Flashcards in Week 2 Multiple Objects Tracking Deck (13):
According to Pylyshyn and Storm (1988) how does the basic serial model of multiple object tracking operate? How does it allow people to track multiple objects simultaneously if they can attend to only one object at a time? What happens when a target is attended and how do tracking errors occur?
The basic serial model of multiple object tracking assumes that attention is directed to each target in turn. When a target is attended, its location is remembered. When it is time to reattend a target it is assumed that the object currently nearest to where the target was when it was last attended is the target. If this assumption is incorrect, an error will occur.
To replace serial models of tracking Pylyshyn & Storm (1988) proposed the FINST model. Does this model propose that tracking occurs in a serial or in a parallel fashion? How does this model work, what are FINSTs, and what do they encode? According to this model, what determines the number of object that can be tracked simultaneously?
In place of serial models, Pylyshyn & Storm (1988) proposed FINST theory. FINSTs are pre-attentive mental indexes/pointers that track the locations of the targets independently and in parallel. These FINSTs do not encode any other information other than location information. The number of objects an observer can simultaneously track is determined by the number of FINSTs the observer has. An observer is thought to typically have 3-5 FINSTs, so can simultaneously track that number of objects.
What theory of tracking did Yantis (1992) propose and what evidence supports his proposition? In particular, how does this theory propose that multiple objects can be tracked simultaneously with just a single focus of attention? What experimental finding supported their theory?
Yantis (1992) proposed that humans group all the targets into a single virtual polygon, with the targets at the vertices of the polygon, and track the targets by tracking this virtual object. This allows multiple objects to be tracked even though the human is attending to only one (virtual) object at a time. In support of his proposition, Yantis (1992) reported that it is easier to track objects when they move as a rigid group, which makes it easier to group them together to form a single virtual polygon.
Alvarez & Cavanagh (2005) investigated the extent to which tracking could occur independently in the left right visual hemifields. What did they find? To what extent do their findings argue in favour or against serial models of tracking? To what extent are their findings consistent with the Yantis (1992) account of tracking, which assumes that all the targets are grouped together to form a single virtual polygon?
Alvarez and Cavanagh (2005) found that observers could track two targets simultaneously in the left and right visual hemifields but could not track four targets in either the left or right visual hemifield. They concluded that observers track independently in the left and right visual hemifields. They suggested that this was strong evidence against a serial model (at least one with a single focus of attention). This finding also argues against the Yantis grouping account, which assumes that there is only a single focus of attention, as opposed to one in each visual hemifield.
Alvarez & Franconeri (2007) investigated how the maximum speed at which objects could be tracked varied with the number of objects that were tracked. What did they find? Did these findings argue for against the Pylyshyn and Storm (1998) FINST model, which assumes that people have a fixed number of pre-attentive mental pointers known as FINSTs.
Alvarez and Franconeri found that the maximum speed at which targets could be successfully tracked continuously decreased as the number of targets to be tracked increased. According the FLEX model, the faster the object moves, the more resource is needed to track the object, so the fewer objects can be tracked. This argues against the FINST account which assumes that the maximum number of targets that an observer can track should be determined by the number of FINSTS the observer has and not by the speed at which the targets move.
How did Franconeri et al. (2010) determine whether speed per se affects tracking accuracy? What did their experimental paradigm keep constant and how did they achieve this? What did they find and what two models did this argue against and why?
Franconeri et al. (2010) took videos of an MOT display and played them back at different speeds, thereby ensuring that the total distance travelled by the objects was kept constant. They found that distance affected tracking accuracy but tracking speed did not affect tracking accuracy (providing the speed was not so fast that data limitations occurred). These findings argued against both the FLEX account and serial models because both models predict that speed per se should affect tracking accuracy.
In Pylyshyn (2004), at the start of each trial each target was briefly labelled with a unique number. At the end of the trial observer were asked to indicate the location of a particular target. What was found and what was the significance of this? Which model is their data most consistent with and which models does this data argue against?
It was found that observers were readily able to distinguish targets from distractors but often could not identify which target was which. This would seem to argue against FINST, FLEX and serial accounts since these assume that targets are tracked individually. FINSTs and FLEXs are meant to act as mental pointers identifying individual targets – so they would keep track of each target’s individual identity. These findings are more consistent with Yantis’s account which assumes that all targets are tracked as a single virtual polygon, thereby allowing them to be mixed up while simultaneously being distinguished from the distractors.
In an otherwise standard MOT experiment, Scholl et al. (2001) joined each target to a distractor with a solid line. What did they find? Which model(s) is their data most/least consistent with?
Objects cannot be tracked when targets and distractors are joined together by the line. Joining targets to distractor forces them to group together, making it harder to group targets together to form a polygon. It is consistent with Yantis’ grouping model and inconsistent with serial models (FINST, FLEX).
How did Scholl & Pylyshyn (1999) investigated to what extent observers could track objects through occlusions. To what extent were their findings consistent with the Pylyshyn and Storm (1988) FINST account?
Objects can be tracked through occlusions, but the findings were inconsistent with the FINST account. This would seem to require that we predict where the object will reappear. It seems unlikely that a preattentive (i.e. “dumb”) mental process like FINSTs would be able to do this. This is debatable but probably not.
Which theory of tracking did Yantis (1992) argue against and what are the confounds of Yantis' (1992) theory?
Not consistent to serial models and FINSTs (they act independently) because they do not take into account grouping. Might be confounded because it might not be able to track all objects by grouping, but grouping helps us to track all of the objects (in an event we lose track of one of the targets).
What did Kunar et al. (2008)'s study find and what was his proposition? Which models does this data argue against?
They had observers perform a standard MOT task except that they had to identify whether a disk turning red was a target or not as quickly as possible. The time taken to respond was recorded. He found that in the group that was simulated (engaged in a conversation vs. not) average RT was slower. In a second experiment, members of one group were asked to do a task that requires attention and effortful thinking vs. a repetition task. RT was even slower for the group that was engaged in an effortful task. This data suggests that MOT requires attention, so argues against pre-attentive accounts e.g. FINST and supports the flexible-resource model of FLEX.
Based on their modelling and their experimental results, why did Pylyshyn & Storm (1988) dismiss serial models of tracking? Why did they feel that serial models could not account for their data?
They had 3 serial models of tracking: recorded, predicted, and guessed. They concluded that for the serial models to actually account for the data, the fact that the observers could track 5 objects at a time, attention would have to be switched between targets faster than was plausible.