Memory for Melodies and Lyrics in Alzheimer's Disease

Abstract

THIS RESEARCH ADDRESSED THE QUESTION: IS MUSICAL memory preserved in dementia, specifically, dementia of the Alzheimer type (AD)? Six tests involving different aspects of melody and language processing were administered to each of five groups of participants: 50 younger adults, 100 older adults, and 50 AD older adults classified into three levels of AD sever- ity—mild, moderate and severe. No test was immune to, but not all tests were equally sensitive to, the presence of dementia. Long-term familiarity for melody was preserved across levels of AD, even at the severe stage for a few individuals. Detecting pitch distortions in melodies was possible for mild and some of the moderate AD participants. The ability to sing a melody when prompted by its lyrics was retained at the mild stage and was retained by a few individuals through the severe stages of AD. Long-term familiarity with the lyrics of familiar melodies was also found across levels of AD. In contrast, detection of grammatical distortions in the lyrics of familiar melodies and the ability to complete familiar proverbs were affected even at the mild stage of AD. We conclude that musical semantic memory may be spared through the mild and moderate stages of AD and may be preserved even in some individuals at the severe stage.

Full text

Musical Memory 479
and for implementing music therapy interventions. Yet the
status of musical memory in dementia is far from clear
(for an extensive review, see Baird & Samson, 2009).
The type of musical memory that we have studied is
the representation of musical knowledge—information
acquired through prior experience outside the labora-
tory and independent of the temporal or spatial context
in which it was acquired. Following the terminology of
memory theory (Tulving, 1983, 1985), an appropriate
term is musical semantic memory (e.g., Platel et al., 1997;
Platel, Baron, Desgranges, Bernard, & Eustache, 2003;
Omar, Hailstone, & Warren, 2012).
Musical semantic memory should be distinguished
from musical episodic memory, which is memory for
specific musical events and the context in which they
were heard. Semantic memory and episodic memory
for music are dissociable in AD, first noted by Bartlett,
Halpern, & Dowling (1995, Experiment 1; see also
Samson, Baird, Moussard, & Clément, 2012; Vanstone
et al., 2012). In their task intended to reflect episodic
memory (Bartlett et al., 1995), AD participants first
heard a study list of melodies followed by a test list
containing the studied melodies randomly mixed with
unstudied melodies. Participants were asked to recog-
nize the studied melodies in the test list. Performance
was very poor. However, by contrast, performance was
nearly perfect on a semantic memory task in which
participants had to decide whether a traditional melody
was previously known in everyday life. Platel et al.
(2003) provided further evidence of dissociation with
healthy undergraduates. They used high-resolution
PET to observe patterns of brain activation while the
participants performed a semantic and an episodic
musical task. Different patterns of activation were
found for the different tasks, a result suggesting distinct
neural substrates underlying the semantic and episodic
components of music.
The current study follows our earlier explorations
(Cuddy & Duffin, 2005; Vanstone & Cuddy, 2010) evaluat-
ing persons with AD. In the first study (Cuddy & Duffin,
2005), we reported preservation of musical memories in
the case of an 84-year-old woman with severe dementia,
EN. To assess long-term familiarity with music, a series of
melodies was presented – melodies from the traditional
LoLa L. Cuddy, JaCaLyn M. duffin, Sudeep S. GiLL
Queen’s University, Kingston, Ontario, Canada
CaSSandra L. Brown
University of Victoria, Victoria, British Columbia,
Canada
ritu Sikka, & aShLey d. VanStone
Queen’s University, Kingston, Ontario, Canada
this research addressed the question: is musical
memory preserved in dementia, specifically, dementia of the
Alzheimer type (AD)? Six tests involving different aspects of
melody and language processing were administered to each of
five groups of participants: 50 younger adults, 100 older adults,
and 50 AD older adults classified into three levels of AD sever-
ity—mild, moderate and severe. No test was immune to, but
not all tests were equally sensitive to, the presence of dementia.
Long-term familiarity for melody was preserved across levels
of AD, even at the severe stage for a few individuals. Detecting
pitch distortions in melodies was possible for mild and some
of the moderate AD participants. The ability to sing a melody
when prompted by its lyrics was retained at the mild stage and
was retained by a few individuals through the severe stages of
AD. Long-term familiarity with the lyrics of familiar melodies
was also found across levels of AD. In contrast, detection of
grammatical distortions in the lyrics of familiar melodies and
the ability to complete familiar proverbs were affected even at
the mild stage of AD. We conclude that musical semantic
memory may be spared through the mild and moderate stages
of AD and may be preserved even in some individuals at the
severe stage.
Received August 8, 2011, accepted September 28, 2011.
Key words: music, melody, lyrics, memory, Alzheimer’s
disease
T he goal of our research is to address the
question: Is musical memory preserved in demen-
tia, specifically, dementia of the Alzheimer type
(AD)? This topic has far reaching implications for
understanding normal and pathological brain organization
Music Perce ption VoLuM e 2 9, iSSue 5, p p. 479–491. iSSn 0730-7829, eLeCtroniC iSSn 1533-8312. © 2012 B y the reGentS of the uniVerSity o f CaLifornia aLL
riG htS reSerVed. pLeaSe direCt aLL requeS tS for perMiSSion to photo Copy or reproduCe art iCLe Content throuGh the uniVerSity of CaLifornia preSS’S
riG htS and perMiSSionS weBSite, h ttp ://www.uCpreSSJournaLS.CoM/reprintinfo.aSp.doi: 10.1525/Mp.2012.29.5.479
MeMory for MeLodieS and LyriCS in aLzheiMer’S diSeaSe

480 Lola L. Cuddy, Jacalyn M. Duffin, Sudeep S. Gill, Cassandra L. Brown, Ritu Sikka, & Ashley D. Vanstone
repertoire (folk melodies, carols, patriotic songs, etc.) ran-
domly mixed with novel unfamiliar melodies. The task was
to discriminate “familiar” from “unfamiliar” melodies. In
an additional test, melodies from the traditional repertoire
were presented, but some of the melodies were played
incorrectly. If incorrect, a sequence of notes within the
melody was raised or lowered in pitch so that the initial
tonality (i.e., key structure) was disrupted. The task was to
indicate whether the melody was correctly played or not.
Because of the severity of her cognitive impairment, EN
was unable to understand formal test instructions. However,
we observed that EN responded to familiar melodies by
singing along and that she often continued singing the
melody after the music stopped. She showed no response
to unfamiliar melodies, usually staring blankly out the
window. She responded to pitch distortions in melodies
with grimaces, shudders, frowns, and sometimes puzzled
laughter. These responses never accompanied melodies
played correctly. Thus, using behavioral gestures as indica-
tors, we found that EN responded to familiar musical
melodies with the same high level of accuracy as normal
age-matched controls and also detected pitch distortions at
the high accuracy level of controls. Her responses were very
easy to score; inter-rater reliability between independent
scorers revealed 100% agreement.
It might be queried whether EN’s case was typical of
Alzheimer pathology (Baird & Samson, 2009). Possibly
EN’s musical sparing was due to (partial) preservation of
medial temporal lobe structures that are typically affected
in AD and implicated in memory loss. However, a later
autopsy report renders this possibility unlikely.
(Neuroimaging was not undertaken because of the likely
confusion and distress she would have experienced.)
According to the report, released to family members
September 13, 2005, examination revealed “a well known
pattern of relatively advanced Alzheimer disease (Braak
Stage V)… [including] significant neuronal loss in medial
temporal lobe structure.” No atypical findings were
reported.
Our second study (Vanstone & Cuddy, 2010) assessed
12 persons diagnosed with moderate or severe AD, and 12
healthy age-matched controls. Again, behavioral
observation was used if formal instructions could not be
communicated to the participant. As in the first study, tests
included differentiation between familiar and unfamiliar
melodies, and detection of a pitch distortion in a traditional
melody. Additional tests involved singing a melody when
prompted by spoken lyrics and detection of a pitch
distortion in a novel tonal melody. The two pitch distortion
tests differ in the strategies available to the listener. When
the melody is traditional and previously known, a
distortion may be detected either because the perceptual
representation of the presented melody does not match the
representation in the lexicon and/or the perceptual repre-
sentation indicates the violation of tonal rules. When the
melody is novel, only the latter strategy can be available.
We found inter-rater reliability to be high (93-100%
agreement across different measures). As a group, AD
participants scored significantly lower than the controls.
However, five AD participants performed impressively,
near or above the control mean on most of the tests,
including both the two pitch distortion tests. Four other
participants showed above chance performance, and only
three were unable or unwilling to complete the test
sessions. There was no consistent relationship between test
scores, levels of dementia, and demographic characteris-
tics of the participants. All told, our findings suggest that,
relative to cognitive decline in other domains, musical
semantic memory may be spared in AD for some
individuals. More recently, Hsieh, Hornberger, Piguet, and
Hodges (2011) found similar performance, on average,
for an AD group and a control group on recognition of
famous tunes and famous tune titles. The range of scores,
however, was somewhat greater for the AD group.
The concept of musical semantic memory is further
illustrated in a music-recognition model proposed by
Peretz and Coltheart (2003). This model, arising from
neuropsychological data, describes a modular system of
music processing components. Parallel and largely
independent processing subsystems handle either music
or speech. Within the music system, the model includes
the notion of a musical lexicon (see Figure 1). The musical
lexicon contains “a perceptual representation system for
isolated tunes, much in the same way as the mental word
lexicon represents isolated words” (Peretz et al., 2009, p.
257). The musical lexicon also contains the rules of musi-
cal syntax that help integrate tunes and phrases into a
high-level representation of a musical context. The tests
that we have administered—assessing the sense of
familiarity for traditional music, and assessing the
detection of error in tonally distorted (i.e., nonsyntactic)
music—are ways of tapping the musical lexicon.
The model has several implications for the present study.
First, it provides a functional architecture for the study of
musical semantic memory. In the case of brain damage, the
musical lexicon, or the flow of information to it, may be
selectively impaired or spared (Peretz, 2001, 2006; Peretz &
Coltheart, 2003). Here we pursue the notion that, in the case
of AD, it may be selectively spared. Moreover, since the
lexicon may be accessed through several routes, the
possibility exists that if one route is impaired, access may
be achieved through other spared routes.
Second, the model assumes that tonal knowledge is
implicitly, or automatically, acquired through exposure

Musical Memory 481
to tonal music. Krumhansl and Cuddy (2010) provide a
recent review of evidence supporting this assumption
and further conclude (p. 62), that “musical memory and
the representation of the tonal hierarchy are intimately
associated.” If so, musical representation in the lexicon
may not necessarily require formal music training or
intervention. Of course, training will lead to an enhanced
lexicon, but the point is that, given appropriate musical
materials, sparing of the lexicon may be found even if
music training is minimal.
This point is important because several reported in-
stances of musical sparing were performances by skilled
professional musicians (e.g., Beatty et al., 1994; Cowles
et al., 2003; Fornazzari et al., 2006). If these instances
are singular and prototypical cases of sparing, then the
involvement of motor skill memory—called procedural
memory—may be paramount (Baird & Samson, 2009).
However, the AD participants for whom we have
rep ort ed spa rin g of musical memor y were no t
professional musicians, and some had very little music
training (Vanstone & Cuddy, 2010). For the present, we
retain the hypothesis that for these participants it is
musical semantic memory that has been spared.
Third, the model invites queries about the relation
between memory for melody and memory for lyrics. The
model assumes that memory for lyrics is represented in
the phonological lexicon (see Figure 1), a component of
the speech system. If the phonological lexicon, or access
to it, is impaired, memory for lyrics may be impaired.
However, the model proposes close associative links
between the musical lexicon and the phonological lexicon
(see also Hébert & Peretz, 2001; Moussard, Bigand,
Belleville, & Peretz, 2012; Peretz, Radeau, & Arguin, 2004).
Thus, even if memory for spoken lyrics is impaired, there
still may be sufficient activation in the phonological
lexicon to co-activate the musical lexicon. Recognition of
the lyrics will occur through association. (Note: Steinke,
Cuddy, & Jakobson (2000) proposed the reverse situation
to account for the case of a man with amusia who
nevertheless could recognize familiar tunes if they had
FIGUR E 1. A modular mode l of music proce ssing. All co mponent s whose domains appear to be specific to music are in green; others are in blue.
Adapted with permissio n from I. Peretz. Permission to repri nt from Macmi llan Publishers Ltd: Peretz, I., & Coltheart, M. (2003). Modular ity of music
processi ng. Nature Neuroscien ce, 6, 688–691 .

482 Lola L. Cuddy, Jacalyn M. Duffin, Sudeep S. Gill, Cassandra L. Brown, Ritu Sikka, & Ashley D. Vanstone
previously been associated with lyrics. The authors
suggested that there may have been sufficient melodic
information to activate associated lyrics in a spared
phonological lexicon and thus recognition occurred.) On
the other hand, the detection of a grammatical distortion
in a lyric would not benefit from a close association of the
lyric to its melody. Detection of a grammatical distortion
would require active analysis in the speech system and as
such might be sensitive to the effects of degenerative
disease on the speech processing system.
The modular account in Figure 1, which is based on
neuropsychological anomalies, differs from accounts
based on neuroimaging and other data (e.g., Besson,
Chobert, & Marie, 2011; Koelsch, 2011; Patel, 2008;
Patston & Tippett, 2011; Strait & Kraus, 2011) that support
overlapping systems for music and language. However,
even if neural resources for activating these networks are
shared, linguistic and musical syntax may be stored in
distinct brain networks (Patel, 2008). Moreover, Rogalsky,
Rong, Saberi, and Hickok (2011), using a within-subject
design, recently reported distinct patterns of spatial
distribution of activation to speech and music processing.
Thus, for the level of processing with which this study is
concerned, various approaches may converge on the
notion that music processing may be separable from other
auditory processing systems.
In the current study, we sought first to substantiate the
reliability of our previous findings by extending our
previous assessment to a considerably larger sample of
AD participants across the full spectrum of dementia
severity and to a considerably larger sample of controls.
Second, we also extended the battery of memory tests to
include tests of memory for spoken materials as well as
for music.
We constructed two new tests, one involving familiarity
with song lyrics, and the other involving grammatical
distortions to song lyrics. No claim is made that these
tests are completely equivalent to tests of familiarity with
melodies and distortions of melodies. Nevertheless, it is
likely that song lyrics are equally familiar as melodies
that accompany them, so that overall familiarity is
controlled, and it is likely that similar attentional
demands are present for tests containing melodies or
lyrics.
Two other tests were also constructed for exploratory
purposes. These were production tests requiring the
completion of a melody, by singing, when prompted by
words, and the completion of a proverb, in words, when
prompted by words. The first taps the ability to retrieve and
produce a related musical sequence, and the second taps the
ability to retrieve and produce a related but nonmusical
( verbal) sequence.
Method
TESTS
Familiarity Decision Test. The Familiarity Decision Test
(FDT; Liégeois-Chauvel, Peretz, Babaï, Laguitton, &
Chauvel, 1998) contains 10-s excerpts from 10 familiar
melodies and 10 novel melodies presented in random order
with piano-tone timbre. The familiar melodies are popular
in Western culture. The novel melodies are reversals of
the pitch sequence of the familiar melodies with small
adjustments of pitch and duration to be musically acceptable
(after Hébert, Peretz, & Gagnon, 1995). Participants are
instructed to listen to each melody and judge whether or
not it is familiar. The FDT served as a Familiar Melodies Test
to accompany the Familiar Lyrics Test.
Familiar Lyrics Test. The Familiar Lyrics Test (FLT)
contains 10 familiar lyrics and 10 unfamiliar lyrics—the
words of the first phrases of songs spoken by a male in a
monotonic voice and presented in random order. Familiar
lyrics are those of highly familiar songs (i.e., those in the
FDT and the DTT below). Unfamiliar lyrics are phrases
from obscure songs similar in length and thematic topics
to the familiar lyrics (selected from Preuss, 1942;
Thompson, 1937). Participants are instructed to listen to
each lyric and judge whether or not it is familiar.
Distorted Tunes Test. The Distorted Tunes Test (DTT;
Drayna, Manichaikul, de Lange, Snieder, & Spector,
2001) contains 26 brief popular melodies, 12 to 26 notes
in length, presented in random order with complex-tone
timbre. Seventeen melodies include pitch distortions—
sequences of pitches one or two semitones higher or
lower than the correct pitches of the original melody—
disrupting the initial tonality of the melody. The contour
(pattern of ups and downs) of the original melody is
preserved. Participants judge whether the melody sounds
correct or not.1
Distorted Lyrics Test. The Distorted Lyrics Test (DLT)
contains 30 lyrics—the first line or two of a popular
song spoken by a male in a monotonic voice and
presented in random order. Lyrics were selected from
those accompanying recognizable melodies (from the
FDT, the DDT and other tests in our laboratory).
Twenty of the lyrics contain alterations contrary to the
rules of standard English usage. For example, the
incorrect version of “Oh say can you see by the dawn’s
early light \ What so proudly we hailed at the twilight’s
last gleaming” is “Oh say can you saw by the dawn’s early
1 The test may be taken at www.nidcd.nih.gov/tunetest/Pages/dtt.
aspx<http://www.nidcd.nih.gov/tunetest/Pages/dtt.aspx>. Copies for
professional work may be obtained from Dennis Drayna, drayna@
nidcd.nih.gov.

Musical Memory 483
light \ What so prouder we hailed at the twilight’s last
gleamed.” Participants judge whether the lyric is correct
or incorrect.
Lyrics Prompt Test. For the Lyrics Prompt Test (LPT)
participants are asked to sing the melody corresponding
to each lyric judged familiar in the FLT. Participants are
assured they are not scored for the quality of their sing-
ing but simply whether they can produce a melody with
sufficient accuracy to indicate memory for the melody.
Proverbs Completion Test. The Proverbs Completion
Test (PCT) contains 21 proverb beginnings. The first half
of the proverb is read aloud to participants, for example
“A bird in the hand...” and participants are asked to
complete the proverb, the correct response being “…is
worth two in the bush.”
Selection of items was made on the basis of a pre-test
with 60 healthy older adults who were presented 60
proverb beginnings, three to seven words in length
(Simpson, 1982). Items selected for the PCT were suc-
cessfully completed by 80% of the pilot participants with
an ending corresponding exactly to the Simpson ending.
PARTICIPANTS
Controls. Controls were recruited from newspaper
advertisements, flyers, or posters distributed in the
university and the local community: 50 healthy younger
adults (university students) and 100 healthy older
adults.2 English was their primary language and all were
born in an English speaking country and/or to an
English speaking family. They received modest remu-
neration for their involvement.
Participants reported good cognitive status and phys-
ical health. However, for older adults, exclusion criteria
2 The large sample of healthy older adults was intended as a data-
base from which to find demographic matches for AD participants,
should demographic characteristics be found to be significant cor-
relates of test scores. However, as no such relationships were found
important, data for the full sample of 100 adults are reported.
applied were a Mini Mental Status Examination (MMSE;
Folstein, Folstein, & McHugh, 1975) less than 28 out of
30, and measured hearing loss greater than 40 dB in the
range 250-1000 Hz as assessed with a manual pure-tone
audiometer (Madsen Electronics Midimate 602
Diagnostic Audiometer) following standard procedures
(American Speech-Language-Hearing Association,
2005). Fifteen older adults, in addition to the 100
reported on here, were excluded, 12 on the basis of
MMSE scores and 3 on the basis of hearing loss.
The upper part of Table 1 shows, for the younger and
older adults, sex distribution, median and range of age,
years of general education, and years of music education.3
AD participants. Recruitment for AD participants was
carried out through community advertising, support
group sessions held by the Alzheimer Society of Kingston,
and a geriatric medicine outpatient clinic. Fifty AD
participants were assessed. English was the primary
language and all were born in an English speaking
country and/or to an English speaking family. They
received modest remuneration for their involvement.
For AD par ticipants, the primary criterion for
inclusion was a diagnosis of dementia of the Alzheimer
type as defined by the Diagnostic and Statistical Manual
for Mental Disorders, Fourth Edition Text Revision
(DSM-IV-TR: American Psychiatric Association, 2000)
and the National Institute of Neurological and
Communicative Disorders and Stroke-Alzheimer’s
Disease and Related Disorders Association (NINCDS-
ADRDA: McKhann et al., 1984). Exclusions were any
co-morbid psychosis, a history of alcoholism, tumor, or
major cerebral vascular accident. Information was
obtained from medical records, including computed
tomography (CT) imaging of the brain where available
3 Occupational status was also collected for all adult participants.
This variable was closely tied to general education, however, so only
the latter is reported here.
TablE 1. Demographic Data for Control and AD Participants.
Sex Age Years of Education Years of Music Training
N M F Median Range Median Range <1 1-3 4-7 7-10 >10
Controls
Younger Adults 50 20 30 20.0 17-24 14.0 12-17 7 13 15 8 7
Older Adults 100 43 57 72.0 65-90 14.0 8-22 51 21 16 4 8
Levels of Dementia
Mild AD 21 10 11 82.0 66-86 12.0 8-21 12 3 2 0 4
Moderate AD 17 8 9 81.0 72-96 12.0 8-21 10 2 2 1 2
Severe AD 12 5 7 82.5 69-94 16.0 8-21 7 3 1 0 1

484 Lola L. Cuddy, Jacalyn M. Duffin, Sudeep S. Gill, Cassandra L. Brown, Ritu Sikka, & Ashley D. Vanstone
(36 of 50 participants), and supplemented by caregiver
interviews.
About two-thirds of AD participants were able to
complete the hearing assessment and meet the criterion
for healthy adults. Others either could not follow
instructions for the hearing assessment or were wearing
hearing aids that interfered with the audiometer
headphones. However, as they could respond to speech at
a conversational level, they were retained in the sample.
AD participants were classified into one of three
groups—mild, moderate, or severe stage of dementia—
on the basis of prior medical evidence, caregiver inter-
views, and MMSE scores at the time of the experimental
tests. The lower part of Table 1 shows sex distribution,
the median and range of age, years of general education,
and years of music education for each AD group.
Table 2 reports for each stage of AD, the median and
range of age at initial diagnosis, MMSE scores, the
number of participants in each stage receiving AD
medication (cholinesterase inhibitors such as donepezil
[Aricept], galantamine [Reminyl], and rivastigmine
[Exelon]) and, where appropriate, Disability Assessment
of Dementia (DAD; Gelinas, Gauthier, McIntyre, &
Gauthier, 1999) scores based on caregiver assessments.
DAD assessment was developed as a tool to measure the
functional abilities in activities of daily living for
community dwelling adults and is not applicable to care-
assisted living; therefore the scores in Table 2 do not
include two mild, seven moderate, and seven severe AD
participants in care homes. Two additional participants
in the Mild AD group said that they had no difficulty
with activities of daily living and declined to answer the
questionnaire.
Most AD participants were medically diagnosed as
pure cases of AD, but three mild cases, four moderate
cases, and one severe case were diagnosed with mixed
dementia with some additional indication of minor
vascular involvement in their medical record. As well,
three (one mild and two moderate cases) had additional
symptoms suggestive of Parkinson disease. Two (one
mild and one severe case) were later given a final
diagnosis of Lewy Body Disease. These persons were
retained in the sample because they met the medical
description of dementia of the Alzheimer type in their
case histories.
APPARATUS AND PROCEDURE
Controls. Controls were tested in sound isolated
chambers at Queen’s University. Data were collected in
the following fixed order: demographic information,
questionnaire on music background, audiometric
assessment (older adults only), FDT, DTT, DLT, MMSE
(older adults only), FLT, LPT, and PCT. Younger adults
completed the entire study protocol in one session
which lasted approximately one hour. Older adults
were tested in two separate sessions, scheduled typically
a week apart, at their convenience. Older adults
completed the FDT, DTT, DLT, and MMSE in the first
session with the remaining three tests completed in the
second. Only one data point was not recorded; one
older adult refused to sing for the LPT.
Participants reported their demographic information
and music background on a paper and pencil questionnaire.
Stimuli for the FDT, DTT, DLT, and FLT were MP3
recordings on an Apple iPod Nano and played through
Logitech Pure-Fi Anywhere Compact Speakers. Participants
responded verbally and the experimenter wrote down their
responses. PCT beginnings were presented on a printed
sheet and read aloud by the experimenter. Responses to the
PCT and the LPT were recorded using a Roland Edirol R-09
MP3 Recorder. Recorded responses were played back and
scored using Audacity Software.
AD participants. Because of the varied cognitive and
physical capacities of our participants, the procedure was
necessarily more flexible. Testing was conducted either in
a quiet room at the participant’s residence, or in a quiet
room at Queen’s University. The sessions were distributed
over as many days as necessary depending on each
participant’s schedule and the length of time he or she was
willing to spend at each session. Participants typically
completed the study protocol in two to four sessions.
Demographic and music background information
were collected first in an interview with both the
participant and a family member. In general, the order
of the tests was the same as for the healthy older adults.
However, some flexibility in test order was allowed to
TablE 2. Characteristics of the Three AD Group s.
Age at Diagnosis MMSE Scores DAD Score Medication for AD
Median Range Median Range Median Range
Number Taking
Medication
Mild AD 79.0 63-85 25 20-30 69 (N = 17) 42-95 17
Moderate AD 77.0 68-94 16 12-21 51 (N = 11) 15-70 13
Severe AD 77.0 66-89 4 0-10 16 (N = 6) 2.5-53 10

Musical Memory 485
keep the AD participant interested and focused. If a par-
ticipant declined to complete a test, the tester moved to
the next test and tried to resume the declined test at a
different session. Two participants in the Moderate AD
group and four in the Severe AD group were unable to
complete the full test protocol.
Stimuli were played using the same test equipment as
for controls. Audiometric assessment was done using a
portable GSI 17 Audiometer when testing was conducted
at a site away from Queen’s University. For AD
participants, all test sessions were video recorded using
a Panasonic PV-GS150 digital camcorder for later
scoring by independent raters. Videos were uploaded
and scored using MotionDV Studio Software.
Results
SCORING
Percent correct was obtained for each participant who
completed each test with one slight modification. For
the LPT, each singing response was scored as 0 (cannot
not sing the corresponding melody), .5 (can sing but
not entirely correctly—for example, begins with the
correct melody and finishes with another melody), or 1
(sings a recognizable rendition of the corresponding
melody). The total score was divided by the number of
lyrics in the FLT correctly recognized by the participant
and converted to a percentage.
RELIABILITY OF TEST SCORING FOR AD PARTICIPANTS
For AD participants, the test administrator and one
independent rater scored each video taped session. The
independent raters were research assistants not present
at the testing session. Mild and moderate AD participants
responded verbally to the test instructions and for these
participants, rater agreement was uniformly high—96%
across the six tests for mild AD and 94% for moderate
AD. Severe AD participants did not readily respond to
test instructions but were frequently able to indicate
familiarity and recognition of distortion with behavioral
gestures (see Cuddy & Duffin, 2005). Rater agreement
was slightly lower at 92% overall; it was higher for the
FDT at 95% but lower for the tests involving lyrics and
words where agreement on average was 90%. Cases of
disagreement were resolved by consultation of the two
raters with a third independent rater.
TEST RESULTS
The distribution of scores in 20 of the 30 conditions (5
groups x 6 tests) departed significantly from normality
(Shapiro-Wilk, p < .05). Thus, nonparametric statistics
were conducted for the analyses below.
Variability within groups: Demographic characteristics. We
first sought to determine whether the variability within
groups was related to sex and level of music education, for
all five groups, and to age and level of general education for
the four adult groups. (Age and general education were not
sufficiently variable to justify analysis within the younger
adult group.) The Bonferroni correction was applied to
determine the significance level required for α < .05.
No significant sex differences were detected by the
Mann-Whitney statistic for any group and any test. No
significant relation was detected by Spearman correlations
between age and general education for any of the four
older groups and for any test. Music education was
predictive of test results in only a few instances. Significant
correlations were found for the DTT for younger adults,
ρ(50) = .56, for the DTT and older adults, ρ(100) =. 42,
and for the LPT and older adults, ρ(99) = .42.
Within each AD group, scores were rank ordered and
the data inspected for evidence of rank differences
betwe en participants receiving or not receiving
medication, and between participants with pure
diagnoses versus mixed diagnoses. No consistent patterns
were found within any group.
Variability between groups: Age and AD. Though the
ranges were comparable, the median age of the older
adults was lower than the median ages for the AD groups.
Therefore, a sample of 25 participants was drawn from
the older adult group so that median age (as well as the
distribution of gender, general education and music
education) was matched to the AD groups. Not
surprisingly given the results in the previous section, no
test differences were found between this selected group
and the remainder of the older adult group. Data will
therefore be presented for the full sample of older adults.
Figure 2 presents median score and interquartile range
for each group and each test. Percent correct is shown on
the y-axis and the five groups—in order younger, older,
mild AD, moderate AD, and severe AD—on the x-axis.
Sample size is given in Table 3. Kruskal-Wallis’ one-way
analysis by ranks revealed significant differences among the
groups on each test, 36.6 < χ2 < 100.4, df = 4, all p <.05.
Of greater interest is the question where breakdown
in performance occurs for the AD groups. Mann-
Whitney statistics for pair-wise comparisons are given
in Table 3. Bolded cells in Table 3 are values significant
for α < .05 according to the Bonferroni correction. The
results in Figure 2 and Table 3 are summarized as follows.
Comparison of control groups. Control performance was
typically very high. Eleven of the 12 median scores (two
groups x six tests) were above 85%. No difficulties for
older adults were detected; in fact, according to Table 3,
scores for older adults were significantly higher than

486 Lola L. Cuddy, Jacalyn M. Duffin, Sudeep S. Gill, Cassandra L. Brown, Ritu Sikka, & Ashley D. Vanstone
FIGUR E 2. Results of five participant grou ps (younger and older adul t controls and three levels of AD—mild , moderate, severe) on each of six tests
(Familiarit y Decision Test [FDT], Familiar Lyrics Test [FLT], Distorted Tunes Test [DTT], Distorte d Lyrics Test [DLT], Lyrics Prompt Test [LPT], and
Proverbs Comp letion Test [PCT]).

Musical Memory 487
TABLE 3. Results of Mann-W hitney Pair-wise Comparison s for Six Tests.
Familiarity Decision Test
(FDT)
Older Adults
(N = 100)
AD Mild
(N = 21)
AD Moderate
(N = 17)
AD Severe
(N = 12)
Younger Adults
(N = 50)
U = 1714.5
p = .001
U = 385.5
p = .07
U = 402.0
p = .74
U = 59.5
p = 1.39E-05
Older Adults
(N = 100)
– U = 1010.0
p = .78
U = 591.5
p = .04
U = 60.5
p = 2.04E-07
AD Mild
(N =21)
– – U = 130.5
p = .15
U = 21.5
p = 7.88E-05
AD Moderate
(N = 17)
– – – U = 29.0
p = .001
Familiar Lyrics Test
(FLT)
Older Adults
(N = 100)
AD Mild
(N = 21)
AD Moderate
(N = 16)
AD Severe
(N = 9)
Younger Adults
(N = 50)
U = 1093.0
p = 2.26E-10
U = 372.5
p = .05
U = 348.5
p = .43
U = 76.0
p = .001
Older Adults
(N = 100) –
U = 685.0
p = .002
U = 396.0
p = 7.34E05
U = 14.0
p = 4.36E-09
AD Mild
(N =21)
– – U = 136.0
p = .31
U = 12.5
p = .0001
AD Moderate
(N = 16)
– – – U = 15.5
p = .001
Distorted Tunes Test
(DTT)
Older Adults
(N = 100)
AD Mild
(N = 21)
AD Moderate
(N = 17)
AD Severe
(N = 10)
Younger Adults
(N = 50)
U = 2217.0
p = .26
U = 322.5
p = .01
U = 138.0
p = 2.83E-05
U = 8.5
p = .0002
Older Adults
(N = 100)
–U = 739.0
p = .03
U = 356.5
p = .0001
U = 38.0
p = 1.10E-06
AD Mild
(N =21)
– – U = 114.5
p = .06
U = 39.0
p = .005
AD Moderate
(N = 17)
– – – U = 68.0
p = .39
(Continued)

488 Lola L. Cuddy, Jacalyn M. Duffin, Sudeep S. Gill, Cassandra L. Brown, Ritu Sikka, & Ashley D. Vanstone
Distorted Lyrics Test
(DLT)
Older Adults
(N = 100)
AD Mild
(N = 21)
AD Moderate
(N = 15)
AD Severe
(N = 8)
Younger Adults
(N = 50)
U = 2360.0
p = .57
U = 167.5
p = 5.90E-06
U = 117.5
p = 5.39E-05
U = 2.0
p = 6.96E-06
Older Adults
(N = 100)
–U = 404.5
p = 8.51E-06
U = 264.0
p = 4.58E-05
U = 11.0
p = 3.95E-06
AD Mild
(N =21)
– – U = 139.5
p = .56
U = 36.0
p = .02
AD Moderate
(N = 15)
– – – U = 26.0
p = .03
Lyrics Prompt Test
(LPT)
Older Adults
(N = 99)
AD Mild
(N = 21)
AD Moderate
(N = 17)
AD Severe
(N = 9)
Younger Adults
(N = 50)
U = 1174.5
p = 5.89E-08
U = 418.0
p = .17
U = 243.0
p = .02
U = 46.5
p = .0002
Older Adults
(N =99)
–U = 836.5
p = .13
U = 276.0
p = 1.02E5
U = 81.0
p = 1.58E-05
AD Mild
(N =21)
– – U = 90.5
p = .02
U = 29.5
p = .003
AD Moderate
(N = 17)
– – – U = 46.5
p = .14
Proverbs Completion Test
(PCT)
Older Adults
(N = 100) AD Mild
(N = 21) AD Moderate
(N = 17) AD Severe
(N = 9)
Younger Adults
(N = 50) U = 472.0
p = 2.61E-16 U = 362.0
p = .04 U = 416.5
p = .90 U = 36.0
p = 6.31E-05
Older Adults
(N = 100) –U = 361.0
p = 1.31E-06U = 213.0
p = 4.15E-07
U = 8.0
p = 5.03E-07
AD Mild
(N =21) ––U = 132.5
p = .18 U = 6.5
p = 6.44E-05
AD Moderate
(N = 17) –– – U = 14.5
p = .0008
Note: Bolded entries are signicant, p < .05 with Bonferroni correction
TablE 3. Continued.

Musical Memory 489
younger adults on three tests—the Familiar Lyrics Test,
the Lyrics Prompt Test, and the Proverbs Completion Test.
Comparison of AD groups with older adults. The
following summaries for six tests are based on the results
in Figure 2 and Table 3.
FDT: In Figure 2, upper left panel, it can be seen for the
Familiarity Decision Test that mild and moderate AD
participants performed comparably to older adults. Table
3 shows that only the severe AD participants scored
significantly lower than the older adults. Even so, however,
the median for the severe AD participants (72.5% correct)
was within, and the range of scores (55-85% correct)
overlapped, the range for older adults (65-100% correct).
FLT: Results for the Familiar Lyrics Test, upper right
panel, showed very high performance for older adults
and the mild and moderate AD groups. The severe AD
group dropped (significantly) to a median of 80%
correct but the range of scores (55-90% correct)
overlapped the range for older adults (80-100% correct).
DTT: The Distorted Tunes Test, middle left-hand
panel, was typically easy for both controls and the mild
AD participants (with variability greater for this test
than the two tests above). Median scores for the moderate
AD and severe AD participants were significantly lower
than for the older adults. For the severe AD group, the
median of 54% correct was essentially at chance.
DLT: The Distorted Lyrics Test, middle right-hand
panel, shows a distinct differentiation between the per-
formance of the older adults and the AD groups. All AD
groups had difficulty with the DLT with even the mild
AD participants differing significantly from the controls.
For the severe AD participants, the median at 44%
reflected chance performance.
LPT: The Lyrics Prompt Test, lower left-hand panel,
yielded similar median performance for the older adults
and mild AD participants. However, the medians for the
moderate and severe AD participants were significantly
below the older adults. Variability was quite high. Despite
the low median score in the moderate and severe AD
groups, several moderate and one severe AD participant
performed near the top of the range for the older adults.
More typically in these groups, however, one or two songs
were attempted with fairly good results but the participant
was unable to respond to the full set. One of the test
administrators noted that “some just seemed to lose the idea
of what they were supposed to be doing with the task as it
progressed.”
PCT: Finally, the Proverbs Completion Test, lower
right-hand panel, distinctly reveals the levels of severity
of the disease. Compared to the high median score for
the older adults, the medians for all the AD groups were
significantly lower; the medians were, in order, mild,
moderate, and severe AD participants.
Discussion
Performance on six tests involving different aspects of
melody and language processing was assessed across two
levels of age and three levels of dementia severity. No test
was immune to the presence of dementia. Overall, the
most important predictor of test performance for AD
adults was the level of cognitive impairment reflected by
group membership. Overall, the role of demographic
variables age, sex, and education was slight (or nonexis-
tent). The findings with the present sample support
earlier observations with much smaller samples
(Vanstone & Cuddy, 2010), but, likely because of the
larger sample in the present study, regular effects of level
of dementia severity were uncovered.
However, not all tests were equally sensitive to dementia,
and several relevant points emerged. First, supporting our
earlier work, long-term familiarity for melody was found
across levels of dementia, even at the severe stage for a few
individuals. Second, persons with mild, and some with
moderate, dementia detected melodic distortions within
the performance range of normal older adults. Third, the
ability to sing a melody when prompted by its lyrics was
retained at the mild stage and was retained by a few indi-
viduals through the severe stages of dementia (though
variability was high). Fourth, long term familiarity with
the lyrics of familiar melodies was also found across levels
of dementia. Fifth, in contrast, however, detection of
grammatical distortions in the lyrics of familiar melodies
and the ability to complete familiar proverbs were affected
even at the mild stage of dementia. Scores decreased
regularly across levels of dementia, and thus supported
the original diagnosis.
Returning to the implications of the model proposed in
Figure 1, we may now offer further support for the notion
that the musical lexicon, as a component of the music
recognition system, may be spared in early to moderate
stages of AD and may be spared for some individuals in
the severe stage. This result stands in stark contrast to re-
sults for tests of episodic memory where performance is
markedly impaired even at the mild stage of AD.
Second, the data revealed that formal music training
may not be necessary for musical sparing in AD. It ap-
pears that advanced motor skills, such as those involved
in playing an instrument, are not required for evidence
of musical sparing. No relation between music training
and test performance was found within any AD group.
However, in this regard, a study quantifying music
exposure and engagement would be welcome, as factors
of music listening experience rather than performance
might be important correlates of sparing.
Procedural memory may play a facilitating role for pro-
fessional musicians, but it is unlikely the whole account.

490 Lola L. Cuddy, Jacalyn M. Duffin, Sudeep S. Gill, Cassandra L. Brown, Ritu Sikka, & Ashley D. Vanstone
In this regard, a recent study, though it deals with semantic
dementia rathe r than AD, uncovered preserved
performance ability in a semi-professional harpsichordist
(Weinstein & McMillan, 2011). What is particularly
noteworthy is that the musician played with expression
and was able to create stylistically appropriate embellish-
ments. The authors conclude that “his retained musical
ability is not procedural (i.e., habitually striking keys in
response to visual stimuli) but reflect his meaningful and
productive control over musical knowledge…the key
elements of musical semantics.” (p. 250)
Finally, regarding memory for lyrics with respect to the
modular model of Figure 1, the performance of the AD
participants on the distorted lyrics test suggests that access
to, or representation of, lyrics in the phonological store is
impaired. However, the remarkably good performance of
AD participants on recognition of familiar lyrics suggests
that there was sufficient activation of the phonological
store to co-activate spared musical lexicon. Because of this
associative bond, familiarity of the lyrics was detected.
In sum, we have put forth the argument that musical
semantic memory may be spared through the mild and
moderate stages of AD and may be preserved even in
some individuals at the severe stage. Tonal knowledge,
that plays an integral role in how musical sequences are
perceived, organized and remembered, may also be
spared in the mild to moderate stages. Musical semantic
memory may be dissociable from episodic memory
where impairment is evident even at the mild stage.
Further work should outline the boundaries of the
sparing of musical semantic memory and should explore
the relation between the sparing of associated lyrics.
Understanding preserved capacities should facilitate
efforts to use music successfully in intervention pro-
grams and programs directed at fostering new learning.
Author Note
This research was supported by grants from the Natural
Sciences and Engineering Council of Canada, the
GRAMMY Foundation®, and the Alzheimer Society of
Canada (Dr. Albert Spatz Award) — principal investigator,
Lola L. Cuddy. We acknowledge the help and facilities of
the Queen’s Biological Communication Centre (Kevin
Munhall, Director), the valuable research assistance of
Avanti Dey, Laura Cuttini, Christopher Trimmer, Tanya
Montenay, Joanne Kim, Sally Suen, Lydia Beck, Jordan
Theriault, Trina Ghiacy, and Afiqah Yusuf, and the
insightful comments of the reviewers. We thank all our
participants, especially those with AD and their care
partners, who kindly assisted in the data collection.
Portions of the research have been presented at meetings
of the European Society for the Cognition of Music (2009)
and the Canadian Psychological Association (2010).
Correspondence concerning this article should be
addressed to Lola L. Cuddy, Department of Psychology,
Queen’s University, Kingston, ON K7L 3N6. e-MaiL: Lola.
Cuddy @queensu.ca
References
aMeriCan pSyChiatriC aSSoCiation. (2000). Diagnostic and
statistical manual of mental disorders (4th ed., text rev.)
Washington, D.C.: Author.
aMeriCan SpeeCh-LanGuaGe-hearinG aSSoCiation. (2005).
Guidelines for manual pure-tone threshold audiometry.
Rockville, MD. Available online from http://www.asha.org/
docs/html/GL2005-00014.html
Baird, a., & SaMSon, S. (2009). Memory for music in Alzheimer’s
disease: unforgettable? Neuropsychology Review, 19, 85–101.
BartLett, J. C., haLpern, a. r., & dowLinG, w. J. (1995).
Recognition of familiar and unfamiliar melodies in normal
aging and Alzheimer’s disease. Memory and Cognition, 23,
531–546.
Beatty, w. w., winn, p., adaMS, r. L., aLLen, e. w., wiLSon,
d. a., prinCe, J. r., & oLSon, k. a. (1994). Preserved cognitive
skills in dementia of the Alzheimer type. Archives of
Neurology, 51, 1040–1046.
BeSSon, M., ChoBert, J., & Marie, C. (2011). Transfer of training
between music and speech: Common processing, attention,
and memory. Frontiers in Psychology, 2, 94. doi: 10.3389/
fpsyg.2011.00094
CowLeS, a., Beatty, w. w., nixon, S. J., Lutz, L. J., pauLk, J., pauLk, k.,
et aL. (2003). Musical skill in dementia: A violinist presumed to have
Alzheimer’s disease learns to play a new song. Neurocase, 9, 493-503.
Cuddy, L. L., & duffin, J. (2005). Music, memory, and Alzheimer’s
disease: Is music recognition spared in dementia, and how can it
be assessed. Medical Hypotheses, 64, 229–235.
drayna, d., ManiChaikuL, a., LanGe, M. de Snieder, h., &
SpeCtor, t. (2001). Genetic correlates of musical pitch recog-
nition in humans. Science, 291, 1969–1972.
foLStein, M. f., foLStein, S. e., & MChuGh, p. r. (1975). “Mini-
mental state.” A practical method for grading the cognitive
state of patients for the clinician. Journal of Psychiatric
Research, 12, 189–98.
fornazzari, L., CaStLe, t., nadkarni, S., aMBroSe, M., Miranda,
d., apanaSiewiCz, n., et aL. (2006). Preservation of episodic
musical memory in a pianist with Alzheimer disease.
Neurology, 66, 610–611.
GeLinaS, i., Gauthier, L., MCintyre, M., & Gauthier, S. (1999).
Development of a functional measure for persons with
Alzheimer’s disease: The Disability Assessment for Dementia.
American Journal of Occupational Therapy, 53, 471–481.

Musical Memory 491
héBert, S., peretz, i., & GaGnon, L. (1995). Perceiving the tonal
ending of tune excerpts: the roles of pre-existing representa-
tion and musical expertise. Canadian Journal of Experimental
Psychology, 49, 193–209.
héBert, S., & peretz, i. (2001). Are text and tune of familiar songs
separable by brain damage? Brain and Cognition, 46, 169–175.
hSieh, S., hornBerGer, M., piGuet, o., & hodGeS, J. r. (2011).
Neural basis of music knowledge: Evidence from the demen-
tias. Brain, 134, 2523–2534.
koeLSCh, S. (2011). Toward a neural basis of music perception -
A review and updated model. Frontiers in Psychology, 2, 110.
doi: 10.3389/fpsyg.2011.00110.
kruMhanSL, C. L., & Cuddy, L. L. (2010). A theory of tonal hier-
archies in music. In M. Riess Jones, R. R. Fay, & A. N. Popper
(Eds.), Springer handbook of auditory research (Vol. 36, pp.
51–87). New York: Springer.
LiéGeoiS-ChauVeL, C., peretz, i., BaBaï, M., LaGuitton, V., &
ChauVeL, p. (1998). Contribution of different cortical areas in
the temporal lobes to music processing. Brain, 121, 1853–1867.
MCkhann, G., draChMan, d., foLStein, M., katzMan, r.,
priCe, d., & StadLan, e. M. (1984). Clinical diagnosis of
Alzheimer’s disease: Report of the NINCDS-ADRDA Work
Group under the auspices of Department of Health and
Human Services Task Force on Alzheimer’s Disease.
Neurology, 34, 939–944.
MouSSard, a., BiGand, e., BeLLeViLLe, S., & peretz, i. (2012).
Music as an aid to learn new verbal information in
Alzheimer’s disease. Music Perception, 29, 521–531.
oMar, r., haiLStone, J. C., & warren, J. d. (2012). Semantic
memory for music in dementia. Music Perception, 29,
467–477.
pateL, a. (2008). Music, language, and the brain. New York:
Oxford University Press.
patSton, L. L. M., & tippett, L. J. (2011). The effect of back-
ground music on cognitive performance in musicians and
nonmusicians. Music Perception, 29, 173–183.
peretz, i. (2001). Brain specialization for music. New evidence
from congenital amusia. Annals of the New York Academy of
Sciences, 930, 153–65.
peretz, i. (2006). The nature of music from a biological per-
spective. Cognition, 100, 1–32.
peretz, i., radeau, M., & arGuin, M. (2004). Two-way interac-
tions between music and language: Evidence from priming
recognition of tune and lyrics in familiar songs. Memory and
Cognition, 32, 142–152.
peretz, i., & CoLtheart, M. (2003). Modularity of music
processing. Nature Neuroscience, 6, 688–91.
peretz, i., GoSSeLin, n., BeLin, p., zatorre, r. J., pLaiLLy, J., &
tiLLMann, B. (2009). Music lexical networks: The cortical
organization of music recognition. Annals of the New York
Academy of Sciences, 1169, 256–265.
pLateL, h., priCe, C., Baron, J. C., wiSe, r., LaMBert, J.,
fraCkowiak, r. S., et aL. (1997). The structural components
of music perception. A functional anatomical study. Brain,
120 (Pt 2), 229–243.
pLateL, h., Baron, J.-C., deSGranGeS, B., Bernard, f., & euStaChe,
f. (2003). Semantic and episodic memory of music are sub-
served by distinct neural networks. NeuroImage, 20, 244–256.
preuSS, t. (Compiler and Arranger). (1942). Christmas in song.
Milwaukee, WI: Hal Leonard Publishing.
roGaLSky, C., ronG, f., SaBeri, k., & hiCkok, G. (2011). Functional
anatomy of language and music perception: Temporal and struc-
tural factors investigated using functional magnetic resonance
imaging. The Journal of Neuroscience, 31, 3843–3852.
SaMSon, S., Baird, a., MouSSard, a., & CLéMent S. (2012). Does
pathological aging affect musical learning and memory?
Music Perception, 29, 493–500.
SiMpSon, J. a. (1982). The concise Oxford dictionary of proverbs.
Oxford, UK: Oxford University Press.
Steinke, w. r., Cuddy, L. L., & JakoBSon, L. S. (2001).
Dissociations among functional subsystems foverning melody
recognition after right-hemisphere damage. Cognitive
Neuropsychology, 18, 411–437.
Strait, d., & krauS, n. (2011). Playing music for a smarter ear:
Perceptual and neurobiological evidence. Music Perception,
29, 133–146.
thoMpSon, G. V. (Compiler). (1937). Sociability songs. Toronto,
ON: Gordon V. Thompson, Ltd.
tuLVinG, e. (1983). Elements of episodic memory. London, UK:
Oxford University Press.
tuLVinG, e. (1985). How many memory systems are there?
American Psychologist, 40, 385–398.
VanStone, a. d., & Cuddy, L L. (2010). Musical memory in Alzheimer
disease. Aging, Neuropsychology, and Cognition, 17, 108–128.
VanStone, a.d., Sikka, r., tanGneSS, L., ShaM, r., GarCia, a.,
& Cuddy, L.L. (2012). Episodic and semantic memory for
melodies in Alzheimer’s disease. Music Perception, 29,
501–507.
weinStein, J., & MCMiLLan, C. (2011). Preserved musical semantic
memory in semantic dementia. Archives of Neurology, 68, 248–250.

Reproducedwithpermissionofthecopyright owner.Furtherreproductionprohibitedwithout permission.