A dialogue between science and conservation: another naïve exploration in response to Salvador Muñoz-Viñas

Abstract This article is a response to Salvador Muñoz-Viñas’ 2022 article in this journal, ‘Conservation science, conservation practice and the conservator’s knowledge: a naïve exploration’. In his article, Muñoz-Viñas outlines 46 questions related to a conservation treatment that, we are told, cannot be answered by using the scientific method. This forms the basis for a vindication of the value of the knowledge of conservators, as well as a critical exploration of the limits of heritage science. In my response, I also present a list of questions related to a hypothetical scientific experiment. I have chosen questions that cannot be answered by science alone, because they require the expertise of conservators. In that way, I showcase a heritage science that respects the tacit knowledge of conservators, indeed, that depends on it. Essentially, the way we ask questions shapes the way we answer them. I argue that we must ask inclusive questions that help us collaborate across perceived disciplinary boundaries. The limitations of heritage science identified by Muñoz-Viñas are genuine, but we disagree in the attitude we should have in front of them. Interesting research happens only at the limits of scientific knowledge, precisely where heritage science has the most acute need for the knowledge of conservators.


Introduction
The recent article in this journal, 'Conservation science, conservation practice and the conservator's knowledge: a naïve exploration' by Salvador Muñoz-Viñas 1 begins as a defence of the value of non-scientific statements in conservation, but quickly becomes an acrid critique of heritage science.Conservation science is termed an 'endoscience', done for and not just by scientists, which is largely 'unapplicable in the real world'.
To some, this could sound plausible.After all, conservation can follow its course without continuous reference to scientific publications.Furthermore, there are very real barriers to the relevance of heritage science research, 2 such as opaque communication styles and a lack of open access publications.However, the fundamental arguments of Muñoz-Viñas are dangerously close to a wholesale disregard of scientific methods in conservation.This is a risky attitude.It hinders the transfer of knowledge between science and conservation, which is the opposite of what the sector needs.This knowledge transfer goes both ways: science needs conservation as much as conservation needs science.If heritage science cannot count on the support of practising conservators, it will grind to a halt.
The core of the article consists of a list of 46 questions of importance in the development of a conservation treatment.'None of these questions'we are told-'can be answered by science or by using the scientific method'.This is true in some cases and debatable in others.In this response, my aim is to demonstrate that these 46 questions, and others like them, far from being a dead end for heritage science, are excellent starting points for its study.It is not the objective of this response to mount a line-by-line rebuttal of the article.For example, to the statement 'so much of [conservation science] appears to have little relevance to conservation treatments', it could be appropriate to respond with a detailed literature review (which could cover anything from laser-cleaning to softcapping).But rather than arguing my case one question at a time, I will take a more constructive route.I want to advocate for the importance of dialogue and collaboration between science and conservation.I will join Muñoz-Viñas in vindicating the conservator's tacit knowledge, which I see not in opposition or as an alternative to scientific thought, but as necessary for its success.In other words, I want to defend a heritage science (or a way of being a heritage scientist) that gives conservation knowledge the highest intellectual authority.
It is useful to begin with a summary of Muñoz-Viñas' argument.The critique of heritage science presented in the paper begins with the impossibility of doing repeatable experiments in heritage science, because 'samples are only representative of some features of the thing represented'.Indeed, many general hypotheses in heritage science are based on experiments that only reflect the behaviour of a specific set of samples.Muñoz-Viñas points out an interesting consequence of this which is that 'if the sample happens not to be representative, it is the conservator who will likely bear the responsibility'.This is summarised as 'the weaknesses of inductive knowledge'.The author cites the philosopher of science Karl Popper to establish that science progresses by 'issuing hypotheses that can be demonstrated to be false', which requires the ability to make falsifiable generalisations.The problem is that conservators work with unique items, and information about them that is hard to codify, therefore, 'establishing a scientific law that could be falsified is rarely within their reach'.
I happen to agree with these fundamental problems.In fact, these are problems that I routinely discuss with my heritage science students, because we encounter them frequently.Creating representative samples or case studies and making general statements that are proportionate to the evidence, are core parts of our job.I disagree profoundly, therefore, with the inference that these problems are so large that they render heritage science useless.
To demonstrate an alternative attitude to science, I will also follow a naïve approach.The following section shows my own version of a simple heritage science experiment, as a way to explore the kind of questions scientists need to ask and how they require different types of knowledge.
A dialogue about a speculative experiment Let us design an illustrative heritage science experiment.I will choose one related to the complex object introduced by Muñoz-Viñas.This will be useful for thematic continuity, but it is not essential, because the same argument could be made with any arbitrary experiment.I have decided to focus on the large tear that divides the artwork in two.Our overall research question would be: how can we predict the appearance of tears in this type of material?This question is not meant to be useful in terms of the choice of treatment as discussed in the original article.It is simply put forward as representative of a heritage science research question.
Before deciding to invest effort, time and resources in answering this research, I would seek the advice of conservation specialists who are familiar with this type of material.I would need help with questions such as: 1. Is it correct to assume that this object shares enough features with other objects, which degrade and develop tears in a similar manner?How can we define a 'type of object' that helps us narrow down the experimental complexity? 2. Is tear formation a prevalent damage process in the type of object? 3. Are there other damage processes that affect the type of object we just defined, and which are poorly understood, and therefore should take priority before tear formation is studied? 4. Is tear formation an active issue, meaning that there are a substantial number of similar objects which may develop tears in the future? 5. Would it be useful to have some predictive capacity on this matter?For example, would it be useful for collection management, transport, storage, access protocols or preventive conservation?
If this dialogue convinces me this is a worthwhile research question, I will proceed to formulate some initial hypotheses about the dynamics of tear formation in this type of material.I would review the literature on material failure and the mechanics of tear formation.But I would also need to speak with a conservator who has first-hand knowledge of the process, and I would ask further questions such as: 6.How would a conservator assess the risk of tearing by inspecting the object?Which visual, tactile and other sensory clues are associated with a higher risk of tearing? 7. What do we know about the dynamics of the process?Do tears tend to appear quickly, because of accidental damage in a single event?Or do they appear gradually, after an accumulation of small events?8. What material properties may be associated with the probability of tearing?I would ask about thickness, type of paper, additives, age of the paper, smell, weight of the artwork and so on.9. What do we know about the prevalence of the damage?Out of all the comparable objects that are conserved, how many display tears?10.What do we know about the rate?How common is it to see this process develop during a human lifetime?
After these conversations, I would be better prepared to formulate a hypothesis about how tears occur and why.The next step is to design an experiment.A good place to start would be working with samples that resemble the object of study.I would select samples that display all the physical properties that are part of my hypothesis: different thickness, additives, sizes, ages, presence of ink and so on.After this process of experimental design, I would ask: 11. Are the model samples representative of the materials that exhibit this type of damage?12. Can data be collected from real, naturally degraded objects, in a way that can be compared with experiments to validate them?13.Are the samples more or less likely to develop a tear than the real materials?14.If they are not representative, but are more likely to develop a tear than real materials, can the samples be considered as a useful 'worst case scenario'?15.If they are not representative, but the experimental variables are already too numerous and the experiment would become impractical, should the 'type of object' be refined to a narrower category, which can be more easily reproduced?
We should note the iterative nature of this process.The creation of test samples can be seen as experiment 'zero'.After the samples have been conceived, it is often useful to re-consider their representativeness.I have been through this process, doing it both well and poorly.
Exhibit A is the occasion where I did it well.I used mock samples of Whatman paper to test the effect of dust deposition on the chemical degradation of cellulose.I relied on the previous experience and knowledge of collaborators who knew a lot more about paper conservation than I do.We exposed the samples in locations around London, and I spoke with conservators again to select locations where the dust would be meaningfully similar to the dust they had seen in their collections.The final experiment combined a material with a simple composition (pure cellulose) and dust that is very representative of real environments.
Exhibit B is the occasion where sample representativeness was less successful.We were attempting to test the effect of the ageing of glues used in canvas lining.Our hypothesis was that, as glue ages, it loses the flexibility required for a successful strip-lining.We created mock samples based on this hypothesis: different common glues, applied in a patch on the back of canvas samples.The glues and canvas were artificially aged at different levels, and the tensile strength tested.Had we spoken with a conservator, perhaps we would have considered other factors that would make our samples more representative of real strip-lining.For example, in designing the experiment I should have considered: (1) the tightness of the weave of the canvas; (2) the properties of the other side of the canvas, such as the thickness and composition of the preparation and the paint; and (3) the adhesion properties of the glue.Crucially, I should have considered how a conservator would adjust this treatment in response to the material properties.Because of these limitations, the experiment was not published.The reviewers rightly pointed out that the assumptions cast a doubt about the significance of the samples (the paper is now available as a preprint). 3The editorial system, for all its flaws, worked well.While I believe the experimental results are valid and show an interesting trend, I understand the difficulty (and risks) of applying our conclusions to specific cases that are not like the experimental samples.Close collaboration between conservation and science would have corrected these limitations.
If we return to our speculative experiment on the formation of tears, we could continue to list questions that require a dialogue between science and conservation.The next block of questions would be about the accelerated ageing protocol, followed by data analysis and the presentation of these data in a way that is meaningful to decision-makers.But I will skip these topics, because the main conclusion would not change: in heritage science, each step of the scientific method provides opportunities to integrate pre-existing knowledge.Good heritage science takes advantage of these opportunities, while poor heritage science is blind to them.
I will add a final element to our tear formation experiment.One of my research interests is the prediction of lifetimes, following the principles of collections demography as defined by Matija Strlic. 4This area of research seeks to create models that describe the ageing of large collections of similar objects, in order to estimate how collection management strategies affect their lifetime.It is possibly the area of heritage science that requires a deeper understanding of materials and their social uses.It also requires a great number of assumptions and simplifications, and a correspondingly high degree of careful interpretation.
Let us suppose that our experiment on tear formation leads us to a usable mathematical model.The output of the model is the rate of tearing, in mm/year.The inputs of the model are some measurable physical properties, let's say: acidity, thickness, size and number of times an object is handled a year.Equipped with this model, we can conduct a collections demography assessment.We can survey the input properties of all the collection items we want, and then predict the rate of tear formation for each of them.We would need to speak with a wide range of experts, including conservators who have handled similar materials as well as preventive conservators who manage relevant collections.We would discuss: 16.Is this model complete and fitting with real-world experience?Is it missing any contributing factors to tear formation?17.Is the rate of tearing, in mm/year, the best output to express this damage process?Would it be better to replace it by the probability of tearing, or the number of tears per year?18.Which aspects of the value of the object are affected by tearing?How does the value depend on the presence of tears?19.Is there a level of tearing in which essential dimensions of the value would be lost?20.If this approach is too reductive, how should it be refined?It could be that the location of the tear matters (i.e.whether it touches the drawing or not) or that the loss of information is a pre-eminent factor.
The nature of this intellectual provocation is that I am pitting a very simple (fictitious) model against very complex questions of value.So complex indeed that some will find the questions not only unanswerable, but perhaps un-askable.The complexity of this conversation is what makes lifetime prediction one of the more exciting areas of development in heritage science.It is close to the limits of what the physical sciences can do.It is an exciting frontier.

Some other relevant questions we could collaborate on
Section II of Muñoz-Viñas' 'naïve approach' article asks several questions about the probability or rate of events.They are: 'What are the chances that the sheet will flatten by itself by simply leaving it on top of the water?';'If the pigments do get wet, what are the chances they will be altered in a noticeable way?'; 'Would water vapour within a humidity chamber humidify the sheet flat enough for float-washing it?';'How much spraying will be needed?';'How long will the pigments stay in place?'; 'How long will the paper remain strong enough to be safely handled?'; and 'How much weakening of the paper is acceptable before becoming too weak to be safely handled out of the wash?'.
These questions strike me as being well-suited to an application of the scientific method.Another way of saying this would be that these questions compel me, as a scientist, to attempt to design an experiment that answers them.For example, when considering the question 'How long will the pigments stay in place?',I cannot help wondering about how we could easily produce or obtain samples with different pigments, wet them with different methods, wait for different lengths of time, and photograph the results.Surely, this is doable and interesting.My answer to this alleged impossibility is simple, and powerfully naïve: to me, it sounds like an exciting challenge.
1.The key problem of inductive reasoning is that every artwork is unique, and therefore it is difficult to make generalisations.Firstly, this is well known among heritage scientists. 5It is widely recognised as something that makes this field intellectually stimulating.Secondly, it is by no means a unique problem of heritage science.Sciences like forensic science or medicine also deal with objects of study that are basically unique but roughly similar to others.The philosophical consequences are largely unexplored in heritage science, and it would be interesting to do so.

The key problem of non-measurable material factors hinges on two
examples.The first one is that 'no device exists that can measure the correct application of adhesive at every point of the surface'.Wouldn't it be interesting, then, to develop such a device in collaboration?Or, if developing it is really outlandish, isn't it fascinating to explore why, and demonstrate that any existing alternative underperforms the tacit knowledge of conservators?The second example is that 'there is no way to measure the adhesive power of the starch paste used before it has cured'.Naturally, there can only be a predictive estimation.However, wouldn't it be interesting to measure how uncertain our estimations would be? 3. The key problem of non-measurable immaterial factors is exemplified by the claim that 'no scientific method can determine if a given degree of alteration is acceptable'.In fact, this has been done with practical results: the development of damage functions for paper and other materials involves a quantification of unacceptable levels of change. 6he research is out there, ready for improvement and debate.It is a very complex type of research.One interesting issue is that the method does not translate easily from one material to the other.It is, probably, the heritage science challenge that most requires multi-disciplinary collaboration.4. The key problem of 'the conservator's influence in reality' states that the same answer can be correct and incorrect depending on the individual using it.It reminds me of a small test we did with colleagues at a UK museum, trying to ascertain which kind of anti-vibration padding was more protective during the transport of artworks.We tested five different padding methods, from simple to high tech, applied by five different conservators.Our findings were exactly as Muñoz-Viñas suggests: conservators had a bigger effect on the outcome than the padding method!Someone with expertise in paper padding could reduce vibration more than someone using a high-tech foam without the required experience.Wouldn't it be interesting to test this for other conservation tasks?Would it not evidence the value of conservation expertise?I therefore discard these four points as objections, and I transform them in my mind into exciting challenges.These are not cracks in the vessel of heritage science, they are the engine that keeps it going.
Heritage science and conservation knowledge I invite the reader to see these questions (my 20 and the 46 by Muñoz-Viñas) not as 'questions that science cannot answer' but as 'questions that science cannot answer alone'.In this response I have outlined a way of thinking scientifically that begins and ends with conservation expertise.The relevance of the questions, the suitability of the methods, the usefulness of the results, the clarity of the presentation: they all benefit from a collaborative attitude.
There is no need to repeat within conservation a small version of the 'science wars' that permeated science studies in the 1990s.Bruno Latour, reflecting on these philosophical encounters, wrote something that can help us here: 'Science studies has become a hostage to a huge shift from Science to what we could call Research […].While Science had certainty, coldness, aloofness, objectivity, distance and necessity, Research appears to have all the opposite characteristics: it is uncertain, open-ended; immersed in many lowly problems of money, instruments and know how, unable to differentiate as yet between hot and cold, subjective and objective, human and nonhuman.' 7Latour's understanding of 'research' is closer to the heritage science I know and advocate for, than the image portrayed by Muñoz-Viñas.A science which embraces doubt, and therefore embraces the expertise of others.
This heritage science is a reality, not an abstract aspiration.The work of many colleagues demonstrates these attributes of collaboration and respect for conservation knowledge.If I may, I would like to cite three examples I am very familiar with: (a) The model of cellulose acetate degradation created by Katherine Curran and her team started and ended with in-depth conversations with subject experts.The outcome of the cross-disciplinary conversations that informed the project design was published in a dedicated article, which may inspire others to follow similar methods.8(b) The study of the complex dynamics of archival collections conducted by Cristina Duran, a highly mathematical piece of research, built its solid foundations on participatory sessions with 25 individuals who had intimate knowledge of the inner working of archives.9(c) As an example of a recent development in conservation treatments, the initial tests around the performance of nanocellulose as a consolidant for canvas paintings were carried out by both scientists and conservators, as seen in the co-authorship of the relevant publications.10 My main claim in this brief article is that this type of collaborative heritage science is possible, and that we must do our best to strengthen it.This means tearing down barriers by asking generous questions that can be answered through partnerships, rather than building walls of suspicion and mistrust.I will, therefore, end this response with a heartfelt agreement with the main conclusion of Muñoz-Viñas's paper: the conservator's tacit knowledge is essential to heritage science.

摘要
学的方法回答。这构成了为保护人员知识价值辩护的基础，以及对 遗产科学局限性的批判性探索。在我的回应中，我也列举了一系列 与假设的科学实验相关的问题。我选择的问题是紧靠科学无法回答 的，因为它们需要保护人员的专业知识。通过这种方式，我展示了 一种尊重保护人员隐性知识的遗产科学，事实上，它依赖于这种知 识。从根本上说，我们提问的方式决定了我们回答的方式。我认为 我们必须提出具有包容性的问题，以帮助我们跨越学科界限开展合 作。穆尼奥斯•比尼亚斯所指出的遗产科学的局限性是真实的，但在 我们面对时，应持何种态度存在分歧。有趣的研究只会发生在科学 知识的边界处，而这恰恰是遗产科学最需要保护人员知识的地方。BiographyJosep Grau-Bove, FIIC, is an Associate Professor of Heritage Science and deputy director of the Institute for Sustainable Heritage, University College London (UCL).He obtained an engineering degree from Rovira i Virgili University, Tarragona, Spain, and a masters in History of Science from the Autonomous University of Barcelona.With his team, he uses scientific and engineering methods to support preventive conservation.He has pioneered the use of various types of simulation in the field of heritage, such as computational fluid dynamics and agent-based modelling.He leads the MSc Sustainable Heritage programme, an innovative course where students specialise in data science, heritage science or heritage management.He also chairs the Heritage Science Group of the Institute of Conservation.