Observational Research Designs

  • Many studies in neuroscience are observations of a phenomenon

    • Scientists replicate, or observe the same phenomena under the same conditions multiple times, to prove the observation is not a one-off finding

    • After replicating, scientists extend their observations by testing the observations in slightly different situations

    • Observations are best when quantified, rather than in verbal descriptions, as they are more precise and reveal patterns

      • To quantify observations, scientists can use event sampling to count the number of times a behavior or event occurs during an observation period

    • There are two types of observations:

      • Structured observations occur when researchers manipulate a situation to make an event or behavior occur more

        • Ex: Making mice pull an “all-nighter” to test beta-amyloid levels

      • Naturalistic observations occur when researchers collect data from the animals’ natural environment

        • Ex: Collecting waste samples to test cortisol levels

        • Miniaturized devices attached to animals allow these observations to occur more precisely

  • Case studies provide an in-depth description of one individual

    • The birth of cognitive neuroscience came from the case study of H.M., who had the medial temporal area of the cerebral cortex removed

      • H.M. was no longer able to form new long-term memories, yet could make short-term memories, maintain motor memory, and access previously formed long-term memories

    • In a variant of the case study called the case-control design, a group of cases with a characteristic of interest is tested against a control group

      • This seeks to falsify data collected from the case group

    • A cross-sectional study measures individuals on the correlation between a risk factor and an outcome measured

      • Ex: Whether beta-amyloid in cerebrospinal fluid is associated with poor sleep

  • Correlational designs measure the relationship between two variables on a continuous scale

    • Ex: Change in brain volume and short-term memory

    • Scatter plots are useful when examining data as they show the relationship of the two variables for each case

      • Once plotted, scatter plots should be examined for the shape of the relationship (linear or curvilinear)

    • The Pearson product correlation coefficient determines the strength and direction of the linear relationship between two variables (sensitive to outliers)

      • It uses the variable r, which lies between -1 and 1, to determine if the relationship is negatively linear (~ -1), not linear (0), or positively linear (~ +1)

        • However, this value is sensitive to outliers, so the Spearman rank correlation coefficient is useful if there is concern about the effects of outliers

  • Correlation does not imply causation, so a third variable is looked for to determine if it causes the correlation between the two variables tested

    • Correlation can also change with an alteration of values

      • A restricted (limited) range can occur if the data is not examined wide enough

    • The square of the Pearson correlation coefficient (r2) is used to compare the coefficients

      • Variation of one axis is predictive of the location of the point on the axis

        • r=1 & r2 = 1 (square between axis intersection, the two points on the axes and the point on the line)

        • (%) variance of one plot is indicative of the variance of the other plot

  • Multi-variate analysis, or using multiple measures from animals, can help link brain activity and behavior