Soft b / d / ɡ · spirantisation Spanish b d ɡ do not fully close between two vowels — they surface as [β ð ɣ] with air flowing throughout. A hard stop in every position is one of the most audible markers of a foreign accent. This page shows the burst-strength difference between the two positions.
Same sound, two positions
The left cell of each pair sits sentence-initially (after a pause, so it must be a stop); the right cell sits between two vowels (so it must be an approximant). Both play the full sentence — listen for whether the burst is there.
Four steps to "no closure"
Hear it first, then change the articulation, then hand it to the sentence.
Borrow the English th
The d in nada is close to the th of English this — it is literally the same ð. Use it to establish "tip near, not pressed".
Feel the airflow
Say bata (a stop): the lips close and release, giving a short puff. Say la bata (an approximant): the lips only come near each other and air keeps flowing. You can feel this with your hand, without listening.
Assign by position
Stop after a pause and after un / el; approximant after la / mi and any other vowel-final word. Practise it mechanically at first.
Run it in sentences
Put the rule into whole sentences: La boca (soft), Beso (hard), el agua (soft — the l of el does not count as a pause). Spanish applies this across the whole phrase, not per word.
Which one are you
The three most common readings.
Every b d ɡ comes out hard and clear
The classic case: your native language only has stops, so the letter maps straight to one. It is also the easiest to fix — the rule is mechanical, not a matter of intuition.
I soften everything
Overcorrection. Word-initial and post-nasal positions must be stops; all-soft is just as foreign-sounding and harder to understand, since bursts are word-boundary cues.
Fine word by word, messy in sentences
Expected: the boundary is phrasal, not lexical. Get article + noun working (la boca, el agua) before extending.
Word list
16 words grouped by where the sound sits in the word. Click a word to hear it. These are recorded as separate words, not cut out of the sentences above — a word cut from a sentence carries its neighbours' transitions and always sounds muddier. All three groups use the same speaker, so what changes between them is position, not voice.
Word-initial8
Word-medial8
Corpus and acoustic measurements
Click any word in a sentence to play from that word. The pale band over the waveform marks the measured boundary of the target word.
Per-token data
Every number below is measured directly from the audio, not hand-annotated.
| Token | Accent | Sex | Target | Word length | Target consonant duration | F0 | Pitch range | HNR |
|---|
Where these numbers come from
Every claim comes from measuring these 12 recordings.
Corpus: 12 sentences, two Peninsular voices. In each pair the left item is sentence-initial (after a pause, necessarily a stop) and the right one is between two vowels (necessarily an approximant). Both are read by the same voice.
Boundaries: MMS_FA forced alignment, words and phonemes from the same pass. The target phoneme is looked up in the word as declared by the page, not taken from the onset — the d in nada and the ɡ in agua are medial, and taking the first phoneme would measure n and a while the page still rendered fine.
Burst strength: peak 10 ms frame energy inside the segment divided by the mean frame energy of the 60 ms on either side. Dividing by the surrounding vowels makes it a question of how much the release stands out. All five pairs point the right way, with stops 1.28–1.92× the approximants.
Why only five pairs: two more were tested and rejected by the data — the p of El pino is voiced by the preceding l (real connected speech, but it hides the contrast), and sentence-initial dime and intervocalic todo came out nearly equal (d has a small contact area and a weak release). Removing them explains more than keeping them.