En réponse à Mark J. Reed : >Okay. Which is then indistinguishable from my original [tj]. :) It is, as much as any affricate is distinguishable from its corresponding cluster. :)) If you can't distinguish them, it's not that there's no difference! ;))) . >[c] - palatal from the beginning >[t_j] - starts out alveolar and moves toward palatal >[tj] - starts out alveolar and moves toward palatal! No! [tj] is [t]+[j], i.e. two phones, one purely alveolar (since I think your [t] is alveolar), the second purely palatal. [t_j] is a single phone, with a palatal release - which means raising the middle of the tongue *while* pronouncing the phone. In the case of the cluster, the raising of the tongue appears *after* you pronounced the [t] -. The acoustic result is quite different (as well as the length of the whole thing ;))) ). >I can't see a three-way difference here . . . That's only because you're not trained to do it. But at Jan's wedding, I heard enough Polish to understand the difference between a palatalised consonant and the cluster consonant+[j] ;)))) . And since Maggel has also palatalised consonants different from clusters consonant + [j], I had to learn to make the distinction. Yet it's true that like the difference between affricate and cluster, the difference between palatalised consonant and consonant + [j] is rather small, and it's not surprising that you don't hear it (especially since historically, clusters consonant+[j] tend to become true palatalised consonants, and even later affricates, like the modern pronunciation of Dutch "tj", originally a cluster, later a palatalised consonant, nowadays usually the affricate [t_S], often in free variation with [t_j]). It's all a matter of training ;))) . Christophe Grandsire. http://rainbow.conlang.free.fr You need a straight mind to invent a twisted conlang.