Experimentation and Modeling: an example journal article

Geoffrey L. Collier and Charles E. Wright

Temporal rescaling of simple and complex ratios in rhythmic tapping

Journal of Experimental Psychology: Human Perception and Performance

Volume 21, No 3, pp 602-627

What do you read anyway?

Science is reported in many places. By the time something makes its way onto CNN, most of the science has gone. Even textbooks necessarily only offer established results, not cutting edge work. Current research is presented mainly in conferences (proceedings), books (collections of papers) and journals. Each of these forms of publication is subject to different levels of peer review, of which journal publication is (should be) the most stringent. The above article is cutting edge work which directly addresses many of the questions we are asking in this course. It is tough reading, even for initiates, so we will be taking it in small doses. You should be able to get at least the following out of the article:

Some facts are already known about how humans produce rhythms. What are the two relevant facts here?
Models of rhythm production need to account for these (and other) facts. Which models are considered here?
What predictions do the models make? What hypotheses do the authors form?
The authors carry out a series of experiments. What are they trying to show? We will look at Experiment 1 in detail. What do they tell us about the DESIGN of the experiment? Why is it designed this way?
What results do they obtain?
Do the results address the questions they set themselves beforehand? Do they help in designing a model?
What questions remain unanswered?

Structure of a journal article

Abstract. The quick version.
Introductory section. Previous work. Hypothesis formation.
Experimental section. Experimental design and procedure. Experimental results.
Discussion section. Pulling it all together.
Maybe, appendices and notes.
Bibliography. Further reading.

Introduction

How is the timing of motor performance controlled?
In particular, does adherance to natural timing patterns (?) determine whether the temporal structure (?) of a sequence is maintained when its overall duration is changed?

Models of rhythm production

There exists a continuum of models, differing mainly in how time is handeled. Time can be Ordinal (sequence matters), Relative (time measured with respect to something else) or Absolute (as measured by clocks of arbitrary precision).
They discuss the Generalized Motor Program with Proprotional Scaling (GMP/PS) model. This is a relative time model.

GMP/PS

Relative durations are prestored, while a single rate parameter specifies absolute durations at the time of execution.

Advantages:
1) Overall durations can be changed while relative durations remain constant (proportional scaling, tempo invariance).
2) Relative durations are unconstrained, therefore the model is very powerful.

Disadvantages:
1) Humans do not display perfect proportional rescaling. (This is mainly a technical point you don't have to worry about. Humans do pretty good in rescaling simple patterns).
2) Relative durations are completely unconstrained, yet humans have been shown to prefer some, simple, ratios to other, complex, ratios. A 2:1 ratio is simple, a 37:17 ratio is very complex.

What is a "natural" temporal pattern?

Hypotheses

Proportional scaling relies on the use of simple ratios.
- Simple ratios can be rescaled (known "fact")
- Complex ratios cannot be rescaled (hypothesis to be tested)

Problem!

It is known that complex rhythms are harder to produce in the first place. If they cannot be rescaled, how do we know whether that is because they cannot be produced properly at any rate, or because they cannot be rescaled?
Tentative solution: Train subjects for a long time on complex ratios, and see if their performance is much worse after rescaling.

Coming soon.....

Experiments and discussion......