In the literature on scientific explanation, there is a classical distinction between explanations of facts and explanations of laws. This paper is about explanations of laws. As our title suggests, our main question will be: how do mechanistic explanations of physical laws provide understanding?.
Hempel’s views on explanation are known as the covering law model. Applied to laws, the idea is that laws have to be explained by subsuming them under other laws. We have to show that the explanandum law could have been expected given the laws in the explanans. Lower-order laws (laws with a relatively narrow scope) are explained by showing that they follow from higher-order laws (more encompassing laws). In many cases, the covering law model represents the explanations physicists give for laws. We explain, for example, the law of refraction in geometrical optics by deriving it as a consequence of Maxwell’s equations in electromagnetic theory. The conviction that laws form such a hierarchy (and the possibility of a unique foundations at the basis of all physical phenomena) has been one of the most powerful driving forces in the research practice of physics.
However, physicists also often explain laws by means of micro-reduction. In that case a physical law, which describes a systemic behaviour at the macro-level, is explained in terms of the behaviour of the constituents of the system at the micro-level. Explanation by subsumption does not require decomposition of systems into lower-level parts, micro-reductive explanation always requires this (by definition). It is not hard to find examples of such explanations in physics. Boyle’s law can be explained in terms of the kinetic theory of gases; the laws of circular movement of rigid bodies is explained by means of the kinetic theory of matter; and Ohm’s law is explained in terms of moving electrons. Our paper deals with explanations of this type.
The mechanistic model of explanation, as it has been developed in the first years of the 21st century, can shed light on the structure of micro-explanations in physics. That will be clarified by means of two examples: the periodic table of elements and Boyle’s law.
These examples show that there are mechanistic explanations in physics. The next step (and main aim of our paper) is to investigate how these mechanistic explanations provide understanding. We will argue that mechanistic explanations of physical laws can work in at least three different ways:
What makes for a scientific explanation? Among philosophers of science, two answers are vibrantly discussed at the moment. The first is that explanations describe mechanisms underlying, producing, or implementing the phenomenon to be explained (e.g. Craver 2007, Craver & Darden 2014). According to this mechanistic view, if scientists want to explain a phenomenon they must discover its mechanism. The other prominent suggestion is that scientific explanations are essentially causal explanations that tell us what would have happened if things had been different (e.g. Woodward 2003). To find these explanations, proponents of the interventionist view suggest, we carry out (under suitable conditions) systematic manipulations of some factor X to observe its effects of another factor Y. Interestingly, there is a connection between mechanism and interventions: proponents of both views argue that scientists need to employ interventionist methodology to discover mechanisms and thus to come up with mechanistic explanation. But how does this square with the apparent contrast between mechanistic explanations and interventionist causal explanations?
In a recent paper, Jon Williamson (2013) usefully highlights that mechanistic and interventionist explanations are quite different in character. An interventionist explanation requires knowledge about which factors can potentially make a difference to a given outcome (the explanandum phenomenon). Often we can have this knowledge without knowing how or why, say, ingesting an antibiotic drug will lead to recovery from streptococcal infection. By contrast, a mechanistic explanation requires mechanistic knowledge about the precise entities and activities at work. For instance, to explain recovery from streptococcal infection mechanistically, we would have to know something about bacterial metabolism and how antibiotics interfere with it. In this context Williamson makes another interesting observation: at times we are more willing to accept a mechanistic explanation as explanatory than one that is based merely on difference-making knowledge. And indeed if we look at scientific practice this seems correct: explanations supported by detailed knowledge about the mechanisms at work seem “more convincing” or “more progressive” than those just based on statistical regularity. This is evident across scientific disciplines. In biology, for instance, explanations of traits making reference to specific genes and how they may be modified are considered superior to those just referring to dominant or recessive chromosomal or autosomal inheritance. Likewise, we can look at cognitive science. Here, well-established psychological findings are replicated in droves with modern neuroimaging techniques, cellular recordings, etc. to support existing psychological theories with evidence about the underlying neural mechanisms. But does this mean mechanistic knowledge is generally superior to knowledge about difference-making relations? And how could this be if we cannot have mechanistic explanations without interventionist explanations? What is the relation between these two kinds of explanation and the two kinds of knowledge they seem to require, after all?
In this paper, we will argue that both the mechanistic and the interventionist view reflect important aspects of scientific practice. Looking at examples from empirical research in neuroscience, we will investigate the relations between interventionist methodology and mechanistic discovery and the explanations they yield.
Philosophy of science offers a rich lineage of analysis concerning the nature of scientific explanation. In recent years, considerable attention has been directed toward the notion of mechanistic explanation, especially in the biological sciences (see, for example the writings of Bechtel and Craver). Much of this work aims to characterize mechanisms or determine how mechanisms need to be described in order to be explanatory. Some of it aims to make explicit contact with contemporary analyses of causation, especially those of Salmon (1984, 1998) or Woodward (2005). This paper examines another scientific context in which appeal to mechanisms is arguably as widespread and central as it is in biological contexts but which has received much less attention: explanatory patterns involving reaction mechanisms in organic chemistry. There are two fundamental aims: (1) to develop a characterization of mechanisms in chemistry as a comparison case for existing analyses of mechanism in the biological sciences, and (2) to use this comparison to highlight certain aspects of explanatory practice across the sciences.
Drawing on recent work by Goodwin (2011, 2012), the paper begins with a general characterization of reaction mechanisms and their role in explanations in organic chemistry. From this characterization, I will argue that mechanistic explanations in chemistry seem different in important respects from their counterparts in biology. Mechanistic explanations in chemistry typically focus on information often lacking in biological cases, specifically information concerning the rate of operation of (some, but only some) various processes that compose a given mechanism, and at the same time, typically omit or suppress information included in biological mechanistic explanations. The next step of the analysis turns more specifically to practice. I argue that the types of information included in chemical mechanisms, as well as the way in which this information is represented in relation to potential energy diagrams, serves to support the largely synthetic aims of organic chemistry. I will suggest that general differences in the aims of given scientific communities may influence what sort of description or information is required for a mechanism to judged explanatory. Finally, I will return to broad issues concerning scientific explanation, arguing that an account of explanation sufficiently oriented toward explanatory practice will be best suited to make sense of the sorts of differences we observe in comparing chemical and biological mechanisms taken to be explanatory by their respective communities. Such an account of explanation stresses the methodological role of explanatory discourse in ways I have discussed elsewhere and will summarize briefly to conclude.