DEVELOPMENT AND PILOT TESTING OF THE IMPLICIT MEASURES USED IN THE STUDIES OF THIS THESIS
6. The Implicit measures of this thesis
6.3. The Sentence Judgment Tasks
Four SJTs with context sentences as primes were designed to tap into the ITs of Opposite sex is dangerous, Relationship entitlement, General entitlement, and Normalisation of relationship violence, using E-Prime 1.1 software.
The task designed here is based on Baldwin, Fehr, Keedian, Seidel, and Thomson’s (1993) similar task which they used to investigate people’s expectations about others, according to their attachment style. They found that participants with a secure attachment style were faster to identify positive outcome words, while participants with insecure attachment style, responded faster to negative outcome words. Keown et al. (2008a) also applied this task to investigate child sexual offenders’ cognitive distortions. They found, however, that child sexual offenders did not differ from controls. The SJTs of this thesis differ slightly from the tasks used in the two above studies and this is explained later in this section.
In the present task, two of each sentence’s possible endings are both plausible, but if the participant holds the specific IT, then one of them is more expected, and response should be faster. The third word is an inappropriate ending of the sentence in terms of meaning and serves as noise. Participants are instructed to decide if the word that follows the sentence stem completes it in a way that makes sense or not. The choice of a real but inappropriate word rather than a non-word (as in a lexical decision task) was used in order to make sure that
participants actually read the sentence and do not just respond to the lexical identity of the target word.
6.3.1. Generation of sentences and target words. For each one of the four ITs an initial large pool of sentences was created, after a review of the literature on the cognitive correlates of IPV and an analysis of the content of items included in standard questionnaires used in IPV research which assess relevant cognitions and offence-supportive beliefs. In order to ensure that the sentences created do, indeed, reflect the concept they were supposed to assess, a pilot study was conducted. Twenty-two undergraduate and postgraduate psychology students from the University of Birmingham (16 female and 6 male) were given a booklet containing a description of each IT, 161 sentences, and rating instructions. The sentences were given complete, with an IT-consistent word. Participants were instructed to indicate which IT each sentence described best and could be an exemplar of it. If participants thought that a sentence belonged to more than one ITs, they were instructed to indicate that, and then order the relevant ITs from the most to the least closely related. A sentence was assigned to one IT if there was an agreement between at least 70% of the raters. There was agreement in the categorisation of 129 (80%) sentences. Of the remaining 32 sentences, 11 were very problematic and were discarded. The remaining 21 sentences where rephrased or altered and were given to six other raters who agreed on the categorisation. Finally, 120 sentence stems were chosen, 30 for each SJT (see Tables 2 to 5 in Appendix B).
Three word completions were assigned to each sentence stem. Two of these words were appropriate completions in terms of meaning: one IT-consistent and one IT-inconsistent (see Tables 2 to 5 in Appendix B for the sentence stems and their completions, and Tables 6 to 29 in Appendix B and for the words’ lexical characteristics for each version of the task). A dictionary-thesaurus was used to identify the sentence completions and a large number of
words and their synonyms were selected in order for each sentence to have alternative endings (still IT-consistent or IT-inconsistent). This was necessary because in the last phase of the design the words in each condition (IT-consistent and IT-inconsistent) should not differ in terms of frequency and length, and therefore alternative endings with different lexical characteristics should be available. The third word ended the sentence in a way that did not make sense (inappropriate word), and random real words were selected for this list. A real but inappropriate word, rather than a nonword (as in a lexical decision task) was used in order to make sure that participants actually read the sentence and did not just respond to the lexical identity of the word. This is one of the differences between this task and the task in Baldwin et al.’ s (1993) and Keown et al.’s (2008a) study, and therefore participants have to decide if the word that followed the sentence stem completed it in a way that made sense or not, instead of responding to its lexical identity. For this reason this task was named SJT.
The four SJTs are administered as one task. The task has three different versions and participants are randomly assigned to each one. There are three versions of the test for male participants and three for female participants. The same ITs are assessed in the male and female versions. Male and female versions have the same sentence stems, with two
exceptions: they differ in one sentence stem in the Normalisation of relationship violence IT, and in six sentence stems in the Opposite sex is dangerous IT, for which the literature
indicates that there are some hostile attitudes and beliefs associated only with men or women (Yodanis & Straus, 1996). The endings of the sentence stems is counterbalanced across the three versions (for example, for the first sentence stem, version 1 has an IT-consistent word ending, version 2 has an IT-inconsistent ending, and version 3 has an inappropriate ending).
Within each SJT version the two word lists (IT-consistent and IT-inconsistent) do not differ in terms of frequency and length (see Table 30 in Appendix B). This also applies to the
word lists of the IT-consistent (see Table 31 in Appendix B) and IT-inconsistent (see Table 32 in Appendix B) conditions across the three versions of each SJT. In order to achieve this, for some sentence stems across the male and female versions, different but similar in meaning words had to be used. For example the sentence in the IT-consistent condition of the Normalisation of relationship violence SJT The idea that violence in the relationship is sometimes acceptable, is… has the word absurd in the male version and the word crazy in the female version. Additionally, 13 sentence stems instead of having one IT-consistent, one IT- inconsistent, and one inappropriate word completion across the three different versions of each SJT, they took, for example, an IT-consistent ending in the two versions and an
inappropriate ending in the third version, or an IT-consistent in one of the versions and an IT- inconsistent in the other two versions etc. This was necessary in order to keep consistency in the mean frequency and length of the target words across the three different versions of each SJT. These sentences can be seen in Tables 2 to 5 in Appendix B, where next to their word completions it is indicated whether they were used as IT-consistent, IT-inconsistent, or inappropriate endings, more than once or not at all. RTs to inappropriate words are not of interest, and therefore, these words were not matched to the other two word lists. Lexical statistics were obtained with the N-Watch program (Davis, 2005).
6.3.2. Procedure of the SJTs. The four SJTs are administered as one task. Before the beginning of the task, participants read on-screen instructions which explain in detail what they are required to do. Additionally, they are informed them that they might read material which they might find distasteful, and they are instructed not be distracted by this, as their task is not to judge the content and appropriateness of the sentences, but to decide on whether the target word is a meaningful ending to the sentence stem. Speed and accuracy are also highlighted.
The task begins with six practice trials, with target words that fit or do not fit the prior sentence stems. These sentences are irrelevant to any of the ITs. In the main task the
sentences are presented randomly and with no replacement by E-Prime software. Each sentence stem is presented only once. Participants read each sentence stem at their own pace and press the space bar when done. A fixation cross is then presented in the middle of the screen for 1,000 ms, superimposed by the target word which remains on the screen for 1,000 ms or until a response is given. A 1,000 ms sentence-word interval was chosen in order to maximise the sentence effect on the words, and this is another difference with Keown et al.’s (2008a) study where the target word was presented immediately after the sentence stem. Participants receive visual feedback (1,000 ms; ‘correct’, ‘incorrect’ or ‘please respond faster’), and additionally, they hear a beep sound for every wrong or late response. Both the sentence stem and the target words are presented in the middle of the computer screen in white lower case letters (except for the first letter of the first word of the sentence stems), in 18 point Courier New font. The colour of the background is black.
The instructions request participants to read each sentence carefully at their own pace, press the spacebar with their left hand when they finish, and then decide if the word that appears in the middle of the screen, after the presentation of the sentence stem, is a word that completes the sentence in a way that makes sense or not by pressing ‘K’ with the right-hand index finger if the target word makes sense, and ‘L’ with the right-hand middle finger if it does not. It is also emphasised that both speed and accuracy are important.
6.3.3. Scoring of the SJT. Only the correct responses’ RTs are considered in statistical analyses. Similarly to the GNAT, a difference score is computed for each
participant, for each one of the four SJTs. Each participant’s mean RT in the IT-inconsistent condition is subtracted from the mean RT in the IT-consistent condition. Positive scores
indicate faster responses in the IT-inconsistent condition (non offence supportive cognition) and negative scores indicate faster responses in the IT-consistent condition (offence
supportive cognition). This task is based on the assumption that the sentence content will facilitate recognition of those target words which complete the sentences in a way congruent with the individuals’ attitudes and beliefs.