As we discuss in more detail here, at the core of our approach is the reliance on “functional localizers”, which ensure that we pick out the same brain regions across individuals, studies, and research groups. This ability to consistently refer to the same regions is key for establishing a cumulative research enterprise, as needed for ultimately understanding the computations that underlie language processing.
We also believe that in order to truly understand the nature of human language, it is critical to examine diverse languages. On the one hand, it has long been known that in bilingual speakers, different languages (including sign languages) activate largely similar brain structures. However, to the best of our knowledge, there has been little or no systematic investigation of neural differences between speakers of languages varying along different key dimensions. For example, some languages rely heavily on word order while others have flexible word orders, some languages have little or no morphology while others have complex systems of morphological markings, some languages use tones to convey differences in meaning while others do not, etc. More recently, it has been further suggested that languages may vary with respect to how complex their syntax is (e.g., Everett, 2005; Jackendoff & Wittenberg, 2014; Futrell et al., 2016). Do brains of speakers of languages with different properties process language differently in some way?
Inspired by this question and the more general desire to examine languages beyond English, we have begun developing language localizers for a diverse set of languages, which we will be making available to other researchers here.
Description of the Materials
We are using Lewis Carroll’s “Alice in Wonderland”. We are taking advantage of the fact that this book has been translated into dozens of diverse languages. Our materials consist of 28 short passages (15-30 sec long) and 3 long passages (~5-6 min long). We chose to use auditory modality in order to make the localizers child-friendly as well as useable in illiterate populations. All the materials are recorded by a female (again, for child-friendliness reasons) native speaker.
The localizer contrast relies on the intact speech > acoustically degraded speech contrast. For each language, the download directory will include: i) 28 short passages, ii) the degraded versions of the 28 short passages, and iii) 3 long passages. For details on the validation of this contrast and showing its similarity to our more traditional contrast (sentences > nonword sequences), please see Scott et al. (2016). There, you can also find details on the procedure we use to create the degraded versions of the materials.
The texts of the English materials can be downloaded here: [long passages; short passages]. Please note that in some languages some of these passages were missing or translated to be quite different in content. When you download the materials for whatever language, please make sure to check the Info_on_x.txt file (where x is the relevant language), which will contain notes on these kinds of discrepancies.
A general MATLAB Psychtoolbox script that works with materials in any language is available for download here. Instructions on how to run the script, including audio file setup, is in the documentation at the top of the file "AliceLocalizer.m". You must have MATLAB 2013b or later with Psychtoolbox installed.
The intact versions of the materials are
Each language contains 3 long passages (~4-5 minutes in length) and 28 short passages
(~12-30 seconds in length). The degraded verstions of the materials
are available here.
The degraded file names are labeled accordingly:
degr_file#_Language, where (e.g.)
degr_file1_Armenian = Armenian_Long_1
degr_file2_Armenian = Armenian_Long_2
degr_file3_Armenian = Armenian_Long_3
degr_file4_Armenian = Armenian_Short_1
degr_file31_Armenian = Armenian_Short28