Adele Abrahamsen's Research
Since moving in 2002-3 to the University of California, San Diego (UCSD), I have been collaborating with William Bechtel on the new mechanistic philosophy of science that he pioneered in a 1993 book with Robert Richardson (reissued in 2010). Like most of the new mechanists we have focused primarily on cases in biology--e.g., chronobiology, cell biology, and cognitive neuroscience--but I have had a particular interest in extending this kind of philosophical analysis to cases in cognitive science involving such domains as language and memory.
What we now call basic mechanistic explanation provides an account of the organized parts and operations responsible for a phenomenon of interest. This is the initial focus in any mechanistic inquiry, but in the cases we have examined, scientists often then turn to the dynamics of the successfully identified parts and operations. In the last decade or so, for example, circadian researchers increasingly have been bringing computational modelers into their research teams or collaborations. We therefore proposed that dynamic mechanistic explanation should be recognized as a partnership of two approaches sometimes regarded as competitors: basic mechanistic accounts and computational models of dynamics. Coordination is achieved by linking properties of parts and operations (e.g., the rate of an mRNA transcription operation) to variables or terms in the model’s differential equations. (Cf. a freestanding dynamical model: the equations may be similar but, with no components identified, their variables characterize only the system and its environment.)
In 2010, William Bechtel and I added to our collaboration two graduate students in the philosophy department: Daniel C. Burnston and Benjamin Sheredos. Our WORking Group on Diagrams in Science (WORGODS) is examining the role of diagrams in scientists’ development of basic and dynamic mechanistic explanations. We have created and tagged a database of over 2000 figures from circadian research, and use it as a resource for identifying different types of diagrams and addressing the issues they elicit.
- Phenomenon diagrams: Since a phenomenon in need of explanation typically makes its appearance as a pattern found in data, it is important to find the formats that make each particular type of pattern especially transparent. Often these are the most familiar of formats: line graphs or bar graphs. While the overall phenomenon of circadian rhythmicity can be conveyed in a line graph by plotting activity against time, specialized formats such as actograms and phase response curves have been developed to make it easier to grasp and work with more specialized phenomena, such as the slightly shorter endogenous period revealed by imposing constant darkness. By tracing the lineages of such formats (a particular interest of Ben’s), we have seen how changes over time have made the target phenomena more transparent, more noticeable, and easier to manipulate mentally. See especially the article Diagramming phenomena for mechanistic explanation.
- Mechanism diagrams: This type of diagram conveys the proposed basic mechanistic explanation of the phenomenon of interest. A typical mechanism diagram uses glyphs, icons or labels to represent the parts of the proposed mechanism; arrows to represent the operations they perform; and layout in a 2D space to indicate their organization. In our Philosophy of Science paper Why do biologists use so many diagrams? we emphasized that one clue to the role of such diagrams in the reasoning of scientists is the frequent use of question marks to flag components whose identity or role is uncertain or unknown and hence requires further investigation. We also examined a case in which different scientists diagrammed the same mechanism in different ways, thereby advancing different ways of understanding how the mechanism generates the phenomenon.
- Augmented mechanism diagrams: When a dynamic mechanistic explanation is on offer, this type of diagram can show explicitly how a particular computational model is anchored in a particular basic mechanistic account by appending the names of variables (and sometimes parameters or terms including both) to the appropriate parts or operations. In circadian models, the variables typically correspond to the concentrations of molecular parts or the rates of the operations that increase or decrease concentrations. Such a diagram conveys the basic mechanistic account and makes it much easier to see how the computational model links to the basic mechanistic account, but it does not convey the computational model itself. Thus, it must be used in conjunction with the relevant equations. See especially the article Roles of diagrams in computational modeling of mechanisms.
- Explanatory relations diagrams: Like many phenomenon diagrams, these may take the form of line or bar graphs, but usually incorporate a comparison, such as a cyclic pattern over time in concentrations of the Cry1 protein for wildtype mice vs. a flat line for a mutant strain in which the Bmal1 gene has been knocked out. Although such relationships sometimes provide crucial evidence for particular parts or operations, Dan has argued that they frequently play a direct explanatory role. In this example, the timing of the cyclic pattern, along with that of other proteins, explains how the whole mechanism achieves a 24-hour cycle. Data graphs have received little attention heretofore in philosophy of science—even in our own initial characterizations of dynamic mechanistic explanation—but we now are seeking to fill that gap. See especially the article Representing time in scientific diagrams.
The other hat I wear is that of cognitive science researcher. In this work I have certain core concerns -- those aspects of cognition that most intrigue me -- but I come at those concerns from a variety of directions. What intrigues me is the onset and trajectory into adulthood of language and its well-orchestrated interactions with other cognitive capacities. My conceptual and research tools have come primarily from psychology but also from linguistics and from some specific, interdisciplinary research areas.
More specifically, my research on language and cognition has addressed early symbolic gestures and words in typically and atypically developing toddlers, the development of language—especially word meaning—at 4-10 years of age, adult psycholinguistics, and augmentative communication. Other publications have addressed interdisciplinary relations, cognitive science as an interdisciplinary field, connectionism, and learning across species in addition to basic and dynamic mechanistic explanation and diagrams. The links above provide access to these publications by date (some with PDF) and, alternatively, by topic.