Sunday, November 9, 2014

Measurement: Do we take it or make it?

In May we took a perspective on measurement theory in the philosophy of science. The current snapshot of measurement theory is that measurement is representational.

The key aspects of representation through measurement are:
  • Measurement tells us what things look like (from a specified vantage point), rather than what they are like. 
  • Measurement involves selective perspectival input. 
Van Fraassen uses the analogy of visual perspective to illustrate (1) and (2). Think of measurement as taking a vantage point on some phenomenon. For example, when we measure evolutionary processes, the scientist decides the perspective (e.g. the view from the gene, epigene, individual, population, niche, etc.).

This view, which I will refer to as the ‘perspectival view,' is informative for analyzing scientific practice. It accounts for the imperfections and limitations of our representational activities while grounding a certain kind of objectivity. The phenomena are presumed to remain stable, even if our theories and practices may vary in successfully representing those phenomena. This is the basis for the appearance-reality distinction discussed in ‘Explanation and Illusion’. However, the perspectival view of measurement places too much focus on the outcome of passive representation, and not enough on the process of measurement.

Van Fraassen is wrong to use perspectival art as an analogy to measurement. Taking a perspective is a passive activity. Much of measurement is not passive. It is messy in terms of the type of interaction that takes place. And, a theory of measurement should account for this interactivity.

Here’s a simple example to bring our attention to measurement interactivity:

How do you measure the boiling point of water? You stick a thermometer in the sample, and, get a reading. This seems simple and representational: the value on the thermometer represents the quantity of temperature. But this perspectival story leaves out the interaction involved in the measurement process.

According to Hasok Chang (2004), the history of standardizing fixed points in thermometry is a history of “manufacturing” fixed points. The point at which water boils depends on the material conditions within our measurement set-ups. Initially it was discovered that boiling point varies with differences in atmospheric pressure (2004, 15). Additionally, the presence of dissolved air in water produced ebulliation-like phenomena at 101.9 degrees C (2004, 19). However, purged water (water without dissolved air) was measured to behave in a phenomenologically similar manner at much higher temperatures (as high as 140 degrees C). According to Chang, scientists began to focus on samples of water without dissolved air (2004, 16-19). Chang presents anecdote about how De Luc walked, slept, ate, etc., for 4 weeks straight all while shaking a tube of water to purge it of the dissolved air. De Luc’s dedication to manipulating the conditions of measurement serves as a good illustration of the care with which the measurement interactions have to be chosen in order to have a stable, reproducible phenomenon. It also illustrates how sensitive the phenomenon is to the interaction between conditions of the measurement set-up.

Now, for the difficult philosophical question: Is this type of measurement representational, or, is it productive? In each of the measurement set-ups, the boiling point is taking shape with the measurement conditions. In other words, the set-up provides the conditions for the production (and re-production) of the phenomenon. This type of language doesn’t have to “sound” quantum mechanic-y or constructivist. We do not have to discuss a pot of water boiling in the forest. In fact, we need not say anything about pre-measurement values and post-measurement results. All we have to focus on is that the interaction of the conditions for measurement matter to the production of the phenomenon. In simple terms, change the conditions, change the phenomenon. Whether you choose to remain a representationalist or a productivist, one thing we have to consider is that much of the measurement process occurs within the measurement set-up and execution. The final representational step, the measurement outcome, is a small slice of the process. A robust theory of measurement should account for the interactions in this process.

While we’re doing some revision, let’s try out a more adequate art analogy for measurement—one that focuses on interaction rather than passive perspective. The art process of Jackson Pollock is a good starting point.

Pollock numbered his paintings so that people would look at them without searching for representational elements in the names of his paintings (Karmel and Varnedoe 1999). For Pollock, the work of art is not a representation of a phenomenon (1999, 68-69). Rather, it is the phenomenon, which is produced by the interactions that take place in the painting set-up (1999, 99). Pollock was resistant to representation in art. He was also resistant to the view that artists should paint things “out in nature”: “When asked whether he painted from nature, Pollock replied: “I am nature”” (1999, 253). Pollock’s painting set-up and the interaction that occurred within this set-up can be summarized as follows: First, paint was carefully selected to have the proper viscosity. Pollock used gloss enamel paint rather than oil-based paint. The paint was sometimes diluted to have little textural effect, and at other times thickened. He used sticks, worn out brushes, and basting devices that looked like giant fountain pens. 

Pollock also used raw, unstretched canvas in order to be able to perform full-body painting (1999, 72). The painting resulted from the interaction that took place within the painting set-up. Moreover, it is fair to say that it is difficult to appreciate the work of art without looking at this process of interaction. As part of the painting set-up, Pollock is interacting with other elements of the set-up to produce the phenomenon. In interviews Pollock describes being “in” his painting when making it (1999, 17). This analogy puts emphasis on the messy interaction that occurs in measurement. It is, however, important to note that the analogy is incomplete. In measurement, we want repeatable, reproducible phenomena. In painting, we want unique, authentic productions. I welcome all of our fellow DR-ers to manufacture a better analogy.

Work Cited:

Chang, Hasok. (2004). Inventing Temperature. Oxford: Oxford University Press.

Karmel, Pepe, & Varnedoe, K. (1999). Jackson Pollock: interviews, articles, and reviews. New York:Museum of Modern Art : Distributed by H.N. Abrams. 
van Fraassen, B. C. (2009). Scientific representation: Paradoxes of perspective. Oxford: Oxford University Press.

Vadim Keyser
Department of Philosophy
Sacramento State


  1. Sports is usually a safe place for an analogy. We can look at a tennis serve. The result that is wanted is for the ball to go over the net. There are many factors going into the production of the serve (toss of the ball, tilt of the racquet, bend in the knees), and it is hard to appreciate the serve (or free throw, or curve ball, measurement, etc) without knowing of the time and practice that had gone into the perfection of process to get the reproducible results.

  2. Michael,
    I like that. It captures the production as well as the precision (possibly accuracy, if we can go that far).

  3. Vadim, thanks this is interesting. I'm not sure I really see the issue between the representationalist and the productivist. What am I missing when I summarize your main point as follows? "In scientific measurement, the phenomena of interest are often difficult to isolate and stabilize. A lot of work is required to produce and re-produce these conditions " I doubt anybody disagrees with that, so what is the real disagreement? Surely it is not whether to characterize the productive work as part of the measurement procedure. That seems like a purely semantic issue. But, as you say, the productivist is not making the more radical claim that phenomena do not exist prior to human activity or involvement. Nor is the representationalist denying that measurements are theory laden.

  4. Randy, thanks for the comment.

    First, a small point: We shouldn't bring in anything about theory-ladenness here. True, data interpretation and theoretical terminology are part of the total measurement process. But these issues open up a very important set of questions about the interaction between theory and practice. We're not dealing with this here. It's just too large.

    Second, let's clarify the "issue" that we're looking for. We're not setting up exclusive options. Representationalism a la van Fraassen (as well as most of the historical tradition) places emphasis on a very specific part of the measurement process--the measurement product, and its relation to the phenomenon. The productivist account brings attention to the large amount of things that occur prior to that point. (Your point above--"surely, it is not whether to charecterize the productive work as part of the measurement procedure"--is a bit off. Representationalism doesn't discuss the measurement procedure independent of the representational aspects. In this way, it is incomplete and inaccurate. This is why productivism is important.) Additionally, the productivist account frames measurement as an active, rather than passive, process. So, this article presents an account that is more adequate to what measurement practice is like. It doesn't present a disagreement between two sides. In fact, there need not be disagreement. One can admit to representational and productive aspects of the measurement process. But, one point to consider: if the representationalist gives in too much to the productivist, then he risks looking too much like a productivist. That's not so bad. Maybe it is. We're not just "re-producing conditions", as you summarize above. We're re-producing phenomena. That doesn't sound too great to representationalists.

  5. Vadim, thanks, that helps.

    As I read what you are saying above, a large part of the difference between the representationalist and the productivist is what they choose to emphasize. I'm not completely sure what that means. I can see it as meaning that there is no real disagreement at all, and that this just reflects what aspect of the process they are interested in, or that they think the aspect they are interested in is in some sense the most important part of the process. Consider as an example, the theory of the swing in baseball. Some theoreticians emphasize the rotational dynamics largely responsible for power; others emphasize the movement of the hands and the bat through the hitting zone, largely responsible for elevating the probability of solid contact. I think it's fair to say that the respective emphases here don't reflect a lot of empirical disagreement; it is more about whether a person values power or average. Anyway, perhaps you could explain the basis of the respective emphases for the two measurement camps.

    Also, I'd be interested for you to explain this statement a little more: "Additionally, the productivist account frames measurement as an active, rather than passive, process." By this, I take you to mean, not just that there is a lot of hard work involved in producing the conditions conducive to a measurement event, but that the event itself is active rather than passive. Can you spell this disagreement out in relation to a particular measurement event? I mentioned theory-ladenness only because that is one obvious way to do it, so setting that to the side, what is the cash value of saying that a measurement event is active or passive?

  6. Randy, great clarification.

    I think we can re-summarize your original point as, there should not be any disagreement. If it's just a matter of emphasis then one can hold a consistent view of measurement that acknowledges the productive aspects of the setup as well as the representational aspects of the interpretation. If that's what you meant, I'm on board.

    But my hunch is that representationalists are not. The reason why is because this would require a major shift in key aspects of representationalism--for example, the views on measurement apparatuses and phenomena. Let's start with apparatuses. Take a look at this quote:

    Measurement is an operation, using something that functions as an instrument, to gather information. The instrument is being used, in Heidelberger's terminology noted above, in a representative role, that is, the opertation has as outcome a representation of the object or situation operated on. (van Fraassen 2009, 374)

    If you are being charitable, you could just say that vF isn't giving us a complete definition of measurement. But you could argue that the measurement setup is being ignored completely. You could also argue that there is a major confusion going on about what measurement consists of: the act, which involves an instrument, or the process, which involves the total measurement setup. Notice that in the quote the focus is on the act of measurement (taking), rather than the process of measurement (making). Here, you also find what I was saying earlier about passive vs. active. Measurement isn't only about putting an apparatus up to an object and matching up representation to reality. In fact, in many cases of biological measurement, the apparatus object distinction are unclear--e.g. manipulating a mouse's diet to measure chromatin patterns. Even worse: when measurement becomes to active--i.e. when phenomena depend on the conditions to a large extent--we might have an even more difficult time getting the representationalist to give up room. More on that below.

    Now about the phenomena. I haven't mentioned this before, but, it is worth noting: There might be a large disagreement about the nature of gathering information about phenomena. Representationalists claim that we can gather information about the phenomena. This is difficult to see, given that when we measure we only obtain appearances. Productivists are all over the place about measurement-independent access to phenomena. Some, like Karen Barad, think that there are no boundaries in phenomena independent of measurement. So, maybe, there is no information to be gotten. Here, we may have a legitimate disagreement about the nature of phenomena, rather than a difference in emphasis.

    But a small point of clarification about emphasis. In the context I presented, emphasis isn't just a choice, but rather it is an issue about adequacy. This relates to your baseball analogy.

    I like the analogy. But I'm not sure it captures the disagreement. In the analogy, the emphasis is on value. Here, we're talking about empirical adequacy (of measurement theory). So, we can't say, if you value the representational aspects more, then side with representationalism. We have to revise measurement theory to account for both represenationalism and productivism. Maybe that's possible with the baseball example too, but it seems like you're setting it up in a way where we have to stick to one or the other. Another way that the baseball analogy falls apart is that both strategies focus on process. Representationalism focuses on product. Productivism focuses on process.