This article was published in an Elsevier journal. The attached copy
is furnished to the author for non-commercial research and
education use, including for instruction at the author’s institution,
sharing with colleagues and providing to institution administration.
Other uses, including reproduction and distribution, or selling or
licensing copies, or posting to personal, institutional or third party
websites are prohibited.
In most cases authors are permitted to post their version of the
article (e.g. in Word or Tex form) to their personal website or
institutional repository. Authors requiring further information
regarding Elsevier’s archiving and manuscript policies are
encouraged to visit:
Author's personal copy
Nonverbal behavior of human addicts: Multimetric analysis
Elena V. Verbitskayaa, Evgeny M. Krupitskyb, Andrey Burakovc,
Marina V. Tsoy-Podoseninab, Valentina Yu. Egorovab,
Natalia Busharab, Olga Yu. Vekovischevad,⁎
aLaboratory of Biomedical Statistics, Scientific Research Center, Pavlov State Medical University, St. Petersburg, Russia
bLaboratory of Clinical Pharmacology of Addictive States, Valdman Institute of Pharmacology,
Pavlov State Medical University, St. Petersburg, Russia
cLeningrad Regional Centre of Addictions, St. Petersburg, Russia
dInstitute of Biomedicine/Pharmacology, Biomedicum Helsinki, University of Helsinki,
P.O. Box 63 (Haartmaninkatu 8), FIN-00014, Helsinki, Finland
Aims: Ethological approach followed by multimetric statistical analysis was applied to characterize and
discriminate alcohol, heroin and dual, alcohol and heroin, dependent subjects.
Design: Heroin, alcohol, and dual dependent patients (n=51) after one month of stabilization of remission and
control volunteers (n=34) without a history of significant drug or alcohol use were interviewed and videotaped
during the interview by approbation. Nonverbal behavioral cues monitored during the interview were analyzed by
means of general linear procedure followed by correlation, factor and discriminant function analyses.
Findings: By using this approach the attempt to discriminate addicted groups between each other failed. Therefore
we found acceptable to combine subjects in one group and to suggest the similarity between alcohol and heroin
dependence. It was found that principal markers of behavioral structure in addicted subjects were higher
responsivity to communicate distance, less expression of affiliation behavioral pattern, low level of correlations
between different behavioral patterns, and unclear factor structure. Behavioral pattern “affiliation” was identified
as discriminate behavior between control and addicted subjects.
Conclusions: Nonverbal cues of human behavior identified clear differences between healthy control and addictive
© 2007 Elsevier Ltd. All rights reserved.
Keywords: Alcoholics; Heroin users; Human nonverbal behavior; Multimetric statistic analysis
Addictive Behaviors 32 (2007) 2260–2267
⁎Corresponding author. Tel.: +358 9 191 25337; fax: +358 9 191 25364.
E-mail addresses: firstname.lastname@example.org (E.V. Verbitskaya), email@example.com (E.M. Krupitsky),
firstname.lastname@example.org (O.Y. Vekovischeva).
0306-4603/$ - see front matter © 2007 Elsevier Ltd. All rights reserved.
Author's personal copy
Ethological, or behavioral, psychiatry is a newly emerging discipline that uses nonverbal behavior cues
in order to improve the assessment of psychopathology (Garb, 2003; Schelde, 2000). A method of
ethological measuring is prevailing in experimental psychopharmacology and might provide psychiatry
with a theoretical framework to integrate animal and human data (Troisi, 1999).
Nowadays the combination of addictive diseases such as alcoholism and drug addiction appears to be
common in human population, and, probably, have common pathways implicating the role of the
endogenous opioid system in the development and maintenance of alcoholism (Modesto-Lowe & Fritz,
2005; Oswald & Wand, 2004). Thus, we compared the nonverbal behaviors of alcoholics, heroin and dual
alcohol–heroin addicts between each other as well as with control subjects. Predictable discriminations
should make clear whether alcohol and heroin addictions have the similar behavioral parameters or not
and might be identified during interview procedure. Multimetric analysis with correlation, factor and
discriminant function analyses were employed to achieve a more complete assessment of nonverbal
behavior (Leighty et al., 2004).
2.1. Subjects and setting
The study was conducted at the St. Petersburg Pavlov State Medical University (PMU) and the
Leningrad Regional Center of Addictions (LRCA), located in St. Petersburg and Leningrad Region,
Russia, respectively. Both sites were trained for identity that showed a nonspecific effect on the results of
the interview (Krupitsky et al., 2004). Drug addicted patients (n=51) (alcohol (age of 23–40), heroin (age
of 22–36) and dual heroin+alcohol (age of 22–35) dependent patients) were tested at the LRCA; control
subjects (n=34) were tested at the PMU. Participants were assessed after one month of detoxification and
remission stabilization, during the short period before discharge when they had been weaned from
medications (only medications for somatic disorders had not been canceled). Axis I psychiatric disorders
were considered as exclusion criteria.
Hospitalized 18 alcohol, 11 heroin and 22 dual addicts were interviewed and videotaped during the
interview as per their approval. Control subjects without a history of significant drug or alcohol use were
recruited from the community surrounding PMU campus. The control group included factory workers,
college students, and unemployed individuals that never met the criteria of harmful alcohol or drugs of
abuse use. They received the same screening and testing batteries administered to the addicted patients
(described below) and were subjected to the same exclusionary criteria. Experiments were approved by
the Ethical Committee of Pavlov State Medical University.
2.2. Design and procedures
Screening procedure: Screening of all subjects prior to participation was performed at two levels. First,
for patients, a record check of the LRCA's current files containing assessments whether the patient was
eligible in terms of current psychiatric status (no dual Axis I diagnoses) using the ICD-X, drug abuse
history, age, head injury (none severe), and IQ (mental retardation [b80 estimated IQ using the Raven's
Progressive Colored Matrices (CPM, Raven et al., 1998)] was exclusionary). Second, recruited patients
2261E.V. Verbitskaya et al. / Addictive Behaviors 32 (2007) 2260–2267
Author's personal copy
Definitions of behavior categories and elements
Behavior categories and elements Definitions of behavior elements
Prosocial AffiliationLook at
Head to side The head is tilted to one side
Bob A sharp upwards movement of the head, rather like an inverted nod
Flash A quick raising and lowering of the eyebrows
Raise The eyebrows are raised and kept up for some time
SmileThe lip corners are drawn back and up
Submission Nod The normal affirmative gesture
Lips in The lips are drawn slightly in and pressed together
Look away Looking away from the interviewer
Look down Looking down at feet, lap or floor
ShutThe eyes are closed
Chin The chin is drawn in towards the chest
Crouch The body is bent right forward till the head is near the knees
Still A sudden cessation of movement, a freezing
Shake The normal negative gesture
Thrust A sharp forward movement of the head towards the interviewer
Frown The eyebrows are drawn together and lowered at the center
ShrugThe shoulders are raised and dropped again
Small mouth The lip corners are brought towards each other so that the mouth looks small
Wrinkle A wrinkling of the skin on the bridge of the nose
DisplacementGroom The fingers are passed through the hair in a combing movement
Hand–face Hand(s) in contact with the face
Scratch The fingernails are used to scratch part of the body, frequently the head
Yawn The mouth opens widely, roundly and fairly slowly, closing more swiftly. Mouth
movement is accompanied by a deep breath and often closing of the eyes and lowering of
Fumble Twisting and fiddling finger movements, with wedding ring, handkerchief, other hand,
Twist mouth The lips are closed, pushed forward and twisted to one side
Lick lips The tongue is passed over the lips
Bite lips One lip, usually the lower, is drawn into the mouth and held between the teeth
Relaxation Relax An obvious loosening of muscle tension so that the whole body relaxes in the chair
Settle Adjusting movement into a more comfortable posture in the chair
Fold arms The arms are folded across the chest
Laugh The mouth corners are drawn up and out, remaining pointed, the lips parting to reveal
some of the upper and lower teeth
Neutral face A face without expression and without particular muscular tension. It is the basic awake
Looking at the interviewer
The corners of the mouth are drawn back but not raised as in smile
Leaning forward from the hips towards the interviewer
Hand(s) in contact with the mouth
2262 E.V. Verbitskaya et al. / Addictive Behaviors 32 (2007) 2260–2267
Author's personal copy
were administered the Addiction Severity Index (ASI) (McLellan et al., 1989), the Risk Assessment
Battery (RAB) (Navaline et al., 1994) and the Time Line Follow-Back (TLFB) instrument, during which
they were extensively probed about their prior drug use experience (Sobell, Toneatto, Sobell, Leo, &
Johnson, 1992). Potential control subjects provided consent for a similar screening process. Behavioral
testing: During the interview participants were videorecorded for 6–7 min and observed subsequently
(sound off) using a computer-assisted data acquisition system (Ethograph, 2.06, Ritec, St. Petersburg,
Russia) (Nyberg, Vekovischeva, & Sandnabba, 2003; Poshivalov, Suchotina, & Verbitzkaya, 1988;
Vekovischeva et al., 2004) to score consecution, duration and frequency of glances and attendant head,
mouth and eyebrow movements and postures (see Table 1). Each behavioral element was characterized
using medial duration (MD) and relative frequency (RF), the parameters that do not depend on each other
or experiment duration.
Statistical analysis was conducted using SPSS v.12.0 statistical package. Multivariate analysis of
variance and analysis of covariance (for age-specific factor) with following Scheffe or Bonferroni (for
homogeny variances) or Games–Howell (for non homogeny variances) post-hoc tests were applied.
Correlation analysis: Nonparametric Bivariate Correlations procedure (Spearman's rho) identified
correlations between each pair of behavioral elements. Highest level of correlations (from 0.6 to 1.0) was
chosen as significant. Factor analysis, a data reduction method to identify underlying relationships
among variances (behavioral elements), was used to organize the variables into the composite entities
called “factors”. Mean durations and relative frequencies of the variables were analyzed based on
principal components extraction method with Equamax rotation and Kaiser normalization (Child, 1971).
We produced factor analysis for both the whole set of subjects (used for following discriminant function
analysis) as well as for controls and for addicts separately. Factors were described by elements with
loadings more than 0.5. Discriminant function analysis, a multivariate statistical technique (based on
Mahalanobis distance and Stepwise-Forward procedure) which is typically used to distinguish between
predefined groups, to identify discriminate variables, as well as to determine group membership of the
unclassified individuals (Pokorny, Shull, & Nicholson, 1999; Stip, Lussier, Ngan, Mendrek, & Liddle,
1999). This technique was applied for two types of discriminant functions: based on the whole set of
relative frequencies and mean durations of behavioral items and based on the factors scoring previously
for the whole set of subjects by factor analysis. We explored whether there is significant discriminability
between all four subsets (control subjects, alcohol, heroin and dual addicts) as well as between two groups
(controls and addicts).
3. Results and discussion
Descriptive analysis. Results of Raven IQ test were lower in alcohol and dual addicts than in controls
(F(3,44)=2.876, p=0.048), whereas marital status and level of education were not significantly different
between any addictive and control subjects. Employment among dual patients was minimal by Fisher's
exact test (pb0.05). Test results suggested that due to high personal variability among addicts we cannot
identify specificity of abusers by demographical characteristics only.
According to self-report, alcoholic and dual patients had higher alcohol consumption than other
subjects (F(3,44)=4.161, p=0.012). Also, in dual patients first alcohol exposure happened earlier than in
controls (F(3,44)=5.268, p=0.004) and the duration of heroin dependence and number of overdoses
were higher than in heroin patients (F(2,28)=5.105, p=0.013 and F(1,26)=7,112, p=0.013
respectively). Data suggests that it is possible to search for behavioral specificity especially in dual
2263 E.V. Verbitskaya et al. / Addictive Behaviors 32 (2007) 2260–2267
Author's personal copy
patients. Analysis of variance. Analysis of covariance (age was calculated as covariate and was 29.2)
showed no significant differences in the mean durations and the relative frequencies of variables between
three addiction groups. Controls demonstrated higher level of “look at” (FRF(3,57)=18.914, pb0.0001;
FMD(3,57)=12.807) and “smile” (FMD(3,57)=12.807, pb0.0001) (Affiliation) as compared with other
addictive groups, “laugh” (Relaxation) as compared with heroin and dual addicts (FRF(3,57)=3.05,
p=0.024; FMD(3,57)=2.967, p=0.027), and “head to side” (Affiliation) as compared with dual addicts
only (FRF(3,57)=4.859, p=0.002; FMD(3,57)=3,49, p=0.024). It seems that Affiliation reflecting a
positive attitude towards an interlocutor and engagement into social interactions (Troisi, 1999), was less
expressed in addicts. Lower level of “look at” in all addicts could be a result of their higher responsivity to
close communicative distance (about 1.8 m) as it has been observed in healthy patients for situations
shorter than 0.8 m (Dixon & Fisch, 1998).
Absence of differences between addictive groups allowed us to combine them all together in
“addiction” group that represented increased spectrum of differences. The mean durations as well as the
relative frequencies of Affiliation (FRF(1,59)=24.071, pb0.001; FMD(1,59)=88.092, pb0.001) based
on “smile”, “look at”, “head to side” and Relaxation (“laugh”; FRF(1,59)=5.76, p=0.005; FMD(1,59)=
6.26, p=0.003) appeared again significantly higher in controls which indicated an emotional arousal in
the control group (Troisi, 1999). Additionally findings such as Assertion, signals of low-level aggression
and hostility (“small mouth”; FMD(1,59)=4.47, p=0.016), and Displacement, concentration on one's
own body (“lick lips”; FRF(1,59)=3.47, p=0.037), were significantly increased in the addiction group.
The relative frequencies of Flight, serving cut-off behavior (“look down”, “shut”; FRF(1,59)=3.60, 0.034;
FRF(1,59)=3.52, 0.036 respectively), were higher in control but the mean duration of Flight (“look
down”; FMD(1,59)=7.21, p=0.002) was longer in addiction group which represents another clear effect
of short communicative distance for addicts (Dixon & Fisch, 1998). In control subjects the combinationof
Flight elements with “smile”, ambivalent positive signal, might be a response to unpleasant, “shame”
questionsduring interview (Dixon &Fisch, 1998).In addicts,effect of shamequestions couldbe reasoned
by increased relative frequency of “lick lips”, since the element is similar to “twist-mouth”, and also
related to Displacement pattern, that was marked in healthy subjects by unpleasant questions (Dixon &
Fisch, 1998). The combination of both increased Assertion and Flight elements might be the expression of
hostility and avoidance behavior that was clearly manifested in abusers (Troisi, 1999).
Despite of resurfaced behavioral differences between control and pooled addiction groups, the
increased responsivity of abusers to communicative distance and their intention to avoid the interview
appeared as the main discriminative behavioral markers that could be implied by the use of additional
statistical methods. Correlation analysis. High correlations between different variances and between
especially behavioral categories were found for the control group only. Shown plasticity of control
behavior indirectly proves inner abnormalities of behavioral structure of abusers. Thus, positive
correlations between Assertion and Displacement (“small mouth”–“bite lips, 0.62) or Flight (“look
down”–”head-shake”, 0.70) might reflect two logical tendencies of the hostility behavior to replace the
reaction or to avoid the situation. Creative behavior such as Submission and Displacement inverse
correlated with Relaxation (“mouth corners back”–”neutral face”, −0.72 and “twist mouth”–”neutral
face”, −0.62, respectively) which opposed low emotional level (“neutral face”) (Troisi, 1999) to
motivated behavior. But positive correlation between Relaxation and Flight (“neutral face”–“look away”,
0.61)suggeststhat flight behavior does notrequire the alteration of emotional status. However, despite the
fact that pair correlations identified could provide the link between two elements, causal-effect relations
between them need further investigation.
2264E.V. Verbitskaya et al. / Addictive Behaviors 32 (2007) 2260–2267
Author's personal copy
Factor analysis. Factor analysis for control subjects and pooled addiction groups is presented in the
Table 2. Behavior structure of control group was based on two principal factors that explain 33.6% of
variances. Factor 1 was loaded mostly by the relative frequencies of elements from Affiliation pattern and
single elements from other categories excepting Assertion that loaded partly Factor 2. The presence of
both hostile and social positive behaviors in Factor structure of healthy subjects might reflect two-
component motivation of their behavior. On the one hand they are open to communicate but some
parameters of this conversation indicate hostility.
Behavioral structure of addicted group consisted of one principal factor that explains only 27.3% of
variances and was loaded mostly by elements of Displacement, Submission and Affiliation. The presence
of Assertion, Flight and Relaxation also might be described as blurry directed behavior. It suggests a
disruption of motivation although aggregation of different groups might have an effect. It will be
investigated in future studies when the number of addicts with different backgrounds will be increased.
Principal factors for control and addiction groups separatelya
Variance explained 18.4%15.2%27.3%
Head to side
Mouth corners back
Mouth corners back
Notes:aExtraction Method: Principal Component Analysis; MD — mean duration; RF — relative frequency.
2265 E.V. Verbitskaya et al. / Addictive Behaviors 32 (2007) 2260–2267
Author's personal copy
Discriminant function analysis. Discriminant function analysis built on factors (13 factors were
revealed for all pooled subjects) or behavioral elements, failed to classify significantly 4 groups (only
61.0% of subjects was classified correctly) but discriminated them successfully into two groups (control
and dual). Discrimination between heroin, alcohol and dual groups failed also. It is possible that the
number of subjects in each addiction group was insufficient to ensure the correct discrimination or that
nonverbal manifestations are similar in addicts of different origin which proves the main idea about
all addicts into one group. Less variability between addicts shown in Fig. 1 for the combined group argues
in favor of similarity despite insufficient numbers. Discrimination between two groups (control and
addiction) based on behavioral elements was successful: 90.6% of original grouped cases and 87.1% of
cross-validated grouped cases. Main discriminant elements appeared Rise (RF), Look at (MD) and Smile
(MD) with coefficients −0.785; 0.694; 0.668 respectively which were related to the Affiliation category.
Discrimination based on the Factors (89.4% of original grouped cases, 87.1% of cross-validated
grouped cases) identified main factors each containing any element of Affiliation: Factor 8 (0.990; highly
loaded by “smile” and “laugh”), Factor 5 (0.470; highly loaded by “laugh”, “head to side” and “shut”) and
Factor 11 (0.440; highly loaded by “frown” and “look at”).
Ethological approach to analyze nonverbal cues of human behavior identified clear differences
between healthy control and addicted patients but not between alcohol, heroin and dual alcohol–heroin
dependent subjects. It might be discussed as the similarity of behavioral expressions in all addicts and,
therefore, homogeneity of the disorders.
Multimetric analysis clarified differences between control and addicted subjects in different aspects
that might be proposed as an algorithm to analyze a set of numerous behavioral variances. Thus, analysis
of covariance identified responsivity of addicted subjects to communicative distance and their weak
affiliation expressions, correlation analysis found less plasticity of their behavioral structure, factor
Fig. 1. Discriminant analysis for control and addicted groups based on MDs and RFs of behavioral elements (A) or based on
factors found for all subjects (B).
2266 E.V. Verbitskaya et al. / Addictive Behaviors 32 (2007) 2260–2267
Author's personal copy Download full-text
analysis has shown diffusiveness of their motivations, and discriminant function analysis identified main
discriminate behavioral elements composing Affiliation pattern. Therefore affiliation behavior could be
used as a marker that might be identified during interview, although the issue concerning cultural or
diagnostic specificity remains open and should be cleared in future studies.
The authors wish to acknowledge the patients and staff of Leningrad Regional Center of Addiction and
volunteers and staff of Laboratory of Clinical Pharmacology of Addiction (Pavlov Medical University, St.
Petersburg, Russia) for their assistance in this research. The study was conducted without external
funding. Dr. Vekovischeva was supported by Finnish Academy fellowship.
Child, D. (1971). Essentials of factor analysis. London: Holt, Rinehart and Winston.
Dixon, A. K., & Fisch, H. U. (1998). Animal models and ethological strategies for early drug-testing in humans. Neuroscience
and Biobehavioral Reviews, 23, 345−358.
Garb, H. N. (2003). Incremental validity and the assessment of psychopathology in adults. Psychological Assessment, 15,
Krupitsky, E. M., Zvartau, E. E., Masalov, D. V., Tsoi, M. V., Burakov, A. M., Egorova, V. Y., et al. (2004). Naltrexone for heroin
dependence treatment in St. Petersburg, Russia. Journal of Substance Abuse Treatment, 26, 285−294.
Leighty, R. E., Nilsson, L. N., Potter, H., Costa, D. A., Low, M. A., Bales, K. R., et al. (2004). Use of multimetric statistical
analysis to characterize and discriminate between the performance of four Alzheimer's transgenic mouse lines differing in
Abeta deposition. Behavioural Brain Research, 153, 107−121.
McLellan, B. A., Koch, J. P., Wortzman, D., Rogers, C., Szalai, J., & Williams, D. (1989). Early identification of high-risk
patients using the “estimated” injury severity score and age. Accident Analysis & Prevention, 21, 283−290.
Modesto-Lowe, V., & Fritz, E. M. (2005). The opioidergic–alcohol link: Implications for treatment. CNS Drugs, 19, 693−707.
Navaline, H. A., Snider, E. C., Petro, C. J., Tobin, D., Metzger, D., Alterman, A. I., et al. (1994). Preparations for AIDS vaccine
trials. An automated version of the Risk Assessment Battery (RAB): Enhancing the assessment of risk behaviors. AIDS
Research and Human Retroviruses, 10(Suppl 2), S281−S283.
Nyberg, J. M., Vekovischeva, O., & Sandnabba, K. (2003). Anxiety profiles of mice selectively bred for intermale aggression.
Behavior Genetics, 33, 503−511.
Oswald, L. M., & Wand, G. S. (2004). Opioids and alcoholism. Physiology & Behavior, 81, 339−358.
Pokorny, L., Shull, R. D., & Nicholson, R. A. (1999). Dangerousness and disability as predictors of psychiatric patients' legal
status. Behavioral Sciences & the Law, 17, 253−267.
Poshivalov, V. P., Suchotina, I. A., & Verbitzkaya, E. V. (1988). Ethological pharmacology of anxiogenic and anxiolytic drugs.
Annali dell'Istituto Superiore di Sanità, 24, 453−459.
Schelde, J. T. (2000). Schizophrenic, depressed, and nonpsychiatric medical patients: An ethologically based comparison.
Journal of Nervous and Mental Disease, 188, 377−380.
Sobell, L. C., Toneatto, T., Sobell, M. B., Leo, G. I., & Johnson, L. (1992). Alcohol abusers' perceptions of the accuracy of their
self-reports of drinking: Implications for treatment. Addictive Behaviors, 17, 507−511.
Stip, E., Lussier, I., Ngan, E., Mendrek, A., & Liddle, P. (1999). Discriminant cognitive factors in responder and non-responder
patients with schizophrenia. European Psychiatry, 14, 442−450.
Troisi, A. (1999). Ethological research in clinical psychiatry: The study of nonverbal behavior during interviews. Neuroscience
and Biobehavioral Reviews, 23, 905−913.
Vekovischeva, O. Y., Aitta-Aho, T., Echenko, O., Kankaanpaa, A., Seppala, T., Honkanen, A., et al. (2004). Reduced aggression
in AMPA-type glutamate receptor GluR-A subunit-deficient mice. Genes Brain Behavior, 3, 253−265.
2267E.V. Verbitskaya et al. / Addictive Behaviors 32 (2007) 2260–2267