Phonological regularity, perceptual biases, and the role of phonotactics in speech error analysis

Abstract

Speech errors involving manipulations of sounds tend to be phonologically regular in the sense that they obey the phonotactic rules of well-formed words. We review the empirical evidence for phonological regularity in prior research, including both categorical assessments of words and regularity at the granular level involving specific segments and contexts. Since the reporting of regularity is affected by human perceptual biases, we also document this regularity in a new data set of 2,228 sublexical errors that was collected using methods that are demonstrably less prone to bias. These facts validate the claim that sound errors are overwhelmingly regular, but the new evidence suggests speech errors admit more phonologically ill-formed words than previously thought. Detailed facts of the phonological structure of errors, including this revised standard, are then related to model assumptions in contemporary theories of phonological encoding. This article is categorized under: Linguistics > Linguistic Theory Linguistics > Computational Models of Language Psychology > Language.

Tone slips in Cantonese: Evidence for early phonological encoding

“…Likewise, the deletion of syllable final [p] in sa [p] → sa results in the short vowel [a] in an open syllable, which is again outlawed by Cantonese phonotactics. The frequency of errors with phonotactic violations is 10.8%, which is a bit higher than what has been reported for English (Alderete & Tupper, 2018). This may be due to the relatively high number of non-native sounds in the Cantonese corpus, which accounts for roughly half of these cases.…”

“…There are, nevertheless, certain kinds of errors that are unambiguously segmental for the simple fact that they are impossible syllables. While sound errors tend to obey phonotactic rules (Boomer & Laver, 1968;Wells, 1951), meaning that sounds generally slip into well-formed syllables, recent evidence has shown that this constraint is far weaker than previously assumed (Alderete & Tupper, 2018).…”

“…As documented in Alderete and Davies (2018), prior studies have detection rates (an error on average every 5 to 6 minutes) that undershoot our detection rates by a wide margin. Our corpus also has a relatively low rate of uncommon but highly salient errors, like sound exchanges, and higher rates of phonotactic violations, which are further indicators of higher data reliability (Alderete & Tupper, 2018).…”

Generative Adversarial Phonology: Modeling Unsupervised Phonetic and Phonological Learning With Neural Networks

“…Human subjects are able to evaluate and produce nonce-words even if a string of phonemes violates language-specific phonotactics, as long as the basic universal phonotactic requirements that treat phones as atomic units are satisfied (for an overview of phonotactic judgments, see Ernestus, 2011 and literature therein). Deleting or inserting segments are also common patterns in both L1 acquisition (Macken and Ferguson, 1981), loanword phonology (Yildiz, 2005), in children with speech disorders (Catts and Kamhi, 1984;Barlow, 2001), as well as in speech errors (Alderete and Tupper, 2018b). For example, #sT clusters are often simplified in L1 acquisition (Gerlach, 2010).…”

The adaptable speaker: A theory of implicit learning in language production.

“…For example, Stemberger (1983) reported that only 1% of the phonological errors in his English slip corpus overtly violated English phonotactics. This 1% rate of violations is likely an underestimate, though, because of perceptual biases that can occur during error collection (Alderete & Davies, 2019; Alderete & Tupper, 2018).…”