Infants have impressive auditory learning capabilities, even from the newborn stage. This study provides neuroimaging evidence that both newborns and 6–7-month-olds already learn and detect grammar-like rules, and that the underlying brain networks a…
Interesting that this study on infant brain development is not done with fMRI but with fNIRS (functional near-infrared spectroscopy [1]). And concerning to read about the complete reliance on software provided opaquely by a company for the analysis: "As all fNIRS data preprocessing was carried out on MATLAB-based toolboxes, no custom scripts are available."
[1] fNIRS measures more or less the same as fMRI: blood flow. "Using fNIRS, brain activity is measured by using near-infrared light to estimate cortical hemodynamic activity which occur in response to neural activity." https://en.wikipedia.org/wiki/Functional_near-infrared_spectroscopy
Why is it concerning that the analysis relies entirely on Matlab toolboxes?
1. Reproducing the analysis requires access to the exact toolbox with the exact version number. May also have platform dependencies (e.g., outputs slightly different results on macos, linux, and windows). 2. Often the version of the toolbox is not reported. It isn't in this paper. 3. Increased use of the toolbox may lead to price gouging, excluding many who can't afford to purchase it. Besides the unfairness, this impacts the ability of just about anyone to reproduce the analysis from the raw data. 4. After some years the toolbox may not be available at all, for example if the company considers its maintenance a burden with negative returns on investment.
If you are a reviewer or an editor: at the very least demand the reporting of the toolbox version and the operating system version where it was run. And ideally demand enough specification of the algorithms used by the functions provided in the toolbox that an independent reimplementation and reanalysis is possible.
Albert Cardona
in reply to PLOS Biology • • •Interesting that this study on infant brain development is not done with fMRI but with fNIRS (functional near-infrared spectroscopy [1]). And concerning to read about the complete reliance on software provided opaquely by a company for the analysis: "As all fNIRS data preprocessing was carried out on MATLAB-based toolboxes, no custom scripts are available."
[1] fNIRS measures more or less the same as fMRI: blood flow. "Using fNIRS, brain activity is measured by using near-infrared light to estimate cortical hemodynamic activity which occur in response to neural activity." https://en.wikipedia.org/wiki/Functional_near-infrared_spectroscopy
#neuroscience
optical technique for monitoring brain activity
Contributors to Wikimedia projects (Wikimedia Foundation, Inc.)Albert Cardona
in reply to Albert Cardona • • •Why is it concerning that the analysis relies entirely on Matlab toolboxes?
1. Reproducing the analysis requires access to the exact toolbox with the exact version number. May also have platform dependencies (e.g., outputs slightly different results on macos, linux, and windows).
2. Often the version of the toolbox is not reported. It isn't in this paper.
3. Increased use of the toolbox may lead to price gouging, excluding many who can't afford to purchase it. Besides the unfairness, this impacts the ability of just about anyone to reproduce the analysis from the raw data.
4. After some years the toolbox may not be available at all, for example if the company considers its maintenance a burden with negative returns on investment.
If you are a reviewer or an editor: at the very least demand the reporting of the toolbox version and the operating system version where it was run. And ideally demand enough specification of the algorithms used by the functions provided in the toolbox that an independent reimplementation and reanalysis is possible.
#reproducibility #academia