Developmental History of the Glover-
Nilsson Smoking Behavioral Questionnaire
Elbert D. Glover, PhD; Fredrik Nilsson, MS; Åke Westin, MS
Penny N. Glover, MEd; Molly T. Laflin, PhD; Birger Persson, MS
Objective: To develop a simple,
easily administered pencil-and-pa-
Delivered by Ingenta to: ? IP: 146.185.203.42 on: Wed, 14 Feb 2018 06:35:04
per questionnaire to determine the
degree to which behavioral pat-
terns play a role in smoking depen-
dence. Methods: A modified Delphi
Copyright (c) PNG Publications. All rights reserved.
technique was used to identify ini-
tial questions and to eliminate ob-
vious duplications. Phase 2 uti-
lized multiple statistical methods
(principal components analysis,
cluster analysis, stepwise multiple
linear regression, cross tables,
Mantel-Haenzel χ 2 -test, and a
Gamma test) to evaluate and re-
T
obacco use has been cited as the
chief avoidable cause of illness and
death in American society, respon-
sible for more than 440,000 deaths annu-
ally in the United States.1 Smoking has
also been known to cause a host of dis-
Elbert D. Glover, Professor & Chair, Public
and Community Health, College of Health and
Human Performance, University of Maryland, Col-
lege Park, MD. Fredrik Nilsson, MS, Senior Clini-
cal Scientist, Pfizer Consumer Healthcare,
Helsinborg, Sweden. Åke Westin, MS, Statisti-
cian, Pfizer Consumer Healthcare, Helsinborg,
Sweden. Penny N. Glover, Public and Community
Health, College of Health and Human Perfor-
mance, University of Maryland, College Park, MD.
Molly T. Laflin, Professor, School of Family &
Consumer Sciences, Bowling Green State Univer-
sity, Bowling Green, OH. Birger Presson, Statis-
tician, Department of Statistics, Univesity of Lund,
Lund, Sweden
Address correspondence to Dr Glover, Public
and Community Health, 2387 HHP Building, Uni-
versity of Maryland, College Park, MD 20742. E-
mail: eglover1@umd.edu
™ 2005;29(5):443-455
Am J Health Behav.™
duce the number of questions from
18. Results: These analyses yielded
an 11-item questionnaire that can
potentially assess behavioral depen-
dence. Conclusion: It is hoped that
the GN-SBQ will assist physicians,
health care providers, and tobacco
interventionists in identifying as-
pects of smoking addiction that are
behavioral in nature. The need for
future research is discussed.
Key words: smoking cessation,
behavioral questionnaire, to-
bacco, smoking
Am J Health Behav. 2005;29(5):443-455
eases including cancer, heart disease,
stroke, complications of pregnancy, and
chronic obstructive pulmonary disease.2
Moreover, estimates indicate that ap-
proximately 25% of the adult population
smoke, and approximately 3200 adoles-
cents initiate smoking daily.3,4
In 1988, after examining the available
evidence, the Surgeon General’s Report
concluded that the principal pharmaco-
logic agent common in all forms of tobacco
is nicotine. Specifically, the report con-
cluded that (a) all forms of tobacco are
addicting, (b) nicotine is the drug in to-
bacco that causes addiction, and (c) the
pharmacologic and behavioral processes
that determine tobacco addiction are
similar to those that determine addiction
to other drugs such as heroin and co-
caine.5 Quantitative measures of physi-
cal dependence can be determined via
nicotine, cotinine, and thyiocynate lev-
els in saliva, urine, and plasma. Al-
though reliable, these assessments may
not be practical measures to utilize in
general medical practice due to sample
443
 Smoking Behavioral Questionnaire
collection, analysis, and cost. The 8-item
Fagerström Tolerance Questionnaire
(FTQ) is the most widely used measure of
nicotine dependence due to its ease of
administration, proven accuracy, and
cost-effectiveness.6 Researchers use the
FTQ to measure levels of physical depen-
dence based on scores ranging from 0 to
11; a score greater than or equal to 6
indicates a high level of dependence, and
a score less than 6 indicates lower depen-
dence.6 Surprisingly, no easily adminis-
tered questionnaire has been identified to
consistently measure the behavioral pro-
cesses that underlie smoking dependence.
Delivered by Ingenta to: ? IP: 146.185.203.42 on: Wed, 14 Feb 2018 06:35:04
It has been well documented that smok-
ing creates physical dependence on nico-
tine; moreover, the Fagerström Toler-
ance Questionnaire (FTQ)6 and its revi-
sion, the Fagerström Test for Nicotine
Copyright (c) PNG Publications. All rights reserved.
Dependence (FTND),7 have been used ex-
tensively for years to estimate this level
of physical dependence in smokers and
more recently in smokeless tobacco us-
ers (modified version). However, there is
no similar instrument to consistently
measure the behavioral aspects of smok-
ing addiction. These behavioral patterns
include the rituals associated with smok-
ing, the feelings or perceptions of security
that smoking provides, and the relation-
ship between the smoker and cigarette.
Traditionally the behavioral aspects of
smoking are addressed via counseling. A
quick, easily administered behavioral
assessment would allow clinicians to tai-
lor behavioral treatments to the indi-
vidual needs of the patient. Ideally, by
administering both a physical (FTQ or the
FTND) and a behavioral (GN-SBQ) ques-
tionnaire, clinicians could match treat-
ment to the behavioral as well as physical
needs of the patient. We would predict
that those with high scores on the FTQ or
FTND would benefit from pharmacologi-
cal intervention. Those with high scores
on the behavioral measure would be more
likely to require intensive behavioral in-
tervention. Patients who score high on
both the physical and behavioral mea-
sures would probably have the most diffi-
culty in quitting smoking and would most
likely require intensive support.
OBJECTIVE
The purpose of this project was to de-
velop a simple, easily administered pencil-
and-paper questionnaire, similar to the
FTQ/FTND, to assess the degree to which
444
behavioral patterns play a role in smoking
dependence. Specifically, this paper re-
veals the process by which the initial ques-
tions were developed, tested, and eventu-
ally reduced to form the Glover-Nilsson
Smoking Behavioral Questionnaire (GN-
SBQ). Reliability of the instrument is not
addressed in this manuscript. Specifi-
cally, the authors wish to share the pro-
cess by which the questionnaire was de-
rived. Current research is underway by
others to determine reliability of the GN-
SBQ and predictive validity with regard to
the need for behavioral treatment compo-
nents for smoking cessation.
METHODS
In 1992, discussion began regarding
development of a questionnaire to mea-
sure the degree to which behavioral pro-
cesses play a role in smoking depen-
dence. A modified Delphi process was
used in the initial phases of this study.
Four international tobacco treatment ex-
perts were asked to develop question-
naire items designed to measure per-
ceived behavioral dependence. All items
were developed independently by the panel
members; no items were selected from
any existing instruments. Round one
produced a total of 39 items.
In the second round, the panel was
asked to ensure that all key components
of behavioral dependence were adequately
addressed by the questions. They were
also asked to reduce the total number of
questions by eliminating those that were
duplicative in nature and omitting those
that were perceived to elicit physical de-
pendence rather than behavioral depen-
dence. This procedure eliminated 21
items, leaving 18 that could possible be
used to identify behavioral dependence.
In the next phase, the authors used a
variety of statistical techniques to elimi-
nate duplicative items and retain the
most critical items that address the be-
havioral processes of smoking. Table 1
presents the 18 items that were ana-
lyzed. Each question contained 5 stem
numerical options (0, 1, 2, 3, 4). A high
numerical response indicated high be-
havioral dependence, and a low numeri-
cal response indicated low behavioral de-
pendence.
This phase of the study involved 2231
individuals from 8 tobacco research trials
using comparable protocols in 4 countries
(Spain [n=1], Sweden [n=2], Switzerland
 Glover et al

Table 1
Initial Glover-Nilsson Smoking Behavioral Questionnaire (GN-SBQ)
Please indicate your choice by circling the number that best reflects your choice.
0=Not at all; 1=Somewhat; 2=Moderately so; 3=Very much so; 4=Extremely so

How much do you value the following (Specific to Questions 1-3).

1. My cigarette habit is very important to me. 0 1 2 3 4

2. I handle and manipulate my cigarette as part of the ritual of smoking. 0 1 2 3 4

3. I handle and manipulate my cigarette pack as part of the ritual of smoking. 0 1 2 3 4

Please indicate your choice by circling the number that best reflects your choice.
Delivered by Ingenta to: ? IP: 146.185.203.42 on: Wed, 14 Feb 2018 06:35:04

(Specific to Questions 4-18).

0=Never; 1=Seldom; 2=Sometimes; 3=Often; 4=Always

4. Do you keep your hands and fingers busy to distract you from smoking? 0 1 2 3 4
Copyright (c) PNG Publications. All rights reserved.

5. Do you place something in your mouth to distract you from smoking? 0 1 2 3 4

6. Do you reward yourself with a cigarette after accomplishing a task? 0 1 2 3 4

7. Do you panic if you find yourself out of cigarettes or if you cannot find your
cigarettes? 0 1 2 3 4

8. If you find yourself without cigarettes, will you have difficulties in concentrating
before attempting a task? 0 1 2 3 4

9. If you are not allowed to smoke in certain places, do you then play with your
cigarette pack or a cigarette? 0 1 2 3 4

10. Do certain environmental cues trigger your smoking, eg, favorite chair, sofa, room,
car, or drinking alcohol? 0 1 2 3 4

11. Do you find yourself lighting up a cigarette routinely (without craving)? 0 1 2 3 4

12. Will just holding a cigarette in your hand (without lighting up) assist you with
reducing stress? 0 1 2 3 4

13. Do you find yourself placing an unlit cigarette or other objects (pen, toothpick,
chewing gum, etc) in your mouth and sucking to get relief from stress, tension
or frustration, etc)? 0 1 2 3 4

14. Does part of your enjoyment of smoking come from the steps (ritual) you take
when lighting up? 0 1 2 3 4

15. Does part of your enjoyment of smoking come from watching the smoke as
you exhale? 0 1 2 3 4

16. Do you light up a cigarette without realizing you have another one burning in
the astray? 0 1 2 3 4

17. When you are alone in a restaurant, bus terminal, party, etc, do you feel safe,
secure, or more confident if you are holding a cigarette? 0 1 2 3 4

18. Do you light up when your friends light up? 0 1 2 3 4

™ 2005;29(5):443-455
Am J Health Behav.™ 445
 Smoking Behavioral Questionnaire
[n=1], United States [n=4]. After elimi-
nating questionnaires that contained
missing or questionable data, the au-
thors were left with 2032 usable ques-
tionnaires. All of the participants were at
least 18 years of age, classified as healthy
smokers (free of disease), and expressed
a desire to quit smoking. All participants
were recruited through physician refer-
rals, advertising via newsprint, televi-
sion, radio, or antismoking clinics. All 8
clinical trials received approval from their
respective institutional review boards.
Reducing the Number of Questions
Delivered by Ingenta to: ? IP: 146.185.203.42 on: Wed, 14 Feb 2018 06:35:04
The average face-to-face time for phy-
sician office visits in the United States is
17.4 minutes8 and 7-16 minutes in many
European countries; 9 therefore, to be
widely accepted, the GN-SBQ needed to be
Copyright (c) PNG Publications. All rights reserved.
brief. The overall objective of this project
was the development of a smoking behav-
ioral dependence questionnaire that could
be self-administered, easy to understand,
and take <90 seconds to complete.
To reduce the number of questions
from the 18 remaining at the end of the
Delphi process, principal components
analysis, cluster analysis, stepwise mul-
tiple linear regression, cross tables, Man-
tel-Haenzel χ2-test, and a Gamma test
were employed. These statistical meth-
ods provided an opportunity to conduct an
accumulated judgment of the fit and ne-
cessity of the various items. Questions
that produced redundant information were
eliminated.
Principal components analysis (PCA).
PCA is a technique that can be used when
a simple representation for a set of
intercorrelated variables is desired. The
technique can be summarized as a
method of transforming original variables
into new, uncorrelated variables. The
new variables are referred to as the prin-
cipal components or factors. One mea-
sure of the amount of information con-
veyed by each principal component is its
variance. For this reason the principal
components are arranged in order of de-
creasing variance. Thus with the PCA,
the authors analyzed principal compo-
nents to see which questions were mea-
suring distinct underlying concepts.
Cluster analysis (CA). CA is a multi-
variate analysis technique for grouping
individuals or objects into groups referred
to as clusters. The elements in a cluster
have relatively small distances from each
446
other and relatively large distances from
elements outside the cluster. Among the
different methods for performing a CA,
the authors chose a procedure in SAS
referred to as Varclus. Varclus clusters
the variables (in this case, questions) by
2 methods: (a) principal component cluster
analysis and (b) centroid component cluster
analysis. CA was used to find clusters of
questions and the distance between re-
sponses provided by the 2032 smokers.
The principal component cluster analy-
sis uses the first 2 cluster components.
The authors used the default value in
Varclus, which means splitting is not
executed until each cluster has only a
single eigenvalue greater than 1. The
centroid component cluster analysis mea-
sures the distance between a variable
(question) and its cluster’s center (cen-
troid). If the variable is closer to its own
cluster’s centroid, it is allowed to stay
within this cluster. Otherwise, it is moved
to another cluster with a centroid closer
to the concerned variable. The difference
between “R-squared with own cluster”
and “R-squared with the next closest”
should be large for well-separated clus-
ters. “R-squared with the next closest”
should be a low value if the clusters are
well separated.
Stepwise multiple linear regression
(SMLR). To further examine the relation-
ship between our variables (questions),
the authors also used a regression model
in which one question at a time became
the dependent variable, and the rest of
the questions were independent variables.
The variable with the highest F was en-
tered into the model first and partial F-
statistics were computed for all of the
remaining variables. The one yielding
the highest F, in the presence of the first-
selected variable was added to the model.
This pattern continued until there were
no variables with significant F-values.
The cutoff for variables entering the analy-
sis was set at P>0.15, the standard default
value when using stepwise regression as
an exploratory tool.
Cross tables with Mantel-Haenzel χ2-
test (CTMH). Cross-tables analysis be-
tween questions allowed the authors to
trace the patterns among the answers.
The χ2 has 1 degree of freedom where r is
the Pearson correlation between the row
and column variable. The Mantel-Haenzel
χ2-test allowed the authors to test the
alternative hypothesis that there may be

a linear association between the row vari-
able and the column variable.
Gamma test (GT). The estimator of
Gamma is based on concordance and dis-
cordance. Concordance means that in-
creased “category belonging” for one vari-
able corresponds to increased “category
belonging” for another variable. Discor-
dance refers to increased “category be-
longing” for one variable corresponding to
decreased “category belonging” for the
other variable.
RESULTS
Table 2 reveals the results of the prin-
Delivered by Ingenta to: ? IP: 146.185.203.42 on: Wed, 14 Feb 2018 06:35:04
cipal components analysis. There were
several groups of questions with similar
eigenvectors. It appeared that some of
the questions might be repetitive and
unnecessary. However, principal compo-
Copyright (c) PNG Publications. All rights reserved.
nents analysis alone provided insufficient
information to eliminate specific ques-
tions; therefore, we continued with clus-
ter and linear regression analyses.
The results of the principal component
cluster analysis and the centroid compo-
nent cluster analysis are presented in
Tables 3 and 4. The principal component
cluster analysis (Table 3) revealed 6 clus-
ters. The centroid cluster component
analysis (Table 4) identified the identical
6 clusters presented in Table 3. The
authors further explored and worked with
the 6 clusters to search for pairs or groups
of questions providing the same or very
similar information about smokers.
To obtain regression coefficients and
correlations between questions, the au-
thors fit the data into a stepwise multiple
linear regression model. One question at
a time was allowed to be the dependent
variable and the rest independent vari-
ables. Table 5 is presented as a sample of
this procedure. This table demonstrates
how the authors used Question 3 (Q3) as
the dependent variable and subsequently
the other 17 questions in this regression
model. Table 6 notes that the regression
analyses identified 8 questions with sig-
nificant F-values. Table 7 reveals regres-
sions with different questions as the de-
pendent variable.
To observe patterns in the 2032 smok-
ing behavior responses, cross tables were
created between the questions from the
linear regression model, the cluster analy-
sis, and the principal components analy-
sis. By selecting these data and knowing
the answer (response) to a question, the
™ 2005;29(5):443-455
Am J Health Behav.™
Glover et al
Table 2
Results of the Principal
Components Analysis
Eigenvectors for
Question Principal Component Number 1
1 0.21
2 0.26
3 0.25
9 0.23
10 0.21
11 0.21
12 0.22
13 0.22
14 0.24
15 0.21
18 0.25
4 0.17
5 0.16
16 0.18
6 0.28
7 0.31
8 0.30
17 0.27
cross tables predicted the potential re-
sponse to another question. The cross
table in Table 8 between Question 2 and
Question 3 was constructed with both a
Mantel-Haenzel χ2 test and a Gamma test.
This table shows a significant P-value
(0.001) and thereby a linear association
between Question 2 and Question 3. The
Gamma test in Table 8 indicates a Gamma
value at 0.664 with an asymptotic stan-
dard error of 0.017. In this case, the confi-
dence intervals (0.631 and 0.697) show
that the questions are not independent.
At least 9 other cross tabulations re-
vealed connections between questions;
more specifically, the Mantel-Haenzel χ2
test revealed linear associations. Due to
space limitations, these data are pre-
sented as cross-table results in Table 9.
Removing 7 Questions
Using the cluster analysis, cross tables,
multiple linear regression, principal com-
ponents analysis, Mantel Haenzel χ2 –
test, and the Gamma test, the authors
removed specific questions in the ques-
tionnaire without diminishing the data
obtained.
447
 Smoking Behavioral Questionnaire

Table 3
Oblique Principal Component Cluster Analysis
(Summary for 6 Clusters)
Variation Proportion Second
Cluster Members explaineda explainedb eigenvaluec
1 5 2.00817 0.4016 0.8804
2 2 1.42360 0.7118 0.5764
3 3 1.72336 0.5745 0.8791
4 3 1.65799 0.5527 0.7400
5 2 1.30052 0.6503 0.6995
6 3 1.83584 0.6119 0.7645
Delivered by Ingenta to: ? IP: 146.185.203.42 on: Wed, 14 Feb 2018 06:35:04

Total Variation Explained = 9.949479;d Proportion = 0.5527e

R-squared With
Own Next
Variable Clusterf Closest g
Copyright (c) PNG Publications. All rights reserved.

Cluster 1
Question 6 0.4492 0.1488
Question 10 0.3900 0.0768
Question 14 0.2427 0.1520
Question 17 0.4473 0.1137
Question 18 0.4789 0.0856
Cluster 2
Question 4 0.7118 0.0699
Question 5 0.7118 0.1207
Cluster 3
Question 2 0.7365 0.0873
Question 3 0.7399 0.0866
Question 15 0.2469 0.0919
Cluster 4
Question 9 0.4803 0.0623
Question 12 0.6156 0.0538
Question 13 0.5621 0.1688
Cluster 5
Question 11 0.6503 0.0876
Question 16 0.6503 0.0680
Cluster 6
Question 1 0.4059 0.0774
Question 7 0.7293 0.1792
Question 8 0.7006 0.1870

Note.
a Variation explained by own cluster.
b The result of dividing the variation explained with the number of cluster members.
c The second largest eigenvalue of the cluster.
d The sum across clusters of the variation explained by each cluster.
e The total explained variation divided by the total sum of members.
f Squared correlation of the variable within its own cluster.
g The next highest squared correlation with a cluster.

Table 10 displays the removed ques- to questions that provide the same or very
tions, the replacement questions, and similar information as the questions re-
the correlations. “Replacement ques- moved, thereby making removal statisti-
tions” do not refer to “new” questions, but cally acceptable.

448
 Glover et al

Table 4
Oblique Centroid Component Cluster Analysis
(Summary for 6 Clusters)
Variation Proportion
Cluster Members explaineda explainedb
1 3 1.81876 0.6063
2 2 1.42360 0.7118
3 4 1.97704 0.4943
4 2 1.30052 0.6503
5 4 1.86767 0.4669
6 3 1.65551 0.5518
Delivered by Ingenta to: ? IP: 146.185.203.42 on: Wed, 14 Feb 2018 06:35:04

Total Variation Explained = 10.04309;c Proportion = 0.5579d

R-squared With
Own Next
Variable Clustere Closestf
Copyright (c) PNG Publications. All rights reserved.

Cluster 1
Question 1 0.4866 0.0753
Question 7 0.6844 0.1708
Question 8 0.6577 0.1831
Cluster 2
Question 4 0.7118 0.0761
Question 5 0.7118 0.1187
Cluster 3
Question 2 0.5590 0.0712
Question 3 0.5550 0.0666
Question 14 0.4727 0.0749
Question 15 0.3994 0.0556
Cluster 4
Question 11 0.6503 0.0882
Question 16 0.6503 0.0688
Cluster 5
Question 6 0.4694 0.1420
Question 10 0.4370 0.0754
Question 17 0.4551 0.1112
Question 18 0.5076 0.0830
Cluster 6
Question 9 0.5111 0.0620
Question 12 0.5908 0.0538
Question 13 0.5550 0.1688

Note.
a Variation explained by own cluster.
b The result of dividing the variation explained with the number of cluster members.
c The sum across clusters of the variation explained by each cluster.
d The total explained variation divided by the total sum of members.
e Squared correlation of the variable within its own cluster.
f The next highest squared correlation with a cluster.

Questions 3, 4, 7, 12, 15, 16, and 18
were removed from the original question-
naire (Table 1). The end variable was
calculated through the linear regression
model where Question 3 corresponded to
0.004 + 0.6 for Question 2, where 0.04 is
equivalent to the intercept and 0.6 is
equivalent to the regression coefficient.
Therefore, Question 2 received the weight
of 0.6 from Question 3. Question 2 had

™ 2005;29(5):443-455
Am J Health Behav.™ 449
 Smoking Behavioral Questionnaire

Table 5
Stepwise Procedure for Dependent Variable Question 3 (Q3)
Step 1 Variable Question 2 Entered R-square = 0.36287009

DF Sum of Squares Mean Square F Prob>F

Regression 1 1198.96619275 1198.96619275 1145.91 0.0001
Error 2012 2105.15346962 1.04629894
Total 2013 3304.11966236

Parameter Standard Type II

Variable Estimate Error Sum of Squares F Prob>F
INTERCEP 0.03916660 0.03867441 1.07309990 1.03 0.3113
Delivered by Ingenta to: ? IP: 146.185.203.42 on: Wed, 14 Feb 2018 06:35:04

Question 2 0.58697120 0.01733969 1198.96619275 1145.91 0.0001

Table 6
Copyright (c) PNG Publications. All rights reserved.

Summary of Stepwise Procedure for Dependent Variable Question 3 (Q3)

Variable Number Cumulative
Step Entered In R2 R2 F Prob>F
1 Question 2 1 0.3629 0.3629 1145.9117 0.0001
2 Question 6 2 0.0151 0.3779 48.7280 0.0001
3 Question 14 3 0.0083 0.3862 27.1366 0.0001
4 Question 8 4 0.0043 0.3906 14.3240 0.0002
5 Question 4 5 0.0018 0.3923 5.8033 0.0161
6 Question 15 6 0.0017 0.3940 5.5161 0.0189
7 Question 10 7 0.0010 0.3950 3.3254 0.0684
8 Question 9 8 0.0009 0.3959 2.9749 0.0847

Note.
All variables left in the model are significant at the 0.15 level. No other variable met the 0.15 level of
significance for entry into the model.

Table 7
Results of Regression Analyses With Different Questions
as the Dependent Variable
Independent
Dependent Variable/ Regression
Variable Variables Coefficient Correlation

3 2 0.6 0.60
4 5 0.4 0.42
7 8 0.5 0.60
12 9 0.3 0.45
13 0.3 0.45
15 14 0.3 0.36
16 11 0.4 0.30
18 6 0.2 0.44
10 0.2 0.44
17 0.2 0.44

450
 Glover et al

Table 8
Cross Tabulation Between Question 2 and Question 3
Q u e s t i on 3
Frequency
Percent
Row Percent
Column Percent 0 1 2 3 4 Total

0 433 18 9 3 2 465
21.31 0.89 0.44 0.15 0.10 22.88
Q 93.12 3.87 1.94 0.65 0.43
44.18 4.80 3.06 1.09 1.87
Delivered by Ingenta to: ? IP: 146.185.203.42 on: Wed, 14 Feb 2018 06:35:04

u
1 201 185 11 3 1 401
e 9.89 9.10 0.54 0.15 0.05 19.73
50.12 46.13 2.74 0.75 0.25
s 20.51 49.33 3.74 1.09 0.93
Copyright (c) PNG Publications. All rights reserved.

t 2 155 83 185 19 2 444

7.63 4.08 9.10 0.94 0.10 21.85
i 34.91 18.69 41.67 4.28 0.45
15.82 22.13 62.93 6.88 1.87
o
3 147 72 64 230 8 521
n 7.23 3.54 3.15 11.32 0.39 25.64
28.21 13.82 12.28 44.15 1.54
15.00 19.20 21.77 83.33 7.48
2
4 44 17 25 21 94 201
2.17 0.84 1.23 1.03 4.63 9.89
21.89 8.46 12.44 10.45 46.77
4.49 4.53 8.50 7.61 87.85

Total 980 375 294 276 107 2032

48.23 18.45 14.47 13.58 5.27 100.00

Note.
Frequency missing = 199, which is the number of respondents who did not respond to both questions.

Asymptotic
Test Df Value P-value Standard Error

Mantel-Haenzel χ2 1 735.380 0.001

Gamma 0.664 0.017

an original weight at 1.0, and with the end
variable, we obtained the value 0.04 + 1.6
for Question 2. Equivalent calculations
were conducted for all questions.
Methods for Evaluating Questions
Essentially there were 2 distinct meth-
ods by which we evaluated the question-
naire and questions. Both were deter-
mined to be acceptable.
Method 1. In this method, each ques-
tion was given the value/weight of 1.0.
This value/weight is the capacity of the
question to identify whether a person has
a behavioral dependence on smoking.
Seven questions were removed because
the same or very similar information
could be obtained from another question
or series of questions; therefore, the re-
maining 11 questions carried a new

™ 2005;29(5):443-455
Am J Health Behav.™ 451
 Smoking Behavioral Questionnaire

Table 9
Linear Associations Between Questions

Asymptotic
Questions Test df Value P-value Standard Error

11 and 16 Mantel-Haenzel χ2 1 181.663 0.001

Gamma 0.382 0.026
4 and 5 Mantel-Haenzel χ2 1 369.8 0.001
Gamma 0.490 0.022
7 and 8 Mantel-Haenzel χ2 1 725.04 0.001
Gamma 0.678 0.016
9 and 12 Mantel-Haenzel χ2 1 213.00 0.001
Delivered by Ingenta to: ? IP: 146.185.203.42 on: Wed, 14 Feb 2018 06:35:04

Gamma 0.490 0.026

13 and 12 Mantel-Haenzel χ2 1 302.79 0.001
Gamma 0.481 0.023
14 and 15 Mantel-Haenzel χ2 1 261.10 0.001
Gamma 0.431 0.023
Copyright (c) PNG Publications. All rights reserved.

6 and 18 Mantel-Haenzel χ2 1 220.95 0.001

Gamma 0.407 0.025
10 and 18 Mantel-Haenzel χ2 1 175.21 0.001
Gamma 0.382 0.027
17 and 18 Mantel-Haenzel χ2 1 223.16 0.001
Gamma 0.380 0.024

weight. The weight from the removed Question 6 : 1.2

question is equivalent to the regression Question 8 : 1.5
coefficient between the replacement ques- Question 9 : 1.3
tion and the removed question. For ex- Question 10 : 1.2
ample, in addition to a weight of (1.0), Question 11 : 1.4
Question 2 receives the weight (0.6) from Question 13 : 1.3
the removed Question 3, resulting in a Question 14 : 1.3
new weight of 1.6. Based on Method 1, the Question 17 : 1.2
new weights for the GN-SBQ are as follows:
Method 2. The original questionnaire
Question 1 : 1.0 (remained unchanged) contained questions that obtained the
Question 2 : 1.6 same or very similar information. These
Question 5 : 1.4 duplicate questions were removed, and
only the distinct questions with no con-
nection to each other were retained.
Therefore, the new and final GN-SBQ
Table 10 contains questions with equal capacity to
Questions Removed identify a smoker’s behavioral depen-
From the GN-SBQ dence; moreover, all questions in Method
2 carry the equal value/weight of 1.0. As
Question Replacement Correlation the choice of method used to reduce the
number of questions will not affect inter-
3 2 0.60 pretation of future studies, Method 2 was
4 5 0.42 utilized to create the final questionnaire.
7 8 0.60
12 9,13 0.45 DISCUSSION
15 14 0.36 So, what does this all mean? Each
16 11 0.30 question was constructed such that a
18 6,10,17 0.44 higher scores represented higher behav-
ioral dependence; therefore, the same

452
 Glover et al

Table 11
Glover-Nilsson Smoking Behavioral Questionnaire (GN-SBQ)
Please indicate your choice by circling the number that best reflects your choice.
0=Not at all; 1=Somewhat; 2=Moderately so; 3=Very much so; 4=Extremely so

How much do you value the following (Specific to Questions 1-2).

1. My cigarette habit is very important to me. 0 1 2 3 4

2. I handle and manipulate my cigarette as part of the ritual of smoking. 0 1 2 3 4

Please indicate your choice by circling the number that best reflects your choice.
(Specific to Questions 3-11).
Delivered by Ingenta to: ? IP: 146.185.203.42 on: Wed, 14 Feb 2018 06:35:04

0=Never; 1=Seldom; 2=Sometimes; 3=Often; 4=Always

3. Do you place something in your mouth to distract you from smoking? 0 1 2 3 4

4. Do you reward yourself with a cigarette after accomplishing a task? 0 1 2 3 4

Copyright (c) PNG Publications. All rights reserved.

5. If you find yourself without cigarettes, will you have difficulties in concentrating
before attempting a task? 0 1 2 3 4

6. If you are not allowed to smoke in certain places, do you then play with your
cigarette pack or a cigarette? 0 1 2 3 4

7. Do certain environmental cues trigger your smoking, eg, favorite chair, sofa, room,
car, or drinking alcohol? 0 1 2 3 4

8. Do you find yourself lighting up a cigarette routinely (without craving)? 0 1 2 3 4

9. Do you find yourself placing an unlit cigarette or other objects (pen, toothpick,
chewing gum, etc) in your mouth and sucking to get relief from stress, tension or
frustration, etc.)? 0 1 2 3 4

10. Does part of your enjoyment of smoking come from the steps (ritual) you take
when lighting up? 0 1 2 3 4

11. When you are alone in a restaurant, bus terminal, party, etc, do you feel safe,
secure, or more confident if you are holding a cigarette? 0 1 2 3 4

TOTAL _______

A high numerical response indicated a high behavioral dependence, and the lower numerical
response indicated a lower behavioral dependence.

Scoring for Behavioral Dependence

<12 Mild
12-22 Moderate
23-33 Strong
>33 Very Strong

can be said for the questionnaire as a
whole – the higher the sum score, the
higher the behavioral dependence. The
distribution of scores at baseline for the 8
studies used in these analyses is the
estimate of the distribution scores for the
normal smoking population and, there-
fore, the best estimate for interpreting
the sum scores.
With the FTQ, 6 is the cutoff between

™ 2005;29(5):443-455
Am J Health Behav.™ 453
 Smoking Behavioral Questionnaire
high and low dependence and is taken to
be the median value for the smoking
population. Adopting the same approach
as the FTQ, the authors separated the
scores into the following behavioral de-
pendence groups: <12 mild, 12-22 moder-
ate, 23-33 strong, >33 very strong. It is
our recommendation that high scores on
the GN-SBQ indicate a need for greater
emphasis by the clinician on behavioral
management and lower scores on the GN-
SBQ suggests a lesser need for emphasis
on behavioral management. Finally, the
GN-SBQ is not intended to guide clini-
cians toward specific types of behavioral
Delivered by Ingenta to: ? IP: 146.185.203.42 on: Wed, 14 Feb 2018 06:35:04
therapy that might be best for the indi-
vidual smoker. Rather the instrument
should serve as a means of estimating
the degree to which behavioral interven-
tion will be helpful.
Copyright (c) PNG Publications. All rights reserved.
Developing a smoking behavior ques-
tionnaire is a difficult process. This pa-
per chronicles the development of the
instrument. There appears to be a strong
interaction between nicotine and behav-
ior. Behavioral dependence may be intri-
cately tied to nicotine dependence; there-
fore, separating the 2 may prove to be too
difficult. For example, Item 5 in the final
GN-SBQ (Table 11) appears to be an addic-
tion question, regarding the need to re-
plenish nicotine. However, this question
was actually selected for its ability to evalu-
ate the extent to which environmental
cues affect the need to light up, not neces-
sarily the need to replenish nicotine.
The FTQ was eventually correlated with
cotinine levels in titration studies. More-
over, an attempt has been made to utilize
FTQ scores with identification of appro-
priate levels of nicotine replacement.
Based on available data to date, the GN-
SBQ can only indicate the level of behav-
ioral dependence of a smoker. Quit rates
have not been correlated with higher vs
lower GN-SBQ scores; however, discus-
sions with investigators who enrolled the
2032 subjects for the study are underway.
We hope to see how the GN-SBQ correlates
with cotinine levels, successful quit rates,
subject assessment of the usefulness of
behavioral interventions, and other se-
lected variables. Because the GN-SBQ has
not been correlated with quit rates, it has
no predictive value at this time.
The intended purpose of this study was
to develop a questionnaire to measure
the degree to which behavior patterns
play a role in continued smoking depen-
454
dence through the development of a
simple, easily administered pencil-and-
paper questionnaire similar to the FTQ/
FTND. That goal has been accomplished.
The 11 item GN-SBQ can be administered
in <60 seconds, well below the upper limit
of <90 seconds we set for the question-
naire at the beginning of the study. The
authors hope that by administering both
the FTQ and the GN-SBQ, clinicians can
better match pharmacological (medica-
tion) and behavioral (counseling) treat-
ments to individual smokers. Table 11
notes the final version of the GN-SBQ.
At the 8th annual meeting of the Soci-
ety for Research on Nicotine and Tobacco
held in Savannah in 2002, a symposium
Reinvigorating Behavioral Therapies for
Smoking Cessation was held.10 The focus of
the symposium was to explore reasons for
the dramatic decline of new innovations
in behavioral therapies. It is well ac-
cepted that beginning with the 1988 Sur-
geon General’s Report5 and the availabil-
ity of pharmacological adjuncts, the em-
phasis shifted from a behavioral empha-
sis to brain changes, genetics, physi-
ological dependence, and other neurobio-
logical explanations to treat and explore
tobacco dependence. Perhaps use of the
GN-SBQ will revitalize research into the
behavioral aspects of treatment for nico-
tine dependence.
The logical progression for future re-
search is to determine how behavioral
patterns are related to continued smok-
ing dependence, and how the GN-SBQ can
be used effectively. It is hoped that the
GN-SBQ will assist physicians, health
care providers, and tobacco intervention-
ists in identifying aspects of smoking
addiction that are behavioral in nature.
The relationship of the GN-SBQ to the
FTQ and the FTND is unknown at this
time. Moreover, the relationship of the
GN-SBQ scores to tobacco quit rates is
also unknown; further research is re-
quired to understand this relationship.
Finally, the following research should be
investigated: (a) modification of the GN-
SBQ for smokeless tobacco use, (b) modi-
fication the GN-SBQ for adolescent use,
and (c) gender11 and special population
differences that all may affect the validity
of the GN-SBQ.
Acknowledgments
The authors wish to thank Agneta
Hjalmarson (Sweden), Scott J. Leischow

(USA), Nina G. Schneider (USA), Cris
Bolliger (Switzerland), Mats Wallstrom
(Sweden), and Carlos Jimenez-Ruiz
(Spain), who administered the question-
naire and provided data from their clini-
cal trials. Also, we acknowledge with
appreciation the work of Birger Persson
(University of Lund, Sweden), who con-
ducted many of the analyses presented.
Finally, we wish to thank the following
individuals who provided valuable sug-
gestions for improving the manuscript:
Myra L. Muramoto MD, MPH (University of
Arizona); Lowell C. Dale, MD (Mayo Clinic);
Joseph Bauer, PhD (Roswell Park Cancer
Delivered by Ingenta to: ? IP: 146.185.203.42 on: Wed, 14 Feb 2018 06:35:04
Institute); Barbara K. Rimer, DrPH (Uni-
versity of North Carolina Chapel Hill);
Mikael Franzon, PhD (Pfizer Consumer
Healthcare); Connie L. Kohler, DrPH
(University of Alabama-Birmingham), and
Copyright (c) PNG Publications. All rights reserved.
Steven Y. Sussman, PhD (University of
Southern California).
All subjects used in this study partici-
pated in clinical trials funded by
Pharmacia AB Consumer Healthcare,
Helsinborg, Sweden. Pharmacia assisted
with the collection of data and supervised
the analyses. However, since the initia-
tion of this study, Pharmacia has been
purchased by Pfizer, Inc. „ Reinvigorating behavioral therapies for smok-
REFERENCES
1.Centers for Disease Control and Prevention.
Annual smoking-attributable mortality, years
of potential life lost, and economic costs -
United States, 1995-1999. MMWR Morb Mortal
Wkly. 2002;51(14):300-303.
2.Department of Health and Human Services,
Public Health Service. The Health Benefits of
Smoking Cessation: A Report of the Surgeon
General. DHHS (CDC) Publication No. 90-
8416, 1990.
™ 2005;29(5):443-455
Am J Health Behav.™
Glover et al
3.Centers for Disease Control Prevention. To-
bacco use among high school students –
United States, 1997. MMWR Morb Mortal Wkly
1998. 1997;47(12):229-233.
4.Centers for Disease Control Prevention. Inci-
dence of initiation of cigarette smoking –
United States, 1965-1996. MMWR Morb Mortal
Wkly. 1998;47(39):837-840.
5.Department of Health and Human Services,
Public Health Service. The Health Conse-
quences of Smoking: Nicotine Addiction. A
Report of the Surgeon General. DHHS (CDC)
Publication No.88-8406. Washington, DC:
Government Printing Office, 1998.
6.Fagerström KO, Schneider NG. Measuring
nicotine dependence: a review of the
Fagerström Tolerance Questionnaire. J Behav
Med. 1989;12:159-18
7.Heatherton TF, Kozlowski LT, Frecker RC, et
al. The Fagerström Test for Nicotine Depen-
dence: a revision of the Fagerström Tolerance
Questionnaire. Br J Addict. 1991;86:1119-1127.
8.National Center for Health Statistics. Ambula-
tory Health Care Data. Physician Office visits.
Available at: http://www.cdc.gov/nchs/
about/major/ahcd/officevisitcharts.htm. Ac-
cessed March 30, 2005.
9.BBC News. GPs demand more time with
patients. Available at: http://news.bbc.co.uk/
1/hi/health/2225316.stm. Accessed March
30, 2005.
10.Hughes JR, Shiffman S, Baker T, et al.
ing cessation. Symposium presented at: An-
nual Meeting of the Society for Research on
Nicotine and Tobacco; February 21, 2002;
Savannah, Georgia. 2002 Proceedings, Soci-
ety for Research on Nicotine and Tobacco,
7600 Terrace Avenue, Ste 203, Middleton, WI
53562-3174.
11.Bohadana A, Nilsson F, Rasmussen T, et al.
Gender differences in quit rates following
smoking cessation with combination nicotine
therapy: influence of baseline smoking behav-
ior. Nicotine & Tobacco Research. 2003;1(5):111-
117.
455