Date: Tuesday 4 June 2024
Time: 13:00 - 14:00
Venue: Room 409, Campus Hub Block O & Online (via Zoom)
The upcoming event will be taking place in a hybrid manner, at Room 409, Campus Hub Block O and online.
Title and Abstract of this Seminar: Phonemic and lexical inventories are shaped by information transmission: two short stories.
The seminar will be conducted by Prof. Andy Wedel (Professor at the Department of Linguistics & PsyCoL Lab, University of Arizona, USA)
Abstract
Trajectories of change in language are shaped by a myriad of constraints and biases, including a bias toward information transmission accuracy in communication. In this talk I’ll tell you about two strands of my research program that investigate the ways that phoneme and lexical inventories are optimized through change to maintain communicative efficiency.
1) Chain Shifts and Transphonologisations are Driven by Homophony Avoidance.
Loss of a phoneme contrast through merger is known to be associated with a low degree of resulting word-level homophony. For example, there are very few English words distinguished by the two low-back vowels /É‘/ as in 'cot', and /É”/ as in 'caught', and this vowel contrast has merged in many dialects of North American English.
In contrast, phonemes that do not merge are characterised by many such 'minimal pairs', that is, words that would become homophonous if the phonemes were to merge. Here, we show that homophony avoidance appears to also drive two superficially distinct, active sound changes. Chain shifts occur when a set of phonemes move in concert within phonetic space. Transphonologisations, on the other hand, occur when the primary cue distinguishing two phonemes merges, while a minor cue expands in concert to become the primary cue.
These two superficially distinct classes of sound change have in common that lexical contrast is maintained throughout the change: in a chain shift, one phoneme moves into the space occupied by another, which concomitantly shifts away into a neighbouring part of the phonetic space. In a transphonologisation, one cue to a phoneme contrast merges, while at the same time another cue to the same contrast expands. Here I show that while phoneme mergers are characterized by a low number of minimal pairs (and therefore low numbers of resulting homophones), chain shifts and transphonologisations are characterised by especially high numbers of minimal pairs.
Finally, I show that this data strongly supports the standard conception of a phoneme inventory as a network of features shared across phonemes, rather than as a set of atomic phonemes.
2) High resource-cost sounds are concentrated in word beginnings and stressed syllables
Sound categories differ in the amount of effort required to produce an easily perceptible signal. For example, /k/ is a relatively low resource-cost sound, while /g/ is higher cost because it requires more physiological effort to maintain voicing at the velar position in the vocal tract.
In an information transfer system in which symbols have different costs, communication is most efficient when costly symbols are less frequent overall, while at the same time being preferentially allocated to positions in which they can convey the most information. Greater use of high cost sounds in information rich positions results in a more balanced range of sound contrasts at these positions, creating a higher entropy (i.e., more informative) lexicon overall. Here I show that higher cost sounds are preferentially allocated to two disparate positions which are high information for entirely different reasons, creating a higher entropy system in positions where it matters most.
(i) Word position: Listeners process sounds incrementally, where each successive sound gives a listener information that helps exclude incompatible words. As a consequence, sounds early in a word convey on average more information.
(ii) Syllable stress: Stressed syllables tend to be more perceptually salient than unstressed syllables, by being longer, louder, and/or higher pitched. Through these attributes, stressed syllables provide a superior platform for the perception of phonetic cues, independently of their position in the word.
In a set of nine genetically diverse phonemically coded languages and using the type of frequency of phonemes in a language as a proxy for resource cost (Everett 2018), we find that indeed, high resource cost sounds are concentrated at the beginnings of words, and in stressed syllables.
