Reaction to Brady, Konkle, & Alvarez (2011) by Richard Thripp
EXP 6506 Section 0002: Fall 2015 – UCF, Dr. Joseph Schmidt
September 7, 2015 [Week 3]
Brady, Konkle, and Alvarez (2011) provide a thorough, though not exhaustive, review of visual memory research, broken down into convenient sections and subsections. Overall, two main sections regarding visual working memory and visual long-term memory investigate the ideas and research about various aspects of these systems, such as memory fidelity (p. 2–5, 13–15), basic units of representation (p. 5–7), interactions (p. 7–10, 23–25), and the effects of stored knowledge (p. 10–12, 15–19). Generally, these topics were treated separately with regard to working memory and long-term memory, which is a traditional distinction that is advantageous for conceptualization, but is of uncertain validity (p. 23–25). Based on an abundance of cited research, several themes emerged. With respect to working memory, capacity may be issue of both quality (fidelity) of memory and quantity of items remembered (p. 12). Structured representations and ensemble effects should be considered, meaning that information is stored in multiple and interacting layers (p. 10, 12). Long-term memory is surprisingly robust, especially with stimuli that are both semantically rich and real-world, supposedly because real-world scenery allows us to employ “passive episodic retrieval”, unlike “semantically impoverished stimuli” such as colored squares (p. 24). Overall, both working visual memory and long-term visual memory are more intricate and inter-dependent than once thought, which makes compartmentalizing any one subcomponent, and experimental research in general, highly difficult.
I found the results regarding long-term memory experiments interesting—I did not recall that our long-term memory is near-perfect for 10,000 items, as long as those items are unique and meaningful (p. 22). The authors did a good job of organizing the subjects and materials into numerous headings and subheadings, which made this review feel less onerous than many experimental research articles regarding cognition.
As a person with some computing knowledge, I liked the analogy of memory to a USB drive (p. 2), but found it constrained by technical inaccuracies. Saying that “the number of files that can be stored is limited only by the size of those files” with respect to a USB drive (p. 2) is inaccurate, given that it is constrained by the cluster size of the file system. USB flash drives typically use the FAT or FAT32 file system with cluster sizes of 4096, 8192, or 16,384 bytes (depending on drive capacity). Any file occupies the nearest higher discrete number of clusters—the author’s example of a 16 × 16 pixel image would typically take up at least 4096 bytes, even though the actual file would be 768 bytes or less. Conceptually, this adds a more complicated layer to the analogy that might actually be relevant to visual working memory—perhaps there is a lower bound on the space an item occupies, regardless of further reductions in fidelity? This characterization offers an appealing middle ground between continuous and discrete visual working memory models, which was not directly addressed by the authors (and may not yet be addressed in the literature).
Expanding on computing analogies, the authors missed a great chance to compare the concept of structured representations to progressive image rendering. Progressive JPEG images look blocky at first, and then gradually appear clearer as more data is received and decoded. This is similar to the idea of a “hierarchically structured feature bundle” (p. 7) where low-level features at the bottom level coalesce into a complete object representation through multiple, structured levels of data. Progressive image rendering shares remarkable commonalities with structured memory models (p. 10), and may even provide a conceptual framework to explore and develop visual working memory models.
Brady, T., Konkle, T., & Alvarez, G. A. (2011). A review of visual memory capacity: Beyond individual items and toward structured representations. Journal of Vision, 11(5), 1–34. doi:10.1167/11.5.4