Recognition of indigenous and foreign speech in a noisy classroom: A case of primary school pupils

Abstract

This study examines the level of speech recognition of English and Igbo utterances by 70 grade four children. The children, whose mother tongue is Igbo and aged between 8 and 10 years, had English monosyllabic words as well as Igbo monosyllabic and disyllabic words dictated to them in noisy and quiet classrooms. The results show that in noise, their level of recognition was significantly higher for English utterances (p < 0.05) while in quiet, it was significantly higher for Igbo (p < 0.05). Thus, the findings confirm those of previous studies that the recognition of Igbo is more affected by noise. In terms of the recognition of phonemes in quiet and noise, for both languages, the ‘t’ test analyses showed significant difference (p < 0.05) in the pupils’ recognition of vowels and consonants in both languages, with vowels being more identified than consonants in both noise and quiet. Hence, though the recognition of both languages is hindered in noise, the study of Igbo requires an acoustically serene environment for maximum results.

Keywords

Speech recognition noise word dictation indigenous language foreign language classroom

Introduction

The negative effects of external and internal noise on learning (specifically language learning) cannot be over emphasized. Although noise in primary schools is a global problem, it is a nagging issue in the Nigerian education system. Most school buildings are unprotected from noise. Some classrooms are demarcated with low walls; worse still, sometimes, halls that are not demarcated accommodate many classes such that pupils are exposed to both audio and visual distractions. The nuisance of noise, generally, is indeed a global issue and WHO,¹ in its document on community health (section 4.3.2) writes thus concerning schools and preschools:

For schools, the critical effect of noise is on speech interference, disturbance of information extraction (e.g. comprehension and reading acquisition), message communication and annoyance. To be able to hear and understand spoken messages in classrooms, the background sound pressure level should not exceed 35 dB LAeq during teaching sessions … (p. 43)¹

Despite the guidelines of WHO,¹ it has been shown that in Europe, classroom noise has remained a problem. A study conducted on 10 primary schools in Madrid shows that 9 out of the 10 schools had noise levels exceeding the permissible noise level of 40 dB(A) as stipulated by the Spanish government. A similar study on nine schools in Chile, South America, reveals that classrooms had noise levels between 50 and 60 dB(A). This is far above the 35 dB(A) recommended by WHO (Anon.).²

Noise has affected the standard of education in Nigeria; pupils have difficulty in perception, comprehension and retention in noisy environments. This has worsened the problem of promotion of indigenous Nigerian languages among native speakers, particularly children. Hence, the learning of these languages, in addition to other subjects, is greatly impeded by noise. There is indeed an urgent need to discover what can be done to facilitate the study of these languages in the midst of both internal and external noise as no immediate solution to noise has been provided in the country. This study, therefore, which addresses the impact of noise on school children’s speech recognition, is geared towards revealing and solving the problem of hindrance to speech recognition caused by noise. In this article, we reveal the difference between the effect of noise in the recognition of speech using two languages – English and Igbo. The study is important because it would help develop a better strategy for teaching the language that is more affected by noise.

The negative effects of noise on both man and the learning environment are obvious. Puglisi et al.³ show that higher background noise levels lead to higher voice levels, resulting in an increased stress for teachers since they have to operate in an environment with much background noise. using questionnaire in their surveys, Bhardwaj et al.⁴ investigated the noise levels in primary and secondary schools in Germany and found that the noise levels in primary schools were higher than those in secondary schools by as much as 2.3 dBA. Hence, the primary school environment has suffered from excessive noise levels.

Having a right environment for teaching and learning (particularly the teaching and learning of languages) is a factor that has been ignored to the detriment of the realization of the desired goal of indigenous language promotion, particularly in Nigeria. Dreossi and Momensohn-Santos⁵ reveal that the learning of the Brazilian Portuguese language has suffered a lot of interference due to ‘competitive noise in the classroom’. In this article, we will also show that the rate of speech intelligibility in noisy environments may be determined by type of language and this has significant implications for indigenous language acquisition.

Ohakamike-Obeka,⁶ in a research on the effect of learning environment on students’ performance in English, discovered that many schools in Nigeria have old and dilapidated buildings which are not conducive for learning. Hence, the negative attitude of students due to the state of the schools resulted in their poor performance in the English language. She therefore recommends that dilapidated and old school buildings across the country should be renovated and repainted to enhance performance in English language.

Ohakamike’s assertion is confirmed by Alexander Graham Bell Association for the deaf and hard of hearing.⁷ They show that buildings get noisier as they age because cracks begin to form and widen, hence external noise comes into the building. Also, house equipment loosen, generating internal noise. They therefore advise that in finding solution to noise in a building, different sources of the noise should be considered and that a qualified acoustician should be hired. The school, studied in the present research, has old buildings, with most of the walls very unattractive and not getting to the roof.

In Nigeria, most children start preschool when they are below 1 year, when they have not acquired the indigenous language. Therefore, in recent times, the language situation in most parts of the country is such that most children first acquire English (Nigeria’s official language and the schools’ medium of instruction) before their indigenous languages. Hence, the test administered on the respondents in this study is the type applied to second-language learners. Flowerdew and Miller⁸ outline a number of tests that can be used for second-language speakers. These include proficiency and placement tests, achievement test and diagnostic test. A combination of achievement and diagnostic tests is used in this study; the former is used to assess the level of knowledge of the respondents, while the latter is used to diagnose their rate of perception in an adverse condition like noise.

However, findings show that standardized tests in any language are biased, favouring native speakers of that language; yet non-native speakers are seen as lacking proficiency when they score low in such tests.⁹ According to Ascher, bilinguals tend to process information more slowly in their less familiar language thus their slower speed in test-taking. Also, she cites Figueroa¹⁰ and Valdes and Figueroa¹¹ as revealing that students with limited English familiarity may be more easily disturbed by noise and other distracting environmental conditions.

Kollmeier et al.¹² reveal from their investigation on intelligibility testing in multilingual situations that the characteristics of speakers generally dominate the differences across languages. Therefore, it is best to use a language test developed for a specific language in testing their speakers. Astolfi et al.¹³ conducted a rhyme test on Italian primary school pupils aged 7–10 using a diagnostic rhyme test (DRT) designed by Fondazione U. Bordoni of Rome. Results obtained from such a test will not be biased since the pupils were Italian speakers. Also, Puglisi et al.¹⁴ used Italian matrix sentence test to test speech intelligibility in Italian native speakers.

Ascher⁹ also mentions five tests that are commonly used for ‘limited’ English speakers. They include nonverbal tests, translated tests, interpreters, tests that are norm-referenced in the primary language as well as assessments by bilingual psychologists. Since none of these can adequately account for the knowledge of English and Igbo languages in the pupils used for this study, we opted to use the dictation method which has been used widely as a language testing method.¹⁵ Myinth shows that dictation is excellent for testing listening skills; hence in this study, which centres on speech perception and recognition, dictation is a good testing option. In this study, we aim at diagnosing how well the languages have been internalized and to what extent noise can hinder the pupils’ perception and recognition of those languages. It is not just centred on hearing in noise but hearing and reproducing language in noise.

In the Nigerian school environment, noise could be generated from many sources – external and internal. In this article, focus is on external noise (noise coming from outside the classroom). It is assumed that the teacher ought to have control over the acoustic environment of the class. A number of factors can give rise to noise from outside the classroom – noise from neighbouring classrooms as well as children’s activities near the classroom. Others include vehicular noise and noise from human voices. All these can suppress children’s perception. While the teacher may have control over noise generated within the classroom, it is difficult to control noise from outside, particularly since the classrooms are not fully walled and some that are fully walled have broken windows. Hence, the issue of ensuring perception and recognition among the pupils is an onerous one.

Cognitive abilities and acoustics

Litovsky and Godar¹⁶ reveal that though noise generally hinders perception in all humans, it poses greater problems to children because their auditory system is not fully developed. They show that in reverberant environments, children may have more difficulty than adults in perceiving signal cues arriving from different sources. The auditory system of an individual becomes fully developed as from 15 years; the hearing system is not fully developed at birth; it matures slowly after the inner ear starts to function.¹⁷ Auditory processing is the ability to perceive, decode and understand sounds. This can be hampered by noise especially in children since they do not have the ability to filter sounds in order to retrieve the necessary information among an array of disarranged sounds. In such situations, therefore, speech recognition is difficult and may be impossible. Klatte et al.¹⁸ reveal from their study, which centres on the effects of classroom noise and background speech on speech perception and listening comprehension, that children are more affected than adults by background sounds in both speech perception and listening comprehension. Hence, classroom noise constitutes a major hindrance to the perception of speech in children. In their study, Klatte et al.¹⁸ discovered that listening comprehension was hampered by background speech. However, though background speech did not affect speech perception as much as classroom noise, it hampered listening comprehension more.

Lewis et al.¹⁹ opine that according to Mody et al.²⁰ phonological awareness, an aspect of phonological processing which relates to an individual’s ability to recognize and manipulate the sound structure of speech (e.g. syllables and phonemes), may be related to children’s ability to understand speech in noise. In their work, Lewis et al.¹⁹ find that age affected the phonological awareness tasks of the children they used for the study (7-year olds performed significantly better than 5-year olds). In word recognition, 7-year olds scored significantly higher than 5-year olds. This goes a long way to confirm that noise hampers speech perception and recognition in young children.

Meyer et al.²¹ analysed intelligibility loss in spoken word lists in a natural background noise. In the work, they evaluated the ability of native French participants to recognize French monosyllabic words. Their results show that the identity of vowels is mostly preserved; that is, vowels could be recognized in noise. The results also confirm the functional role of consonants during lexical identification. Their analyses of recognition scores, confusion patterns and associated acoustic cues reveal that sonorant, sibilant and burst properties are the most important parameters influencing phoneme recognition. However, they observe that apart from environmental noise disturbances, another factor that can determine the rate of speech perception is type of language.

Uguru²² analysed the harmonics-to-noise ratio of English and Ika Igbo languages and discovered that lexical tone (which exists in Ika Igbo), when compared to intonation, reduced the degree of periodicity (voicing) and that led to the difference in the harmonic-to-noise ratio of English and Ika Igbo utterances used for the study. The fact that each syllable in a tone language has its own pitch introduced some noise which reduced the periodicity in Ika Igbo utterances. The result of the present research (as will be seen in subsequent sections) is a confirmation of the findings of Uguru. Also, Qin and Mok²³ cite Bent²⁴ and Huang²⁵ as showing that the perception of some Mandarin tone pairs, T2–T3 and T1–T4, were difficult for both native and non-native speakers because of their acoustical similarity. Hence, not all tones are perceived equally. Some tones, which are acoustically similar, may be more confused by speakers than others and this equally affects the perception of the words bearing them.

Tabri et al.²⁶ show from their research that bilinguals’ perception is more affected by noise than monolinguals’, while trilinguals perceive less than bilinguals. In their experiment, they discovered that even bilinguals, highly proficient in both languages, were still more affected by noise than monolinguals. They conclude that this discovery, which is a confirmation of previous findings, has educational implications because most children are bilingual and study in schools with classrooms that are not acoustically serene. These implications are also very important in the Nigerian environment (and particularly in this study) since most Nigerian children (including the ones used for this study) are also bilingual and their learning environments are normally not protected from noise. Tabri’s findings are corroborated by Schmidtke.²⁷ His study involved 48 Spanish–English bilinguals and 53 monolingual English young adults. They were tested on the recognition accuracy for words presented in noise; results showed that monolinguals recognized more words than bilinguals.

Golestani et al.²⁸ have revealed that bilinguals are better able to perceive speech-in-noise in their native language compared to their non-native language. However, this is not confirmed in this study as will be seen although the result of the study may also be attributed to the relegation of indigenous languages in Nigeria.

Warzybok et al.²⁹ reveal that non-natives show native-like speech recognition thresholds in noise only for the linguistically easy speech materials; however, the limitation of phonemic-acoustical cues in digit triplets affects speech recognition to the same extent in both non-natives and natives. For more complex and less familiar speech materials, non-natives, regardless of the degree of proficiency, require an average better signal-to-noise ratio to obtain 50% speech recognition compared to native listeners.

From the work of Ebem et al.,³⁰ Igbo subjects (adults) perceived American English more than their native Igbo. They conclude that it is likely that low-level signal parts of the Igbo tone language affect perception more than low-level signal parts of American English. Hence, in judging the quality of their own language, Igbo listeners need more signal level and more signal-to-noise ratio for perception than American subjects require in their own language.

In addition to the tonal features of Igbo, the working memory capacity (WMC) of listeners is important if speech can be identified in an adverse condition like noise. Working memory (WM) represents the ability to simultaneously store and process information.³¹ Rudner et al.³² show that the WM model for ease of language understanding (ELU) proposes that understanding language under challenging conditions is related to cognitive capacity. To develop this capacity, sufficient exposure to the language is necessary and this is lacking in the subjects used for this study. Most of them are not sufficiently exposed to the use of their indigenous language, Igbo. They speak English more frequently, responding in the same language when a question is asked in Igbo.

According to Rudner et al.,³³ the involvement of WM in speech perception varies. They show that in favourable circumstances, the processing of language input in the mental reservoir is automatic and implicit. However, in noise for instance, the input is weak or distorted. This causes a mismatch such that explicit processing is needed to match the input with the lexicon already existent in the listener’s memory. High WMC is important in such a situation. Rudner et al.³³ reveal that ‘efficient cognitive function and good cognitive capacity’ could be necessary for effective remedial explicit processing of language input. We argue that a good way of ensuring this ‘good’ cognitive capacity is by having a good knowledge of the language. The results of this work (to be seen later) therefore reveal that this is lacking in the subjects as far as the knowledge of Igbo is concerned.

Experiment

The experiment is centred on the impact of noise on the intelligibility of English and Igbo in a quiet and a noisy classroom. Especially, the degree of the negative impact of noise on the recognition of English and Igbo speech is evaluated. Also, phonemic (consonantal and vocalic) recognition is evaluated. In all, 70 grade four children, aged between 8 and 10 years, were used for the study. The children’s indigenous language is Igbo. They were drawn from three classes. Their school, situated inside a university campus, was chosen because it is rated as a fairly good school (in terms of the teaching standard; also, it is relatively exposed to less noise). The classrooms are oblong in shape, have wooden chairs, tables and chalkboards. The floors are cement floors, while the walls are halfway to the roof. The number of children per class was not taken; however, the 70 children, from three classes, sat in their normal sitting positions. The acoustic characteristics of the classroom are as follows:

Reverberation time 1.0 s;

C50 0 dB;

STI values } Middle position 0.60;

Back position 0.56.

A list of randomly selected words (20 monosyllabic English words and 10 Igbo words, half of which were monosyllabic and the other half disyllabic) were dictated to the children at an average noise level of about 70 dB(A) and also in a quiet environment. It was important to typify the actual learning environment which the pupils are usually exposed to; public address system is not used in their school hence using it for the experiment will yield a false result, thereby negating the aim of the research. The sound level, 65–70 dB(A), is usually the noise level range (at some periods during school hours) in the classrooms as the voices of teachers and pupils in adjoining rooms filter into other classrooms. This range of noise level is higher than the level of 35 dB(A) recommended by the WHO. It is also far above 40–50 dB(A) which some countries have adopted as acceptable.^34,35

The words that were used in the experiment were carefully selected to ensure that they are simple. Also, the words were read aloud to the pupils. Educational Testing Service³⁶ opines that testing modification – for instance, allowing a reading test to be read aloud – if it does not change the construct being measured (decoding of text) is permitted. Furthermore, since standardized tests in any language are usually framed in favour of its native speakers,⁹ we adopted the dictation method instead of existing tests like achievement and language proficiency tests which are more basically meant for English. According to Heaton,³⁷ dictation brings out the following skills: auditory discrimination, auditory memory span, spelling, recognition of sounds and so on. In our analysis, phonological mistakes were ignored since our focus was on perception hence any clue that showed a pupil identified an utterance correctly was taken as right recognition. The teacher dictated the words from the front of the class.

The words are as follows:

English Words

Toy /tɔɪ/

Boy /bɔɪ/

Noise /nɔɪz/

Girl /ɡɜl/

Cry /kɹaɪ/

Lip /lɪp/

Light /laɪt/

Like /laɪk/

Lice /laɪs/

Life /laɪf/

Bread /bɹɛd/

Deep /dip/

Gum /ɡʌm/

Cat /kӕt/

Road /ɹəʊd/

House /hɑʊs/

Feet /fit/

Pot /pɒt/

Clay /kleɪ/

Class /klɑs/

IGBO WORDS

Òké /oke/ ‘rat’

Jí /ʤi/ ‘yam’

Àlà /ala/ ‘ground’

Nyé /ɲe/ ‘give’

Ńrí /nri/ ‘food’

Gbá /ɡba/ ‘run’

Ókwú /okwu/ ‘speech’

Tí /ti/ ‘strike’

Ị̀gbà /ɪɡba/ ‘drum’

Gụ́ /ɡʊ/ ‘read’

Analyses, results and discussion

The number of utterances recognized by particular number of participants in the noisy and quiet environments was determined. These were also expressed in simple percentages. Furthermore, phonemic recognition was analysed through the evaluation of the pupils’ consonantal and vocalic identification in both noise and quiet. That is, assessing if the right vowel or consonant was perceived, particularly the first phoneme of the first syllable (in the case of disyllabic words) or those words with closed syllables. The analysis was centred on revealing which of the two (consonants or vowels) were most recognized. This was deduced from the scores. To reveal the significance in the difference between the pupils’ recognition of the languages and the phonemes, student ‘t’ test analysis was conducted.

The results of the study can be seen from the tables below which show the pupils’ perceptions and identification of phonemes in both noise and quiet for the two languages.

From Table 1, although the Igbo utterances are fewer in number than the English ones, it can be observed that 66 respondents (94%) perceived all the utterances while only 4 (6%) could not perceive one of the utterances. Conversely, 45 (64%) respondents perceived all the English utterances, while 25 (36%) failed to recognize one utterance. This therefore shows that the children are able to recognize their mother tongue better when in a quiet environment. A student ‘t’ test carried out shows that this difference in identification is highly significant, p < 0.05.

Table 1.

Correct perceptions for English and Igbo utterances in quiet environment.

English	Igbo
1. 20	10
2. 20	10
3. 20	10
4. 20	10
5. 20	10
6. 20	10
7. 20	10
8. 20	10
9. 20	10
10. 20	10
11. 20	10
12. 20	10
13. 20	10
14. 20	10
15. 20	10
16. 20	10
17. 20	10
18. 20	10
19. 20	10
20. 20	10
21. 20	10
22. 20	10
23. 20	10
24. 20	10
25. 20	10
26. 20	10
27. 20	10
28. 20	10
29. 20	10
30. 20	10
31. 20	10
32. 20	10
33. 20	10
34. 20	10
35. 20	10
36. 20	10
37. 20	10
38. 20	10
39. 20	10
40. 20	10
41. 20	10
42. 20	10
43. 20	10
44. 20	10
45. 20	10
46. 19	10
47. 19	10
48. 19	10
49. 19	10
50. 19	10
51. 19	10
52. 19	10
53. 19	10
54. 19	10
55. 19	10
56. 19	10
57. 19	10
58. 19	10
59. 19	10
60. 19	10
61. 19	10
62. 19	10
63. 19	10
64. 19	10
65. 19	10
66. 19	10
67. 19	09
68. 19	09
69. 19	09
70. 19	09

Test statistic t = −3.02 on 138 d.f.

Probability = 0.003.

A look at Table 2 shows that the pupils’ recognition of Igbo utterances is highly impeded in noise, while that of English fared better. Only three pupils (4%) recognized all the Igbo utterances. However, although none of the pupils recognized all the English utterances, 14 (20%) recognized 19 out of the 20 utterances. On the extreme, nine pupils (13%) could not recognize any of the Igbo utterances while the least number of English utterances recognized by a pupil was four. The ‘t’ test conducted shows that the difference between the recognition of English and Igbo in noise is highly significant (p < 0.05).

Table 2.

Correct perceptions for English and Igbo utterances in a noisy environment.

English	Igbo
1. 19	10
2. 19	10
3. 19	10
4. 19	09
5. 19	09
6. 19	09
7. 19	09
8. 19	09
9. 19	09
10. 19	09
11. 19	09
12. 19	09
13. 19	09
14. 19	09
15. 18	08
16. 18	08
17. 18	08
18. 18	08
19. 18	08
20. 18	08
21. 18	08
22. 18	08
23. 18	08
24. 18	07
25. 18	07
26. 18	07
27. 18	07
28. 18	07
29. 18	07
30. 18	07
31. 18	07
32. 17	07
33. 17	07
34. 17	07
35. 17	06
36. 17	06
37. 17	06
38. 17	06
39. 17	06
40. 17	06
41. 17	06
42. 17	06
43. 17	05
44. 16	05
45. 16	05
46. 16	05
47. 16	04
48. 16	04
49. 15	03
50. 15	03
51. 15	02
52. 15	02
53. 15	02
54. 15	02
55. 15	02
56. 14	02
57. 14	01
58. 14	01
59. 14	01
60. 14	01
61. 13	01
62. 13	0
63. 13	0
64. 13	0
65. 12	0
66. 11	0
67. 11	0
68. 11	0
69. 6	0
70. 4	0

Test statistic t = 6.38 on approximately 95.76 d.f.

Probability < 0.001.

Furthermore, it was necessary to analyse the recognition of the individual languages in both noisy and quiet environments. Table 3 shows that in quiet, English utterances were perceived better than in noise. In all, 45 pupils recognized all the English utterances in quiet but none of them could recognize all the utterances in a noisy condition. Also, the least number of utterances perceived in quiet was 19, while the least number of utterances perceived in noise was four. Student ‘t’ test shows that the disparity in the perception of English utterances in quiet and in noise is significant (p < 0.05).

Table 3.

English utterances correctly perceived by individual pupils in quiet and noisy environments.

Quiet	Noise
1. 20	19
2. 20	19
3. 20	19
4. 20	19
5. 20	19
6. 20	19
7. 20	19
8. 20	19
9. 20	19
10. 20	19
11. 20	19
12. 20	19
13. 20	19
14. 20	19
15. 20	18
16. 20	18
17. 20	18
18. 20	18
19. 20	18
20. 20	18
21. 20	18
22. 20	18
23. 20	18
24. 20	18
25. 20	18
26. 20	18
27. 20	18
28. 20	18
29. 20	18
30. 20	18
31. 20	18
32. 20	17
33. 20	17
34. 20	17
35. 20	17
36. 20	17
37. 20	17
38. 20	17
39. 20	17
40. 20	17
41. 20	17
42. 20	17
43. 20	17
44. 20	16
45. 20	16
46. 19	16
47. 19	16
48. 19	16
49. 19	15
50. 19	15
51. 19	15
52. 19	15
53. 19	15
54. 19	15
55. 19	15
56. 19	14
57. 19	14
58. 19	14
59. 19	14
60. 19	14
61. 19	13
62. 19	13
63. 19	13
64. 19	13
65. 19	12
66. 19	11
67. 19	11
68. 19	11
69. 19	6
70. 19	4

Quiet – 64%.

Noise – 36%.

Test statistic t = 9.48 on approximately 72.70 d.f.

Probability < 0.001.

From Table 4, it is seen that the recognition of Igbo utterances in quiet is high while that of noise is very much hindered. Whereas as many as 66 pupils (94%) recognized all the Igbo utterances in quiet, only three (4%) of them could recognize all the utterances in noise. Also, the least number of Igbo utterances perceived in quiet was nine. On the contrary, nine pupils (had zero recognition) could not recognize any of the utterances at all in noise.

Table 4.

Number of Igbo utterances correctly perceived by individual pupils in quiet and noisy environments.

Quiet	Noise
1. 10	10
2. 10	10
3. 10	10
4. 10	9
5. 10	9
6. 10	9
7. 10	9
8. 10	9
9. 10	9
10. 10	9
11. 10	9
12. 10	9
13. 10	9
14. 10	9
15. 10	8
16. 10	8
17. 10	8
18. 10	8
19. 10	8
20. 10	8
21. 10	8
22. 10	8
23. 10	8
24. 10	7
25. 10	7
26. 10	7
27. 10	7
28. 10	7
29. 10	7
30. 10	7
31. 10	7
32. 10	7
33. 10	7
34. 10	7
35. 10	6
36. 10	6
37. 10	6
38. 10	6
39. 10	6
40. 10	6
41. 10	6
42. 10	6
43. 10	5
44. 10	5
45. 10	5
46. 10	5
47. 10	4
48. 10	4
49. 10	3
50. 10	3
51. 10	2
52. 10	2
53. 10	2
54. 10	2
55. 10	2
56. 10	2
57. 10	1
58. 10	1
59. 10	1
60. 10	1
61. 10	1
62. 10	0
63. 10	0
64. 10	0
65. 10	0
66. 10	0
67. 9	0
68. 9	0
69. 9	0
70. 9	0

Test statistic t = 11.59 on approximately 69.70 d.f.

Probability < 0.001.

From the foregoing, it is observable that, in noise, the pupils’ level of recognition was higher for English utterances while they recognized Igbo more than English utterances in the quiet classroom. The analysis of the data (Table 1) reveals that none of the children could perceive all the English words dictated despite the fact that the words were simple (below their level). Hence, noise interfered with the pupils’ word recognition in both languages although that of Igbo was more. This confirms the findings of Tabri et al.²⁶ that noise affects speech perception in bilinguals to a great degree.

The details of the pupils’ perception of the English utterances in noise are as follows: 14 children were able to perceive 19 of the 20 utterances correctly. Conversely, the lowest perception was that of one pupil who perceived only 4 of the 20 words correctly. In between these two perception rates, we have 17 children perceiving 18 utterances, 12 perceived 17 utterances, 5 perceived 16 utterances, another 5 pupils recognized 14 utterances, 4 perceived 13 utterances, 1 perceived 12 utterances, 3 perceived 11 utterances while 1 pupil perceived 6 utterances.

However, 19% of the English utterances were not identified in noise while as much as 46% of the Igbo utterances were not identified in noise. One can see that the perception/recognition of Igbo is more hindered by noise than that of English. In the quiet classroom, the pupils’ recognition of the utterances of both languages improved tremendously; 45 pupils identified all the 20 English utterances, while 25 perceived 19. Hence, only 2% of the utterances were not identified. For the Igbo utterances made in the quiet classroom, 66 pupils perceived all the 10 utterances while four recognized nine utterances; only 1% of the utterances were unidentified. It can be seen that though noise hampered the recognition of Igbo more, it also made a negative impact in the recognition of English utterances.

In the case of the Igbo utterances, in noise, three pupils were able to recognize all the 10 words used for the study. However, the average correct identification rate for Igbo in noise was very much lower than that of English probably due to the fact that the pupils are more accustomed to speaking English. It could have been caused by some linguistic factors – the fundamental properties of the Igbo speech may have played a role in the lower identification rate. Uguru²² reveals that lexical tone in Igbo incorporates some level of noise that makes Igbo speech less perceptible or voiced than that of an intonation language. This noisy component of Igbo speech is confirmed in Ebem et al.³⁰ This factor may have been worsened by the noisy environment in which the words were dictated, hence the low recognition of Igbo utterances. This therefore establishes the findings of Meyer et al.²¹ that speech perception can be determined by type of language.

Analyses of the recognition of consonants versus vowels

The analyses of the recognition of consonants versus vowels in noisy and quiet classrooms for both languages appear in Tables 5 –8.

Table 5.

Recognition of English consonants versus vowels by respondents in noise.

Utterance	No. of respondents who correctly identified vowel/diphthong	No. of respondents who correctly identified consonant
Toy	63	42
Boy	70	70
Noise	70	70
Girl	70	70
Cry	42	56
Lip	42	56
Light	70	70
Like	63	49
Lies	70	49
Life	70	63
Bread	63	70
Deep	70	35
Gum	63	63
Cat	63	70
Road	70	56
House	70	70
Feet	56	56
Pot	56	56
Clay	70	70
Class	70	70
Total	65	58

Vowels – 91%.

Consonants – 86%.

t = 1.19.

Probability = 0.244.

Not significant (p > 0.05).

Table 6.

Recognition of Igbo consonants versus vowels by respondents in noise.

Utterance	No. of respondents who correctly identified vowel/diphthong	No. of respondents who correctly identified consonant
Òké	70	70
Jí	63	63
Àlà	63	63
Nyé	49	49
Ńrí	63	63
Gbá	42	42
Ókwú	63	56
Gụ́	56	35
Ị̀gbà	56	63
Tí	70	70

Vowels – 85%.

Consonants – 82%.

t = 0.43.

Probability = 0.672.

Not significant (p > 0.05).

Table 7.

Recognition of English consonants versus vowels by respondents in quiet.

Utterance	No. of respondents who correctly identified vowel/diphthong	No. of respondents who correctly identified consonant
Toy	70	70
Boy	70	70
Noise	69	70
Girl	69	68
Cry	70	70
Lip	69	68
Light	68	68
Like	70	70
Lice	69	69
Life	69	67
Bread	70	70
Deep	70	69
Gum	69	65
Cat	70	69
Road	70	70
House	70	70
Feet	70	69
Pot	70	69
Clay	70	67
Class	70	70

Vowels – 99%.

Consonants – 98%.

t = −2.09.

Probability = 0.046.

Significant (p < 0.05).

Table 8.

Recognition of Igbo consonants versus vowels by respondents in quiet.

Utterance	No. of respondents who correctly identified vowel/diphthong	No. of respondents who correctly identified consonant
Òké	70	70
Jí	70	70
Àlà	70	70
Nyé	70	70
Ńrí	70	70
Gbá	70	70
Ókwú	70	70
Gụ́	70	66
Ị̀gbà	70	70
Tí	70	70

Vowels – 100%.

Consonants – 99%.

t = 1.00.

Probability = 0.331.

Not significant (p > 0.05).

In terms of the recognition of consonants and vowels in noise and quiet, for both languages, student ‘t’ test analysis was carried out. There was significant difference (t = −2.09, p < 0.05) in the pupils’ recognition of consonants and vowels for English utterances made in the quiet classroom. On the contrary, the difference in the pupils’ recognition for consonants and vowels for English utterances made in a noisy classroom did not differ significantly (t = 1.19, p > 0.05). The difference in the recognition of consonants and vowels in Igbo utterances made in quiet and noisy environments followed a similar trend, that is, no significant difference.

However, it was discovered that most of the vowels were well recognized even in cases where the consonants were not well perceived. This is also in line with the findings of Meyer et al.²¹ who show that vocalic identities are maintained in noise and that they are not usually confused with another. In the case of consonants, some pupils replaced some of the consonants with wrong phonemes; the plosives /k, t/ in word medial or final positions were at times replaced with the nasal, /n/. Recognition of place of articulation was also a problem. For instance, the alveolar plosive, /d/ was replaced with the bilabial plosive /b/ by some subjects; hence, ‘deep’ was identified as ‘big’. It must be pointed out that in Nigerian English, the same vowel, /i/ is used for both words, ‘deep’ and ‘big’; hence, the recognition problem here centres on the consonants and not the vowel since we have mostly adopted the Nigerian English pronunciation in our analysis. Most of the pupils identified the diphthong, /ɔɪ/, while some replaced it with /ɒ/. The pupils also identified the vowel, /i/ correctly but some replaced /d/ with /g/.

Generally, it was observed that some of the pupils tended to replace word final plosives with nasals hence rendering ‘light’ and ‘like’ as ‘line’. Voiced plosives were also frequently replaced with other voiced plosives having different places of articulation as seen in pupils identifying ‘give’, ‘bic’ and ‘be’ instead of ‘deep’. It can be observed that the vowels are mostly identified correctly and this confirms the report of Meyer et al.²¹ More details of the pupils’ identification of consonants and vowels can be seen from Tables 5 –8 above.

Further analyses are based on the pupils’ recognition of consonants in quiet and noise, on one hand, and their recognition of vowels in quiet and noise on the other hand for both English and Igbo languages. The results are presented below in Tables 9 –12.

Table 9.

Recognition of English vowels in noise and quiet (values indicate the number of pupils that recognized the consonants/vowels).

	Noise	Quiet
Toy	63	70
Boy	70	70
Noise	70	69
Girl	70	69
Cry	42	70
Lip	42	69
Light	70	68
Like	63	70
Lies	70	69
Life	70	69
Bread	63	70
Deep	70	70
Gum	63	69
Cat	63	70
Road	70	70
House	70	70
Feet	56	70
Pot	56	70
Clay	70	70
Class	70	70

Noise – 91%.

Quiet – 99%.

t = 2.79.

Probability = 0.012.

Significant (p < 0.05).

Table 10.

Recognition of English consonants in noise and quiet.

	Noise	Quiet
Toy	42	70
Boy	70	70
Noise	70	70
Girl	70	68
Cry	56	70
Lip	56	68
Light	70	68
Like	49	70
Lies	49	69
Life	63	67
Bread	70	70
Deep	35	69
Gum	63	65
Cat	70	69
Road	56	70
House	70	69
Feet	56	69
Pot	56	69
Clay	70	67
Class	70	70

Noise – 86%.

Quiet – 98%.

t = −3.44.

Probability = 0.003.

Significant (p < 0.05).

Table 11.

Recognition of Igbo vowels in noise and quiet.

	Noise	Quiet
Òké	70	70
Jí	69	70
Àlà	63	70
Nyé	49	70
Ńrí	63	70
Gbá	42	70
Ókwú	63	70
Gụ́	56	70
Ị̀gbà	56	70
Tí	70	70

Noise – 85%.

Quiet – 100%.

t = 3.35.

Probability = 0.004.

Highly significant (p < 0.05).

Table 12.

Recognition of Igbo consonants in noise and quiet.

	Noise	Quiet
Òké	70	70
Jí	69	70
Àlà	63	70
Nyé	49	70
Ńrí	63	70
Gbá	42	70
Ókwú	63	70
Gụ́	56	66
Ị̀gbà	56	70
Tí	70	70

Noise – 85%.

Quiet – 99%.

t = 3.19.

Probability = 0.011.

Significant (p < 0.05).

From Tables 9 – 12 it can be seen that the perception and identification of consonants and vowels were significantly worse in noise.

Discussion

The analyses of the study are based on four platforms – first, the recognition of English and Igbo utterances in noise and quiet; second, the recognition of the individual languages in noise and quiet. Third, we analysed the recognition of the consonants versus vowels (for both languages) in noise and quiet and fourth, the pupils’ recognition of consonants in quiet and noise, on one hand, and their recognition of vowels in quiet and noise on the other hand (for both English and Igbo languages).

Under the first platform, it can be observed from our analysis on Table 1 that though both English and Igbo were perceived better in quiet, more Igbo utterances (94%) were recognized in quiet than those of English (64%). On the other hand, in noise, the recognition of Igbo was highly hampered, with only three (4%) of the pupils recognizing all the Igbo utterances (Table 2). The pupils perceived the English utterances more than the Igbo ones – 14 (20%) recognized 19 of the utterances. Nine pupils (13%) could not recognize any of the Igbo utterances. The ‘t’ tests conducted reveal that these differences are significant (p < 0.05). This implies that the study of Igbo language requires a very serene environment for learners to achieve maximum learning.

On the second platform, the recognition of the utterances of the individual languages in noise and quiet was analysed. For English in quiet and noise (Table 3), the pupils’ recognition rates of the utterances are 64% and 36%, respectively, and the ‘t’ test analysis conducted showed this difference to be significant (statistic t = 9.48 (p < 0.05)). In the same vein, the pupils’ recognition rates of Igbo utterances in quiet and noise (Table 4) are as follows: 66 pupils (94%) recognized all the utterances in quiet while only 3 (4%) recognized all the utterances in noise. This difference is also significant (test statistic t = 11.59 (p < 0.05)). Thus, neither of the languages should be studied in noisy environments if maximum learning is to be achieved. This has a great implication for other subjects which are studied in English throughout the country. There is a need to limit noise as much as possible in our schools.

On the third platform, the recognition of the consonants versus vowels (for both languages) was analysed. The recognition rates for English consonants versus vowels in noise (Table 5) are 86% and 91%, respectively. However, this difference is not significant as revealed by the ‘t’ test results (t = 1.19; probability = 0.244; not significant: p > 0.05)). Similarly, the recognition rates of Igbo consonants and vowels in noise (Table 6) are 82% and 85%, respectively. Their difference is also not significant as shown by the ‘t’ test result (t = 0.43; probability = 0.672; not significant (p > 0.05)). The pupils’ recognition rates of English consonants and vowels in quiet (Table 7) are 98% and 99%, respectively. The ‘t’ test conducted shows that their difference is significant (t = −2.09; probability = 0.046; (p < 0.05)). The recognition rates for Igbo consonants and vowels in quiet (Table 8) are 99% and 100%, respectively. The ‘t’ test result for the analysis is t = 1.00; probability = 0.331 (not significant: (p > 0.05)).

Fourth, the pupils’ recognition of consonants in quiet and noise, on one hand, and their recognition of vowels in quiet and noise on the other hand (for both English and Igbo languages) were analysed. From the ‘t’ test analyses, the pupils’ recognition of consonants on one hand and vowels on the other, in noise and quiet in both English and Igbo utterances, indicate significant differences as shown below.

The analysis of the pupils’ recognition of English vowels in noise and quiet (Table 9) shows the rates of 91% and 99%, respectively. The difference is significant (t = −2.79; probability = 0.012 (p < 0.05)). The analysis of the pupils’ recognition of English consonants in noise and quiet (Table 10) shows the following results: 86% and 98%, respectively; ‘t’ test result is significant (t = −3.44; probability = 0.003 (p < 0.05)).

Table 11 displays the pupils’ recognition of Igbo vowels in noise and quiet. 85% of the vowels were recognized in noise, while 100% were recognized in quiet. The ‘t’ test result (t = 3.35, probability = 0.004 (p < 0.05)) shows difference to be highly significant.

Igbo consonants in noise and quiet (Table 12) reveal the recognition rates of 85% and 99%, respectively. The difference is significant as revealed by the ‘t’ test result (t = 3.19, probability = 0.011 (p < 0.05)).

From the foregoing, we conclude that noise is unfavourable for the recognition (and consequently) and the study of both English and Igbo languages. Although a greater percentage of Igbo utterances were not recognized, the study of both languages is significantly hindered by noise. Hence, since other subjects are studied using language, we recommend that schools be equipped with language laboratories that have necessary acoustic installations which prevent or reduce both internal and external noise. This will facilitate language study, thereby enhancing the study of other subjects.

Solutions to classroom noise and indigenous language study

Since speech perception can be determined by type of language, it is important to determine how the perception of some languages (e.g. tone languages) can be enhanced. This fact is most pertinent in Nigeria where most schools have noisy environment, in addition to the noise in the tonal features of indigenous Nigerian languages. First, there is a need to enhance and modify the perception of pupils as discovered by Akahane-Yamada et al.³⁸ Their findings show that training in the perception domain produces long-term modifying of both perception and production. Thus, pupils need to be trained on the perception, particularly, of tone languages.

Also, students should be acquainted with oral materials that are beyond their abilities. This can help them perceive language better.³⁹ This is because such materials will enable them to be conversant with the sounds of the language such that they can recognize the sounds even in noise.

Another way of ensuring easy recognition of languages in noise is encouraging the pupils to speak the language constantly. This is a major problem in Nigeria because most children speak English rather than their native languages. Hence, they may not be conversant with the sounds of their native languages; this has led to their having low WM as revealed in this research. The trend has to be reversed to ensure the survival of indigenous Nigerian languages.

A simple solution to increasing speech recognition rate in noisy classrooms so as to enhance language learning especially for tone languages like Igbo, which has noisy components hindering perception in normal cases, is to prevent the noise from being generated. Reducing the background noise level in classrooms is a major factor towards ensuring speech recognition.³⁵ Acoustic stoplights can be used to reduce the speech production of children in classrooms. Di Blasio et al.⁴⁰ show that S&N-S Light, Speech & Noise Stop-Light, a device with a warning light activation determining sound levels limits, hence aiding personal voice control, has been successfully used in classrooms, restaurants, and urban squares (i.e. densely occupied spaces) to achieve reduction in noise. This is because the device reduces chatting by occupants.

However, noise from outside the classroom (which is the focus of this study) is difficult to control. Hence, Shield and Dockrell⁴¹ advise that the exterior shell of school buildings should be shaped to deflect and absorb traffic and other external noises. Barrett et al.⁴² show that considering the naturalness principle in constructing school buildings is important since it contributes to 50% progress in learning. This may be because the frequency of noise disturbance is reduced.

Also, one can reduce the effect of noise by applying insulation between the source of the noise and the classroom. If noise is generated in rooms adjacent to the classroom, one can apply sound absorbing materials (particularly in ceilings) in these rooms and/or improve the attenuation of the sound transmission between rooms. Furthermore, sound absorbing materials can be applied in the noisy classroom itself; this will decrease the annoyance of sound from outside, decrease the sound transmission between rooms and also decrease the impact of noise generated in the classroom. A further significant improvement in speech intelligibility can be achieved using sound diffusers which prevent the generation of sharp resonances that decrease intelligibility (Puglisi et al.,³ Choi⁴³ and Shtrepi et al.⁴⁴). Secchi et al.⁴⁵ reveal that in Italian school buildings, the average noise level could be significantly reduced by applying façade sound insulation. In their measurements, background noise levels dropped from a range between 30 and 40 dB(A) towards a range between 25 and 30 dB(A).

In terms of improving intelligibility, a simple method that a teacher can apply is to raise his or her voice in order to increase the signal-to-noise ratio at the listener’s side. This process is more or less automatically induced by the background noise and is known as the Lombard effect.⁴⁶ The only disadvantage, however, is that the teacher’s voice will be strained. To forestall this problem, audio materials should be used; these will not only ensure that the pupils hear the teacher equally well but will also go a long way to improve the perception of languages like Igbo which have noisy components. Bottalico et al.⁴⁷ assert that the Lombard effect has a ‘starting point’. That is, there is a Lombard effect change-point at a background noise level. This change-point is anticipated by noise disturbance, and is followed by a high magnitude of vocal discomfort. It results from the efforts put in to speak loudly in a noisy environment.

In order to be able to objectively quantify possible improvements in intelligibility, an objective speech intelligibility measure should be used. Two different approaches are available, the STI⁴⁸ (IEC) and the newly developed POLQA Intelligibility, an extension of POLQA (ITU⁴⁹) and a follow-up of PESQ intelligibility.⁵⁰

WM training will help pupils learn to identify speech in noise. Such a training will help them learn to listen and decipher information from various sources at the same time. This measure will suffice for the present since Nigerian schools are still faced with the problem of noise; the enactment as well as the implementation of noise policies is not in sight. Discovering measures of planning and executing WM training will make a good research topic and is hereby suggested for further research. Also, Crandell and Smaldino⁵¹ reveal that using sound field (SF) amplification systems, which use a wireless microphone to transmit the teacher’s speech signal to an amplifier-loudspeaker system will enhance speech recognition.

Conclusion

The most important conclusion that can be drawn from the experiments is that there is a significant difference between the impact of noise on the intelligibility of English and Igbo, with the recognition of the latter being more hindered in noise. In noise, the pupils’ level (percentage) of recognition was higher for English utterances while they recognized Igbo more than English utterances in the quiet environment. Our analysis shows that 81% of English utterances were correctly identified, while 54% of Igbo utterances were correctly identified in noise. The lower percentage in the correct identification of Igbo utterances could be due to the neglect of Igbo language by the children. Also, the tonal features of the Igbo language, which incorporates noise in its perception, may have played a role. This would be in line with the result found by Ebem et al.³⁰ in which noise had a bigger impact on the perceived speech quality of Igbo than English. Although noise impacts more negatively on the recognition of Igbo, there is need to also improve the environment acoustically for the study of both languages. This call springs from the fact that the pupils’ recognition for both languages improved tremendously in the quiet environment, with only 2% of English utterances being unidentified and only 1% of Igbo utterances not identified.

We also suggest that increasing the WMC of the children through constant use of their native language, Igbo, may improve the recognition of speech in adverse conditions like noise. Furthermore, differences in the recognition rates of pupils used in this study imply that children do not perceive or identify information alike; hence, teachers need to find a way (e.g. the SF system) of helping those pupils whose rate is lower than average (particularly in noise). Since most school environments in Nigeria are noisy, it then means that there is a need to device a way of teaching so as to ensure that the foreign and indigenous languages, as well as other subjects, are perceived by learners.

Footnotes

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship and/or publication of this article.

References

WHO. Guidelines for community noise (ed Berglund

Lindvall

Schwela

). Geneva: WHO, 1999.

Anon, http://www.hear-it.org/classrooms-far-too-noisy (accessed 17 September 2017).

Puglisi

Bolognesi

Shtrepi

. Optimal classroom acoustic design with sound absorption and diffusion for the enhancement of speech intelligibility. J Acoust Soc Am 2017; 141: 3456.

Bhardwaj

Baum

Markevych

. Are primary school students exposed to higher noise levels than secondary school students in Germany? Int J Occup Environ Med 2013; 4(1): 2–11.

Dreossi

RCF

Momensohn-Santos

. Noise and its interference over students in a classroom environment: literature review. Pró-Fono R Atual Cient 2005; 17(2): 251–258.

Ohakamike-Obeka

The school learning environment and students’ attitude and achievement in English language. Res Humanit Soc Sci 2016; 6(2): 31–37.

Alexander Graham Bell Association for the Deaf and Hard of Hearing. Reducing noise in learning spaces, https://www.agbell.org/professionals/hearing-and-hearing-technology/reducing-noise-in-learning-spaces/ (accessed 17 September 2017).

Flowerdew

Miller

Second language listening: theory and practice. Cambridge: Cambridge University Press, 2005.

Ascher

Assessing bilingual students for placement and instruction. ERIC/CUE Digest 1990; 65, https://www.ericdigests.org/pre-9217/placement.htm (accessed 17 September 2017).

10.

Figueroa

RA.

Psychological testing of linguistic-minority students: knowledge gaps and regulations. Except Children 1989; 56(2): 145–152.

11.

Valdes

Figueroa

The nature of bilingualism and the nature of testing: towards the development of a coherent research agenda. Prepared for the National Commission on Testing and Public Policy. Unpublished manuscript, University of California, Berkeley & University of California, Davis, 1989.

12.

Kollmeier

Warzybok

Hochmuth

. The multilingual matrix test: principles, applications, and comparison across languages: a review. Int J Audiol 2015; 54: 3–16.

13.

Astolfi

Bottalico

Barbato

Subjective and objective speech intelligibility investigations in primary school classrooms. J Acoust Soc Am 2012; 131(1): 247–257.

14.

Puglisi

Warzybok

Hochmuth

. An Italian matrix sentence test for the evaluation of speech intelligibility in noise. Int J Audiol 2015; 54: 44–50.

15.

Myinth

MK.

Dictatory: a new method of giving dictation. English Teach Forum 1998; 36(1).

16.

Litovsky

Godar

SP.

Difference in precedence effect between children and adults signifies development of sound localization abilities in complex listening tasks. JASA 2010; 128(4): 1979–1991.

17.

Kappel

Moreno

Buss

CH.

Plasticity of the auditory system: theoretical considerations. Braz J Ortohonolaryngol 2011; 77: 4, https://dx-doi-org.web.bisu.edu.cn/10.1590/S1808-86942011000500022 (accessed 17 September 2017).

18.

Klatte

Lachmann

Meis

Effects of noise and reverberation on speech perception and listening comprehension of children and adults in a classroom-like setting. Noise Health 2010; 12(49): 270–282.

19.

Lewis

Hoover

Choi

. The relationship between speech perception in noise and phonological awareness skills for children with normal hearing. Ear Hear 2010; 31(6): 761–768.

20.

Mody

Studdert-Kennedy

Brady

Speech perception deficits in poor readers: auditory processing or phonological coding?

J Exp Child Psychol 1997; 6(4): 199–231.

21.

Meyer

Dente

Meunier

Speech recognition in natural background noise. PLoS ONE 2014; 8(11): e79279.

22.

Uguru

JO.

Harmonics-to-noise ratio in Ika Igbo and English utterances: implications and inferences. In: Proceedings of the 19th international congress on acoustics, Madrid, Spain, September 2 – 7, 2007, pp. 1–5.

23.

Qin

Mok

PPK

. Discrimination of Cantonese tones by speakers of tone and non-tone languages 2014. http://hdl.handle.net/1808/12864 (accessed 17 September 2017).

24.

Bent

The perception and production of non-native prosodic categories. Unpublished PhD Thesis, Department of Linguistics, Northwestern University, Evanston, IL, 2005.

25.

Huang

Language-specificity in auditory perception of Chinese tones. Unpublished PhD Thesis, Ohio State University, Columbus, OH, 2004.

26.

Tabri

Chacra

KMSA

Pring

Speech perception in noise by monolingual, bilingual and trilingual listeners. Int J Lang Commun Disord 2011; 46: 411–422.

27.

Schmidtke

The bilingual disadvantage in speech understanding in noise is likely a frequency effect related to reduced language exposure. Front Psychol 2016; 7: 678.

28.

Golestani

Stuart

Scott

SK.

Native-language benefit for understanding speech-in-noise: the contribution of semantics. Biling Lang Cogn 2009; 12(3): 385–392.

29.

Warzybok

Brand

Wagener

. How much does language proficiency by non-native listeners influence speech audiometric tests in noise? Int J Audiol 2015; 54: 88–99.

30.

Ebem

Beerends

Vugt

. The impact of tone language and non-native language listening on measuring speech quality. JAES 2011; 59(9): 647–655.

31.

Füllgrabe

Rosen

. Investigating the role of working memory in speech-in-noise identification for listeners with normal hearing. In: van Dijk

Baskdent

Gaudrain

. (eds) Physiology, psychoacoustics and cognition in normal hearing (Advances in Experimental Medicine and Biology vol. 894). New York: Springer, 2016, pp. 29–36.

32.

Rudner

Foo

Rönnberg

. Cognition and aided speech recognition in noise: specific role for cognitive factors following nine-week experience with adjusted compression settings in hearing aids. Scand J Psychol 2009; 50(5): 405–418.

33.

Rudner

Rönnberg

Lunner

Working memory supports listening in noise for persons with hearing impairment. J Am Acad Audiol 2011; 22: 1–12.

34.

Guidini

Bertoncello

Zanchetta

. Correlations between classroom environmental noise and teachers’ voice. Rev Soc Bras Fonoaudiol 2012; 17(4): 398–404.

35.

Chan

KMK

EPM

. Noise levels in an urban Asian school environment. Noise Health 2015; 17(74): 48–55.

36.

Educational Testing Service. Guidelines for the assessment of English language learners. Educational Testing Service, 2009, https://www.ets.org/s/about/pdf/ell_guidelines.pdf (accessed 17 September 2017).

37.

Heaton

JB.

Writing English language tests. London: Longman, 1988.

38.

Akahane-Yamada

Tohkura

Bradlow

. Does training in speech perception modify speech production? In: Fourth international conference on spoken language (ICSLP ’96), Philadelphia, PA, 3–6 October 1996, vol. 2, pp. 606–609. New York: IEEE.

39.

Price

Rypa

Speech technology and language learning: some examples from VILTS the Voice Interactive Language Training System. Proceedings of the AATOLL conference, Honolulu, HI, 1998, No other one located.

40.

Di Blasio

Calosso

Giuseppina

. S&N-S Light: an anthropic noise control device to reduce the noise level in densely occupied spaces encouraging personal control of voice. J Acoust Soc Am 2017; 141: 3663.

41.

Shield

Dockrell

JE.

External and internal noise surveys of London primary schools. J Acoust Soc Am 2004; 115(2): 730–738.

42.

Barrett

Davies

Zhang

. The impact of classroom design on pupils’ learning: final results of a holistic, multi-level analysis. Build Environ 2015; 89: 118–133.

43.

Choi

YJ.

Effects of periodic type diffusers on classroom acoustics. Appl Acoust 2013; 74: 694–707.

44.

Shtrepi

Astolfi

Puglisi

. Effects of the distance from a diffusive surface on the objective and perceptual evaluation of the sound field in a small simulated variable-acoustics hall. Appl Sci 2017; 7(3): 1–23.

45.

Secchi

Astolfi

Calosso

. Effect of outdoor noise and façade sound insulation on indoor acoustic environment of Italian schools. Appl Acoust 2017; 126: 120–130.

46.

Lane

Tranel

The Lombard sign and the role of hearing in speech. J Speech 1971; 4: 677–709.

47.

Bottalico

Passione

Graetzer

. Evaluation of the starting point of the Lombard effect. Acta Acust United Ac 2017; 103(1): 169–172.

48.

IEC 60268-16:2003. Sound system equipment, part 16: objective rating of speech intelligibility by speech transmission index.

49.

ITU-T Rec. P.863: perceptual objective listening quality assessment. Geneva: International Telecommunication Union, 2011.

50.

Beerends

van Buuren

van Vugt

. Objective speech intelligibility measurement on the basis of natural speech in combination with perceptual modelling. JAES 2009; 57: 299–308.

51.

Crandell

Smaldino

JJ.

Classroom acoustics for children with normal hearing and with hearing impairment. Lang Speech Hear Serv Sch 2000; 31: 362–370, http://www.hear-it.org/classrooms-far-too-noisy (accessed 17 September 2017).