Percentages, normalities, abnormalities, and the limits of statistics

It is obviously useful to assess and measure degrees of disorder, if only to compare different degrees of need.

Tests measure what it is possible to measure in some given domain with some given degree of accuracy. A test characteristically involves a number of preset tasks, each representing some arbitrary degree of complexity. But no set of preset tasks can determine how far the child’s capacities stretch. How children say particular words is just one of the more easily testable aspects of speech. But there are many domains of speech, the way sounds are formed, the ‘phonetics‘ the way they are put together, the ‘phonology‘  , the way they are formed into clusters, the ‘phonotactics‘, the way they are formed into words, the ‘morphology‘, and so on. These often comprise the main issues which get noticed in children and young people’s speech. But how do measures of the various domains weigh up against one another? Serious abnormalities are mostly rather obvious. At a given point on a scale they count as impairments. But what defines the point? What criteria are used to define it? And what defines the scale?

One widely used scale is the Percentage of Consonants Correct or PCC. Now, the obvious reason for prioritising consonants is that children have many more developmental difficulties with the consonants than with the vowels. But in respect of any given consonant what counts as ‘correct’? In the considerable literature on children’s mastery of R there is a wide variation in the ages at which a ‘normal’ or ‘correct’ R is said to be acquired (from four to seven). It seems likely that the criteria vary from study to study. And what if a given consonant seems to be pronounced correctly in all but a particular context or a set of words? Do they count in the computation? A less used scale is the Percentage of Vowels Correct  or PVC. But this is subject to the same problems as as the PCC. Some children have difficulties with both vowels and consonants or, in rare cases, more difficulties with vowels than with consonants.

But speech is not just about ordering sounds. Speech is also about knowing when and how words change – like the fact the and a both change when the next word begins with a vowel, and fitting sounds into syllables. A child may just fail to recognise the syllable as a category, treating syllables as just complex sounds.. Such aspects of speech are not generally tested by standardised tests. But speech entirely based on a linear sequence of speech sounds would be barely recognisable as speech. This became obvious when the first efforts to build speech synthesisers were made in the early 1950s using the new technology of tape recording. Percentages of correctness may give rise to misleading estimates of competence.

Similar issues arise with respect to children’s language, but with the additional issue that it often seems that they understand more than they can say. This critically involves their grasp of what is known as ‘syntax‘ – or how words fit together.

My clinical focus is on disorders of speech and language and the various interactions between syntax , morphology, phonology and phonetics. I am not confident about the representation of these domains in terms of points on scales or percentages. The number of words or the ‘lexicon‘ can in principle be counted, but only in principle. The number relates only to the words that are said, not the number of words that are understood.

It is possible that speech and language are just statistically intractable.

Unusual or abnormal development is not, in itself, a disorder. The diagnostic question relates only to the degree of abnormality in any given case. Sometimes a major issue resolves sponaneously and faster than by normal development. But there are issue which do not resolve spontaneously, which without treatment the individual may be left with for life. Such issues can be revealed by extreme degrees of abnormality. To this degree, abnormality can be diagnostic.

Beyond statistics

Outside the scope of statistical evaluation, there are a number of fundamentals. One of these concerns the development of syllables as parts of words. English happens to have a relatively rich inventory of both vowels and consonants and an unsually complex foot structure. And English is very permissive about the phontactics.

Children are not given any guidance or privileged information on which bits of linguistic structure influence which other bits. Nor could they be. Where English stands in the learnability space is not a matter for discourse with children.

In competently spoken English there are interactions between words and between the parts of words inside the words which define the language – like those between an and the and the next word. And there are things which typically developing children commonly way which are most definitely not part of English. Some of these interactions are like points on a pathway. So there is a natural pathway from little, often hard to say even at four, to monopoly, often hard to say at seven or eight or even in adulthood. The T in little and the P in monopoly both get pronounced with more contrast between them and two other sounds – as LIKU and MONOKOLI. This is known linguistically as ‘disharmony’ or ‘dissimilation’. By nine most children can say most words, even the more difficult ones like monopoly. Those who can’t most likely need help. But this may be outside the scope of useful statistical representation.