tr / dr / str · one tap, not a trill The classic error in trabajo is a vowel slipped between t and r: te-ra-ba-jo. In Spanish the stop's release carries the tongue tip into one tap — no vowel, no rolling. This page measured twelve cluster r's: 50–61 ms, one tap each.
Four steps to the cluster r
Structure first, then the cause of the error, then practice.
The shape is stop + one tap, with no vowel between
trabajo starts with t + ɾ + a, not t + ə + ɾ + a. The release burst is exactly what flicks the tongue tip once, and the vowel follows immediately — the three stages are almost simultaneous.
Why a vowel appears anyway
Mandarin has no tr cluster: a consonant must be followed by a vowel, so your system inserts ə to bridge the gap. This is not sloppy pronunciation — it is your syllable structure at work. Knowing that tells you what to suppress.
How long one tap is: 50–61 ms
All twelve cluster r's land between 50 and 61 ms, averaging 55. In the same voice, written rr runs 121 ms — the difference is not speed, it is how many taps. A cluster never has more than one, so do not roll it.
Practice: from ta to tra
Say ta, then ra (tongue tip taps the alveolar ridge once), then push them together with no gap. For estrella (s-t-r), master tr first and add the s in front afterwards — do not start with three consonants at once.
Spelling: one r, one tap
One rule: a single r taps once; only rr rolls.
Stop followed directly by the tap. The most common cluster, and the one most often broken by a vowel.
Intervocalic d is soft in Spanish (see the spirantisation page), so padre is pa-ðɾe, not pa-də-re.
Three consonants in a row. The s only starts the airflow — do not read “es-trella” with an extra e.
Three typical readings
Pick the one that sounds like you; fix one thing at a time.
I say “te-rra-ba-jo”
A vowel plus a roll is the usual combination error. Drop the roll first (one tap), then drop the vowel — changing both at once usually changes neither.
My tr sounds like English ch
English truck is often [tʃɹ], so trabajo comes out “chrabajo”. In Spanish t and r are separate sounds; the r is one tap with no retroflex colouring.
I use the English r
English r never touches the palate and often rounds the lips. Spanish ɾ actually touches the alveolar ridge once, very briefly, with the lips neutral.
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-medial8
Word-final8
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 | target word | cluster r duration | F0 | Pitch range | HNR |
|---|
Where these numbers come from
Twelve sentences from one Spanish female voice; boundaries from MMS_FA forced alignment.
Corpus: 12 Spanish sentences whose target word contains tr / dr / str, with the target never first or last in the sentence (edge position skews duration through pause and sentence stress).
Boundaries: after MMS_FA alignment, every target in all twelve is ɾ — not one was labelled r. Durations are the target segment, not the word and not the whole cluster.
Result: 50–61 ms, averaging 55 ms. In the same voice, written rr measures 121 ms (see the trill page).
What this page does not do: r after l / n / s (carne, alrededor, honra) measures 65–71 ms — slightly longer than a cluster tap, nowhere near rr; word-final r (comer, hablar) is 40–50 ms, the same order as a tap; /ɾ/ vs /d/ (pero / pedo) separates on neither duration nor burst strength. None of the three supports a page of its own, so they are not here.
An honest caveat: in real read speech trills shorten a lot — in our human corpus, segments labelled r have a median of 40 ms, and 76% are under 60 ms. So “a trill is twice as long as a tap” holds only for these synthetic recordings; judge rr by how many taps, not by length.