Acoustic Variation of Syllable-Initial [ŋ] in Zhengding

The study explores the acoustic properties of syllable-initial [ŋ] in Zhengding dialect, to see whether the younger generation shows the same pattern with the senior group. 60 items with vowel realizations [ʌ, a, ɑ, ə, ɤ] and [ai, ɑo, ou] in „[ŋ]-V‟ and „[g]-V‟ structures are produced by 8 native speakers. Three experiments are conducted. Experiment I compares „[ŋ]-V‟ and „[g]-V‟ structures in senior speeches. Three acoustic effects due to the initial [ŋ] are established: vowels become less distinctive from each other by decreasing the first formant (F1), increasing the second formant (F2), and shrinking the gap between the second formant (F2) and the third formant (F3). Experiment II is conducted between „[ŋ]-V‟ and „[g]-V‟ in the younger speakers, investigating whether they have a similar pattern with the seniors. Experiment III is supplemented to compare the younger speeches in Zhengding dialect and Mandarin, to explore whether the generational variation in Zhengding dialect is relevant to dialect contact, i.e., whether the younger speakers are largely influenced by Mandarin. The result shows the younger generation does not produce the initial [ŋ] with the vowel realizations [ʌ, a, ɑ, ə, ai, ɑo, ou], which traditionally have an initial [ŋ], with an exception in [ɤ]. A fusion process is assumed in [ɤ] in the younger pattern, in which the initial nasal [ŋ] and the following vowel [ɤ] are combined into the single nasalized vowel [ɤ ], with the nasal effects remained, but the initial nasal then deleted. From the sociovariationist perspective, the nasal-initial pronunciation is a partial variation in Zhengding dialect. Not all speakers pronounce with the velar-initial [ŋ]. The older generation largely remained the velar-initial variant, but the younger generation preferred the zero-onset, which might be due to the influence of dialect contact with Mandarin.

The velar nasal [ŋ] is also found in syllable-initials in Zhengding dialect (Annals of Zhengding County 1992;Song 1999), a subtype of northern varieties. Zhengding county is located in central Shijiazhuang, in the southwest of Hebei province in China. It is 15 km to the north of the urban area of Shijiazhuang city, and 258 km to the south of Beijing, the capital city of China (Zhengding County Annals, 1992). In Zhengding dialect, three nasal initials are licensed: [m, n, ŋ]; and five vowels are allowed to co-occur with the initial [ŋ]: /a, e, ai, ɑo, ou/ (Annals of Zhengding County 1992), among 9 vowel inventories, as seen in (2).
(2) Vowel Inventories in Zhengding dialect (Note 2) The precise realization is depending upon the phonetic environment. The vowel phoneme /ɑ/ is phonetically realized as [ʌ], when it follows a consonant or used independently. It is realized as [ɑ] in compound finals (Fuyunmu in Chinese) ao and ang; and realized as [a] in compound final ai and an. The vowel phoneme /e/ is realized as [ə] in the compound final en, and as [ɤ] when following a consonant or used by itself (based on http:// web.csulb.edu/~txie/461/ Readings/PinyinandIPA), although there were arguments on different vowel variations in oral Chinese dialects (Flege, Bohn and Jang, 1997;Li and Thompson, 1981;Chen & Wang, 2013). Thus eight vowel realizations [ʌ, a, ɑ, ə, ɤ, ai, ɑo, ou] are licensed with initial [ŋ] in Zhengding dialect. Some examples on the word-initial contrast between Zhengding dialect and Mandarin are illustrated in (3), in which the initial [ŋ] is selected in Zhengding dialect, but prohibited in Mandarin. ( "hungry" (Zhu 2007: 47) The goal of the study is to see whether all natives in Zhengding dialect produce with the initial [ŋ], and how this sound affects the phonetics of the following vowels in different generations. 8 native speakers, four senior (>/= 45 years old) and four younger (<45 years old), are recruited for participation. It is hoped to deepen the perceptional recognition of the nasal onset [ŋ] in Zhengding dialect, and to specialize the phonological consequence of the generational variations.
The remainder of the paper consists of five sections. Previous analyses on initial [ŋ] are presented in section 2. Section 3 introduces fusion process in sound changes, which is applied to explain the sound variation in the young speakers. Section 4 describes the methodology, involving the experiment setup and data analysis. Section 5 presents the results of three acoustic experiments, to investigate whether the young generation shows the similar pattern with the senior speakers, traditionally with the initial [ŋ]. I propose a fusion process undergoes in the young speaker in the vowel [ɤ]. It also suggests dialect contact between Zhengding dialect and Mandarin, contributes to the generational variations in Zhengding dialect. Section 6 presents the conclusion and the limitations about this study.

Previous Studies of Initial-[ŋ]
The literatures on initial [ŋ] is outlined in this section, involving the distributional and phonotactic characteristics of syllable-initial [ŋ], the origin of the word-initial [ŋ] in cross-linguistically, the current studies on the initial [ŋ] in Cantonese and Shijiazhuang urban dialect from the socio-variationist perspectives, and the acoustic measurements for vowel formant changes.

Distributional and Phonotactic Studies of Initial-[ŋ]
The velar nasal [ŋ] was acknowledged as one of the most definable sounds in areal and phonotactic distributions across the world (Anderson 2005: 217). Anderson (2013) claimed the sound [ŋ] existed in 48% of the 100-language sample in the World Atlas of language structures. It was included in "languages of North America (especially California), Australian languages, also in Burushaski, Munda languages and the Tibet-Burman family", and many others. It lacked in "most Dravidian and Indo-European language of South Asia" (Anderson 2013). It may be controversial whether this sound is phonemic (contrastive) or phonetic in some languages. For example, in Japanese, Martin (1952: 12), Jorden (1955: 3), and Bloch (1958: 334) proposed [ŋ] as a separate phoneme (Goodman 1968: 153); while Goodman (1968: 154;156) argued it was the allophone of [g], and [ŋ] only occurred medially in phonological words.
For the phonotactic characteristics of [ŋ], Anderson (2013) reported [ŋ] might occur in onset or coda word-peripherally (word-initially or word-finally) or word-medially. The velar nasal [ŋ] in Polish was fully determined by contexts. It could appear only before the velar plosive [k, g], such as in kongo, tango, ręka and drąg. However, in English it was determined by phonetic contexts to a lesser extent. Although it was impossibly initial, it might occur medially, such as before consonants in England, anchor, and anxiety. It was also licensed finally in tongue and among. Cross-linguistically, the nasal [ŋ] in the word-initial was more restrictive than that in the word-final or word-medial position. The word-initial [ŋ] was found "in languages as Bagirmi, Canela-Krahô, Chukchi, Indonesian, Mangarrayi, Meithei, Oksapmin, and Rama among the 100-language sample, but forbidden in a large number of European, western Asian, western and central Siberian languages" (Anderson 2013).
For the initial [ŋ], Blevins and Kaufman (2012) studied its origins in Palauan, from morphological instead of phonological perspectives. Palauan is "a Malayo-Polynesian language with a long history of independent development". Blevins et al. (2012) argued that the nasalization at word or phrase boundaries in Palauan did not result from natural phonetic motivations, but from a reanalysis of a particle *ŋ. A third person singular marker "ŋ" in the word-initial position is reanalyzed as a clitic. This marker combined with the following vowel into a phonological word, as a result of pro-cliticization, as seen in (4).
(4) Syntactic word Phonological word (Blevins 2012: 23)  The morphological analysis on the origin of the word-initial [ŋ] in Palauan provides a new approach, different from the regular sound changes with phonological motivation. However, for Chinese, which is impoverished in morphology, the phonetic variation from zero-initial to velar nasal [ŋ] in some dialects is widely considered as a process of sound change across time.

Socio-variationist Perspective for Phonetic Variation
After introducing the origin of the word-initial [ŋ] in particular languages, the socio-variationist perspective is presented in the following part, followed by the case studies on Hong Kong Cantonese (a southern Chinese dialect) (Carol 2015), and Shijiazhuang dialect (a northern Chinese dialect) (Zhu 2007) respectively.
The socio-variationist study involving social factors of age and gender could be dated from Labov (1966Labov ( , 1972, which analyzed social stratification of -r in New York City (N.Y.C) department stores. The variants of the phonological variable -r were either presence or absence of post-vocalic [r], depending on the social class of speakers. Higher socioeconomic status produced more frequently than those in lower socioeconomic status. Labov"s (1966) analysis on post-vocalic [r] presented a variationist perspective for my study. But only the social factor-age, will be mainly focused in my study. Carol (2015) studied the phonological variation in Cantonese and argued that the zero-initial and the nasal onset [ŋ] were used in strictly different contexts in standard Cantonese in the 20th century. In Cantonese, there are 8 inventories of vowels: [i, y, ɛ, ɵ, u, ɔ, ɐ, a] (Matthews & Yip 1994). Before vowels [a, ɐ, o, ɔ], the nasal [ŋ] was chosen in standard Cantonese. Carol (2015) further mentioned age was relevant with this sound variation. Speakers preferred to speak [ŋ] in syllable-initial before 1950s; however, the zero-initial began to be more frequently chosen by speakers born after 1960 till now. The results indicated that the zero-initial increased popularity in the younger generation (Carol 2015: 343). However, Wong (2005)  To sum up, the velar-nasal onset [ŋ] in Cantonese was undergoing the process of sound variation to nasal-initial [n] (Wong 2005), or to zero-initial (Carol 2015). Zhu (2007) studied Shijiazhuang urban dialect (one of northern Chinese dialects) from a socio-variationist perspective. Zhu (2007) divided 44 subjects into two groups according ages: senior generation (> 60 years old) and young generation (</= 60 years old). The whole senior generation was observed speaking with the nasal [ŋ], but the young generation spoke with the zero-initial sounds. Zhu (2007) further argued syllable-initial [ŋ] was a contrastive phoneme in Shijiazhuang dialect, which lacked in syllabic onsets in Mandarin. Shijiazhuang urban area was geographically adjacent to Zhengding county. The current studies mainly pointed out this phenomenon without specific analysis on Zhengding dialect (i.e., the Annals of Zhengding County 1992; Song 1999). My study will explore whether there is also a generational difference in Zhengding dialect, as that in Cantonese and Shijiazhuang urban dialect. Before going to the analysis, a brief introduction for acoustic measurement of formants is presented below.

Acoustic Measurement for Nasalization
Acoustic theories of nasalization reveal that vowel nasalization primarily influences the first formant region F1 in the vowel spectrum. A set of major acoustic effects of nasalization is calculated in House and Stevens (1956), involving the reduction in amplitude of the first formant, the concomitant broadening of the bandwidth, the upward shifting of the frequency of F1, and an overall reduction in the energy of the vowel. Wright (1986) further suggests that nasalization may be expected to cause the raising of low vowels and some mid vowels, or lowering of high vowels and other mid vowels in perceptual findings. This is known as the centralization of nasal vowels in Beddor (1983), and also known as the bi-directional shift of vowel height in perceptual findings in Krakow et al. (1987). That is, the perceptional nasalized vowel quality is largely relevant to the first nasal formant (N1), and the first vowel formant (F1). For instance, for the nasalized/nasal vowel [a], the frequency of N1 is less than the frequency of F1, so the energy is distributed lower than usual. A raising effect is perceived. For the nasalized/nasal vowel [e] and [o], the frequency of N1 exceeds the frequency of F1, then a lowering effect is perceived in quality. The underlying trigger is due to the low-frequency region of the first nasal formant (N1), and the low-frequency of the zero formant (i.e., anti-formants in nasals), which reduced the prominence of F1 in vowels (Krakow et al. 1987). In my study, I will test whether Zhengding dialect shows the same acoustic tendency between the senior and younger speakers, in terms of the frequency of the vowel formants.
In the following section, the theoretical background about "fusion" in phonological change is provided, which will be applied to explain the phonological variation in Zhengding dialect.

Fusion and Phonetic Variation
Changes in pronunciation are known as phonological changes, or sound changes in a traditional term. Fusion, as one of the triggers for sound changes, occurs when a feature is added to a segment, or removed from a segment, or spread from one segment to another. The redistribution of the features is found in most types of phonological changes (Trask 1996: 52). Trask (1996: 62) illustrated that this phenomenon widely existed in French historically. The nasal consonant disappeared when vowels preceded the syllable-coda [n] and [m]. This was the origin of pain "bread" ([pɛ ]), langue "tongue" ([lɑ :g]), and bon "good" ([bɔ ]), etc. The nasal vowels an > ã can be also observed in the word blanc > blã in French (Crowley & Bowern 2010: 34). This process is known as fusion, in which two segments (an oral vowel and a nasal consonant) combine into single segment. Since the segments are made up of a number of features, as seen in (5a), fusion can be understood as a process that combines different features of the two original sounds into a new set of features upon a single sound.  (Trask 1996: 62). Another fusional analysis in Indonesian about nasal substitution is provided in (5b) (Halle & Clements 1983: 125;cited in Key 2008: 5). The [+nasal] feature in the prefix-final consonant and the [-voiced] feature in the root-initial obstruent are combined upon a single segment.
(6)a (Key 2008) The first derivation (a) stands for normal fusional operation in coalescence. Two units S 1 and S 2 are combined into a single unit S 1,2 , with features of S 1 and S 2 remained. The second derivation (b) in coalescence is resulted from the segmental spreading (known as "assimilation"), followed by a segmental deletion. It may be realized through completely independent processes. The third derivation of coalescence (c) is known as stability phenomenon. Segmental deletion occurs, but the feature does not lose itself. Instead, the feature attaches to a neighboring segment. The fourth derivation (d) in coalescence firstly undergoes the classic fusion as seen in (a), and then followed by the deletion of features associated with either of the input segments. To sum up, the first derivation of coalescence in (a) is the classic fusion, which combines different features of the two original sounds into a new set of features upon a single segment.
Two phonological characteristics of fusion are mentioned in previous literatures: word-hood, and sequential permissibility (Mackay 1987: 66-67). Cutting (1973) argued fusional operation had "a longer and linguistically more complex perception than either of the two input speech stimuli". For instance, when the pair Banket/Lanket was mentioned, the listening subjects usually suggested hearing Blanket instead. Here the "word-hood" is one of the phonological factors that largely influence the occurrence of fusion. The result of the fusion is usually words, no matter whether the stimuli are words or non-words (MacKay 1987: 66). Another International Journal of Linguistics ISSN 1948-5425 2021 www.macrothink.org/ijl feature of phonological fusion is "sequential permissibility". The fusion is allowed only when the fused results are licensed in listeners" language (MacKay 1987: 67). For instance, the sequence bdad and dbad is not permitted as a result of fusion of bad and dad. Regarding the fusion between a nasal and a vowel, the original oral vowel changes into a nasal vowel or a nasalized vowel. The nasal vowel and oral vowel are produced in different manner, as seen in (6b). Nasal vowel is produced "from a lowering of the velum during the vowel, which opens the velopharyngeal port and allows air to flow out through the nose and nostrils" (Styler, 2008:3), as illustrated in French historically. Nasalized vowel has no phonemic distinction from oral vowel, and the vowels, adjacent to nasal consonants, are produced partially or fully assimilated, with a lowered velum, as shown in English.
(6)b The production of an oral vowel and a nasal vowel (Styler 2008: 3) In this study, I will examine the loss of initial [ŋ] in Zhengding dialect, involving a process of fusion, in which the syllabic onset [ŋ] and the following vowel [ɤ] are fused into a single nasalized vowel in the young speakers in Zhengding dialect.

Methodology
This section describes the methods on how to collect data, to segment and to measure the frequency of the formants in Zhengding dialect.

Data Collection
The recording task is proceeded on 8 native speakers in Zhengding dialect. It includes four senior (>/=45 years old) and four younger (<45 years old), with four females and four males in total, as seen in (7). The participants were informed of the research goal and the process of the reading. The recording consists of 60 words traditionally in [ŋ]-V context, and 60 paired items in [g]-V context, including both monosyllabic and multisyllabic words. The Mandarin pronunciation are also recorded by the younger speakers. All of the participants are allowed to review and practice the words before the recording is proceeded. Data are recorded using a recording software "byly.exe", developed by Beijing Language and Culture University. The recording software is downloaded in an Acer laptop, at 16-bit sampling size, 44,100 Hz sampling rate, with a high quality noise cancelling head-mounted microphone. The recordings are saved in (uncompressed) WAV format. The environment for recording the initial [ŋ] and initial [g] is hold constant, by keeping the distance between the participants and the microphone unchanged. A file was then allocated with their ages and genders for each International Journal of Linguistics ISSN 1948-5425 2021 www.macrothink.org/ijl 45 participant. A code number (N=1, 2…n) is further linked with each participant, to protect their private information.

Data Segment and Analysis
The pronunciations are measured by using the acoustic analysis software Praat (Boersma and Weenink 2012). Since the first three formants are widely acknowledged to differentiate sounds, especially the first and second formants, representing the high-low and front-back distinctions of vowels (cited in Chen et al. 2013: 2), so in this study, the first two formants are plotted in this study. In monophthongs [ʌ, a, ɑ, ɤ, ə], the frequencies at the mid-point, are plotted in [ŋ]-V, [g]-V and zero-initial structures respectively, assuming consonantal effects to be minimal at the mid-point. But a different measurement point is determined for the diphthongs [ai, ɑo, ou]: the frequencies at the assuming average transitional space are plotted, permitting comparisons of dynamic vowel features between two single vowels. Additionally, values of the velar pinch (i.e., F3-F2) are also plotted, to compare the velar effect between the nasal [ŋ] and the plosive [g], both of which are velar segments. Multiple paired t-tests are also conducted, to see whether the formant changes in vowels are significantly distinct between different generations.

Experimental Analyses of Initial-[ŋ] in Zhengding Dialect
Three acoustic experiments are conducted on Zhengding dialect in this section. Experiment I compares the dialect speech between two syllabic structures in senior speakers: vowels traditionally preceded by initial velar nasal [ŋ] (i.e., [ŋ]+V), and vowels preceded by initial velar plosive [g] (i.e., [g]+V). The goal is to explore the acoustic properties in the traditional [ŋ]+V context. Experiment II is conducted in the young speakers, to see whether the young speakers present the same speech pattern with the seniors, i.e., whether the initial velar nasal [ŋ] is retained in the younger generations. Experiment III compares the younger Zhengding dialect and the Mandarin pronunciation in the younger speakers. The goal is to see whether the younger generation in Zhengding dialect is influenced by Mandarin, the dominant official language spoken in China.

Experiment I: Acoustic Effects in Senior Patterns
This experiment compares the acoustic characteristics of vowels produced by 4 senior speakers, traditionally with the initial velar nasal [ŋ] and with the initial velar plosive [g]. The frequencies of F1, F2 and the pinch values (F3-F2) in eight vowel realizations [ʌ, a, ɑ, ɤ, ə, ai, ɑo, ou] are plotted in (8), which corresponds to the pinyin labels "a1, a2, a3, e1, e2, ai, ao, ou" respectively. For F1, it seems the frequencies in [ŋ]+V context are lower in four senior speakers, with no deviation from this tendency. This is true for the eight vowel realizations, and supported by multiple paired t-tests across the four speakers. The specific details are presented in (9)  International Journal of Linguistics ISSN 1948-5425 2021 www.macrothink.org/ijl , the tongue dorsum raises during the articulation, and the velum also lowers, and these two factors contribute to the velar effect, as seen in (12a) (Baker et al. 2008). That is, the velar nasal [ŋ] is "more velar" than the velar plosive [g], resulted from the double articulation factors.

(9) Frequencies of F1 in [ŋ]+V and [g]+V in senior speakers
(12)a Articulation differences between the velar plosive and the velar nasal (Baker et al. 2008: 61) The similar explanation is found in the different realizations of short-a, before oral voiced stops, before [m] and [n], and before the velar nasal in Western American English (Baker et al. 2008). The spectrogram in (12b) shows the velar pinch of short-a before [ŋ] is more enlarged, when compared the formant trajectories of short-a before [g]. In the right figure, we can see the velar International Journal of Linguistics ISSN 1948-5425 2021 www.macrothink.org/ijl 52 pinch starts almost from the midpoint of the vowel in the short-a right before [ŋ], which indicates a stronger velar effect acoustically.
(12)b Velar pinches of vowel preceding velar plosive and velar nasal (Baker et al. 2008: 61) Here I adopt the idea and propose that the initial [ŋ] is more velar than initial [g] in Zhengding dialect. So it is not surprising there is a smaller pinch values in [ŋ]-V structure, which indicates a stronger velar effect in the minimal pairs. Summary of Experiment I: based on the F1, F2 and the pinch values between contexts, three acoustic characteristics are calculated in the traditional [ŋ]-V context in senior speakers: a lower F1, a higher F2, and a small F3-F2 value (indicating a strong velar effect).
For my purpose, in the following experiment, I will test whether the younger generation shows the same speech pattern reported above for the seniors.

Experiment II: Nasal Deletion in Young Patterns
This section focuses on the phonetic characteristics of younger speakers, examining whether the onset [ŋ] is also present in their speeches. 60 items in ([ŋ]-)V and [g]-V structures are produced by four young speakers (2 males and 2 females). The mean frequencies of F1, F2, and velar pinch (F3-F2) in [ʌ, ɑ, a, ɤ, ə, ai, ɑo, ou] are plotted in (13), with the pinyin labels "a1, a2, a3, e1, e2, ai, ao, ou". International Journal of Linguistics ISSN 1948-5425 2021 www.macrothink.org/ijl 54 younger generation has different patterns from the senior group, then different frequencies in F1, F2, and the velar pinch are supposed to be observed. For F1, it seems the frequencies in the traditional [ŋ]-V context are not lowered, as the senior pattern. Instead, a significant (p<0.05) higher F1 is found in [ʌ, ɑ, a, ə, ɑi, ɑo, ou] (labelled a1, a2, a3, e2, ai, ao, ou) in the traditional [ŋ]-V context. A deviation is found in [ɤ] (labeled e1, in the bold fonts): without significant difference between the two syllabic structures across four younger speakers (p>0.05), as seen in (14) ISSN 1948-5425 2021 www.macrothink.org/ijl a1 Turning to deviational [ɤ] (labelled e1), the pinch values do not have any significant difference between the traditional [ŋ]-V context and the [g]-V context. This may suggest that the velar (nasal) effect is still retained upon this vowel in the younger generation, since both contexts have a velar segment, or have features upon a velar segment.
When further observing the spectrogram trajectories in [ɤ] in the younger group, we cannot obviously visualize the low-frequency of the nasal preceding the darker vowel bands on the spectrogram, as seen in left figure in (17). However, in the senior group, a comparatively obvious low-frequency transition is found preceding the vowel bands, as seen in the right spectrogram in (17).
(17). Vowel [ɤ] in the traditional NV words in the younger (left) and senior (right) speakers In this study, I tentatively hypothesize a fusion process occurs between the initial nasal [ŋ] and the subsequent vowel [ɤ], contributing to the nasalized vowel [ɤ ] in the younger Zhengding dialect. This hypothesis is mainly based on two empirical observations on [ɤ]: (a) no low-frequency of the initial nasal is obviously visualized on the spectrograms; (b) three established nasal effects are still tested on this vowel: a lower F1, higher F2, and smaller velar pinch values, when compared with the [g]-V context. That is, the nasal effects are attached on the subsequent vowel [ɤ], and then the initial nasal [ŋ] has been deleted in the younger Zhengding dialect. The fusion in this study covers two phonological processes: nasalization of the vowel, followed by deletion of the initial nasal. The fusion on [ɤ ] does not occur systematically in other vowels (or vowel realizations) in Zhengding dialect, and the number of the distinctive sounds in this dialect is not changed, so this sound change is a non-phonemic (allophonic) change.
However, it is sometimes difficult to attest whether it is a nasalized vowel resulted from the fusion (such as [ɤ ] in the younger Zhengding speakers), or an oral vowel in nasal coupling (such as [ŋ]+[ɤ] in the senior speakers in Zhengding dialect). Ohala (1975) discussed the acoustic similarity between nasal vowels and velar nasals. Ohala argued that nasal vowels and velar nasals share more acoustic similarity than they are with labial or coronal nasals. Although nasals are produced with closure of the oral cavity, and the sound is radiated through the nasal cavity, the coronal and labial nasals are different from the velar nasal in manner of articulation. For instance, [m] and [n] are produced as a side branch in the vocal tract, which can form a low-frequency zeros or anti-formants. However, the velar nasal [ŋ] does not have obvious acoustic anti-formants, which makes [ŋ] more similar to vowels in the spectrum than nasals [m] and [n] (Ohala 1975;Mackenzie, et. al. 2007: 535). Alternatively speaking, if the nasal vowels are misperceived as a nasal, the nasal is highly supposed to be a velar nasal (Mackenzie, et. al. 2007: 535), rather than coronal or labial nasals. ISSN 1948-5425 2021 www.macrothink.org/ijl 59 Following Ohala (1975), I hypothesize the nasalized vowel (such as [ɤ ] in the younger Zhengding speakers), are also not that easy to distinguish from the oral vowel in nasal coupling (such as [ɤ] in [ŋ]+[ɤ] structure in the senior Zhengding dialect), since the low-frequency anti-formants usually cannot be visualized on the spectrograms.

International Journal of Linguistics
Summary of Experiment II: in this experiment, different speech patterns on F1, F2, and velar pinch are examined in the younger generation. The multiple paired t-tests indicate the initial velar nasal [ŋ] is not characterized in [ʌ, ɑ, a, ə, ai, ɑo, ou] (labelled a1, a2, a3, e2, ai, ao, ou) in the younger Zhengding dialect. The deviation is found in [ɤ] (labelled e1), in which a fusion (Note 5) process is proposed, with the nasal effects retained upon the vowels.
Considering the differences between the senior and the younger speakers in experiment I and experiment II, we can see the younger generation does not have the similar pattern with the senior speakers. However, a deviation is found in [ɤ] (labelled e1): the nasal effects are examined in both generations.
A question may be further asked: what is the socio-variationist factors, related to the generational differences in Zhengding dialect? Is this relevant to the dialect contact with Mandarin Chinese, the dominant official language, spoken in China. The following experiment will mainly focus on this issue.

Experiment III: Dialect Contact in Young Patterns
In this section, 60 items with vowel realizations [ʌ, a, ɑ, ɤ, ə, ai, ɑo, ou] are respectively produced by the young speakers in Zhengding dialect and Mandarin. The goal is to investigate whether the younger Zhengding dialect has the same speech pattern in Mandarin, and finally contributes the generational variation in Zhengding dialect. The frequencies of F1, F2 and velar pinch (F3-F2) are compared in (18), with the pinyin labels "a1, a2, a3, e1, e2, ai, ao, ou" between two contexts. International Journal of Linguistics ISSN 1948-5425 2021 II). The multiple paired t-tests do not indicate significant differences between these two contexts, with p>0.05, as seen in (19) Zhengding generational variation is expected to be part of the experimental sociolinguistics, allowing the sociolinguists and laboratory phonologists to study correlation between the social contexts and the language perception.

Conclusion
In this paper, three acoustic experiments have been conducted, showing that the younger generation in Zhengding dialect has been influenced by Mandarin, and they have largely lost the traditional initial-[ŋ] nowadays. Experiment I has defined and calculated three phonetic characteristics in the traditional [ŋ]-V context: a lower F1, a higher F2, and a smaller pinch value (F3-F2). Experiment II suggests that different patterns are produced by the younger generation, who have largely deleted the initial [ŋ]. A deviation is observed in [ɤ] (labelled e1), and a hypothesis on the fused vowel [ɤ ] (labelled e1) is proposed. The assumption is based on two major observations: (a) no obvious low-frequency of initial nasal is visually perceived in the spectrogram trajectories; (b) three established nasal effects are attested on this vowel. Experiment III further explores the correlation between the younger Zhengding dialect and Mandarin. Multiple paired t-tests suggest that the younger Zhengding dialect is correlated to Mandarin Chinese, both having the deleted-initials (i.e., zero-initial).
From the socio-variationist perspective, the pronunciation with the initial nasal [ŋ] is a partial variation in Zhengding dialect. Not all natives speak with this sound. The older generation largely remains this traditional variant, but the younger generation preferred the deleted-initial, which might be correlated with the dialect contact, especially when the younger generation starts to learn Mandarin.
For the limitation of this study, the data base for this acoustic analysis is not large enough.
Only 8 native speakers are recruited for the participation. The recording is conducted in one-time reading, although participants are allowed to pre-read and practice in advance. All of the limitations might cause the analyses not very precise to some extent. These limitations need to be improved for the future investigation.