Acquisition of Fricatives in Icelandic and French

I point out in this blog a recent paper by Bernhardt and colleagues describing fricative production by Icelandic preschoolers with “protracted phonological development”. Icelandic-speaking children with speech problems produced /f,s,θ/ with about 70% accuracy by four years of age. The English-speaking children with speech problems achieved less than 15% accuracy on the single word naming task (those of us who work with those children are not surprised!). Error patterns were different as well (for example, the Icelandic children substituted [t] for /f/ whereas the English children substituted [p] for /f/). The study is a nice reminder of the need for language specific articulation tests for children with intelligibility problems whose first language is not the majority language. There are no universal developmental patterns. Therefore we can be grateful to Sharynne McLeod and the International Working Group on Multilingual Children’s Speech for their great site.

A point of theory in the paper was the failure to support the hypothesis of “higher overall fricative accuracy for English, because English has more fricative types.” This hypothesis was put forward because Edwards, Beckman and Munson concluded that “the more words containing a sound that a child has learned to say, the more practiced the child becomes at recognizing and reproducing the sound abstracted away from the phonological contexts of a few specific words.” One example given is the case of /ð/ in English. I was surprised that an alternative hypothesis, considering the prosodic structure of Icelandic, was not tested. Perhaps the authors didn’t want to attempt this because it is so hard to understand the prosodic structure of Icelandic! I certainly will not try to do this. However I point out that French and Icelandic are characterized by earlier acquisition of fricatives compared to English:

Acq of Fric

Françoise and I have previously pointed out that the prosody of French helps children learn the segments. Several studies now confirm that French-speaking children learn consonants early: they have a complete consonant repertoire at an early age and they master accurate production of the consonants at an early age (compared to English). Furthermore, French-speaking children with speech problems produce more syllable structure and fewer segmental errors (compared to English-speaking children with speech problems). How does the prosody of French explain these findings? French is a syllable timed language with a tendency toward long words made of several syllables having equal stress. The syllables tend to have a simple CV structure even though complex onsets and codas are allowed. The difference between the two languages can be observed in the ratio of consonants to vowels, being 1.6 in French and 2.1 in English.

Recall that when children first learn to speak, they learn to produce “whole words” and not strings of individual segments; these first whole words act as templates for the production of more words. These initial templates are often made up of reduplicated or partially reduplicated syllables. Possibly, if the initial word shape template is easy to learn or construct, more processing power is left over to acquire the segments inside the syllables. Furthermore, simple stressed syllables in French may increase the perceptual salience of those segments. Vihman provides wonderful examples of this process in her very excellent paper on templatic phonology. The French child, Gael, saying words like ‘accroché’ [χʁoʃe] at age 21 months is rather fun to see!

Possibly the same process occurs in Icelandic which has a very particular prosody that may also increase the perceptual salience of postvocalic consonants, fricatives in particular. I will not attempt an explanation, the phonology of Icelandic being quite beyond me, but any explanation for the early acquisition of these phonemes must take the prosodic (word shape) characteristics of early child input into account.

In the meantime some more general points strike me. It is impossible for us as speech-language pathologists or academics to understand the process of speech development without data and it is disheartening how little data there is on these languages. Basic information such as the types and tokens of fricatives that are presented to children in the input across languages is difficult to find. We need to know much more about the prosody of all these languages but prosody receives scandalously little attention compared to, for example, consonants.

We also need to know more about the processes by which children select and construct their early word templates. This requires painstaking small sample detailed longitudinal work and large, large sample laboratory work using creative paradigms, some probably not invented yet. However, all our work is perilously underpowered and underfunded.

Finally we need huge diversity in academia – lots and lots of movement of students across the world, bringing new techniques to new languages. Unfortunately in the past few days, in the aftermath of Brexit, I have encountered many depressing conversations on the internet about the control of borders. Even in the domain of “highly qualified personnel”, some graduates are deemed more worthy than others and we all know where the humanities fall on the ranking! I prefer not to counter this sort of thinking with purely utilitarian arguments, but clearly it is short sighted given the importance of communication and communication disorders. On a more positive note I close by congratulating the excellence of child phonology projects sustained on a wing and a prayer such as the Cross-linguistic Child Phonology Project led by Bernhardt and colleagues and also the Paidologos project led by Edwards and colleagues.



Bernhardt, B. M., Másdóttir, T., Stemberger, J. P., Leonhardt, L., & Hansson, G. Ó. (2015). Fricative acquisition in English- and Icelandic-speaking preschoolers with protracted phonological development. Clinical Linguistics & Phonetics, 29(8-10), 642-665. doi:10.3109/02699206.2015.1036463

Brosseau-Lapré, F., & Rvachew, S. (2014). Cross-linguistic comparison of speech errors produced by English- and French-speaking preschool age children witih developmental phonological disorders. International Journal of Speech-Language Pathology, 16(2), 98-108.

Edwards, J., Beckman, M. E., & Munson, B. (2015). Frequency effects in phonological acquisition. Journal of Child Language, 42(02), 306-311. doi:doi:10.1017/S0305000914000634

MacLeod, A. A. N., Sutton, A., Trudeau, N., & Thordardottir, E. (2011). The acquisition of consonants in Québec French: A cross-sectional study of preschool aged children. International Journal of Speech-Language Pathology, 13, 93-109.

Rvachew, S., Marquis, A., Brosseau-Lapré, F., Paul, M., Royle, P., & Gonnerman, L. M. (2013). Speech articulation performance of francophone children in the early school years: Norming of the Test de Dépistage Francophone de Phonologie. Clinical Linguistics & Phonetics, 27(12), 950-968. doi:doi:10.3109/02699206.2013.830149


I am getting questions about our PHOPHLO project now that a part of it has inexplicably made Science Daily  so I will provide a summary of the project here with a few of the outcomes. This project is a collaboration between me and Laura Gonnerman at McGill University and Phaedra Royle at the University of Montréal. The testing of the children was coordinated by Phaedra’s post-doctoral student Alexandra Marquis under my supervision. The purpose of the project was to develop a screening test that could be administered to francophone children at the beginning of first grade to predict difficulties in the achievement of written language skills at the end of second grade. The project proceeded in three phases.

Pilot Phase. We began by administering a battery of tests of oral language skills to children in kindergarten and first grade classrooms in two schools. The tests included two tools developed by Françoise Brosseau-Lapré and me for the ECRIP trial, specifically a version of SAILS assessing the children’s perception of the word “gris” (French word for ‘grey’) and the Test de Conscience Phonologique Préscolaire, a measure of rime and onset matching that does not require any spoken responses. The third was a test of articulation accuracy developed by a former masters student Marianne Paul and subsequently validated by me and Françoise, Test de Dépistage Francophone de Phonologie . The fourth test was a measure of spoken morphological skills, Jeu de Verbes,  assessing the ability to produce appropriate past tense morphemes (a bit like a French ‘wug’ test), developed by Phaedra Royle at the University of Montréal. We also collected ratings of the children’s abilities from the teachers and conducted item analyses to develop a screening test from all this data that was a much shorter version of these four measures (44 items in total, about 20 minute administration time compared to 80 minutes). We have published two papers describing the children’s performance during Phase I, with comparisons across the unilingual French children and the multilingual French children (about 40% of the Phase I children spoke a language other than French at home; the language at school is 100% French). Rvachew et al. (2013), published in Clinical Linguistics & Phonetics  , reported that (1) there were no statistically significant differences in articulation accuracy across these two groups; (2) there was a slight tendency toward more errors involving the features [+voice] and [-anterior] nonetheless; and (3) all groups produced more errors in unstressed syllables. Marquis et al. (2012), published in Travaux interdisciplinaires sur la parole et le langage, reported that there were significantly fewer errors for the verb group [é] than the other verb groups tested.

Phase I. During this phase we administered the new screening test to 91 first grade children with approximately half being unilingual speakers of French and the remainder being multilingual. We also asked for teacher ratings of the child’s risk for future difficulties in the acquisition of writing skills and we piloted a measure of spelling abilities with a subset of these children in the latter half of the first grade year. Kendall Kolne, a doctoral student co-supervised by Laura and I, has just submitted a manuscript to Language and Literacy describing the relationship between the children’s performance on the Phase I tests, some other demographic measures,  and the teacher ratings. First, we found that language background, the education levels of parents, home literacy practices, and oral language skills (as measured by our screening test) all differentiated whether or not the children were identified by their teachers as at-risk for future writing difficulty. With respect to predicting first grade spelling skills, teacher ratings accounted for more variance than our screening test but the screening test and the teacher rating combined predicted the most variance in spelling ability (52% overall),and there was some unique information provided by the screening test. I leave it to the reader to decide whether it is good news or bad news that our 20 minute screen is as good as or is only as good as the teacher’s opinion! We did find a few interesting exceptions where the teachers relied overly much on family background information, leading to over and under estimation of risk for a few children. We have not yet done an analysis to find out how well the teacher ratings hold up over the longer term.

Phase II. In the final phase of the project we administered three tests of written language skills to the Phase I children when they reached the end of second grade (we were able to test 78 of the original sample at this time). The primary outcome measure was a standardized test of spelling that included nonwords, real words and sentences (BELO; Georges, F., & Pech-Georgel, C. (2006). BELO – Batterie d’évaluation de lecture et d’orthographe: Éditions Solal). We found that the screening test administered in first grade was a reasonably good predictor of BELO scores: the sensitivity of the PHOPHLO (i.e., proportion of true positives identified) was 66% while the specificity (i.e., proportion of true negatives identified) was 92%. I expect to submit a manuscript describing these results next year. We also administered two additional measures of written morphology skills developed by Phaedra Royle and Laura Gonnerman. Alex Marquis presented some of these results at the Romance Turn IV Congress in September 2014 and it is this presentation that made the big splash. They reported on the relationship between oral morphosyntax skills measured in first grade and written morphosyntax skills measured at the end of second grade. The written task involved reading sentences and choosing the correct one of three alternatives, e.g., Paul a/as/à une amie. (in this case the first alternative is the correct one). The interesting finding was that the multilingual children performed much more poorly than the unilingual children on the oral language task but there were no differences between these two groups on the written language tasks including the morphosyntax task or the spelling task. More interpretation is found in the Science Daily piece.

Post study phases: Laura Gonnerman is following these and other children into the later grades and confirms a mixed pattern of strengths and weakness in the multilingual group relative to the unilingual French group. She is also looking at the efficacy of certain teaching strategies. I am working with a company (iLanguageLabs) to create a software version of the PHOPHLO. It is nearing completion and I am hopeful that it will be available on the chrome store in 2016 for use by teachers and orthophonistes (SLPs). As they say, the proof of the pudding is in the eating, and in this case the proof will be in cross validation with a new and larger sample of children. I will seek funding and move forward with this phase when the app is fully operational. For myself, I am prone to be a bit cautious with my conclusions until the “proof” is in, which could be years yet.