An Acoustic Study of Romanian Stressed Vowels with Special Reference to Mid Central [ɨ] and [ə]

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The manuscript describes an acoustic analysis of Romanian vowels, with a focus on the high and mid central vowels in comparison to the peripheral vowels. The dataset is quite small, involving just three talkers, and the results are presented descriptively without any statistical analysis. The interpretation of the results in the context of previous work on Romanian and other languages is quite vague, making it difficult to assess the novel contribution of the work. I elaborate on these points below and then provide a few specific comments.

 

The phonological status of the central vowels is critical for predicting their acoustic properties, both for Romanian and for the other languages to which the Romanian data are compared throughout the manuscript. Are the two central vowels phonemic in Romanian? How many phonemic central vowels are found in Khmer, Sundanese, Thai, and the other languages to which Romanian is compared? The contrastive status of the two central vowels in Romanian is finally stated at the end of p. 2, but this information should come much sooner, ideally in the very first paragraph of the paper. Information about the number and phonemic status of central vowels in other languages should be presented whenever the languages are first mentioned. And, given that greater variability is predicted for non-phonemic central vowels that result from vowel reduction, as in English, only languages with phonemic central vowels should be compared to Romanian.

 

The results are very strongly interpreted without any statistical support, based on a relatively small data set. Some examples of this likely over-interpretation of the data include: What is the evidence for greater variability in F2 than F1 for the back vowels? This pattern is not clearly shown in any of the figures. What is the evidence that the observed variability in F2 for the back vowels is due to consonantal context? No F2 data are provided separately by consonantal context in the paper. These are just two examples, but there are many more throughout the results section.

 

Throughout the results and discussion sections, the Romanian data are compared to other languages, without either specification of those languages or citations to previous work on those languages. As a result, it is not clear if the Romanian data are really different from what we already know about Romanian or other languages and/or what the novel contribution of the study is. Much greater specificity about previous work is needed throughout the manuscript to understand the current data and its significance for the field.

 

Specific comments:

p. 1, lines 38-41, The predictions that F1 of schwa is 500 Hz and that F2 of schwa is 1500 Hz only apply to vocal tracts of a particular length (roughly 16cm). The more general prediction is that a neutral vocal tract, such as that for schwa, will produce an F2 that is three times the value of F1, independent of vocal tract length and, therefore, any particular F1 value.

Figure 1, The bottom panel is missing vowel (IPA) labels for each point.

p. 7, lines 253-256, Can you provide citations for the claim here about trends in vowel variability across languages?

Comments on the Quality of English Language

Careful proofreading of the entire manuscript is needed to correct typos.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

see attached. 

Comments for author File: Comments.pdf

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

Please see pdf attached.

Comments for author File: Comments.pdf

Comments on the Quality of English Language

There are occasional disfluencies (see comments) and individual sentences can be difficult to read due to their length. However, for the most part, the paper is written in good academic English. 

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

The authors have added some additional information and citations about the various languages they discuss, as well as data from three additional speakers and a statistical analysis of the duration data.

 

However, I remain confused about the primary purpose of the paper. It is now clear from the first paragraph that Romanian has two central phonemic vowels, like many other languages of the world. The phonemic status of these vowels clearly differentiates them from vowels with similar quality that may be transcribed with the same IPA symbol (e.g., schwa) in languages such as English, where they are non-phonemic and occur only as reduced vowels in unstressed syllables. And, indeed, the data from the previous studies shown in Figure 1 clearly show that the Romanian central vowels pattern in acoustically sensible ways relative to the other vowels in the system, as would be expected given their phonemic status. Given the established phonological and acoustic-phonetic facts about Romanian vowels, what is the central question that this paper seeks to address? And to what extent are data from other languages, especially languages where these vowels are unstressed, relevant to that central question? Put another way, what is the novel contribution of the work? 

 

In considering these questions, it is essential to bear in mind that the same IPA symbol is often used to indicate phonetically different sounds across languages and that acoustic properties such as formant frequencies vary across talkers due to physiological differences in things like vocal tract length, so we shouldn’t expect that particular acoustic values reported for one sound in one language in one study will correspond directly to acoustic values for a sound transcribed using the same IPA symbol in another language in another study.

 

The formant results are still interpreted qualitatively, without any statistical support, including the consonant context effects shown in Figure 2 (although it’s not actually clear in this figure which symbols correspond to which consonant contexts) and the overall variability effects shown in Figure 5. The interpretation of the results also does not seem to follow clearly from Figures 2-4, which show considerable overlap in the ellipses, both in general and for the central vowels, but which are interpreted as showing “very little” overlap (line 422) and “the absence of overlap” (lines 519-520).

 

Specific comments:

Lines 97-99, Please provide citations for the claim that the schwa symbol is “usually used for reduced vowels” in languages of the world.

Lines 204-206, Please provide a citation for the claim about relative vowel duration in high vs. mid vs. low vowels.

Author Response

REVIEWER 1

The authors have added some additional information and citations about the various languages they discuss, as well as data from three additional speakers and a statistical analysis of the duration data.

However, I remain confused about the primary purpose of the paper. It is now clear from the first paragraph that Romanian has two central phonemic vowels, like many other languages of the world. The phonemic status of these vowels clearly differentiates them from vowels with similar quality that may be transcribed with the same IPA symbol (e.g., schwa) in languages such as English, where they are non-phonemic and occur only as reduced vowels in unstressed syllables. And, indeed, the data from the previous studies shown in Figure 1 clearly show that the Romanian central vowels pattern in acoustically sensible ways relative to the other vowels in the system, as would be expected given their phonemic status. Given the established phonological and acoustic-phonetic facts about Romanian vowels, what is the central question that this paper seeks to address? And to what extent are data from other languages, especially languages where these vowels are unstressed, relevant to that central question? Put another way, what is the novel contribution of the work? 

The goals of this investigation are well defined in the Introduction section.

(line 55) “The research goals of the present investigation are as follows”.

(lines 56-105) “Firstly, we will focus on the placement of mid and central vowels in the vowel space…”.

(lines 111-203) “A second research issue concerns acoustic variability. A salient characteristic of stressed schwa in the world’s languages is its high degree of spectral variability as a function of segmental context and therefore its high degree of sensitivity to coarticulatory effects in tongue and lip position exerted by the flanking consonants…The possibility that stressed central vowels including schwa may be produced with some active tongue configuration and consequently turn out to be less variable than expected should hold precisely in languages like Romanian where [ɨ] and [ə] have phonemic status…”.

(lines 209-228) “A final goal of the present investigation is to compare the duration of the two central vowels…”.

Taking into consideration the reviewers’ advice, the central goal of the study is summarized in a new paragraph (lines 204-208).

 

Regarding another comment made by the reviewer, it is important to indicate that the present study deals exclusively with central vowels in stressed position, not with unstressed schwa.  Acoustic data from the literature show that phonemic stressed schwa does not differ much from non-phonemic unstressed schwa irrespective of the number of stressed central vowels in a given language. Indeed, as reported in the Introduction section (lines 111- 118), in most languages with stressed schwa and no [ɨ] schwa differs from peripheral vowels in being, very much like unstressed schwa, highly variable and shorter. In the few languages and dialects with (phonemic) stressed [ɨ] and [ə] for which we have acoustic data, stressed schwa turns out to be less variable and overlaps minimally with [ɨ] (lines 174-182). Moreover, it is commonly the case that the formant frequency and acoustic variability data reported in previous studies have not been obtained in highly controlled experimental conditions, which is why we decided to undergo the present investigation by focusing on a single language (Romanian).

 

In considering these questions, it is essential to bear in mind that the same IPA symbol is often used to indicate phonetically different sounds across languages and that acoustic properties such as formant frequencies vary across talkers due to physiological differences in things like vocal tract length, so we shouldn’t expect that particular acoustic values reported for one sound in one language in one study will correspond directly to acoustic values for a sound transcribed using the same IPA symbol in another language in another study.

We are aware that the same IPA symbol is often used to indicate sounds that may differ acoustically as well as articulatorily across languages.  This is why in order to determine which is the most suitable IPA symbol for transcribing stressed schwa it is so important to have formant frequency values for all peripheral vowels available in the language vowel system. This should allow us to ascertain whether stressed schwa is rather front or rather back as well as higher or lower. Based on these considerations, we evaluate how suitable are the phonetic symbols which have been proposed elsewhere for transcribing the Romanian stressed mid central vowel. Some authors, such as Renwick, have proposed [ʌ]. Based on the F1/F2 data for six speakers reported in our investigation, we consider the use of [ə] to be more appropriate, which is in accordance with the traditional practice and with modern authors such as Chitoran (2001: 8).

Chitoran, Iona. 2001. The Phonology of Romanian. Berlin/New York: Mouton de Gruyter.

 

The formant results are still interpreted qualitatively, without any statistical support, including the consonant context effects shown in Figure 2 (although it’s not actually clear in this figure which symbols correspond to which consonant contexts) and the overall variability effects shown in Figure 5. The interpretation of the results also does not seem to follow clearly from Figures 2-4, which show considerable overlap in the ellipses, both in general and for the central vowels, but which are interpreted as showing “very little” overlap (line 422) and “the absence of overlap” (lines 519-520).

               Statistical results are now provided showing that the F1 frequency contrast between the two mid vowels is highly significant (see lines 440-454). The consonant context effects in F1 and F2 frequency have also been evaluated statistically for the two central vowels (lines 455-470) while the corresponding effects for the peripheral vowel are still characterized descriptively (running statistics on the C-to-V effects for all vowels would involve changing substantially the formulation of the main goal of the paper in the Introduction section). In order to render all these contextual effects clearer to the reader we now include the F1/F2 values for the relevant consonants in the body of the text since phonetic symbols cannot possibly be inserted in the figures (see lines 385-388, 391-393, 401-403). Differences in formant frequency variability among vowels may be ascertained from the standard deviation values provided in Figure 5 in a more straightforward way than from the ellipses displayed in Figures 3 and 4.

We are now being more precise regarding the issue the F1 frequency overlap between the two central vowels in the paragraph coming just before Figure 5 and have also made changes regarding the same issue in the 3rd paragraph of the Discussion,

Specific comments:

Lines 97-99, Please provide citations for the claim that the schwa symbol is “usually used for reduced vowels” in languages of the world. Done.

Lines 204-206, Please provide a citation for the claim about relative vowel duration in high vs. mid vs. low vowels.Done

Reviewer 2 Report

Comments and Suggestions for Authors

I appreciate the opportunity to review this revision of "An acoustic study of Romanian stressed vowels with special 2 reference to mid central [ɨ] and [ə]." The authors have addressed or responded to my previous comments, including by doubling the number of speakers in the study and elaborating their methods for acoustic analysis. 

My remaining comment regards the authors' advocacy for reconsideration of /ʌ/ or [ʌ] as the IPA symbol of choice for the Romanian mid central vowel. Some phoneticians have argued that [ʌ] is not appropriate, because the vowel's phonetic realization is not sufficiently backed, with lowered F2; therefore the Romanian vowel does not meet the phonetic criteria of the International Phonetic Alphabet for the use of [ʌ]. Are the authors advocating (potentially) for [ʌ] on phonetic grounds, or because there is a precedent in Renwick's work for using [ʌ]? Since, as Recasens (2021) points out, there are many languages with stressed schwa, why should Romanian be treated differently, using [ʌ] rather than [ə]? (Or, are the authors arguing that schwa should be restricted to unstressed/reduced syllables in all languages?) 

I hope the authors will be able to address this question, because it is the linguistic debate that their paper can speak to. 

Author Response

REVIEWER 2

I appreciate the opportunity to review this revision of "An acoustic study of Romanian stressed vowels with special 2 reference to mid central [ɨ] and [ə]." The authors have addressed or responded to my previous comments, including by doubling the number of speakers in the study and elaborating their methods for acoustic analysis. 

My remaining comment regards the authors' advocacy for reconsideration of /ʌ/ or [ʌ] as the IPA symbol of choice for the Romanian mid central vowel. Some phoneticians have argued that [ʌ] is not appropriate, because the vowel's phonetic realization is not sufficiently backed, with lowered F2; therefore the Romanian vowel does not meet the phonetic criteria of the International Phonetic Alphabet for the use of [ʌ]. Are the authors advocating (potentially) for [ʌ] on phonetic grounds, or because there is a precedent in Renwick's work for using [ʌ]? Since, as Recasens (2021) points out, there are many languages with stressed schwa, why should Romanian be treated differently, using [ʌ] rather than [ə]? (Or, are the authors arguing that schwa should be restricted to unstressed/reduced syllables in all languages?) I hope the authors will be able to address this question, because it is the linguistic debate that their paper can speak to. 

In order to account for the reviewer’s comment we have made a few changes in the 5th paragraph of the Introduction section (lines 93-98), the 2nd paragraph of the Results (lines 352-354) and the first paragraph of the Discussion (lines 546-558). Based on phonetic, not phonological, grounds we state that [ə] appears to be the most suitable IPA symbol for Romanian stressed schwa and that the symbol [ʌ] could also be used in the event that data from more speakers reveal that F2 for this vowel overlaps quite often with F2 for the mid back vowel. The placement of the vowel ellipses in Figure 2 suggests that [ʌ] rather than [ɤ] could perhaps be chosen in addition to [ə].

Reviewer 3 Report

Comments and Suggestions for Authors

Not many comments, but see attached file

Comments for author File: Comments.pdf

Comments on the Quality of English Language

The disfluencies mentioned in the earlier draft were addressed. The writing style is appropriate and and mostly fluent.

Author Response

REVIEWER 3.

(lines 95-98) The use of [] does not necessarily imply that the transcription is motivated on phonetic grounds. As such, I once again would like to point that the findings of this paper do not contradict the transcriotion standard relayed from Renwick.personally I am not convinced that [ʌ] is an inappropriate symbol to transcribe Romanian schwa, if this transcription is motivated phonologically. I would revise this comment slightly  to indicate that Renwick’s transcription is not (necessarily) faithful to the phonetic realization of Romanian schwa, which is the purpose of the current paper.

In order to account for the reviewer’s comment we have made a few changes in the 5th paragraph of the Introduction section (lines 93-98), the 2nd paragraph of the Results (lines 352-354) and the first paragraph of the Discussion (lines 546-558). In any case we find it hard to understand why a different symbol from [ə] should be used just for phonological reasons mostly since Renwick uses /ʌ/ and [ʌ] for schwa not only in stressed position but in unstressed position as well and F2 values about 1200-1300 Hz for this Romanian vowel are only available in those consonantal contexts which cause the lowest vowel F2 to occur.

(line 246) “All subjects who took part in the recrding procedure were familiar with some regional words which were available in the real word list reproduced in the Appendix A”. I Imagine that the authors’ meant to say ‘all regional words’ (as in: for every regional word, the subects were familiar with it). Done.

Round 3

Reviewer 1 Report

Comments and Suggestions for Authors

The authors have softened their conclusions to better reflect the data and added some clarification related to the goals of the paper, as well as a statistical analysis of the formant data to support their claims.

Author Response

Some changes have been carried out regarding the contextualization of the present study with respect to previous contributions (see lines 34-38, 46-48, 76-77, 121-124, 215-216 and 176 of the Introduction section), the formulation of the major research goals (lines 235-244) and the description of the results and their summary and discussion (lines 360-363, 394-396, 399, 444-445, 448-450 and 625-626).