Mechanisms and the unity of science

Phyllis McKay

University of Kent

Craver

Craver’s Thesis

• Craver rejects classical approaches to the unity of science as reduction.

• He argues instead: ‘different fields integrate their research by adding constraints on a multilevel description of a mechanism. Mechanistic integration may occur at a given level or in the effort to build a theory that oscillates among several levels.’ (p373)

Typical classical reductionism

• Rejecting view that unity of science comes from reduction, from collapsing the laws of ‘higher’ levels to those of ‘lower’ levels to finish in the grand unified theory of everything.

• Classical model of reduction: ‘one theory is reduced to another when it is possible to identify the theoretical terms of the first with those of the second and to literally derive the first from the second.’ (p374)– Original classical view of reduction you deduce the higher-level

theory from the lower-level theory.– Later classical accounts, you derive an approximation of the

higher-level theory from the lower-level theory.– Accepted that the vocabularies of the reduced and reducing

theories will be quite different, so we require something like bridge laws to link the two.

Typical classical reductionism (p376)

1. Intertheoretic relationship– Relation between theories, not items in the world.– Focuses attention on vocabularies and meanings.– Doesn’t exist within a single theory.

2. Interlevel relationship– Theories about higher-level phenomena are reduced

to theories about lower-level phenomena.– Levels relation understood as a part-whole relation.

3. Formally specified– Presupposes a formal analysis of theory structure.

Craver’s Argument

• Development of Learning and Memory research program reveals three limitations of reduction models:

1. ‘they neglect upward-looking aspects of interlevel interfield integration’

2. ‘they ignore intralevel forms of interfield integration’3. ‘they gloss over the fact that scientific progress has sometimes

been achieved by abandoning reduction as an explanatory goal.’ p375

• ‘peripherality is the primary drawback of reduction models: they ignore most of what is interesting about the development of the contemporary LM research program and the integration of fields in the search for mechanisms.’ (p377-8)

History of reduction in LM

• Standard scientific history of the discovery of LTP: ‘suggests that researchers started out believing that the hippocampus was an organ of memory and then went searching for evidence of plasticity in its synapses. The history then depicts progress in this research program as proceeding to ever deeper and ultimately molecular levels. But as we will now see, this story is a rational reconstruction, not an accurate history.’ (p380)

History of reduction in LM

• 1950s electrophysiologists produce and report synaptic plasticity in the hippocampus.– Didn’t then last longer than 10mins and not connected by

scientists to learning and memory.– At the time, the hippocampus was not associated with memory.

Were looking at it because eg suspected role in epilepsy.– Anatomists and neurophysiologists had different techniques for

studying the neurons.• LTP only became very gradually associated with learning

and memory after discovery of long-lasting potentiation. But mechanistic shift ‘involved coming to see LTP not as identical to memory or as a kind of memory, but rather as a component in a multilevel memory mechanism.’

History of reduction in LM

• Dislodges 3 reductionist assumptions about the LM research program

1. In the 1960s, the development of the LM research program was upwards-looking: looking for a way to connect this synaptic phenomenon into a higher-level mechanism. Current work looks both upwards and downwards.

2. Interfield integration at a single level can be important – electophysiologists and anatomists use different techniques to study the cells of the hippocampus.

3. The LM research program abandoned reduction as an explanatory goal and adopted a mechanistic approach.

Craver on mechanistic integration

Integration of neuroscience

• ‘Understanding the structure of contemporary neuroscience requires understanding how these multiple fields [anatomy, biochemistry, computer science, developmental, evolutionary and molecular biology, electrophysiology, experimental psychology, ethology, pharmacology, psychiatry, radiology…], embodying distinct perspectives, techniques, and vocabularies, manage to integrate their work.’ (p374)

• Suggestion is that you should see all these disciplines as integrated in the LM project because they collaborate to work on a multilevel mechanism.

Intralevel integration• ‘Mechanistic theory building typically proceeds through the

piecemeal accumulation of constraints on the space of possible mechanisms for a given phenomenon. A constraint is a finding that either shapes the known boundaries of the space of possible mechanisms or changes the probability distribution over that space.’ Do that by identifying the relevant entities and activities and detailing their organization. (p388)

• (p388) ‘Scientists in different fields use different techniques to investigate different kinds of constraints on different components of the same mechanism. When the findings of two fields cooperate as constraints on a mechanism, the fields are, perhaps only rather locally, integrated.’– Eg anatomists and elecrophysiologists investigating different components

or stages of the same mechanism.– Craver notes: ‘The terms describing the different constraints are not

translated into one another. Nor are the different constraints identified with one another. Rather, the constraints open or close different portions of the space of possible mechanisms for a given phenomenon.’ (p388)

Interlevel integration

• Levels in LM include at least: learning/memory tasks – hippocampus generating spatial maps – synapses – NMDA receptor.

• The levels in such cases are levels of mechanisms. ‘levels are related as parts to wholes and … the parts are components in a mechanism.’ (p389)

• Interlevel integration:– Downwards-looking: detailing lower-level mechanisms for a

phenomenon. Character of phenomenon imposes constraints on any mechanism that will produce it, and description of mechanism must be responsive to changes in how the phenomenon is understood.

– AND Upward-looking: showing an item is a component in a mechanism and describing its role. Higher-level character of the phenomenon often must be accommodated to findings about lower-level mechanisms. Eg now recognise different forms of memory.

Interlevel integration

• See the mutual information between levels most clearly in experimental work. Interlevel experiments: ‘use the techniques of different fields to intervene into and to detect the activities of mechanisms at different levels.’ We wiggle the bottom and see what happens at the top, and vice versa. (p392)

• P393 ‘Findings at different levels constrain the links between levels accommodatively, spatially, temporally, and experimentally.’

• P393 ‘This gradual process of accumulating constraints on mechanisms at multiple levels in no way resembles efforts to translate one theory to another or to create a homomorphic image of one in terms of another. Instead, different fields elaborate the multilevel mechanistic scaffold with a patchwork of constraints on its organization, thereby revealing different hints as to how the mechanism can and cannot be organized.’

Issues, ramifications and worries

Interesting ramifications

• Have long thought of science in terms of the classical picture of reduction. Other ways to think about science:

• Right that fields mutually constrain – ‘upwards’ and ‘downwards’.

• In other work Craver rejects the idea of the world as organised in monolithic levels. Idea that the only levels are the levels of mechanisms is expanded. Only things in a single mechanism are regarded as related at all in terms of levels.

• Federica’s work on not only multilevel but genuinely mixed mechanisms as another reason for rejecting monolithic levels view. Need information from various disciplines to build explanations of some phenomena. See the way evolutionary biology is interested in some highly specific biochemistry.

• Levels of causation: the elephant and the flea.

A worry

• Craver is concerned with the explanatory unity of science. Interesting to consider implications for more traditional reduction worries.

• How are we to think of the relations between mechanisms in multilevel mechanisms? Or: what are we supposed to think about all the arrows?

What scientists say about their work

• ‘The main purpose of evolutionary biology is to provide a rational explanation for the extraordinarily complex and intricate organization of living things. To explain means to identify a mechanism that causes evolution and to demonstrate the consequences of its operation.’ (Bell 1997 and 2008 p1, emphasis added.)

• ‘Uncovering the cellular mechanisms resulting in sequential transfer of information from DNA (our genes) to RNA and then to protein represents one of major achievements of biochemistry in the 20th century.’ (Whitford p247, emphasis added.)

PS and NS

Protein synthesis and natural selection: a fair comparison

• Mechanisms and their sub-mechanisms are partially individuated in terms of their functions.– Protein synthesis is the mechanism for decoding DNA to produce

proteins.– Natural selection is the mechanism for adaptation.

• Components in protein synthesis (objects, their structures, and sub-mechanisms) are also partially functionally individuated.

• For natural selection to be a decompositional mechanism and yield the same kind of explanation of its phenomenon we are looking for functionally individuated components – NOT barely physically similar components.

Function and explanation in protein synthesis

• ‘In addition to the promoter-like elements, the Xenopus intergenic spacer contains repetitive short sequences, the 60/81 bp elements, that are clustered in tandem (Fig. 9.10) and which stimulate transcription when placed at a variable distance from the promoter and when placed in either orientation… . They are thus analogous in their activity, but perhaps not in their mechanism, to the enhancers of RNA polymerase II …’ (Adams et al. p361.)

• The obsession with structure in protein synthesis exists because in this field, structure is most often a good guide to function.– Note aminoacyl-tRNA synthetases example of an exception.

Protein synthesis and natural selection: a fair comparison

• Mechanisms and their sub-mechanisms are partially individuated in terms of their functions.– Protein synthesis is the mechanism for decoding DNA to produce

proteins.– Natural selection is the mechanism for adaptation.

• Components in protein synthesis (objects, their structures, and sub-mechanisms) are also partially functionally individuated.

• For natural selection to be a decompositional mechanism and yield the same kind of explanation of its phenomenon we are looking for functionally individuated components – NOT barely physically similar components.

The components of natural selection are also functionally individuated

• First division: directional selection, stabilizing selection, disruptive selection.

• Second division in directional selection: sorting, recombination, both sorting and recombination.

• Third division in recombination: structure of DNA and morphogenesis crucial. Consider gene linkage, epistasis and pleiotropy.

• Dozens of well-understood sub-mechanisms in the field.• Population structure the kind of structure functionally

relevant in this field.• At its lowest levels, evolutionary biology collides with

biochemistry in, for example, meiosis or autoselection.