Project 2: Godden & Baddeley (1975) Data Analysis

GENERAL INSTRUCTIONS: Store your work as a .m file named pj2_<YOUR STUDENT    ID>.m (replace <YOUR STUDENT ID> with your own 9-digit student ID number). Upload this  file to CCLE as your homework submission. Your entire .m file should run without errors. The  script should begin with the command clear and, when it completes, all variables listed at the end of this document should be in the workspace (extra credit is optional).

Background

The context-dependence of memory, or difference in memory performance across contexts, has now been widely accepted in cognitive science. Importantly, it is the concurrence or difference   in context between learning and recall, rather than the specific context at either learning or         recall, which is often most impactful on memory performance. In a famous study (Godden &       Baddeley, 1975), researchers presented scuba divers with a list of words to remember and later tested their ability to recall those words. Scuba divers studied the words either underwater or on land and were later tested either underwater or on land. This was a fully crossed design with a   total of four conditions:

1.   Learning on land (dry) and recall on land (dry)

2.   Learning on land (dry) and recall underwater (wet)

3.   Learning underwater (wet) and recall on land (dry)

4.   Learning underwater (wet) and recall underwater (wet)

Below I’ve put these conditions into a table to illustrate the design:

The researchers found that recall was highest when the study and test environment were the  same. If scuba divers studied the words underwater, they remembered the most words when  they were tested underwater. Likewise, if scuba divers studied the words on land, they            remembered the most words when they were tested on land. This is an example of a “cross-   over” interaction, in which the effect of one independent variable (here, recall environment)      reverses depending on the level of the other independent variable (here, learning                     environment). This cross-over interaction is evidence of context-dependent memory: when the recall environment is the same as the learning environment, recall is enhanced; equivalently,  when the recall environment is different from the learning environment, recall is impaired.

The term “context” in “context-dependent memory” is intentionally vague, and the general          finding of a cross-over interaction has been replicated across many definitions of “context” . One interesting example is “Alcohol and recall: State dependent effects in man.” (Goodwin et al.,       1969, Science), in which participants who studied while intoxicated performed the best on the    test when intoxicated. In particular, recall tends to me more context-dependent than recognition. I couldn’t find the raw data for this study, so instead I constructed a simulated data set which     reproduces the pattern of results from Godden & Baddeley (1975). This data file is on CCLE in  the Week 5 folder as godden_baddeley.xls. In brief, your objectives are to reproduce (most of)

Table 1 from Godden & Baddeley (1975) using the data provided in godden_baddeley.xls. For extra credit, your objective is to make a graph of the results.

Part 1: Reproduce Table 1 (cell array)

The first part of the project is to reproduce (most of) Table 1 from Godden & Baddeley (1975)     using the data provided in godden_baddeley.xls, storing your result as table1a. This table should be a 4 by 6 cell array. The cell array should only include this part of Table 1:

For your convenience, and because the first row is potentially ambiguous, here’s some code    you can use to start the table (keep this code as-is in your submission; these cells are done for you):

table1a=cell(4,6);

table1a (1,:)={'Learning environment' 'Mean recall score (dry)' ...

'S.D. (dry)' 'Mean recall score (wet)' 'S.D. (wet' 'Total'};

Reproduce the other labels exactly as above. Insert a character vector containing a single space ( ' ') into any cells in table1a which are empty (here, marked with – ) in Table 1 in Godden & Baddeley (1975).

Part 2: Reproduce Table 1 (table)

Also, store a similarly-designed table as a table object named table1b in MATLAB. For this  table, the values provided in the code above must be used as the column names; you’ll need to edit table properties to do this. The first variable may be a string array or a cell array of      character vectors, the rest should be numbers (type double), and any missing values (– )          should be NaN.

Both tables should have the same values (but possibly in different formats):

•    Text labels: 1 point

•    Empty cells: 1 point

•    Means: 4 points

•    Standard deviations: 4 points

•    “Total” row and column values: 2 points

Each point value is split between table1a and table1b, such that, if you only complete one of them, you’d get 6 of these 12 points.

Project 2 will require calculating two statistics for each group, mean and standard deviation,       which are calculated in MATLAB using the functions mean and std, respectively. For those who have taken 100A, note that the std function divides by n-1 by default so you can use its output as your estimate of standard deviation. Unlike in Psych 100A, you don’t need to know the           formulas, but knowing what these statistics are can help you determine whether the values you

obtain from the data are reasonable. Here are some Simple English Wikipedia (a highly valuable resource for everyone) pages for both:

Mean:https://simple.wikipedia.org/wiki/Mean

Standard deviation:https://simple.wikipedia.org/wiki/Standard_deviation

Both the mean and std functions accept a vector input and return a scalar output. I recommend using the split-apply-combine workflow described in W3P3 lecture (functions findgroups and splitapply), but you can calculate these statistics some other way if you’d like. The marginal values (“Total” row and column in Table 1) are sums of the corresponding cell values in the       table; for instance, the cell for row “Total” in column “Recall Environment: Wet” is the sum of the cell in row “Learning Environment: Dry” , column “Recall Environment: Wet” and the cell in row   “Learning Environment: Wet” , column “Recall Environment: Wet” .

PLEASE NOTE: The values in table1a and table1b will not match the values in Table 1 in Godden and Baddeley (1975). Instead, the pattern of results for the means (rank order of the means in the four conditions) should be the same as in the paper.

You should round your numbers to one decimal place, as in Table 1 in the paper. Otherwise, you don’t need to match the other formatting aside from capital letters on words and spaces between words.

Project 2 Extra Credit: Bar Graph with Error Bars

In relation to graphs versus tables, my own personal experience with reviewers of academic       journals is that it depends on the expectations of the field. Some fields, journals and reviewers   expect bar graphs with error bars for this kind of 2x2 design, some hate them and will ridicule     you (but probably not reject outright if you include one) , and likewise some fields and journals    will feel the same about a statistical table like what you’ve already created and ask that you put  things into bar graphs with error bars. It just depends. In any case, graphing data, especially this kind of 2x2 design, is an important skill in academia and the private sector, to the extent that (I   believe) you need to make this kind of graph in Psych 100B. In that class, you’ll use Excel (I       think), but here, for extra credit, you’ll use MATLAB.

For extra credit, adapt the code from W4P3 Slide 30 to graph your data. Your graph should look pretty much the same as this graph, with more specific details provided below the graph (3        points):