Assessing Temporal Relational Responding in Young Children

Relational frame theory (RFT) sees temporal relational responding (e.g., A is after B; B is before A) as a key operant skill involved in the understanding of time. From this perspective relating events temporally is important for everyday life situations such as sequencing events, planning, and talking about the past or future. The aim of the present research was to assess performance on a test of temporal relational responding in young children at increasing levels of complexity. Twenty-five typically developing children between 3 and 8 years were assessed on tasks of nonarbitrary (i.e., based on physical events) and arbitrary (i.e., based on contextual cues only) temporal relations. Results showed a correlation between overall performance across temporal relational responding tasks and age. Performance on nonarbitrary “before” and “after” trials improved similarly with age whereas with arbitrary relations, participants performed much more poorly on “after” trials than on “before” trials and some interesting cohort specific patterns were also seen. Implications of the results and future research directions are discussed.

Acquiring an understanding of how events are temporally related to each other, both with respect to actual experienced events as well as with respect to events in a more abstract sense (e.g., as represented in a calendar) is critically important in young children's development, allowing them to communicate more precisely about events in their world and to advantageously organize their behavior through sequencing, planning, problem solving, and time management. Examining how temporal understanding is acquired in childhood may help inform practical intervention to teach such skills when they are deficient.
Research on this area with children has mainly stemmed from cognitive developmental and psycholinguistic perspectives. Clark (1971) examined comprehension of "before" and "after" in 3-to 5-year-old children by requiring participants to perform sequences of actions with toys when instructions using these terms were given. Findings indicated that comprehension of both terms improved with age and that based on error patterns, children appeared to acquire an understanding of "before" earlier than "after." Clark (1971) reported that when unsure of the meaning of the temporal term, the children appeared to use an orderof-mention strategy-responding as if the order of events matched the order of mention in the sentence. Using an order-of-mention strategy would suggest that less errors would likely occur with sentences in which the actual order of events and the order in which they are mentioned match (e.g., "The girl kicked the ball before she patted the dog," or "After the girl kicked the ball, she patted the dog") compared to sentences in which the event order and the order of mention do not match ("Before the girl patted the dog, she kicked the ball," or "The girl patted the dog after she kicked the ball").
More recent research has extended findings on "before" and "after" comprehension in children. McCormack and Hanley (2011) showed 3-to 5-year-olds pairs of video clips of two noncausal actions in a particular order (e.g., brushing teeth, then washing face) and the experimenter said a sentence using either "before" or "after" describing the sequence. Participants were then asked to point to the video pair on the screen that either matched or not with the sentence they had been given. Children performed better with matching-sentences and only 5-year-olds performed above chance levels with nonmatching sentences. Blything 1 3 et al. (2015) sought to determine the age at which children begin to shift towards relying on "before" or "after" terms themselves as cues for temporal order instead of relying on background knowledge of events or an order of mention strategy. They showed young children sequences of an animated character performing two actions while the experimenters manipulated the sentence structure that narrated the order of the actions. Children were then asked to select the picture of the activity that came first. Results suggested that 3-to 4-year-olds performed better on "before" sentences compared to "after," whereas sentence variation and performance on a test for working memory (digit-span task) predicted performance in 4-and 6-years-olds. The authors noted the difficulty of discriminating effects of memory from those of language development.
The production of temporal terms has been shown to appear as young as 2 years, although the comprehension and accurate use of these terms gradually improves over childhood (Busby Grant & Suddendorf, 2011;French & Nelson, 1985). Errors in comprehension of these terms in particular contexts have been demonstrated up to early adolescence. For instance, Pyykkönen and Järvikivi (2012) tested understanding of multiple events in children between the ages of 8 and 12 years. Children were given a questionnaire on order of events that included a variety of sentences depicting simultaneous or sequential events although the temporal terms' sentence placement varied across trials. Results demonstrated that the children's accuracy on comprehension questions about the order of events was affected by the position of the temporal terms "before" and "after" within the sentences. Although performance generally increased with age, even the 12-year-old children did not perform to levels of accuracy shown by an adult comparison group.
Some developmental theorists have pointed to the acquisition of language as critical in a child's acquisition of temporal understanding. For example, Nelson (1996Nelson ( , 2007aNelson ( , 2007b has argued that it is through verbal interactions, such as by sharing past and future experiences or through direct instruction of conventional time systems, that children learn how to temporally organize events and reference themselves in time. Moore et al. (2014) evaluated Nelson's interpretation of the relation between time concepts and language and intellectual skills in an aspect of their study with children between 5 and 10 years of age. Participants were tested on time concepts and their performance compared with that on standardized measures of verbal and intellectual abilities . A variety of tasks were used to evaluate temporal event ordering abilities such as, placing cards depicting events on a spatial layout indicating how near or far they were from the present, labeling time concepts, sequencing picture cards of events in forward and backward order, and answering questions about the relative order of related events. Results showed that performance on the temporal cognition tasks were significantly predicted by verbal abilities, thus providing empirical support for a relationship between language ability and temporal understanding.
It is critical to note a limitation of the current literature: minimal research has been done investigating the environmental variables that influence the acquisition of temporal understanding. Although some research has shown that feedback and other instructional arrangements can improve understanding of temporal order in sentences (Amidon & Carey, 1972;Ehri & Galanis, 1980), studies are largely descriptive in nature. Turning towards a behavioral approach may assist with addressing this limitation on training these skills due to an emphasis on both prediction and influence of behavior (Biglan & Hayes, 1996). One particularly relevant approach in this regard is relational frame theory (RFT; Hayes, Barnes-Holmes, & Roche, 2001a;Stewart, 2016), a contextual behavior analytic account of human language and cognition. Substantial empirical research now attests to the efficacy of RFT as a basis for teaching basic and complex language skills to children and adults with and without disabilities (Gilroy et al., 2015;Kirsten et al., 2022b;Ming et al., 2018;Newsome et al., 2014;O'Connor et al., 2017;Tarbox et al., 2011;Wymer et al., 2016).
RFT suggests that arbitrarily applicable relational responding (AARR) is a key generalized operant underpinning human language and cognition. Relational responding refers to responding to one stimulus in terms of its relation to another stimulus. Nonarbitrary relational responding (NARR) is based on the formal properties of the stimuli (such as selecting a stimulus because it looks larger than another stimulus) and has been seen in many nonhuman species as well as in humans (Giurfa et al., 2001;Harmon et al., 1982;Ming et al., 2018). AARR, in contrast, is based not on the formal properties of the stimuli being related but instead on contextual cues that specify the relation irrespective of formal properties; for example, if I am told that coin A is worth "more than" coin B then I will treat coin A as larger in value and preferable to coin B even though they are similar in size. AARR has thus far been shown unequivocally only by humans (Dugdale & Lowe, 2000;Hayes, 1989;Lipkens et al., 1993). RFT suggests that humans show a variety of different patterns of AARR (frames) including coordination (sameness), distinction (difference), opposition, comparison (more/less), spatial relations (e.g., above/below), temporality (e.g., before/after), hierarchy (e.g., member of/class of), and deixis (Hayes, Fox, et al., 2001b). Framing allows the child to derive novel untaught relations between events (e.g., if told that A is more than B and B is more than C, I could derive that A is more than C and C is less than A) and hence is key to acquiring generative language (Stewart et al., 2013). From both theoretical and practical standpoints, RFT studies have provided substantial evidence that AARR is closely linked with language and cognitive abilities (Barnes et al., 1 3 1990; Barnes-Holmes et al., 2005;Cassidy et al., 2011;Cassidy et al., 2016;Colbert et al., 2018;Devany et al., 1986;Dickins et al., 2001;Hayes & Stewart, 2016) and that it is possible to establish a variety of different patterns of AARR or frames (Barron et al., 2019;Belisle et al., 2020;Berens & Hayes, 2007;Dunne et al., 2014;Kirsten et al., 2022a;Mulhern et al., 2018;Rehfeldt & Root, 2005).
Although there are a variety of types of relational framing, all frames share three core properties. Mutual entailment refers to a bidirectional relation between two stimuli, such that if a participant is taught that A is before B, then they derive that B is after A. Combinatorial entailment refers to the combining of relations, allowing for new derivations. For example, if a participant is taught that A is before B and that B is before C, then they may derive that A is before C and that C is after A. The third property of relational framing is the transformation of function, in which the psychological function of a stimulus participating in a derived relation is changed or transformed based on the other stimuli to which it is related as well as the nature of the relation. For instance, if stimulus event B has an appetitive function and is derived as being after event A, then the functions of A may be transformed such that anticipatory responding occurs during this event.
From an RFT perspective, children acquire particular patterns of AARR through multiple exemplar training (MET), which involves exposure to a variety of exemplars in which the stimuli and/or situations differ but in which the specific relational cues for that pattern remain the same. Through a history of such reinforcement from their socioverbal community, children abstract the relevant relational cues that indicate how stimuli are to be related (Hughes & Barnes-Holmes, 2016) and thus acquire generalized patterns of AARR.
In addition, although NARR and AARR can be described as distinct classes of behavior, acquisition of AARR is typically preceded by and supported by the acquisition of NARR (see Stewart & McElwee, 2009). In other words, the child might first learn to relationally respond in the presence of particular contextual cues based primarily on nonarbitrary properties of the stimuli involved but with sufficient exemplars, their relational response can come to be applied to arbitrary stimuli irrespective of the physical properties involved. Indeed, there is now empirical evidence in the case of a number of frames showing that assessing and/or training NARR can promote acquisition of AARR (Berens & Hayes, 2007;Zygmont et al., 1992).
With respect to temporal AARR, as in the case of other relational frames, RFT suggests that this repertoire is established through MET and also that children learn temporal NARR before learning AARR. In particular, RFT suggests that children likely first learn to relationally respond with respect to events actually experienced in time (NARR) and subsequently learn to derive temporal relations between events without having to actually experience them; for example, if given two hypothetical events A and B, and told that A is after B, they could derive that B is before A (AARR; Hughes & Barnes-Holmes, 2016).
With regard to RFT research on temporal relational framing, O'Hora et al. (2004) provided the first relevant study, which was focused on relational responding in the context of instructional control. They trained and tested 12 university students on a computer-based relational responding protocol that involved "before/after" and "same/different" relations. Two arbitrary symbols were trained as contextual cues for "before" and "after" by reinforcing participants' selection of a statement that included one of the arbitrary symbols that accurately described an observed sequence of shapes. Participants were subsequently tested for instructional control on a task that involved pressing computer keys in a particular order specified by the arbitrary relational networks involving the "before/after" and "same/different" cues. Eight of the 12 participants showed responding in accordance with the arbitrary relational networks and participants who demonstrated instructional control in a second experiment also showed generalized responding to novel stimulus sets. A set of follow-up studies (O'Hora et al., 2005(O'Hora et al., , 2008 extended this work by showing that performance on the relational responding protocol correlated with performance on standardized measures of intellectual ability in university students. Other research relevant to temporal relational responding has focused on comparing accuracy and response latencies to arbitrary "before" and "after" cues on tasks of nonarbitrary sequenced responding. Overall, studies have shown that adults respond slower on "after" trials than on "before" trials (Hyland et al., 2012(Hyland et al., , 2014 and with less accuracy (Brassil et al., 2019;McGreal et al., 2016).
The studies just reviewed provide useful investigations of temporal relational framing; however, these studies have focused on typically developing adults with already developed relational framing repertoires. Research on the acquisition of temporal relations in young children without such advanced repertoires on the other hand is almost nonexistent. Thus far, just one study has looked at temporal framing in children. Kirsten and Stewart (2022) examined a number of different varieties of relational framing in a cross-section of young children between 3 and 7 years and in doing so provided some evidence that temporal framing may emerge later in development than other types of framing. However, one of the limitations of this study was that, because it involved surveying performance of young children on a number of different frames, the scope of testing in the case of particular types of framing was limited. As such, these data on temporal framing in young children are perhaps not as informative as they might otherwise be.

3
Despite the limitations of past research on temporal framing, however, this work has provided at least a starting point for further exploration, such as by evaluating temporal relations in different populations. The aim of the present study is to investigate performance on tasks of temporal relational responding in young, typically developing children between the ages of 3 and 8 years. The present study utilized an RFTbased protocol to assess temporal relational responding at increasing levels of complexity. This involved testing children on tasks of nonarbitrary temporal relations and arbitrary temporal relations, including assessing for properties of mutual and combinatorial entailment, and transformation of function. We intended that this exploratory study would be an initial step that might help facilitate future research to examine the acquisition of temporal relational responding in more detail and to design appropriately targeted training procedures if this repertoire was found to be deficient based on the level of performance typically observed at a given age.
The present study, which was conducted online, utilized adaptations of existing RFT methods to assess temporal relations. The relational evaluation procedure (REP) is an effective and efficient method to assess various patterns of relational framing (Cassidy et al., 2011(Cassidy et al., , 2016. The REP involves presenting a particular contextual cue (e.g., SAME) in between two or more arbitrary stimuli on the screen (e.g., "QUB is the SAME as TIG") and asking participants questions about the presented relation (e.g., Is TIG the SAME as QUB?"). Kirsten and Stewart (2021) recently adapted the REP for younger children who were not yet readers or had minimal reading skills by using colored shapes instead of nonsense words and single letters as contextual cues instead of full words, alongside audio options (e.g., [Red circle] [S] [Blue circle] to indicate that "red is the same as blue."). A similar format was used in the present study as the age of participants was comparable. In addition, the present study adapted the test for instructional control used by O'Hora et al. (2004) to test for a transformation of function effect in accordance with temporal relations. Instead of using computer keyboard presses in an order specified by a relational network, observable and discriminable motor actions were included. We anticipated this as an advantageous modification given the ages of participants and as the study was conducted online.

Participants
Twenty-five typically developing children (13 female, 12 male) between the ages of 3 and 8 years (M = 5 years 6 months, SD = 17.87 months) participated in the study. These included five 3-year-olds (M = 42.49 months, SD = 3.63 months), six 4-year-olds-olds (M = 56.26 months, SD = 3.77 months), four 5-year-olds (M = 66.35 months, SD = 1.56 months), four 6-year-olds (M = 77 months, SD = 3.82 months), five 7-year-olds (M = 87.92 months, SD = 3.23 months), and one 8-year-old (100.5 months). All participants were recruited through personal contacts in the United States and Ireland. Ethical approval for the study was first obtained through the institutional research ethics committee prior to conducting the study. Verbal and written consent was obtained from each participants' caregivers and verbal assent was obtained from the children themselves before starting a session.

Setting and Materials
All sessions took place online through the use of a video conferencing platform with end-to-end encryption. Sessions were conducted at times preferred by each family and with the child and an accompanying family member participating via a networked device in a suitable room in their living residence. All materials were created in Microsoft Power-Point and displayed on a 13-inch MacBook Pro laptop via the screen share feature of the video-conferencing platform.
Each child was tested on an initial preassessment and a temporal relations assessment (TRA). The preassessment was used to test the child's ability to vocally label all stimuli used in the TRA, answer yes/no questions, and follow single and multistep directions. The TRA included two subtests: nonarbitrary temporal relations and arbitrary temporal relations. The nonarbitrary subtest materials included images of eight different common items (e.g., an apple, ball, cat). The arbitrary subtest included images of four different colored circles (blue, green, red, and yellow), as well as four images of simple motor actions (clapping, raising arm, touching nose, and waving) (see Procedure for complete description of materials used across each test).

Procedure
At the onset of each session, the researcher first tested the audio, video, and screen sharing features with the participant and their caregiver to ensure all features functioned properly before proceeding. Caregivers were instructed that the purpose of the test was to explore how children would respond to the presented activities and that they as caregivers should not provide answers or guidance to the children at any point during the assessment. Caregivers were asked to intervene only in the event of problems with the audiovisual display. For all test sections, the researcher first waited for participants to demonstrate an observing response (e.g., looking at the screen) prior to beginning the trial. It was determined that if any break in attending to the stimuli occurred (e.g., lag in video call, participant looking away), the researcher would regain the participant's attention by stating their name and asking if they are ready to see the example again before proceeding. Once attending to the stimuli was reestablished, the trial presentation for each test section continued as described.

Preassessment
During the first session, children were first administered the preassessment to determine suitability for inclusion in the study. The preassessment included three stages: tacting the stimuli used in the TRA, answering Yes/No questions, and performing simple motor responses. To test for tacting of the TRA materials, the child was shown common items or colors (one stimulus on the screen at a time) and asked by the examiner on each trial, "What is this?" A correct response was scored if the child accurately vocally labelled the stimulus shown on the screen (e.g., said "cat" when an image of a cat was shown). A total of 12 trials was conducted consisting of eight pictures of common stimuli that were used in the nonarbitrary section of the TRA, and images of four colored circles that were used in the arbitrary section of the TRA. The child was then exposed to the same stimuli but tested on their ability to answer Yes/No questions. The examiner showed one stimulus on the screen at a time and asked the child a Yes/No question such as, "Is this an apple?" A total of 12 questions was presented, 6 to which "yes" was the correct response and 6 to which "no" was the correct response.
Following this, children were tested on following single and multistep instructions with four different motor actions (clap hands, raise arm, touch nose, wave) relevant to the transformation of function section of the TRA. A total of 6 trials was presented, including each of the four single actions, one probe for two-step actions, and one probe for three-step actions. In the case of each of the individual actions, the child was shown an image on the screen depicting that action and asked to demonstrate it (e.g., "Show me clapping"). A correct response was scored if the child acted out the action specified. If an error occurred, the researcher provided a model of the action and presented another opportunity to correctly perform the action. After each individual action was correctly demonstrated, the child was then asked to perform more than one (i.e., either two or three) of the four actions in succession. A correct response was scored if the child performed each of the actions stated by the examiner but they were not required to perform them in a particular order. In total, the preassessment consisted of 30 individual trials. Participants were required to score 11 out of 12 trials correct on each the tacting and Yes/No sections, and 5 out of 6 trials correct on the motor actions section (total criterion of 27 out 30 trials correct, 90% correct) in order to continue with the study. During initial recruitment, one child did not pass the preassessment and was therefore excluded from the study. All the remaining children (n = 25) passed the preassessment.

Temporal Relations Assessment (TRA)
The Temporal Relations Assessment included two major subsections: (1) nonarbitrary temporal relational responding (i.e., responding under the control of physically temporally related events) and (2) arbitrarily applicable temporal relational responding (i.e., responding under the control of contextual cues that specify temporal relations between events). The assessment was delivered using a combination of visual materials, auditory recordings, and orally presented instructions. For each trial, the particular visual pictures and/ or words relevant to answering the question were presented to the participant. Once the materials were displayed, the principal researcher then vocally presented the predetermined question.
Following completion of the preassessment, children were assessed on the TRA over the course of one to two sessions that lasted in total between 20 and 45 min per child. Session durations varied depending on breaks given during testing and the age of the child. The TRA was used to test responding to temporal relations at nonarbitrary and arbitrarily applicable levels across a total of 64 trials. Children were first tested on 16 trials of nonarbitrary relations, consisting of eight questions with two-image sequences and eight questions with three-image sequences. The testing of arbitrarily applicable relations was carried out across 48 trials, consisting of 16 questions for mutual entailment, 16 questions for combinatorial entailment, and 16 questions for transformation of function (8 questions for transformation of function with mutually entailed relations and 8 questions for transformation of function with combinatorial entailment relations). For all trials, each correct response was scored as "1" whereas an incorrect response was scored as "0." Throughout all sections of the assessment, no feedback was provided following correct or incorrect responses. In order to maintain engagement, nonspecific verbal praise was provided on a schedule of around every four trials (e.g., "you're working really hard"). In between each section, the researcher briefly engaged the participant in conversation about preferred topics unrelated to the test. This was done to maintain a level of rapport and to provide brief breaks between tasks. In addition, after test sections the researcher asked participants if they would like a brief break before resuming or if they would prefer to continue with the activities.
Nonarbitrary Testing At the onset of the nonarbitrary section, participants were instructed that they would be shown a series of images and be asked questions about which image came before or after the other(s). Participants were instructed to give a "yes" or "no" vocal response to each question and were told that they could request to see the image sequence again if needed. The 16 trials included both two-and three-image trials and included four trial types. There were four "before" trials with "yes" as the correct response, four "after" trials with "yes" as the correct response, four "before" trials with "no" as the correct response, and four "after" trials with "no" as the correct response (Table 1). Each of the stimuli used in the nonarbitrary section appeared in all sequential positions and in all trial types (see Appendix 1 for a full list of the nonarbitrary trials used).
In the two-image nonarbitrary relations trials, a participant was shown a single stimulus (e.g., fish) on the screen for 1 s with a simultaneous audio clip producing the name of the item. The image was then removed and a blank screen was shown for 1 s, and then a second stimulus was displayed on the screen for 1 s (e.g., shoe) again with an accompanying audio clip, and then this second image was removed and a blank screen shown again (Fig. 1). The experimenter then presented a question regarding the order of the observed sequence using either a "before" or "after" cue (e.g., "Was the shoe before the fish?"). The same procedure was used for trials of three-image sequences with the only difference being the addition of a third image. That is, an image was shown on the screen for 1 s accompanied by an auditory stimulus, followed by a blank screen for 1 s, and then a second stimulus appeared for 1 s, followed by a blank screen for 1 s, and then a third stimulus appeared for 1 s, followed by a blank screen (Fig. 1).

Arbitrary Testing: Mutual and Combinatorial Entailment
Prior to beginning the arbitrary relations section, participants were shown sample items of the contextual cues used to indicate "before" and "after" respectively. For all arbitrary trials, the letter "B" was used to cue a "before" relation, and the letter "A" was used to cue an "after" relation. The researcher presented the letter "B" on screen and said, "B means before," and subsequently asked, "What does B mean?" This was repeated by a similar procedure involving the letter "A." Participants were then presented with a sample relational statement for both of the contextual cues. In the case of the "before" cue, the researcher showed, for example, [blue circle] [B] [green circle] on the screen and read it aloud to the participant and a similar procedure was conducted in the case of the "after" cue. The researcher then presented two trials in  1 Example stimulus presentations. Note. From left to right, the figure illustrates an example stimulus arrangement for a two-stimulus nonarbitrary trial, a three-stimulus nonarbitrary trial, a mutual entailment trial, and a combinatorial entailment trial which the participant was asked to read aloud the relational statement (one trial with a "before" cue and one trials with an "after" cue). If participants failed to accurately read aloud the relational statement, the researcher modeled the correct response, reviewed the contextual cues in isolation, and then re-presented a sample relational statement for each contextual cue. Participants had to read aloud the statement accurately on a single-trial probe for each cue before proceeding. Participants were then tested for arbitrary relations. For all arbitrary trials, the relational statements were presented in a multimodal format as they were both visually displayed on the screen as well as being stated aloud. During trials of mutual entailment, the participant was shown an image consisting of two colored circles with a contextual cue in between them (Fig. 1). An accompanying audio clip was played which stated the relational statement out loud (e.g., "blue is before red"). The researcher then asked the participant a question about the relation between the A-B stimuli (e.g., "Is red after blue?"). Participants were required to give a "yes" or "no" vocal response. During trials of combinatorial entailment, participants were shown an image consisting of three colored circles with interspersed contextual cues (Fig. 1). An audio clip stated the relational statement aloud (e.g., "blue is before red, red is before yellow") and then the researcher asked the participant a question about the A-C relation (e.g., "Is yellow after blue?"). Again, participants had to give a "yes" or "no" vocal response. For both mutual entailment and combinatorial entailment sections, the 16 trials for each section were broken down into 8 unique trial types, examining both directly trained and entailed relations ( Table 2). Each of the trial types was presented twice as the stimuli were systematically rotated so that they each appeared in all stimulus positions and in all trial types (see Appendix 1 for a full list of arbitrary trials used).
Arbitrary Testing: Transformation of Function In order to assess for transformation of function (ToF), the experimenter first trained a discriminative function in each of the arbitrary stimuli. Each of the four colored circles was shown one at a time alongside an image of a particular action. For example, the experimenter showed the blue circle, and said, "This means clap," while modeling clapping. Participants were then instructed to copy the action modeled. This was repeated with the other three colors. Participants were subsequently probed across four trials (i.e., one trial per color) on their ability to demonstrate the correct action when shown the corresponding color. The researcher presented a colored circle on the screen alongside a visual that showed each color-response combination and instructed the participant, "Do this one." Social praise was delivered for correct responding and corrective feedback was used if an incorrect response was emitted. Participants were required to respond correctly to each of the four trials before proceeding to ToF testing. If any errors occurred, trials were re-presented until the participant independently demonstrated the corresponding action for each stimulus.
Once the participant had demonstrated each of the stimulusresponse combinations correctly, they proceeded to the ToF testing. The 16 ToF trials included eight mutually entailed ToF trials (ToF-ME) and 8 combinatorially entailed ToF trials (ToF-CE). For all trials, a visual sample was presented on the left side of the screen that showed each color-response combination and the relational statement was presented on the right side of the screen (see Fig. 2). During the ToF-ME trials, a relational statement with either the "before" or "after" cue (e.g., red before blue) was displayed while an audio clip stated the relational statement out loud. The participant was then told to do the actions in the correct order. If the participant demonstrated the two motor actions in the order that the relational statement indicated then this was recorded as a correct response. For ToF-CE trials, the same procedure was used with the difference that the relational statement involved three stimuli and participants were required to perform three motor actions in the order that corresponded with the relational statement. "Before" cues were used in 8 trials (four ToF-ME trials and four ToF-CE trials) and "after" cues were used in 8 trials (four ToF-ME trials and four ToF-CE trials).

Interobserver Agreement and Procedural Fidelity
Interobserver agreement (IOA) and procedural fidelity data were collected by trained research assistants for five of the

Preassessment
Age cohort results from each section of the preassessment and overall score are shown in Table 3. Within the 3-4 years cohort, each participant scored 100% correct on the tacting section. One participant from the 3-4 years group scored 92% correct (11 out of 12 trials) on the Yes/No responding section, and two participants scored 83% correct (5 out of 6 trials) on the motor responses section. All other participants scored 100% correct on all sections of the preassessment. Table 4 shows Spearman's rank correlations between age at the date of testing and scores on the TRA, including total score and subscale scores for nonarbitrary and arbitrary sections. 1 Significant correlations were found between age and total score on the TRA (r s = .72, p < .001), between age and nonarbitrary temporal relational responding (r s = .72, p < .001), and between age and arbitrarily applicable temporal relational responding (r s = .64, p < .001). A significant correlation was also found between scores on the nonarbitrary and arbitrarily applicable temporal relations subtests (r s = .77, p < .001).

Relational Responding Scores and Age
Figure 3 graphs the correlation between the total score for each participant on the TRA (out of a total of 64 possible points) and age in months. The R 2 value indicates the proportion of variance in total score that can be attributed to age (.53, low to moderate). Note that the 4-5-year-old cohort featured two of the most extreme outliers one of which was 2 standard deviations below the best-fit line. The percentage of correct responses across all sections of the TRA for each participant is shown in Table 5. Total scores on the TRA total assessment ranged from 30% to 83% correct (M = 57%, SD = 8.05). On the nonarbitrary section, scores ranged from 31% to 100% correct (M = 75%, SD = 3.5); on the mutual entailment section, scores ranged from 44% to 81% (M = Fig. 2 Example stimulus presentation for transformation of function trials  61%, SD = 2.20); on the combinatorial entailment section, scores ranged from 31% to 69% (M = 50%, SD = 1.35); and on the transformation of function section, scores ranged from 0% to 100% (M = 44%, SD = 3.58). Table 5 shows the average percentage of correct responses for each cohort both for the TRA as a whole and for each assessment section. The mean scores on the nonarbitrary temporal relations section showed a clear improvement between the youngest and oldest participants (M = 55% to 94% correct, respectively). Performance on the nonarbitrary section was similar for the 3-4 years and 4-5 years cohorts (M = 55% and 58% correct, respectively) but then a substantial improvement is seen in the 5-6 years cohort (M = 80% correct). Performance continued to improve in the 6-7 years (M = 89% correct) and ≥ 7 years cohorts (M = 94% correct). Mean scores on the mutual entailment section also showed a general trend of improvement with age, with the exception of the 6-7 years cohort, who scored the lowest out of any age cohort (M = 53% correct). However, between 3-7 years, group means on the mutual entailment section remained near chance-level responding. For the combinatorial entailment section, no clear improvement with age was apparent and the means for all age groups were near chance-level responding. Mean scores on the ToF-ME and ToF-CE sections each showed improvement between the youngest and oldest participants (M = 38%-65% correct and 28%-65% correct, respectively). The 3-4 years and 4-5 years cohorts scored similarly on the ToF-ME section (28% and 29% correct, respectively) and the ToF-CE section (28% and 21% correct, respectively). The 5-6 years and 6-7 years cohorts also scored similarly on the ToF-ME section (56% and 53% correct, respectively) and on the ToF-CE section (47% correct for both cohorts).

Age-Related Performance across Contextual Cues
An analysis of performance across temporal trial types (i.e., "before" vs. "after") was also conducted to further investigate the results found, in particular with respect to the arbitrarily applicable relations sections of the assessment. Figures 4, 5 and 6 show the average correct responses per age cohort and relational response type (nonarbitrary, mutually entailed, combinatorially entailed, and transformation of function) based on the contextual cue used (i.e., "before" vs. "after"). The NARR data in Fig. 4 refer to which word was used in the question for that section of the assessment. For mutual entailment, combinatorial entailment, and transformation of function, the "before" data in Figs. 5 and 6 represent trials in which the relational statement read, "A before B" and "A before B before C" (see trial types in top half of Table 2). "After" data represent trials in which the relational statement read, "A after B" and "A after B after C" (see trial types in bottom half of Table 2).

Age (Months)
TRA Total Score by Age 4-5 years 5-6 years 6-7 years 7-8 years 8-9 years 3-4 years 1 3 the ≥ 7-year-old cohort scored higher on "before" compared to "after" trials (M = 100% and 92% correct, respectively). For the nonarbitrary-three stimuli section (Fig. 4), the 3-4-year-old cohort scored higher on "after" trials (M = 55% correct) compared to "before" (M = 50% correct). The 4-5-year-old cohort performed better on "before" trials (M = 67% correct) compared to "after" trials (M = 54% correct). The mean scores for the 5-6-year-old cohort were the same across both trial types (M = 88% correct). The 6-7-year-old group performed slightly better on "before" trials than "after" (M = 88% and 81% correct, respectively) as did the ≥ 7-year-old group (M = 96% and 88% correct, respectively). On the mutual entailment section (Fig. 5), there was an improvement on "before" trials with age with the exception of the oldest group (M = 60%, 71%, 78%, 91% 77% correct, respectively moving from youngest to oldest). On "after" trials performance remained near chance level responding in the three youngest age groups (55%, 44%, and 50% correct, respectively) but then in the 6-7-year-old cohort the mean scores were substantially lower than had been the case for the previous three groups (M = 16%) and then performance for the oldest cohort was the highest among the cohorts (M = 65%). Regarding a comparison between performance on "before" and "after," it might be noted that difference in mean scores across "before" and "after" mutual entailment trials became increasingly pronounced from the 3-4-yearold cohort to the 6-7-year-old cohort. As regards combinatorial entailment (Fig. 5), performance on "before" trials was lowest in the 3-4-year-old group (48% correct) and somewhat better in the 4-5 and 5-6-year-old groups (67% and 63%, respectively). The two oldest cohorts both showed high levels of correct responding on "before" trials, with the 6-7-year-olds performing better (M = 97%) than the ≥ 7 years group (M = 81%). In contrast, performance on "after" trials showed a gradual disimprovement with age moving from the 3-4 to the 5-6-year-old cohorts (M = 45%, 40%, 34% respectively) before showing a much lower performance from the 6-7-year-old age group (3% correct) and then a slight increase again in the oldest cohort (21% correct). As in the case of mutual entailment trials, the difference in mean scores on "before" and "after" trials became increasingly pronounced between the 3-4years group up to the 6-7-years group.
Results on both transformation of function trial types (ToF-ME and ToF-CE; Fig. 6) show that the participants 5 years and older demonstrated high levels of accuracy with "before" trials whereas mean scores on "after" trials for the transformation of function sections remained low across all age cohorts. It is noted that only two participants (P20 and P25 who were both in the ≥ 7 years group) scored high on the "after" trials in the transformation of function sections (100% and 75% correct, respectively).

Discussion
The present study aimed to examine the development of temporal relational responding in young children across a range of ages. In particular, children between 3 and 8 years of age were tested on an assessment of nonarbitrary and arbitrary temporal relations. To date, RFT research has minimally investigated temporal relational frames and the research that has been done has primarily investigated this repertoire in adults with an existing repertoire. The present study expands on prior RFT research in this area and provides new empirical data on the acquisition of temporal relational responding patterns in children. Overall performance on the TRA was shown to increase with age. Significant correlations were found between age and total score on the assessment as well as between age and performance on both the nonarbitrary and arbitrary relations sections. Performance on the nonarbitrary temporal relations section steadily increased with age as average scores for each age group tended to be higher than for younger groups. Children 5 years and older in particular showed high levels of responding on nonarbitrary trials. Other patterns were observed on the arbitrary relations section, which included trials for mutual entailment, combinatorial entailment, and transformation of function. Performance on mutual entailment and transformation of function trials generally increased between the youngest and oldest participants, though errors were still made by the oldest participants. Accuracy on combinatorial entailment trials did not appear to improve with age as all cohorts performed near chance level responding. However, when analyzed according to trial types, notable differences in performance between "before" and "after" trials were observed (Figs. 4, 5 and 6). This pattern will be elaborated on below.
One key finding from the present study was that different patterns were observed across ages when comparing responding in accordance with nonarbitrary temporal relations and responding in accordance with arbitrary temporal relations. Namely, with the nonarbitrary section of the assessment, accuracy on "before" and "after" questions improved similarly with age (see Fig. 4). However, on arbitrary sections of the assessment, better accuracy was consistently observed across all ages on trials that included "before" as the cue for the trained relation compared to trials that included "after" as the cue for the trained relation (see Figs. 5 and 6). For example, trials that showed a relational statement such as, "red before blue" and included questions like, "Is blue before red?" or "Is blue after red?" were on average responded to with greater accuracy than trials that showed a relational statement such as, "green after yellow" with questions like, "Is green before yellow?" or "Is green after yellow?" Whereas performance on arbitrary trials with "before" as the trained contextual cue generally improved with age, the pattern of performance on arbitrary trials with "after" as the trained contextual cue was more complex. For the mutual entailment section, mean scores on "after" trials remained around chance level responding between 3 and 6 years. A substantial drop in mean accuracy on these trials was observed in the 6-7-year-old cohort (M = 16% correct) followed by a substantial improvement to the highest-level

Transformation of Function Trials
ToF-ME Before ToF-ME After ToF-CE Before ToF-CE After performance of any cohort in the ≥ 7-year-old group (M = 65% correct). For the combinatorial entailment section, a somewhat similar pattern was observed for the younger age groups to the extent that the 3-6-year-olds showed low levels of correct responding, and the 6-7-year-olds again showed a substantial drop in correct responding (M = 3% correct) compared to the 3-6-year-olds. However, in this case although the oldest cohort did show an increase in mean accuracy (M = 21% correct) compared to the 6-7-year-old group, their level of performance remained low. For transformation of function trials, children 5 years and older demonstrated high levels of accuracy with "before" trials but all ages had low levels of accuracy on "after" trials. These results suggest firstly that trials in which "after" was the stated relation were more challenging than trials in which "before" was the stated relation across all arbitrary sections. This coheres with previous work with adults which also found that participants tended to respond slower and with less accuracy with "after" relations (Brassil et al., 2019;Hyland et al., 2012Hyland et al., , 2014McGreal et al., 2016). One possible reason for why both children and adults might be less accurate with "after" than "before" relations is as follows. Temporal framing is acquired gradually, as with all frames. In particular, there may be functionally different aspects of the relational pattern to be acquired. Children might learn "before" prior to learning "after," or in general, children may receive more exposure to "before" relations than "after" relations. It is also plausible that children receive more practice deriving an entailed "after" relation based on "before" than vice versa. Further research is suggested to investigate potential differences between the acquisition of these contextual cues in children.
One specific issue affecting the acquisition of temporal relational responding that might contribute to differential performance with respect to "before" and "after" contextual control (and one previously hinted at in cognitive developmental research by Clark, 1971) is the potential of a conflict between a nonarbitrary temporal relation controlling the response and the arbitrary temporal relation controlled by the contextual cue. For instance, when a participant was told the trained relation, "red is before blue, blue is before yellow," the experienced temporal order of the stimuli in terms of what was heard auditorily (and possibly observed visually) was "red" . . . "blue" . . . "yellow." This experienced sequence coheres with the stated arbitrary temporal relation. However, when told, "red is after blue, blue is after yellow," the experienced sequence is the same but in this case, the arbitrary relation conflicts with the nonarbitrary relation. Previous RFT research has shown that such conflicts can lower levels of correct derivation of the arbitrary relation (Kenny, Barnes-Holmes, & Stewart, 2014a;Kenny, Devlin, et al., 2014b;Stewart et al., 2002).
If the relational statement is presented visually then the same potential issue arises but in that case there is also a spatial dimension involved. To illustrate, Fig. 7 shows relational statements with both "before" and "after" contextual cues with the added question, "Is red before yellow?" With the "before" relational statement, a person may correctly derive that yellow is after red based purely on the arbitrary cue, or they might also relate yellow as after red based on the spatial relation of left-to-right (because yellow is further to the right of red). Either way, there is congruence between the temporal order according to the contextual cue and the spatial order. This congruence is not the case when "after" cues are used in a statement. For example, in Fig. 7 on the right side, a person may correctly derive that red is after yellow if they respond in accordance with the arbitrary contextual cue. If, however, they are responding in accordance with a spatial relation of left-to-right, they will answer this question as though yellow is after red (because yellow is further to the right of red). In other words, there is incongruence between the order based on the arbitrary contextual cue and the spatial order.
An interaction between spatial and temporal relations may be sensitive to one's individual reinforcement history. For instance, although terms like "before" and "after" are often used for temporal relations, they are also used for spatial relations. Barnes-Holmes et al. (2001) suggested that young children's misuse of prepositions (e.g., "can I have any reading behind dinner," Pinker, 1990) might be explained by children initially developing more general relational frames, which are then shaped to be more specific through interactions in the verbal community. Prepositions for temporal and spatial relations in particular bear similarity and may initially participate in a more general frame before they are acquired as contextual cues for distinct temporal and spatial relations.
In English-speaking socioverbal communities, people read from left-to-right and children are typically taught to sequence items from left-to-right, thereby possibly Fig. 7 Comparison of relational statements using "Before" and "After" contextual cues establishing a learning history that stimuli to the left may come "before" stimuli to the right. Indeed, some recent work has shown that a directional preference (e.g., left-to-right) for ordering events gradually emerges in childhood (Tillman et al., 2018;Tillman et al., 2022). This is also arbitrary because a left-to-right order is based on social convention, and research from other domains has showed humans spatially represent time in other ways such as right-to-left (Fuhrman & Boroditsky, 2010) top-to-bottom (Boroditsky, 2001) and even according to cardinal directions like east-to-west (Boroditsky & Gaby, 2010). Further, research has shown that people can be trained to spatially represent time in different ways (Hendricks & Boroditsky, 2017). Nevertheless, if a history of reinforcement is established for reading or ordering from left-to-right, then it is likely the person will actually experience the stimuli in this order in a nonarbitrary sense. This means that they may orient to the image on the left first, and then scan towards the right. Correctly answering "after" trials is potentially more difficult, as the arbitrary temporal relation is incongruent with how the stimuli are experienced nonarbitrarily if reading from left-to-right.
If a person were to respond in this manner (based on the spatial order), this would likely produce a response pattern where questions that involve stimulus arrangements like that on the left of Fig. 7 are answered correctly and stimulus arrangements like that on the right of Fig. 7 are answered incorrectly. That is, a child could score 50% correct on all arbitrary trials by responding to the spatial order of presentation without using the "before" and "after" cues. Thus, the relational response may be under stimulus control of an observed spatial relation instead of under the control of the contextual cue. In addition, the same pattern would occur on a transformation of function trial. If relating the stimuli based on a spatial relation (i.e., further left is "before" and further right is "after"), a person may act out the actions in sequence based on the colors from left-to-right regardless of the arbitrary cue presented. Indeed, participants 5 years and older showed high levels of accuracy to ToF "before" trials, whereas nearly all failed the "after" trials. As an anecdotal example, P12 accurately read aloud the relational statement during ToF trials, (i.e., vocally stated that "blue is after green") yet performed the actions according to the spatial left-to-right order (blue . . . then green).
Apart from the difference in performance on "before" and "after" trials one other interesting aspect of the current data is the decrement in performance on "after" trials in the 6-7-year-old group in particular compared to the younger groups. For the ME trials this substantial drop in performance for the latter cohort is juxtaposed with a significant improvement in the oldest cohort. In the CE trials there is a similar drop in performance in the 6-7-year-olds that again is followed by an improvement in the oldest group, though in the case of the CE trials this improvement was not a substantial one. One possible explanation for this pattern is that the youngest groups are simply guessing (as evidenced by the fact that they are showing a correct rate of around 50%) whereas the 6-7-year-olds are responding systematically but under the wrong stimulus control. In particular, it might be that they are in fact responding under the pattern of stimulus control discussed in the previous paragraphs. For example, if participants were responding on "after" trials not in accordance with the arbitrarily applicable temporal relation but in accordance with the nonarbitrary relation between the stimuli in the trained statement (whether temporal or spatial) then they would get zero or close to zero correct. In contrast, the obvious improvement in performance on ME "after" trials in the oldest cohort could be due to the fact they have by now acquired the repertoire of appropriate derivation in accordance with temporal ME and this overrides to some extent potential influence by nonarbitrary relations. In the case of the CE relations even the oldest cohort did not show correct responding for "after" relations, suggesting that in this case this cohort had yet to acquire this repertoire.
The present study adds to the body of RFT research more broadly in a number of ways. It expands work on temporal relations to a new population and provides data for the development of temporal framing in children. It expands on the results by Kirsten and Stewart (2022) relative to temporal frames by assessing this frame in particular and adding a test for transformation of function with temporal relations. The test for transformation of function was adapted from O' Hora et al. (2004) and provides an initial approach to assessing this relational frame property with temporal relations in children. As has been discussed, the current work also highlights potential key differences in the acquisition of "before" versus "after" with respect to contextual control. Finally, a secondary benefit of the present study is that the relational assessment was carried out entirely online. As the use of online learning models continues to grow, future RFT research may find utility in further exploring online formats (e.g., synchronous, asynchronous) for testing and training of relational skills in children.
The current data also provides new insight on non-RFT based research on "before" and "after" comprehension in children. Research on the comprehension of "before" and "after" has historically been limited to examining how their placement in a sentence affects performance. The present study however, sought to examine these terms in the context of contextual control over relational responding (both nonarbitrary and arbitrary). From an RFT view, the cognitive developmental or linguistic studies that involve answering questions about observed actions coming "before" or "after" others (Blything et al., 2015;McCormack & Hanley, 2011) are likely assessing responding under the control of nonarbitrary relations, whereas studies involving questionnaires or hypothetical scenarios rely on AARR (Pyykkönen & Järvikivi, 2012). Distinguishing between NARR and AAAR offers new insight on the acquisition of temporal understanding in children as the primary source of stimulus control over temporal NARR is the actual experience of sequential change, whereas temporal AARR is under control of the contextual cue itself. In the present study, the NARR data indicated that performance with both terms similarly improved with age. The AARR data however, showed that deriving temporal relations in the presence of "before" cues was more strongly established in participants than deriving temporal relations in the presence of "after" cues.
There are a number of limitations of the current study that could be addressed in future research. First, this study did not include any norm-referenced measures of cognitive or language abilities to which performance on the TRA could be compared. Other RFT studies have shown strong correlations between performance on relational tasks and standardized measures of cognitive or language abilities (Kent et al., 2017;Kirsten & Stewart, 2022;Mulhern et al., 2017). By including similar measures in future work, this line of research could be extended specifically to temporal relations with children.
A second consideration is the relatively small sample size and relatively limited range of participant ages in the present study. Using a larger sample across all age cohorts would improve the external validity of results and might provide further insight into developmental patterns. In addition to the size of the sample, future research should also look at expanding the assessment to older age groups. Although overall performance on the temporal relations improved with age, error patterns with arbitrary relations were still present in the oldest participants within the present study. As previously mentioned, even adults show some differences between derivations based on "after" versus "before" relations. Conducting this assessment with a wider age range might help to further clarify at what age(s) children might begin to more reliably demonstrate mastery of these types of relational tasks.
Another consideration for future research is to address the relatively limited trial types used. In the present study, only Yes/No questions were used for the nonarbitrary, mutual entailment, and combinatorial entailment sections. Future research might include additional response topographies, such as tacting the correct contextual cue on trials or selection-based responses to more fully examine the relational repertoires. Second, the nonarbitrary section used only short (1-s) durations between stimuli. Future work could present questions of nonarbitrary temporal relations with events that happened over a longer time frame (e.g., a few minutes ago). In addition, the current study exclusively used multimodal relational statements (e.g., participants saw the relational statement while hearing it spoken aloud). It remains unclear to what extent each of these modalities might have facilitated performance. Future work might compare performance across visual, auditory, or multimodal presentations of relational statements.
The current study also included different stimuli across test sections with more concrete stimuli used on nonarbitrary trials and only colored shapes on arbitrary trials. Concrete stimuli were selected for the nonarbitrary trials to provide more salient exemplars and a clearer discrimination between stimuli within and across trials because the task involved the presentation/removal of stimuli. The use of colored shapes for the arbitrary trials was however adopted from recent empirical work using this REP format (Kirsten et al., 2022a, b;Kirsten & Stewart, 2022). We suggest that future research utilize the same stimuli across both nonarbitrary and arbitrary tasks, whether it be more meaningful or abstract stimuli. Further, the present study only used contextual cues between the relata ("A before B"; "B after A") to be consistent with the REP format. However, in natural environments, these cues may also be contacted in the first part of sentences (e.g. "Before B, A"; "After A, B"). As previously noted, other psychological research has showed performance differences with these arrangements in children (McCormack & Hanley, 2011) and this could be evaluated from an RFT perspective.
In addition, only "before" and "after" cues were used to assess nonarbitrary and arbitrary temporal relations. Although these two words are important to contextually controlled temporal relational responding, investigating performance with additional words might be desirable to closer examine the flexibility, fluency, and breadth of the relational repertoire (Luciano et al., 2009). Words like "first" and "last" are other commonly used words to which young children may often be exposed. Some linguistic research has shown that young children may more easily follow instructions with "first" and "last" (e.g., "Move the red car first; move the blue car last,") compared to "before" and "after" ("Move the red car before you move the blue car") due to the absence of a subordinate clause (Amidon & Carey, 1972). In the present study, "before" and "after" were specifically selected because they advantageously describe a relation between stimuli in a manner similar to that seen in other RFT studies on relational framing, whereas "first" and "last" more so describe a feature of a stimulus or deal with quantification of the (temporal) dimension. Nonetheless, these terms could be used for example, by giving a relational context (e.g., "if A is after B and B is after C . . .") and assessing for a transformation of ordinal functions using "first" and "last" (". . . Which is first/last?"). Other types of temporal relations could also be examined by using experienced past events ("Which was a longer time ago, your birthday or Christmas?") or hypothetical future events ("Which comes sooner, your birthday or Christmas?"). Future research might integrate these terms as well as other commonly used words into protocols of assessment and training of temporal relational responding.
Based on the present results, future research might seek to investigate the relationship between cues for spatial and temporal relations further as time is often represented spatially. In the present study, visual stimuli were presented horizontally and read from left-to-right. As previously discussed, notable differences in performance were observed between the arbitrary temporal contextual statements in the current study. For instance, did the participants that responded correctly to "after" trials (such as those to the right in Fig. 7) respond under control of the contextual cue for a temporal relation or might they have responded based primarily on a reversed spatial relation? In other words, was a rule derived about the spatial order (e.g., right-to-left order) which thereby altered responding? More investigation is needed to examine to what extent arbitrary spatial relations might interfere or compete with stimulus control under the temporal contextual cues in children.
A final recommended research direction is to examine training procedures for temporal relational responding. The aim of the present study was to explore how children at different ages perform with relational tasks when assessed and did not include a specific training component. One direction might be to investigate directly training both temporal NARR or AARR repertoires directly. For instance, the current study found that deriving arbitrary relations when given an "after" cue to be particularly difficult for participants. Future research could look to train this skill in particular. Another direction might consider to what extent training a temporal NARR repertoire facilitates acquisition of temporal AARR. In the current study, the NARR repertoire appeared to hit mastery prior to the AARR repertoire (which, in fact, with respect to some specific patterns, was not seen in any of the cohorts) and they strongly correlated with each other. RFT research on other relations has shown that NARR is an important precursor ability to AARR and may help support the acquisition of particular relational frames (Berens & Hayes, 2007). Likewise, another direction could be to examine whether training other particular frames help facilitate the acquisition of temporal framing. Investigating further which relational repertoires are important precursors could help pinpoint optimal relational training sequences that involve temporal relations.

Conclusion
Despite the limitations, the present study provides insight into children's performance on tasks of nonarbitrary and arbitrary temporal relational responding. The current data expands the research base on temporal frames to a new population and adds to the growing RFT literature on relational responding abilities in general. The results can be used as a starting point for further investigation into how contextual control is acquired for temporal relations and how these repertoires may be best trained in children.

Data Availability
The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.

Ethics Approval
The study was approved by the NUI Galway Research Ethics Committee. The study was performed in accordance with the ethical standards of the research institutional research ethics committee and with the 1964 Declaration of Helsinki and its later amendments or comparable ethical standards.

Consent to Participate
Informed consent was obtained from all participants included in the study.

Conflicts of Interest
The authors have no known conflicts to disclose.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.