Few acoustic studies of the intervocalic consonant lenition in central Italian dialects (a process known as Gorgia Toscana) have been undertaken. This study examines speech data from Florentine Italian in order to describe the process of Gorgia Toscana quantitatively and to assess the roles of physiological, perceptual, phonological, and social factors in the process. Results of acoustic and statistical analysis indicate gradient and variable output, with certain patterns occurring in the variation. The observations that emerge from the data cannot all be accounted for if Gorgia Toscana is characterized as a purely phonetic, phonological, or socially driven process of sound change. Rather, different aspects of the process are attributed to different motivators: gradience and velar preference to articulator movements, resistance of nonvelar lenition to perceptual constraints, targeting of a natural class and categorical weakening to abstract featural representations, and intersubject variation in velar lenition to external social factors.

A surprising dissimilarity is attested in the perception of approximants by speakers of American and Standard Southern British English. Eighteen subjects (6 AE and 12 SSBE) performed an identification task in which they judged whether stimuli were more like /r/ or /w/. The stimuli comprised five sounds, copy-synthesised from a source /r/, where the values of F2 and F3 we adjusted to fall between the frequencies typical for ["turned r"] and [w]. The only significant difference between the two dialect groups' performance occurred with a token in which F3 was at a frequency typical for /r/ and F2 was lowered to that of /w/. AE and SSBE speakers identified this token as /r/ 90% and 59% of the time, respectively. This is unexpected as /r/ in both dialects is characterised by a low F3. However, the difference may be due to the existence of the 'labiodental' variant of /r/ in SSBE. As this variant does not have a low F3, SSBE speakers must tolerate a wider diversity of /r/-types than AE speakers. Therefore, the /r/ category in SSBE may be becoming increasingly defined by F2, rather than F3, which also has implications for the future production of /r/ in this accent

A safe, non-invasive imaging technique which captures articulatory movements in real time is the Holy Grail of articulatory analysis. Ultrasound imaging of the tongue seems suitable, albeit for a single articulator. However, ultrasound requires positioning a probe under the chin which may inhibit jaw movement. It is well known that speakers compensate rapidly for similar articulatory perturbations such as bite-blocks. Consequently, the images obtained by ultrasound may show compensatory articulations due to jaw inhibition, rather than natural articulations. This study assesses the effects of articulatory compensation during ultrasound imaging using acoustic analysis. The experiment compares acoustic data from three speech conditions: probe-free, probe held manually, probe fixed non-manually. It is hypothesised that sounds requiring a low jaw position, such as low vowels, velars, and velarised laterals, will be worst affected. High jaw sounds such as /s/ and /i/ are also assessed. Any differences between conditions are assumed to be due to compensation for jaw immobility rather than learned segment-specific register effects and are therefore unlikely to diminish as the subjects relax. Variability may diminish, however, as compensatory strategies become more practised. We expect to find some effects on certain segments, but that manual probe holding allows for more natural speech.