Why I Chose the Paper
First of all, the paper provides insights on the interaction of memory and vision. In addition I was able to learn very interesting facts from the four research experiments as well as previous findings elaborated in the literature review or background of the study section. Such an interesting finding is one from previous research: that semantics influences location of objects (objects that are semantically consistent are found faster compared to the ones that are inconsistent.
From their own research particularly the last experiment I was drawn to the finding that even when things remain in the same position, people rely on visual stimulus instead of the available memory of location of such items. For example instead of extending a hand to pick a common item like a phone which has been in the same position, it was found that individuals look before reaching out for the phone probably due to the suspicion that although the world has been stable, anything could happen.
Experimental Questions/Hypotheses Addressed in the Paper
- Vision first: observers might always choose to perform a visual search, even if memory search is possible
- Inefficient memory: memory search and visual search might be of similar efficiency giving no advantage to memory search.
- Pragmatic choice: repeated memory search might be efficient but might impose other costs that favor visual search.
Importance of the Research Questions/Hypotheses
Question one draws its inference from habitual tendencies: that observers rely on visual display. This enabled the researchers to configure panoramic search which is a wise choice as it allows concurrent assessment of visual search and memory search strategies.
Question two allows for repeated memory search in order to see if it’s efficient or not. It guides the researchers in providing hundreds of trials in order for them to draw valid and reliable inferences.
The third question considers the other side of repeated memory search: that it might be efficient. It provides the antithesis to question two and seeks to investigate the reason behind inefficiency of memory search.
The following experiments were performed to answer the research questions
Wolfe et al conducted four experiments
- Experiment 1 used static stimuli where observers confirmed the presence of an object in a scene with they became familiar. The configuration of objects within a scene never changed. The experiment was aimed at replicating the pattern of performance observed with previous repeated search displays. The task was to respond to the presence of a visible object in an unchanging scene. In the unrepeated condition, observers searched for a new object in a novel scene on each trial.
- In experiment 2, using panorama procedure Wolfe et al had observers respond ‘yes’ if the probe item was present. If it was present it was always in the visible set. Observers responded ‘no’ if the item was not present in the scene at all. Hidden items were never probed in the experiment. Observers could perform either a visual search or a memory search, because both sources of information were reliable.
- In experiment 3, observers responded ‘yes’ if the probe item was present in the visible set, ‘no’ if it was not present in the scene at all. In this case observers had to consult the visual stimulus, because a memory search would not necessarily produce the correct answer. Compared to experiment 2, experiment 3 could reveal interference of memory on visual search.
- In experiment 4 , observers responded ‘yes’ if the probe item was present in the visible set, ‘yes’ if it was in the hidden set, and ‘no’ if it was not present in the scene at all. In this case, observers had to consult memory because a visual search, by itself, would not necessarily have produced the correct answer.
Experiment 1: with minimum error range, a 2 factor within subject analysis of variance (ANOVA) performed on the correct mean response time of present trials showed that observers responded faster in the repeated condition (626 ms) than in the unrepeated condition (834 ms). Results indicated that although participants were faster in the repeated task, repeated search slopes remained inefficient even after 200 searches through the same, static scene. Concerning the effect of memory it was found that search patterns that incorporated memory did not alter the pattern of results in any way.
Experiment 2: search performance was slightly more efficient in the panorama procedure (slopes of about 20 ms/item). The relatively steep panorama slopes indicates that that the participants did not make use of an automatized visual memory in this task. Faced with a task that permitted either visual or memory search, observers seem to have searched the set of visible objects rather than based their responses on memory.
Experiment 3: the goal of this experiment was to force observers to perform a visual search. The results suggest that observers did perform a visual search as expected, because all classes of respondents were dependent on the visible set size. Observers were asked to determine if the probe was in view. The most important finding is that there was a strong visual component to the completely absent trials, even though memory would have sufficed for those trials. The experiment showed that observers could be coerced in to performing a visual search even when a presumably more efficient memory search would have been adequate.
Experiment 4: the results of experiment 4 are broadly consistent with an automatized memory. The visible target-present trials had a significant slope in the first half that vanished in the second. This was the only significant change and it suggests that responses were dominated by the visual information in the earlier trials but not later ones.
Generally, results of the four experiments argue that observers make a pragmatic choice and memory, with a strong bias toward performing visual search even in the presence of well known visual stimuli.
Did the Results Answer Research Questions?
Two of the three hypotheses that were proposed at the beginning were falsified. The inefficient memory hypothesis was falsified by the efficient search where the probe was either hidden or an absent item. The search became “automatic”, mirroring the results of other well-learned, consistently mapped memory searches. The vision first hypothesis was falsified by efficient search for visible target probes. Given that visual searches of this sort did not become efficient after a few hundred trials, Wolfe et al concluded that the response to visible probes became independent of the visible set size because observers chose to rely on the reliable, consistently mapped memory search rather than a variably mapped visual search. The same effect was reflected in the priming benefit when observers turned to memory in experiment 4, hidden trials benefited much more from identical twins of trials than did visible objects
However results of experiment 4 are consistent with the hypothesis that observers made a pragmatic choice to use memory rather than vision in this task. Initially, the responses to visible targets showed evidence of a dependence on the set size, as if observers were searching though the display. After conducting many versions of repeated visual search, the visual search slopes became shallow in experiment 3 but flattened in experiment 4.
Suggestion for Alternative Experiments.
In my view there could have been another experiment to investigate the reluctance of observers to use a reliable, efficient memory search. The researchers used panoramic procedure to show evidence of reliance on visual stimuli even where observers made a quick decision (within 500ms). Another experiment would have been set up to accommodate behavior that takes more time. This, in my view, would address the real reason behind falling back to visual search even when memory search is adequate. It would in essence clear the air whether observers rely on memory when making quick decisions or it happens even for when time is longer.
Validity of Conclusions
In as much as I desire alternative experiments, I believe that the paper has to a great extent made valid conclusions and that they are well supported. To begin with, the conclusions tend to agree with earlier studies. Results of panorama experiments agree with those of Hayhoe et al; that search for objects in real scenes involves a great deal of visual checking of information even for well trained tasks. Their findings were also consistent with contextual cuing work of Chun and his colleagues; that observers implicitly learn to direct attention to the target if it appears in a specific contextual display.
The fact that some hypotheses were falsified validates the researcher’s conclusions. Falsification of the hypotheses was done after a careful and thorough analysis of results (through statistical and tested means).
Lastly conclusions are arrived at in a logical process after a thorough examination of results.
Did I Like the Paper?
I think the paper is superb due to its structure and content. I particularly liked the informative nature of the paper in results and discussion sections. Concerning the structure I admired the logical nature in which arguments are presented. After displaying assumptions made by researchers, experiments that they performed are explained after which results are logically put forward. Simply put it is easy for a reader to get evidence for every proposition made or refuted.
In as much as the paper is well written I however think that efforts could have been made to simplify the mathematical details which I guess would be difficult to comprehend to those, like myself, that do not bear a mathematical background. After presenting such figures, the authors could have in my view interpreted them having readers of diverse background in mind. It is a great paper though.