Moving and sensing without input and output: early nervous systems and the origins of the animal sensorimotor organization

It remains a standing problem how and why the first nervous systems evolved. Molecular and genomic information is now rapidly accumulating but the macroscopic organization and functioning of early nervous systems remains unclear. To explore potential evolutionary options, a coordination centered view is discussed that diverges from a standard input–output view on early nervous systems. The scenario involved, the skin brain thesis (SBT), stresses the need to coordinate muscle-based motility at a very early stage. This paper addresses how this scenario with its focus on coordination also deals with sensory aspects. It will be argued that the neural structure required to coordinate extensive sheets of contractile tissue for motility provides the starting point for a new multicellular organized form of sensing. Moving a body by muscle contraction provides the basis for a multicellular organization that is sensitive to external surface structure at the scale of the animal body. Instead of thinking about early nervous systems as being connected to the environment merely through input and output, the implication developed here is that early nervous systems provide the foundation for a highly specific animal sensorimotor organization in which neural activity directly reflects bodily and environmental spatiotemporal structure. While the SBT diverges from the input–output view, it is closely linked to and supported by ongoing work on embodied approaches to intelligence to which it adds a new interpretation of animal embodiment and sensorimotor organization.

Keywords: Early nervous systems, Skin brain thesis, Sensorimotor organization, Embodiment, Input–output view, Coordination view

It remains a standing problem how and why the first nervous systems evolved. From the nineteenth century onward, scientists proposed various evolutionary scenarios but the issue was never resolved. Nowadays, molecular and in particular genomic data have made large inroads into the phylogenetic relations between different animal lineages but still no clear evolutionary sequence for the origins of the first nervous systems has emerged.1 Instead, it has become a debate whether neurons, nerve nets and central nervous systems evolved several times independently in separate lineages (Moroz 2009; Northcutt 2012).

An important way to make progress here is by improving our scenarios of how very early nervous systems may have functioned. Sophisticated scenario building should address the proximate and macroscopic operation of plausible early nervous system configurations as well as of protonervous systems that span the gap between nervous systems proper and other tissues. So far, almost all proposals targeted a too limited domain of functional options. Our appreciation of the functional options available for early nervous systems can be improved by drawing a contrast between two broad interpretations of primitive functioning. The first interpretation sees nervous systems primarily as connecting devices that link sensors to effectors. In this case, the simplest nervous system that is conceptually possible consists of a single neuron that connects a sensor to an effector (Parker 1919; Mackie 1990; Lichtneckert and Reichert 2007; Miller 2009, Jékely 2011). This interpretation that will be referred to as the input–output view is closely linked to the standard interpretation for modern nervous systems (e.g. Braitenberg 1984; Greenspan 2007). A contrasting interpretation, the coordination view, sees early nervous systems primarily as coordination devices enabling motility by increasingly large multicellular organisms (Pantin 1956; Passano 1963; Keijzer et al. 2013). Conceptually, the simplest nervous system possible here would be a diffuse nerve net consisting of a significant number of neurons spread out over a large portion of the animal’s body, such as can nowadays be seen in Hydra (Grimmelikhuijzen 1985). While both interpretations should be considered as purely hypothetical at present, much more attention has been given to the first, skewing research towards this interpretation, while disregarding the issues raised by the second interpretation.

Both an input–output view and a coordination view can take many different forms. The present paper will develop a particular coordination interpretation in some detail. The view developed here consists of the skin brain thesis2 or SBT (Keijzer et al. 2013). The SBT proposes an evolutionary progression from excitable and contractile epithelia to full nerve nets as increasingly sophisticated ways to coordinate contractile tissues such as muscle. However, as remarked by Monk (2014), the SBT as formulated earlier seems to stress only motility and to neglect sensory aspects that are obviously central to nervous system organization and evolution as well. Thus the specific aim of the present paper will be to sketch how the SBT deals with sensing as well as motility.

According to the SBT, the use of tissue contraction for motility was the initial key feature that provided both an evolutionarily new form of motility for early animals as well as an evolutionary new means to sense extended surface structures in the environment. In this view, tissue contraction provided the basis for a specific animal sensorimotor organization or ASMO. In this account, the evolution of nervous systems and contractile tissues are fully intertwined and together they turned the animal body into the motile and sensing unit that became so evolutionarily successful at the start of the Cambrian.

The ASMO is specific to animals and will be contrasted to the idea that an animal sensorimotor organization at heart similar to a robotic device equipped with various sensors and effectors (e.g. O’Regan and Noë 2001) as implied by the input–output view. The aim is not to criticize such an input–output view, but only to develop the SBT as a viable and attractive alternative view on early nervous system evolution. Obviously, by elaborating the soundness of the SBT while stressing the differences with an input–output view, the result may not be completely even-handed. However, critique is not the aim.

Formulating a coordination-based alternative for the dominant input–output view will expand the conceptual options that are available for making sense of early nervous system evolution. While having such an alternative is a good thing in itself by extending the range of options, it also provides competition that will force defenders of the dominant view to develop their own range of scenarios in more detail. At another level, the coordination view developed below connects more widely with cognitive and neuroscience theorizing. More specifically, the SBT and the ASMO notion are closely related to embodied approaches to cognition (e.g. Van Gelder 1995; Brooks 1999; Keijzer 2001; Pfeifer and Bongard 2007; Thompson 2007; Clark 2008; Chemero 2009; Barrett 2011) as well as sensorimotor approaches to cognition and consciousness (e.g. O’Regan 2011; O’Regan and Noë 2001; Hurley 2001; Chemero 2009; Buhrmann et al. 2013). This embodied and sensorimotor work provides a wider conceptual and empirical backing for the SBT, while the latter provides an outline how these embodied and sensorimotor approaches invite new conceptual and empirical interpretations of nervous systems more generally.

The text has the following structure. The next section introduces the Precambrian setting in which early nervous systems can be assumed to have evolved. The “The input–output view on nervous systems and the animal sensorimotor organization” section sketches the input–output view to which the coordination view will be contrasted. “The skin brain thesis: coordinating contraction-based motility” section sets out the main features of the skin brain thesis as the most recent version of a coordination view. “Pantin patterning instead of output” and “Sensing bodies rather than input” sections provide the argument for an animal sensorimotor organization, ASMO, by sketching respectively how the skin brain proposal provides the ingredients for self-induced motility that is not output, and for a form of sensing that is not input. The concluding section will highlight the main features of the ASMO by contrasting them with those of the input–output view.

What can we say about the context in which the first nervous systems evolved? A first port of call is the famous Cambrian explosion, a comparatively short period lasting from about 530 to 515 Ma (million years ago) during which most of the modern animal phyla first appeared in a recognizable form in the fossil record (Valentine 2004). From here on ‘standard’ animals with brains, multiple appendages, eyes, heads, tails and hard skeletons roamed the Earth. Many innovations in morphology and behavior first appeared during this period, and these could very well have driven the ‘explosion’. For example, Andrew Parker’s (2003) ‘light switch’ theory takes the evolution of the first eyes as the key change that initiated a new predatory life-style for arthropods. This idea can be combined with prey-predator co-evolution, which is indicated by the evolution of shells, burrowing and hard mouth parts (e.g. Vermeij 1989; Conway Morris and Bengtson 1994; Dzik 2007). Ginsburg and Jablonka (2010) suggest that these prey-predator interactions were driven by associative learning together with a centralization of nervous systems, while Trestman (2013) stresses that complex cognitive embodiment enabled tracking the spatial properties and relations within the environment. In all cases, increasingly sophisticated behavior is cast as a key reason behind the Cambrian explosion.

However, all these proposals require the presence of nervous systems and this implies that their evolution must have taken place earlier (e.g. Fedonkin et al. 2007; Brasier 2009; Erwin et al. 2011; Zhang et al. 2014). The lower boundary of the Cambrian itself (541 Ma) is marked by the presence of the trace fossil Treptichnus pedum (Erwin et al. 2011; Landing 1994) which consists of burrowing traces similar to those of modern priapulid worms (Vannier et al. 2010). Preceding periods lack such bioturbation of the sea floor, making the presence of active burrowing animals before the Cambrian unlikely (Fedonkin et al. 2007; Brasier 2009). As active burrowing requires a hydrostatic skeleton, muscle tissue and a nervous system, the logical conclusion is that nervous systems originated before the Cambrian, either in the Ediacaran (635–541 Ma) or even the Cryogenian (850–635 Ma).

Current knowledge about animal life during the Precambrian period is rather tentative. It derives from various sources that do not all converge nicely. First, macroscopic Ediacaran body fossils have been found that are dated from 579 to 541 Ma. Most are difficult to relate to modern animal phyla or to interpret behaviorally. For example, there is an ongoing debate whether the emblematic Dickinsonia did (Gehling et al. 2005) or did not (Brasier and Antcliffe 2008) actively graze and change its position on the substrate, and also whether they can be presumed to have some form of nervous system. In addition Ediacaran trace fossils are found (Droser et al. 2005). Unquestioned examples of millimeter wide surface traces go back to at least 560 Ma (Jensen et al. 2006; Liu et al. 2010). These traces suggest that these organisms remained very small and worm-like, possibly moving by ciliary creeping rather than muscle (Valentine 2007) and possibly without a nervous system until late in the Ediacaran.

Molecular data add to this picture. A lot of the important genetic and molecular ingredients for animal life were already present at a very early stage, often going back to the last common shared ancestor, single celled choanoflagellates (King 2004). For example, key components of the postsynaptic density—itself an important part of the mammalian synapse—were already present in sponges (Sakarya et al. 2007) and some of them even in fungi and choanoflagellates (Ryan and Grant 2009). A similar story can be told for the evolution of sodium channels, also present in choanoflagellates (Liebeskind et al. 2011). In addition, the gene repertoire and genomic organization of cnidarians—such as jellyfish and sea anemones—is already remarkably complex and matches many human sequences (Putnam et al. 2007; Technau et al. 2005), while extensive gene loss can be observed in classic model organisms such as Drosophila and Caenorhabditis (Kortschak et al. 2003). This ancestral genetic complexity also applies to neurodevelopment as cnidarians possess an almost complete set of the signaling molecules that have a critical role in bilaterian neurodevelopment (Watanabe et al. 2009). Together, these molecular data suggest significant morphological complexity at an early stage, although the morphology itself remains unknown (Erwin 2005).

Work on molecular clocks aims to position these molecular changes and inventions in time by setting dates for the splits between animal lineages. However, the dates arrived at by these molecular clocks are very early compared to the fossil record (e.g. Blair 2009), despite efforts at calibrating the two (Peterson et al. 2008; Erwin et al. 2011). For example, the deep divergences between animal lineages are set in the Cryogenian, thus before 635 Ma. As some of the split lineages already share the presence of nervous systems, this suggests that nervous systems evolved very early and were present in different animal lineages for tens of millions of years without leaving any fossil evidence, presumably because they remained very small (e.g. Brasier 2009). Alternatively, one may reject these implications of the molecular data and stick to the fossil evidence that positions the relevant evolutionary changes right at the Ediacaran-Cambrian boundary (Monk and Paulin 2014). One way to do this is to accept that the important lineages split very early, but also argue that nervous systems evolved much later and independently in different lineages (Moroz 2009, 2014; Northcutt 2012). Without additional confirmation for any of these options, the absolute timing of early nervous systems evolution should be considered as contested. This wide time range makes it also difficult to be specific about the ecological conditions and the specific selective environment under which nervous system evolution occurred: ecological conditions, such as global temperature and oxygen levels, changed dramatically during the Proterozoic (Gaidos et al. 2007).

Thus, while our knowledge at the molecular level has vastly increased, much less is known about the macroscopic functioning and organization of the organisms that first evolved nervous systems, or about the ecological circumstances under which they lived. Given the scarcity of tangible evidence, specifying schemes that sketch how a non-neural organization could have been transformed into a nervous system would help to further constrain research on early nervous systems.

Both the input–output view and the coordination view provide scenarios for this transition. Many of these proposals do not involve highly specific adaptive constraints. Instead, these scenarios can be better interpreted as lineage explanations (Calcott 2009). Such explanations specify a series of phenotypic stages; each stage shows how a particular phenotypic organization acts as a mechanism that fulfils some particular task; the differences between adjacent stages demonstrate how these mechanisms could change into the next stage’s mechanism through minor modifications. Nilsson and Pelger’s evolutionary sequence of eye-organizations that increasingly enhance visual acuity constitutes a classic example of such a sequence (ibid.). Given their proximate focus, neither the currently unknown adaptive conditions under which these mechanisms functioned, nor the particular animal in which they were embedded need to be addressed by such explanations.

Lineage explanations for the evolution of early nervous systems thus involve specifying a sequence of neural organizations, all acting as a mechanism that performs a plausible adaptive task. While both the input–output view and the coordination view propose such sequences, they differ in their take on what counts as the plausible task that early neural organizations performed to enable a motile life-style. The dominant input–output view interprets nervous systems foremost as a mechanism that allowed organisms to react (more) adequately to the environment by connecting sensors and effectors. In contrast, the lesser known coordination view stresses the initial need to organize bodily coordination as the main purpose. This difference in view on what nervous systems initially were for has important implications for the scenarios that are built on these views.

Developing scenarios for early nervous system evolution has a long history. Most of them fit within an input–output view. To position the contrasting coordination view developed here, a short sketch of what is typical for the input–output view will be helpful. The input–output view is an idealized general interpretation of nervous systems’ functioning that also provides the background for most proposals for the origins of nervous systems. Two issues must be stressed. First, it is important to differentiate the wider use of input–output terminology from the specific view on nervous systems and their evolution. Only the latter will be discussed here. Second, the input–output view is so generally accepted that it is usually only implicit. For example, the idea that brains process information is so widespread within psychology, cognitive- and neuroscience that this particular claim is hardly ever argued for but just assumed as given. In the following, I aim to explicate some of the most typical and relevant characteristics of the input–output view to provide a contrast with a coordination view.

According to the input–output view, nervous systems are at heart information processing input–output devices: nervous systems receive information through sensors, process this information, and subsequently use this information to instruct and modulate effectors. In basic forms, this is envisioned as an almost direct pathway as exemplified by the classic spinal reflex. In more complex cases, a central brain intermediates but the input–output view for whole nervous systems remains in place as brains are conceived of as way stations between afferent and efferent fibers that connect it to sensors and effectors. Feedback and reafference relations add to the basic picture but do not change it fundamentally. The view on nervous systems is closely tied to an input–output view on animal embodiment and its sensorimotor organization. Under the input–output view this sensorimotor organization consists of the sensor and effector capabilities of animals that connect the nervous system to the world. The animal body constitutes a platform that puts the senses and effectors in specific configurations that enable and constrain the specific sensorimotor interactions any animal is capable of (e.g. Clark 2008).

Most scenarios for early nervous system evolution start from here and aim to provide an account how nervous systems came to be interposed between a set of effectors and a set of sensors. For example, George Parker (1919) made the still influential proposal for an evolutionary sequence of (a) independent effectors, (b) direct sensor-effector systems, and (c) a sensory-interneuron-effector organization. Since then there have been many variations on this basic theme (Lichtneckert and Reichert 2007) with Jékely’s (2011) as an ingenious recent version. These proposals are well reviewed (see footnote 1) and will not be discussed here. Instead it will be useful to specify some of the central characteristics of the input–output view.

An iconic illustration of these characteristics is provided by Braitenberg’s sequence of hypothetical artificial vehicles (Braitenberg 1984). These vehicles are small carts with sensors at the front and one or more motors driving it forward. Braitenberg envisioned a systematic progression of these vehicles to illuminate the evolutionary change towards more complex nervous systems and eventually the human brain. The most basic nervous system that is conceptually possible within this view consists of a single sensor, a single motor and a single neural connection between the two. This basic configuration is extended in a quasi-evolutionary way by providing more extensive sensor arrays, more complex effectors and more complex wiring systems between the two. This enabled increasingly complex behavior and complex sensitivity to environmental features.

The image of Braitenberg’s vehicles highlights a number of typical features of the input–output view on nervous systems and the sensorimotor organization to which it is connected. The following list specifies some of these features, selected here with an eye to provide specific points of contrast for the coordination view developed below.

a direct connection between a sensor and an effector is the most basic form nervous systems can take;

the resulting sensorimotor system is mostly passive and put into motion by external stimuli;

activity within nervous systems is primarily through-conducting, traveling from sensors to effectors;

the animal body constitutes a stable platform for sensors and effectors, the precise placement of which is central to behavior;

the animal body houses a nervous system but the latter remains separate, except through its sensors and effectors.

All these features follow directly from the imagery provided by Braitenberg’s artificial vehicles: the similarity of the artificial and the biological; the priority of sensors and effectors; the minimal case cast as a direct connection between them; the inherent passivity and through-conducting organization and so on. The relevance of some of them—in particular 7 and 8—will only become apparent after discussing the alternative.

These features are meant to be typical and conceptually influential rather than a rigid set of necessary conditions that specify the input–output view. For example, the feature of passivity may be questioned given the central role of internal pacemakers (e.g. Selverston 2010) or the observation that it is not a necessary requirement. Nevertheless, passivity remains an influential background assumption in large parts of neuroscience where reactions to stimuli are held central rather than endogenous activity (Raichle 2010) and thus can be considered typical. Finally, this sketch of the input–output view and the following comparison is not to criticize this view. The aim is to articulate some of its central features in order to show in what way the skin brain thesis, as a coordination view, differs on these features, and therefore how it adds to the conceptual options available for understanding nervous systems and their evolution (Jékely et al. in preparation).

The skin brain thesis (SBT) provides a general outline how early nervous systems could have evolved as organs that coordinate contraction-based motility (Keijzer et al. 2013). It sketches how and why a multicellular organization evolved that turned clusters of cells into integrated behavioral units. This section provides a short overview of the main ideas.