The art of observation: bridging science and art to see the unexpected

Every iteration of the scientific process boils down to this: making observations, interpreting and integrating observations, and communicating observations to others. Yet the biology that we seek to observe contains more complexity than we could ever capture in our data and models. Even when we attempt to grasp that complexity with increasingly sophisticated technologies and analytical approaches, ever-looming deadlines, the increasing bar of what a ‘complete’ scientific story looks like, and the natural human desire for simplicity can tempt us into looking for singular, straightforward conclusions within our data. We can also feel pressured to move on quickly to the next dataset or analysis after finding that one straightforward conclusion, or after struggling and not finding any conclusion at all. This cycle results in collections of data that have only been superficially explored, and that contain answers to many more questions than we have yet asked. Analyzing these datasets and approaching a fuller understanding of biology requires systematic approaches, ample time, and a balance between the desire for simplicity and an appreciation of complexity.

In the Greco lab, we access the complexity of biology through visual information – colors and shapes and cellular movements under the microscope. Specifically, we use intravital imaging to understand the properties and behaviors of cells and tissues in the skin of living mice. The images and movies we produce are rich with information, and have allowed us to discover phenomena we had not imagined before doing the experiments, such as cells pushing out, eating up or actively coordinating with their neighbors (Greco, 2016). Discoveries such as these have only been possible when we have really lingered with our data, giving ourselves time to notice the unexpected, and remaining willing to dismantle our initial interpretations. In our experience, this approach leads to the most impactful results and the most fulfilling research process. We try to nurture this approach with regular ‘big-picture’ brainstorming sessions and comprehensive six-month data reviews. In 2023, inspired by the ways art has impacted our colleagues in the clinic, we experimented with a new approach: connecting with one of our university's art galleries to learn new ways of overcoming our preconceptions and exploring our data deeply.

In the late 1990s, Linda K. Friedlaender, Head of Education at the Yale Center for British Art, and Dr Irwin Braverman, Emeritus Professor of Dermatology at the Yale School of Medicine, developed an art-based program using representational paintings to enhance medical students' ability to see visual details (Friedlaender and Friedlaender, 2013; Dolev et al., 2001). In this program, students learn observation by cataloguing the constituent parts in a painting and resisting premature interpretations. Linda and others have found that this systematic practice of slowing down to focus on observation before coming to conclusions is particularly important for healthcare practitioners, who are continually integrating explicit and implicit information about patients to quickly identify the ‘correct answer’ about what their patients need. ‘Enhancing Observational Skills’ is now a required class session for all Yale medical students, and similar programs have been adopted by dozens of medical and nursing schools around the globe (Perry et al., 2011).

When our lab learned about this program, it struck a chord with the need for careful, open-minded observation in our work with microscopy images. We connected with Linda and, over several months, we designed a version of the ‘Enhancing Observational Skills’ workshop and follow-up activities involving paintings and our own microscopy images (Box 1). The goal was to use unfamiliar art-based objects to learn how to explore microscopy data beyond expected patterns and initial answers. As Jennifer L. Roberts puts it in her article, ‘The Power of Patience’ (Roberts, 2013): ‘just because you have looked at something does not mean you have seen it’. We hoped to get closer to truly seeing the biology in our own data.

One morning in early May 2023, our lab spent two hours with Linda at the Yale Art Gallery. Getting out of our lab space was crucial to stepping out of our comfort zone. We found ourselves surrounded by soft-lit artworks and the quiet murmuring of fellow museum visitors, as opposed to the usual fluorescent-lit bays and the relentless hum of lab machines. Linda immediately took us to sit in front of John Constable's ‘Stratford Mill’ study (Fig. 1) before giving us a brief overview of the ‘Enhancing Observational Skills’ class she teaches to medical students. She then instructed us to spend ten minutes silently observing the painting. We were given minimal instructions on how to do this, except that we should try looking at the painting from different perspectives (far away and close up, from the left side and from the right, squinting or with a tilted head), and we were encouraged to roughly sketch what we saw on provided paper. After about ten minutes of silent observation, each of us separately orbiting the painting, we sat back down to describe our observations verbally. For many of us, this is where the challenge began. Linda pushed us to describe the shapes, colors and relative positions of objects in the painting in as much detail as we could before articulating any interpretations or even naming objects directly. For example, instead of saying, ‘there is a man fishing’, we might say, ‘there is a figure in the lower-center of the painting with a red fabric near their neck or back, facing mostly away from us, holding a long, thin object that projects in front of them and towards the blue-gray region of the painting’. Any time we slipped up – stating an interpretation or pointing at the object rather than describing it – she re-directed our words. She also prompted us to see some of the techniques Constable had used to draw attention to different features of the painting, such as distributing pops of red throughout an otherwise mostly green, blue and brown image to move the viewer's eyes across the painting and towards specific figures. Each lab member contributed at least one observation of the painting, and this mode of description seemed to become easier over time.

After about 25 minutes of collectively describing the painting in this way, drawing out as many of these abstract observations as we could, Linda invited us to begin articulating the story we perceived in the painting. The simplest version of that story was a group of country-dwellers on the banks of a river, some fishing or doing laundry or just relaxing. However, different people saw subtly different versions of the story (Was a storm coming in or leaving? Were those horses or cows in the background?). These different stories often stemmed from very subtle aspects of the painting – things that we would not have noticed if we had spent less time with the painting. Finally, by giving us additional historical context, Linda helped us to see that this painting was one of John Constable's attempts to romanticize the English countryside of his youth while Britain was rapidly industrializing. Hearing this context colored our group's story of the painting and, even more importantly, began to reveal how our own expectations and cultural context colored the observations and interpretations we arrived at.

Linda's lesson can be extrapolated into a three-part process: (1) individually making observations, (2) collectively describing our observations without opinions or interpretations, and (3) collectively building a narrative within the painting based on our observations and contextual information. Crucially, we learned to separate the initial observation from the eventual interpretation (i.e. adding time between each step). Preliminary observations constrain what we can observe (Yanai and Lercher, 2021, 2020; Koehler, 1993). Although this constraining effect can sometimes lead us in a productive direction (e.g. presenting new questions to ask about the data), it can also prevent us from seeing things in the data that do not neatly fit the initial interpretation. Being able to recognize and separate observation from interpretation can give us control over when we are closing the door to more observations, and when we are opening that door.

Our activity also reinforced the power of observing and interpreting as a collective group. Each person brings a different perspective, augmenting our capacity to observe and to understand the painting with more depth. Importantly, no one in our group had any expertise in art, let alone John Constable's paintings from the early 19th century, but we were all able to speak about shapes and colors. As a result, each observation or interpretation was voiced and considered with equal weight, and all the observations together formed our group's collective story about the painting. This feature resonated with our group's commitment to team-based science and our frequent conversations about the need to welcome more voices into the scientific process, themes which we continued to explore in subsequent activities.

After practicing this deep observational approach in the art gallery, we applied it to microscopy images from our lab. Lab members were asked, first, to individually observe objects in the image, then to describe their observations in an abstract manner, and finally to collectively integrate the observations together with known contextual information to interpret what the image contained. We practiced this in two stages: in small groups immediately after the gallery experience (Box 1), and in pairs at the lab retreat two months later (Box 2). In both cases, lab members were asked to prepare images from their own data that stuck out to them in some way (e.g. they found the data beautiful or ugly, or they were proud of those data). We wanted to see whether we could observe anything new in our microscopy images using the observational approach Linda had taught us, and to develop concrete strategies for integrating this observational mindset into our daily research practices.

In some ways, describing our own images of mouse skin cells in terms of colors and shapes was even more challenging than it had been with Constable's painting. For example, rather than describing an epidermal stem cell layer interspersed with collagen bundles and hair follicles, we described interconnected cyan polygons blending into seemingly tangled white fibers surrounding distinct circular arrangements of the cyan polygons (Fig. 2). We spend much of our daily lives looking at and discussing such images and one another's projects, so we had to actively resist connecting the images to known research stories in the lab. Describing our familiar images in this abstract way forced us to slow down and created opportunities to see more in the images, including phenomena that did not match our experience or expectations (an example is described below).

Importantly, when we observed and described microscopy images from our lab during these activities, we had the ‘artists’ in the room with us, and they could explain how they generated the images and what they saw in them. This prompted us to realize that we always make observations about our microscopy data while we are in the process of generating images, rather than simply trying to observe the ‘final’ images. Indeed, this process starts as early as choosing which region of skin to image and extends through choosing exposure levels and colors for each channel. We played with this theme in our lab retreat activity (Box 2). First, each lab member chose one image to manipulate in several ways (changing the colors assigned to one or multiple channels, changing the crop or orientation of the image) and put those images side-by-side in one document (Fig. 3). We refer to these compilations as ‘Microscopy Pop Art’ because they reminded us of Andy Warhol's famous multi-colored works. We divided into pairs and swapped our Microscopy Pop Art without explanation. Each ‘observer’ then described the image to the ‘artist-scientist’. Often, the observer's attention was drawn to something different from what the artist-scientist had been focusing on when generating the images, and multiple lab members commented that hearing the observer's perspective allowed them to access information that they had not previously noticed. For example, one artist-scientist gained a new curiosity about the pattern of collagen fibers relative to folds in the epidermis (Fig. 3). She commented that she had initially not focused on the pattern of collagen fibers, but it caught her attention during our activity for two reasons: (1) changing the collagen signal from blue to magenta made the pattern stand out, and (2) the observer, a collagen expert, had noticed and described it. This practice of having someone else describe your own image to you is similar in some ways to discussing your data in a lab meeting context, but a crucial step that we added was having the observer describe the image before the artist-scientist explains anything about it – this gives the observer a better chance of avoiding the artist-scientist's own biases about their data.

This set of activities also allowed us to relate to one another in new ways. For example, some lab members more readily participated in the group conversation when describing microscopy images in abstract terms, in contrast to a typical lab meeting setting. We suspect that this is because describing shapes and colors de-emphasizes the viewer's subject matter expertise and how expected or unexpected patterns in the image are – two things that are valued in scientific research for good reason, but that can sometimes inhibit full participation in a group of varying experience levels. The pair-and-swap Microscopy Pop Art activity was also explicitly designed to allow each artist-scientist to choose images that emotionally resonated with them in some way: images they found beautiful or ugly, fascinating or bewildering, representing experimental struggles or successes. These emotions are almost never discussed in formal scientific venues, but we created a space to become more aware of how these emotional experiences affect our willingness to accept or reject parts of our data. The group activities also provided peer support and collective language that will support our daily research far beyond the bounds of these specific activities.

We came away from this practice with a set of tools to explore and make observations in our microscopy images (Table 1). Some of these tools overlap with practices many of us had already picked up in our scientific training (e.g. changing exposure levels of each channel), but others were less familiar (e.g. the art-inspired observation approach). This list of tools makes it easier to draw on them as needed during our data exploration, and to teach them to new members of the lab.

Through partnering with the Yale Center for British Art, our lab confronted unfamiliar objects and practiced slowing down our observational processes, allowing us to connect with our data and each other in new, profound ways. Although we applied this approach to microscopy data, researchers in any field, working with any type of data, could benefit from these activities and the fundamental lesson: look at the data in simple terms before drawing interpretations. Slowing down and separating observation from interpretation will give you a better chance of seeing more in your data, and seeing the data for what it really is rather than what you want it to be.

Repeatedly throughout this process, we kept returning to the power of the collective in gathering observations and creating deep, authentic interpretations of those observations. This concept is already embedded in so much of our scientific culture, from researching in collaborative teams, to presenting our data in front of many different groups, to the peer-review process itself. Our activities highlighted the power of a group containing many perspectives: if our lab had been more homogeneous, in terms of scientific background as well as lived experiences, we would have likely all observed the same things in the paintings and microscopy images. Moreover, if our lab did not continually work to create a culture of open communication and valuing different perspectives, the varied observations within the group would not have been voiced and could never have been integrated into the rich stories we created together.

We can never observe everything, and our interpretations will never perfectly capture biology. Nevertheless, collaborative experiences like the ones described here can provide us with new tools to explore and make sense of our data, coming closer to a fuller understanding of the biological systems that so fascinate us.

We acknowledge Sangwon Yun for initially connecting the Greco lab with L.K.F., and all members of the Greco lab for contributing to the activities. We also thank Sara Gallini and Ingrid Heumann for providing the microscopy images used as figures in this article, the Yale Center for British Art, Paul Mellon Fund for the use of their collection, and the Yale Art Gallery for the use of their space. Finally, we thank Ingrid Heumann, Shuangshuang Du, Elizabeth Black, Ryan Wepler and Nils Neuenkirchen for providing valuable feedback on this manuscript.