Holger Mitterer

Mir fehlte die Option, dass das die Schuld der Grünen ist!

Going to be a Bayesian nitpicker here, re the body-image study. The fact that some tests were ns does NOT mean they were inconsistent with the hypothesis. For any study, if we have 50% power and half of the tests are significant, this is in fact very STRONG evidence for the underyling hypothesis.

This thread was useful (=close to maximal praise in German)

Why younger Maltese people are now rolling their Rs less A study of recordings of 60 Maltese speakers finds signs of a generation divide

There's an advantage to work in a small (really small, bro) country. It's not that difficult to make it into the main national newspaper:
timesofmalta.com/article/why-...

Bike cleaning day. If I bring out the bike stand, let's make that count.

We'll have a last minute poster at 'r-atics 8 that shows that Maltese listeners, who are used to both trilled and approximants rhotics, associate both with roughness to the same degree.

While everyone is on about how good AI is, Google Scholar suggests to me to add a publication to my profile called
"Max Planck Institute for Psycholinguistics".

I am honored to have (co-)authored the institue, but I will keep away from automatic profile updates for now.

I was using forced alignment short utterances from a speech-production experiment. Using recordings that had been noise-reduced, the results were unusuable, so I tried the original, noisy recordings. Those worked very well.

Can anyone here explain why noise-reduction makes forced-alignment worse?

YouTube video by Sabine Hossenfelder I was asked to keep this confidential

I want to read you an email that I was asked to keep confidential because I think it explains some of my worries about academia.

www.youtube.com/watch?v=shFU...

Did I mention that I do triathlon?
psycnet.apa.org/record/2025-...

This one is much better:

NLP = Natural Language Processing
NLP = Neuro-Lingusitic Programming

(just in case anyone is wondering, the latter, despite being called programming, has nothing to do with coding)

That piano break from MiTierra

Generting images for eye-tracking via API requests Hosted on the Open Science Framework

Most visual-world eye-tracking study need hundreds of images. That's quite a laborious task. Sine openAI can generate images (chatGPT + Dall-E), I managed to batch generate images via an R-script, which I shared via osf (osf.io/apcyb/). I hope this is helpful.

Fancy a trip to Malta during the best time of the year there?
- Submit an abstract to 'r-atics 8, to be held in Valletta 22-23 May.
- Extended deadline: 21st of January
www.um.edu.mt/research/cog...

Question on the ethics of using ChatGPT in science writing.

As an L2 speaker of English, I sometimes find a sentence not sounds well. In that case, I ask ChatGPT to reformulate (providing the original sentence).
Is that cheating or OK?
Do I need to disclose this at submission (Elsevier asks...)?

One issue with X is that it seems rather naive to me to assume that the algorithm is not rigged to favour the political opinions of Elon Musk. To me, X is quite likely-just has been shown for TikTok-a propaganda machine. I'd pre-register that prediction!