Standard Hungarian (Uralic)

Hungarian is analyzed as having the alveolar and postalveolar/palatal affricates in Siptár and Törkenczy's (2000) book, Phonology of Hungarian. The arguments, we think, are not super clear, since Hungarian is fairly permissive phonotactically. There are stop-fricative sequences in the language, and while they are etymologically foreign, they can occur in the same environments as the purported affricates. The weakness of the arguments is in line with the inconsistent findings of the learner.

The closest the learner comes to the inventory posited by traditional analyses of the langugae is when it is trained on the list of stems/roots. This could be due to the agglutinative morphology of the language; the clusters found in morphologically complex words are more diverse than those in morphemes, so the affricates' share of the distribution is lower in words than in morphemes.

Simulation data at a glance

Click on simulation name to view additional simulation details.

Simulation name	Initial state Learning Data	Initial state features
Roots	LearningData.txt	Features.txt
Words Broad	LearningData.txt	Features.txt
Words Narrow	LearningData.txt	Features.txt
Gem_Rts_Narrow	LearningData.txt	Features.txt

---

Simulation details for Hungarian roots

The simulation run on roots finds the less controversial affricates of Hungarian, [ts] and [tʃ], even without resorting to narrow transcription of the latter. The more controversial affricate [dʒ] just misses the threshold on the 2nd iteration. See Siptár and Törkenczy's discussion in their section 4.1.4, where they also weigh the arguments for [dz] (nowhere near the threshold for unification in any of our simulatoins for the language).

Summary of inventory changes

Stage	Consonant set
Input	p b t d c ɟ k ɡ f v s z ʃ ʒ h m n ɲ l r j pː bː tː dː cː ɟː kː ɡː fː vː sː zː ʃː ʒː hː mː nː ɲː lː rː jː
Output	p b t d c ɟ k ɡ f v s z ʃ ʒ h m n ɲ l r j pː bː tː dː cː ɟː kː ɡː fː sː zː ʃː ʒː mː nː ɲː lː rː jː ts tʃ

---

Simulation details for Hungarian words broad

The simulation trained on the broadly transcribed words does not find any complex segments at all.

Summary of inventory changes

Stage	Consonant set
Input	p b t d c ɟ k ɡ f v s z ʃ ʒ h m n ɲ l r j pː bː tː dː cː ɟː kː ɡː fː vː sː zː ʃː ʒː hː mː nː ɲː lː rː jː
Output	p b t d c ɟ k ɡ f v s z ʃ ʒ h m n ɲ l r j pː bː tː dː cː ɟː kː ɡː fː vː sː zː ʃː ʒː hː mː nː ɲː lː rː jː

---

Simulation details for Hungarian words narrow

The narrow transcriptions allow the learner to find the postalveolar voiceless affricates but nothing else. In particular, [ts] is well below the threshold. This could be because morphological complexity dilutes the frequency of affricates, as it does in Quechua (see the discussion in the paper).

Summary of inventory changes

Stage	Consonant set
Input	p b t d c ɟ k ɡ f v s z ʃ ʒ h m n ɲ l r j pː bː tː dː cː ɟː kː ɡː fː vː sː zː ʃː ʒː hː mː nː ɲː lː rː jː
Output	p b t d c ɟ k ɡ f v s z ʃ ʒ h m n ɲ l r j pː bː tː dː cː ɟː kː ɡː fː vː sː zː ʃː ʒː hː mː nː ɲː lː rː jː cʃ cʃː

---

Simulation details for Hungarian gem_rts_narrow

The simulation run on roots finds the less controversial affricates of Hungarian, [ts] and [tʃ], even without resorting to narrow transcription of the latter. The more controversial affricate [dʒ] just misses the threshold on the 2nd iteration. See Siptár and Törkenczy's discussion in their section 4.1.4, where they also weigh the arguments for [dz] (nowhere near the threshold for unification in any of our simulatoins for the language).

Summary of inventory changes

Stage	Consonant set
Input	p b t d c ɟ k ɡ f v s z ʃ ʒ h m n ɲ l r j
Output	p b t d c ɟ k ɡ f v s z ʃ ʒ h m n ɲ l r j ts cʃ

View the source code on GitHub