Database of Dutch diphone perception

Material -- Diphone Selection

paper preprint

DCS project

The phonemic vowel and consonant inventory used for the diphone selection was based on the Dutch CELEX lexical database and on "The phonology of Dutch" by Booij (Oxford: Oxford University Press, 1995).  Not all possible phonemes were used, however. Reasons for the selection or exclusion of particular phonemes were the following: 
  1. Besides the voiceless velar fricativ /x/, CELEX recognizes the voiced velar fricativ / gamma /. The phoneme was excluded because many Dutch speakers neutralize the distinction, maintaining only / chi /. 
  2. The vowels / i_colony_colonu_colonopeno_colonoe_colonepsilon_colon / occur only in unassimilated loanwords (such as "analyse, centrifuge, cruise, oeuvre, serre" respectively) and differ from native vowels phonemes only in length. These vowels were excluded because they only occur in a small number of recent loanwords. 
  3. Nevertheless a number of consonants which occur in Dutch only in unassimilated loanwords were included, such as the voiced velar stop /g/, the fricative / yogh /, and the affricate / d-yogh /, since they appear in a relatively large number of loanwords, many of them quite frequent (e.g. "goal", "jam", and "jazz"). 
  4. There are certain inconsistencies in the CELEX inventory, such as the fact that [t-esh] is treated as a sequenceof a stop and a fricative, /t-esh/, while [d-yogh] is treated as a single affricate segment /d-yogh_overtiebar /. In these cases it has been adhered to the CELEX standard. 

For the combination of phonemes into diphones the following general rules were applied: 

  1. For each sequence of two phonemes containing a vowel other than / schwa / (which is never stressed), one diphone was included with the vowel stressed, and another with it unstressed. For vowel-vowel diphones, all four stress combinations (stressed-stressed, unstressed- unstressed, stressed-unstressed, unstressed-stresses) were included. 
  2. Diphones which can only occur across word or morpheme boundaries in Dutch were included (e.g. / eng-p /), whereas such diphones were excluded which could never occur even across word boundaries, because of phonotactic constraints. 
  3. In cases where phonotactic constraints were violated by large numbers of loanwords the diphone were included. 
  4. Excluded were certain diphones which are possible (at least across morpheme boundaries) according to a phoneme transcription, but unlikely ever to be produced as a sequence of the two sounds (e.g., / s-eshesh-st-d-yogh_overtiebar /). 
  5. All sequences of identical consonants were excluded, since Dutch phonology requires that these be degeminated within the prosodic word, and they are likely to be reduced to a singel consonant even across word boundaries unless produced with a pause. 
  6. A few diphones which probably never occur in Dutch, e.g. / scriptaepsilonsmall-capital-y / followed by / yogh /, were included simply because no known phonotactic constraint excludes them. 
The table below gives a full list of  rules defining our final set of diphones. 

CV diphones C V
stop, affricate, nasal, liquid,
or glide
p, t, k, b, d, g, d-yogh, m, n, eng, r, l, j, w all full vowels stressed,
all vowels unstressed 
fricative f, v, s, z, eshyogh, x, h all full vowels stressed
f, v, s, z, yoghyogh, x all vowels unstressed
h all full vowels unstressed

Exclusion: / h-schwa / within the syllable, and /h/ cannot be syllable-final

VC diphones V C
stop, affricate, liquid, or glide all full vowels stressed,
all vowels unstressed
p, t, k, b, d, g, d-yogh, r, l, j, w
fricative all full vowels stressed,
all vowels unstressed
f, s, eshyogh, r, l, j, w
all long vowels and diphthongs stressed;
all long vowels, diphthongs, and / schwa / unstressed
v, z

Exclusion: short vowel before /v, z/ is not possible within the syllable, and short vowels cannot be syllable-final 

nasal all full vowels stressed m, n
all short vowels stressed; all short vowels and / schwa / unstressed eng

Exclusion: / eng / cannot follow long vowels within the syllable and cannot be syllable-initial

VV diphones V1 V2
stressed-unstressed all long vowels & diphthongs all vowels
unstressed-stressed all long vowels, diphthongs, &  schwa all vowels except  schwa
unstressed-unstressed all long vowels, diphthongs, &  schwa all vowels
stressed-stressed all long vowels & diphthongs all vowels except  schwa

Exclusion for all VV categories: short vowel cannot be V1 because they cannot be syllable-final

CC diphones C C
voiceless stop, nasal, liquid, or glide p, t, k, m, n, eng, l, r, j, w all consonants except C1=C2 and / eng /

Exclusion: / eng / cannot follow a stop or another sonorant within the syllable or be an onset 

voiced stop b d, g, d-yogh, v, z, yogh, n, l, r
d b, g, v, z, yogh, n, j, w

Exclusion for /b, d/: /bw, bj, bm, dl/ in syllable onset, and voiced stops must devoice if not on onset unless followed by a voiced obstruent; cannot be followed by / eng / because / eng / cannot be an onset 

g b, d, v, z

Exclusions: syllable-final /g/ without devoicing is only followed by these consonants, and /g/ is never word-final

fricative f all consonants except f, v, eng

Exclusion: /fv/ is too difficult for speakers to produce without assimilation

s, esh all consonants except s, esheng

Exclusion: / s-esh / and / esh-s / are unlikely, unless assimilated
x all consonants except x, eng
v b, d, g, z, yoghd-yogh, n, l, r

Exclusion: /vj, vw, vm/ as onsets and /v/ must devoice if not in onset

z b, d, g, v, d-yogh, m, n, j, w

Exclusion: /zl, zr/ as onsets and /z/ must devoice if not on onset; / z-yogh / is likely to assimilate
Exclusions for /v, z/: cannot be followed by a voiceless fricative within the syllable, and will devoice in coda position unless followed by a voiced obstruent

yogh w

Exclusions: / d-yogh / never occurs syllable-finally and in onset occurs only before vowels or /w/ (e.g. "bourgeois")
Exclusions for all fricative: / eng / cannot follow a fricative within the syllable and cannot be an onset 

affricate d-yogh m

Exclusions: / d-yogh / never occurs word-finally, occurs syllable-finally only in the word "management", and cannot be followed by any other consonant within an onset
Exclusions for all CC diphones: no geminates 

Many of the diphones were recorded in several stress conditions.
Stress for consonant C (i.e., the syllable the consonant was part of) was chosen as follows (the dash symbol indicates syllable boundary): 

  • CV: one syllable, stressed or unstressed 
  • C1C2: one stressed syllable, or 
  • C1-C2: two syllables with stress on the first 
  • V-C: two syllables with either stress on first or on second syllable
Stress for consonant V was chosen as follows: 
  • CV: one syllable, stressed or unstressed 
  • V-C: two syllables with either stress on first or on second syllable
  • V1-V2: two syllables with stress on first or second or both or neither.