The Brief-LAPS should be relevant and easily applicable for researchers studying attachment to companion animals. We therefore initially focus on how the full LAPS has been used in practical research, including an overview of the species in which the owner-animal attachment has been studied, and the type of research in which the LAPS has previously been used. We also focus on how the LAPS variable has been calculated in practical research. The reason for this is that there are different ways in which the scores on latent scales, such as the LAPS, can be calculated. These include the composite score method (i.e. all observed item scores are summed together), and, factor scores, which are linear combinations of the observed items (21). Johnson and colleagues used the composite approach: specifically, they developed a composite variable for the full LAPS scale ranging from 0 (lowest level of attachment) to 69 (highest level) (3).

A literature search was conducted in four electronic databases: Scopus, BASE, Web of Science, and Google Scholar. The search query used for all databases was “Lexington Attachment to Pets Scale”. Initially, the frequently used abbreviation “LAPS” was also used as a search term, but this was abandoned, as it generated a very large number of hits, because LAPS is an abbreviation and concept that is used in other research fields. Since all manuscripts included in the review mention the full name at least once before using the abbreviation, we believe that we missed very few, if any, papers.

Initial searches in Scopus, BASE, and Web of Science (January 2024) focused on titles, keywords, and abstracts. In February 2024, Google Scholar was used, as this electronic database also included full-text searches. Duplicate records were excluded, i.e., a publication identified in one database was not included again if found in another. Author S.S. reviewed the initial results from Scopus, BASE, and Web of Science, while author V.H. reviewed the Google Scholar results. Publications were included when they met the following criteria: (i) research reported in a peer-reviewed journal, (ii) in English language, (iii) the LAPS questionnaire data were used and reported in the article. Publications were coded by S.S. and V.H. to extract information about journal, study country, studied animal(s), approach to calculating LAPS scores, types of analyses performed, and the overarching research themes of the manuscripts (for more details about the literature search, see Supplementary File 1).

We present the results of the literature review using cross-tabulations with publication counts, and depict temporal trends in the use of the full LAPS scale and its three sub-dimensions using a line graph.

One hundred and seventy-five peer-reviewed papers that made use of the scale in the period January 1992 to February 2024 were identified. It can be seen in Figure 1 that use of the LAPS has grown steadily in recent years.

The LAPS was used in research published in the major journals that focus on anthrozoology. The study populations were from varied geographical regions, although most studies featured Western populations. Only 19 of the studies involved owners from Asia or South America (see Table 2).

Sixty-six studies included only dogs, 18 included only cats, and 38 included both species. In 23 studies, multiple animal species were covered, although these studies typically also included dogs, cats or both dogs and cats.

Most studies (n = 159) conducted descriptive statistics on, or reported the internal consistency of, the full LAPS (see Table 3). In many studies (n = 133) the full LAPS was used in associational analysis. Thirty-eight studies reported descriptive statistics for, or the internal consistency of, 1 or more of the 3 sub-dimensions. There were 46 studies that used 1 or more of the sub-dimension scales in associational analyses.

A clear majority of studies (n = 127) used the composite approach to calculate LAPS scores. Specifically, they followed the scoring method laid out by Johnson and colleagues in the original paper, which produces a composite variable for the full LAPS scale ranging from 0 (lowest level of attachment) to 69 (highest level) (3).

Just six studies reported the use of factor scores, and 4 studies used a structural equation approach. In 37 of the studies, the scoring approach was not described in the methods section, and it is not possible to infer the approach based on the presented results and analyses. Group comparisons of the means of the LAPS were conducted in 70 of the studies, and correlational tests were carried out in 132 studies (see Table 4).

Five overarching research themes were identified. The first theme, human health and conditions, explores how the emotional attachment between owners and their pets affects human mental health (22–25), physical health (26–28), perceived loneliness (12, 29), and lifestyle choices, such as exercising together (30). The second theme, human behavior and attitudes toward pets, involves studies that use the LAPS to investigate how emotional attachment influences owners' behavior toward pets, such as caring for aging animals (31, 32), and spending more time with the animal (30, 33), owners' decision-making about the use of veterinary services (34–36), or owners' willingness to invest in pet products and services (37). The third theme examines emotional attachment as a specific human socio-demographic characteristic (24, 38, 39). The fourth theme focuses on human personality, investigating how traits like neuroticism impact human-animal attachment (40). Finally, the fifth theme uses the LAPS to compare emotional attachment to different animal species, such as the differences between dog and cat owners (36, 41, 42) (Table 5).

The English language studies that have used the LAPS were predominantly from North America and Europe, and the majority had respondents with either a dog, a cat or both. Therefore, it is important to ensure that the Brief-LAPS scale is psychometrically sound for studies of attachment to dogs and cats. Inclusion of other species appears to be less important as they are seldomly studied, although it would still be desirable if the Brief-LAPS could be used in such cases.

The preferred scale construction method was to use composite scores (range from 0 to 69), and the composite score was often used in group mean comparisons and in analyses of correlations.

The LAPS was used in many research fields, including to study associations between animal attachment and human health and behavior, human socio-demographic characteristics and human personality, as well as treatment decision-making in veterinary care. Other studies have explored levels of human attachment across different animal species.

Since we are not developing a new scale but adapting an existing one, many of the steps involved in the development and validation of a scale are irrelevant here (43). It is important, however, to maintain the content of the original construct. We therefore aimed to include items from each of the 3 sub-dimensions of the LAPS. To guide us in the choice of items to retain from the 3 sub-dimensions, the idea of content validity is central. We use content validity as a term following, for instance, Boateng et al. (43) and Furr (44). The idea of content validity is to evaluate whether there is a match between what researchers are claiming that a scale measures and the actual scale content. As Furr puts it, “…if scale scores are to be interpreted in terms of a particular psychological phenomenon, then its content must reflect important facets of that phenomenon—no more, no less” (44).

It is recommended in the scale development literature that content validity should be evaluated by experts in the field with a solid understanding of the relevant psychological phenomenon to avoid bias, preconceptions, and flawed opinions from the scale creators (43, 44). In our effort to include the most accurate items from each of the sub-dimensions of the full LAPS in the Brief-LAPS version, we therefore invited experts in human-animal attachment to evaluate the content validity of the 23 LAPS items.

Study 2 was approved by the Institutional Review Board at the Faculties of Science and of Health and Medical Sciences at the University of Copenhagen (journal no.: 504/0465/24-5000).

We sought experts from the International Society for Anthrozoology (ISAZ), as a community of scholars and scientists working in the relevant study field (https://isaz.net/). ISAZ agreed to distribute an e-mail on May 7, 2024, via its voluntary membership listserv inviting a total of 237 members to participate in the survey.

Fifty-four of the 237 invitees completed the survey (response rate~23%). Table 6 presents an overview of the participant characteristics. Most respondents were female. There was considerable variation with respect to age (26% < 40 years; 48% 40–59 years; 26% >59 years). Most participants had a master's degree or a PhD or equivalent doctoral degree. The respondents also came from diverse occupational backgrounds, but most worked in academia or animal-assisted interventions. Participants were asked to self-rate their knowledge about HAI and owners' attachments to their pets on a scale from 1 (very low) to 7 (very high). The majority considered their knowledge to be high (self-reported knowledge about HAI: M = 5.7; and about owners' attachments to their pets: M = 5.1). Eight of the participants did not know the LAPS scale at all, 24 had heard about the scale, but knew little about it. Thirteen knew the scale quite well, and nine stated that they knew the scale very well.

Participants were asked to consider each of the 23 LAPS question items (which are formulated as statements), in the context of six response options. The first response gave participants the opportunity to flag statements that they found to be unsuitable measures of attachment (“Unsuitable statement to describe pet owners' attachment”). Three response options covered the LAPS sub-dimensions (“People Substitution”, “Animal Rights/Welfare”, “General Attachment”). One response option was available for statements where the LAPS sub-dimension was unclear or included several of them (“Unclear or several sub-dimension(s)”). Finally, an option was available if the statement was considered to belong to another dimension of attachment (“A different attachment dimension”) (instruction text details are appended in Supplementary File 2).

The frequency of respondents' evaluations of all 23 items is reported. We also examined whether the evaluation of five statements that many experts rated as “unsuitable” or belonging to a different attachment dimension differed across expert characteristics. We further examined whether the general assignment of the 23 statements to all six response options differed across expert characteristics. In both examinations, chi² tests were conducted, and a p < 0.05 was considered statistically significant.

The experts' evaluations of the 23 LAPS items are presented in Figure 2. The figure consists of 3 graphs, one for each sub-dimension of the LAPS. Each item was placed in the sub-dimension graph that it was assigned to by Johnson et al. (3).

There were five items where a relatively large share (> 20%) of the experts indicated that the statement was unsuitable to assess attachment or did not belong to one of the 3 LAPS sub-dimensions (i.e. that it belonged to a different attachment dimension). The items were: “Owning a pet adds to my happiness”, “I am not very attached to my pet”, “I think that my pet is just a pet”, “I believe that loving my pet helps me stay healthy” and “Quite often, my feelings toward people are affected by how they react to my pet”. We identified no differences in the evaluation of the five statements as unsuitable or belonging to a different attachment statement across age, educational levels, occupational characteristics (comparing people working in academia and animal-assisted interventions with other occupations), self-reported expertise in HAI and owner's attachment to pets, and knowledge about the LAPS.

Figure 2 also shows that many items were not unequivocally assigned to the correct sub-dimension by the experts. Just 13 items (6 from General Attachment, 5 from People Substitution, and 2 from Animal Rights/Welfare) were assigned to the same sub-dimension by more than 60% of experts. Of these, only 11 items (5 from General Attachment, 4 from People Substitution, and 2 from Animal Rights/Welfare) were assigned to the original sub-dimension laid out in Johnson et al. (3) by more than 60% of experts. Five of the 11 items belonging to the General Attachment sub-dimension were considered indicators of People Substitution by more than 30% of experts. For example, the General Attachment statement “My pet knows when I am feeling bad” was considered to belong to General Attachment by 40.7% and to People Substitution by 42.6% of experts. Conversely, 3 of the 7 items belonging to People Substitution were considered to be indicators of General Attachment by more than 30% of experts. For instance, “I love my pet because it never judges me” was considered an indicator of General Attachment by 33.3%, and of People Substitution by 46.3% of experts.

We examined whether the assignment of the 23 statements to all 6 response options differed across expert characteristics and found almost no differences. Only 1 statistically significant difference (at the 0.05 level) was identified. Specifically, experts from academia were more likely to evaluate the item “I think my pet is just a pet” as an unsuitable statement to describe pet owners' attachment, while experts not working in academia were more likely to evaluate the item as belonging to the General Attachment dimension.

More than 20% of the participants considered 5 of the 23 LAPS items to be unsuitable indicators of owner's attachment to pets or to belong to other domains of attachment that are not intended to be covered by the LAPS. Also, experts did not unequivocally assign many of the 23 items to the same sub-dimension.

Experts agreed that just 2 (of the original 5) items were good indicators of Animal Rights/Welfare. The remaining 3 items were seldomly assigned to this sub-dimension. For many of the remaining items, experts had difficulties separating whether they were indicators of General Attachment or People Substitution. In some cases, items that according to the original study (3) are supposed to be indicators of General Attachment were evaluated as indicators of People Substitution by most of the experts, and vice versa. It seems reasonable to conclude that many of the General Attachment items and People Substitution items reflect both dimensions.

A limitation of this expert data is that only 54 of the 237 invitees responded to the questionnaire. This could introduce nonresponse bias where those who did not respond differ from those who did in a way that is relevant to the study findings. However, apart from one statistically significant difference between experts from academia and other experts, no differences in the evaluation of items were identified across respondent characteristics such as age, gender, occupation, education, and self-reported knowledge. Assuming that we were able to compare relevant expert characteristics, this suggests that nonresponse bias is limited.

To develop the Brief LAPS scale we utilized findings from Study 2 to identify the items that HAI experts evaluated as having the strongest content validity.

Additionally, Study 1 delivered insights that are consequential for how we develop the Brief-LAPS. Firstly, in practical research, composite scores are predominantly used to calculate the LAPS variable. Secondly, many studies (70 of 175) use these composite scores to conduct group comparisons. When composite scoring is used, the psychometric literature recommends that a scale needs to possess what is referred to as scalar invariance (45, 46) (For technically interested readers, a description of scalar invariance and other types of measurement invariance levels is provided in Supplementary File 4).

If a scale does not have this level of measurement invariance, comparison of group differences may be biased. Examples from HAI research where a lack of scalar invariance could lead to biased results include differences in attachment to pets between cat and dog owners, and between individuals who live alone and individuals who live with others.

For this reason, our aim was to develop a brief scale that achieved scalar invariance.

Study 3 was approved by the Institutional Review Board at the Faculties of Science and of Health and Medical Sciences at the University of Copenhagen (journal no.: 504/010300/22-5000).

Pet owners in three European countries were surveyed: Austria, Denmark, and the UK. A sample of participants were recruited from pre-existing panels of citizens. The sampling scheme was intended to be representative of the overall populations in all three countries, i.e. both participants with and without pets were included. The study data is fairly representative for the target populations with regard to age, gender and living areas in the 3 countries. However, in this paper we only utilize data from respondents that self-identified as pet owners. Since the samples were drawn from pre-existing panels, it should be noted that actual random sampling was not carried out. Details about the study data, including panel recruitment criteria and average error rates have been described in detail elsewhere (47). The questionnaire included several themes (see the full questionnaire in Supplementary File 3). In this paper we included 884 pet owners from Austria, 688 from Denmark, and 767 from the UK.

The 23 LAPS statements and response options in English were replicated from the original study (3), and we used existing translated versions of the LAPS in German (19) and Danish (48). Table 7 presents the number of species that were declared as the favorite pet for each country and in total.

Dogs and cats were clearly the favorite pets; there were very few cases where horses, birds, reptiles, fish, rodents, rabbits, or other animal species were selected. We therefore decided to exclude the latter species in the development of the Brief-LAPS and only focus on respondents that mentioned dogs and cats as their favorite pet (n = 2,037).

Item exclusion strategy: The exact number of items to include in the Brief-LAPS was not decided in advance, but we hoped that we could reduce the scale to include a maximum of 8 items, and preferably 6–8 items. The item removal process involved the use of confirmatory factor analysis (CFA) (49). We examined modification indices which provide information about the items in a CFA that contribute to a poor model fit (50, 51). Items that contributed the most to a poor model fit were incrementally removed until we reached 8 remaining items. At that point we started evaluating several candidate Brief-LAPS scales (consisting of different combinations of items) in an iterative procedure where measurement invariance (see next sub-section) was also evaluated.

The development of the Brief-LAPS scale involved several steps and trials. First, we removed the 5 items that many experts evaluated as unsuitable. Among the remaining 18 items, we then attempted to include only the items that were consistently assigned to the same sub-dimension by the experts. We set the threshold for a consistent assignment at >60%, resulting in 11 consistently assigned items. We used modification indices to remove 4 items after which model fit for the remaining 7 items was acceptable. However, the pool of 7 items mainly included People Substitution items (5 of 7 items), and just 1 item from General Attachment and 1 from Animal Rights/Welfare. Further, the item from the General Attachment scale that was retained was: “I consider my pet to be a friend”. This was predominantly considered a People Substitution indicator by the experts. We concluded that this content representation of the original LAPS scale was too unbalanced, and we therefore rejected it as a suitable brief scale.

We then expanded the model search with items that were not unequivocally assigned to 1 sub-dimension by the experts, thus giving 18 items as input variables to the CFA. An item removal strategy was initiated where modification indices helped flag items contributing the most to a poor model fit. Using this removal strategy, we identified several 7-item scales that exhibited acceptable model fit. However, for most of these models the subsequent assessment of measurement invariance failed to meet the cut-off points multiple times. One model, however, did meet the cut-offs for measurement invariance across almost all of the compared subgroups. Model fit for this scale was acceptable [Chi² = 69.69 (15); RMSEA=0.044; CFI=0.996; n = 2,037] and Cronbach's α was satisfactory (0.86 (95% CI: 0.85–0.87). Further, the item pool was quite balanced, as it included items from all 3 sub-dimensions without any single sub-dimension dominating the pool.

The resulting Brief-LAPS is presented in Table 8. It consists of 7 items where all factor loadings are >0.600 and commonalities are in the range 0.400–0.805. The sub-dimension that the particular item was assigned to by Johnson et al. (3) is presented in a separate column.

The evaluation of measurement invariance is laid out in Supplementary File 4 where a table with model fit for all invariance levels along with overall RMSEA and CFI, as well as ΔRMSEA and ΔCFI values is presented. We detected invariance at scalar level for the following variables: animal species, gender, age, income, household composition, perceived loneliness, UCLA loneliness scale, tangible and emotional support. Thus, composite scores can be used for such variables with limited risk of biased estimations. However, in the assessment of measurement invariance for countries, we did not detect scalar invariance. Therefore it was explored if it was possible to reach partial scalar invariance for country comparisons. Indeed, we identified partial scalar invariance for countries by freeing 8 (of 21) threshold parameters. See details in the notes to the table in Supplementary File 4.

We consider the implications of the lack of full scalar invariance between countries in the Discussion. However, when we conducted CFA and calculated Cronbach's alpha internal consistency coefficients for each country separately, model fit was acceptable [Austria: Chi² = 34.323 (14); RMSEA = 0.043; CFI = 0.996; n = 773]; Cronbach's α=0.86 (95% CI: 0.84–0.88). Denmark: Chi² = 38.834 (14); RMSEA = 0.054; CFI = 0.995; n = 607; Cronbach's α = 0.84 (95% CI: 0.82-0.86) UK: Chi²= 34.328 (14); RMSEA=0.047; CFI = 0.996; n = 657; Cronbach's α=0.86 [95% CI: 0.84–0.88)].

A 7-item Brief-LAPS scale was developed in which all 3 of the LAPS sub-dimensions are represented. Still, it should be noted that the item coverage of the three sub-dimensions is a little unbalanced. There is only 1 Animal Rights/Welfare item, as this was the only item that ensured acceptable model fit of just 2 possible items considered suitable by the experts in Study 2. There is also a slight overrepresentation of General Attachment items (cf. Table 8), but we note that one of these items (“I consider my pet to be a friend”) is mainly considered by the experts to be an indicator of People Substitution. Furthermore, one of the General Attachment items that is included (“My pet knows when I'm feeling bad”) is considered to be a General Attachment indicator by some of the experts and a People Substitution indicator by other experts. Thus, the 3 attachment domains from the full LAPS are covered in the Brief-LAPS.

To ensure that results obtained using the Brief-LAPS are compatible with those obtained using the full LAPS, we compared their associations with a number of variables, including socio-demographic factors, attitudes to veterinary services targeting pets, loneliness, and social support.

The same data employed in Study 3 were used in this analysis (see previous details). The following variables were developed for the analysis below:

Gender: (0 = male; 1 = female). Two other gender response options were available (“Other” and “I do not wish to tell”), but these were treated as missing data, since they were very rarely selected.

Age: Number of years (range 18–86).

Income: Respondents were asked about the household's annual income in the relevant currency of each country. For each country, we divided annual income into five equally sized groups. Where “I don't know” was selected, this was treated as missing data.

Household composition: whether the respondent lives alone, with 2 or more adults, or in a household where there are children.

Attitudes to advanced veterinary services targeting pets: here we used 3 items from a recent paper where the same data was used as in this study (47). The items are: “My pet should have access to the same diagnostic tests that are available to human patients”, “My pet should have access to the same treatment options that are available to human patients” and “The advanced care available in modern veterinary medicine is unnecessary—animals should not be treated in the same way as humans.” Response options ranged from 1: strongly disagree to 7: strongly agree.

Emotional and tangible support: This is a 6-item measure of social support that covers two dimensions, namely emotional and tangible support (60). We calculated composite scores for both dimensions.

UCLA 3-item loneliness scale: This 3-item scale measures loneliness (61).

Perceived loneliness: this single-item measure is used in national surveys in the UK (Community Life Survey, England 2017 to 2018: Statistical bulletin) and the Nordic/Baltic countries (62) as a brief measure of perceived loneliness.

For each of the abovementioned measures we report the strength and direction of the association with the full LAPS and with the Brief-LAPS. Depending on the type of variable used in the analysis, we used either Spearman's correlation coefficients, Pearson's correlation coefficients, or analyses of variance for categorical variables with more than 2 values.

The full LAPS and the Brief-LAPS have a different score range (Full LAPS = 0–69; Brief-LAPS = 0–21). Therefore, when reporting group differences in analyses of variance, we rescaled both LAPS measures into the same metric, so that they range from 0 (lowest attachment) to 1 (highest attachment).

The full LAPS and the Brief-LAPS are very highly correlated (Pearson's r = 0.95).

The strength and direction of associations between sociodemographic variables and the full and Brief-LAPS are presented in Table 9 (for gender, age, income). The patterns are very similar.

In Table 10, the scores on the Full and Brief-LAPS are presented across different household compositions (living alone, 2 or more adults, and households with children). In all countries, the score patterns move in similar directions for both scales across the 3 household types. Results from F-tests are also similar in that all p-values are significant.

Table 11 presents the strengths and directions of association between attitudes to advanced veterinary services targeting pets and the full and Brief-LAPS. The patterns appear very similar. In the final comparisons (see Table 12), we also find that the patterns between tangible support, emotional support, and loneliness are very similar for the Full and Brief-LAPS.

The Brief-LAPS produces associations with other variables that are very similar to those identified with the full LAPS. The variables that were used to compare the full and Brief-LAPS covered a great deal of the content of variables that are typically correlated with the LAPS in applied research (cf. Table 5).