Relevant and irrelevant distinctions

As discussed in Section 2.4, the speech sounds in spoken human languages do not carry meaning — there is only a limited number of sounds we can produce and distinguish, so that we could only express a limited number of meanings if they did. Instead, we combine sounds into larger units and then associate those larger units with particular meanings that are unrelated to the individual sounds that occur in them.

Of course, those larger units also have to be distinguishable from each other. It is not difficult to distinguish between units that are made up of completely different sounds. Take the units [bɛɹ] and [hʊf]. As becomes evident when we represent their segments as bundles of features, they do not share any features at any position:

Figure 4.2.1: The words bear and hoof represented as bundles of phonetic features (American English).

The two sequences of speech sounds would be different even if we ignored two of the three features at any position.

This is not the case with the sequences [bɛɹ] and [pɛɹ]. In order to associate them with different meanings, like BEAR and PEAR in English, we must pay attention to voicing.

Figure 4.2.2: The words bear and pear represented as bundles of phonetic features (American English).

If we did not pay attention to voicing, the two sequences would be identical, which would lead to potentially life-threatening situations if we thought that someone had asked us to pluck bears from a tree instead of pears. The words bear and pear differ in terms of aspiration, too: the first segment of bear is unaspirated, the first segment of pear is aspirated. However, we do not have to pay attention to this fact, because the words are also distinguished by the voicing difference of the first segment.

This is not the case in Hindi, where we have to pay attention to aspiration in order to associate the sequences [bɑːluː] and [bʰɑːluː] with the meanings ‘sand’ and ‘bear’ respectively.

Figure 4.2.3: The words baaloo and bhaloo represented as bundles of phonetic features (Hindi).

If, as speakers of Hindi, we did not pay attention to aspiration, our brains would categorize the two sequences as identical, which could lead to potentially life-threatening situations if we thought that someone suggested building a [bʰɑːluː kaː ˈmɛɦɛl] (बालू का महल, bear castle) instead of a [bɑːluː kaː ˈmɛɦɛl] (भालू का महल, sand castle).

Identifying phonemes using minimal pairs

What we have done above is use minimal pairs — two different words that differ in just one phone — in order to identify phonemes and their relevant properties: [bɛɹ] and [pɛɹ] are identical except for the first slot: [ _ ɛɹ]. Since [p] and [b] can both occur in the first slot resulting in two different words, they are distinct phonemes (our brains put them into different categories). To distinguish phonemes from phones, they are enclosed by slashes rather than square brackets: /p/, /b/.

We can use this procedure to identify other phonemes of English: for example

• [f] and [v] must be distinct phonemes, because foal and vole are different words;
• [t] and [d] must be distinct phonemes, because tear and deer are different words;
• [s] and [z] must be distinct phonemes, because seal and zeal are different words.

These phonemic distinctions also confirm that voicing is a phonologically relevant feature in English. We can also find minimal pairs that confirm the relevance of places and manners of articulation. For example,

• [s] and [f] must be distinct phonemes, because soul and foal are different words, confirming the relevance of the features alveolar and labiodental;
• [t] and [s] must be distinct phonemes, because teal and seal are different words, confirming the relevance of the features plosive and fricative.

We will not find a minimal pair for [bʰ] and [b] or for [pʰ] and [p] in English. In the first case, this is simply due to the fact that voiced plosives are never aspirated in English, so that there are no words containing a [bʰ]. Languages differ in their inventories of speech sounds.

The second case is more interesting: voiceless plosives can be aspirated or unaspirated in English, so that there are words containing [p] and words containing [pʰ] — for example, spare and pear. However, the two phones cannot occur in the same position in an otherwise identical sequence of phones, so there cannot be any pairs of words that are identical except for aspiration. We will discuss this in the next section.

Contrastive distribution

If two phones can occur in the same phonetic contexts and be perceived as different, they are said to contrast with each other, to be in contrastive distribution. Minimal pairs are the clearest evidence of a contrastive distribution: if changing one phone in a word gives us a different word, then the two phones obviously contrast with each other. So, finding a minimal pair is the simplest and most convincing way of showing that two phones are separate phonemes. Sometimes, we even find whole sets of words that differ by just one speech sound. For example, for American English, the words beet, bit, bait, bite, bet, bat, but, bot, bought, boat, boot and bout provide evidence for the phonemic status of almost all vowel phones in a single step: [i], [ɪ], [eɪ], [aɪ], [ɛ], [æ], [ʌ], [ɑ], [ɔ], [oʊ], [u], and [aʊ] (AmE).

However, there are cases where finding convincing minimal pairs is very difficult. Take the voiced postalveolar fricative [ʒ] in English. It is very rare consonant. It occurs mostly in words that end end in -ion (for example, [vɪʒən] vision), -ual (for example, [ˈvɪʒuəl], visual), and -ure (for example, [mɛʒəɹ] measure). Such endings are a special type of sign that can attach to other words to form new words, we will discuss them further in Chapter 5. There is only a handful of other words that contain it, for example, [ɹuʒ] rouge and [ɡəˈɹɑʒ] garage, [ˈbuː.ʒi] bougie, [ˈʒɑːn.dɑːrm/] gendarme, [ˈʒɑnɹə] genre, [ʒæˈboʊ] jabot and [ʒʊʒ] zhoosh — all of them French loanwords, except for the last one, which is from Polari. It is also somewhat unstable, with a tendency to change to [dʒ], especially in British English, where some of the AmE pronunciations shown here do not exist or have alternatives with [dʒ]. Given its rarity, it is not surprising that it is difficult to find minimal pairs in which [ʒ] contrasts, for example, with the voiceless postalveolar fricative [ʃ] or with the voiced postalveolar affricate [dʒ].

For [dʒ] and [ʒ], we find [ˈliː.dʒən] legion versus [ˈliː.ʒən] lesion and [ˈplɛʒəɹ] pleasure vs. [ˈplɛdʒəɹ] pledger ‘someone who pledges something’, for [ʃ] and [ʒ], we find [əˈluʒən] allusion versus [əˈluʃən] Aleutian or [ˈmɛʒəɹ] measure versus [ˈmɛʃəɹ] mesher ‘a device for creating a mesh’. But the second member of both pairs is very rare, many English speakers may not know them. And they are words containing the special endings mentioned above, so we should be careful when accepting them as evidence.

In such cases, or in cases where we do not find a minimal pair at all, it is important to understand that it is not the existence of a minimal pair that determines whether two phones are distinct phonemes, but the question whether they are in contrastive distribution, i.e., whether they can create a contrast in the same phonetic environment. If there is a minimal pair, we can immediately conclude that the phones involved are in contrastive distribution. If there is not, we can look for words that have the two phones in the same immediate context — the same position in the syllable, the same stress pattern and the same phones occuring to the left and right. For [ʒ], [ʃ] and [dʒ], for example, we find [ɹuʒ] rouge and [duʃ] douche and [hjudʒ] huge, where they occur after a [u] at the end of the word.

← Previous section Next section →

CC-BY-NC-SA 4.0, Written by Anatol Stefanowitsch (Examples for [ʒ], [ʃ] and [dʒ] from Anderson et al.)