Working Memory is a critical component of human cognition and can be defined as the ability to store and process information simultaneously.
A complex span task is the commonly used tool by researchers and practitioners to assess working memory capacity; it does so by interleaving to-be-recalled material (e.g., digits, letters, visual patterns) with processing tasks (e.g., verbal, spatiovisual). One example would be to have participants read a series of sentences and deciding whether they were correct while remembering the last word of each sentence for serial recall at the end of each trial (Daneman & Carpenter, 1980).
However, research and clinical settings currently face a few issues regarding the use of complex span tasks:
- Whether they are measuring what they should be (i.e. working memory capacity); many studies operationalize the interleaved processing tasks as a binary yes/no response - this encourages random guessing and fails to maximize cognitive load.
- Multiple complex span tasks are often used in one single session, which results in increasing fatigue for participants.
The goal is to build a short, single task that measures working memory capacity. We aimed to move away from the use of large-number-trials, multiple span tasks to reduce proactive interference throughout successive tasks, as this can significantly compromise the accurate assessment of working memory capacity.
Softwares needed:
- PsychoPy (https://www.psychopy.org/download.html)
- R 3.6.1 (https://cran.r-project.org/bin/windows/base/)
- R Studio (https://rstudio.com/products/rstudio/download/)
Packages needed in R:
- Download and open the experiment in Psychopy: source("https://github.com/neuropsychology/ComplexSpan/blob/master/psychopy/task.psyexp")
- Click on the green button to begin the task
- Enter participant number and session number
- Run task
- Instruct participants not to press "Enter" at any point during the task unless required, as the experiment would quit if so
- Task outputs 3 files: a csv file, a text document, and a psydat file
This is a Psychopy implementation of a composite complex span task adapted from Gonthier et al. (2015). It comprises of 2 components, the simple span and the complex span:
Simple Span: Presentation of consonants to be remembered
- Consonants to remember
- Set size increases after 2 correct answers in a set size
- Simple span determined as set size attained after 3 incorrect answers
Complex Span: Components include the following:
-
Processing: Practice block containing open-ended arithmetic operations to respond to
- All correct answers are one-digit
- Feedback provided with each processing trial
- Processing block to repeat if cumulative score attained < 65%
-
Complex Span Practice: Practice block containing open-ended arithmetic operations and to-be-recalled consonants
- Consonant presented after each correctly-answered arithmetic operation (if incorrectly answered, a new arithmetic operation is presented)
- There is a time limit (mean RT + 2.5 standard deviations of the processing practice trials) on the processing trials where the next arithmetic operation is presented if no answer is given and the trial is recorded as incorrect
- Set size of 2 only
- Feedback provided with each processing trial and each recall trial
- Block ends after 3 consecutive correct recall
-
Complex Span Task: Test block containing open-ended arithmetic operations and to-be-recalled consonants
- Consonant presented after each correctly-answered arithmetic operation (if incorrectly answered, a new arithmetic operation is presented)
- There is a time limit (mean RT + 2.5 standard deviations of the processing practice trials) on the processing trials where the next arithmetic operation is presented if no answer is given and the trial is recorded as incorrect
- No feedback provided
- Start with set size of 2; set size increases by 1 with each correct recall
- Block ends after 2 consecutive wrong recall
Open R studio and start by running the following:
# Download the function used to preprocess the data
source("https://github.com/neuropsychology/ComplexSpan/blob/master/statistics/preprocessing.R")
source("https://github.com/neuropsychology/ComplexSpan/blob/master/statistics/processing.R")
Then, set your working directory to the source file location (Session > Set Working Directory > To Source File Location). Create a folder in your working directory called "data" and create the path for your output csv file, for example:
path_to_csv <- "./data/pilot_1.csv"
This will allow you to use the functions created to preprocess the datafile. You can then simply run:
preprocess_WM(path_to_csv)
# This creates a single dataframe for the participant comprising of the simple span and complex span task components.
process_WM(path_to_csv)
# This creates 3 dataframes:
# The participant's span scores and mean reaction time of the processing trials;
# The participant's proportion of correct recall, mean processing reaction time, and proportion of correct processing across set sizes;
# The participant's mean processing reaction time and proportion of correct processing across different distractors
To create figures, create a folder called "figures" in your working directory and download the make_figures.R file; running the code will save the figures in this folder.
# ---- Example Script ---- #
# > getwd()
# [1] "/Users/lauzenjuen/Documents/GitHub/ComplexSpan/statistics"
# Load packages
library(tidyverse)
library(easystats)
# Load functions
source("preprocessing.R")
source("processing.R")
# Combine Data Frames
data <- data.frame()
data_subject <- data.frame()
data_within <- data.frame()
data_distractors <- data.frame()
for(participant in list.files("./data/")){
data <- rbind(data, preprocess_WM(paste0("./data/", participant)))
processed <- process_WM(paste0("./data/", participant))
data_subject <- rbind(data_subject, processed$between)
data_within <- rbind(data_within, processed$within)
data_distractors <- rbind(data_distractors, processed$distractors)
}
data_subject # produces a table of scores across participants
The analysis script computes scores according to different scoring schemes as per Conway et al. (2005)
- Partial-Credit Unit Scoring (PCU): Mean proportion of elements within an item that were recalled correctly
- All-or-Nothing Unit scoring (ANU): Proportion of items for which all elements were recalled correctly
- Partial-Credit Load Scoring (PCL): Sum of correctly recalled elements from all items, regardless of whether the items are perfectly recalled or not (also without respect to serial order within items)
- All-or-Nothing Load Scoring (ANL): Sum of correctly recalled elements from only the items in which all elements are recalled in correct serial order
This task can be referenced as follows:
- Lau, Z. J., Pham, T. T., Makowski, D., & S H Chen, A. (2019). A Psychopy Implementation of the Complex Span for Working Memory Assessment. Retrieved from http://doi.org/10.5281/zenodo.3529329
@misc{complexspan2019,
title={A Psychopy Implementation of the Complex Span for Working Memory Assessment},
author={Lau, Zen J and Pham, Thanh T and Makowski, Dominique and S H Chen, Annabel},
month = {nov},
year = {2019},
doi = {10.5281/zenodo.3529329},
version ={0.0.1},
publisher={Zenodo},
url = {http://doi.org/10.5281/zenodo.3529329}
}
These graphs were produced by make_figures.R
Here are some other resources that may be helpful!
- Follow PsychoPy updates here
- Well-documented memory tasks