Method and Analysis
Sampling
Advertisements for the study were taken in two tranches from online Australian-based sale websites: Gumtree.com.au, Horsedeals.com.au and Horsezone.com.au. Search criteria for each website included all advertisements for equines for sale within Australia. Search string criteria are listed in the table below.
An initial analysis selected a random sample of 300 advertisements: 100 from each website from the 2.6.2020 to the 30.6.2020. The data was analysed to extract words describing equine personality. From the 300 advertisements, 408 words or compound words were extracted. Of the 408 entries, 162 adjectives or compound adjectives were identified that related to equine personality.
When contrasted with personality descriptors used in equine personality studies
A number of exclusion criteria were used to ensure the integrity of the study and to ensure the sample adhered to the study’s parameters. Given the study was designed to examine personality descriptors within an Australian context, advertisements listing animals residing outside Australia were not included in the data set. The study also aimed to gauge language used to describe a single equine’s personality as in previous research into equine personality domains (Hausberger, Muller, & Lunel 2011: Lloyd, 2008: McGrogan, Hutchison, & King, 2008). As such, advertisements containing listings of more than a single equine were not included in the study. The definition of sale is the exchange of a commodity for money (Oxford Dictionary, 2020). For homogeneity, advertisements listing equines for trades that did not adhere to the Oxford Dictionary definition were not included in the study. These exclusions included advertisements giving equines away, trades, leases and advertisements for stud services. Finally, 28 duplicate advertisements were excluded from the final analysis leaving a total of 1719 samples in the dataset.
For Personality Descriptors
The list of 162 personality descriptors identified in the initial data set were parsed to determine the prime adjectives for tokenization in the second data set and main analysis. Initial word list in Appendix A. Adjectives which did not provide information on equine personality were excluded. For example, the word amazing is defined as causing great surprise or wonder; astonishing. (Oxford Dictionary, 2020). Examples of some of the 60 excluded words included amazing, awesome, beautiful, best, cool, excellent, and exceptional. Once excluded, 102 adjectival terms remained. In order to measure the construct of personality dimensions within the online advertising space, adjectives preceded by a negation: not hot, not spooky, were excluded from the data set (seven items).
A number of words were removed due to data analysis requirements. These included 36 compound adjectives such as bottom-of-pecking-order, and craves-attention. In order to efficiently target adjectives that directly described equine personality factors, a further 19 adjectives were removed that did not correlate with personality domains identified in either Morris, Gale, & Howe, (2015) study into the factor structure of horse personality or Lloyd’s, (2008) Horse Personality Questionnaire. These included words such as smart, cheeky, charming and foolproof. A final list of 40 adjectives remained. List of words removed in Appendix B.
Outline of Study Procedure
Advertisements were taken from three public Australian online websites in two data sets between 2.6.2020 and 31.7.2020. Advertisements were in the public domain, however all identifying information of people involved with the advertisement was excluded from the datum including email addresses, phone numbers and full names. In order to identify the language used to describe equine personality advertisements were used as the source of target words. Advertisements for equines were placed by equestrian professionals including breeders and trainers and lay or amateur equine owners. The study was given approval to proceed without submission to the Human Research Ethics Committee due to the public nature of the samples.
Method of Data Collection
Advertisements were downloaded from three Australian websites: Gumtree.com.au, Horsedeals.com.au and Horsezone.com.au. For both initial and main data sets, all sampled advertisements were saved and downloaded to the researcher’s laptop.
Data Analysis
Data were analysed using LIWC (2015) a text analysis program that counts words in psychologically meaningful categories (Tausczik & Pennebaker, 2010). From the initial dataset, the 40 target words were tokenised and categorised into personality domains established in Lloyd’s (2008) Horse Personality Questionnaire of sociability, anxiousness, excitability, dominance and inquisitiveness.
Table 2
Equine personality adjectives and corresponding personality domain as per Lloyd’s Horse Personality Questionnaire (2008).
Sociability Anxiousness Excitability Dominance Inquisitiveness
affectionate anxious bombproof bold brave
friendly chilled calm bossy curious
grumpy docile cruizy confident inquisitive
personable nervous enthusiastic dominant
piggy nervy excitable pushy
smooch skittish hot stubborn
sook spooky lazy territorial
gentle terrified quiet willing
unsure sensitive
unflappable
The second data set collected 1719 advertisements categorising data on age, price, height, sex, colour and target words present.
Data were visually inspected using scatterplots, and Q-Q plots. Using the Shapiro- Wilk’s test, the assumption of normality was found to be violated for age: W(1669)= 0.94, p< .000, sex: W(1657)=0.65, p<.000, height: W(1616)=0.56, p<.000, colour: W(1715)=0.89, p<.000, price: W(1670)=.78, p<.000 and personality descriptors: W(1719)=.64, p<.000. Linearity and homoscedacity were both normal. Although normality was violated, Central Limit Theorum dictates that in a large enough sample normality can be assumed as normal (Field, 2016). However, given the violation of the assumption of normality in the Shapiro-Wilk’s test, the non-parametric test, Kendall’s tau was chosen as the appropriate method of bivariate correlation analysis. Kendall’s tau was appropriate as the data set contained a high number of items with data points with the same score or tied ranks on the personality descriptor variable (Hanna & Dempster, 2012). Bivariate correlations of personality descriptors (23 levels) and variables of age (scale), height (2 levels), sex (3 levels), colour (7 levels) and price (5 levels) were analysed using a two-tailed Kendall’s correlation test excluding pairwise cases with an alpha level of p<0.05 and correlations significant at the 0.01 level.
Significant correlations were followed with a non-parametric one-sample test split file layered by personality descriptor to identify which equine personality domains or combinations had significant effects.
Results
Descriptive Statistics
Of the 1719 advertisements sampled, 839 (50.7%) listed a mare or filly, 111 (6.7%) a stallion or colt and 706 (42.6%) a gelding. Equines are measured in hands with one hand equal to four inches (Equestrian Australia, 2020). In order to classify heights, categories were split at the 14.2hh mark in line with breed and showing standards for ponies such as the Riding Pony Stud Book Society (2020), and the Australian Pony Stud Book (2020) that state ponies must not exceed 14.2hh. 1166 (72.2%) advertisements were for equines 14.2hh or over and 450 (27.8%) advertisements listed equines 14.1hh and under. Advertised equines varied in age from 1 to 25 years with a mean age of 7.49 and standard deviation of 4.923. Equines have a broad range of coat colours which are determined by relatively simple modes of Mendelian inheritance most of the time (Rieder, 2009). Coat colour nomenclature varies culturally and between breed associations and environmental factors can modify coat colour, such as sun exposure, age, sex, and nutritional status of the animal (Neves, Schwengber, Albrecht, Isola & van der Linden, 2017). Advertised equines were classified into seven colour groups: bay, 597 (34.8%), chestnut, 387 (22.6%), other, 248 (14.5%), brown, 146 (8.5%), black, 144 (8.4%), grey, 127 (7.4%) and palomino, 66 (3.8%). Advertised price were split into five categories: $1-$999, 170 (9.9%), $1000- $4999, 971 (58.1%), $5000-$9999, 301 (18%), $10,000- 14,999, 110 (6.6%) and $15,000 and over, 118 (7.1%). Of the total 1719, 666 (38.7%) were recorded as having at least one of the 40 target personality adjectives used in the wording of the advertisement. Personality descriptor adjectives were categorised into corresponding equine personality domains as per Lloyd’s (2008) Horse Personality Questionnaire. Combinations of the five categories of sociability, excitability, anxiousness, dominance and inquisitiveness led to a total of 23 categories, frequency of which can be found in table 2. Target words appeared predominantly in categories: excitability 16.4%, dominance, 5.7% and sociability, 5.6%. The mean word count of advertisements was 119.20.
Figure 1 provides the frequency of target words in each personality domain.
Table 2
Categories of personality domain variants and frequencies
Personality domain variant Number of advertisements Percentage (%)
I will put this in
Figure 1
Frequency of target words in personality domains
I will put this in
Correlations and Relationships
Using a two-tailed Kendall’s correlation with an alpha level of p<0.05 and correlations significant at the 0.01 level, there was a negative, significant but weak correlation between personality descriptors and sex; r (1654)= -.07, p=0.001. Follow up tests using chi square test were used to determine which personality descriptor values had a significant relationship with sex. There was a significant relationship between personality descriptors in the sociability domain and sex, X2(2, N=93) = 24.00, p< .000, excitability and sex, X2(2, N=271) = 92.02, p< .000, dominance and sex, X2(2, N=89) = 38.71, p< .000, inquisitiveness and sex, X2(2, N=21) = 10.28, p= .006, sociability/excitability and sex, X2(2, N=24) = 7.00, p= .03, sociability/dominance and sex, X2(2, N=22) = 9.90, p= .007, and excitability/dominance and sex, X2(2, N=20) = 13.30, p= .001.
Using a Kendall’s correlation test there was a significant but weak negative correlation between height and personality descriptors: r(1615)= .06, p=0.004. Follow up tests were used with one-sample binomial tests to determine the personality descriptor domains in which there were significant relationship with height. The null hypothesis was rejected in the case of sociability and height, (N=92) p<.000, anxiousness and height, (N=27) p=.007, excitability and height, (N=256) p<.000, dominance and height, (N=95) p<.000, inquisitiveness and height, (N=19), p<.000, sociability/dominance and height, (N=21) p=.02, excitability/dominance and height, (N=19), p<.000, and excitability/inquisitiveness and height, (N=12) p=.009.
A Kendall’s test also found a significant but weak negative correlation between personality descriptors and price, r(1669)= .06, p=0.003. These were followed up with a one sample chi square test which found the null hypothesis was rejected with sociability and price and, X2(4, N=95) = 72.63, p< .000, anxiousness and price, X2(4, N=30) = 31.66, p< .000, excitability and price, X2(4, N=276) = 275.26, p< .000, dominance and price, X2(4, N=96) = 45.66, p< .000, sociability/excitability and price, X2(3, N=27) = 25.88, p< .000, excitability/dominance and price, X2(3, N=21) = 10.04, p< .01, and excitability/inquisitiveness and price, X2(3, N=11) = 9.00, p< .02.
Discussion
The aim of this exploratory study was to examine the type and frequency of personality descriptors used in Australian online advertisements of equines, compare and contrast these with personality domains in the existing literature and investigate correlations with age, price, height, sex and colour.
An initial analysis revealed 162 personality terms descriptive of equine personality which when refined, yielded 40 adjectival terms that were aligned with the five personality domains of Lloyd’s (2008) Horse Personality Questionnaire. The 40 terms included a range of descriptors, some aligning closely to terms used in previous scientific literature of equine personality domains (in parenthesis): inquisitive (Inquisitiveness), dominant (Dominance), excitable (Excitability), anxious, calm, nervous (Anxiousness), personable and friendly (Sociability), while others were informal, or colloquial terms: affectionate, piggy, grumpy, smooch, sook, gentle (Sociability), chilled, nervy, skittish, spooky (Anxiousness), cruizy, unflappable, enthusiastic, sensitive, hot, lazy, quiet, bombproof (Excitability), bold, bossy, confident, willing, pushy, stubborn, territorial (Dominance) brave and curious (Inquisitive). In the data set of 1719 advertisements these 40 adjectives describing equine personality were found in 38.7 per cent of samples. Thirty-two of the 40 adjectives were colloquial terms. This identifies that equine personality is frequently described in vivo in a range of terms, a minority of which are standardised in current literature in the field of equine personality. The majority are colloquial, informal and unstandardized terms that do have a relationship with established personality domains. This may reflect the limitations of previous research that have utilised NEO based assessments and small sample sizes of equine trainers asked to provide descriptors of equine personalities.
There was a significant but weak negative correlation between personality descriptors and sex, height and price. The domains which correlated with sex were: sociability, excitability, dominance, inquisitiveness, sociability and excitability, sociability and dominance and excitability and dominance. The domains which correlated with height were: sociability, anxiousness, excitability, dominance, inquisitiveness, sociability and dominance, excitability and dominance, and excitability and inquisitiveness.
The domains which correlated with price were: sociability, anxiousness, excitability, dominance, sociability and excitability, excitability and dominance and excitability and inquisitiveness.
Sociability descriptors correlated to all three variables of height, sex and price. Sociability can be described as how an equine relates to people and other equines: its friendliness, gentleness, and how affectionate it is as per the Oxford Dictionary definition of sociable: willing to talk and engage in activities with other people; friendly (Oxford Dictionary, 2020). It was used in (insert number value here) advertisements to describe an equine’s personality as can be seen in the examples in table 3.
Table 3
Given there is a correlation with sociability and these variables, we can conclude that these areas of personality are a domain which are observed characteristics by sellers. Describing and therefore understanding an equine’s sociability may be highly desirable for sellers and purchasers, to understand how the equine relates to people and how easy it is to handle.
Excitability descriptors also correlated to all three variables of height, sex and price. Excitability can be described as an equine’s response to new stimulation as per the Oxford definition of excitability: responding too readily to something new or stimulating (Oxford Dictionary, 2020). Words within the excitability variable contain adjectival terms that describe equine personality responses to stimuli. These adjectives appear across a spectrum from calm, lazy and quiet to describe an equine low in excitability, to sensitive, excitable and hot to describe equines high in excitability. Examples of the use of excitability descriptors in advertisements can be seen in table 4.
The significant correlation of excitability and sex, height and price demonstrates the importance of the personality domain of excitability to sellers and buyers. Rationales for this may include suitability for rider types, for example novice, beginner or young riders’ who have preference for equines low in excitability. This may be for safety reasons or due to limited skill levels. Other reasons for the prevalence of excitability adjectives may be for response to novel stimuli in work or competition, ie showjumping horses’ excitability in competition atmosphere. This may be of relevance to buyers and sellers seeking equines able to perform within specific contexts.
Dominance descriptors were correlated to all three variables of height, sex and price. Dominance, according to the Oxford Dictionary, can be described as: having power and influence over others (Oxford Dictionary, 2020) and the variable contains adjectives that described the expression of an equine’s dominance over people or other equines. These expressed behaviours were described with descriptors such as: bold, bossy, dominant, pushy, stubborn and willing. Examples of dominance descriptors used in advertisements can be seen in table 5.
Table 5
The significant correlation of dominance descriptors and sex, height and price demonstrates the importance of adjectives describing a suite of temperamental attitudes and behaviours demonstrated by equines and listed in online advertisements. The dominance adjectives varied across a spectrum of behaviours from describing equines low in dominance (willing) to neutral (confident, bold) to high in dominance (bossy, dominant, pushy, stubborn, territorial).
Implications
Humans seek to understand and codify equine personality in various ways. Previous studies have developed a number of domains into which equine personality can be divided (Lloyd, 2008,). In the online world, the language used to describe equine personality appeared to support domains found in Lloyd’s 2008 study. But the question this study sought to explore was, how is equine personality described in adjectival terms online and were there correlations with age, sex, height, price, colour and sex? The answer was that there were correlations with sex, height and price, and these were within the domains of dominance, sociability and excitability. So what does this mean? A correlation with these variables – even a weak one – illustrated the import of language of dominance, sociability and excitability to sellers and buyers.
These are the domains which appeared most frequently. Hence, how equines relate to people and other equines, their level of dominance or alternatively submissiveness, and the level of excitability are the personality domains of primary importance to buyers and sellers. Further research could examine buyer preferences for these personality domains and qualities. Furthermore, the preference for these personality domains may be examined among specific groups or types of equines for example buyers seeking equines for beginner riders or children.
The sellers who composed the advertisements used many more adjectives to describe equine personality than those provided by the trainers in Lloyd’s 2008 study. This study’s initial data set found 162 adjectival terms to describe aspects of equine personality that was honed down to a list of 40 for analysis. In vivo, language describing equine personality contains greater variety, depth and heterogeneity than previous scientific studies. Language may be colloquial, cultural or provide finer details of equine personality domains. Further research into language used by owners, trainers and breeders via web advertising and online platforms may continue to illuminate elements of equine personality not previously studied. The way people describe their equine’s personality may be an area of qualitative research in the future – an area which has received little to no attention previously.
Finally, is the subject of standardisation. If it is accepted that dominance, excitability and sociability are of importance to buyers and sellers, adjectives used to describe equine personality need to be used in a standardised form in order to convey the same meaning to all buyers and sellers and to measure constructs evenly. Establishing a standardised set of terminology describing equine personality would allow buyers and sellers to use a frame work of homogenous adjectives.
Limitations
Current linguistic software has limitations within academic applications. Tokenisation of language is limited to single words or phrases. The simplistic word count approach fails to recognise the context of each word or the meaning with which it has been applied. While software packages such as LIWC can apply psychological factors to text they are yet to be able to analyse complex sentence structures. Many linguistic packages are designed for web-based applications and for marketing, with a simplistic approach of whether a word appeared rather than the context of that word. Linguistic software packages and their current limitations make studies such as the current one difficult to conduct and take many research hours.
Without a current standardised language for equine personality, it is unknown if the traits identified within the online advertisements are reflective of the actual personalities of the equines listed. While this is an area of further research, it is also a limitation of the current study.
Conclusions
Within online advertising, language used to describe equine personality is diverse. Unlike in previous scientific literature, it is not restricted to a limited range of descriptive adjectives. It includes a range of colloquial and informal terms not previously identified. However these terms can be aligned with the equine personality domains previously recognised by Lloyd (2008) and support the idea of stable personality domains. Weak but significant correlations were identified between height, sex, price and personality descriptors. Further examination identified the domains of excitability, sociability and dominance as the most prominent domains in which personality descriptives appeared indicating their importance to both buyers and sellers.