Figure 1: Results of the laboratory studies of Marshall , Elliott  and previous in-situ tests .
Figure 1 illustrates some examples of the divergence of results from previous studies. Marshall used the WIPI test on children aged 5, 7, 9 and 11 . The speech and noise signals were presented monaurally using headphones. Elliott used a different type of speech test but again used monaural presentation for a wide range of subject ages [1,2]. Bradley carried out tests using the Rhyme test with complete classes of students in actual classrooms but only produced results for one age group (12-13 year olds) . The results are quite divergent and often differ from the results obtained by students listening binaurally in actual classroom situations.
We need to know how children's ability to recognize speech as a function of S/N, and under completely realistic conditions, varies with age in order to better establish ambient noise criteria for their classrooms. To meet this need, the present work carried out speech recognition tests in actual classrooms for grades 1, 3 and 6 students in schools near Ottawa, Canada.
The WIPI test was used because it is easy to explain to listeners of a wide range of ages [3,9]. It consists of simple test words said to be familiar to 5 year olds and these were presented in the carrier phrase, "Please mark the _____ now." The students responded by placing a sticker on one of 6 pictures to indicate the correct word. The students sat at their desks in their regular classroom. The tests were carried out in 41 classrooms evenly distributed among grade 1, grade 3, and grade 6 students (6, 8, and 11 year olds). A total of 840 students were evaluated in 41 classrooms. Grade 1 students were tested at 2 different S/N values and the other students at 3 different S/N values to give a total of about 2200 individual speech recognition tests.
The sound source was a small loudspeaker with similar directionality to that of a human talker. Digital recordings of the WIPI test material, made in an anechoic room, were edited to use exactly the same version of the carrier phrase for all test words and to have the same sound levels for all test words. Varied S/N were obtained by changing the playback level of the speech material relative to the existing ambient noise.
Speech and noise levels were recorded during the tests at 4 positions in each classroom. There were about 5 students near each microphone. These recordings were used to determine speech and noise levels during the tests by statistical analysis of the distribution of recorded sound levels . Room acoustics parameters were also measured from impulse responses obtained at the same locations . These included decay times, energy ratios, Useful/Detrimental ratios and STI values.
The same WIPI test was used to evaluate conditions intended to simulate those in classrooms, but with young adult listeners. The simulations were achieved with an 8- channel electro-acoustic system in an anechoic room. The simulated sound fields consisted of a direct sound, and early reflections followed by a reverberant tail representative of those found in the real classrooms. These speech sounds were combined with a 48dBA ambient noise having a spectrum shape representative of ventilation noise. The 8 test conditions included the combinations of 4 S/N values and two different room acoustics conditions. The two room acoustics conditions corresponded to: (a) the average condition measured in the real classrooms with a 0.5 s reverberation time and (b) the other representing a more reverberant classroom with a 1.0 second reverberation time.
This part of the work also used the Rhyme test and Difficulty ratings . It was intended, that these results for young adults, would provide baseline data for the effects of listener age and also allow us to compare with other previous results.
3.1. Classroom Tests
Figure 5: Comparison of subjective Difficulty ratings with the speech intelligibility test scores.
Although the young adults got very high scores for S/N> +1 dB, they still thought they had difficulty in perceiving the speech sounds as indicated by the Difficulty ratings in Figure 5. These are the results of a third test that the young adult subjects performed in the simulated sound fields in which they gave subjective ratings of the Difficulty of understanding speech material . In the range of S/N values from +1 to +20 dB, where the younger students show increasing intelligibility scores, the adults expressed decreasing difficulty. Thus, the adults have developed the skills to understand speech in more difficult conditions but they still require extra effort to do this and they find it more difficult to listen in these conditions where there are less than ideal S/N values. It is not clear how children would rate the Difficulty of the same situations because only young adults were tested.
4. Room Acoustics Effects
It was hoped that the selection of classrooms would include significant variations in room acoustics characteristics. Unfortunately this was not the case and mid-frequency reverberation times varied only between 0.3 and 0.7 s for the occupied classrooms . As a result, initial analyses of possible relationships between speech intelligibility scores and room acoustics parameters were inconclusive. For example, when the speech intelligibility scores were plotted versus A weighted Useful-to-Detrimental sound ratios, the resulting relationships were no better than those in Figure 2. Because S/N was deliberately manipulated over about a 40 dB range it had by far the dominant influence of speech recognition scores.
The results of the speech intelligibility tests in classrooms with grade 1, 3, and 6 students show clear effects of the age of the students. Grade 1 students are seen to require, on average, conditions with 7 dB better S/N than grade 6 students to achieve the same 95% correct speech intelligibility scores.
Young adults had, on average, substantially better speech intelligibility scores than the students for conditions with the same S/N value. While the adults obtained higher speech intelligibility scores, they still expressed difficulty in understanding the speech for the conditions of less than ideal S/N values. It may be that young students would have even higher levels of difficulty than the adults.
The young adults were tested in more ideal laboratory conditions without the additional distractions of real classrooms. Further experiments are required to more completely connect the results for the young adults with those for the students.
The measurements of S/N values in these classrooms during normal teaching activities had a mean S/N of 11 dB . For this average condition, the grade 1 students would understand only 92% of the teacher's speech. Of course, a significant number of the grade 1 students would understand much less than this average. The grade 3 students would perform only 2% better. Clearly many common classroom situations do not provide ideal acoustical conditions where younger students can understand all that is said by their teacher.
The authors are grateful for the financial support and collaboration of the Canadian Literacy and Language Research Network and for the help of Ms. Kimberlee Cuthbert in carrying out these experiments. They are also very appreciative of the help of the audio group at the Banff Centre for their help in editing the speech recordings.
 Elliott, L.L., "Effects of noise on speech by children and certain handicapped individuals," Sound and Vibration 16, 10-14, (1982).
 Elliott, L.L. "Performance of children aged 9 to 17 years on a test of speech intelligibility in noise using sentence material ~ with controlled word predictability", J. Acoust. Soc. Am., 66,651-653 (1979).
 Marshall, N.B., "The effects of different signal-tonoise ratios on the speech recognition scores of children", Ph.D. Thesis, University of Alabama, Tuscaloosa Alabama, (1987).
 Nabelek, A.K., and Pickett, J.M., "Reception of onsonants in a classroom as affected by monaural and binaural listening, noise, reverberation andhearing aids," J. Acoust. Soc. Am. 56, 628-639, (1974).
 Finitzo-Hieber, T, and Tillman, T.W., "Room acoustics effects on monosyllabic word discrimination ability for normal and hearing impaired children," J. Speech Hear. Res. 21, 440 - 458, (1978).
 Bradley, J.S., Sato, H. and Picard, M., "On the importance of early reflections for speech in rooms", J. Acoust. Soc. Am. 113 (6) 3233-3244 (2003).
 Carhart, R., Tillman, T.W., Greetis, E.S., "Perceptual masking in multiple sound backgrounds", J. Acoust. Soc. Am. 45, 694-703 (1969).
 Bradley J.S., "Speech intelligibility studies in classrooms", J. Acoust. Soc. Am., Vol. 80, No. 3, 846-854, (1986).
 Ross, M., and Lerman, J., "A picture identification test for hearing-impaired children", J. Speech and Hearing Research 13, 44-53 (1970).
Sato, H. and Bradley, J.S., "Evaluation of acoustical conditions for speech communication in active elementary school classrooms", Proceedings of ICA, Kyoto (2004).
 Sato, H., "Subjective measures to evaluate speech intelligibility, quality and difficulty in rooms for young and elderly listeners", Canadian Acoustics, 30 (3) 50-51 (2002).