On Atypical Errors and the Risk of Phonological Awareness Deficits

Jon Preston is publishing some amazing research on speech sound disorders (see for example this nice imaging study). He has recently published a paper showing a good sized correlation between atypical speech errors observed in preschoolers and phonological awareness skills three years later. The data are rather nice and remind us that preschool aged children with speech delay are at risk for delayed acquisition of phonological awareness (PA) skills. We care about this because the PA deficit likely explains the  association between speech sound disorders and spelling and reading disability in this population. I am going to comment further on the paper for two reasons: First, Jon repeats his claim that his findings are inconsistent with some that I published with Pi-Yu Chiang in 2007 and I would like to add some nuance to that; second, he presents some scatter plots to support the notion that atypical errors are of clinical utility as a marker for persistent delays in phonological awareness whereas I find this to be a somewhat dangerous idea.

First the nuance regarding my own findings on atypical errors: Pi-Yu and I found that PA deficits are associated with atypical speech sound errors, a finding that was significant and large for concurrent correlations in kindergarten aged children. The longitudinal (predictive) relationship from the prekindergarten assessment to the kindergarten assessment was weaker and not significant after applying the Bonferroni corrections which is why Jon says that his results and mine conflict. I would like to point out that one can never compare these kinds of results across studies in this fashion. First of all the strength of associations across variables is always dependent upon the shape of the distributions which is determined by the nature of the sample of children and the nature of the tests that are given.  With respect to the speech tests, Jon’s measure was better than mine and thus more sensitive to atypical errors. On the other hand, I used a “purer” measure of PA than he did: Bird, Bishop and Freeman’s Phonological Awareness Test was designed to require NO SPOKEN RESPONSES from the children whereas the test in Jon’s study, the CTOPP requires spoken segmenting and blending responses. It is known to be very difficult for preschool aged children, even those who do not have speech delay. Notwithstanding the difference in the difficulty levels of the two PA tests, it is no surprise that a measure that is a proxy for the severity of one’s speech disorder (frequency of atypical errors) is more strongly correlated with a PA test that requires SPOKEN RESPONSES. The CTOPP is like a test of phonological processing and articulation skills all rolled up into one. So it is completely inappropriate to compare the strength of the associations reported in these two studies and declare that they conflict because the findings are different. If the findings were the same (given that the tests were measuring different constructs) I would be worried because I have previously published findings showing that measures of implicit versus explicit PA give very different results even when testing the same children with the same items.

Even though we cannot compare the strength of the statistical outcomes in these two studies we can draw conclusions across all the studies that have looked at these variables and conclude without any doubt that there is some kind of relationship between atypical errors and PA (Jon and I both find that). Does that mean that SLPs should use the presence of atypical errors as a clinically significant marker for the risk of PA deficits? Not at all. I demonstrated this using logistic regression analyses in my study with Pi-Yu Chiang. We found that error type profiles were not a reliable predictor of PA outcomes one year later. Jon shows essentially the same thing with his lovely scatter plots (Figure 1). Six out of seven kids with lots of atypical errors as preschoolers failed the CTOPP 3 years later (good sensitivity). Ten out eighteen kids who passed the CTOPP in school had lots of atypical errors at the earlier testing (poor specificity).  It is true that none of the kids who scored above normal limits on the CTOPP had atypical errors!

What is most interesting about the data in the tables and figures in Jon’s paper is that the PA performance of these children is less than would be predicted by their Peabody Picture Vocabulary Scores at both time points (e.g, as preschoolers on average:  PPVT = 109, CTOP scores are 91, decoding scores are 96, situation 3 years later essentially unchanged – huge gaps between receptive vocabulary and PA and literacy performance). These children are not achieving their potential for emergent literacy during the preschool period or for reading skills in elementary school.

What is the SLP to take away from all of this?

1.  Phonological awareness is an important emergent literacy skill. It is our job to help the school implement high quality explicit instruction to all junior and senior kindergarten age children using an RTI approach because it is critical to have these skills in place before the children learn to read because interventions implemented after the kids have failed are not all that effective.

2. If there are any children in these class rooms with a current or past history of speech delay, watch these kids like a hawk and provide interventions as necessary. As a group, all of these kids are at risk for reading disability (Jon and I have each independently reviewed this literature and come to the same conclusion).

3. When you assess children with a speech problem, assess everything – articulation, language, and phonological processing (speech perception, phonological awareness, phonological memory, rapid automatic naming). Don’t make assumptions about any of these children based on the types of errors that you hear in their surface speech patterns. See my book for case studies – dyslexia can be associated with all manner of profiles, including distortion errors. Ironically, Jon Preston has published some nice work on phonological awareness problems on kids with distortion errors.

4. Remember that, given problems with phonological awareness, the risk of actual reading delay is moderated by other factors including quality of reading instruction, the child’s language abilities and persistence of the speech deficit. The SLP needs to help the school marshal all of its resources to handle all aspects of the problem in the most efficient manner. The SLP should treat the speech and phonological processing deficit in a coordinated fashion using an approach that targets all levels of phonological representation, as described in our book.