The evolution of life-history theory: a bibliometric analysis of an interdisciplinary research area

The term ‘life-history theory’ is a familiar label in several disciplines. Life-history theory has its roots in evolutionary models of the fitness consequences of allocating energy to reproduction, growth and self-maintenance across the life course. Increasingly, the term is also used in the conceptual framing of psychological and social-science studies. As a scientific paradigm expands its range, its parts can become conceptually isolated from one another, even to the point that it is no longer held together by a common core of shared ideas. Here, we investigate the literature invoking the term ‘life-history theory’ using quantitative bibliometric methods based on patterns of shared citation. Results show that the literature up to and including 2010 was relatively coherent: it drew on a shared body of core references and had only weak cluster divisions running along taxonomic lines. The post-2010 literature is more fragmented: it has more marked cluster boundaries, both between the human and non-human literatures, and within the human literature. In particular, two clusters of human research based on the idea of a fast–slow continuum of individual differences are bibliometrically isolated from the rest. We also find some evidence suggesting a decline over time in the incidence of formal modelling. We point out that the human fast–slow continuum literature is conceptually closer to the non-human ‘pace-of-life’ literature than it is to the formal life-history framework in ecology and evolution.


Section 1. 'Life history theory' and alternative search terms
We based our study on the topic search term 'life history theory' because we were interested in the emergence and use of that particular label. However, it is clear that this search term does not capture all of the literature on life history evolution. Indeed, many works commonly thought of as establishing canonical 'life history theory' ideas did not in fact use that term and so are not captured by the search (e.g. [1,2]). Stearns [3] estimated that there had been 52 papers on life history evolution by 1980, whereas our search captured only one [4]. Our strategy therefore under-samples the early theoretical work on which life history theory is based. Our study is therefore best thought of as a study of the label 'life history theory', rather than a review of the whole literature that has contributed to our understanding of the evolution of life history traits.
We did, however, also explore a broader search strategy designed to capture more of the literature on life history evolution. We searched Web of Science for '("life history theory" OR "life history evolution" OR "life history strategies")'. This search was performed on February 21 st 2019, several months later than the searches reported in the main paper. This search returned 7185 documents, almost four times as many as the 'life history theory' search. We also constructed maps of this broader literature, again dividing into two time periods, up to and including 2010, and post-2010. These maps are shown in figure S1.
The broader search appears to return a much greater range of research on plants, invertebrate animals, reptiles, amphibians and fishes than the 'life history search'. Moreover, research on humans constitutes a smaller proportion of the data. This suggests that the label 'life history theory', as opposed to 'life history evolution' or 'life history strategies', has become particularly widespread in research on humans, whereas related ideas appear in other parts of biology under slightly different labels.
The map up until 2010 is basically radial in structure, though its 'arms' are longer than in the 'life history theory' search, whilst the post-2010 map is more linear in structure. This shift over time echoes figure 2 of the main paper. Once again we see human research concentrated at one end of the continuum and non-human work at the other, though the human research constitutes a smaller proportion of the total on this broader search than for 'life history theory'. There are seven clusters in both time periods for this broader search. Establishing what these contain would be a full study in its own right. Some of them are taxonomically based: in both time periods there appears to be a cluster of work on ocean fish and fisheries management (dark blue); and one on freshwater fish ecology and evolution (light blue). Our impression is that that in this broader search, more of the clusters are genuinely taxonomically mixed. For example, the research in the purple clusters in both maps spans multiple taxa. However, in both time periods the research on humans is concentrated into a single cluster, shown in green. The three distinct clusters of human research of figure 2B in the main paper do not separate here, though the relative positions of different styles of human research does correspond to that figure (i.e. more 'psychological' research is generally positioned further away from the non-human clusters). A B Figure S1. Bibliographic coupling maps of the broader literature search '"life history theory" OR "life history evolution" OR "life history strategies", using the same parameter values as the maps in figure  Although the results of the broader search are different in detail from those of the 'life history theory' search, they do support the notion of an increasingly linear literature, with 'life history' research on humans sequestered at one end and not strongly connected to corresponding work on non-human organisms.
We also searched Web of Science for 'pace of life' (see Discussion of main paper). This returned 407 hits, all only 22 were also in the 'life history theory' set. The papers in the 'pace of life' set tend to draw on the 'fast-slow continuum' and related ideas. For example, 123 of the 407 mention 'fast' in their title or abstract, and 113 mention 'slow'. As we argue in the main paper Discussion, the research programme of clusters B2 and B5 more logically belongs with the non-human 'pace of life' literature than it does with the literature on life history theory.

Section 2. Mapping the whole of the literature
For our analysis in the main paper, we divided the literature into two time periods, up to and including 2010, and post-2010. If instead we use the whole of the 'life history theory' literature for mapping, we obtain a map with a similar general shape to the post-2010 map in the main paper (figure S2). As in the main paper, the human research is concentrated along one end of the line, with the 'dark triad' research most distant from the rest of the literature. This analysis produces four clusters using the same cluster resolution as in the main paper. These clusters appear to predominantly correspond to: birds and mammals (red); fish, insects and other taxa (blue); human evolution (dark green); and human developmental/personality psychology (mid green). The human developmental/personality psychology cluster amalgamates clusters B2 and B5 of the post-2010 map from the main paper: spatially, these two research areas still separate, but they are not assigned separate clusters in this overall analysis.

Section 3. Finer-grained division of time
As well as the binary division of time used in the main paper, we also divided the 'life history theory' data into four time categories containing roughly equal numbers of records: up to and including 2004 (n = 458); 2005-2010 (n = 453); 2011-2014 (n = 423); and 2015-2018 (n = 508). Figure S3 shows bibliographic coupling maps of each of these four time bins using the same parameter values as the maps in the main paper. It also shows the increasing variegation of the human research: in the first and second time periods, there is only one cluster of human research. In the third time period, the 'dark triad' cluster separates from the human development cluster, although the two are combined again in the fourth time period. The division between the human evolution and human developmental/personality psychology is only detected by the clustering algorithm in the most recent time period. However, although not yet marked enough to be detected by the clustering algorithm, the difference between the three kinds of human life history theory research on humans existed from the very earliest time period. For example, in the period prior to 1995, papers by Helle et al. [7], Chisholm [8] and Rushton [9] respectively already exemplified the styles of research that would later become the three human clusters. The relative positions of these three types of research on the map, and their relative distance from non-human research, are completely consistent across the time periods.

Section 4. Co-citation maps
An alternative to the bibliographic coupling method we use in the main paper is co-citation analysis.
Here, the nodes on the map are the papers cited by the papers found in the literature search. The links are formed by being cited by the same source. Co-citation analysis should produce similar representations of the structure of the field as bibliographic coupling [10]. We repeated our mapping exercise from the main paper using co-citation analysis instead of bibliographic coupling. We used a minimum number of citations of eight for a node to be included. We also set the cluster resolution parameter to 1.0 instead of 0.8 as for the bibliographic coupling: the number of clusters found at any given resolution was fewer in the co-citation analysis. Other parameter values were as for the bibliographic coupling. The resulting maps were as shown in figure S4.
The same basic generalizations hold as for the bibliographic coupling maps, especially the increasing linearity of structure of the map. In the earlier time period, five clusters were detected; as in the main paper, all the human research was in one of these. In the later time period, the clustering identified six rather than five clusters. Three of these were identical to the bibliographic coupling clusters (the three human clusters, B2, B4 and B5). The non-human literature was divided into three clusters here rather than the two in the bibliographic coupling map.

Section 5. Varying the cluster resolution parameter
The cluster detection is controlled by a resolution parameter (with higher resolution representing greater sensitivity to clustering). We chose 0.80 for the main analyses as this gives a reasonable number of interpretable clusters in both time periods. This was a non-pre-registered researcher decision not based on any a priori rationale. We also experimented with other cluster resolution values (table S1). The maximum cluster resolution that can be used before distinct clusters were detected was different in the two time periods: 0.26 in the literature up to 2010, but only 0.15 in the more recent period. This, along with the other evidence discussed in the main paper, suggests a greater degree of internal separation in the more recent period. In both time periods, the first cluster division to appear was 'human' versus 'non-human'. At a cluster resolution of 0.60, again in both time periods, the 'non-human' cluster split into one focussed on birds and mammals, and another focussed on fish, invertebrates, reptiles, plants, etc. The non-human clusters further divided more finely as the resolution was increased beyond 0.80.
We also examined the maximum cluster resolution that can be used before distinct clusters are detected in the four finer-grained time categories detailed in section 3 of this document. The Section 7. Information on the clusters (tables S1 and S2)