ABSTRACT
Developmental biologists can perform studies that describe a phenomenon (descriptive work) and/or explain how the phenomenon works (mechanistic work). There is a prevalent perception that molecular/genetic explanations achieved via perturbations of gene function are the primary means of advancing mechanistic knowledge. We believe this to be a limited perspective, one that does not effectively represent the breadth of work in our field. We surveyed a representative and diverse group of colleagues to share their views on what it takes to infer mechanism. Here, we briefly examine the factors that have shaped the dominant view of mechanism, summarize responses to the survey, present our views, and suggest a path forward that embraces a broad outlook on the diversity of studies that advance knowledge in our field.
The challenge
All scientific research begins by identifying an observation of interest, which could be an object such as an organism or a planet, or a phenomenon such as the migration of a bird or the rising and setting of the sun. A key first step in understanding these observations is to obtain a full description or characterization. This ‘descriptive’ work seeks to address the ‘what’ questions about an observation. Further investigations can seek to answer the ‘how’ questions, and these studies that unveil how a phenomenon works are generally referred to as ‘mechanistic’ work. Although descriptive studies provide the foundation for mechanistic work (enabling the generation of mechanistic hypotheses), mechanistic studies are perceived to be more important in modern day developmental biology. Further, although many types of explanations are possible for understanding how a biological phenomenon works (e.g. on a molecular, genetic, cellular, organismal, biophysical or systems level), molecular/genetic mechanisms are widely perceived to be the gold standard for a mechanistic explanation of development. We wondered how mechanistic questions, particularly at the molecular/genetic level, came to dominate developmental biology, questioned whether there truly is strong consensus amongst our colleagues, and considered whether a revision of our attitudes about mechanism is needed.
In this context, it is important to note that a long history of philosophies of science have defined what we as developmental biologists seek as explanations of the biological world (Baedke, 2021; Bechtel, 2005; Cleaver et al., 2023; Gilbert and Sarkar, 2000; Glennan, 2002; Lillie, 1945; Lynch, 2017; Machamer et al., 2000). Notably, advances in genetics and molecular biology in the previous century brought a significant shift in philosophies – from (w)holism to reductionism (Gilbert and Sarkar, 2000). Reductionist philosophy argues that biological systems can be understood by studying the workings of their constituent parts. Given that genetic information in DNA carries the instructions for building cells, tissues, and organisms, genes have become a focal point for reductionist studies of development. In this view, identifying genes that underlie the process of gastrulation would be seen as a satisfying mechanistic finding, but understanding how population density or environmental variables impact gastrulation would not be valued as a mechanistic explanation. In contrast to the bottom-up approach of reductionism, (w)holism considers that biological systems are greater than the sum of their parts, they exist in complex environments, and both bottom-up and top-down approaches are needed for a full understanding. In this view, considering the environment, other embryos, cells and genes would all be seen as important for a full understanding of how gastrulation works. The shift toward reductionism has meant that developmental biology has come to value mechanistic studies at the molecular and genetic level over explanations at other levels.
Different biologists place different values on descriptive and mechanistic work, and there is further disagreement on which level of mechanistic understanding qualifies as satisfactory. Depending on which level we focus on, we are subcategorized as structural biologists, geneticists, cell biologists, evolutionary biologists, biophysicists, and so on. Within our subfields, we have established frameworks for questions and standards for experimental approaches, which have boxed us into silos that think differently and feel strongly about levels of insight. For example, a developmental geneticist might find a population-based study unsatisfying because their interest lies in a detailed understanding of how a gene of interest impacts a particular phenotype, which requires using isogenic lines in controlled lab conditions. However, an evolutionary developmental biologist might be unfulfilled by the geneticist's study identifying one gene's function in one process in one strain of an organism as their interest may lie in understanding processes that occur in nature – in polymorphic populations in uncontrolled environments. Although these differences in perspective are innocent, emerging purely from our different interests, there is palpable pressure pushing one type of insight. In our experience as authors, editors, and reviewers, and in discussions with colleagues in our field, we have noticed ever-increasing requests for more experimentation to establish molecular/genetic mechanisms.
Accepting the dominance of one level of mechanistic explanations has significant repercussions. We risk not achieving a fuller understanding of how developmental processes work and we risk harming the careers of colleagues who pursue questions at different levels. Prominently, many of us have witnessed the use of terms such as ‘descriptive’ or ‘not mechanistic’ as pejoratives in the assessment of grants and manuscripts. We argue that being hyper-focused on molecular/genetic mechanisms is unjustifiable: developmental systems have many levels of organization. Should explaining these systems not require knowledge of how things work at many levels? To achieve these multi-level explanations, should we not value a diversity of approaches, well beyond perturbing genes? We wondered who should define the level of mechanism at which a study needs to be conducted: the investigator(s) who sets out to study the question in the first place, or the collective community? We sought input from scores of developmental biologists and below we summarize their perspectives on what is, and should be, meant by ‘mechanism’. Finding that many of our colleagues take a broad view of mechanism, we propose a path for incorporating this diversity to advance our field.
The community view
We solicited answers to three questions from developmental biologists at varied ranks and career stages, including postdoctoral researchers and faculty. The first question asked the respondents to describe how they think their field defines ‘mechanism’ (Fig. 1A), the second asked them to articulate their own view of the term (Fig. 1B), and the third asked them to provide advice to future generations of scientists on how they should think about mechanism (Fig. 1C). Of the 226 biologists the survey was sent to, 71 (31%) responded. In response to the first question, despite the respondents identifying themselves as having diverse subfield associations (Fig. 1D), we observed broad consensus among the respondents that mechanistic work, as defined by their field, requires molecular or genetic perturbations. This matched our expectations based on our own experience in this field. However, what drew our interest was the response to the second question. We observed a divergence in views among the respondents as they elaborated their personal perspectives – many survey participants stated that mechanistic understanding can also be achieved by studying biology at higher organizational levels, such as cells or tissues, as well as interactions across biological scales. In fact, over 70% of the respondents expressed a broad view of how mechanistic understanding can be achieved, while only 16 (23%) fell into a ‘strict molecular’ category, requiring molecular or gene perturbation studies to define a study as mechanistic. Below, we summarize the three major themes that stood out to us from the views articulated in the survey.
Word clouds summarizing answers to survey questions. (A) Question 1: In your area of work, what is the most common interpretation of the term ‘mechanism’ as applied to research questions? (B) Question 2: When you assess work in your field, what do you look for in terms of a mechanistic understanding of development? (C) Question 3: How would you recommend new generations of scientists to think about ‘mechanism’? (D) Question 4: What is your field of study?
Word clouds summarizing answers to survey questions. (A) Question 1: In your area of work, what is the most common interpretation of the term ‘mechanism’ as applied to research questions? (B) Question 2: When you assess work in your field, what do you look for in terms of a mechanistic understanding of development? (C) Question 3: How would you recommend new generations of scientists to think about ‘mechanism’? (D) Question 4: What is your field of study?
Levels of mechanism and biological organization
A key point that emerged from the survey is the importance of being mindful of the organizational level at which a mechanism is being studied. Many respondents emphasized the significance of understanding phenomena across multiple levels or scales, recognizing that often a single study cannot address these different levels simultaneously. It is crucial to determine the specific level at which the mechanism is being explored and to formulate hypotheses accordingly. Most importantly, the responses were thoughtful in identifying current inequities that developmental biologists are subjected to, in terms of what counts as sufficient advance. One respondent stated: ‘My feeling is that when focusing on a process at the scale of the organism, we are expected to go to molecular details, all in one paper. However, when people work on a molecule they are not expected to move scales up to the organism level in one paper.’ Another researcher expressed that ‘understanding the complete picture from gene sequence to organisms in their environments is an ideal, but such breadth is not feasible to achieve for most labs’. Most importantly though, everyone agreed that each of us as a researcher finds a certain level of reductionism satisfying enough for defining ‘mechanism’, whether it's the molecular details or broader organismal functions. Thus, it is important to appreciate that others may focus on different levels of this spectrum: ‘We need to appreciate and learn from all styles of research endeavors. The forest is important; as are the trees and even the leaves!’
The need for perturbation for mechanistic inference
There was a broad range in respondents' views on the need for perturbation experiments as an approach for inferring mechanisms. A few respondents suggested that perturbation does not need to be confined to the genetic level, but can include changes in parameters such as environment, developmental time, or age. Others emphasized that mechanistic explanations should extend beyond perturbations, and noted that observational techniques such as live imaging are also approaches that enable mechanistic inferences. These views reminded us of the words of Theodor Boveri, the German zoologist who studied chromosomes and was one of the founders of cell biology: ‘Some people think that it is not a real experiment if no sections have been made, or a new stuff injected into the material, or a new gadget has been built. But the essence of experimentation is that one knows with certainty that certain typical conditions have been changed in a definite way. It does not matter whether the experimenter or nature makes these changes. Actually, the scientist will prefer what nature did without man's crude methods of interfering with the material’. Indeed, in the years preceding the publication of On the Origin of Species, Darwin made many observations that led him to propose the Theory of Evolution, which included a mechanistic explanation of how the process of evolution works (via the action of natural selection on heritable traits) (Darwin, 1859). We need to be open to the possibility that observation-based work can also yield mechanistic insight.
The value of descriptive studies
We found that several respondents (13/71) specifically indicated that descriptive studies are essential for advancing biological understanding and are crucial prerequisites for mechanistic research. For example, the advent of new technologies that enable profiling developmental stages or across tissues at single-cell resolution has helped generate unprecedented large-scale maps of organs and organisms. Although these studies are descriptive in nature, they generate hypotheses for mechanisms of controlling gene expression. Thus, a bias which tilts the scales in favor of only perturbation-based mechanistic studies as being important is limiting – we should remember that descriptive studies lay the groundwork, providing foundational descriptions of phenomena. One respondent perspicaciously noted: ‘When I review a paper, I ask whether the work has significantly advanced our understanding of the gap in knowledge. I don't bother myself with trying to define and assess “mechanism” since this word means different things to different people’. Respondents also suggested that the rigor of descriptive studies can be enhanced by applying a quantitative framework to these investigations.
A definition of ‘mechanism’ by the community today
Overall, we attempted to integrate community views into a definition of mechanism, and we propose the following as a starting point to be discussed, shaped, and improved upon: a mechanism in developmental biology can be defined as the causal explanation of a phenomenon, encompassing the interactions of components and processes at various biological levels (molecular, genetic, cellular, organismal, or systems) that drive the observed outcome. Mechanistic understanding can be achieved via different approaches, such as perturbation or observation, and descriptive studies are often crucial predecessors to mechanistic insights.
A way forward
The survey results corroborated our view that the idea of ‘mechanism’ is complex and revealed that the perception of the field's expectations differs from individual definitions of the concept. They also drew attention to the dichotomy between ‘descriptive’ versus ‘mechanistic’ work and their widely-perceived associations with observation- and perturbation-based approaches, respectively. We argue that descriptive work probes the ‘what’ aspect of a biological system and is, by definition, an important first step when learning about a new biological system/phenomenon. Moreover, we should be open to the possibility that observation-based work can deliver mechanistic insight. For example, consider John Sulston's work on the Caenorhabditis elegans embryonic cell lineage (Sulston et al., 1983), in which following embryonic cell divisions revealed key mechanisms of development – that fates are segregated early and that apoptosis is important for shaping the embryo.
We urge grant and manuscript reviewers to question reductionism, to be open-minded about the approach used and the level of mechanism under study, and to consider each study relative to what is already known in that system, assessing the potential of the work to advance knowledge. In the same vein, we call upon researchers to invest effort in effectively defining the context of the study and explaining how their study advances knowledge in the field. Our proposal should not result in reduced rigor, it is charging authors and reviewers to delve deeper into the work to identify and articulate its context and impact. These efforts should reduce our reliance on reductionism and allow for conceptual advances at any level to be recognized based on their contribution to the field. This shift has been advocated by philosophers of science (Brigandt and Love, 2008; Gilbert and Sarkar, 2000; Sarkar et al., 2018) as well as by scientists, e.g. see Cooper (2024) for a thoughtful argument against seeking simple genetic explanations.
Developmental biology is at an exciting juncture. Recent advances such as single-cell RNA sequencing, genomics, genome-editing, novel imaging techniques and other quantitative approaches have expanded our toolkit. These advances allow us to study new organisms and to integrate across disciplines, opening up new questions. At the same time, our planet faces unprecedented challenges, bringing an opportunity for developmental biology research programs to be adapted to study the impacts of climate change on biodiversity and organismal health. We can meet this moment by creating space and respect for work at all levels of mechanism, harvesting the richness of developmental biology more fully. In all aspects of our professional lives, be it as designers of a new study or as reviewers of others' work, we recommend focusing on whether a study advances the field, not on whether it fits a predefined framework of mechanistic insight.
The excitement for broadening the work ahead of us is shared by many members of our community, and it is evident in the advice our survey respondents provide to younger generations of scientists (Fig. 1C; Box 1). Instead of doubling down on the primacy of molecular/genetic work, their words implore trainees to ‘think’ (deeply and broadly) and to consider ‘levels’ of organization. We cannot wait to see where the broad-minded, question-focused, and interdisciplinary work of our trainees takes the field in the years to come.
Appreciate diversity in approaches and interdisciplinary work. Recognize the value of different research approaches, from observation to computational analysis to genetic perturbation to modeling. In depth mechanistic understanding may require interdisciplinary studies.
Acknowledge complexity and integrate across multiple scales. Be mindful that developmental systems involve complex, multiscale interactions, and a full mechanistic explanation would ultimately include many levels of biological organization, ranging from organisms to cells to genes and even populations and the environment. While reductionism is a useful framework, avoid becoming too narrowly focused.
Push for advancing knowledge. Look for the question being addressed and how a particular study advances knowledge. In some studies, this would involve establishing causality at a clearly-defined level of biological organization. In some studies this will involve making predictive models based on the interactions of system components. In some studies, this could involve descriptive work that provides foundational insights that are essential for building towards mechanistic explanations.
Methods
To identify prospective survey participants, we acquired invited speaker lists for several international meetings focused on developmental biology, including Society for Developmental Biology, Pan-American Society for Evolutionary Developmental Biology, European Society for Evolutionary Developmental Biology, and Gordon Research Conference in Developmental Biology, over the past five years. This resulted in a list of 226 faculty and postdoctoral researchers from around the world. Our survey was emailed to all these people in the list.
Word clouds in Fig. 1 were generated using wordart.com. The full text of all answers received for each question, redacted for the term ‘mechanism’, was used as input. Word.art automatically removes common words (e.g. are, we, etc…). For Box 1, we used ChatGPT-4o to summarize the advice given in answer to Question 3. This summary list was extensively edited by all authors and cross-checked by manual assessment of the original answers. We also used ChatGPT-4o to obtain a preliminary definition of ‘mechanism’ based on survey answers. This definition was also extensively edited by all authors for the final text.
Acknowledgements
We thank 71 developmental biologists who took the time to answer our survey, many colleagues who have engaged in conversations about the definition of mechanism in developmental biology, and many colleagues who read and gave feedback on this manuscript.
Footnotes
Funding
B.D.O. is supported by the National Institute of General Medical Sciences (1R35GM138008-01). M.S. is supported by the National Institutes of Health (GM153252) and National Science Foundation (IOS-2401057). Deposited in PMC for release after 12 months.
References
Competing interests
The authors declare no competing or financial interests.