Interspeaker variations in V-to-V coarticulation: effects of Motor Speech Disorders, age, speech tempo changes, and boundary type

Anticipatory coarticulation refers to the anticipation of articulatory movements for the achievement of forthcoming speech segments and can be considered an index of planning in speech. Indeed, despite the accounts of speech production differ in modeling coarticulation, it can be taken as a starting assumption that if a segment is anticipated into a preceding one, the movements for producing it have to be already planned at the moment of executing the whole sequence. If anticipatory coarticulation would thus reflect the coordination of segments that have been planned and sent to execution as a whole, the presence of coarticulation in the acoustic output can be considered an index of cohesion within a single unit of planning. This dissertation attempts at addressing issues on speech planning by investigating interspeaker and intraspeaker variability of acoustic cues of coarticulation in French, focusing in particular on anticipatory V-to-V coarticulation. Throughout all studies, the linearity of the relationship between coarticulation and features linked to the temporal organization of speech such as vowel length, vowel-to-vowel lag or speech rate is also explored. In the first study, V-to-V coarticulation is investigated in different motor speech disorders, namely Apraxia of Speech (AoS) and Dysarthria associated with different underlying pathologies (Parkinson Disease, ALS, Wilson Disease), and it is compared to C-to-V carryover coarticulation, which could be considered to depend more on biomechanical constraints due to the unfolding of speech segments in time. The first goal of this study is to investigate whether anticipatory V-to-V and carryover C-to-V coarticulation patterns can distinguish AoS speakers, who present an impairment at the level of speech planning, and Dysarthric speakers, who present an impairment at the execution of motor programs. However, since AoS and Dysarthria, although different as for pathomechanisms, are both characterized by a deficit in the temporal organization of speech, presenting slower movements and/or syllabification, the second goal of this study is to investigate whether differences in coarticulation patterns between healthy controls and MSD patients can be accounted for by differences in the temporal characteristics of speech. The results show a reduction of V-to-V coarticulation and unimpaired C-to-V coarticulation for all groups of patients, suggesting that a reduction of V-to-V coarticulation is not specific to a deficit in planning, but it is more largely related to disordered speech. Even though this reduction is accompanied by vowel lengthening, at least for three out of four groups, a reduction in coarticulation in MSDs cannot be uniquely attributed to a slowing in movements. In the second study, Anticipatory V-to-V coarticulation and its relationship with articulation rate is further examined in adults aged from 20 to 93. Indeed, changes in coarticulation with age are well attested in childhood development, but less in adulthood. Since one well-known effect of aging is a slowing of speech, a reduction in coarticulation with age could stem from this decrease of speech rate. The results show that the amount of coarticulation continuously changes throughout adulthood, although not linearly: after a gradual reduction of coarticulation up to 50 y.o., coarticulation reduces more sharply after 70 y.o. More variable patterns emerge for the middle-aged speakers. If the expected slowing of speech with age is found, the reduction of coarticulation can be put in relationship with speech rate only at young age, while for older speakers, a decrease in the amount of coarticulation is not always linked to a decrease in rate. In the last two studies, inter- and intra-speaker variability of V-to-V coarticulation is investigated across different conditions. In the third study, coarticulation is examined in the speech of five speakers across two instructed speech tempo, slow and fast, compared to a comfortable self-paced rate, in order to isolate more directly the effect of speech rate on coarticulation. The results show that the effects of tempo changes on coarticulation are not straightforward: the expected reduction of coarticulation at slow rate, and increase of coarticulation at fast rate, is not found for all speakers, but coarticulation patterns are rather speaker dependent. In the fourth study, individual patterns of coarticulation are further studied in the same five speakers depending on boundary type (within a word, across word and clause boundary) over multiple repetitions, in order to investigate whether coarticulation within a word differ from coarticulation between two words, and whether inter and intraspeaker variability in coarticulation between two words is accountable for by individual difference in the prosodic rendition of the boundary. Results show individual differences in the amount of coarticulation across word boundary which cannot be entirely accounted for by differences in prosody. Moreover, inter and intraspeaker variability of coarticulation between two words and two clauses cannot be entirely accounted for by the degree of boundary strength, measured perceptually. Overall, the results evidence variations of V-to-V coarticulation due to pathology and age but also to speaker’s identity: speakers seem to adopt individual coarticulatory patterns, which cannot be related entirely to speech rate or to the degree of juncture between the two syllables. These results challenge a view assuming that anticipatory coarticulation is fully planned. In order to account for the variability found in anticipatory V-to-V coarticulation, one has to considered to introduce at the level of encoding of the speech plans, either (i) a mechanism which allows variability in the coordination of elements pertaining to a same speech plan, or (ii) a control on the size of the speech plans allowing flexibility according to speakers, prosody and speech conditions. In some cases, a certain variability in anticipatory coarticulation seems to reflect aspects of coordination in speech that are not planned in the sense of the first assumption.