Turkish adaptation of the Fear of Spiders Questionnaire: Reliability and validity in non-clinical samples

Abstract The rapid, objective measurement of spider fear is important for clinicians, and for researchers studying fear. To facilitate this, we adapted the Fear of Spiders Questionnaire (FSQ) into Turkish. The FSQ is quick to complete and easy to understand. Compared to the commonly used Spider Phobia Questionnaire, it has shown superior test-retest reliability and better discrimination of lower levels of spider fear, facilitating fear research in non-clinical samples. In two studies, with 137 and 105 undergraduates and unselected volunteers, our adapted FSQ showed excellent internal consistency (Cronbach’s α = .95 and .96) and test-retest reliability (r = .90), and good discriminant validity against the State–Trait Anxiety Inventory—Trait (r = .23) and Beck Anxiety Inventory—Trait (r = .07). Most importantly, our adapted FSQ significantly predicted 26 students’ self-reported discomfort upon approaching a caged tarantula; however, a measure of behavioural avoidance of the tarantula yielded little variability, so a more sensitive task will be required for future behavioural testing. Based on this initial testing, we recommend our adapted FSQ for research use. Further research is required to verify that our adapted FSQ discriminates individuals with and without phobia effectively. A Turkish-language report of the studies is included as supplementary material.

ABOUT THE AUTHORS Robert W Booth and Müjde Peker are Assistant Professors of Psychology at MEF University in Istanbul; Robert W Booth researches affective cognition, while Müjde Peker focuses on intergroup relations. Together, they study the psychopathological consequences of moralitybased isolation from society. Pınar Öztop is studying for her PhD at Plymouth University in the UK, where she works on a CogNovo project investigating collaborative creativity.

PUBLIC INTEREST STATEMENT
Spider phobia, known colloquially as arachnophobia, is very common and treatable, but many sufferers do not seek treatment. Measuring spider-related fear is important for clinicians, and also important for more theoretically focused researchers who want to understand specific fears. Unfortunately, most research into these topics has been conducted in English-speaking countries; very little has come from Turkey or its neighbouring countries. This is a problem because different cultures may experience and report fears and phobias in different ways. To facilitate such work, we have adapted a popular and effective questionnaire measure of spider fear into the Turkish language. Our adapted questionnaire performs well: we found that people scoring high on the questionnaire were less comfortable when encountering a securely caged tarantula. We hope our scale will help psychologists to assess and study spider fear in the Turkish context.

Introduction
Specific phobias represent an extreme and irrational fear, even panic, in response to a specific object or situation, or in anticipation of encountering that object or situation. Specific phobias are common, and commonly untreated (Magee, Eaton, Wittchen, McGonagle, & Kessler, 1996). Animal phobias tend to appear during childhood and are associated with strong disgust responses as much as fears of physical harm (LeBeau et al., 2010). Fears of animals, especially 'bugs' and spiders, are among the most common phobias (e.g. Chapman, 1997) with an annual prevalence of up to 5% (LeBeau et al., 2010), but also respond well to exposure-based treatment (e.g. Wolitzky-Taylor, Horowitz, Powers, & Telch, 2008). For these reasons, it is important that clinicians have tools to help them quickly identify spider-related fears. Furthermore, academic research with non-clinical samples is increasingly interested in the effects of specific relative to general fears and anxieties: for example, findings suggest that individuals high in spider fear show selective attentional bias towards spider images (Lipp & Derakshan, 2005;Watts, McKenna, Sharrock, & Trezise, 1986), and that this bias operates early (Larson et al., 2006;Mogg & Bradley, 2006) and may be followed by an attentional avoidance of the spider information (Pflugshaupt et al., 2005;Rinck & Becker, 2006). However, most of this research has originated from English-speaking or Western European nations; very little research into cognitive features of fear and anxiety has been conducted in Turkey or its neighbouring countries. This is a problem because different cultures may experience and/or express fear and anxiety in different ways (Davey et al., 1998;Prokop, Tolarovičová, Camerik, & Peterková, 2010); in particular, Prokop and colleagues (Prokop, Özel, & Uşak, 2009;Prokop, Uşak, Erdoğan, Fancovičová, & Bahar, 2011) have found differences between Turkish and Slovakian students' attitudes to snakes and parasitic invertebrates such as lice and ticks. For these reasons, an objective measure of spider fear is desirable for the Turkish psychological community, and for the broader international research community.
There are two common questionnaire measures of spider fear in English. The Spider Phobia Questionnaire (SPQ; Klorman, Hastings, Weerts, Melamed, & Lang, 1974;Watts & Sharrock, 1984) is often used because early studies found it discriminated clinically diagnosed phobic individuals from healthy, low-fear non-phobic individuals (Fredrikson, 1983;but see O'Donohue & Szymanski, 1993), but other studies found it to have poor test-retest reliability  and some have questioned its ambiguous timeframe (it asks the respondent about their symptoms without specifying when they experienced those symptoms; Packer, Bond, & Siddle, 1987). The Fear of Spiders Questionnaire (FSQ; Szymanski & O'Donohue, 1995) showed greater stability across time and good test-retest reliability in early testing (three-week r = .91, ; onemonth r = .63, Szymanski & O'Donohue, 1995), and in one study showed more reliable measurement of lower levels of spider fear (Cronbach's α = .91, as opposed to .43 for the SPQ in non-phobic control participants, , perhaps because it uses a Likert scale rather than the SPQ's true-false format. Better reliability at low levels of fear makes it more suitable than the SPQ for research with analogue as well as clinical samples. No Turkish adaptations or validations of either the SPQ or the FSQ have previously been published. We elected to adapt the FSQ, due to its superior reliability and flexibility. The resulting scale may be called the Örümcek Korkusu Ölçeği ("Spider-Fear Scale") in Turkish.
The original English FSQ consists of 18 items, assessing the examinee's thoughts, feelings and behaviours regarding spiders. The examinee is asked to rate how true each item statement is of them, on a seven-point scale. Early testing found it to have good internal consistency and test-retest reliability, and suggested it could discriminate diagnosed phobic from healthy non-phobic examinees Szymanski & O'Donohue, 1995). Furthermore, the FSQ has shown high utility: at the time of writing, Szymanski and O'Donohue's original report has over 250 citations in Google Scholar. The scale has been used as an outcome measure in treatment studies, and also in theoretical work from abnormal psychology, neuroscience and endocrinology.
The FSQ was translated by a native Turkish-speaking bilingual (P. O.), taking care to preserve the meaning of the questions while maintaining a clear and easily comprehensible Turkish. The translated scale was back-translated blind by another Turkish bilingual (M. P.); the back-translation was judged acceptable by all authors. The adapted scale retains the response scale (1-7 Likert, anchored by 1 = Definitely disagree, 4 = Agree, 7 = Definitely agree) and 18 items of the original. The items and instructions are presented in Appendix A.
We conducted three studies to test our adapted FSQ. In Studies 1 and 2a, we assessed its internal consistency, and discriminative validity against general measures of non-specific trait anxiety. In Study 2a, we also assessed its test-retest reliability. Finally in Study 2b, we assessed the adapted FSQ's predictive validity using a Behavioural Approach Test with a real tarantula. All studies were approved by the relevant institutional ethics panel, and were conducted in accordance with the Helsinki Declaration of 1975, as revised in 1983.

Study 1: internal consistency and discriminant validity
Study 1 was designed to test the internal consistency of our adapted FSQ in a large sample. We also wanted to assess whether the FSQ would discriminate specific fear of spiders from more general anxiety.
We administered our adapted FSQ to 137 Turkish speakers. To check its ability to discriminate spider fear from more general anxiety, we also administered the trait scale of the State-Trait Anxiety Inventory (STAI-T; Spielberger, Gorsuch, & Lushene, 1970). We chose the trait anxiety scale of the STAI because the FSQ attempts to measure spider fear as a stable trait rather than as a transient mood. This is more useful because, while anxiety may fluctuate, the specific stimuli which can increase anxiety for a specific individual are likely to be fairly consistent. This is especially true of animal fears, which often result from prior learning (Merckelbach, de Jong, Muris, & van Den Hout, 1996).

Participants and procedure
One hundred and thirty-seven native Turkish-speaking undergraduates (N = 46) and university open day visitors (N = 91; mean age = 21.32, SD = 3.02) participated. Participants were approached during class time or visits to the university, and asked to complete the scales. Students were compensated with course credits; others participated voluntarily. Participants were not screened for psychiatric disorders in Study 1. In the interests of collecting as large a sample as possible, participation was not restricted by age or other variables. Participants were assured their data would be confidential. Participants signed informed consent before the study, and were thoroughly debriefed. Data were collected in an informal setting, but participants were asked to not discuss or share their responses.

Measures
The adapted FSQ, as described above, was administered first, followed by the STAI-T in its Turkish adaptation (Öner & LeCompte, 1985). The STAI-T asks how one feels generally in one's life, and consists of 20 items (e.g. "I worry about unimportant things"), eight of which are reversed. Participants respond on a four-point Likert scale, where 1 represents "Not at all" and 4 represents "Completely". Like the English original, the Turkish STAI-T has good internal consistency in student and non-clinical samples (typically Cronbach's α > .8), and shows good convergent validity with other anxiety measures across different sample types.

Internal consistency
All analyses were conducted using SPSS v.22 (see Table 1 for a summary of all studies). The adapted FSQ showed excellent internal consistency in this study (Cronbach's α = .96). Item total correlations were generally acceptable, ranging from .38 to .91. Inter-item correlations were all positive, ranging from .17 to .88. The mean score was 36.21 (SD = 23.87) in this sample.

Discriminant validity
To assess the scale's ability to discriminate between spider fear and more general anxiety, we tested the correlation between FSQ scores and STAI-T trait anxiety scores. The study has a power of .95 to detect a potentially problematic correlation of .30. The STAI-T (M = 50.41, SD = 9.48) also showed good internal consistency in this sample (Cronbach's α = .87). One participant did not provide complete data on the STAI-T and is excluded from this analysis. There was a small-to-medium correlation between FSQ and STAI scores, r (136) = .23, 95% CI [.07, .39], p = .01. It is not unexpected that spider fear should be related to trait anxiety; both represent a stable tendency to respond to aversive situations with anxious states. However, the small-to-medium size of this correlation suggests that the FSQ and STAI-T are not redundant; the FSQ captures variance quite distinct from that measured by the STAI-T. Therefore, we interpret Study 1 as showing the adapted FSQ has acceptable discriminant validity.
In this initial exploratory study, the adapted FSQ performed well. We next turned to a more careful assessment of its discriminant validity, its test-retest reliability, and its predictive validity.

Study 2a: test-retest reliability and discriminant validity
Study 2 consisted of two parts. Study 2a was another correlational study, designed to further assess the psychometric properties of the adapted FSQ. A subset of the participants from Study 2a was invited to participate in additional behavioural testing in Study 2b.
In Study 2a, we assessed the FSQ's test-retest reliability in 105 undergraduates. We also further assessed its discriminant validity by administering the Beck Anxiety Inventory-Trait (BAIT; Kohn, Kantor, LeCicco, & Beck, 2008), a more recent but arguably more valid measure of trait anxiety than the STAI-T.

Participants and procedure
One hundred and five undergraduates (89 females, M age = 21.08, SD = 1.61), who had not participated in Study 1, completed the FSQ and BAIT for course credit. Of these, 77 (65 females, M age = 20.94, SD = 1.59) subsequently completed the FSQ again two to three weeks later. Data were collected en masse following class; 26 participants were absent from class two weeks after initial data collection, and their data were collected the following week. A further 28 absent participants were again absent at this time, their data are excluded from the test-retest reliability analyses only. Participants were asked about medical or psychiatric conditions; none reported a diagnosis of spider phobia. Participants signed informed consent before the study, and were thoroughly debriefed.

Measures
The FSQ was administered first, followed by the BAIT. The BAIT is a simple adaptation of the standard Beck Anxiety Inventory (Beck, Epstein, Brown, & Steer, 1988). While the standard scale samples respondents' symptoms in the past week, the BAIT samples symptoms from respondents' lives in general. It therefore measures trait anxiety as does the STAI, except that it is superior in its discrimination of anxiety from depression (Kohn et al., 2008). In Kohn et al.'s original testing, the BAIT showed good internal consistency (Cronbach's α > .85) and convergent validity with other measures of anxiety (e.g. r = .74 with Depression Anxiety Stress Scale-Anxiety). The Turkish version of the Beck Anxiety Inventory (Ulusoy, Şahin, & Erkman, 1998) lists 21 symptoms (e.g. "Fear that very bad things will happen"); participants rate their frequency on a four-point scale from 0, "Not at all", to 3, "To a serious extent".

Internal consistency and test-retest reliability
Again, the FSQ showed excellent internal consistency (Cronbach's α = .95). Participants' FSQ scores correlated very highly between administrations, r (75) = .90, 95% CI [.85, .94], p < .001. The FSQ therefore has shown excellent test-retest reliability in this study, which is important as it is designed to measure spider fear as a stable trait. The mean score was 45.77 (SD = 25.27) in this sample.

Discriminant validity
The study has a power of .88 to detect a potentially problematic correlation of .30. FSQ scores did not correlate with BAIT scores, r (103) = .07, 95% CI [−.12, .26], p = .46, supporting our conclusion from Study 1 that the FSQ is not substantially sensitive to more general trait anxiety. Note that the BAIT also showed excellent internal consistency in this sample (Cronbach's α = .94; M = 23.67, SD = 13.84). Although Study 1 found a significant correlation between trait anxiety and spider fear, the correlations from the two studies fall within each other's 95% confidence intervals, so these two studies do not greatly contradict one another. The STAI-T appears to be less able to discriminate anxiety from general negative affect than is the BAIT (Kohn et al., 2008), so the difference between these two studies' findings may indicate that the adapted FSQ correlates more with general negative affect than it does with trait anxiety. Future research should investigate this possibility since it may suggest a weakness in the adapted FSQ's discriminant validity. However, as stated in the Discussion to Study 1, since this correlation between the FSQ and the STAI-T has a modest effect size, such a potential weakness would probably also be modest in size. Szymanski and O'Donohue (1995), in their large initial testing, found that the original English FSQ decomposed into two factors, namely avoidance/help seeking and fear of harm. We compiled our data from Studies 1 and 2a to form a combined sample of 242. Factor analysis on these data suggested a single dominant factor (eigenvalue = 10.25), which explained 56.94% of the variance. Szymanski and O'Donohue's sample included a large proportion (64%) of phobic individuals, who had been recruited for spider phobia therapy studies; however, our opportunity samples did not include this proportion of highly fearful individuals, therefore fear of harm might be limited in this sample. The second factor may appear when we extend testing to clinical samples (see Section 5).

Study 2b: predictive validity
Since no formal spider fear scales exist in Turkish, it is currently impossible to assess the convergent validity of the adapted FSQ. We therefore assessed its predictive validity against participants' willingness to approach a caged tarantula, and their felt discomfort upon learning about this approach test. A subset of participants from Study 2a was invited to participate in this further testing.

Participants
Participants from Study 2a were invited to participate in Study 2b if their FSQ score was in the lowest (18-24, M = 20.75, SD = 2.20, N = 13) or highest (60-118, M = 82, SD = 17.27, N = 13) quartile for the sample. Twenty-six (24 women, M age = 20.73, SD = 1.31) participated. We selected these groups because selecting extreme scorers in this way increases the power of correlational analyses. Note that analyses are based on all 26 participants, as extreme groups sampling increases the power of correlational analyses (see Preacher, Rucker, MacCallum, & Nicewander, 2005); group mean comparisons are underpowered compared to correlations.

Materials and procedure
Participants signed informed consent before the study, and were thoroughly debriefed. Participants were greeted by the researcher, who introduced the Behavioural Approach Test (BAT; see e.g. O'Donohue & Szymanski, 1993). Participants were told that they would be physically safe during the study, and reminded that they could leave the study at any time. They were informed that the purpose of the test was to assess when they wished to discontinue the study, and asked not to do anything which made them uncomfortable.
Participants were positioned, standing, outside a closed laboratory cubicle. They were informed that on the table in the cubicle was a closed and sealed vivarium containing a tarantula. They were told that the spider was not aggressive and not lethally venomous, and could not escape; they were assured that they would not be asked to physically touch the spider. They were then asked to rate their level of discomfort on a 100-mm visual analogue scale, anchored with "I am completely comfortable" and "I am completely uncomfortable".
Our BAT consisted of five tasks; participants received one point for completing each task. The researcher described and modelled each task. The first task was to open the door to the cubicle. The second task was to place both hands flat on the table opposite to the vivarium, approximately one metre away. The third task was to place both hands flat on the table adjacent to the vivarium, approximately 30 cm away. The fourth task was to touch the vivarium with one finger, and the final task was to place one hand flat upon the ventilation holes in the vivarium lid; participants were assured the spider was unable to jump and bite through these holes. After completing as many tasks as they were able, participants were escorted from the cubicle, locking the door behind them, and asked to rate their discomfort on a second visual analogue scale.
The vivarium measured approximately 30 cm × 20 cm × 20 cm, and was equipped with a deep soil substrate, water dish and an overturned flowerpot hiding place. The vivarium contained a tarantula with a leg span of approximately 8 cm. For practical and safety reasons, a taxidermied tarantula was used; no participant expressed suspicion that the spider was not alive.

Results and discussion
Of the 26 participants, 23 completed all five tasks of the BAT. One was unable to touch the vivarium, and two were unable to lay their hand flat on its ventilation holes. Statistical analysis of such a skewed distribution is not really appropriate; however, it is worth noting that these three individuals had scored 77, 83 and 105 on the FSQ, and so were all in the highest quartile of the Study 2a sample. Importantly, FSQ scores correlated significantly with participants' visual analogue scale ratings of their discomfort before, r (24) = .47, 95% CI [.11, .73], p = .01, and after the BAT, r (24) = .50, 95% CI [.13, .74], p = .01. BAIT anxiety scores did not correlate with either visual analogue scale score, r (24) = .06, 95% CI [−.33, .43], p = .79; r (24) = .12, 95% CI [−.28, .48], p = .57. The mean visual analogue scale rating before the BAT was 15.96 mm (SD = 24.83 mm); for the low FSQ group this mean was 5.93 mm (SD = 8.70 mm), for the high FSQ group this mean was 27.67 mm (SD = 32.16 mm). The mean visual analogue scale rating after the BAT was 31.81 mm (SD = 24.49 mm); for the low FSQ group this mean was 22.36 mm (SD = 21.27 mm), for the high FSQ group this mean was 42.83 mm (SD = 24.13 mm).
These results suggest that the adapted FSQ is valid, in that it quite strongly predicted individuals' self-reported discomfort upon (safely) encountering a tarantula. Importantly, the FSQ predicted discomfort much more than did the BAIT, a measure of general trait anxiety. Unfortunately, we were unable to analyse participants' actual behavioural avoidance of the tarantula. We were surprised that so many participants completed all five stages of our BAT, but perhaps our BAT was too easy: generally such tests proceed until the examinee is handling the spider (e.g. O'Donohue & Szymanski, 1993), but we could not do this for practical (our tarantula was dead) and safety reasons (even a dead tarantula's urticating hairs can be intensely irritating). Our BAT therefore ended early, and did not place spider-fearful participants in as intensely stressful a situation as is usual in a BAT. However, this does not threaten our visual analogue scales' validity because a too easy BAT biases our study against our hypothesis. In future, we would like to repeat this study using a more challenging BAT, ideally with a live spider, or with the requirement that participants open the vivarium.

General discussion
The FSQ is a validated (and popular) index of spider-related fear Szymanski & O'Donohue, 1995). Our adapted FSQ, in our undergraduate samples, seemed to validly assess self-reported fear upon encountering a large spider and demonstrated excellent reliability; its reliability and discriminant validity scores were similar to those reported for the original English version. We therefore feel confident in recommending our adapted scale to emotion researchers seeking to measure specific fears in Turkish-speaking samples. However, it should be noted that the present studies used relatively small, non-representative samples. Most of our participants were female psychology undergraduates. Furthermore, our samples contained few participants with high levels of spider fear. Further research is required to establish the psychometric properties of the scale in larger, more representative samples, and ideally in samples of spider phobic patients. Furthermore, since the FSQ is often used as an outcome measure in treatment studies, it would be instructive to assess our adapted FSQ's ability to detect spider fear pre-and post-treatment in diagnosed spider phobic individuals. Finally, it would be instructive to assess the equivalence of the English and Turkish FSQs in a bilingual sample.
Although the FSQ was not primarily designed for clinical assessment, it may also be possible to use our scale to assess spider fear in phobic patients for more clinically oriented research. Although no scale can replace a formal diagnostic interview, our adapted FSQ may be useful in clinical settings for rapidly screening patients, since it consists of only 18 questions and is easy to understand. However, as stated above it will be important to thoroughly test the psychometric properties of the adapted scale in psychiatric populations first. In particular, it remains unclear how well the FSQ would discriminate spider phobics from individuals with high levels of more general fear, fear of harm, or disgust sensitivity. In future research, we aim to assess our scale's validity for discriminating spider phobic patients, and develop norms.