A Pilot study of the Online Assessment of Self-Regulated Learning in Preschool Children: Development of a Direct, Quantitative Measurement Tool

a,*Corresponding Author: Lisa Jacob, Department of Educational Sciences, Saarland University, Saarbrücken. Campus, Building A4 2, Room 4.19, 66123 Saarbrücken, Germany. E-mail: li-sa.jacob@uni-saarland.de b Sandra Dörrenbächer, Department of Educational Sciences, Saarland University, Saarbrücken, Germany. E-mail: sandra.doerrenbacher@uni-saarland.de c Franziska Perels, Department of Educational Sciences, Saarland University, Saarbrücken, Germany. E-mail: f.perels@mx.uni-saarland.de Abstract


Introduction
Self-regulated learning (SRL), that is, the ability to learn through the autonomous and self-directed application of strategies, is crucial for dealing with the fast-moving challenges and demands of everyday life (Brunstein & Spörer, 2010). Given the undeniable link of such self-regulatory, strategic skills to academic success (e.g. McClelland, Acock, Piccinin, Rhea, & Stallings, 2013), self-regulated learning has been examined primarily in students (e.g. Dörrenbächer, Russer, & Perels, 2018, Fadlelmula, Cakiroglu, & Sungur, 2015Schnell, Ringeisen, Raufelder, & Rohrmann, 2015) and (under-) graduates (e.g. Mega, Ronconi, & De Beni, 2014;Tabuenca, Kalz, Drachsler, & Specht, 2015, Leidinger & Perels, 2012. However, from a developmental perspective, it seems essential to expand the research focus to earlier stages of the lifespan, too: importantly, preschool age has been shown to mark the sensitive period for the maturation of SRL and closely related abilities, such as the development of executive control functions (EF; Hofmann, Schmeichel, & Baddeley, 2012;Lockl & Schneider, 2007;Zelazo, 2015). However, the assessment of SRL, or EF, respectively, in preschool age requires specific measurement tools that match the particular characteristics of this age group (Hoyle & Dent, 2018), e.g. the restricted reading abilities or limited memory capacity (Van Den Broek, Kendeou, Lousberg, & Visser, 2011). Kindergarten teacher ratings have become established for these age ranges (Howse, Lange, Farran, & Boyles, 2003). The sole application of such external ratings has to be been deemed critical due to the missing possibility for countervalidation. There is, however, a lack of instruments that measure SRL directly at the child level. Hence, the aims of the present study are (1) to take the first steps towards the development of an instrument for the direct and quantitative (online-) assessment of SRL in preschoolers at the child level and (2) to evaluate its psychometric quality.

Self-Regulated Learning at Preschool Age
SRL (precursor) skills at preschool age ('Preschool age' describes the age span between 5 and 6 when German children generally attend their last year of kindergarten) predict SRL abilities in later life (e.g. McClelland et al., 2013), which implies that early limitations in SRL may amplify poor learn-ing control in adulthood. Hence, given the predictive power of SRL competencies for later 'life success', it seems worthwhile to start early in life with the measurement and instruction of stra-tegic, self-regulated activities (Montroy, Bowles, Skibbe, Mcclelland, & Morrison, 2016). Moreover, considering the high correlations with other fluid intellectual abilities (Brydges, Reid, Fox, & Anderson, 2012), well-developed SRL (precursor) skills as well as practiced EF may also foster cognitive development in general. Preschool age in particular represents a sen-sitive period for cognitive maturation: there appears to be a general shift from emotion-driven regulation to a more cognitive regulation where complex learning processes like SRL can be built upon (Zelazo, 2015). Moreover, a qualitative shift from an external regulation to a more internally guided self-regulation style can be observed (see Montroy et al., 2016). In this age range, SRL is also closely related to core developmental tasks, such as good habits and adequate peer behaviour (McClelland, Morrison, & Holmes, 2000). Thus, SRL helps to reduce im-pulsive behaviour and allows for 'thinking before acting' (Eisenberg et al., 2005), both support-ing the development of social appropriateness that may be particularly important in educational settings. In summary, self-regulation (precursor) skills seem to be a hallmark for everyday-functioning in the age group of preschoolers (Bronson, 2000;Kochanska, Coy, & Murray, 2001). sumed to be composed of a set of cognitive, metacognitive and motivational learning strategies that can be applied within this dynamic learning cycle. Regarding the latter structural dimension of SRL, it has to be consid-ered that at preschool age, the cognitive system is still poorly differentiated (Brydges et al., 2012;Shing, Diamond, & Davidson, 2010), implying that the unity of SRL components or strategies disproportionately may outweigh the diversity of strategies. Consequently, SRL at preschool age may be defined more sharply on the temporal axis than on the structural axis, which should be considered when constructing a measurement tool for preschoolers.

Theoretical Considerations Regarding Self-Regulated Learning in Preschool Age
In the present study, we rely on the social-cognitive framework of SRL by Zimmerman (2000) that emphasizes the temporal, dynamic character of SRL. In this framework, SRL is defined as 'self-generated thoughts, feelings, and actions that are planned and cyclically adapted to the attainment of personal goals' (Zimmerman 2000, p.14). According to Zimmerman (2000), throughout the SRL process, a multitude of strategies can be applied. However, given the primacy of the temporal compared to the structural (strategy-related) dimension of SRL in preschoolers (see previous section), we propose an adapted version of the framework with a focus on the temporal differentiation and a looser, more parsimonious differentiation of SRL strategies within the SRL process than in the Zimmerman model (2000; see figure 1). When selecting relevant SRL strategies, the state of development of children at preschool age was considered: (a) Children of that age group are capable of goal setting and adjustment of thinking and acting towards goals (Blaye & Chevalier, 2011;Hendry, Jones, & Charman, 2016), which represent important skills concerning the forethought phase and the performance phase of SRL; (b) Furthermore, preschoolers already show inhibitory control (Carlson, 2005;Lewis, Reeve, Kelly, & Johnson, 2017), and attention focusing skills (Bronson, 2000;Lewis et al., 2017) which are relevant abilities, especially for the performance phase; (c) preschoolers can reflect their own learning process (Zelazo, 2015) which is essential for the self reflection phase of SRL.

Executive Functioning at Preschool Age
Preschool age is simultaneously the critical period for the maturation of another class of basic cognitive skills, namely EF such as shifting, updating or inhibition (Erb, Moher, Song, & Sobel, 2017;Shaul & Schwartz, 2014). Such EF enable individuals to perform higher level cognitive operations, such as planning, problemsolving and target-oriented acting (Miyake et al., 2000) and can be thus considered comparable to SRL skills (Perry, Hutchinson, Yee, & Määttä, 2018; see also the next section for a discussion of the relationship of both concepts). For the assessment of EF in preschoolers, in contrast to the assessment of SRL, there already exists a considerable number of quantitative, direct measurement tools (Ackerman & Friedman-Krauss, 2017). A common instrument to measure higher level EF is the well-established Tower of London Test (ToL Test, Shallice, 1982) that has also been successfully administered to preschoolers (Byrd, Van Der Veen, McNamara, & Berg, 2004) and has proven to be a useful instrument to capture interindividual differences in children (Raizner, Song, & Levin, 2002). The ToL Test consists of reconstructing target configurations with the aid of coloured balls by considering a predetermined number of action steps (also see method section).

The Relationship Between Self-Regulated Learning and Executive Functioning
Higher level EF shows some conceptual overlap with SRL, both having a heterogeneous structure, and encompassing a wide and diffuse range of interrelated, yet unique abilities (Jurado & Rosselli, 2007;C. C. Ponitz, McClelland, Matthews, & Morrison, 2009). However, the precise nature of the relationship among both constructs remains elusive so far (e.g. Hofmann, Schmeichel, & Baddeley, 2012). Some authors argue that SRL can be interchangea-bly with EF (Gaskins, Satlow, Presseey, & Meltzer, 2007), while some argue that SRL can be considered as a superordinate construct, including executive basis operations (Barkley, 2001). In a recent review by Hofmann et al. (2012), EF has been subdivided into a set of subcompo-nents, including (a) working memory operations, (b) behavioural inhibition and (c) task-switching, that are directly related to a number of self-regulatory mechanisms, such as (a) the active representation of self-regulatory goals, (b) the active inhibition of 'mindless' behaviour, and (c) self-regulatory goal shifting and balancing. Lockl and Schneider (2007) consider the development of EF as a precondition for the emergence of SRL abilities.
Given the undoubtedly close relationship between both constructs, of whatever nature, and the greater availability of direct measurement tools for higher level EF for preschoolers (such as the ToL test, Shallice, 1982), such EF tasks seem well suited as a criterion, against which a newly developed SRL measure tool needs to prove itself.

Previous Attempts to Measure Self-Regulation In Preschoolers
There are at least four general challenges associated with the direct measurement of SRL in preschoolers: (a) their restricted reading and writing skills, (b) their fragile memory for past events which may impede retrospective recall of strategy knowledge (Maylor & Logie, 2010), (c) misjudgements of their own performance (Schneider & Büttner, 2008), and (d) a low test compliance for standard instructions (Stephenson & Hanley, 2010).
Previous attempts to measure self-regulation in preschoolers can be classified into two major approaches: offline and online measures (e.g. Winnie & Perry, 2005). Regarding offline methods that assess SRL before or after a learning activity, structured interviews seem feasible for the special age cohort of preschoolers, yet they need to be applied in child-appropriate ways due to the restricted language proficiency, the low test compliance in preschoolers and the lack of metacognitive abilities that enable them to verbalize their own learning process (Whitebread et al., 2009). A rare example of a successful application of such a structured interview is the study by Perels, Merget-Kullmann, Wende, Schmitz and Buchbinder Figure 1. Zimmerman's (2000) process model of SRL, adapted for preschoolers (2009). They provided preschoolers with interviews embedded into a puppet show to enhance the children's treatment compliance. However, preschoolers' tendency to overestimate the abilities they are asked about limits the results of the interviews (Schneider & Büttner, 2008). Another useful measurement method seem to be external assessment as rated by trained kindergarten teachers (Howse et al., 2003). External rating scales that are frequently used include the Child Behavior Rating Scale (Bronson, 1994), the Strengths and Difficulties Questionnaire (Goodman, 1997), or the CHILD-Checklist (Whitebread et al., 2009). However, there is a lack of reliable self-assessment in pre-schoolers to allow for a cross-validation of these indirect measures.
A useful online measure of SRL at preschool age are thinkaloud protocols. It has been shown that four year old children are already capable of articulating their own thoughts when viewing a picture book (Paris & Paris, 2003;Tompkins, Guo, & Justice, 2013). However, a relevant disadvantage of thinking aloud methods as a measure of SRL is the high demand on productive language skills and metacognitive skills. Another measurement tool is observational inventories, including the Bronson's Social and Task Skills Profile (Bronson, 1994), the C.Ind.Le Coding Framework (Whitebread et. al., 2009) and the SRL observation tool using the Train Track Task (Bryce & Whitebread, 2012). However, a disadvantage of behavioural observation is that learning strategies that are known implicitly but not demonstrated during obser-vation may not be captured, thus underestimating children's strategy knowledge (Landmann, Perels, Otto, Schnick-Vollmer, & Schmitz, 2009).
In the present study, we aimed to develop a measurement tool that counteracts these disadvantages of established instruments. The online character of our tool may limit bias due to the insufficient self-estimation skills of children at that age, allows for direct measurement on child level, requires little productive language skills, captures a set of SRL learning strategies which are considered important for preschoolers, and provides quantitatively interpretable data based on a standardized evaluation protocol. The measurement tool should be evaluable by realizing cross-validation against an external SRL measurement tool and a direct EF measure-ment tool.

The Present Study: Development, Evaluation and Conceptual Alignment of a Direct, Quantitative SRL 'Online' Measurement Tool for Preschoolers
To summarize we aimed to develop an SRL online test procedure for preschoolers that meets the following criteria: First, the temporal SRL dimension should be focused. Second, we aimed to develop a quantitative rather than a qualitative tool to allow for clear interpretability and comparability. Third, the tool should address the children directly (i.e. online measure-ment), thus requiring some adaptions to the particular needs of this age group. Fourth, our tool should be cross-validated against established instruments from a familiar domain of research that already provides adequate quantitative online measurement tools for preschoolers, namely research on EF. Our developed SRL test procedure was inspired by the direct and quantitative test of metacognition for first graders used in the German National Educational Panel Study that meets most of these criteria (Lockl, Händel, Haberkorn, & Weinert, 2016). Metacognition (to-gether with cognition and motivation) represents a core component of SRL (Boekarts, 1999) and describes the availability of 'meta-information' about many different types of cognitive processes. Similarly, knowledge about SRL means specific 'meta-information' about learning processes. Lockl et al. (2016) presented groups of first graders orally with twenty child-oriented everyday problem scenarios involving different settings and characters. Each scenario required the selection between three more or less strategic response options to solve the presented problem. This test of metacognition is tailoring towards the particular needs of the age group of first graders may still present some disadvantages for preschool age: First, the test of metacognition only focuses on the metacognitive component of self-regulation, and does not specifically address the temporal, processual character of self-regulated learning that is of primary interest in preschoolers. Second, the asymmetric reply format may allow for error of central tendency, which seems especially harmful in five to six preschoolers with a large degree of decision uncertainty (Hembacher & Ghetti, 2014). Third, alternating protagonists could lead to excessive demands on preschoolers with regard to their only partially developed ability to grasp perspectives (Gamannossi & Pinto, 2014). For the development of our instrument, we aimed to adapt the test of metacognition to the specific needs of preschoolers.

Research Questions
This study examines whether our newly developed SRL measurement tool is a reliable and valid instrument to measure self-regulated learning in preschoolers. It addresses the following research questions: 1. Does the application of a newly developed SRL measurement tool yield data indicating a sufficient reliability when regarding (a) the tool as a whole and (b) three theoretically based subscales (along the timeline of the self-regulated learning process)?
2. Can scores from the newly developed SRL measurement tool be considered valid indicators of preschoolers' SRL abilities when using two kinds of cross-validation strategies, namely (a) 'near cross-validation': comparison with the results of a SRL measurement tool that is rated externally and (b) 'far cross-validation': comparison with the performance in a EF measurement tool that is also applied at child-level?

Participants
For this study, 183 preschoolers were recruited from 11 kindergartens in Germany. Nineteen children had to be excluded from the analysis, either because they refused to undergo the testing procedure (n= 17) or due to excessive missing data (n= 2). Hence, the final sample for analysis consisted of 164 children (51.5 % male and 47.3% female; mean age: 5.9, age range: 4.9 -6.7). All parents gave their written consent for the participation of their children in the study, in accordance with national law and the protocols approved by the local ethics committee. All children had normal or corrected to normal vision, no hearing impairment and no history of any neurological diseases. Based on parents' statements, none of the children had a known learning disability or suffered from developmental delay.

Procedure and Design
The testing procedure consisted of two sessions and was conducted in separate, quiet rooms in kindergarten buildings. All tests were conducted by competent test administrators which underwent training sessions and adhered to a standardized test manual. On average, the test session took 18.4 minutes. The relevant instruments were embedded in a series of tasks. The order of task presentation was always the same: At first, the SRL measurement tool for preschoolers was administered; subsequently, children worked on an EF test (Tower of London Test; Shallice, 1982). Each child was tested individually with two administrators present. As a thank you for their participation, children received a child-friendly certificate of attendance after they had completed all tasks.
An external assessment of preschoolers´ SRL ability in the form of a questionnaire was given to the kindergarten teachers at a first informative meeting in the run-up of the testing procedure and recollected at the day of the first testing session.

Measures
Self-regulated learning in preschoolers (SRL measure tool for preschoolers). The present study applied this newly developed direct and quantitative test tool to investigate SRL in preschool children.

Cover story, structure of measurement tool and items
The testing script revolves around a background story in which a fictional protagonist called Lennie the Lion is confronted with everyday problems and tries to solve them using various strategies. The story is partly based on the cover story of an earlier study of Perels (2009). Children are instructed to assist Lennie in 'real time' in planning, performing and reflecting a chosen way of proceeding according to a predefined overarching goal (i.e. finding a present for his friend Ellie the duck on the occasion of her first day in school). Hence, the narrative is about different problem scenarios within the story's framework, throughout which the protagonist evolves and develops solution strategies in a step-wise fashion. The measurement tool captures the following SRL strategies: planning, using prior knowledge, dealing with deflectors, self-efficacy, monitoring, breaks and self-motivation, reflection and causal attribution (see figure 1, Zimmerman, 2000). Children are instructed to rate the usefulness of Lennie's proposed solution strategies to master the problem scenarios, similarly to Lockl et al. (2016). Children should rate the usefulness, in a dichotomous response format, as 'not very beneficial' or 'highly beneficial' to the objectives of the protagonist ('Do you think Lennie´s idea is a good/bad idea or a bad/ good idea'? in varying order) by tapping a happy or an unhappy face (child oriented display format of response scale). Highly beneficial solution strategies represent one amongst eight SRL strategies (i.e. 'SRL+' items) supposed to be relevant for preschool age (c. figure 1, adapted version of Zimmerman´s [2000] process model of SRL). In contrast, less beneficial solution strategies represent non-SRL strategies (i.e. 'SRL-'items). The number of SRL (+) and SRL (-) was counterbalanced across the test procedure, and items were presented in randomized order. An example problem scenario for the SRL strategy planning (forethought phase) with two corresponding items is shown in Figure 2. Further example problem scenarios can be found in appendix A.
In sum, the measurement tool consists of 24 items. A total score was calculated by following the signal detection theory (Swets, 1996) alarms. This procedure should prevent distortions caused by guessing or systematic response bias (Coe, Waring, Hedges, & Arther, 2012). The total score ranges from -24 (exclusively misses and false alarms) to +24 (exclusively hits and correct rejections).

Age-appropriateness
In consideration of preschoolers' lacking reading abilities, the scenarios and proposed solution strategies were presented orally by the test leaders. To improve understanding, children could follow the explanations by looking at coloured drawings. In contrast to Lockl et al. (2016), a dichotomous response format (happy vs. unhappy face) was ap-plied to prevent the error of central tendency. The protagonist and the scenarios are designed in a way that should foster self-identification with the cover story to enhance treatment motiva-tion.

Test construction process
Initially, the first version of the measurement tool to as-sess SRL in preschoolers was structured differently, since it consisted of a smaller number of problem scenarios with a higher number of corresponding items (two SRL (+) items and two SRL (-) items). The following SRL strategies were captured: planning, using prior knowledge, dealing with deflectors, self-efficacy, monitoring, breaks and self-motivation and reflection. The aim of the construction process was a preferably balanced measurement tool with a solid theoretical foundation. Consequently, the number of SRL (+) and SRL (-) items was balanced as well as the total number of items representing each learning strategy. The temporal dimen-sion of SRL was considered (c. Zimmerman, 2000). The relevant learning strategies were embedded in this structure and build on each other, leading towards an overarching goal.
Before the presented study, a pilot study was conducted with a smaller sample of N= 15 preschoolers of two German kindergartens who agreed on request to participate. These children were between 5 and 6 years old. The analyses of the data of this pilot study led to the following modifications which resulted in the used, modified test tool: First, three SRL (+) items, which were used to capture the SRL strategies of planning, dealing with deflectors and reflection, showed ceiling effects. A further three items showed poor discriminatory power. These were constructed to measure the SRL strategies of planning (SRL (+) item) and monitoring (SRL (+) and SRL (-) item). Second, the data indicated that children were not able to keep the initial scenario in mind when four response alternatives Figure 2. Zimmerman's (2000) process model of SRL, adapted for preschoolers were presented. Consequently, the number of response alternatives was reduced from four to two responses (one SRL + and one SRL -for each scenario). Third, answering texts were shortened and a linguistic revision was made with the help of a team of experts consisting of educational scientists and psychologists, in order to ameliorate the understanding. As a consequence of the pilot study, changes at the structural and single item level were made and led to the final instrument (see section Cover story, structure of measurement tool and items).

Synthesis
In summary, different elements were useful to build the final SRL meas-ure tool presented in this study. First, Zimmerman's (2000) model of SRL served as theoretical base. Second, the adapted story used in Perels (2009) served as cover story. Third, the child-appropriate presentation of SRL problem scenarios and corresponding response was inspired by Lockl et al. (2016). Fourth, the test evaluation is based on the signal detection theory of Swets (1996). Fifth, the results of our pilot study influenced item selection, number of responses and linguistic elaboration of the final SRL measure tool.

External rating-scale of self-regulated learning (SRL rating scale)
The SRL rating scale is an external assessment of SRL in preschoolers, filled out by their kindergarten teachers as experts. On the one hand, the item pool consists of a selection of items used in two previous studies examining SRL in children, namely studies by Otto (2007) and Merget-Kullmann and Wende (2004). On the other hand, it consists of items from two established measurement tools: the Children's Independent Learning Development (CHILD 3-5) checklist (Whitebread et al., 2009) and the Child behavioural ratings scale (Rowley, 2015). Item selection is based on con-tent considerations and the results from item analysis of a former version of the SRL rating scale, used in Venitz & Perels (2018). All items of the composed measurement tool were rated on a Four Point Likert scale that ranges from 1 (never) to 4 (always). The questionnaire, which is structurally related to the SRL measure tool for preschoolers, contains 35 items, which are grouped into eight subscales operationalizing SRL learning strategies. The structure and relia-bilities of the SRL rating scale are shown in Table 1. We used a modified version of the ToL Test (Shallice, 1982), a well-known and valid neuropsychological test of satisfactory to high reliability with α= .78 (Tuche & Lange, 2004), used to measure EF, in particular test takers' planning ability and problem solving ability. The version we used in this study was shortened from 20 to 10 items because the examined age cohort was of a younger age than the target group of the original ToL Test. One example problem and ten problems (or items) were administered. For each item, a stimulus card with a target configuration was presented to the child.
Children were asked to rearrange three different-coloured balls on three bars in different sizes so that the target configuration turns out. The number of ball movements was predefined, so the child has to make a plan before starting the action. The range of total performance ranges from 0 to 10 points (1= Problem solved correctly in compliance with the specified number of ball move-ments, 0= Problem not solved).

Statistical Procedure
The internal consistency of the SRL measure tool for preschoolers was estimated with the Kuder-Richardson formula which can be regarded as an antecedent of Cronbach's alpha and is used to deal with dichotomous data. Concurrent validity was estimated by correlating (a) an indicator of an external measurement tool of the domain of SRL; that is, scores of the SRL external rating scale, and (b) an indicator of a measurement tool of the related domain of EF that is also applied on the child level; that is, performance and planning time in the ToL Test.

Sample Test Results
The descriptive data of the SRL variables and the EF variable are shown in Table 2. Data is based on the analytic sample of N= 164.  Table 3 illustrates the descriptive item statistics, including item difficulty of the SRL measure tool for preschoolers. In eight SRL (+) items (1a, 3a, 4b, 6b, 7a, 9a, 10b,11a) ceiling effects (P i > 80) were found, whereas one SRL (-) item (11b) showed floor effect (P i < 20). Therefore, those items were dropped for further analysis.
Both types of items -those items with ceiling/ground effects and those items with low discriminatory power -would impair the reliability and consequently the validity of the SRL measurement tool. In sum, all 12 SRL (+) items and one SRL (-) item were removed from further analysis. Consequently, further analyses are based exclusively on the remaining 11 SRL (-) items. Nevertheless, all SRL strategies are still captured by the measurement tool, which also means that items from all SRL phases (forethought phase, performance phase, reflection phase) were preserved.

Reliability Analysis
The internal consistency of the final SRL measurement tool for preschoolers (including 11 items) was α= .72. The corrected item-total correlations varied between r it = .17 and r it = .55 (see table 3) with an average of r it = .35. The internal consistency of the three theoretically plausible subscales representing the three phases of learning (each comprising four items) was α= .42 for forethought phase of SRL, α= . 58 for performance phase of SRL and α= . 40 for self-reflection phase of SRL.

'Near Cross-Validation' Using an External Srl Measurement Tool
The Pearson correlation coefficient between the total score of the SRL measurement tool for preschoolers and the SRL rating scale (M= 98.10, SD= 17.29) was r= .20 and reached significance at the .05 level (p=.03). Because of insufficient reliabilities of the assumed subscales (phases of learning) in the SRL measurement tool, further analysis on subscale level was not conducted. Based on theoretical assumptions, correlation coefficients between the single items of the SRL measure tool for preschoolers and the corresponding reliable subscales of the SRL rating scale were calculated and are shown in Table 5.

'Far Cross-Validation' Using an EF Measurement Tool Applied At Child-Level
The Pearson correlation coefficient between the total score of the SRL measurement tool for pre-schoolers and the performance in the ToL Test (M= 6.81, SD= 1.90) was r= .18 (p= .018).

Discussion
The aim of this study was to initialize the development and evaluation of a direct, quantitative (online-) measurement tool of SRL for preschoolers by considering the cognitive and linguistic development status of children at this age. This topic has been subject to little research to date. For this reason, we intended to expand the repertoire of existing SRL method-ology for the preschool age ranges (Spörer & Brunstein, 2006). The SRL measurement tool presented in this study is a time-efficient, child-friendly instrument that was developed for ap-plication in field settings such as kindergartens. The results reported in this paper indicate ini-tial empirical support for the psychometric quality of the instrument, yet may also especially indicate possibilities for optimization.

The Reliability Analysis
Statistical analysis indicated a satisfactory reliability of the overall scale of the SRL measurement tool for preschoolers after the initially constructed instrument had undergone some changes. Importantly, the SRL (-) items proved to be of appropriate item difficulty in contrast to the SRL (+) items, where many items suffered from ceiling effects. A possible explanation could be response biases in the form of acquiescence phenomena (Coe, Waring, Hedges & Arther, 2012) which describes children´s tendency to agree with the test leader or protagonist of a cover story. For instance, Cleveland, Quas, and Lyon (2016) demonstrated children's sensitivity for acquiescence in interviews at preschool age. Even though we intended to reduce distortions by adequate formulations of test instruction ('Do you think Lennie´s idea is a good/bad idea or a bad/ good idea'? in varying order) we, presumably, could not counteract the sensitivity for response bias in this age group. Evaluating items as 'highly beneficial' (or 'good idea') could explain the ceiling effects in specifically those SRL (+) items, where 'highly beneficial' is the correct answer and consequently rated as a hit. Potentially, the SRL (-) items, which showed a good item difficulty, were able to counteract response biases and, additionally, made greater demands on children -they need to inhibit an automatic response tendency such as 'agreeing'. Probably, SRL (-) items represent a type of item that is more differentiating and appropriate to explore SRL on the child level with the kind of measure tool we used.
Contrary to our assumptions, the reliability of subscales, also on the time axis based on Zimmerman's (2000) process model of SRL (forethought phase, performance phase, self reflection phase), was limited in our preschool sample. The specific developmental status of the exam-ined age cohort may serve as an explanation. To be precise, there are two possible explanations as to why the application of our measurement tool might have failed to yield the expected subscale structure: (a) The sensitive period of preschool age is characterized by critical cognitive maturation processes that enable individuals to regulate their thinking and behavior (Montroy et al., 2016;Zelazo, 2015). Potentially, our tested children might, however, have just started to apply particular learning strategies which may have crossed the sequential order as assumed in the SRL process model. The integration into a holistic learning process could then evolve at a later point in time. Consequently, the process model of SRL (Zimmerman, 2000) would not be as easily applicable as in other age groups such as, for example, students (Leidinger & Perels, 2012) or university students (Dörrenbächer & Perels, 2016;Schmitz, Klug, & Schmidt, 2011). (b) The assumption that the use of intuitive learning strategies does not fit the process model (Zimmerman, 2000) does not necessarily render preschoolers unable to integrate the learning strategies into a holistic process. Rather, they may be in need of special support through programmes with a focus on SRL. Training studies in combination with multi-methodological measurement of change are essential to test this hypothesis and may shed more light on the developmental progress of SRL in preschoolers.

The Validity Analysis Using 'Near and Far Cross Validation'
Results of the validity analysis, using two established instruments with good psychometric quality as criterions, suggest that the newly constructed tool as a whole seems to measure SRL-like abilities: Both the total scores of a structurally similar external measure and the structurally dissimilar EF measure show significant correlations with the newly developed test. Additionally, analyses on the single item level gathered more information on the validity of our instrument: In sum, four items (1b /using prior knowledge, 3b/planning, 6a/self-motivation and 10a/reflection) showed significant correlations with the corresponding sub-scales of the external rating scale. This demonstrates that the at-  tempt to cover different SRL-relevant learning strategies may have been successful and supports the idea that the particular SRL strategies -possibly without support -evolve in a more independent manner than as-sumed in the SRL process model (Zimmerman, 2000). In this context, it might be of interest to examine particular learning strategies more deeply by using a larger item pool with the aim to (a) replicate our findings that certain SRL strategies (using prior knowledge, planning, self-motivation and reflection) are indeed already measurable on the child level and (b) ameliorate problem scenarios and corresponding items to operationalize the particular SRL strategies of our SRL measurement tool for preschoolers.

Limitations and Outlook
Naturally, there are limiting factors and unanswered questions regarding this study. First, the test results on the child level are based on only one measurement point. Consequently, the results possibly represent a 'snapshot' which may be influenced by the way children felt on the day of testing as well as outside factors such as, for example, interruptions while playing for the purpose of testing. Contrary to the direct measure of SRL, the indirect measure via external ratings by the kindergarten teachers is based on many observation moments in the every-day life of kindergarten. In future research, assessing the constructs of interest on at least two occasions would be useful. Second, properties of the kindergarten teachers limit the results: (a) the external ratings are subject to the response behaviour of the respective kindergarten teacher which filled out the questionnaire; (b) even though kindergarten teachers were introduced to the topic of SRL in the run-up to the filling out of the questionnaire, their knowledge after this instruction was not systematically examined, so a different degree of expertise and sensitivity to the detection of SRL abilities is conceivable; and (c) the time period during which the kindergarten teachers attended the respective child was not recorded, but could assumedly have influenced the validity of the rating. Third, the age range of the examined preschoolers is approximately two years because the status of 'preschoolers' was chosen as the inclusion criterion instead of the numerical age. This procedure may have resulted in a loss of information concerning interindividual developmental differences and corresponding SRL abilities in the wide age range. Therefore, future research in the field of SRL in preschoolers should restrict the age range.

Summary and Practical Implications
In summary, the present findings indicate that it seems both plausible and possible to assess SRL online at the child level in an objective and quantitative manner. Further research is needed (a) to make valid statements about the fit of the underlying theoretical model that was considered in our test construction process, (b) to justify an appropriate selection of SRL strat-egies that are already measurable at preschool age (intuitively or after SRL intervention programs) via direct measurement, and (c) to optimize test instructions and item construction with regard to difficulties that may arise from the particular response behaviour in preschool children.
The further development of the SRL test tools is important for the realisation of adapted educational tasks in kindergartens which do not dispose of a structured, well-defined curriculum for the preschool year (as it is the case in Germany). A preferably simple and standardized applicability of such a tool is important to allow a) the execution by external trainers visiting kindergarten in the context of intervention programs as well as b) the execution by kindergarten teachers. An uncomplicated application with no need of special equipment and less spatial conditions (only a quiet room to test children one by one) should allow for usage in kindergarten setting. Direct SRL assessment tools of high psychometric quality are a necessary precondition to develop and empirically verify programmes aiming to promote SRL in intervention settings (Hoyle & Dent, 2018). In this context, the usage of an SRL test tool as a perfor-mance measure before and after an SRL interventions is thinkable. Furthermore, a SRL test tool allows to track progress in SRL through repeated measurement (formative assessment).
Finally, the overarching goal to do research on early SRL assessment in preschool children is to facilitate the transition from kindergarten to primary school. The assessment and promotion of basic skill such SRL represents an important pillar due to the documented association between SRL with both academic and life success (e.g. Eisenberg et al., 2005;McClelland et al., 2013).