The notion of order in disorder is absurd. There should not be any patterns at all.
Some things that go wrong
As children’s speech develops, the error patterns change in a number of ways. But not randomly. Particular words are mispronounced in particular ways at particular stages of development. As we shall see, the patterns demonstrate order in disorder. But this should not be.
• In the speech of some normally developing two year olds, stressed syllables are often doubled or ‘reduplicated’ as in water as WAWA. In speech which is significantly disordered, this doubling is sometimes both prolonged and carried out in more complex ways, as in donkey as DODONG, plastic as PLAPLAK, Indian as HEEJING
•. By one common form of early incompetence, key is said as TEA, and Guy as DIE. This is to say that even where there is just one consonant in the word, back-of-the-tongue articulations in stops are lost in favour of tongue-tip articulations. So this is often known as ‘fronting’. The opposite happens in some children, and is usually known as ‘backing’, but this is quite rare.
•. By another common form of early incompetence, sea is also said as TEA, but for a different reason. What should be near complete closure of the vocal tract becomes a complete closure instead. In both cases the gesture is momentary. Since sounds with a complete closure are known as ‘stops’, this is known as ‘stopping’. The opposite also happens, in other words, saying tea as SEA. But this, like ‘backing’ is rare.
Stopping and fronting are perhaps the two commonest signatures of speech said to be delayed or disordered.
But while stopping is almost unattested in competently spoken language, the opposite process, known as ‘spirantisation’ (fricatives were once called spirants) is common. This in the history of English the K in electric became an S when the word became used as a noun in electricity.
•.Where D and G or T and K occur in the same word, doggy is often said as GOGI, and ticker as KIKKER, with the polarity of the loss by fronting seemingly reversed. This is generally known as ‘harmony’ or ‘assimilation’ – where one sound seems to harmonise with or assimilate to another. But doggy as DODI and ticker as TITTER are seemingly never heard. The loss of an articulator when it occurs on its own goes the opposite way when it occurs with another articulator. The oddness of this apparent polarity reversal between fronting and the apparent articulator harmony was noted by Alan Cruttenden in 1978.
• But in child-speech sounds don’t just harmonise. Often for as long as two years many normally-developing children say little as LIKU and middle as MIGU. Here the tongue-tip T and D makes no pretence of harmonising. But rather it goes to K and G, seemingly disharmonising with the tongue tip articulation of the neighbouring L. Even in in adult-speech there is not much of a tongue tip gesture in the actual pronunciation here. The L is mainly signalled by a colouring similar to the lip-rounded vowel in hook. And in children’s speech the colouring is all that’s left in the spoken form of the L. The only real sign of the tongue-tip is when it initiates a full syllable on its own in littler, middling, and other such forms. But hidden though its tongue-tip credentials are, the L in little and middle is seemingly detected by children so as to effect a seeming disharmony.
In child-speech this effective disharmony is in fact quite common, even though in lists of children’s processes it is hardly ever mentioned.
‘Disharmony’ or ‘dissimilation’ does sometimes happen in competent adult speech in the history of a dialect. The most well known case is the contrast between the L in the ancestral classical Latin form of naval and in the R of the Latin form of solar. But unlike the case with child speech, in fully developed, competently-spoken languages this is quite rare.
• Many children say soldier so that it sounds like shoulder. Correctly, in most current varieties of English, the second syllable should begin with the same sound as jaw or jeep, with the tongue moving from a position of closure to one of near-closure. In soldier as shoulder, the feature defining the articulation of the fricative edge of the affricate gets moved leftwards, but only this feature. This is the feature making sheep sound different from seep. And the fricative edge of the affricate is lost. Instead of S we get an SH, making the word sound like shoulder. The air-stream is still squeezed through a narrow gap in the roof of the mouth, but in the middle of the tongue rather than at the tip, as for soldier. Interestingly, children don’t say soldier as JOLDER with the J sound moving as a whole, but only with one of its features seeming to move. Nor do they say the word as SOLDER, just losing the fricative edge. When such speech is investigated it often turns out that the child is well aware of the difference between soldier and shoulder, but unable to say soldier in the intended way. Soldier seems to be the only word in English with these consonants in these positions and an L in between. Sausages is often problematic in a similar way, sometimes said as SHOSIDIZ and sometimes as SHOSIJIZ, with the articulator from the right edge of the affricate in the final syllable getting moved to the beginning of the first syllable, but again only just this one feature of the sound. The L in soldier somehow increases the vulnerability of the fricative edge of the affricate in its original position.
• Many children say spaghetti as BASKETTI, with the S moving from the beginning of an unstressed syllable to the beginning of the stressed syllable, and the stranded lip-action sound becoming a B. A few children go a step further, leaving out the first vowel, saying the word as PSKETTI. In this case it is as though there were two steps, the first moving the S one syllable to the right and the second losing the first vowel. There is no way a syllable beginning with PSK could be considered as anything other than highly deviant. It certainly doesn’t make the word easy to say in any simple or obvious sense.
•. Much later in speech development, as late as seven or eight, it might seem that articulators are vulnerable once more, but in a different way, in a very small number of words, with very particular characteristics, in cardigan as KARDIDAN, never as KARGIGAN, in hippopotamus commonly as HITTAPOTTAMUS, never as HIPPAPOPPAMUS, in calculator commonly as KALTALATOR, but never as as KALKALAKOR, and archaeopteryx commonly as ARTIOPTERIX, occasionally as ARPIOPTERIX, but never as ARKIOKTERIX. But this is odd. The polarity would seem to be the opposite from what happened in very early speech where the tongue tip articulator was systematically vulnerable to the others, as in doggy as GOGGI. Why should a new sort of error suddenly emerge at this late stage?
• At the same developmental stage, the lip-action P seems to dissimilate to back-of-the-tongue K next to stressed lip-rounded O in monopoly as MONOKOLI. I haven’t found another case where this happens systematically, as it does in monopoly. So it may be worth noting that in this case in the M and the N and the P and the L there is first a lip action and then a tongue tip action. And the lip action gets lost.
Harmony and disharmony seem like contradictory processes. They clearly happen in quite different environments. But there is a commonality between them. In both the degree of contrast is being manipulated, reduced in the case of harmony, and increased in the case of disharmony.
•. In a more complex way, some children with clinical speech problems say cardigan as KARDINTON. Here the G sound has turned into a D. And then that has turned into a T. And the final N has got copied into the previous syllable. That is seemingly by three steps, each one clear and simple in itself.
•. In a similarly complex way, I once heard a child of four and a half saying cardigan as KARDALUNG, with some of the G sound moving right, some of the N sound moving left, and then turning into a sort of L, in other words becoming more different from the target sound structure than it should have been. The speech could hardly be understood at all.
What is happening to the articulators in these child-speech forms has not equivalent in competent phonology. In competent phonology there are many instances of harmony or assimilation between adjacent stops and many instances of long range phenomena between vowels. But there no known cases of harmonic or disharmony phenomena between the articulators of consonants over a long range as in the case of hippopotamus and calculator.
Something wrong with the theory
Not only should there not be any patterns in disorder. There should not be any forced ‘derivations’ where processes apply one after another, as in the case of cardigan as KARDINTON. There is something wrong with the theory here. Now it might be said that these mispronunciations are ‘making the words easier to say’. But this is circular. And spaghetti as PSKETTI is not not easier than the lexical word in any natural sense of ease.
Alternatively, it is often said, almost as an axiom of clinical linguistics, that the characteristic mispronunciations listed above should be characterised as displaying ‘processes’. But this doesn’t rescue the situation at all.
In all varieties of English, including what is known as ‘Received pronunciation’ or RP, as the speech of privilege in Britain, there are certainly processes, as aspects of competent adult speech, such as these, most familiar to the author, as a native speaker of British English:
• There are different T’s in top, litter, little, stop, and spot. If there is such a thing as a phoneme, and the overwhelming majority of linguists think there is, and if T is a phoneme in these words, there have to be processes governing the different pronunciations in these and other cases. In RP, for example, there is no release of the tongue-tip closure for the T before the gesture changes to one involving just the centre line of the tongue for the L.
• In Southern varieties of British English, an R is inserted between India and and in India and Pakistan.
• In RP, in the word prince, there is a closure of the larynx between the N and the S, making the word fractionally distinct from prints, in which the T is commonly pronounced.
• In ways that vary across dialects and generations, there is what is known as ‘lenition’ in a T after a stressed vowel in spot, little and huntsman, where the closure by the tongue tip is replaced by a closure of the larynx. In RP, the most narrowly defined of these cases occurs in huntsman and appointment. Here the T occurs between an N and an M. In other varieties, the lenition is more general.
• In at least most varieties of English, there is what is known as ‘mutual harmony’ or ‘coalescence’ between the T and the R at the beginning of a stressed syllable, as in try and triangle; the T is articulated further back in the mouth and with a gradual release and the R loses its voicing. This does not happen in pastry and triangulate.
All these processes and many others are characteristic of competent adult speech. What happens to the T in huntsman and appointment and the glottal gesture in prince go unnoticed. If children’s speech is defined on processes, the development of fully competent speech depends on learning one group of processes and unlearning another. How does the child learner know which is which?
In the case of stopping, incompetent speech is in the opposite polarity from competent spirantisation. In a way quite unlike competent speech, incompetent harmony does not hugely prevail over disharmony. This raises the suspicion that such forms, and perhaps child speech generally, are by a different sort of mechanism.
In early speech tongue-tip articulations give way to lip and back-of-the-tongue articulations. Later on, exactly the opposite happens. So the overall pattern of incompetence cannot be characterised in terms of the internal properties of the features.
The more serious a child’s issue, the more processes there seem to be, and the more complex they become. The speech of children with the greatest problems commonly falls outside any recognisable schema of processes. So no listing of developmental processes in child speech will ever be complete or explanatory.
An alternative approach
By my proposal in What makes us human, Glue, Doing things with words, and Phase, what is characteristically happening in child speech is that the child learning a system which makes the processes of competent speech possible. But the two things are not the same. Child speech is by a process of learning, not by a set of processes.
The learning is happening in a space, which is factored into two main parts, the assembly of the elements by Glue, and the manipulation of these assemblies for the sake of what they do. In other words, what the language learner has to learn is what is built where.
The fronting and the incorrect instances of D and T in cardigan, hippopotamus, calaculator and archaeopteryx are by failures in building the G, P and K phonemes. The building of the lip articulation of the first P in hippopotamus, and the back of the tongue articulation in the second K in calculator, and the final G in cardigan, should all precede the building of the tongue tip articulation in T and D as a late default. But instead the default tongue tip articulation is wrongly brought forward. This is by virtue of the very particular distribution of contrasts in these words.
The replacement of the tongue tip articulation in magnet as MAGNIK is triggered by the close contrast between the lip and tongue tip articulations of the M and the N and the weaker contrast between the G and the T on the right edges of the two syllables. In such cases the back of the tongue articulation is brought forward by an erroneous copying.
On this approach, every spoken word in every language is by a sequence of steps, assembling the different constituent parts of every sound. The characteristic errors in child speech as by failures to assemble the sound structures in full in the right positions. These failures can occur at any of a number of points. There are more of these possible points of failure than by any list of processes.