Syllabic complexity effects in phonological speech errors: The role of articulatory-phonetic impairment

Research with native Italian people with aphasia (PWA) argues that syllabic complexity effects in phonological speech errors are based in articulatory encoding impairments [1, 2]. This important claim has yet to be substantiated for English. Indeed, one influential study reported no effects of syllabic complexity (a.k.a, “markedness”) on phonological error production [3]. We present evidence from an ongoing study that corroborates and extends the main claims of the Italian studies.

In [1], markedness effects were contrasted in PWA with high vs. low proportions of phonetic-articulatory errors in single word repetition. We measured this “phonetic error proportion” (PEP) in 121 diverse, English-speaking PWA and found a graded distribution ranging from .00 to .55 with Mn = .10 and SD = .12. PEP scores correlated positively with apraxia of speech (AOS), WAB AQ, and lesion size (all r > .35; p < .001).

From the sample of 121, we identified those who made > 20% phonological errors on the Philadelphia Naming Test. This identified 22 PWA, with a suitable range in PEP (.03-.55), AQ (25-84), and lesion volume (8-376 cc) and ample phonological naming errors to analyze (Mn. 47; range18-94). Half had AOS (measured as in [4]). For each individual, we calculated several structure-change proportions: we counted how many of their phonological errors instantiated a particular syllable-structure change and divided the number by the opportunities for that change present in the PNT targets. We report on two changes affecting the pre-vocalic (onset) position of a syllable:
• fills an empty onset (onset creation) (e.g., VC→CVC) – decreases complexity
• deletes a filled onset (e.g., CVC→_VC) – increases complexity
and two changes affecting consonant clusters:
• deletes a consonant cluster (e.g., CVCC→CVC_) – decreases complexity
• creates a consonant cluster (e.g., CV→CCV) – increases complexity

Fig. 1 shows the median change proportions, broken down by locus of change (1st, 2nd, or 3rd syllable). Predictably, there were far fewer opportunities for changes that decrease syllable complexity (see insert), yet the corresponding change proportions (onset creation; cluster deletion) were found to be high relative to those that increase complexity. Moreover, onset creation and cluster deletion proportions correlated strongly (r=.66), suggesting a common underlying mechanism.

To investigate this mechanism, we computed an overall effect size for complexity reduction, reflecting the difference between structure decreasing and structure increasing changes:

(onset creation + cluster deletion) – (onset deletion + cluster creation)

We then used simultaneous multiple regression to predict this effect size from PEP, WAB AQ and lesion volume. Adjusted R2 for the model was .50 (F = 7.8; p = .001), and the strongest predictor was PEP (beta = .73; t = 4.43; p < .001). WAB was marginally significant (p=.05).

In conclusion, our study corroborates the influence of syllabic complexity on phonological errors in naming and shows that the tendency for errors to transform marked structures to unmarked ones correlates with the severity of the phonetic-articulatory involvement, after controlling for aphasia severity and lesion size. This has implications for the theoretical treatment of phonological error production and its relation to AOS.