9.8: Learning Phonetics and Phonotactics in a Later Language

If your first language is a spoken language, you’ve had a lot of practice at articulating the speech sounds in your phonetic inventory: the set of segments that are present in the grammar. For example, most varieties of English have somewhere around 25 consonants and a dozen or so vowels. That’s a lot of vowels, compared to many other languages, so learning to articulate English vowels can be particularly challenging for adult learners of English.

In contrast, Hawaiian, the Indigenous language spoken in the Hawaiian islands, has only eight consonants and five vowels. And on the other end of the spectrum is Adyghe, one of the Circassian languages spoken in Russia and Türkiye. Adyghe has only three vowels, but (depending on the dialect), between 50 and 60 consonants, including 18 different plosives, ten affricates, and 24 fricatives!

If the language you’re learning includes segments you haven’t yet learned to articulate, you might make a substitution with a segment from the phonetic inventory of your L1. If you have a name that isn’t English, you’ve probably had the experience of English-speakers making substitutions in your name. For example, the Hebrew name Baruch [baɾʊx] ends with a velar fricative [x], but English speakers often pronounce it with a velar stop instead [baɹuk]. Likewise, the Tamil name Kavitha [kavit̪a] has a dental stop in the onset of the last syllable, which English speakers often turn into a flap [kəviɾʌ] or into a dental fricative [kəviθʌ].

Another challenge for adult learners comes from your L1 phonotactics. Phonotactic constraints are restrictions in the mental grammar on what sounds can appear in what positions, and what syllable structures are possible. For example, the velar nasal [ŋ] is part of the phonetic inventory of English, but it never appears in the onset of a word, only in coda position, like in lung, tank, and singer. The phonetic inventory of Vietnamese also includes the velar nasal, but it’s grammatical in onset position, like in the very common surname Nguyen [ŋwɪn] and the word nghe [ŋɛ] which means listen.

Phonotactics also constrain the possible syllable structures in a grammar. In Chapter 3 we learned that English can have a whole lot of consonants in syllable onsets and in syllable codas – consider the word strengths, [strɛŋkθs] which has three consonants in the onset and four in the coda! English also allows much simpler syllables, like nice [naɪs], or odd [ɑd], and even syllables with nothing in the onset or coda, like eye [aɪ].

But some languages have much tighter phonotactic constraints on their syllable structure. When you’re learning a later language, you often adapt the shape of the words to fit the phonotactics of your L1. Likewise, when a language borrows words from another language, the loanword gets adapted.

A famous example comes from Hawaiian, which has only two possible syllable structures: a syllable can have either one or zero consonants in its onset, no coda consonants and no consonant clusters. So when Hawaiian borrows the two-syllable English word Christmas [krɪs.məs], there’s a lot of adaptation to do.

The only position a consonant can occupy is the onset, and onsets can’t have more than one consonant. So the [krɪ] from English becomes [kali] with the [l] substituting for English [r] and [i] for English [ɪ]. Then the [s] from the coda of [krɪs] gets a substitution and its own syllable to become [ki]. The [s] from the coda of the second English syllable [məs] becomes another onset, and result is that the Hawaiian adaptation of the English word Christmas is [kalikimaka].

Besides loanwords, English speakers also tend to adapt proper names to conform to English phonotactics. For example, when English speakers are talking about visiting the Polish city of Gdańsk [ɡdaj̃sk], they usually epenthesize an extra [ə] between the first two consonants, because [ɡd] isn’t a grammatical onset in English. Or when English borrows the German name Pfeiffer [pfaɪfɐ], the plosive [p] at the beginning gets deleted because [pf] isn’t a possible onset.

In Chapter 11 we learned that babies have set up the phoneme categories of their L1 by the time they’re only twelve months old. So by the time you’re an adult learning a new language, those phoneme categories are pretty deeply embedded in you mental grammar. Psycholinguist Catherine Best has proposed a theory, called the Perceptual Assimilation Model (Best, McRoberts & Goodell, 2001), that predicts which phoneme contrasts will be hard for adult language learners, and which will be easy. The theory centres around the concepts of phonemes and allophones.

Best predicts positive transfer for phoneme contrasts happens if the new language has a phoneme contrast that maps onto a contrast in the adult learner’s L1. In this case, the new phoneme contrast should be easy to learn. She also predicts that it should be easy to learn a new contrast in the case where there’s no transfer at all: if the new language has contrastive phones that don’t exist at all in the learner’s L1, then the learner can set up two new phoneme categories from scratch.

The situation that leads to negative transfer arises when two contrasting phonemes in the later language map onto a single phoneme category in the learner’s L1. In this case, the learner will have spent a lifetime treating the phonetic difference as allophonic variation, and not a meaningful contrast, so it’s a challenge to learn to pay attention to the difference as meaningful.

Best and her colleagues (Best, McRoberts & Sithole, 1988) tested this theory by investigating how English-speaking adults learn phonemic contrasts in Zulu. Zulu is a language that has about 27 million speakers, most of them in South Africa. First, researchers asked the English-speakers to notice the difference between voiced and voiceless lateral fricatives [ɬ and ɮ] in Zulu. English doesn’t have lateral fricatives, but English does have lots of pairs of fricatives that contrast in their voicing, so the theory predicts that it should be easy for English listeners to map the voicing difference between the Zulu fricatives onto those English voicing contrasts and recognize this phonetic difference. And that prediction was upheld: The English listeners were about 95% correct.

Then they asked the English speakers to tell the difference between three Zulu clicks: a dental [ǀ], an alveolar [ǃ], and a palato-alveolar [ǂ] click. English doesn’t have any clicks at all, so the English listeners should be able to simply pay attention to the phonetic differences between these segments, without any interference from their English phonology. The English listeners were about 80% correct at these sounds.

Last, they asked the English listeners to notice the difference between two different kinds of bilabial stops in Zulu: the plosive stop is similar to the English [b] sound. The other is an implosive [ɓ], which is made by obstructing airflow at the lips, but when the stop is released, air flows into the mouth instead of out of the mouth. The English adults were not much better than chance at hearing this difference.

These results are consistent the predictions of the Perceptual Assimilation Model: Because the English listeners have only one phoneme category for voiced bilabial stops, their mental grammar simply treats the implosive as an allophone of the plosive. It’s difficult for the adult learners to hear the phonetic difference between the two sounds in their new language because the mental grammar of their L1 categorizes both segments as members of the same phoneme.