David Shanks
David Shanks is an experimental psychologist at University College London and a long-standing critic of claims about unconscious influence on behaviour. He published the 2017 reply showing that Payne's data are reproduced by a model containing no distinction between automatic and controlled processing, and he is a co-author of the 2015 thirteen-experiment failure to reproduce the disfluency-and-reasoning effect.
Who they are
His characteristic move is not to say that a result is wrong but to show that it does not require the explanation attached to it. The 2017 reply is the cleanest example. The data are not disputed; what is disputed is whether they license a claim about two separable processes. A model that contains no such distinction reproduces the same pattern, which means the data underdetermine the theory. That is a different and stronger form of criticism than a failed replication, because it applies even when the result holds.
The 2015 work is the other kind. Thirteen experiments testing whether making text harder to read improves reasoning, and thirteen experiments finding nothing. Effects of that shape were widely reported and widely taught before the multi-experiment tests were run.
Both are statements about what published data support. Neither is a statement about the researchers whose work is being re-analysed, and this site describes them as published scientific disagreement, which is what they are.
Key papers
Shanks, D. R. (2017). Reply showing that Payne's data are reproduced by a model containing no automatic and controlled processing distinction. The underdetermination argument in its clearest form.
Shanks, D. R., and colleagues (2015). Thirteen experiments finding no disfluency-and-reasoning effect.
Key ideas
Underdetermination. Data consistent with a theory do not establish it if a theory without its central distinction fits the same data.
A modelling reply is stronger than a failed replication in one respect. It bites even when the original result reproduces.
Thirteen experiments is the answer to one experiment. Multi-experiment tests are the level at which effects of this size have to be settled.
Criticism of what data support is not criticism of anyone. Keep the two apart when writing about this literature.
Sources
- Shanks, D. R. (2017). Reply on the automatic and controlled processing distinction.
- Shanks, D. R., and colleagues (2015). Thirteen-experiment test of the disfluency-and-reasoning effect.
- Vault note: [[David Shanks]], which records that this page is confined to what the cited work supports.