Dear CSG members,
I’d like to draw your attention to a new article that Warren, Bruce Abbott and I have recently had published in the journal Attention, Perception & Psychophysics. Which I will briefly summarise below. The article can be found at the following URL: https://link.springer.com/article/10.3758/s13414-017-1398-2
The University of Manchester media relations office should be making a press release in the next couple of days which we’ll also post to CSGNet. We welcome your comments and discussion.
Many thanks,
Max Parker
···
The primary aim of the study was to directly and rigorously test the proposal put forward by Tom Bourbon that PCT models can be used to make idiographic predictions of individual performance. First we wanted to know whether
models characterised idiographic performance, so we measured individual differences in participants’ model parameters, and intra-individual consistency in participant’s model parameters. Second, to determine the degree to which these estimated model parameters
predicted model simulation accuracy. We then tested whether models simulated individual participant performance with new validation data and Finally whether models fit the data of the participant from which they were derived more accurately than they fit other
participant’s data.
We used the pursuit tracking task provided by the Living Control Systems III to which Bruce made some small adaptations, the targets to track moved in a pseudorandom pattern. We trained participants on the pursuit task with targets and then tested their
performance immediately following training, and one week later. A PCT model was optimised to the training data of each participant, yielding 20 models. These models were fit to the test data of the participants from which they were derived, and also to each
other participant’s data.
We found:
a) Estimated model parameters significantly predicted model performance, curvilinear relationship. Each parameter was a significant predictor, including the unique reference value parameter.
b) Significant individual differences in the estimated model parameters of each individual
c) Significant intra-individual consistency in all estimated model parameters
d) Models simulated the performance of the participant who’s training data they were optimised to with a high degree of accuracy. In fact, this ‘self’ simulation was significantly more accurate than when models simulated the performance of participants
who’s data they were not optimised to.
From: Warren Mansell [wmansell@gmail.com]
Sent: 29 August 2017 12:52
To: csgnet@lists.illinois.edu
Cc: Control Systems Group Network (CSGnet)
Subject: Re: Warren replicating Tom’s results
Thanks Bruce, I was planning to send that out when we put out a press release. Yes we cite Tom and the study extends his findings in a number of ways…
On 28 Aug 2017, at 15:58, Bruce Nevin bnhpct@gmail.com wrote:
[From Bruce Nevin (2017.08.28.10.58)]
It’s nice to see mention of
“Perceptual control models of pursuit manual tracking demonstrate individual specificity and parameter consistency”
The abstract describes a nice replication of results that Tom Bourbon demonstrated over a 20-year span (IIRC).
/Bruce