Article Review - "Findings from the Nun Study"

     Glenn Mason-Riseborough (19/10/1998)

Review of article submitted by D. A. Snowdon, S. J. 
Kemper, J. A. Mortimer, L. H. Greiner, D. R. Wekstein, and W. 
R. Markesbery, under the title of Linguistic Ability in Early Life 
and Cognitive Function and Alzheimer’s Disease in Late Life: 
Findings From the Nun Study.  The article appeared in the 
journal JAMA on February 21, 1996.  The intention of this 
review is to overview the article and determine whether or not it 
is suitable for publication in the journal, Neuropsychology.  In 
addition, should the article be deemed suitable, this review will 
determine whether or not any modifications are necessary before 
publication.

Introduction:  This section introduces a number of theories 
that attempt to explain why low attained education is associated 
with dementia in older adults.  It cites a previous theory that 
suggests that one reason may be lifestyle differences associated 
with education such as nutrition, alcohol consumption, and 
occupational exposures.  A second mentioned theory is the 
threshold model of dementia.  On this theory, dementia is only 
evident once the cognitive or neurological reserve falls below a 
certain level.
The article under review suggests that while education may 
reflect cognitive ability, linguistic ability may be a better indicator 
than education for determining this cognitive ability and 
neurological reserve.  The suggestion is that a high level of 
linguistic ability in early life may slow cognitive decline in later 
life by providing tools for memory encoding, processing, and 
retrieval.  It states previous research in which linguistic ability 
may be measured by analysing the form and content of language 
samples.  The research indicated that a decline in grammatical 
complexity in later life was associated with working memory 
limitations in healthy adults and their ability to use and develop 
complex grammatical constructions.
The study presented in the article under review examined a 
subset of the Nun Study population who had handwritten 
autobiographies from early life.  The relationship between the 
linguistic ability examined in the autobiographies and cognitive 
function was examined with the objective of finding a correlation 
between low linguistic ability and Alzheimer’s disease.
While it is not explicitly stated, it is the assumption of this 
reviewer that if this correlation is significant in the Nun Study 
sample, then the first theory discussed will be strongly doubted 
and further evidence will be provided for the second theory.  This 
should be explicitly stated in the introduction.
Overall, the introduction provides adequate background 
and rationale for the research documented.  The intended 
investigation of the research is clear.  However, it may be useful 
to clarify which theories would be affected by the outcomes of 
the research.

Method:  The study presented in the article under review 
used a subset of 93 participants of the Nun Study.  Participants 
did not differ significantly by country of birth, age, race or annual 
mortality rate.  The autobiographies were written just before the 
sister took her religious vows and ten sisters confirmed that they 
had written their autobiographies without assistance.  While we 
can probably surmise that all other sisters were likewise without 
assistance in the writing of their autobiographies, it would be 
preferable to have all participants confirm this, or explain the 
rationale for not researching this.  In addition, it should be 
explained further why the typewritten autobiographies were not 
included as part of the data set.
Two indicators of linguistic ability were used – idea density 
and grammatical complexity.  The mean of each per sentence was 
derived from the last ten sentences of each autobiography.  There 
was no explanation as to why only the last ten sentences were 
used.
The article under review explains how the two indicators 
were defined.  However, in a letter to the editor of JAMA dated 
June 26, 1996, Gary Miranda (1996) notes that there is a 
difference in reading and writing linguistic ability.  He suggests 
that while it is true that the ability to read complex sentences 
indicates linguistic ability, it is precisely the opposite when 
writing.  In other words, writing grammatically complex 
sentences with multiple ideas tends to be a result of poor writing 
skills.  Although Snowden et al. (1996) address this issue in a 
response also published on June 26, 1996 in JAMA, this is not 
addressed in the article itself.  This is an issue that needs to be 
addressed in further detail before publication should proceed.
The scoring for the autobiographies on the indicators was 
made by a single coder (blinded to age and cognitive function).  
Ten autobiographies were also scored by a second coder, with a 
resultant high intercoder correlation.  To provide additional 
verification it may have been useful for the second coder to score 
all autobiographies and an average over both coders taken, and it 
was not clearly explained why ten were chosen and how this was 
achieved (eg random or extremes).  47 of the 93 sisters also 
provided new autobiographies to check for reliability.  There was 
a high correlation between the earlier and later autobiographies.  
However, it was not clearly explained why the other 46 sisters 
did not produce new autobiographies and the method for 
determining which sisters provided new autobiographies.
The article under review also discusses the testing methods 
of cognitive function.  A battery of seven tests was used based 
on previous studies of testing Alzheimer’s disease patients.  The 
article also explained the neuropathological evaluation and how 
Alzheimer’s disease was diagnosed in the sisters.  In addition 
there was a brief inclusion of the statistical tools used.  All of 
these subsections were sufficient and no modifications need to be 
made before publication.
In general the method section provided sufficient 
information to address the questions posed.  Except for the areas 
indicated, the methodology and reasons were sufficiently clear.  
However, these should be addressed and the explanations 
expanded on before publication proceeds.

Results:  The findings are clearly presented and complete.  
All analyses are appropriate to the data and the questions.  In 
addition, the tables provide useful and relevant information and 
examples to the findings.  The statistics are completed 
appropriately to the study’s questions.

Conclusions:  The conclusions explicitly relate back to the 
questions posed in the introduction.  It states that there is a 
strong relationship between cognitive ability in early life, as 
indicated by linguistic ability, and cognitive function and 
Alzheimer’s disease in late life.  Relating back to a subset of 
results of college-educated sisters who were teachers, the 
conclusions follow that it is unlikely that there were confounding 
variables of education or occupation.  The conclusions also 
address the wider issue of lifestyle, which was mentioned in the 
introduction.  The sisters had similar lifestyles and while the 
article under review admits that this reduces the generalisability, 
it also avoids many confounding variables.  Thus, the results of 
the study do not give evidence for the first hypothesis mentioned 
in the introduction, but rather support the second hypothesis.  
This is explicitly stated.  The interpretations of the results are 
appropriate to arrive at these conclusions.

Overall:  The study discussed in the article under review 
provides new and useful information regarding possible 
indicators of Alzheimer’s disease.  The work is focused and 
concise and there are no sections that do not contribute to the 
paper as a whole.  As such, no sections need to be deleted.  
However, some sections, (specifically the introduction and 
method) could be elaborated on as summarised below.

It is the opinion of this reviewer that the manuscript should 
be published.  However, a number of issues need to be addressed 
and some minor modifications made.  Firstly, the introduction 
needs to relate the current study more explicitly with the 
background theories.  Secondly, it should be detailed why not all 
sisters were asked if they wrote their autobiographies without 
assistance.  In addition, it should be elaborated as to why 
typewritten autobiographies were not included in the data.  
Thirdly, it is questioned why only the last ten sentences were 
used for the analysis of the autobiographies.  Fourthly, the issue 
brought up by Miranda should be explicitly discussed in the 
article.  Fifthly, the method of choosing the data for the second 
coder to score was not explained, and also that of choosing 
which sisters to provide new autobiographies for comparison.

References:
Mirander, G. (1996). Letter to the editor. JAMA, 275, 1879.

Snowdon, D. A., Greiner, L. H., Wekstein, D. R., Danner, D., 
Markesbery, W. R., Kemper, S. J., & Mortimer, J. A. 
(1996). Letter to the editor. JAMA, 275, 1879.

Snowdon, D. A., Kemper, S. J., Mortimer, J. A., Greiner, L. H., 
Wekstein, D. R., & Markesbery, W. R. (1996). Linguistic 
ability in early life and cognitive function and Alzheimer’s 
disease in late life: Findings from the Nun Study. JAMA, 
275, 528-532.