Verbal and visuospatial short-term and working memory data across a 43-year period (1973–2016) worldwide: Flynn and anti-Flynn effects

Secular gain and drop in cognitive test performances over time have been observed and called respectively the Flynn and anti-Flynn effects. The current datasets include raw data from an investigation of the Flynn and/or anti-Flynn effects on verbal and visuospatial short-term and working memory reported in ‘The Flynn effect for verbal and visuospatial short-term and working memory: A cross-temporal meta-analysis’ (Wongupparaj, Wongupparaj, Kumari, Morris, 2017) [1]. Specifically, the datasets totally contain 1754 individual samples (n = 139,677) across a 43-year period from forward/backward digit span (F/BDS) and forward/backward Corsi-block span (CBS) tests. Mean memory test scores, standard deviation scores, types of memory tests, years of publication, mean ages, male percentages, types of publication, types of countries, platforms of memory tests, and sample sizes were collected and included in the datasets. DS and CBS data are unique in that they can provide a rich source of trends concerning changing short-term and working memory test scores across memory types, test platforms, age groups, gender, and countries. Further, these data can be of use for investigation of psychometric properties for the memory tests.

The obtained data can be used for estimating the Flynn and anti-Flynn effect on test interpretation across the additionally provided variables. The datasets will be of use to investigate changes on verbal and visuospatial short-term and working memory scores overtime and also provide a beginning point for conducting further meta-analysis or computing any updated metaanalytic estimates from 1973 to the present [1].

Data
The datasets consist of four SPSS.sav files that contains key parameters for FDS, BDS, FCBS, and BCBS tests. All four datasets generally share a similar structure and number of contained variables, that is, authors of each sample, years of publication, mean ages of each sample, male percentage in each sample, types of literature, types of country, tests platform, age groups, sample sizes, mean scores of memory tests, standard deviation scores of memory tests, standard error scores of memory tests, and variance scores of memory tests.
A complete list of references for included sample and studies across the four memory tests is openly available at https://osf.io/3wa94/

Data and literature search strategy
The search terms used were "digit span", "Wechsler's digit span", "forward digit span", "backward digit span", "Corsi block*" (* represents multiple spellings and endings), "forward Corsi block", and "backward Corsi block". These search terms were employed individually and in combination with the Boolean OR function in order to increase search sensitivity. The target literatures were systematically searched through 15 scientific databases that contained journals and/or theses and dissertations (unpublished materials or grey literature) from 1910 to 2016. The following scientific databases were BioMed Central, BMJ, Cambridge Journals, Frontiers in Psychology, HighWire Journal, JSTOR, Karger Online Journals, Oxford Journals, ProQuest dissertations and theses, PsycARTICLES & PsychINFO, Sage Journals, ScienceDirect, SpringerLink, Taylor & Francis Online, and Wiley Online Library. All search outputs from databases were exported to a reference management software, EndNote, and duplicates were then removed.

Inclusion and exclusion criteria
The final datasets for all memory tests were only from studies which reported the mean and/or standard deviation of raw scores from FDS, BDS, FCBS, and BCBS. Thus, samples were excluded if they reported the target parameters for age-scaled and standardized scores. Further, only mean and/or standard deviation scores and relevant variables for pre-test or baseline were collected if these studies adopted a test-retest method or a time series analysis. In addition, only mean and/or standard deviation scores and relevant variables were treated as a single data point for multiple studies with the same sample or shared datasets. Also, studies with clinical research participants were excluded with the exception if they reported data for health controls or control groups. The following samples were excluded because no target parameters were reported; book chapters, review articles, systematic reviews, meta-analyses, research protocols, and case reports.
Finally, outlier and influential case diagnostics were performed on all target parameters across all studies using Cook's distances, DFBETAS, and DFFITS. Accordingly, final datasets include 742, 594, 307, and 111 independent samples for FDS, BDS, FCBS, and BCBS respectively, resulting in 139,667 participants overall.

Data extraction
Two research assistants independently reviewed all relevant parameters across samples. The process of identification, screening, and eligibility check for including target parameters in the final datasets was conducted according to the PRISMA guideline [2].