40].

All these explanations suggest that developing a theoretical and evolutionary framework for soil ecology should benefit grandly soil ecology and ecology in general as already mentioned [7]–[9]. What are the historical reasons of the absence of such a framework and the scarcity of links between soil ecology and general ecology?

5 Historical explanations

Soil ecology is historically much more based on the ecosystem paradigm of ecology than on its population paradigm [34]. One reason might be linked to the close dependence of soil organisms on chemical and physical constraints [14] and the importance of physical and chemical processes in soils. This is confirmed by a study on trend words in ecological journals which shows that “below-ground” is associated to the abiotic pole of ecology [47]. This is also confirmed by the quantitative importance of articles published in journals emphasizing chemical and physical soil processes among the articles cited by and citing SBB (Fig. 2). The close link between ecosystem ecology and soil ecology can explain partially the poor use made by soil ecologists of mathematical models and evolutionary knowledge since population ecology has always been more theory- and evolution-orientated. In the same vein, soil ecology is historically linked to agronomy [48] as confirmed by the high number of citations between SBB and agronomy journals (Fig. 2). This contributes to explain the relative independent development of general ecology and soil ecology [48], [49]. Moreover, agronomy has so far mostly aimed at gains in production and is consequently poorly evolution-oriented.

6 Partial responsibility of the publication system

We finally acknowledge the fact that our approach, i.e. studying scientific thinking through counts of articles published in different journals using very broad thematic categories, is slightly naïve and simplistic. However, all scientists know that choosing the journal to submit a manuscript is of paramount importance to get published and to get a wide readership. Whether we want it or not, in an academic scientific world where we have to “publish or perish”, journals influence greatly the way we work and probably the way we think. We suggest three general drawbacks of the classification of journals according to their degree of generality and their Impact Factor.

First, this classification might deter ecologists who are specialists of a given field of ecology from reading and quoting more theoretical works or results from other fields of ecology. Conversely, it might deter more generalist ecologists and theoreticians to read and quote more specialized studies. This classification opposes “particularity” and “generality” while our goal, as scientists, should be to derive as many links as possible between these two poles [34]. Of course, generalist journals could be the place where “particularity” and “generality” should be confronted but we have shown that it does not work efficiently, at least in soil ecology. Of course, many bibliographic databases are generalist and interrogating them with a given keyword yields articles published in many journals whatever their degree of generality and their Impact Factor. However, the superabundance of published papers impeding to read exhaustively the relevant literature, the temptation is great to focus on a handful of journals. For example, soil ecologists might already have difficulties to read all relevant articles published in specialized journals and might not have the time to read articles published in other journals. If, by chance, they do so they are not likely to focus on theoretical and evolutionary articles because it is not in the “style” of the specialized journals to which they intend to submit their own work to quote such studies.

Second, we know that to publish in a journal it is important to submit to the “style of the journal”. Journals delimit their respective scopes and styles (through intentional choices, self-organization and competition) so that it might be difficult to publish articles that do not fall within one of the categories journals have created. It might for example be difficult to publish models in soil ecology journals or soil ecology studies in generalist journals because soil ecology has historically developed partially independently from general ecology and theory. In other words, scientific structuring by traditional journals is likely to slow down changes in the structure of sciences and limit interdisciplinary studies. This might be particularly detrimental for the development of ecology which is by nature interdisciplinary because it aims at linking (1) the physical and biological worlds, (2) natural and human sciences (because mankind is one of the most influential component of the biosphere and to develop applications), (3) many different scales (from genes and molecules to atmospheric circulation).

Third, the classification of journals according to a gradient of generality, as denoted by their Impact Factors, and the fact that these Impact Factors are often considered as an index of intrinsic scientific quality might be counterproductive. The Impact Factor of a journal has been demonstrated to be a poor predictor of the number of citations an article will finally get [50]. Indeed, reviewers and editors are able to control the scientific soundness of a study, the relevance of the protocols and the consistency between results and conclusions. It is much more difficult to predict a priori the level of generality of the study and its real usefulness.

Taken together, our bibliographic study and these last paragraphs emphasize the utility of generalist interdisciplinary journals, such as PLoS ONE, that select their articles on a technical base and not because they fit to the journal style, approach, subject and supposed level of generality. Such journals are likely to follow scientific progresses in a quick and flexible way and foster more connections between disciplines. In such a publication system articles will no longer be ranked according to the Impact Factor of their journals but according to the number of citations they really get. Articles will thus be ranked according to their “usefulness” and not according to an a priori level of generality. We finally think that the development of this type of journal should be especially profitable for soil ecology that has developed, as we have shown, too independently from the rest of ecology, theoretical ecology and evolutionary ecology.