Evolutionary Biology

, Volume 35, Issue 2, pp 97–99

From Marshalling Yards to Landscapes to Triangles to Morphospace


    • Department of BiologyDalhousie University

DOI: 10.1007/s11692-008-9021-z

Cite this article as:
Hall, B.K. Evol Biol (2008) 35: 97. doi:10.1007/s11692-008-9021-z

Metaphors have long played a vital role in channeling our thinking about organismal structure, development, evolution, function, and behavior, but especially, how the genotype maps to the phenotype and how the role of the environment can (or should) be visualized as influencing development and evolution. Great chains of being, trees of life, machines, programs, progress, all conjure up images of organisms, and organismal relationships that are not necessarily congruent and that may bear more or less similarity to the organisms or organismal traits they seek to portray. Morphospace is another metaphor. Landscapes, triangles, and marshalling yards are three others.

Marshalling yards are the oldest, first used by Joseph Needham with his analogy of decision-making in the early totipotent egg as a freight wagon moving down a sloping railway track, in this case into a marshalling yard of the London & North Eastern Railway, shown as an addendum figure in his 1936 book Order and Life (Needham 1936). The tracks bifurcate. The track taken by the wagon is set by an external switching system. So the passage of an animal through development is seen as a series of successive responses to changing signals, each response involving a binary choice; go left or right for the freight wagon, become mesoderm or ectoderm, neural tube or neural crest, cartilage or muscle for the cell(s).

Waddington turned the marshalling yard into a landscape, initially as a valley as the frontispiece to his book The Epigenetics of Birds (Waddington 1952) painted by his close friend John Piper, then as the famous and often reproduced epigenetic landscape (Fig. 1). The egg or organism is a ball rolling down a bifurcating valley. It is hard to move the ball up the sloping walls of the valley, but once the ridge is crossed, the ball (egg) moves into a new valley (the egg takes a different developmental trajectory).
Fig. 1

The classic depiction of the epigenetic landscape from Waddington (1957)

The importance of the environment in marshalling yards and landscapes (signals within the egg or outside the organism) is paramount. Evolution was not so readily incorporated into the landscape metaphor, although genetics was: many interpreted Wright’s (1968) adaptive landscape as the underside of Waddington’s phenotypic landscape. Had Waddington used his epigenetic landscape as a metaphor for evolutionary change—which he did not—the bifurcations of each valley would have represented nodes on a cladogram. But Waddington was aware that evolution was much more subtle than the responses of organisms to binary switches. Evolution was epigenetic. Interestingly, the latest version of the epigenetic landscape is a pinball machine as a metaphor for the epigenetic molecular machinery involved in decision-making within cells. It is figured in the leading edge essay by Goldberg et al. (2007) introducing a series of essays in Cell devoted to epigenetics, which in the “modern” usage is defined as “a change in phenotype that is heritable but does not involve DNA mutation…the change in phenotype must be switch-like..,” (Shi 2007, p. 639) or “the study of any potentially stable, and ideally heritable change in gene expression or cellular phenotype that occurs without changes in Watson–Crick base-pairing of DNA” (Goldberg et al. 2007, p. 635).1

Seilacher (1970) used a triangle onto which he plotted morphology, the three sides of the triangle representing environmental influences (as reflected in ecological adaptation), phenotype (rules of fabrication/morphogenesis), and phylogenetic legacy (evolutionary history). Seilacher (1979) used his ‘triangle’ as a predictive tool in assessing the range of possible morphologies in sand dollars and the major processes contributing to those morphologies.

Landscapes, marshalling yards, and Seilacher’s triangles explicitly include the environment as playing a central role in directing development. Landscapes and marshalling yards reflect dynamical processes and development (or evolution) through time; embryos are balls rolling down a bifurcating valley or trains whose destinations are set by external signals. Triangles give environment, rules of development, and evolutionary history equal weight.

Morphospace, on the other hand, has primarily been used to visualize the end result of processes by plotting elements of the phenotype in three-dimensional space. We owe the popularization of morphospace to the palaeontologist David Raup who plotted aspects of the phenotype of related organisms in three-dimensional space (morphospace). Each axis represented an element of morphology, allowing visualization of whether phenotypic variation was unbounded or limited (constrained). Plot extant and extinct taxa and you could follow phenotypic changes through the history of a group. When Raup (1966) did this for four groups of shelled invertebrates—gastropods, brachiopods, ammonites, pelecypods—much of theoretical morphospace was unoccupied. The obvious conclusion—morphological variation is limited. The extrapolation to theory—morphology is constrained. Why? Because of physical limits, modes of growth, adaptation to environments.

In the paper to which this is a commentary David Polly treats morphospace as landscape, an aim summed up in the question that forms the title of his paper (Polly 2008): Can morphometric spaces be used to model phenotypic evolution? Conrad Waddington who gave us the epigenetic landscape, would have answered with a resounding yes; “The evolution of organisms must really be regarded as the evolution of developmental systems” (Waddington 1975, p. 7).

Polly is concerned with bridging approaches that plot the phenotype as a gradual continuum (morphospace as extrapolated trajectories) and those that allow discontinuity, even novelty, to be visualized. The cladistic analysis of Burgess Shale arthropods by Briggs et al. (1992) supports the existence of many body plans forming a continuum at the early stages of metazoan evolution. Subsequent loss of groups leaves what now appear as intermediates in the fossil record, but that are (or were) in reality, parts of a continuum. Polly seeks a phenotypic landscape that will enable adaptive landscapes to be related to phenotypes, all the time being conscious that the phenotype does not equate to a one-to-one readout of the genotype, and certainly not to a simple genotype-environment map.

The theoretical basis for morphospace is the evolutionary quantitative genetic theory developed for allele frequency in populations, one of the basic assumptions of which is that phenotypic change is continuous. But evo-devo tells us (discussed by Polly) as does QTL analysis (also discussed by Polly) that change can be discontinuous, canalized (as visualized in the epigenetic landscape and in the concept of chreods (a canalized developmental sequence or developmental trajectory) introduced by Waddington (1961). Developmental interactions can limit the extent of phenotypic variation/change. The genotype may set a range of phenotypic potentials, the actual phenotype reflecting responses of the organisms to environmental signals; the number of caudal vertebrae in zebrafish (Danio rerio) can range from 16 to 18, the number that form within that range is determined by, among other factors, the temperature at which embryos develop when somites are being established (Connolly and Hall 2008). Neither genetic, developmental nor environmental factors (or temporal and spatial interactions between any two or all) map simply to the phenotype. Even a continuous distribution in a chemical (morphogen) can result in a discontinuous phenotype if the responding cells “see” only thresholds of concentrations and not continuous changes in concentration—formation of a discrete number of digits in response to a continuous gradient in concentration in Sonic hedgehog across a limb bud, for example. Developmental processes proceed along only a limited number of pathways, deviations from which are controlled genetically and developmentally by threshold reactions. Polly’s paper reviews and analyzes how the phenotypic landscape provides the middle ground between the adaptive landscape and morphospace. In showing that the phenotypic landscape can visualize processes underlying the maintenance and origin of phenotypes, Polly demonstrates a means to explore “the evolution of organisms…as the evolution of developmental systems.”


This is not a new metaphor. At Waddington’s 50 birthday party, held at the Genetics Institute at Edinburgh University in 1955, a pinball machine was used to represent the epigenetic landscape.


Copyright information

© Springer-Verlag 2008