Plan
The three groups of children were used for specific purposes. Whilst the comparison subjects were clearly a control group with normally developing oral motor skills, a proportion of children with NOFT were expected to have some oral motor deficits. In contrast, the children with cerebral palsy were chosen specifically because of their known oral motor skill deficits and therefore used as validating criteria for the instrument. Before proceeding with data analysis it was crucial to define the status of each individual behaviour.
Procedures
Defining abnormality
For the purposes of the validation exercise it was necessary to make decisions regarding the status of each DOM behaviour, that is, whether the presence or absence of a behaviour could be considered a failed or passed response. An exhaustive list of over 700 discrete oral motor behaviours was compiled, between 75-90 for each oral motor challenge category. The task of ensuring that each behaviour could be defined as normal or abnormal was complicated. As the literature review highlighted there was a paucity of research in the area. Whilst a large standardisation sample would be ideal in order to accurately describe abnormal verses normal oral motor function it was not possible within the constraints of this thesis.
A number of approaches were therefore considered. However none of these approaches could be adopted in isolation and a combination was used to establish a normal - abnormal classification for each behaviour. They included;
1. Theoretical approaches. That is, was knowledge of normal and abnormal development sufficient to be able to clearly define the status of each discrete behaviour?
Where possible, decisions were supported by references in the literature which defined normal and abnormal oral motor behaviour in infants aged 12-18 months. However, comprehensive data did not always exist and the data available, were
often confounded by numerous factors known to affect the development of oral motor skills in infancy.
2. The use of developmental norms was considered Important.
Almost no developmental data were available. Developmental age, as in other areas of child development, was considered an important factor in the decision making process. For example, the absence of a particular behaviour in a child aged 6 months may not necessarily be considered a failed response, whereas absence of the same skill during the latter stages of infancy would be considered as a failed response. Young infants are often unable, for example, to use their upper and lower lips to clean the spoon and do not use rotary jaw movements when munching solid textures. However, by the later stages of infancy these DOM behaviours would be part of the infant's oral motor repertoire.
Similarly, texture was also an important factor; the absence of some DOM behaviours, such as lateral tongue and jaw movements for purée or semi-solids, would not be considered failed responses whereas their absence when dealing with a more challenging texture, such as the solids or crackers, would be considered as a failed response. There is evidence to suggest that normal children use the method requiring the least effort for dealing with food orally. They will for example, often ingest foods such as semi-solids by sucking or munching instead of using a more mature chewing pattern.
3. Clinical insight could be used to judge the significance of individual behaviours.
Few clinicians use standardised methods for recording and judging oral motor performance in their clinical practice. Furthermore few can reliably agree on distinctions regarding normal and abnormal behaviour. Where no data were available from previous studies, decisions were made on the basis of the rater's clinical experience in evaluating the oral motor skills of more than 100 normally developing children.
4. Statistical methods could be considered to make distinctions, between normal and abnormal behaviour.
A variety of statistical methods were considered. They included: ■ Factor analysis
■ Discriminant analysis and ■ Cluster analysis
Factor analysis is normally used to identify a relatively small number of factors that represent relationships within or among sets of many interrelated variables. Factor analysis requires the identification of a set of 'not-easily-observable' or underlying factors based on a set of readily observable variables (NoruSis 1990). The basic assumption in factor analysis is that these underlying factors can be used to explain more complex phenomena. In the current study the interest
initially was not in the relationships between or among variables but rather in a technique that would discriminate between relatively homogenous groups of cases. Furthermore, the SOMA resulted in a rather large number of variables which could not have been entered into a factor analysis without considerable data reduction. Therefore, factor analysis was not considered an appropriate method to adopt in order to prove/disprove the proposed hypotheses.
Discriminant analysis, which involves computing 'discriminant scores' for each individual case in order to predict group membership, was also considered. Discriminant scores are obtained by establishing linear combinations of the independent variables. However, a prerequisite of discriminant analysis is that there is prior knowledge of group membership for the cases used to derive the classification rules (Norusis 1990). Although the group membership of all the children in the sample was known, the status of their oral motor skills was not known.
Cluster analysis was the chosen method for analysis for 2 reasons:
■ First, group membership for all 127 children included in the analysis was 'technically' unknown. That is, how the cases would be distributed in terms of normal/abnormal oral motor function was unknown.
■ Second, by using cluster analysis it was possible to identify homogeneous groups or clusters of cases and to study the characteristics or behaviours shared by each cluster. Finally, how each cluster differed from each other could also be studied.
The procedures adopted will be discussed in greater detail in the second part of this chapter, which details the results of the development study. Full details of the steps undertaken in the analysis are described in Skuse et al (1995) and Reilly et al (1995) in appendix 1. Statistical analysis were undertaken using SPSS-PC, version 4.0.