Behaviour Research and Therapy 42 (2004) 1163–1182

www.elsevier.com/locate/brat

Measurement of compulsive hoarding: savinginventory-revised

Randy O. Frost a,�, Gail Steketee b, Jessica Grisham c

a Department of Psychology, Smith College, Northampton, MA 01063, USAb Boston University School for Social Work, 264 Bay State Road, Boston, MA 02215 USA

c Department of Psychology, Boston University, Boston, MA 02215 USA

Received 26 February 2003; received in revised form 21 July 2003; accepted 23 July 2003

Abstract

Four studies examined a new measure of compulsive hoarding (Saving Inventory-Revised; SI-R). Fac-tor analysis using 139 hoarding participants identified 3 factors: difficulty discarding, excessive clutter,and excessive acquisition. Additional studies were conducted with hoarding participants, OCD parti-cipants without hoarding, community controls and an elderly sample exhibiting a range of hoardingbehavior. Internal consistencies and test–retest reliabilities were good. The SI-R distinguished hoardingparticipants from all other non-hoarding comparison groups. The SI-R showed strong correlations withother indices and methods of measuring hoarding (beliefs, activity dysfunction from clutter, observer rat-ings of clutter in the home) and relatively weaker correlations with non-hoarding measures (positive andnegative affect and OCD symptoms). The SI-R appears to be an appropriate instrument for assessingsymptoms of compulsive hoarding in clinical and non-clinical samples.# 2003 Elsevier Ltd. All rights reserved.

Keywords: Acquisition; Clutter; Discarding; Hoarding; Obsessive–compulsive disorder

1. Introduction

Compulsive hoarding, ‘‘the acquisition of, and failure to discard, possessions which appear to

be useless or of limited value’’ (Frost & Gross, 1993) occurs in 20–30% of OCD patients

(Frost, Krause, & Steketee, 1996). Hoarding has been observed in a number of other disorders

� Corresponding author. Tel.: +1-413-585-3804; fax: +1-413-585-3786.E-mail address: rfrost@email.smith.edu (R.O. Frost).

0005-7967/$ - see front matter # 2003 Elsevier Ltd. All rights reserved.doi:10.1016/j.brat.2003.07.006

as well, including schizophrenia (Stein, Laszlo, Marais, Seedat, & Potocnok, 1997), anorexia(Frankenburg, 1984), organic mental disorders (Greenberg, Witzhum, & Levy, 1990), anddepression (Shafran & Tallis, 1996). Severity of hoarding behaviors range from mild to severeand even life threatening (Frost & Gross, 1993; Frost, Steketee, & Williams, 1999). Althoughrecent research has identified a number of important features of the problem (Frost & Hartl,1996; Frost & Steketee, 1998), to date little emphasis has been placed on the measurement ofhoarding and related symptoms.Until recently, existing research has relied on unsystematic descriptions of hoarding (Frank-

enburg, 1984; Greenberg, 1987) or yes/no responses to 2 items on the Yale Brown ObsessiveCompulsive Inventory checklist (Baer, 1994; Black et al., 1998; Calamari, Wiegartz, & Janeck,1999; Leckman et al., 1997; Summerfeldt, Richter, Antony, & Swinson, 1999). These indices ofhoarding problems have not been validated and are not complete enough for the task of assess-ing this complex problem. In an earlier investigation of compulsive hoarding, we (Frost &Gross, 1993) developed a 22-item Hoarding Scale based on the available, albeit limited, infor-mation about the topic. This measure was subsequently revised to a 24-item scale (Frost et al.,1998). Items assessed agreement with statements reflecting difficulty discarding, emotional reac-tions to discarding, problems with deciding to discard, infrequent use of saved items, concernthat discarded items may be needed in the future, and sentimental attachments to possessions.The Hoarding Scale has excellent internal reliability, and a number of studies have establishedits validity (Frost & Gross, 1993; Frost, Hartl, Christian, & Williams, 1995; Frost, Krause, &Steketee, 1996; Frost et al., 1998; Frost, Steketee, Williams, & Warren, 2000; Frost, Steketee, &Williams, 2002). However, it has significant limitations.The items on the Hoarding Scale were generated early in research on hoarding and conse-

quently do not reflect important symptoms unknown at the time the original scale wasdeveloped. For instance, compulsive acquisition has been found to be an important aspect ofcompulsive hoarding (Frost et al., 1998, 2002), but is not reflected among the original items onthe Hoarding Scale. Second, a number of items on the hoarding scale measure beliefs aboutpossessions (their nature and function) rather than symptoms of the disorder (e.g., ‘‘I see mybelongings as extensions of myself, they are a part of who I am.’’). While we have identified anumber of relevant beliefs, no single set of beliefs characterizes all people with this problem.Rather than including all relevant beliefs in this measure and having a number of them notapply, it seems prudent to separate symptoms of hoarding from beliefs thought to be associatedwith it. Separate assessment of hoarding beliefs may help to explain hoarding symptoms (Ste-ketee, Frost, & Kyrios, 2003). Third, several items on the Hoarding Scale refer to specifictypes of possessions (e.g. papers, containers) which are likely to have relevance to some, but notall people with this problem. Finally, the wording of the items on the existing scale does notcapture the distress or impairment associated with clinically significant pathology. While theoriginal Hoarding Scale has been useful in studying both nonclinical and clinical populations(Frost et al., 1998, 2002), it is likely to be less sensitive at the more severe ranges of the problem.The purpose of the present study was to develop and validate a new measure of compulsive

hoarding in which the items (1) adequately sample the content domain of the prominent symp-toms of compulsive hoarding, (2) reflect symptoms and not beliefs about possessions, (3) do notrefer to specific types of possessions, and (4) adequately capture the levels of distress andimpairment seen in severe hoarding cases.

R.O. Frost et al. / Behaviour Research and Therapy 42 (2004) 1163–11821164

Elsewhere, we (Frost & Hartl, 1996) have defined clinically significant compulsive hoardingas ‘‘(1) the acquisition of and failure to discard a large number of possessions that appear to beuseless or of limited value; (2) living spaces sufficiently cluttered so as to preclude activities forwhich those spaces were designed; and (3) significant distress or impairment in functioningcaused by the hoarding’’ (p. 341). This definition emphasizes 3 key symptoms, as well as the dis-tress and impairment associated with each. The first symptom is excessive acquisition of pos-sessions. Recent research has shown that compulsive acquisition (including both compulsivebuying and the compulsive acquisition of free things) characterizes people who have hoardingproblems (Frost et al., 1998), and that hoarding behavior is significantly elevated in people whomeet criteria for compulsive buying (Frost, Steketee, & Williams, 2002). The second key symp-tom is difficulty discarding. People with compulsive hoarding problems tend to place higher val-ues on possessions and have great difficulty discarding them as a result (Frost, Hartl, Christian,& Williams, 1995). Third, the hallmark of serious compulsive hoarding problems is clutter thatprevents normal use of living spaces (Frost & Hartl, 1996). Although ‘‘clutter that prevents nor-mal use of living spaces’’ can be considered a form of interference, we have given it more prom-inence in the definition because it is ubiquitous in this population. Other forms of interferencevary from person to person. While each of these 3 components exists to some extent in non-clinical populations (Frost & Gross, 1993), the factors that make hoarding clinically significantare the distress and impairment/interference associated with the symptoms.To develop a scale that adequately represented all of the factors that define hoarding, items

from earlier versions of the scale and from pilot testing were generated to measure each of thecomponents of compulsive hoarding outlined above: Problems with acquisition (5 items), dif-ficulty discarding (5 items), clutter (6 items), as well as distress (5 items) and impairment/inter-ference (5 items) associated with each of the 3 substantive symptoms. Thus, the original SavingInventory- Revised (SI-R) contained 26 items in a self-report format that used a Likert-typescale from 0 to 4. Distress and impairment/interference items were modeled after elements ofthe Y-BOCS (Yale–Brown Obsessive Compulsive Scale, Goodman et al., 1989). Subjects wereinstructed to circle the most appropriate response.In the present study we examined the factor structure of the 26-item SI-R among a sample of

people with compulsive hoarding problems in order to identify symptom-relevant subscales(Study 1). Following this, we examined known groups validity of the SI-R by comparing hoard-ing participants with non-hoarding participants in both OCD and non-OCD samples (Studies 2and 3). We also studied the construct validity of the scale in relation to other measures ofhoarding and of other less closely related psychopathology (Studies 2 and 3). Finally, in a sam-ple of elderly people we examined the validity of the SI-R in relation to observer ratings of clut-ter in the home (Study 4). An elderly sample was chosen since hoarding has been identified asan especially difficult problem among the elderly (Steketee, Frost, & Kim, 2001). Four studiesexamining the psychometric properties of the SI-R are described below followed by a discussionof the overall findings.

1.1. Study 1: Factor analysis of the SI-R in a clinical sample

In a previous study, a factor analysis of the 26-item SI-R in a large college student samplerevealed a 4-factor solution that included 3 strong factors corresponding to the 3 symptom

1165R.O. Frost et al. / Behaviour Research and Therapy 42 (2004) 1163–1182

components, and a somewhat weaker factor in which distress and impairment items were com-bined (Coles, Frost, Heimberg, & Steketee, 2003). We sought to determine the scale’s factorstructure in a sample of people who self-identified as having hoarding problems.

1.1.1. Method1.1.1.1. Participants. The factor analysis sample included 139 participants from 4 separatehoarding projects. All subjects suffered from problems with compulsive hoarding and repre-sented a range of severity. Among these, 32 came from a sample of people who reported receiv-ing a diagnosis of OCD and who reported hoarding as one of their symptoms (see Study 2 fordetails), 70 were recruited from among people across the country seeking help from the authorsfor their hoarding problem and from several self-help groups devoted to compulsive hoarding(see Study 3), 24 came from a study of hoarding, trauma and ADHD recruited in a similar fash-ion in the Eastern Massachusetts and Central Connecticut areas (Hartl, Duffany, Allen, Ste-ketee, & Frost, 2003), and 13 subjects were people seeking treatment for hoarding problems at theCenter for Anxiety and Related Disorders at Boston University. Ages ranged from 18 to 75with a mean of 50.7 (s:d: ¼ 11:1). Thirty-six men and 101 women completed the SI-R (2 sub-jects failed to report their gender). Due to missing items, complete SI-R data were availablefor only 131 of these subjects.

1.1.1.2. Measures. The 26-item SI-R described earlier was completed by participants as part ofthe study in which they were involved.

1.1.2. ResultsBecause this was the first study examining the psychometric properties of the SI-R in a clini-

cal sample, there was insufficient empirical basis to merit a confirmatory factor analysis. There-fore, initial analyses were conducted in an exploratory fashion. The data to be submitted tofactor analysis were found to be normal (Bartlett’s Test of Sphericity <0.0001) and haveadequate sampling variance (KMO index >0.5). An exploratory maximum likelihood factoranalysis with Promax rotation and Kaiser normalization initially revealed 4 factors. Twenty-three SI-R items loaded strongly (>0.4) and uniquely (difference >0.2) onto one of the 4 factors.Items 24, 25, and 26 had no significant loading. When these items were dropped from the analy-sis, a 3-factor solution emerged that accounted for 61.2% of the variance. The scree plot alsosuggested a 3-component solution (see Table 1). These 3 factors were similar to those found in anon-clinical population, except no Interference/Distress factor emerged (Coles, Frost, Heim-berg, & Steketee, 2003). The three factors mirrored the 3 hypothesized domains of compulsivehoarding: Clutter (9 items), Difficulty Discarding (7 items), and Acquisition (7 items).Of six items originally selected to represent the clutter dimension, five loaded on a Clutter fac-

tor (items 3, 6, 12, 18, 21), and the sixth (item 24) did not load on any factor. Four additionalitems loaded on the Clutter factor, three representing impairment/interference associated withclutter (items 7, 14, 20) and one describing distress from clutter (item 13). For Difficulty Dis-carding, all five originally selected items loaded on this dimension (items 1, 11, 16, 17, 23), aswell as 2 additional ones from the original distress domain (items 2, 4), both referring to distressassociated with discarding. For Acquisition, 4 of the original 5 items generated for this factorloaded on it (items 8, 9, 10, 15), whereas item 25 did not load on any factor. In addition, two

R.O. Frost et al. / Behaviour Research and Therapy 42 (2004) 1163–11821166

distress items (items 5 and 19) and one impairment item (item 22), all of which related to acqui-sition, loaded on this factor. Thus, distress and impairment/interference items were dividedamong the 3 factors as described above, and one impairment item (item # 26) did not load onany factor.Because the 3-factor solution was consistent with the three observed types of hoarding symp-

toms, analyses in subsequent samples were conducted on the reduced 23-item measure (seeAppendix) and its unit-weighted subscales. Internal consistency of the reduced scale was high(0.92), and alpha coefficients for the three subscales exceeded 0.87 (see Table 2). The three sub-scales were correlated (see Table 2) suggesting some overlap among subscales.The Difficulty Discarding subscale showed a small but significant correlation with age

(r ¼ 0:17, p < 0:05), indicating that older hoarding participants had more difficulty discardingpossessions. The SI-R total score, Clutter and Acquisition subscales were not correlated with

Table 1Results of Maximum Likelihood Factor Analysis of the Saving Inventory-Revised (SI-R) for 136 participants withhoarding problems

Percentage of variance accounted for 1 2 3

37.97 14.77 8.50

SI-R item content Factor 1Clutter

Factor 2 DifficultyDiscarding

Factor 3 Acquisition

18. clutter restricts navigation through home 0.90 �0.27 0.1220. clutter prevents normal activities 0.83 0.00 0.006. clutter in living areas 0.82 0.00 �0.1712. clutter prevents use of home 0.80 0.00 0.0014. clutter prevents inviting people over 0.75 0.00 0.007. clutter interferes with functioning 0.72 0.00 0.003. extent of clutter 0.61 0.16 �0.1621. lack of control over clutter 0.53 0.27 0.0013. clutter causes distress 0.50 0.11 0.231. extent of difficulty throwing things away �0.11 0.83 0.004. avoid discarding because of distress 0.00 0.77 0.002. how distressing is discarding 0.00 0.75 0.0016. strength of urge to save 0.00 0.73 0.0011. frequency of saving things not needed 0.00 0.69 0.0023. inability to discard 0.00 0.66 0.0017. lack of control over urges to save 0.19 0.59 0.008. feel compelled to acquire? 0.00 0.00 0.8810. lack of control over urges to acquire �0.11 0.00 0.875. distress if can’t acquire �0.14 �0.15 0.829. strength of urge to acquire 0.00 0.17 0.7115. frequency of buying unneeded items 0.00 0.26 0.5722. financial difficulties from saving/buying 0.00 0.00 0.5519. distress over acquiring habits 0.23 0.00 0.54

Note: Items 24, 25, and 26 loaded <0.40 on any factor or double loaded and were dropped from the final solution.61.2% of the variance was accounted for by the 3 factor solution. Kaiser–Meyer–Olkin measure of samplingadequacy = 0.89. Bartlett’s test of sphericity chi square ðdf 253Þ ¼ 1795, p < 0:0001.

1167R.O. Frost et al. / Behaviour Research and Therapy 42 (2004) 1163–1182

age (p0s > 0:05). This is somewhat surprising since earlier research has found age-related effects.However, the present sample contained only people suffering from compulsive hoarding whichwould restrict the range of scores on the SI-R.There were no differences between males and females on SI-R total, Difficulty Discarding or

Clutter subscales. However, females had significantly higher Acquisition scores (M ¼ 15:3,s:d: ¼ 5:7) than males (M ¼ 12:5, s:d: ¼ 4:9), tð133Þ ¼ 2:59, p < 0:05.

1.2. Study 2: Comparison of self-identified OCD subjects with and without hoarding symptoms

To examine the validity of the SI-R, a second study compared self-identified hoarding parti-cipants with non-hoarding participants and examined correlations of the SI-R with a measureof hoarding-related beliefs and attitudes. This was part of a larger study that included a seriesof self-report measures and an experimental study of attachment to possessions which isreported elsewhere (Kim, Frost, Steketee, Turkoff, & Hood, 2003).

1.2.1. Method1.2.1.1. Participants. Participants were 66 attendees of the Obsessive Compulsive FoundationConference (OCF). OCF is a consumer-based organization whose conference is designed forOCD sufferers and their families. Fifty-eight of the subjects reported having received an OCDdiagnosis from a mental health professional and were retained for analysis. Thirty-two reportedhoarding as one of their symptoms, 26 did not. Among the participants for whom hoarding wasa symptom, 11 (34%) reported that it was their primary symptom. Other frequent primarysymptoms for the 32 hoarding participants included contamination/washing (N ¼ 6, 19%) andmultiple obsessions and compulsions (N ¼ 7, 22%). Among the non-hoarding OCD parti-cipants, the most frequent primary symptoms were contamination/washing (N ¼ 9, 35%) andmultiple obsessions and compulsions (N ¼ 5, 19%). Eighteen of the participants were male(31%), and 40 were female (69%). Ages ranged from 17 to 71, with a mean of 43.2 (s:d: ¼ 13:1).

1.2.1.2. Measures. All participants completed the SI-R described above. The 23-item total scoreand the 3 subscales corresponding to the 3 factors of the SI-R were used in this study. Subjectsalso completed the Saving Cognitions Inventory (SCI; Steketee et al., 2003), a 24-item scalereflecting beliefs related to possessions. Subjects indicate the extent to which they had eachthought when deciding whether to throw something away; ratings were done on a Likert-typescale range from 1 (not at all) to 7 (very much). Factor analyses of the SCI indicate four

Table 2Reliabilities (diagonals) and intercorrelations among Saving Inventory-Revised (SI-R) subscales

SI-R Total Difficulty discard-ing

Clutter Acquisition

SI-R Total 0.92Discarding 0.80 0.88Clutter 0.82 0.56 0.91Acquisition 0.73 0.45 0.31 0.87

All p0s < 0:001.

R.O. Frost et al. / Behaviour Research and Therapy 42 (2004) 1163–11821168

subscales: Emotional Attachment (10 items, a ¼ 0:96), Memory (5 items, a ¼ 0:82), Control (3items, a ¼ 0:89), and Responsibility (6 items, a ¼ 0:91). The SCI has been found to be a validmeasure of hoarding beliefs and is correlated with hoarding symptoms after controlling for age,other OCD symptoms and other cognitive variables (Steketee et al., 2003).

1.2.1.3. Procedure. Participants were solicited from a table in the conference exhibition area.The study was described to potential participants as one concerning saving behavior and allconference attendees were invited to participate. Participants were given a set of questionnaireswhich included the SI-R and the SCI to complete and return to the investigators. In addition,participants were asked their age, education level, whether they had received an OCD diagnosisfrom a mental health professional, if hoarding was one of their symptoms, and to list their pri-mary symptom.

1.2.2. ResultsHoarding and non-hoarding samples did not differ in the proportion of males (Chi Sq: ¼ 0:8,

ns) or in education level, tð55Þ ¼ 1:34, ns. Consistent with previous research (Frost, Steketee,Williams, & Warren, 2000), however, hoarding participants were significantly older, M ¼ 47:2,s:d: ¼ 11:2, than non-hoarding ones, M ¼ 38:2, s:d: ¼ 13:7, tð56Þ ¼ 2:77, p < 0:01, and the SI-Rtotal score (23 items) was correlated with age (r ¼ 0:33, p < 0:05). Examination of the SI-R sub-scales indicated that age was correlated with Difficulty Discarding (r ¼ 0:37, p < 0:01) and Clut-ter (r ¼ 0:38, p < 0:01), but not with Acquisition (r ¼ 0:08, ns).A t-test revealed that hoarding subjects scored significantly higher on the SI-R total than

non-hoarding subjects, tð55Þ ¼ 8:19, p < 0:001. They also scored significantly higher on each ofthe subscales, Difficulty Discarding, tð55Þ ¼ 8:62, p < 0:001; Clutter, tð55Þ ¼ 6:66, p < 0:001;and Acquiring, tð55Þ ¼ 4:66, p < 0:001. Table 3 contains the means, standard deviations andrange of participants’ scores. Examination of the frequency distribution of SI-R total scores foreach group indicated that no non-hoarding participants scored at or above the mean for thehoarding group (53.7), and no hoarding subjects scored at or below the mean for the non-hoarding group (24.0). Further, only 4 non-hoarding participants scored above one standarddeviation below the mean for the hoarding group (38.8). Only two subjects in the hoardinggroup scored at or below one standard deviation above the mean for the non-hoarding group

Table 3Means, standard deviations and range on the Saving Inventory-Revised (SI-R) and its subscales for hoarding andnon-hoarding OCD participants

Scale Hoarding participants Non-hoarding participants

(N ¼ 32) (N ¼ 26)

Mean s.d. Min. Max. Mean s.d. Min. Max.

SI-R Total 53.7 14.9 29 88 24.0 12.0 5 51DifficultyDiscarding

18.5 4.9 8 28 7.8 4.5 2 19

Clutter 20.9 7.3 7 36 8.1 7.1 0 31Acquisition 14.3 5.6 4 26 8.1 4.1 2 17

1169R.O. Frost et al. / Behaviour Research and Therapy 42 (2004) 1163–1182

(36.0). Thus, despite the fact that for some of the hoarding subjects, hoarding symptoms werenot their primary or even a prominent symptom, there is clear separation of these subject sam-ples on SI-R scores.There were no differences between participants who listed hoarding as their primary symptom

versus those who listed hoarding as a symptom but not their primary one any of the SI-R varia-bles (SI-R totaltð30Þ ¼ 1:02; DDtð30Þ ¼ 1:78; CLtð30Þ ¼ 0:62, ACQtð30Þ ¼ 0:47, all p’s > 0:05).Correlations between the SI-R and its subscales and the Saving Cognitions Inventory and its

subscales were all significant (see Table 4). Correlations between SI-R Acquisition and hoardingbeliefs were somewhat lower (r’s ranged from 0.38 to 0.55; mean ¼ 0:47) than correlationsbetween the other SI-R dimensions and hoarding beliefs (r’s ranged from 0.52 to 0.70;mean ¼ 0:62).

1.3. Study 3: Comparison of hoarding participants and non-clinical controls

This study further examined the validity of the SI-R by comparing self-identified hoardingparticipants to a community control sample and by examining correlations of the SI-R withmeasures of clutter severity in the home, mood state, and OCD symptoms.

1.3.1. Method1.3.1.1. Participants. Participants for this study consisted of 70 people with hoarding problemsand 23 community controls. The hoarding participants were solicited from self-help groups forproblems with clutter and from among people contacting the first author seeking assistance witha hoarding problem. Community controls were solicited by nomination from hoarding parti-cipants of friends and family who did not have hoarding problems, and from newspaper adsasking for volunteers to participate in a psychological study.There were 17 males and 74 females in the sample (two people did not specify their gender).

The mean age was 49.1 (s:d: ¼ 11:3) and ranged from 24 to 72. The mean education level wasnearly 4 years of college (15.6, s:d: ¼ 2:0). Thirty-six of the subjects were married or cohabitat-ing, 31 were single, 18 were separated or divorced, and 4 were widowed. Thirty-three subjectslived alone and 57 lived with at least one other person. Three people did not report their livingsituation.

Table 4Correlations among the Saving Inventory-Revised (SI-R) and its subscales and the Saving Cognitions Inventory(SCI) and its subscales for 58 individuals with OCD

SCI Total Emotionalattachment

Responsibility Control Memory

SI-R total 0.73 0.59 0.70 0.61 0.6 2Difficultydiscarding

0.72 0.57 0.70 0.59 0.65

Clutter 0.67 0.54 0.65 0.61 0.54Acquisition 0.53 0.47 0.49 0.38 0.46

All correlations significant at p < 0:005.

R.O. Frost et al. / Behaviour Research and Therapy 42 (2004) 1163–11821170

1.3.1.2. Measures and procedures. Participants in both groups completed the SI-R and a demo-graphic information sheet that inquired about age, education level, gender, marital status andwhether anyone else lived in the home.The self-report version of the Yale Brown Obsessive Compulsive Scale (YBOCS; Goodman

et al., 1989) was completed. This a 10-item scale assesses the severity of obsessive and compul-sive symptoms on 0–4 rating scales, with scores of 16 or greater generally used to identify clini-cal samples. Both the clinician-rated and self-report versions of the YBOCS have goodreliability and validity (see Taylor, 1998). For the present study, the hoarding obsession andcompulsion items were removed from the Y-BOCS Checklist to ensure that subjects’ ratings ofOCD symptom severity were independent of hoarding symptoms.Subjects also completed the Positive Affect Negative Affect Scale (PANAS; Watson, Clark, &

Tellegen, 1988) which contains 20 items measuring positive affect (PA) and negative affect (NA)during the last week. Both scales have adequate reliability and validity for measuring these dis-tinct affective structures (Watson, Clark, & Tellegen, 1988).A subsample of 38 hoarding participants were interviewed via telephone using a structured

interview to determine the extent to which clutter in the living areas of their home impairedtheir ability to carry out basic activities. All 38 participants reported at least moderate levels ofclutter that would be considered clinically significant, and 34 of the 38 reported extreme clutter.Participants indicated whether they could complete activities on a 3-point scale (1 ¼ can dowithout difficulty, 2 ¼ can do only with difficulty, 3 ¼ unable to do) for each of 12 routinehousehold activities (e.g., using the stove, eating at a table, sitting on a sofa). Scores for allitems were summed to provide a total activity dysfunction (AD) score. This scale had adequateinternal reliability (alpha ¼ 0:83).In addition, a small subsample of hoarding participants were sent a follow-up questionnaire

containing a second SI-R to complete and return. Twelve of the 25 participants returned theretest SI-R. Timing of the retest ranged from 2 weeks to 4 weeks.

1.3.2. ResultsConsistent with other research on hoarding, the hoarding participants were significantly older

(M ¼ 50:7) than the community controls (M ¼ 44:4; tð90Þ ¼ 2:36, p < 0:05). The groups did notdiffer on levels of education (t½83� ¼ 0:52, n.s.) or income (t½85� ¼ 0:78, n.s.), but a higher per-centage of hoarding participants lived alone (44.8%) compared to community controls (13.6%;Chi Sq ð1Þ ¼ 5:6, p < 0:05).Test–retest reliability was studied for a small subset of 12 participants who completed a

second SI-R two to four weeks after the original questionnaire. Correlations were high: 0.86 forthe SI-R total, 0.89 for Difficulty Discarding, 0.90 for Clutter, and 0.78 for Acquisition. Pairedt-tests examining changes over time on the SI-R and its subscales failed to reveal any significanteffects (all p0s > 0:15).T-tests comparing hoarding participants to community controls revealed a significant effect

for the SI-R total and subscale scores (SI-R total t½91� ¼ 12:4; Difficulty Discarding t½91� ¼ 9:0;Clutter t½91� ¼ 11:5; Acquisition t½91� ¼ 7:3; all p’s < 0:001). In each case, the SI-R measure wassignificantly higher for hoarding participants than for community controls (see Table 5). Again,there was relatively little overlap in the SI-R total score distributions. No community controlparticipant scored above the mean for the hoarding group, and only one community control

1171R.O. Frost et al. / Behaviour Research and Therapy 42 (2004) 1163–1182

subject scored within one standard deviation below the hoarding mean (49.3). Similarly, nohoarding participant scored below the mean for the community control group, and no hoardingparticipant scored within one standard deviation above the SI-R total mean for the communitycontrol group (36.9).A further test of the construct validity of the SI-R was provided by correlations between SI-R

scores and the interview measure of activity dysfunction on a subsample of 38 hoarding parti-cipants. Significant correlations were evident for SI-R total (r ¼ 0:54, p < 0:001), and for Dif-ficulty Discarding (r ¼ 0:36, p < 0:05) and Clutter (r ¼ 0:52, p < 0:001), but not Acquisition(r ¼ 0:23, ns). This finding supports the validity of the SI-R subscales, particularly the Cluttersubscale, in predicting other relevant variables, such as interference from clutter. The pattern ofcorrelations also supports the discriminant validity of the subscales of the SI-R. The Cluttersubscale correlated with a different measure of clutter (i.e., interview-based assessment of inter-ference from clutter), while the Acquisition subscale did not.As evident from Table 6, among hoarding participants only the Difficulty Discarding sub-

scale of the SI-R was significantly and negatively correlated with Positive Affect from thePANAS. SI-R total, Clutter, and Acquisition were significantly and positively correlated withNegative Affect, and the correlation with Difficulty Discarding approached significance(p ¼ 0:08). These correlations were low to moderate in size (0.21 to 0.38). SI-R subscales

Table 5Means, standard deviations and range on SI-R and its subscales for 70 hoarding participants and 23 community con-trols

Scale Hoarding participants Community controls

(N ¼ 70) (N ¼ 23)

Mean s.d. Min Max Mean s.d. Min Max

SI-R total 62.0 12.7 37 91 23.7 13.2 0 51Difficulty dis-carding

19.8 5.0 4 28 9.2 5.0 0 18

Clutter 26.9 6.6 6 36 8.2 7.1 0 26Acquisition 15.2 5.4 3 28 6.4 3.6 0 13

SI-R=Saving Inventory-Revised.

Table 6Correlation of hoarding symptoms (SI-R) with positive and negative affect (PANAS) and with OCD symptoms (Y-BOCS) among 70 hoarding participants

Positive affect Negative affect Y-BOCS

SI-R total �0.22 0.38��� 0.25�

Difficulty discarding �0.26� 0.21 0.24�

Clutter �0.15 0.26� 0.06Acquisition �0.08 0.38��� 0.29�

�p < 0:05, ��p < 0:01, ���p < 0:001:SI-R = Saving Inventory-Revised, Y-BOCS = Yale-Brown Obsessive Compulsive Scale, PANAS = Positive AffectNegative Affect Scale.

R.O. Frost et al. / Behaviour Research and Therapy 42 (2004) 1163–11821172

showed a differing pattern of correlations with OCD severity on the YBOCS (see Table 6). Thetotal SI-R score and both Difficulty Discarding and Acquisition subscales were moderately posi-tively correlated with the YBOCS, whereas the Clutter subscale was not. Comparisons of thesecorrelations with correlations with activity dysfunction provide additional evidence that the sub-scales measure slightly different phenomena. For example, correlations of the Clutter subscalewith activity dysfunction were significantly larger than the correlations between Clutter andNegative Affect or the YBOCS (t½35� ¼ 2:71, p < 0:05).Research on OCD has suggested that hoarding is a distinct subtype of OCD and may reflect

a separate disorder (Calamari, Wiegartz, & Janeck, 1999; Leckman et al., 1997; Summerfeldt,Richter, Antony, & Swinson, 1999). Interestingly, 23 of the hoarding subjects had YBOCSscores below 5 suggesting no other OCD symptoms at all.

1.4. Study 4: Behavioral assessment of hoarding

Study 4 examined the validity of the SI-R in relation to a self-rating and an independentobserver’s rating of clutter in the main rooms of people’s homes. The SI-R was also studied inrelation to self-rated dysfunction in daily activities as a result of the clutter. We hypothesizedthat the SI-R would be correlated with each of these measures, and in particular, that the Clut-ter subscale would show the strongest correlation since the observational measures were of clut-ter. Hoarding has been found to be particularly problematic in elderly populations (Steketee,Frost, & Kim, 2001), so to increase the generalizability of findings related to the SI-R, an eld-erly sample was chosen for this study.

1.4.1. Method1.4.1.1. Participants. As part of an interview study examining features of hoarding among eld-erly clients, 25 individuals from the Boston area were recruited for a study of saving behaviorfrom elder service agencies who agreed to advertise the study among their clients. Included wereclients with no evident hoarding problems (N ¼ 13), as well as those whose homes were seri-ously cluttered (N ¼ 12). Participants’ mean age was 75.0 (s:d: ¼ 7:9) and 19 (76%) werewomen. Most participants were widowed, divorced or separated (68%), 16% were married and16% had never married.

1.4.1.2. Measures and procedures. All interviews took place in participants’ homes; theseincluded rented or owned apartments (68%) and single family homes (32%). On arrival in thehome, a trained master’s level psychologist obtained informed consent and was given a brieftour of the main rooms of the home, including living room, dining room bedroom(s), kitchenand bathroom. After demographic information was collected, the interviewer read each SI-Rquestion and the possible responses and recorded the participant’s numerical rating.Following this, the interviewer inquired about the person’s perception of the extent of clutter

in each of the rooms in the home, on a scale from 0 ¼ none to 4 ¼ very severe. The intervieweralso rated these rooms using the same scale based on his observation of the rooms. The inter-viewer was trained by the authors (GS and RF) in rating severity of clutter using photographsthat illustrated a range of clutter severity in various rooms. Because all participants’ homes con-tained a living room, kitchen, and bedroom, the severity ratings for these three rooms were

1173R.O. Frost et al. / Behaviour Research and Therapy 42 (2004) 1163–1182

summed and divided by 3 to provide a mean clutter severity score (CS). The interviewer assistedsubjects in completing the activity dysfunction (AD) scale as described in Study 3, in addition toseveral other questions not relevant to the present project.

2. Results

The mean score for all subjects on the SI-R was 28.3 (s:d: ¼ 18:1), scores that are slightlyhigher than those of non-hoarding OCD participants and community controls in Studies 2 and3. The internal consistency of the SI-R total score in this sample was 0.94. Reliabilities for thesubscales were similar to those observed in Study 1 (Difficulty Discarding ¼ 0:93; Clutter ¼0:88, Acquisition ¼ 0:80). Comparison of the 2 groups indicated that the hoarding participantsscored significantly higher on the SI-R total and each of the subscales than the non-hoardingparticipants. See Table 7.Pearson correlations were computed for the SI-R total and subscale scores and the average

self and observer ratings of clutter severity, as well as dysfunction in daily activities (AD). Theseare presented in Table 8. Findings indicated that both the residents’ and the observer’s rating ofclutter were highly correlated (0.70–0.80) with the SI-R total score and with the Difficulty Dis-carding and Clutters subscales. Correlations with the Acquisition subscale were only moderate(0.5 range). The SI-R total score and Difficulty Discarding and Clutter subscales were also mod-erately correlated with activity dysfunction (0.52–0.58). As anticipated, acquisition was notrelated to this measure of the effect of clutter.

3. Discussion

The items for the Saving Inventory-Revised were selected so as to adequately sample therange of hoarding symptoms identified by recent research, especially with respect to problemswith compulsive acquisition (Frost et al., 1998). Items were also selected to minimize the assess-ment of beliefs associated with hoarding which can be assessed separately (Steketee et al., 2003).SI-R items were constructed to avoid references to specific types of possessions since thesevary from person to person with hoarding problems and none may apply universally. Finally,items were designed to measure distress and impairment associated with each of thethree dimensions of compulsive hoarding. A factor analysis of the SI-R among a sample ofcompulsive hoarding participants (Study 1) revealed a 3-factor solution using 23 of the 26 orig-inal items. The 3 factors correspond to the 3 major symptoms hypothesized in Frost and

Table 7Means, (standard deviations), t and p-values for hoarding and non-hoarding elders in Study 4

Hoarding elders Non-hoarding elders t(23) p

SI-R total 44.6 (10.1) 13.3 (7.2) 8.97 0.001Difficulty discarding 15.9 (3.7) 6.8 (4.5) 5.53 0.001Clutter 18.9 (6.2) 3.2 (3.8) 7.70 0.001Acquistion 9.8 (4.2) 3.3 (2.6) 4.65 0.001

R.O. Frost et al. / Behaviour Research and Therapy 42 (2004) 1163–11821174

Hartl’s (1996) model of compulsive hoarding: excessive clutter in the home, difficulty discardingpossessions, and excessive acquisition of purchased and free items. Impairment and distress onthe SI-R were worded to reflect one of the three behavioral symptoms and this may account forwhy they loaded onto symptom factors rather than onto a separate factor, as was found byColes, Frost, Heimberg and Steketee (2003). Discarding items that did not load onto a factordid not appear to reduce the utility of the measure, and thus we retained only the 23 items thatloaded clearly onto one of the factors.This factor analysis was similar to one conducted using university undergraduates (Coles

et al., 2003). The Coles et al. factor analysis generated 3 strong factors corresponding to the 3 majorsymptom dimensions (clutter, difficulty discarding, acquisition), and one additional somewhatweaker distress/interference factor. Differences in the subject samples may account for thevariation in factor structure. In university students, hoarding behaviors may be displayed, butrelatively few undergraduates suffer significant distress or interference (Frost & Gross, 1993).In the compulsive hoarding sample, however, all of the subjects experience significant distressor interference related to the symptoms.Internal consistency was high for the 23-item total scale and each of the subscales. Test–retest

reliability was very good, but needs replication with a larger sample. Some overlap among thesubscales was evident, as would be expected from clinical observation of hoarding behavior.Interestingly, the Acquisition subscale was less closely associated with the total hoarding scoreand with Difficulty Discarding and Clutter, suggesting that this component of hoarding may beslightly more independent.As findings from Studies 2 and 3 indicate, scores on the SI-R total and its subscales clearly

distinguished self-identified hoarding participants from OCD participants without hoarding, aswell as from community controls. Hoarding participants scored significantly and substantiallyhigher than non-hoarding participants on all SI-R scales, usually by a factor of 2 or more.Examination of scores also indicated that rarely did individuals from one group score withinone standard deviation of the mean of the other. This minimal overlap in the distributionspoints to the specificity of the scale.Hoarding and non-hoarding samples also differed in age, as has been found consistently in

other studies (e.g., Frost, Steketee, Williams & Warren, 2000); hoarding participants were sev-eral years older than other clinical and non-clinical samples drawn from similar sources. InStudies 2 and 3, the mean age for adult hoarding participants approached 50 whereas the meanage for non-hoarding adults was nearer 40. Gender was associated only with acquisition:

Table 8Correlations of the Saving Inventory-Revised (SI-R) and averaged self and observer ratings of clutter severity (CS-Total) and observer rated Activity Dysfunction for a sample of 25 elders

Self-rated CS-total Observer CS-total Activity dysfunction

SI-R total 0.79�� 0.78�� 0.54��

Difficulty discarding 0.70�� 0.71�� 0.52��

Clutter 0.79�� 0.79�� 0.58��

Acquisition 0.52� 0.52�� 0.22

��p < 0:001, �p < 0:01.

1175R.O. Frost et al. / Behaviour Research and Therapy 42 (2004) 1163–1182

women acquired more things than men, perhaps due to cultural expectations that women buymore things and derive emotional satisfaction from buying. The gender difference observed hereis consistent with the literature on compulsive buying (Lejoyeux, Ades, Tassain, & Solomon,1996). Findings from numerous studies report higher levels of compulsive buying amongfemales (Frost, Steketee, & Williams, 2002). In Study 3, hoarding participants tended to livealone (44%) more often than community controls (14%). That more hoarding participants livedalone is consistent with previous findings that a large percentage of elderly people with hoardingproblems had never married and lived alone (Steketee, Frost, & Kim, 2001). As Table 8 shows,living alone versus with someone led to a difference only on the Clutter scale of the SI-R, butnot on Difficulty Discarding or Acquisition. Our findings do not clarify the direction of therelationship between clutter and living situation: does hoarding lead to social isolation and atendency to live alone or does having no one else in the home increase the likelihood that clutterwill be unrestricted and become a problem? Further research on the causes and consequences ofliving alone for people suffering from compulsive hoarding is warranted.Examination of the correlations with other measures of hoarding indicates good convergent

validity of the SI-R and its subscales. The Total score and the Difficulty Discarding and Cluttersubscales correlated significantly and strongly (rs ¼ 0:54 to 0.75) with a measure of hoarding-related beliefs and attitudes (Study 2). The Acquisition subscale was less strongly related to thismeasure, again suggesting that acquisition may be a less central and less consistent componentof hoarding behavior. In Study 3, the SI-R total, Difficulty Discarding and Clutter subscaleswere correlated with measures of activity dysfunction resulting from the clutter. Differential sub-scale validity can be seen in the fact that the highest correlation was between the clutter subscaleand the activity dysfunction from clutter, and lowest for Acquisition and activity dysfunctionfrom the clutter. This pattern would be expected since the activity dysfunction measure is analternate measure of clutter. The magnitude of the correlations with activity dysfunction (rs inthe 0.5 range) is noteworthy since the SI-R and its subscales clearly have a restricted rangeamong the hoarding subjects used for these correlations.Further evidence for the convergent validity of the SI-R was provided in Study 4 in which the

SI-R, Difficulty Discarding, and Clutter subscales were significantly and highly correlated withboth self-rated and observer-rated clutter in the home. Again, since these observational meas-ures focus specifically on how cluttered the homes were, we expected them to be less highly cor-related with the Acquisition dimension of compulsive hoarding. Indeed, the correlations withAcquisition, though still significant, were significantly smaller than those between Clutter andobservational measures of clutter.Convergent validity of the SI-R is demonstrated by the fact that the SI-R and its subscales

are strongly related to other indices of compulsive hoarding. Evidence of discriminant validitycomes from the absence of (or smaller magnitude of) correlations with measures less central tothe construct of hoarding, particularly positive affect, and to a lesser extent, negative affect andYBOCS. For the latter two scales, the correlations were significant, but small in magnitude.While part of the reason for the small correlations may be the restricted range of SI-R scoresamong the hoarding subject, the limited range did not result in low correlations with measuresof hoarding-related behavior (e.g., activity dysfunction, observer ratings of clutter).In examining the relationship of the SI-R to other measures of psychopathology, it is interest-

ing to note that, although hoarding has been considered a symptom of OCD, the correlation of

R.O. Frost et al. / Behaviour Research and Therapy 42 (2004) 1163–11821176

the SI-R and its subscales with the YBOCS were relatively small (rs ¼ 0:24 to 0.29) and a largenumber of hoarding subjects in Study 2 did not display any other OCD symptoms. Moreresearch is needed to examine the possibility that hoarding associated with OCD may have dif-ferent causes or characteristics than hoarding without OCD.These findings suggest that the SI-R is a reliable and valid measure of compulsive hoarding.

The subscales appear to reflect real and meaningful phenomena related to compulsive hoarding.Further work using this measure is warranted. For instance, adding a time frame to the instruc-tions for the SI-R (e.g., in the past month) may make the measure more sensitive to change.One limitation of the research reported here is the use of self-report to determine hoarding

status and severity. Independent verification of hoarding status and its significance would havestrengthened these findings. Further research on this problem is needed to establish a clearerdelineation of when hoarding is clinically significant.

Acknowledgements

This research was supported in part by a grant from an anonymous donor.

Appendix. Saving Inventory Revised

(Please circle the response that is most appropriate.)

1. To what extent do you have difficulty throwing things away?

0 = Not at all1 = To a mild extent2 = To a moderate extent3 = To a considerable extent4 = Very much so

2. How distressing do you find the task of throwing things away?

0 = No distress1 = Mild distress2 = Moderate distress3 = Severe distress4 = Extreme distress

3. To what extent do you have so many things that your room(s) are cluttered?

0 = Not at all1 = To a mild extent2 = To a moderate extent

1177R.O. Frost et al. / Behaviour Research and Therapy 42 (2004) 1163–1182

3 = To a considerable extent4 = Very much so

4. How often do you avoid trying to discard possessions because it is too stressful or time-consum-ing?

0 = Never avoid, easily able to discard items1 = Rarely avoid, can discard with a little difficulty2 = Sometimes avoid3 = Frequently avoid, can discard items occasionally4 = Almost always avoid, rarely able to discard items

5. How distressed or uncomfortable would you feel if you could not acquire something you wanted?

0 = Not at all1 = Mild, only slightly anxious2 = Moderate, distress would mount but remain manageable3 = Severe, prominent and very disturbing increase in distress4 = Extreme, incapacitating discomfort from any such effort

6. How much of the living area in your home is cluttered with possessions? (Consider the amountof clutter in your kitchen, living room, dining room, hallways, bedrooms, bathrooms or otherrooms.)

0 = None of the living area is cluttered1 = Some of the living area is cluttered2 = Much of the living area is cluttered3 = Most of the living area is cluttered4 = All or almost all of the living area is cluttered

7. How much does the clutter in your home interfere with your social, work or everyday function-ing? Think about things that you don’t do because of clutter.

0 = Not at all1 = Mild, slight interference, but overall functioning not impaired2 = Moderate, definite interference, but still manageable3 = Severe, causes substantial interference4 = Extreme, incapacitating

8. How often do you feel compelled to acquire something you see (e.g., when shopping or offeredfree things)?

0 = Never feel compelled1 = Rarely feel compelled

R.O. Frost et al. / Behaviour Research and Therapy 42 (2004) 1163–11821178

2 = Sometimes feel compelled3 = Frequently feel compelled4 = Almost always feel compelled

9. How strong is your urge to buy or acquire free things for which you have no immediate use?

0 = Urge is not at all strong1 = Mild urge2 = Moderate urge3 = Strong urge4 = Very strong urge

10. How much control do you have over your urges to acquire possessions?

0 = Complete control1 = Much control, usually able to control urges to acquire2 = Some control, can control urges to acquire only with difficulty3 = Little control, can only delay urges to acquire only with great difficulty4 = No control, unable to stop urges to acquire possessions

11. How often do you decide to keep things you do not need and have little space for?

0 = Never keep such things1 = Rarely2 = Occasionally3 = Frequently4 = Almost always keep such possessions

12. To what extent does clutter prevent you from using parts of your home?

0 = All parts of the home are usable1 = A few parts of the home are not usable2 = Some parts of the home are not usable3 = Many parts of the home are not usable4 = Nearly all parts of the home are not usable

13. To what extent does the clutter in your home cause you distress?

0 = No feelings of distress or discomfort1 = Mild feelings of distress or discomfort2 = Moderate feelings of distress or discomfort3 = Severe feelings of distress or discomfort4 = Extreme feelings of distress or discomfort

1179R.O. Frost et al. / Behaviour Research and Therapy 42 (2004) 1163–1182

14. How frequently does the clutter in your home prevent you from inviting people to visit?

0 = Not at all1 = Rarely2 = Sometimes3 = Often4 = Very often or nearly always

15. How often do you actually buy (or acquire for free) things for which you have no immediateuse or need?

0 = Never1 = Rarely2 = Sometimes3 = Frequently4 = Almost always

16. How strong is your urge to save something you know you may never use?

0 = Not at all strong1 = Mild urge2 = Moderate urge3 = Strong Urge4 = Very strong urge

17. How much control do you have over your urges to save possessions?

0 = Complete control1 = Much control, usually able to control urges to save2 = Some control, can control urges to save only with difficulty3 = Little control, can only stop urges with great difficulty4 = No control, unable to stop urges to save possessions

18. How much of your home is difficult to walk through because of clutter?

0 = None of it is difficult to walk through1 = Some of it is difficult to walk through2 = Much of it is difficult to walk through3 = Most of it is difficult to walk through4 = All or nearly all of it is difficult to walk through

19. How upset or distressed do you feel about your acquiring habits?

0 = Not at all upset1 = Mildly upset

R.O. Frost et al. / Behaviour Research and Therapy 42 (2004) 1163–11821180

2 = Moderately upset3 = Severely upset4 = Extreme embarrassment

20. To what extent does the clutter in your home prevent you from using parts of your home fortheir intended purpose? For example, cooking, using furniture, washing dishes, cleaning, etc.?)

0 = Never1 = Rarely2 = Sometimes3 = Frequently4 = Very frequently or almost all the time

21. To what extent do you feel unable to control the clutter in your home?

0 = Not at all1 = To a mild extent2 = To a moderate extent3 = To a considerable extent4 = Very much so

22. To what extent has your saving or compulsive buying resulted in financial difficulties for you?

0 = Not at all1 = A little financial difficulty2 = Some financial difficulty3 = Quite a lot of financial difficulty4 = An extreme amount of financial difficulty

23. How often are you unable to discard a possession you would like to get rid of?

0 = Never have a problem discarding possessions1 = Rarely2 = Occasionally3 = Frequently4 = Almost always unable to discard possessions

References

Baer, L. (1994). Factor analysis of symptom subtypes of obsessive compulsive disorder and their relation to person-ality and tic disorders. Journal of Clinical Psychiatry, 55, 18–23.

Black, D. W., Monahan, P., Gable, J., Blum, N., Clancy, G., & Baker, P. (1998). Hoarding and treatment responsein 38 nondepressed subjects with obsessive–compulsive disorder. Journal of Clinical Psychiatry, 156, 1409–1416.

Calamari, J. E., Wiegartz, P. S., & Janeck, A. S. (1999). Obsessive–compulsive disorder subgroups: a symptom-basedclustering approach. Behaviour Research and Therapy, 37, 113–125.

1181R.O. Frost et al. / Behaviour Research and Therapy 42 (2004) 1163–1182

Coles, M. E., Frost, R. O., Heimberg, R. G., & Steketee, G. (2003). Hoarding behaviors in a large college sample.Behaviour Research and Therapy, 41, 179–194.

Frankenburg, F. (1984). Hoarding of anorexia nervosa. British Journal of Medical Psychology, 57, 57–60.Frost, R. O., & Gross, R. C. (1993). The hoarding of possessions. Behaviour Research and Therapy, 31, 367–381.Frost, R. O., & Hartl, T. L. (1996). A cognitive-behavioral model of compulsive hoarding. Behaviour Research andTherapy, 34, 341–350.

Frost, R. O., Hartl, T. L., Christian, R., & Williams, N. (1995). The value of possessions in compulsive hoarding:Patterns of use and attachment. Behaviour Research and Therapy, 33, 897–902.

Frost, R. O., Kim, H.-J., Morris, C., Bloss, C., Murray-Close, M., & Steketee, G. (1998). Hoarding, compulsive buy-ing and reasons for saving. Behaviour Research and Therapy, 36, 657–664.

Frost, R. O., Krause, M., & Steketee, G. (1996). Hoarding and obsessive compulsive symptoms. Behavior Modifi-cation, 20, 116–132.

Frost, R. O., & Steketee, G. (1998). Hoarding: Clinical aspects and treatment strategies. In M. A. Jenike, L. Baer, &Minichiello (Eds.), Obsessive–compulsive disorders: practical management. (3rd ed.) St. Louis: Mosby.

Frost, R. O., Steketee, G., & Williams, L. F. (1999). Hoarding: a community health problem. Health and Social Carein the Community, 8, 229–234.

Frost, R. O., Steketee, G., & Williams, L. F. (2002). Compulsive buying, compulsive hoarding, and obsessive com-pulsive disorder. Behavior Therapy, 33, 201–214.

Frost, R. O., Steketee, G., Williams, L. F., & Warren, R. (2000). Mood, personality disorder symptoms and dis-ability in obsessive–compulsive hoarders: A comparison with clinical and nonclinical controls. Behaviour Researchand Therapy, 38, 1071–1081.

Goodman, W. K., Price, L. H., Rasmussen, S. A., Mazure, C., Fleischmann, R. L., Hill, C. L., Heniger, G. R., &Charney, D. S. (1989). The Yale–Brown Obsessive Compulsive Scale. I. Development, use and reliability. Archivesof General Psychiatry, 46, 1006–1011.

Greenberg, D. (1987). Compulsive hoarding. American Journal of Psychotherapy, 41, 409–416.Greenberg, D., Witzhum, E., & Levy, A. (1990). Hoarding as a psychiatric symptom. Journal of Clinical Psychiatry,51, 417–421.

Hartl, T. L., Duffany, S. R., Allen, G. J., Steketee, G., & Frost, R. O. (2003). Relationships among compulsive hoard-ing, trauma, and attention deficit hyperactivity disorder. Submitted.

Kim, H-J., Frost, R. O., Steketee, G., Turkoff, A., & Hood, S. (2003). The formation of attachments to possessions.Manuscript in preparation.

Leckman, J. F., Grice, D. E., Boardman, J., Zhang, H., Vitale, A., Bondi, C., Alsobrook, J., Peterson, B. S., Cohen,D. J., Rasmussen, S. A., Goodman, W. K., McDougle, C. J., & Pauls, D. L. (1997). Symptoms of obsessive–com-pulsive disorder. American Journal of Psychiatry, 55, 18–23.

Lejoyeux, M., Ades, J., Tassain, V., & Solomon, J. (1996). Phenomenology and psychopathology of uncontrolledbuying. American Journal of Psychiatry, 153, 1524–1529.

Shafran, R., & Tallis, F. (1996). Obsessive–compulsive hoarding: A cognitive–behavioural approach. Behavioural andCognitive Psychotherapy, 24, 209–221.

Stein, D. J., Laszlo, B., Marais, E., Seedat, S., & Potocnok, F. (1997). Hoarding symptoms in patients on a geriatricpsychiatry inpatient unit. South African Medical Journal, 87, 1138–1140.

Steketee, G., Frost, R. O., & Kim, H.-J. (2001). Hoarding by elderly people. Health and Social Work, 26, 176–184.Steketee, G., Frost, R. O., & Kyrios, M. (2003). Cognitive aspects of compulsive hoarding. Cognitive Therapy andResearch, 27, 463–479.

Summerfeldt, L. J., Richter, M. A., Antony, M. M., & Swinson, R. P. (1999). Symptom structure in obsessive–com-pulsive disorder: a confirmatory factor-analytic study. Behaviour Research and Therapy, 37, 297–311.

Taylor, S. (1998). Assessment of obsessive–compulsive disorder. In P. Swinson, M. M. Antony, S. RachmanM. A. Richter (Eds.), Obsessive–compulsive disorder: Theory, research, and treatment (pp. 229–257). New York:Guilford.

Watson, D., Clark, L. A., & Tellegen, A. (1988). Development and validation of brief measures of positive and nega-tive affect: The PANAS scales. Journal of Personality and Social Psychology, 54, 1063–1070.

R.O. Frost et al. / Behaviour Research and Therapy 42 (2004) 1163–11821182