Dec 6 ===== Johnson:2004 ------------ Why are speakers able to produce clusters in reduced phrases that are illegal in regular speech? For instance, it is known that English native speakers cannot say the /ks/ cluster in the onset "even if they try". Why can they do that in [kʰsif]? Why was the sample limited to middle-class Caucasians? The author introduces the concept of lexical islands of reliability. How is this determined in a word, and can it be predicted? Not a question but a comment on how Johnson's findings align with what is observed in historical language change in Portuguese. The European variety has undergone a strong massive reduction in conversational speech, to the extent that sometimes it's practically unintelligible for speakers of the Brazilian variety (ex. 'telefone', which in BP is pronounced syllable-by-syllable, but in EP it would sound something like [tfon]). It seems that EP aligns more closely with Johnson's observations of casual American English speech, which suggests that speakers of this variety rely more heavily on phonetic reduction strategies and contextual cues to parse speech. It'd be interesting to do a case study to see how these changes took place over time and whether it was more linguistic (coarticulatory efficiency and articulatory economy) or sociolinguistic factors (speech rate, formality, language contact) that has conditioned this change. Since coarticulation and massive reduction are so widespread in everyday speech and across many languages, why has it taken so long to recognize that traditional auditory word recognition systems do not adequately consider and account for these phenomena? Is this related to the interaction/tension between linguists and computer scientists? "The sample was stratified for age (under 30 and over 40) and sex" -> Would have been interesting to follow up on this and see if these factors had any impact on reduction rate. Table 2 -> The word until is an interesting case because the abbreviated form till/'til exists and is perhaps a bit more “accepted”/mainstream than deviations of other words. "[wɛn] from when is different from [wɛn] from went" -> If two forms are actually phonetically different, as is claimed in this example, shouldn’t there be a way to express this in the IPA representation through something like diacritics? I've always thought that foreign language classes should focus more on teaching common reductions, since this seems to cause a lot of comprehension difficulties for L2 speakers but is rarely explicitly an object of study (in my experience). Why did function words have higher deviation and deletion rates than content words? What is the relation between word length and deletion/deviation rate? Why was the sampling limited to middle-class Caucasians in the 'Variation in Conversation' (ViC) corpus? Why are 1- and 2-syllable words "relatively immune" to massive reduction? The experiment involved analyzing conversations between participants and an interviewer who had been specifically trained to conduct sociolinguistic interviews. Could it be possible that the experimental setting (i.e., speaking with an unfamiliar researcher) had an influence on the results, potentially reducing any observable reduction effects? The formality of the situation could have encouraged participants to adopt a more formal speaking style than they typically use in everyday conversational speech. The findings of the paper suggest that function words exhibit a greater degree of segmental deviation compared to content words. To what extent might word frequency contribute to this pattern? That is, deviating from a word's citation form could require less effort during speech production. Consequently, adopting such deviations for high-frequency words, like function words, might offer a more effective reduction in production effort compared to applying similar deviations to low-frequency words, such as content words. Noting that in the current study, only natives of Central Ohio were considered, how generalisable is massive reduction across all speakers of American English? Surely reductions of this degree would be highly influenced by regional dialects and other factors like age and social status. How often would we find the same patterns in reduction among people from different regions? How exactly are certain consonant clusters in english deemed illegal? Is it purely because they do not usually appear in natural non-reduced speech? Often within the Examples section, it is seen that a lot of vowels get replaced by aspiration/breathy voice, when followed by p,t,k stops. What is this behaviour attributed to? This is observed to a great extent in the case of "particular", where all three stops are followed by aspiration. Is it just another way of stressing the consonant whenever the vowel has been deleted? Schatz/etal:2017 ---------------- The authors state that "the existing acoustic measures of articulation suffer from a lack of robustness and are not systematic, and both acoustic and direct articulatory measures are more descriptive than functional". The measure based on the Mutual Information (which should've been known by the authors) despite its flaws, seems pretty robust and systematic. How, if at all, is their method better than that? I wonder if the data collection method for the two languages had any impact on the results. Since the English data comes from reading newspaper articles out loud whereas the Japanese data comes from spontaneous speech, I would suspect it’s possible that the English data was biased towards having somewhat less coarticulation, which would line up with the consonant results albeit not with the vowel results.