Full Terms & Conditions of access and use can be found at
Anthrozoös
A multidisciplinary journal of the interactions of people and animals
ISSN: 0892-7936 (Print) 1753-0377 (Online) Journal homepage: http://www.tandfonline.com/loi/rfan20
Psychometric Evaluation of the Lexington Attachment to Pets Scale (Laps)
Timothy P. Johnson, Thomas F. Garrity & Lorann Stallones
To cite this article: Timothy P. Johnson, Thomas F. Garrity & Lorann Stallones (1992)
Psychometric Evaluation of the Lexington Attachment to Pets Scale (Laps), Anthrozoös, 5:3, 160-175
To link to this article: http://dx.doi.org/10.2752/089279392787011395 Published online: 27 Apr 2015.
Submit your article to this journal
Article views: 4
View related articles
Citing articles: 40 View citing articles
PSYCHOMETRIC EVALUATION OF THE LEXINGTON ATTACHMENT TO PETS SCALE (LAPS)
Timothy P.Johnson,1 Thomas F.Garrity,2 and Lorann Stallones3
Abstract.This paper reports on the devel- opment and psychometric evaluation of a scale for assessing emotional attachment of individuals to their pets. Previous attach- ment scales have suffered variously from low internal consistency and reliance on small or nonrepresentative samples for their development. Telephone interviews of a random, representative sample of 412 pet owners in Fayette County, Kentucky, were completed in September 1990; a 69.5 percent response rate was achieved.
From a preliminary set of 42 questions, a final 23-question instrument, the Lexington Attachment to Pets Scale (LAPS), was de- veloped, having excellent psychometric properties. The scale is suitable for use with dog and cat owners. Data on internal consistency, factor structure, and item re- sponse theory (IRT) modeling are pre- sented, along with correlations between the LAPS and several domains of variables known to relate to pet attachment.
INTRODUCTION
Eight years ago, a review of the literature concerned with the role that pet attach- ment may play in the maintenance of hu- man health concluded that most of the available evidence was lacking in scien- tific rigor (Marx 1984). Since then, the re- search community has taken an active in- terest in the hypothesized link between human-animal interaction and both physi- cal and psychological well-being (Culliton 1987; Department of Health and Human Services 1988). Evidence has been re- ported that pet ownership may be benefi- cial in both younger (Poresky et al. 1987) and older populations (Akiyama, Holtzman, and Britz 1986–87; Bolin 1987;
Lago, Connell, and Knight 1983; Ory and Goldberg 1983).
Studies by our research group have found that emotional attachment to com- panion animals was associated with de- creased reports of depressive symptoms among elderly respondents (Garrity et al.
1989). Using a similar methodology, we also studied a national probability sample of adults aged 21–64. We again observed a significant relationship between pet at- tachment and depressive symptoms, but only among those aged 35–44, and this association was reduced to nonsignificance when other respondent characteristics were controlled (Stallones et al. 1990). The research conducted to date suggests that, although some evi- dence does support the hypothesis that pets contribute to emotional health, the relationship between pet attachment and well-being is complex.
1 University of Illinois, P.O. Box 6905, Chicago, IL 60680.
2 University of Kentucky, Lexington, KY.
3 Colorado State University, Fort Collins, CO.
REVIEWS AND RESEARCH REPORTS
This research was supported by a grant from the Delta Society and the American Pet Products Manufacturers Association to the authors.
Downloaded by [University of Windsor] at 23:31 25 September 2015
Research undertaken since Marx’s (1984) review reflects a marked improvement in the quality and rigor of the methods used. However, additional efforts are needed to more adequately measure the concept of pet attachment.
After examining available instruments, Wilson, Netting, and New (1987) concluded that few measures existed for which adequate evaluations of reliability and validity have been conducted.
Similarly, Poresky (1988) argued that
“there is a clear need to improve the utility of the measures used to assess the human- animal relationship.” During this period, several important measures were developed to address many of these concerns. Prominent among these were the Companion Animal Bonding Scale (Poresky et al. 1987), the Companion Animal Semantic Differential (Poresky et al. 1988), the Pet Attitude Inventory (Wilson, Netting, and New 1987), the Pet Attitude Scale (Holcomb, Williams, and Richards 1985), and the Pet Relationship Scale (Lago et al. 1988). Most of the developmental work these scales were based upon, however, involved nonrandom convenience samples. Our group has reported the development of a six-item index of pet attachment that was evaluated using data from a national sample of the elderly (Stallones et al.
1988). This measure was found to have a moderate level of internal consistency.
During a subsequent study, we developed an eight-item scale that improved upon our earlier measure (Stallones et al. 1990).
Although these efforts were promising, given that the work was based upon national random probability samples, we have been concerned that the available measures remain imperfect indicators of the degree of affection that may exist between individuals and their companion animals. This article reports on the
development of a new pet attachment instru ment, to be referred to as the Lexington Attachment to Pets Scale (LAPS), that has psychometric properties superior to our previous efforts.
METHODS Data Collection
Data were collected via a telephone sur- vey conducted in Fayette County, Kentucky, a metropolitan area with a population of approximately 250,000. The survey sample was generated using random-digit dialing, a methodology that gives all residential tele- phone numbers in the geographic area of interest an equal probability of selection (Waksberg 1978). Each number contacted was screened to determine if there was a pet in the household. Within those house- holds with a pet, an adult respondent (18+) was selected using a random respondent selection procedure. Call-back appoint- ments were made whenever the selected respondent was unavailable. In addition, selected individuals refusing to be inter- viewed were called back at a later date and asked to reconsider participation. In all, 412 interviews were completed, and the survey’s response rate (completed inter- views/total eligible) was 69.5%.
All interviewing was conducted by professional interviewers employed by the University of Kentucky Survey Research Center. Interviewing was 100 percent supervised, and a sample of each interviewer’s work was monitored for accuracy. Completed questionnaires were edited prior to data entry. All interviewing was conducted during September 1990.
Instrument Development
Pet attachment items were developed based upon our review of a number of
Downloaded by [University of Windsor] at 23:31 25 September 2015
sources. We adapted a number of items from the two previous attachment scales developed by our research team. The work of other investigators, including the Com- panion Animal Bonding Scale (Poresky et al. 1987), the Pet Attitude Inventory (Wil- son, Netting, and New 1987), and the Pet Attitude Scale (Templer et al. 1981), were also examined. In developing candidate items we made an effort to include ques- tions that might be representative of weak attachment to pets. Previous analyses have reported an insufficient number of items that measure weak attachment (Stallones et al. 1990), suggesting that the full range of human-animal bonds may not have been adequately assessed.
The development of items for this scale was also guided by theoretical consider- ations. Research into the association be- tween human social support and health suggests that emotional ties is the dimen- sion of support that is most closely related to health (House and Kahn 1985). Several researchers have suggested that a similar mechanism may underlie human-pet rela- tionships (Garrity et al. 1989; Lago et al.
1988; Ory and Goldberg 1983). Although a variety of candidate items were devel- oped for this study, we emphasized ques- tions that appeared to be indicative of re- spondent affection for the companion ani- mal. Our theoretical approach makes the assumption that, as with human relations, it is this aspect of owner-pet relations that is most closely related to well-being.
A total of 42 attachment items were included in the instrument. These items were introduced by the following statement:
I’d like to ask you whether you agree or disagree with some very brief statements about your favorite pet. For each statement, please tell me whether you strongly agree, somewhat agree, somewhat disagree, or strongly disagree.
The sequence in which items were asked was randomized in an effort to minimize order effects (Groves 1989).
In addition to the candidate items for the attachment scale, the questionnaire also contained information regarding types of pets in the household, favorite type of pet, and care of favorite pet. These items were positioned in the survey instrument prior to the candidate LAPS items. The final section of the instrument contained several baseline demographic items, including age, race, sex, education, marital status, and income.
Finally, interviewers were asked to record their own opinion of how attached the re- spondent was to his or her pet.
Analysis Methods
Items to be included in the LAPS were se- lected using both traditional item analysis procedures (Kline 1986) and item response theory (IRT) models (Hambelton and Swaminathan 1985). The traditional proce- dures utilized included calculation of al- pha internal consistency coefficients (Cronbach 1951), examination of itemtotal-scale correlations, and factor analysis techniques to examine the scale’s dimensionality. For these analyses all vari- ables were coded in a continuous manner.
The response categories for each item were coded as follows: 0=strongly dis- agree; 1=somewhat disagree; 2=somewhat agree; and 3=strongly agree. This coding scheme was reversed for items worded in a negative direction (items h and u in Table 1).
The IRT model utilized was the two- parameter binary logistic IRT model. Our decision to use this particular model was guided by the fact that it had been successfully utilized in previous efforts to apply IRT methods for similar applications (Duncan-Jones, Grayson, and Moran 1986; Reiser 1989; Schaeffer 1988).1 This
Downloaded by [University of Windsor] at 23:31 25 September 2015
procedure models responses to individual items as being a function of item discrimination and threshold. Discrimina- tion is the extent to which each item is associated with the underlying trait of interest, pet attachment in this case. Item discrimination is also referred to as item slope and can be considered the reliability with which each item measures the scale’s underlying characteristic (Schaeffer 1988).
Higher discrimination is indicative of greater reliability. Threshold is the probability that a given individual will endorse each item. It is considered the
point at which each respondent has a 50–
50 chance of item endorsement. By ranking item thresholds, we will be able to observe how adequately the items in the LAPS cover the range of affective ties that pet owners feel for their animals. This work was accomplished using the BILOG 3 program (Mislevy and Bock 1990).
The IRT analysis required that the four response categories of each scale item be collapsed into binary values. In examining item statistics, it was found that a plurality of respondents endorsed the response category reflective of highest attachment
Downloaded by [University of Windsor] at 23:31 25 September 2015
for a majority of the items (see Table 2). As a consequence, each item was recoded for the IRT analysis such that each indicated whether or not the respondent’s answer represented strong attachment to the pet.
This was accomplished by recoding to a value of 1 items worded in a positive direction if the respondent strongly agreed with them and items worded in a negative direction if the respondent strongly
disagreed. Otherwise, each item was recoded as 0, reflecting an absence of strong attachment. It is important to emphasize that these attachment items were dichotomized for IRT analysis only.
The two-parameter IRT model was also used to examine the structure of the LAPS across two groups of pet owners (dog versus cat owners). In analyzing pet attachment scales using heterogeneous
Downloaded by [University of Windsor] at 23:31 25 September 2015
samples of pet owners, the assumption is commonly made that the scale is equally efficient in discriminating levels of attachment for all pet owners. By conducting separate analyses for various groups, we will be able to determine if there are large differences in the ability of individual scale items to discriminate differently among dog and cat owners (other types of pet owners were not considered in this part of the analysis because of small numbers). Large differences in the characteristics of various questions across type of pet ownership would, of course, make it necessary to eliminate them from our final scale.
Finally, preliminary exploration of the construct validity of the LAPS using other available data was examined. As part of this effort, responses to the LAPS were correlated with interviewer assessments of each respondent’s attachment to his or her favorite pet. After each survey was completed, interviewers were asked to rate their subjective assessment of each respondent’s attachment to his or her favorite pet as being “very attached, somewhat attached, not very attached, or not at all attached.” Although interviewer assessments were undoubtedly based in large part on the respondent’s answers to the LAPS items, they also reflect an additional assessment of the respondent’s relationship with the favorite pet. In addition, oneway analysis of variance (ANOVA) was used to examine relationships between the LAPS and three categories of respondent characteristics for evidence of construct validity: respondent demographics, social network ties, and other characteristics related to respondent- pet interaction. Variables included within each of these three categories have been previously reported to be associated with pet attachment (Kidd and Kidd 1989;
Serpell 1981; Stallones et al. 1990).
RESULTS
The final sample for this study was 59%
female, and 94% white. The average age was 43.0 years, and ages ranged from 18 to 83 years. The typical respondent had 12 years of education (range 5 to 25 years) and reported annual family income of be- tween $20,000 and $30,000. In regard to marital status, 62% were married, 12%
had never been married, 4% were cohabitating, 13% were separated or di- vorced, and 8% were widowed; 87% re- ported having grown up with pets; 63% of the respondents indicated that dogs were their favorite pets, and 28% said that cats were their favorite. Smaller percentages said other types of pets were their favorite (4%), and 5% were unable to specify which of their pets was their favorite. An average of 16 minutes was required for interviewers to complete each question- naire.
Internal consistency of the 42 attachment questions was estimated using coefficient alpha. The coefficient for these items was 0.937, indicating a high degree of internal consistency. The purpose of this study was to develop a scale with a more practical (i.e., smaller) number of items.
Therefore the corrected item-total correlations for each of the 42 questions were examined, and 24 questions were found to have values greater than 0.50.
These 24 questions were subsequently included in a two-parameter IRT analysis to determine the extent to which each question provided a reasonable “fit” to the latent concept of “pet attachment.” A good fit is indicated by low (nonsignificant) chi-square values. Although the likelihood ratio chi-square for the overall model suggested an acceptable fit (G2=102.5, df=93, P=.235), one question had a significant chi-square value. The analysis was repeated with this item eliminated,
Downloaded by [University of Windsor] at 23:31 25 September 2015
resulting in an acceptable fit both for the model as a whole (G2=72.2, df=88, P= .739), and for each of the remaining 23 questions included in the model.
The alpha coefficient for these 23 questions (0.928) was close to the coefficient for the entire set of questions, indicating that the number of items in the scale could be reduced by nearly half without a noticeable loss in reliability. The wording of each question is shown in Table 1, and frequency distributions for each
question are provided in Table 2. The scale had a range of values from 0 to 69. The mean value for the 322 respondents who answered all questions was 47.99 (standard deviation=12.65, median value=50, modal value=54). Based upon the findings of these analyses, we further investigated the properties of the 23-item scale, which will subsequently be referred to as the LAPS.
The parameters obtained from the final IRT analysis are presented in Table 3. The
Downloaded by [University of Windsor] at 23:31 25 September 2015
units in which these parameters are pre- sented are to a large extent arbitrary. A common practice is to center the mean threshold value to 0 and the estimated variance to 1. Scales calibrated in this manner have a theoretical threshold range from +3 to -3. The thresholds shown in Table 3 indicate that 14 of the 23 items in the model are indicators of above-average (i.e., strong) pet attachment. These data also indicate that the theoretical range of
pet attachment is more adequately cov- ered by scale items at the high attachment end of the spectrum (i.e., those items with positive threshold values). Although the remaining nine items in the scale could be endorsed by individuals with less than av- erage levels of pet attachment, these items have a more restricted range. In general, these findings suggest that the LAPS is more successful in measuring strong at- tachment than weak attachment.
Downloaded by [University of Windsor] at 23:31 25 September 2015
Next, the IRT model was examined separately for respondents who indicated their favorite pet was a dog and those whose favorite pet was a cat. As mentioned earlier, this procedure was undertaken as a first step in assessing whether or not there are differences in how individual responses in the LAPS are affected by types of pets owned. These
results are presented in Table 4. There appears to be a reasonable amount of congruence between the parameters estimated for dog versus cat owners, as the correlation between the 23-item thresholds is 0.94. The distribution of the item threshold values is quite similar, suggesting the LAPS covers a similar range of affective ties for each group. The largest
Downloaded by [University of Windsor] at 23:31 25 September 2015
difference in item threshold was for item d (“I believe my pet is my best friend”), which had an estimated value for cat owners that was more than twice that of dog owners. This suggests that this item is an indicator of stronger attachment among cat owners. Table 4 also indicates that most of the items in the LAPS inventory (16 of 23) are somewhat more reliable indicators of pet attachment among cat owners.
A principal-components analysis was conducted to examine the structure of the LAPS. Results of this analysis are presented in Table 5. Using a varimax rotation procedure, three orthogonal factors were
identified. Examination of the items loading most highly on each of these factors suggested that the first represents a
“general attachment” dimension. Items included in factor 2 indicate the pet in question occupies a more central position in the respondent’s life. This factor has thus been labeled “people substituting.” The third factor contained items that appear to be mostly concerned with the pet’s status within the respondent’s household, and has been designated as the “animal rights/
animal welfare” factor. Cronbach’s alpha coefficients for these three factors were 0.90, 0.85, and 0.80, respectively.
The associations between the LAPS and
Downloaded by [University of Windsor] at 23:31 25 September 2015
a number of respondent characteristics were also examined. Table 6 shows the average pet attachment for five demographic items. All were found to be significantly related to the LAPS. Female and black pet owners tended to be more highly attached to their favorite pets, as were older respondents. In addition, those
with less education and lower income also reported stronger attachment.
The relationship between pet attach- ment and personal social networks was next evaluated (Table 7). Several of these variables were also found to vary significantly with attachment to pets.
Respondents from smaller households
Downloaded by [University of Windsor] at 23:31 25 September 2015
(those with one to two members) and those in households with no children were found to be more attached to their animals. Marital status was also related to degree of pet attachment. Those never married, divorced, and who were cohabiting expressed the highest attachment, and those who were married expressed the least. Inverse relationships between the LAPS and (1) numbers of persons the respondent felt close to, (2) the number of groups and organizations the respondent was active in, and (3) the number of friends and relatives the respondent saw monthly were also observed. The first two of these variables were significantly related to the LAPS.
The relationship of the LAPS with other pet-related variables is presented in Table 8. Respondents who had grown up with pets reported greater degrees of
attachment, although this association was only borderline significant (P=.08). Those who indicated that a dog was their favorite pet were found to have higher attachment than those who reported a cat to be their favorite pet. No relationship was observed between pet care and attachment.
Additional questions (not shown) asked about responsibility for feeding and for cleaning up after the pet. No association was found between these and the LAPS.
Finally, average LAPS scores for various interviewer ratings of respondent attachment to pets revealed that a strong association existed between respondent and interviewer assessments of respondent attachment to a favorite pet. The correlation coefficient between these two variables was 0.64.
The association between these demographic, social network, and
Downloaded by [University of Windsor] at 23:31 25 September 2015
pet-related variables and each of the three LAPS factors reported earlier was also examined (Appendices available from first author). With few exceptions, a pattern of results very similar to that reported for the total LAPS scale (Tables 6–8) was observed in these analyses.
DISCUSSION
Reviews of instruments designed to mea- sure various aspects of social support have concluded that these instruments are far from adequate indicators of what they pur- port to represent (House and Kahn 1985;
Leavy 1983; Payne and Jones 1987). Com- pared with the development of measures of affective ties with companion animals, however, the social support literature ap- pears well established. Indeed, research into the link between pet attachment and personal health would appear today to be where social support research was in the mid-1970s—with abundant anecdotal in- formation but insufficient empirical sup- port.
The purpose of the research presented in this article has been to develop and evaluate a measure of pet attachment that is reliable and for which some preliminary evidence of validity can be provided.
Previous efforts, reviewed earlier, suffered either from less than ideal internal consistency, as measured by coefficient alpha, or from reliance on small or nonrepresentative samples. The measure reported here has excellent reliability and was constructed using data collected from a random probability sample of pet owners in a medium-sized metropolitan area.
Validation of any instrument is an ongo- ing process, but the LAPS appears to have content validity in that the items all appear to represent some level of attachment to a companion animal. Some initial evidence also exists for the construct validity of this
scale, as it is strongly correlated with sub- jective ratings of respondent pet attach- ment that were made by interviewers. Ide- ally, of course, these ratings would have been made by individuals unaware of the respondent’s answers to the LAPS items—
something that would not have been prac- tical in this study. Future research will need to correlate the LAPS with one or more of the other available attachment scales before its construct validity can be more firmly established.
Evidence for the construct validity of the LAPS also comes from its association with other respondent characteristics previously demonstrated to be related to pet attach- ment. In particular, gender, education, in- come, marital status, the presence of chil- dren in the household, having grown up with pets, and being primarily responsible for pets have all been significantly corre- lated with pet attachment in other studies (Kidd and Kidd 1989; Lago et al. 1988;
Stallones et al. 1990). Perhaps most inter- esting, though, was the association between the LAPS and indicators of social relation- ships. For both primary and secondary so- cial relationships, respondents reporting fewer ties were found to have greater at- tachment to their pets. To the extent that pets might be expected to play a more central role in the lives of individuals with smaller social networks, these results represent evi- dence of construct validity. They also sug- gest an important direction for future re- search: the degree to which pet attachment is a complement to or a substitute for social relationships. The answer is most likely some of each. Our investigations, however, do suggest that pet attachment is greater among those with fewer social ties, a finding that appears to lend support to the “substitute”
hypothesis. Further work is needed in this area.
The IRT analysis demonstrated that the LAPS items do not cover the entire
Downloaded by [University of Windsor] at 23:31 25 September 2015
potential range of levels of attachment to pets. We were more successful in developing items that measure strong attachment. This is particularly frustrating given the fact that items reflecting lower levels of attachment were found, on average, to be more reliable indicators of attachment, as evidenced by higher discrimination parameters. Yet, the LAPS items do provide a reasonable distribution of attachment levels, and can be used as a benchmark for further improvements.
Although some differences were observed, it was also useful to determine that the IRT parameters for the 23 LAPS items were reasonably stable when examined separately for dog versus cat owners. Had estimates varied substantially, it would have called into question the utility of the instrument for measuring attachment to different types of pets. Although other research (Miller and Lago 1990) suggests differences in how persons interact with pets of varying species, we nonetheless conclude that the LAPS is a useful indicator of pet attachment for both dogs and cats.
Because so few other types of pets were reported to be the respondent’s favorite in this study, examination of other groups could not be conducted. Questions of applicability to other types of pets must therefore await further investigation.
Most previous attempts to develop a pet attachment measure have reported the scale to contain more than one dimension (Poresky et al. 1987; Stallones et al. 1988;
Templer et al. 1981). The LAPS was found to have three well-defined factors: general attachment, people substituting, and animal rights/animal welfare. Not surprisingly, examination of the item thresholds in Table 3 reveals that the items in factor 1 represent lower levels of pet attachment and the items in factor 2 are representative of higher levels of
attachment. The Item thresholds for factor 3 were somewhat more randomly distributed. These findings suggest that items from each factor (or at least factors 1 and 2) are necessary to adequately represent the overall construct. We therefore conclude that attempts to develop a unidimensional measure of pet attachment would not be appropriate.
Several limitations of this research should be acknowledged. First, the state- ment used to introduce the LAPS items asks respondents to answer questions re- garding their “favorite” pet. The term “fa- vorite” was included to provide respon- dents who had more than one pet with a frame of reference for responding. We used this approach because our own per- spective assumes the most salient, or “fa- vorite” pet, is most important for the re- search purposes we have identified. In retrospect, directing respondents to an- swer regarding their “favorite” may have a social desirability effect, influencing them to answer in a more positive manner than they might otherwise. This may in part ac- count for the inability of the LAPS items to assess weak attitudes.
In addition, the LAPS has to date been evaluated using responses collected only via telephone interviews. Although other research suggests that attitude scales are largely robust to variations in data collec- tion (van Tilburg and de Leeuw 1991), the utility of this scale in self-administered and in-person interviews has yet to be estab- lished. Finally, although a random prob- ability sample was used to develop this instrument, the survey was conducted in a metropolitan county. Baseline information regarding responses from rural inhabitants is therefore unavailable.
In conclusion, this study has presented details on the development and psychometric evaluation of a measure of pet attachment referred to as the Lexington
Downloaded by [University of Windsor] at 23:31 25 September 2015
Attachment to Pet Scale (LAPS). As stated earlier, attempts to address the question of how emotional attachment to companion animals may be linked to personal well- being have only recently begun. Although continued psychometric evaluation of the LAPS will be necessary, it is our hope that this instrument will contribute to the development of research in this field.
NOTE
1. Consideration was also given to utilizing one- and three-parameter IRT models. The one- parameter model was not selected because it makes the assumption that all items have equal discriminatory power. Examination of these and previously collected data strongly suggested this assumption was not valid. The three-parameter model adds a parameter for
“guessing,” a factor that we considered to be a negligible consideration for the self-report indicators used in this study.
REFERENCES
Akiyama, H., J.M.Holtzman, and W.E.Britz. 1986–
87. Pet ownership and health status during bereavement. Omega 17: 187–93.
Bolin, S.E. 1987. The effects of companion animals during conjugal bereavement. Anthrozoös 1:
26–35.
Cronbach, L.J. 1951. Coefficient alpha and the internal structure of tests. Psychometrika 16:
297–334.
Culliton, B.J. 1987. Take two pets and call me in the morning. Science 237: 1560–61.
Department of Health and Human Services. 1988.
Health benefits of pets: Summary of working group. U.S. Government Printing Office: 2988–
216–107.
Duncan-Jones, P., D.A.Grayson, and P.A.P.Moran.
1986. The utility of latent trait models in psychiatric epidemiology. Psychological Medicine 16: 391–405.
Garrity, T.F., L.Stallones, M.B.Marx, and T.P.
Johnson. 1989. Pet ownership and attachment as protective factors in the health of the elderly.
Anthrozoös 3: 35–44.
Groves, R.M. 1989. Survey errors and survey costs.
New York: Wiley.
Hambelton, R.K., and H.Swaminathan. 1985. Item
response theory: Principles and applications.
Boston: Kluwer-Nijhoff.
Holcomb, R., C.R.Williams, and P.S.Richards.
1985. The elements of attachment: Relationship maintenance and intimacy. Journal of the Delta Society 2: 28–33.
House, J.S., and R.L.Kahn. 1985. Measures and concepts of social support. In Social support and health, ed. S.Cohen and S.L.Syme. Orlando, FL:
Academic Press.
Kidd, A.H., and R.M.Kidd. 1989. Factors in adults’
attitudes toward pets. Psychological Reports 65:
903–10.
Kline, P. 1986. A handbook of test construction:
Introduction to psychometric design. New York:
Methuen.
Lago, D., C.M.Connell, and B.Knight. 1983. A companion animal program. In Mental Health and Aging, ed. M.A.Smyer and M.Gatz. Beverly Hills: Sage Publications.
Lago, D., R.Kafer, M.Delaney, and C.Connell. 1988.
Assessment of favorable attitudes toward pets:
Development and preliminary validation of the Self-Report Pet Relationship Scales. Anthrozoös 1: 240–54.
Leavy, R.L. 1983. Social support and psychological disorder: A review, Journal of Community Psychology 11: 3–21.
Marx, M.B. 1984. The salubrious endearment:
Review of New Perspectives on Our Lives with Companion Animals by A.H.Katcher and A.M.
Beck (Eds.). Contemporary Psychology 29:
902–3.
Miller, M., and D.Lago. 1990. Observed pet-owner in-home interactions: Species differences and association with the Pet Relationship Scale.
Anthrozoös 4: 49–54.
Mislevy, R.J., and R.D.Bock. 1990. BILOG 3: Item analysis and test scoring with binary logistic models, 2nd ed. Mooresville, IN: Scientific Software.
Ory, M.G., and E.L.Goldberg. 1983. Pet possession and life satisfaction in elderly women. In New perspectives on our lives with companion animals, ed. A.H.Katcher and A.M.Beck.
Philadelphia: University of Pennsylvania Press.
Payne, R.L., and J.G.Jones. 1987. Measurement and methodological issues in social support. Stress and health: Issues in research methodology, ed.
S.V.Kasl and C.L.Cooper. Chichester: Wiley.
Poresky, R.H. 1988. Analyzing human-animal relationship measures. Anthrozoös 2: 236–44.
Poresky, R.H., C.Hendrix, J.E.Mosier, and M.L.
Samuelson. 1987. The Companion Animal
Downloaded by [University of Windsor] at 23:31 25 September 2015
Bonding Scale: Internal reliability and construct validity. Psychological Reports 60: 743–6.
Poresky, R.H., C.Hendrix, J.E.Mosier, and M.L.
Samuelson. 1988. The companion animal semantic differential: Long and short form reliability and validity. Educational and Psychological Measurement 48: 255–60.
Reiser, M. 1989. An application of the item- response model to psychiatric epidemiology.
Sociological Methods & Research 18: 66–103.
Schaeffer, N.C. 1988. An application of item response theory to the measurement of depression. In Sociological Methodology 1988, Volume 18, ed. C.C.Clogg. San Francisco:
Jossey-Bass.
Serpell, J.A. 1981. Childhood pets and the influence on adults’ attitudes. Psychological Reports 49:
118–24.
Stallones, L., T.P.Johnson, T.F.Garrity, and M.B. Marx.
1990. Quality of attachment to companion animals among U.S. adults, 21 to 64 years of age.Anthrozoös 3: 171-76.
Stallones, L., M.B.Marx, T.F.Garrity, and T.P. Johnson.
1988. Attachment to companion animals among older pet owners. Anthrozoös 2: 118–
24.
Stallones, L., M.B.Marx, T.F.Garrity, and T.P.
Johnson. 1990. Pet ownership and attachment in relation to the health of U.S. adults, 21–64 years of age. Anthrozoös 4: 100–12.
Templer, D.J., C.A.Salter, S.Dickey, R.Baldwin, and D.M.Veleber. 1981. The construction of a pet attitude scale. The Psychological Record 31:
343–48.
Van Tilburg, T., and E.de Leeuw. 1991. Stability of scale quality under various data collection procedures: A mode comparison on the ‘De Jong-Gierveld Loneliness Scale. International Journal of Public Opinion Research 3: 69–85.
Waksberg, J. 1978. Sampling methods for randomdigit dialing. Journal of the American Statistical Association 73: 40–7.
Wilson, C.C., F.E.Netting, and J.C.New. 1987. The Pet Attitude Inventory. Anthrozoös 1: 76–84.
Downloaded by [University of Windsor] at 23:31 25 September 2015