Transcript Noise, memory and learning

Bo i Ro, Stockholm Oct 20th, 2015
Staffan Hygge
Noise, memory and learning
(Buller, minne och inlärning)
Staffan Hygge
Environmental Psychology
Department of Building, Energy and Environmental Engineering
Faculty of Engineering and Sustainable Development
University of Gävle
Contents of my talk
Text based reading and learning in noise
• Classroom experiments on noise and memory of a text
• A natural experiment on chronic aircraft noise and
cognition in children (the Munich study)
• Laboratory experiments on noise and memory systems
• WHO – Burden of Disease from environmental noise
Speech based listening and learning in noise
or degraded speech
• Speech intelligibility, recall and memory – Signal-toNoise Ratio (SNR) and Reverberation time (RT)
Road-Air 66
Air-Road 66
Train-Air 66
Road 55
Aircraft 55
Verbal 66
Train 66
Road 66
Road-Air 66
Air-Road 66
Train-Air 66
Air-Train 66
6
Air-Train 66
Road 55
Aircraft 55
Verbal 66
Train 66
Road 66
7
Aircraft 66
8
Aircraft 66
Mean score recognition
Mean score recall
2.5
Classroom Experiments - Results
2.0
1.5
1.0
Quiet
Noise
0.5
0.0
5
4
Quiet
Noise
3
2
1
0
The Munich airport noise study on
children's cognition
Nürnberg
A 92
Pulling
Hallbergmoos
Neufahrn Erdinger
Moos
A9
S8
Dorfen
Isen
Pastetten
B 12
A 94
München
Laim
Passau
Trudering
0
5 km
N
The Munich airport noise study on children
Children's cognition, some of the results
Difficult word list
Difficult paragraphs
Old airport
10
Long-term recall
Old airport
7
Old airport
8
9
No aircraft noise
No aircraft noise
5
Mean errors
Aircraft noise
6
5
4
3
Aircraft noise
4
3
Aircraft noise
2
1
0
Wave 1
Wave 2
Wave 2
Wave 3
Wave 1
New airport
7
Wave 2
Wave 3
Old airport closed
Old airport closed
New airport
10
3
0
Wave 1
Old airport closed
4
1
0
Wave 3
5
2
2
1
No aircraft noise
6
Mean score
7
Mean errors
7
6
8
New airport
8
9
6
7
6
5
4
3
Mean score
5
Mean errors
Mean errors
7
6
8
4
3
2
1
0
Wave 2
New airport opened
Wave 3
3
1
1
Wave 1
4
2
2
0
5
0
Wave 1
Wave 2
New airport o pened
Wave 3
Wave 1
Wave 2
New airport opened
Wave 3
Laboratory experiments on noise and
memory systems – Memory of the text
Recall of the text
Recognition of the text
10.0
12.0
10.0
8.0
8.0
Means
Means
6.0
4.0
6.0
Noise groups
Noise groups
4.0
No noise
No noise
2.0
2.0
Road traffic noise
0.0
Road traffic noise
Irrelevant
0.0speech
13-14 years
Agegroup
18-20 years
35-45 years
55-65 years
13-14 years
Agegroup
Irrelevant speech
18-20 years
35-45 years
55-65 years
Laboratory experiments on noise and
memory systems – Aggregated measures
8.0
Means
7.0
Aggregated m easure
M e m txt
6.0
M e m se n t
Wo rd fl
Wo rd co m p
5.0
Fa ce m e m
Quiet
Noise groups
Road traf f ic nois e
Irrelev ant s peec h
Structural Equation Model
Conceptual diagram of a
modified Model M4 for a single
sample with the addition of a
separate 1st-order factor (Text)
for the text memory items
(RCLtxt, RCGtxt) loaded on the
2nd-order factor Episodic, with
the corresponding indicators,
standardized loadings, and free
error covariances (2 = 90.17,
df= 74, p = .097, RMSEA =
.028).
Text
RCLtxt
RCGtxt
FRwE
.90
Episodic
.86
FRwoE
Recall
.67
.44
.76
.61
.48
.51
CRCwE
CRCwoE
CRNwE
.56
CRNwoE
Face rcg
.86
Recogn
Noise conditions did not
significantly change the
fit of the Structural
Equation Model!
.71
.61
.69
.50
.59
GN IncL
FN IntL
Knowl
.83
W comp
.97
Semantic
WF A
.88
Fluency
.79
.58
.55
WF M
WF prof
WHO, 2011
Chapter:
Environmental
noise and cognitive
impairment in
children
http://www.euro.who.int/__data/assets/pdf_file/0008/136466/e94888.pdf
Three field studies combined
Percent impairment
100
80
60
40
3 Rd
20
1 Rcl
2 Rcl
5 Rd
4 Rd
6 Rcl
0
40
45
50
55
60
65
70
75
80
85
90
95 100
Ldn
Figure. Dose-effect curves from epidemiological field studies with chronic noise exposure
Rd = Reading, Rcl = memory, recall
1 Recall, children, old airport, Hygge et al. (2002) - longitudinal
2 Recall, children, new airport, Hygge et al. (2002) - longitudinal
3 Reading, children, old airport, Hygge et al. (2002) - longitudinal
4 Reading, children, new airport, Hygge et al. (2002) - longitudinal
5 Reading, children, airport, Stansfeld et al. (2005) – cross-sectional
6 Free recall, children, road traffic, Lercher et al. (2003) – crosssectional
EUR-A countries can be calculated (Table 3.3).
The absolute DALY for the EUR-A countries,
with an estimated total population of 420 503
million, is therefore 45 036.
Speech based listening and learning
in noise or degraded speech
Exploring how spoken information, in contrast to
written information, is processed in working memory
(WM) and mediated to long-term memory
How much does SNR and RT influence recall and
memory?
Are the effects of SNR and RT additive or do they
interact?
Knowing more about their additive or interactive
effects has both a theoretical and practical value.
Speech intelligibility and memory –
Reverberation time and background noise
• Direct sound: The part of the sound that goes the shortest
distance from speaker to listener
• Reflected sound: The part of the sound that reaches the
listenser after having been reflected by different surfaces
in the room. Reaches the ear later than the direct sound
Sound pressure level, dB(A)
Indoor sound pressure level as a
function of the distance to the
sound source
-6dB with doubling distance
Constant sound level
Near field
Reverberation field
Distance from source
An example:
When a teacher
speaks at 66 dB(A),
that level drops to
 52 dB(A) 6 m out
in an ordinary
classroom. This is
only 7 dB(A) above
a background level
of 45 dB(A)
A long reverberation time may
disturb speech perception
• Reflected sounds that arrive later than 35-40 ms
after the direct sound interfere with the direct
sound and disturb speech perception by masking
the direct sound
• The largest and most critical masking effects are
vowels that mask subsequent consonants, which
have a lower sound level higher frequency, but
are more important for speech intelligibility
• At the same SNR-level a shorter reverberation
time therefore gives a clearer signal and better
speech perception
Does reverberation time cause
a problem in schools?
Measurements and recordings in two
classrooms – A short lecture
• A classroom with a long reverberation time
1.6 - 2.0 s in the lower frequency bands
• Another classroom with a reverberation time
around 0.3 s in all frequency bands
Results: Means 2.16 (long RT) and 4.00 (short RT), F(1,17) =
16.60, p < .001
To hear but not remember
Reverberation time (RT) and signal-to-noise ratio (SNR)
Participants heard word lists and repeated the words directly
Number of correctly recalled words
SNR 27 dB
Good
SNR 4 dB
Bad
Short RT
(0.5 s)
Long RT
(1.2 s)
11.0
8.5
13.0
10.8
p <.001
p =.025
Conclusion: Fewer word are remembered after
hearing them under them under a low signal-to-noise
(SNR) ratio and a long reverberation time
The role of Working Memory (WM)
• WM is employed for temporary storage of
information, to elaborate the information, to
link it together with information in long-term
memory and to transfer the information to
long-term memory
• WM has a limited capacity
• The more of that capacity that is taken up by
identifying the words that are heard, the less
is left for elaboration of the information, for
stroring, and for memory and learning
Recent studies
Recall of words spoken in the first
and second language:
Effects of signal-to-noise ratios and
reverberation times for school children
in grade 4 and College students
Materials and Procedure
Free recall of words in wordlists
Children Grade 4, N = 72, run as a group in their classrooms
College students, N = 48, run individually in the lab with
headphones and computer
Signal-to-Noise ratio (SNR) +3 and +12 dB
Reverberation time (RT)
1.2 and 0.3 s
Language (Lang)
English and Swedish
12 wordlists in each language
For children grade 4 - 8 words in each list
For college students - 12 words in each list
Words with high ranks were taken from language specific
category norms for 24 categories, which were sorted by GraecoLatin squares into 24 lists with equal average ranks of the words
Materials and Procedure
The presentation orders of the lists were counterbalanced
across experimental conditions and subjects
Examples of the words
Swedish RT = 1.2, SNR = +3 dB
English RT = 1.2, SNR =+12 dB
Swedish RT = 0.3, SNR = +3 dB
English RT = 0.3, SNR =+12 dB
Results
Source
Lang
Lang*Study
Total By Study
Total
X
X
Separate follow-ups
Grade 4
College
X
X
---------
SNR
SNR*Study
X
X
X
X
---------
X
X
---------
Lang*SNR
Lang*SNR*Study
RT*SNR
RT*SNR*Study
X
(p = 0.063)
X
---------
Lang*RT*SNR*Study
(p = 0.083)
---------
Lang*Study
For Grade 4 there is a more marked improvement with Swedish
compared to English than for the College students
SNR*Study
There is a larger gain in recall from +3 dB to +12 dB for Grade 4
than for the College students
Lang*SNR*Study
Looking at the slopes, there are more marked increases in recall
with improved SNR for Grade 4 than for the College students,
and the most marked increase for Grade 4 is with the Swedish
words
Lang*SNR*Study
Looking at the slopes, there are more marked increases in recall
with improved SNR for Grade 4 than for the College students,
and the most marked increase for Grade 4 is with the Swedish
words
Increases
Value: 0.131
Ratio: 1.482
Value: 0.110
Ratio: 1.932
Increases
Value: 0.070
Ratio: 1.172
Value: 0.093
Ratio: 1.278
Relative improvement of recall with
a +5 dB increase of SNR
Recall - Improvement
per SNR 5 dB, %
Grade 4
English
Swedish
50.4
26.0
College
English
Swedish
15.4
9.5
Conclusions
• SNR is a stronger determinant of recall than RT is
• A SNR improvement to +12 dB from +3 dB is more
important for Grade 4 than for College students, in
particular for the Swedish words
• There are no strong indicators of an interaction
between SNR and RT, and a long RT at SNR +3 dB
may improve, not impair, recall