Research Roundup: Published papers by CMU faculty

Article: “Fine-grained temporal coding of visually-similar categories in the ventral visual pathway and prefrontal cortex.”

Published: Frontiers in Psychology, October 2013

Who: Researchers in the departments of machine learning, psychology, and statistics at Carnegie Mellon, as well as researchers in the Center for the Neural Basis of Cognition and United States Air Force Academy.

Background: Humans can distinguish visually similar categories very well. A good example is the ability to distinguish between poisonous berries and ripe ones. Scientists refer to this phenomena as “subordinate-level categorization,” which is a higher skill than “basic-level categorization,” which is distinguishing between two obviously different objects, such as an airplane and a computer. If one becomes an expert in a particular subject, they perform subordinate-level categorization just as intuitively as basic-level categorization. An example the authors give is an expert bird-watcher who can tell bird species apart as easily as a layperson could tell a bird from a chair.

Hypothesis: The scientists were aware of two regions in the brain associated with the process of categorization: the ventral visual pathway (VVP) and the prefrontal cortex (PFC). However, current understanding of the mechanisms behind categorization is poor. The researchers proposed using magnetoencephalography (MEG), an imaging technique that can track brain activity, to test two leading hypotheses.

The first hypothesis emphasizes the role of the PFC and states the role of the VVP as helpful to visual identification but not categorization. The other leading hypothesis states that the PFC and VVP play complementary and nearly equal roles in visual processing and categorization. The Carnegie Mellon team hypothesized that there are necessary contributions from both brain regions.

The Experiment: The researchers created two similar “shape categories.” They devised a unique series of blob-like shapes that were in one of two categories, designated “A” and “B.” Although each blob was unique, blobs in categories A and B had distinguishing characteristics. The MEG monitored neural activity as the participants were trained in how to discriminate between the two categories. The MEG gave researchers a chance to observe what happens in the brain when participants discriminate between blob categories.

Results: The research suggested that the VVP is integral to discriminating between two categories of similar-looking objects. Surprisingly, the researchers found no evidence to support the idea of explicit coding of information in the PFC. The researchers admit that they used a fairly small sample size, and that further research is needed. Future applications of similar research could help determine the mechanisms behind visual procession and learning.

Article:" Source Water Changes and Energy Extraction Activities in the Monongahela River, 2009−2012"

Who: Jeanne Van Briesen, professor of civil engineering at Carnegie Mellon, and Jessica Wilson of the department of civil and environmental engineering at Manhattan College.

Published: Environmental Science and Technology, October 2013

Background: The Monongahela River runs through southwestern Pennsylvania and north-central West Virginia. Over the course of the 20th century, the river has been exposed to drainage from abandoned coal mines and waste-water from conventional oil and gas production. In addition, unconventional methods of gas extraction has dramatically expanded, especially in the Marcellus Shale gas reserve. This is of interest as the Monongahela River is the source for 17 drinking water plants that serve over 1 million people.

Hypothesis: The study looks at water properties pertinent to drinking quality between September 2009 and September 2012. The researchers aimed to determine if changes posed a threat to drinking quality and if the changes were due to seasonal variability or due to waste-water disposals of oil and gas.

The Experiment: Water samples were taken from six drinking water treatment plants on the Monongahela River. The samples were analyzed for total dissolved solids, pH, sulfate, chloride, and bromide concentrations and analyzed within two weeks of collection. The samples were 500 milliliters in size. The researchers also collected data on the seasonal flow of the river. This was necessary as they needed to compare concentration levels relative to flow rate.

Results: The researchers looked at both constituent loads, which are determined by multiplying concentrations of pollutants in the water by the flow, and constituent ratios, which compare concentration of pollutants to the flow. It was expected that the highest constituent loads would occur during periods of high flow and that similar constituent loads would occur during periods of similar flow.

Looking at constituent loads, they found that observed increases in bromide and chloride occurred during different time periods, indicating that the ions came from different sources. Accounting for the ratios, they found that overall bromide levels increased in May of 2010, and decreased in 2012. The researchers concluded that levels maintained at these concentrations would not pose a threat in drinking water. In terms of sources of ions, it would be nearly impossible to determine if ions came specifically from unconventional natural gas.