Chapter 11 Progress and recommendations
The study of multiple object tracking has traditionally been done in the laboratory exercise with only a few dozen participants in an experiment. Only recently, then, has one of the greatest virtues of the MOT task become clear: it’s high test-retest reliability, of approximately .8 or .9, among the highest among attentional tasks and far higher than many common laboratory tasks (see chapter 9). The outlook for future work, then, is very positive both from the prospect of new results being credible (because with a non-noisy task, less data is needed to have high statistical power) and for the use of MOT as a tool to study other things, such as individual differences.
To further highlight how much we have learned, next I will describe how much findings have allowed us to disconfirm multiple aspects of the first and most influential theory of tracking.
11.1 The decline of Pylyshyn’s FINST theory
At the time of writing, Pylyshyn’s FINST theory described in the very first paper on MOT is the featured theory on the “multiple object tracking” Wikipedia page (Editors 2021), and it is frequently invoked in the scholarly literature as well. However, some of the main features of the theory have been rebutted (Brian J. Scholl 2008).
Core to the theory is the idea that tracking is mediated by a fixed set of discrete, preattentive indices. As we have seen, however, as speed increases, the number of targets that can be tracked rapidly decreases, to one target, which is hard to explain with a fixed set of discrete indices (G. A. Alvarez and Franconeri 2007; Alex O. Holcombe and Chen 2012).
Adding to the evidence that tracking draws on attentional resources are dual-task studies. I did not have space to review them in this book, but they indicate that tracking draws on attentional resources shared with some other tasks, and pupil data (Oksama and Hyönä 2016; Alnaes et al. 2014). One should add, however, that such studies do not seem to have ruled out the possibility that these findings were caused entirely by a C=1 resource rather than the hemifield-specific tracking processes; unfortunately, none of these studies appear to have tested for the hemifield specificity of their findings.
An important prediction of Pylyshyn’s theory was also that participants should be aware which target is which among the targets they are tracking. Pylyshyn himself found strong evidence against this. Again, I have focused on the evidence against Pylyshyn’s theory to showcase that the literature has come a long way in thirty-four years. This is also worth emphasizing because we share a problem of many other literatures, that large numbers of researchers know something about tracking but can be way out of date on theory and recent findings.
A salient case of more mixed progress is that while early work had concluded that MOT could not reflect serial position sampling (Z. W. Pylyshyn and Storm 1988; Yantis 1992), this recently has been questioned, and now is a favored account of tracking in some circumstances, as reviewed in Chapter ??.
11.2 Topics not covered by this book
Many topics were not covered in this book (I was given a word limit by the publisher). Some of the most important are whether tracking operates in a retinotopic, spatiotopic, or configural representation (see (Yantis 1992; Bill et al. 2020; Piers D. L. Howe, Pinto, and Horowitz 2010; Meyerhoff et al. 2015; G. Liu et al. 2005; Maechler, Cavanagh, and Tse 2021)), the role of distractors and possible distractor suppression, the role of surface features (Papenmeier et al. 2014), and the literature on dual-task paradigms, although a few such papers were mentioned along the way (e.g., G. A. Alvarez et al. 2005).
11.3 Recommendations
Here are a few recommendations for multiple object tracking researchers that have emerged from this journey through the literature:
To dilute the influence of C=1 processes, use several targets, not just two or three. But remember that even with several targets, a small effect could be explained by a C=1 process, so consider that. Test for hemifield specificity as that can also rule out C=1 processes.
Always test for hemifield specificity. In addition to it helping to rule out a factor having its effect only on C=1 processes, we know very little about what limited-resource processes are hemisphere-specific, so any results here are likely to be interesting.
- For computational modelling, I don’t recommend modelling only data from standard, fairly unconstrained trajectories, as such data may not constrain models enough. Show your model successfully mimics both spatial crowding and temporal interference effects, including that temporal interference is more resource-intensive than spatial interference. Another strategy is to model large sets of data at an individual trial level; previous efforts seem to have modelled data from only one or a few different MOT experiments.