Categories For later read

7 days 30 days All time Recent Popular
From the Middle of the Night File, up again after a few hours decent sleep, I was doing some organizing for tomorrow's thread. And what do I find? Once again Twitter has broken one of my threads. Yesterday morning's. It was a 4-tweet thread. I can only see 27 tweets.


2) I don't know what I'm going to do about it. That's a bridge too far. If I didn't have the thread reader app unroll above, I'd be checking my memory of my work. Thank goodness at least that's there!

45 - 27 = 18

Twitter disappeared 18 of my tweets. And they're not random.

3) At the same time, look what I find in my feed. Two announcements composed in identical format. I'll point out what's wrong with these "clarifications" below. This is from


4) And this from


5) Isn't it interesting that both clarifications follow the exact same series of questions and answers? Obviously this is from the legal department, as if it were information to be shared. It is not. It is very, very carefully NOT information. Focus on that.
Amazigh (Berber) languages are quite close to each other and in most places nearby varieties are mutually intelligible. They function like a discontinuous dialect continuum.

A loooong thread with maps (and no memes☹️).

So can we do sub-classification?

Many people say “no”, like André Basset’s famous quote: “cette langue s’éparpille directement ou à peu près en une poussière de parlers de 4 à 5 mille peut être” (1952:1) and Alfred Willms (1980). Others are a bit more nuanced.

For example, the fantastic studies by Lafkioui (
https://t.co/hHKMgEPjK2) give a synchronic classification of Tarifiyt dialects. To cut a very long story irresponsibly short: all variables are counted the same.

However, we should ask ourselves: Does this continuum hide a more discontinuous past? Has there never been major disruption, or has much of it been smoothed out by later convergence?

In order to study this, one has to classify variables and their isoglosses. Some variables are continuous and can be assigned to the latest convergence period. Others are clustered in a group unrelated to the continuum. Still others are scattered.