A convergence warning from lme4 does not by itself show that a model is wrong. One check that the lme4 authors recommend is to refit the model with every available optimiser through allFit and compare the estimates that matter. This post records the candidate publications found by a reproducible search of Scopus, Europe PMC and OpenAlex, and explains why the list can only be a starting point for screening.
This reproducible descriptive analysis uses annual Scopus counts to compare how often records in nine query-defined literatures contain a form of “speculate” in their titles, abstracts or keywords. The count is a proxy for explicit wording in indexed records and says little about the quality of the evidence.
Figures often change their visual language as an analysis moves from exploration to reporting. This post uses depictr to examine a simulated lexical-decision experiment, plot its mixed-model estimates, audit a figure and correct the accessibility problems that the audit finds.
Greenhouse gas emissions can be reported as raw totals or per-capita values, each answering a different descriptive question. Using the 2025 release of the Emissions Database for Global Atmospheric Research, the World Bank income classification and the Human Development Index, this post compares totals, per-capita figures, sectors and a descriptive association with human development.
Languages lay time out along different spatial axes and build number words in different orders. Studies with speakers of Turkish, Aymara, Mandarin, German, English and other languages examine associations between these habits and temporal or numerical tasks. The review finds effects that are context-sensitive or developmental, and no evidence that language rigidly determines thought.
Selecting words is only the first step in preparing a psycholinguistic experiment. This post uses lexsync to match high- and low-frequency items, balance them across lists, and export timed PsychoPy and OpenSesame scripts that send electroencephalography (EEG) triggers, plus a jsPsych version that logs event markers.
A simulated semantic-priming study shows how precision, sign errors, magnitude errors and singular fits can be examined alongside power before choosing a design.
A literature-search export does not preserve the decisions that produced it. This post uses scopusflow to define a search in advance, resume it from checkpoints and generate a PRISMA-S search record from the resulting data.
A theory written only in prose cannot be checked mechanically. This post represents a theory of modality switching as structured data, derives the conditional independencies it implies, and tests them against data simulated from the theory and from a rival account.
In 1957, the linguist J. R. Firth observed that ‘you shall know a word by the company it keeps’. The principle that words sharing contexts share meaning underlies generative AI, from early Latent Semantic Analysis to today's largest Transformers. This post traces that lineage with three interactive LSA-to-PCA visualisations in R, built from Reuters newswire, State of the Union addresses and IMDB reviews. They show where simple co-occurrence models succeed and where they fail, and the post explains how scale and the Transformer turned a modest insight into the technology behind ChatGPT. It then examines why LLMs, optimised for fluency, cannot be rid of hallucination entirely, and argues that careful prompt engineering is the best tool we have for steering a fundamentally heuristic machine.
How transparency, preregistration and open data are becoming the standard for credible research, and what the shift offers researchers, funders and the wider scientific community.
A transcription pipeline, written with the help of GitHub Copilot, runs OpenAI's Whisper on local machines or high-performance computing clusters. Hundreds of hours of sensitive interview audio can be transcribed this way without sending any data to the cloud, and each hour of recording now takes minutes to process.
A production-ready local transcription workflow leveraging OpenAI's Whisper models that addresses the limitations of cloud-based solutions through complete data sovereignty, unlimited scale, reproducible processing and advanced quality control, while maintaining GDPR compliance.
4authors-year-doi-url is a CSL style designed to be as compact as possible while retaining the three most critical pieces of information for a reference: who (authors), when (year), and where to find it (DOI/URL).
Understanding the interplay between speech and gesture is crucial for linguistic and cognitive research. The current prototype, available on GitHub, aims to automate the analysis of temporal alignment between spoken demonstrative pronouns and pointing gestures in video recordings. By integrating computer vision (via Google’s MediaPipe) and speech recognition (using language-specific Vosk models) using Python, the workflow provides enriched video annotations and alignment data, offering valuable insights into deictic communication.
Reducing the impedance in electroencephalography (EEG) is crucial for capturing high-quality brain activity signals. This process involves ensuring that electrodes make optimal contact with the skin without harming the participant. Below are a few tips to achieve this using a blunt needle, electrolyte gel and gentle wiggling.
Researchers often make participants jump through hoops. Because of our own blind spots, it seems easier to realise the full extent of these acrobatics when we consider the work of other researchers. In linguistic research, the acrobatics are often spurred by unnatural grammatical constructions.
Say, you need to set up a makeshift EEG lab in an office? Easy-peasy—only, try to move the hardware as little as possible, especially laptops with dongles sticking out. The rest is a trail of snapshots devoid of captions, a sink, a shower room and other paraphernalia, as this is only an ancillary, temporary, extraordinary little lab, and all those staples are within reach in our mainstream lab (see Ledwidge et al., 2018; Luck, 2014).
Electroencephalography (EEG) has become a cornerstone for understanding the intricate workings of the human brain in the field of neuroscience. However, EEG software and hardware come with their own set of constraints, particularly in the management of markers, also known as triggers. This article aims to shed light on these limitations and future prospects of marker management in EEG studies, while also introducing R functions that can help deal with vmrk files from BrainVision.
Electroencephalographic (EEG) signals are often contaminated by muscle artifacts such as blinks, jaw clenching and (of course) yawns, which generate electrical activity that can obscure the brain signals of interest. These artifacts typically manifest as large, abrupt changes in the EEG signal, complicating data interpretation and analysis. To mitigate these issues, participants can be instructed during the preparatory phase of the session to minimize blinking and to keep their facial muscles relaxed. Additionally, researchers can emphasize the importance of staying still and provide practice sessions to help participants become aware of their movements, thereby reducing the likelihood of muscle artifacts affecting the EEG recordings.