Archive for September, 2011


Explanation/Models in Chemistry

September 30, 2011

Julia Bursten

I want to describe how chemists reason when they are confronted with conflicting models of a given phenomenon. To meet this aim, I study the phenomenon of hypervalent compounds, that is, compounds whose central atom is bound to more than four ligands. Standard models used to predict molecular structure, such as VSEPR and Lewis structures, hold that four is the expected maximum number of ligands that can be affixed to a central atom. So hypervalent compounds are anomalous with respect to standard molecular-structure models, and if the models are to be preserved, this anomaly demands an explanation. Moreover, the fact that hypervalence cannot obtain in compounds whose central atom is in the first row of the periodic table suggests that there is a principled reason why hypervalence occurs in some compounds and not others. This reason should factor into a satisfactory explanation of the phenomenon.

Explanations of non-hypervalent, four-ligand molecules centered on atoms that can express hypervalence refer to the participation of s- and p-orbitals. Atoms have one s-orbital and three p-orbitals in their outer (valence) shells that are available to form bonding orbitals. Bonding orbitals accept electrons from ligand atoms to form bonds; hence, the expected maximum number of bonds around a central atom is four.
The received view among chemists has held that a central atom can express hypervalence just in case the outer electrons of the atom are of sufficiently high energy to involve d-orbitals as well as s- and p-orbitals. In such cases, the central atom can form up to seven bonding orbitals. This explanation of hypervalence involves direct reference to the systematic molecular-structure modeling tool of molecular orbitals, which is accepted as generally accurate for predicting molecular geometry and explaining relationships between geometric and energetic features of sets compounds. As such, it provides a strong basis for generalization and prediction, and it relates intimately to explanations of other bonding phenomena.

A competing view, which is rapidly gaining traction, holds instead that a central atom can express hypervalence just in case it is sufficiently large; that is, just in case there is enough room around the nucleus to fit in additional electrons. This explanation is simpler, and some research suggests it accords better with empirical data. It provides a rationale for why some atoms and not others can express hypervalence. And yet, it is somehow unsatisfying—it does not reveal new capabilities of the bonding model, nor does it confirm or disconfirm beliefs about the model. And it fails to provide as strong a basis for predicting molecular geometries of hypervalent compounds as the d-orbital explanation does.

So which explanation is better, for chemists? It turns out that neither explanation should be thrown out full-stop. The d-orbital explanation is more useful in predicting geometries and reinforcing the systematic relationships that underlie the bonding model, where the size explanation is more useful in predicting bond energies and accommodating empirical results. That both explanations survive is a testament to the diverse and context-sensitive explanatory needs of chemists, as well as to intricate and nonreductive relationships between theoretical models.


Is Psychology Mechanistic?

September 16, 2011

Catherine Stinson

This will literally be a work in progress talk where I present the current state of a (part of a) chapter of my dissertation. The wider context is a discussion of whether cognitive psychology and neurobiology can be integrated (unified, reduced one to the other, related by multi-level mechanistic explanation, …) and if so how. Whether cognitive psychology is a mechanistic science, and its particular way(s) of either being so or failing to be so has implications for its compatibility with neurobiology in an integrative project like cognitive neuroscience.

I’ll focus on cognitive psychology, and in particular, information-processing models. I’ll try to characterize information processing, relating it to functional analysis, computation, flow-charts, and maybe some other related notions, then compare examples of information-processing models to prototypical mechanistic models. Some nitpicking about what the definition of a mechanism should be will be necessary here; in particular I’ll try to clarify what a mechanism schema is.

If there’s time I’ll discuss a couple of recent/forthcoming papers by Piccinini and Craver, and Weiskopf that argue back and forth about whether psychological explanations are mechanistic, whether non-mechanistic or non-componential explanations can still be good explanations (yes, they can), and whether Craver-style unification can swallow up cognitive psychology (probably not).