Ways of composing teaching teams and their impact on teachers’ perceptions about collaboration

Abstract

The present study examined the impact of teacher team composition on characteristics and attributes regarded as necessary for effective cooperative teaching. The study focused on potential differences between self-selected teacher teams and teams composed by the school administration. The central assumptions were that teachers working in self-selected teacher teams show more positive ratings of enjoyment, shared responsibility, job satisfaction and collective self-efficacy expectations than teachers who worked in institutionally composed teams. In order to investigate these hypotheses, an online survey was created. 321 language arts teachers participated in the survey. MANCOVA revealed significant differences in the dimensions ‘shared responsibility’ and ‘enjoyment with the co-teaching process’, where teachers from self-selected teaching teams showed significantly more positive ratings. These results support the assumption that self-selection of the team-mate is helpful for establishing compatible teaching teams, but does not necessarily lead to a higher quality of collaborative teaching.

Keywords:

• Teaching teams
• teacher collaboration
• team teaching

1. Introduction

Collaborative and cooperative teaching is used with increasing frequency in many educational systems and is often seen as an appropriate response to the challenges of individualised schooling in heterogeneous groups (Ford and Gray 2011; Kliegl and Weaver 2014; Liston, Nevin, and Thousand 2010; Moolenaar, Sleegers, and Daly 2012).

Numerous forms of cooperation between teachers are implemented in daily school work. They differ in the degree of cooperation, in the level in which the cooperation is performed, and in the extent of cooperation itself. It is therefore not astonishing that various models and terms exist, which describe and summarise cooperation between teachers. The spectrum ranges from different concepts of professional learning communities (plcs) which focus on the degree of cooperation within the staff of a certain school, to various concepts which focus on the cooperation of usually two or more teachers in daily class work (Bonsen and Rolff 2006; Perry and Stewart 2005; Reed and Groth 2009).

One of the foremost used models to describe cooperation and the degree of cooperation between teachers was introduced by Little in the early 1990s. According to Little (1990) the degree of dependence or interdependence between the teacher and his/her cooperation partner is the distinctive feature of the kind of cooperation. In this regard Little (1990) differentiated between four hierarchical levels or degrees of cooperation: (1) storytelling and scanning for ideas, (2) aid and assistance, (3) sharing and (4) joint work. At level (1) storytelling and scanning for ideas only a very informal exchange of experiences and ideas exists. Teachers at this level are largely independent from each other. Therefore, it represents the lowest degree of cooperation between teachers. The degree of dependence increases at the other levels and reaches a maximum at the fourth level ‘joint work’. At this level the teachers work closely together and are strongly depending on each other (Little 1990). Teachers cooperating at this level usually work and teach together in teacher teams by doing Co-teaching or Teamteaching. (Dizinger 2015; Gräsel, Fußangel, and Pöbstel 2006; Little 1990).

Even in this regard, no definitive terminological and conceptual clarity exists (Welch, Brownell, and Sheridan 1999). Scholars, for example, often use the terms ‘Co-teaching’ and ‘Teamteaching’ synonymously without describing concretely what kind of collaborative work teachers are doing (Da Fonte and Barton-Arwood 2017; Dugan and Letterman 2008; Friend et al. 2010; Ronfeldt et al. 2015; Vangrieken et al. 2013). Usually the term Teamteaching is used for the joint work of two general educators in a double sized classroom, whereas the term Co-teaching is used in Special Education for the joint work of a Special Educator and a General Educator in a single sized classroom (Friend 2008). This has changed during recent years. Especially outside the US the term Co-teaching is nowadays also used for the joint teaching of two general educators in a single sized classroom. For example Villa, Thousand, and Nevin (2004, p.5), refer to the term Co-teaching as, ‘Co-teaching […] is defined as two or more people sharing responsibility for teaching some or all of the students assigned to a classroom.’ In this definition the teacher team not necessarily consists of a General Educator and a Special Educator. Rather the sharing of responsibility for the class is crucial.

Beside the terms Co-teaching and Teamteaching the terms disciplinary teacher teams and interdisciplinary teacher teams were increasingly used in recent years. These terms usually refer to the joint teaching of two general educators with similar or different subject backgrounds.

However, since the present study refers to the joint work of two general educators with the same subject background in a single sized classroom and a teacher-student ratio of 2:25 or even 3:25, the term joint teaching of teacher teams is used. Moreover, despite the usage of different terms to describe the joint work of teachers, the actual working realities of these teachers are likely to show some similarities, and usually the same characteristics are considered as important for delivering high quality instruction by the joint teaching of two or more professional teachers, regardless of what it is called (Vangrieken et al. 2013).

1.1. General framework

To determine the quality and effectiveness of joint teaching, personal as well as structural components are seen as important by different scholars. Drach-Zahavy and Somech (2002) argued that structural elements have an influence on the effectiveness of the cooperation in teacher teams. These are characteristics such as functional heterogeneity, frequency and duration of interaction as well as gender heterogeneity. Functional heterogeneity refers to the kind of teachers working together in one team – for example, a math teacher and a language teacher are a functionally heterogeneous team. Drach-Zahavy and Somech (2002) pointed out, that functionally homogeneous teams – for example, two English teachers – report higher ratings on effectiveness of the cooperation than functionally heterogeneous teams. To a far lesser extent, mixed gender groups reported also slightly higher levels of effectiveness of the cooperation. In contrast, age heterogeneity and educational heterogeneity of the team members seem to have no significant effects on the effectiveness of the cooperation (Drach-Zahavy and Somech 2002). These results are similar to the results of more general organisational and management research. In a meta-analysis on teambuilding research, Biemann and Weckmüller (2012) asserted that frequency and duration of interaction have a significant impact on the quality of cooperation in a specific teacher team.

It is important to note, however, that personal characteristics, attitudes and perceptions of the team members are considered as more important for team quality than structural characteristics (Biemann and Weckmüller 2012; Honingh and Hooge 2014; Krammer et al. 2017). These personal characteristics refer to a considerable number of prerequisites which must be fulfilled in order to provide good quality joint teaching. The most important ones for this study are: parity, personal compatibility and shared responsibilities for a series of classroom activities, for lesson planning and for assessment (Brinkmann and Twiford 2012; Brouwer et al. 2012; Brown, Howerter, and Morgan 2013; Cramer and Nevin 2006; Forte and Flores 2014; Sileo 2011).

Shared responsibility for classroom actions is considered particularly important for the delivery of high quality instruction by teacher teams (Simmons and Magiera 2007). These shared responsibilities encompass activities like shared delivery of instruction, shared assessment, shared responsibility for differentiated teaching and shared responsibility when deciding how to teach (Brinkmann and Twiford 2012; Friend 2008; Villa, Thousand, and Nevin 2004). In successful teacher teams, a negotiation process takes place where such shared responsibilities are brokered (Shibley 2006). In this context, the concerned teachers start a shared learning procedure where they become familiar with the perspectives and attitudes of their team partner. In an optimal case, the two teachers no longer count these elements individually, but rather as a ‘we’ result, where these elements are viewed from a joint perspective and where the team members work together on an equal footing (Rytivaara and Kershner 2012).

If the establishment of parity fails to happen, personal relationships between teachers can be strained. Moreover, the incompatibility of teachers hinders a successful collaborative process and the sharing of responsibilities. This limits the effectiveness and quality of the joint teaching (Baeten and Simons 2014; Pratt 2014). Not only differences in teaching style, but also personal differences between teachers often create tensions that must be addressed in order to achieve good quality joint teaching (Pratt 2014).

Teachers who have been interviewed have a solution to this problem. They often mention that it may be helpful if teachers could select their team partners themselves (Friend 2008; Friend et al. 2010; Murawski and Swanson 2001; Scruggs, Mastropieri, and McDuffie 2007). In their study, Rytivaara and Kershner (2012) concluded that concerned teachers regard an adequate level of autonomy in the partner selection process as a ‘must’. Experienced teachers are convinced that they know whom they are compatible with. Unsurprisingly, interviewed teachers often mention that they feel more effective and report higher ratings of job satisfaction when they could choose their team partner on their own (Wobak and Schnelzer 2015). This leads to serious implications regarding the impact of the team composition on the success and the efficacy of joint teaching. According to Bandura (1997) positive beliefs about collective self-efficacy have a positive impact on the academic achievement of students. In fact, extensive research supports the claim that teachers’ collective self-efficacy beliefs influence their teaching and their students’ motivation and behaviour (Bandura 1997; Klassen and Chiu 2010). The same is true for job satisfaction (Caprara et al. 2006). Here, teachers who are more satisfied with their jobs show better teaching performances based on the academic outcomes of their students. Following this argumentation, it must be assumed that voluntary colleague selection should have a positive impact on the collective self-efficacy perception and job satisfaction of the concerned teachers and thus also on the effectiveness and quality of their joint teaching.

Nevertheless, due to administrative reasons, self-selection of the team partner is often not possible. In fact, teams are often put together by the school administration due to the limited human- and time resources of a particular school. Moreover, serious concerns were pointed out on a theoretical level regarding voluntary partner selection. Some scholars argued that self-selection is not a guarantee of effective joint teaching because it does not ensure that the specific skills necessary for good quality Co-teaching are actually available (Friend 2008). Along these lines, Kain (2006) highlights concerns about this kind of team composition. According to him, experiencing social pleasure and social aspects, rather than purpose or teaching style, often determines the composition of self-selected teacher teams. Building teams based on social factors does not support the quality of joint teaching, particularly if teaching partnerships are then seen as a ‘party.’ Admittedly, experiencing social pleasure in occupational activities may have a positive influence on motivational aspects for a person and can positively influence the outcomes of a specific work process (Comelli and von Rosenstiel 2011). So, enjoyment with the joint teaching process should be seen ambivalently. On the one hand, it can positively influence the quality of the joint taught lessons, but on the other, it can also be just fun.

Unfortunately, only a relatively small number of peer-reviewed articles exists which are based on empirical data (Bacharach, Heck, and Dahlberg 2010). In contrast, many articles dealing with this issue can be characterised as ‘How to do’ instructions that rely more on logical conclusions than on sound empirical data (Brinkmann and Twiford 2012; Noonan, McCormick, and Heck 2003). Moreover, a larger amount of existing research can be defined as single case studies (i.e. Kain 2006). Nevertheless, these studies allow a deeper and more specific understanding of joint teaching and the group processes in teacher teams. However, sound empirical data would allow for a more extensive understanding of Co-teaching/ Team teaching and would enable a critical understanding of the transferability of the examined patterns on a more general level.

1.2. Specific framework in Austria

Joint teaching of two subject teachers was officially implemented in the Austrian compulsory school system at the lower secondary level with the introduction of the New Middle School (Neue Mittelschule) by the Austrian Ministry of Education in 2008/2009 (bm:bf 2014a). According to policymakers, the New Middle School represents a joint school for 10-14 year-olds and uses modern pedagogical concepts like Co-teaching, inclusion and other social modes of learning (bm:bf 2014b).

Joint teaching in the New Middle School was implemented as a strategy for better and more differentiated teaching (Grössinger 2014). Six hours per week (one class hour lasts 45 or 50 min.) are reserved for jointly taught lessons in the main subjects of German language arts, English lessons and Mathematics. In the ideal case, every class in the New Middle School is jointly taught in every main subject for two hours per week (bm:bf 2014a; Specht 2009). In daily school life, every teaching team consists of two regular teachers of the relevant subjects. In addition, due to the inclusive character of the New Middle School, special educators and paraprofessionals may also belong to the teaching team depending on demand. Therefore, if students identified as having Special Educational Needs (SEN) attend the class, one Special Educational Needs teacher or paraprofessionals participate in the teaching team alongside with the two subject teachers (Specht 2009).

Regarding the educational background of the teachers, it was originally intended by the Ministry of Education that one secondary level subject teacher and one subject teacher of a Gymnasium (which is a kind of school in Austria comparable to British sixth form colleges and mainly academically oriented. Hence, the main function of Gymnasium is preparing students for higher education) work together in a teaching team. This could not be implemented in daily school life because hardly any teachers of Gymnasiums were willing to teach at New Middle Schools. Hence, the teaching teams consist mainly of secondary level subject teachers of the New Middle School (Vogtenhuber, Lassnigg, and Bruneforth 2012; Author et al. submitted).

Teachers working in teacher teams in Austrian New Middle Schools face many of the same challenges as their colleagues in other countries – a dearth of adequate training, lack of adequate administrative support, incompatibility between the teachers and a lack of parity in the classroom (Grössinger 2014; Vogtenhuber, Lassnigg, and Bruneforth 2012). As a consequence, concerned teachers at the New Middle School also emphasise the preference to choose their teammate on their own (Wobak and Schnelzer 2015; bm:ukk 2013). However, this is often impossible because of administrative reasons. (Author et al. submitted).

2. Research hypotheses

In this study, differences between self-selected teaching teams and teams that were put together by the school administration were examined

First of all, it was examined if the kind of team composition has any effects on specific teaching skills and characteristics regarded as important for high quality joint teaching itself. Not only the shared responsibility for classroom activities, but also the knowledge of specific teaching methods and the enjoyment with the joint teaching process, were considered. The first three research hypotheses rely on these dimensions:

(i)	Teachers working in self-selected teams show higher ratings of sharing responsibility in the classroom than their counterparts in teams composed by the school administration.

This research hypothesis relies on the assumption that teachers know whose teaching styles they are more likely compatible with. It is easier for teachers to work together with colleagues who share their own views and conceptions.

(ii)	Teachers working in self-selected teams do not show higher ratings regarding specific teaching skills and the usage of external resources than their counterparts in institutionally composed teams.

As already mentioned, the voluntary team composition does not guarantee successful teamwork. In fact, self-selection of the team partners does not ensure that external resources are used nor does it guarantee that specific teaching skills are adopted.

(iii)

Teachers working in self-selected teams show higher ratings of enjoyment with the joint teaching process than their counterparts in institutionally composed teams.

This research hypothesis focuses on Kain’s assumption (2006) that self-selected teacher teams show higher levels of experiencing social pleasure and social meeting. Importantly, this research question does not necessarily concentrate on the risk of teacher teams being more determined by social factors. As mentioned earlier, the enjoyment factor can have a positive influence on motivational aspects as well. Nevertheless, it should be kept in mind that the theoretical underpinning of this research hypothesis is in some way ambiguous.

In addition, this study examined possible effects of different kinds of team compositions on collective self-efficacy expectations and job satisfaction of the surveyed teachers. The research hypotheses are:

(iv)	Teachers working in self-selected teams show higher ratings of collective self-efficacy expectations than their counterparts in institutionally composed teams.
(v)	Teachers working in self-selected teams report higher ratings of general job satisfaction than their counterparts in institutionally composed teams.

Due to the fact that positive collective self-efficacy and job satisfaction are related to higher motivation and student achievement, these research hypotheses should provide a very cautious and conservative indication of teaching quality of the teaching teams. Moreover, because teachers mentioned the importance of adequate autonomy in the partner selection process, self-selection was expected to have generally positive effects on these two dimensions.

3. Method

In order to answer the research questions, an online survey, using the software LimeSurvey®, was conducted. The sampling was conducted through an online-platform (www.nmsvernetzung.at) for secondary level teachers of the New Middle School and by sending invitations for participating in the survey to New Middle Schools all over Austria. Participants were also personally recruited during advanced training courses for English language teachers at the University of Teacher Education. The participation in the survey was voluntary and anonymity and confidentiality were guaranteed. Thus, teachers who replied to the invitations received an email from the first author of this study with the access data of the online-survey. The only precondition was that the participants had experience in joint teaching and had taught in at least one teaching team in language arts (i.e. German language arts or English) during the last three months. This procedure was chosen in order to ensure that only participants with current joint teaching experience took part in the survey. Language arts teachers were chosen because only functionally homogeneous teams were examined.

The implementation of the survey was carried out in two steps. First, the piloting of the questionnaire took place from January 2014 until March 2014. During this step, the instruments used were tested, validated and adapted. In sum, eighty teachers participated during the piloting phase. Secondly, after the revision of the questionnaire, the main survey was carried out. 321 teachers participated in the main survey, which lasted from May 2014 until November 2014.

3.1. Sample

Ethic permission was requested at the Austrian School authorities and granted under the conditions that involvement in the study was voluntary and anonymity and confidentiality were assured. The sample collection consisted of two different stages:

3.1.1. Pilot phase

Eighty teachers participated in the pilot phase. 76.7% of the teachers in this group were female. The average age was 48.3 (SD: 9.6) years. The teachers had an average general teaching experience of twenty-five years and an average experience of four years in joint teaching within disciplinary teacher teams. As already mentioned, the pilot sample was only used for scale revision.

3.1.2. Main survey

321 teachers participated in the main survey. 280 of them completed the questionnaire entirely with all variables necessary for the construction of indices and for the following Multivariate Analysis of Co-Variance (MANCOVA). In the 41 questionnaires which were not entirely completed, usually only a single value (out of over 50) was missing. To manage the missing data, Expectation Maximisation (EM) was used. Little’s MCAR test was performed and proved not to be significant (χ ² : 99.45, df: 123, p = .941), suggesting that the data in the sample were missing completely at random. EM was then used to estimate missing values for the continuous variables. After completing the EM-procedure, data from 294 participants were available for the calculations used in this study. The missing of 27 participants can be explained by the fact that Special Needs teachers were excluded from further analysis and, moreover, missing values for categorical variables were not estimated. Hence, participants with missing values in the categorical variables were excluded from further analyses as well.

81% of the teachers were female. The average age was 46.8 (SD: 9.8) years. The teachers had an average general teaching experience of 22.1 (SD: 11.8) years and an average of 2.8 (SD: 2.7) years’ experience in joint teaching. Regarding the educational background of all participants, 83.5% were secondary level teachers of the New Middle School and approx. 11.3% of the participants were teachers from a Gymnasium. The rest were Special Needs teachers who were excluded from further analysis. The situation is quite similar for the educational background of their team partners, 81% of them were secondary level teachers of the New Middle School. 14.1% were Gymnasium teachers and the rest were Special Needs teachers. All of these characteristics can be regarded as representative for this kind of school in Austria (Statistik Austria 2013). As this study focuses on teacher teams in language arts, almost all teachers were German language arts or English teachers. Over 88% exclusively taught these subjects. A small number of teachers also mentioned miscellaneous in this part, without more precisely stating in what subject they are teaching in a teacher team beside the language subjects.

3.2. Instruments

To estimate the quality of the sharing of responsibility, the availability of skills and external resources, the enjoyment with the joint teaching process, collective self-efficacy expectations and general job satisfaction three different instruments were used:

To examine the joint teaching related skills and actions, the ‘Are we Really Co-Teachers Scale’ (Villa, Thousand, and Nevin 2004) was used. Since this scale was developed based on a review of the literature on Co-teaching and relied only on a weak empirical basis of the responses of forty teachers, this scale was tested and adapted during the pilot phase of the present study to ensure empirical reliability and validity (Cramer and Nevin 2006). The original ‘Are we Really Co-teachers’ scale consists of thirty-four items. The response possibilities of the scale ranged from 1 (not true at all) – 5 (completely right). Two of these items were excluded due to inconsistencies during the translation process. Exploratory factor analysis of the pilot sample revealed statistically insufficient characteristics, such as low starting communalities (<.40) for eight items and a textually ambiguous factor structure. After excluding these eight items from the scale, it showed very good internal consistency with Cronbach’s α of .93. Exploratory factor analysis according to Kaiser-criterion (eigenvalue >1) revealed a theoretically suitable three factor solution which explains 57% of the variance. Kaiser-Mayer-Olkin criterion for sample suitability was good with .84. Bartlett-test for sphericity was also significant, with p < .001, therefore, sphericity could be ensured. This adapted scale was then used in the main survey. To further ensure reliability and validity of the scale, calculations of internal consistency and exploratory factor analysis were again performed for the main survey sample. In the main survey the scale showed good to very good statistical characteristics for internal consistency (Cronbach’s α = .95). Kaiser-Mayer-Olkin criterion for sample suitability was very good with .91. Bartlett-test for sphericity was also significant, with p < .001, therefore, sphericity could be ensured. Moreover, exploratory factor analysis according to the Kaiser-criterion revealed a three-factor solution which is identical to the factor structure of the explanatory factor analysis of the pilot sample and explains 61.2% of the variance. The first factor summarised items on different classroom activities like: ‘We share responsibility what to teach’, ‘We share responsibility for how to teach’ or ‘We share responsibility for deciding who teaches what part of a lesson’. This factor was therefore called ‘Shared Responsibility’. The second factor encompassed items like ‘We seek and enjoy additional training to make our Co-teaching better’, ‘We can use a variety of Co-teaching approaches’ and ‘We communicate our needs for logistical support and resources to our administrators’. This factor, thus, referred to ‘Skills and External Resources’ helpful for high quality Co-teaching. Finally, the third factor could be summarised as an enjoyment factor, encompassing items like ‘We celebrate the process of Co-teaching and the outcomes and successes’, ‘We have fun with the standards and each other when we co-teach’ and ‘We communicate freely our concerns’. This factor, hence, was called ‘Enjoyment with the Co-Teaching Process’. The first factor consists of 9 items (Cronbach’s α = .94), the second consists of twelve items (Cronbach’s α = .86) and the third and last one consists of three items (Cronbach’s α = .75). This three factors were the basis for the subsequently built subscores. In order to build mean scores the ratings of all items belonging to a certain dimension were averaged.

To assess the collective self-efficacy expectations, the ‘Teacher Collective Self-Efficacy Expectations Scale’ (Schwarzer and Jerusalem 1999) was used. This scale consists of twelve items. The range of the scale was from 1 (not true at all) – 4 (completely right). Calculations for internal consistency and exploratory factor analysis were performed based on the data collected in the present study. Internal consistency (Cronbach’s α = .94) was very good. Exploratory factor analysis revealed a unidimensional factor structure. The calculated statistical coefficients were comparable with the statistical characteristics reported in the scale publication (Cronbach’s α = .92, unidimensional factor structure, N = 300).

Finally, to determine the general job satisfaction, the ‘General Job Satisfaction Scale’ (Bos et al. 2009) was used. The scale was initially used to determine the job satisfaction of secondary level teachers in Germany and consists of nine items (Bos et al. 2009). Calculations of internal consistency and exploratory factor analysis were performed based on the data collected in the present study. Internal consistency (Cronbach’s α = .91) was very good. Exploratory factor analysis revealed a unidimensional factor structure. Statistical characteristics were comparable with those reported in the scale publication (Cronbach’s α = .88, unidimensional factor structure, N = 774).

Besides these three instruments, some socio-demographic variables were collected, such as gender and age. The survey also asked in how many teaching teams the teachers currently worked and whether these teams were self-selected or institutionally composed. Finally, the survey also asked about the duration of the partnerships of every teaching team.

3.3. Procedures

Because it was possible that a teacher worked in multiple joint teaching arrangements, each teacher was provided with an adaptive survey, where the ‘Are we Really Co-Teachers Scale’ and the time spent in a specific team were questioned depending on the number of arrangements in which one teacher worked. This means if one teacher worked in two teaching teams he or she received two versions of the ‘Are we Really Co-Teachers Scale’, one for each team. This was done in order to allow a more precise and specific determination of classroom activities. In fact, the 321 teachers participating in this study worked in 692 teaching teams. More precisely, 91 teachers worked in one teaching team, 123 teachers worked in two teaching teams, 73 teachers in three and, finally, 34 teachers in four teaching teams. These teachers were jointly teaching at least two hours per team, per week and per subject in front of a specific classroom (bm:bf 2014a; Specht 2009). From these 692 teams, 182 were self-selected and 500 had been formed by the school administration. Ten teams lacked responses for this issue and were therefore excluded from further analysis. Moreover, 22 teams with Special Needs teachers were excluded as well. Thus, finally, only ratings concerning the team characteristics of 648 teacher teams were used in this study.

In order to ensure independence of the sample, which is one of the underlying assumptions for MANCOVA, the data depending on the teams (i.e. the factors of the ‘Are we Really Co-teacher scale’ and the working experience in the teacher teams) was aggregated on the teachers where they belonged. In other words, if one teacher worked in two self-selected teacher teams, the three factor scores of the ‘Are we Really Co-Teachers Scale’ for the two teams and the data concerning the joint teaching experience were averaged. This implicates that the status of the teachers was broken into three different groups (self-selected, institutionally composed, mixed). The first group (self-selected) consisted of 66 teachers, the second group (institutionally composed) consisted of 187 teachers and the third group (mixed) consisted of 39 teachers, who were working in more than one teacher team and had experienced both self-selection as well as assignment by the school administration.

4. Results

MANCOVA was used to reveal effects of different kinds of team composition on the subscores of the ‘Are we Really Co-Teachers Scale’, the ‘Teacher Collective Self-Efficacy Expectations Scale’ and the ‘General Job Satisfaction Scale’. Age, gender, joint teaching experience, and number of teaching teams where the participant worked were used as covariates in order to ensure that no structural characteristics of the teacher teams distorted the results of the analysis. Skewness and kurtosis of the dependent variables were examined and assured normality. The Box’s M test testing equality of the variance-covariance matrix was significant (p = .02). However, Tabachnick and Fidell (2001) have pointed out that significant test results can be ignored, since the test is not robust as soon as the group sizes differ. Levene-Tests were performed to assess the equality of the error variances. The test was significant for the subtest ‘Shared Responsibility’. Levene-Tests are also very sensitive particularly with unequal group sizes. Significant test results do not prevent following MANCOVA calculations as long as some adjustments are performed. Therefore, Pillai’s criterion was used to assess statistically significant differences. Pillai’s trace is recommended for MANCOVAs with unequal group-sizes and violations of assumptions (Tabachnick and Fidell 2001). Additionally, according to the recommendations of Tabachnick and Fidell (2001), a more conservative critical alpha level of α = .01 was used in this study.

There was a highly significant effect of the kind of team composition on the dimensions ‘Shared Responsibility’ and ‘Enjoyment with the Co-Teaching Process’ (Pillai’s V = .088, F(_10,566) = 2.62, p = .004).

Means, standard deviations of the three groups and MANCOVA results are presented in Table 1.

Table 1. Means, standard deviations and MANCOVA comparison for the three kinds of teacher teams (N = Self-selected 66; Mixed 39; Institutionally composed 187).

Dimension	Team composition	M (SD)	F	p	η ^2
Shared responsibility	Self-selected	4.42 (.59)	5.82	.001	.036
	Mixed	4.08 (.63)
	Institutionally composed	4.02 (.80)
Skills and external resources	Self-selected	3.63 (.77)	2.24	.109	.015
	Mixed	3.57 (.62)
	Institutionally composed	3.38 (.79)
Enjoyment with the Co-teaching process	Self-selected	4.28 (.72)	7.02	.001	.047
	Mixed	4.11 (.67)
	Institutionally composed	3.85 (.84)
Self-efficacy expectations	Self-selected	3.07 (.64)	2.16	.075	.018
	Mixed	3.16 (.68)
	Institutionally composed	2.93 (.51)
Job satisfaction	Self-selected	2.86 (.21)	3.07	.048 ^*	.021
	Mixed	2.87 (.29)
	Institutionally composed	2.79 (.27)

Notes:

df: 2/292; scaling = 1: lowest – 5: highest. For Self-Efficacy Expectations and Job Satisfaction 1: lowest – 4: highest.

^* Please note that a more conservative alpha level of .01 is used instead of .05, to determine significant effects.

According to the presented results, significant effects of the kind of team composition were revealed for the dimensions ‘Shared Responsibility’ and ‘Enjoyment with the Co-Teaching Process’. No significant effects were found for the other dimensions. Post-hoc tests (Bonferroni corrected) confirmed these results. In the dimension ‘Shared Responsibility’ the self-selected group has significantly higher mean scores than the group composed by the school administration (p = .001, SE = .107). The same was true for the dimension ‘Enjoyment with the Co-Teaching Process’: the self-selected teams showed significantly higher ratings than the institutionally composed teams (p = .015, SE = .117). The highest proportion of explained variance was found in the dimension ‘Enjoyment with the Co-Teaching Process’ (η ²  = .047).

To examine the differences within the mixed group, the means regarding the different kind of teams from teachers who work in both kinds of teams are presented in Table 2. The ratings of the 39 teachers who worked in both kinds of teams were separated for the self-selected and institutionally composed teams.

Table 2. Means and standard deviations for the teams of teachers engaged in the mixed group (N = 39).

Dimension	Team composition	M (SD)	Number of teams
Shared responsibility	Self-selected	4.37 (.55)	57
	Institutionally composed	3.84 (.84)	61
Skills and external resources	Self-selected	3.62 (.63)	57
	Institutionally composed	3.27 (.70)	61
Enjoyment with the Co-teaching teaching process	Self-selected	4.38 (.73)	57
	Institutionally composed	3.75 (.72)	61

The results of paired t-tests supported the findings of the MANCOVA. In the dimensions ‘Shared Responsibility’ (t ₍₃₈₎= 4.138; p < .001) and ‘Enjoyment with the Co-Teaching Process’ (t ₍₃₈₎= 4.353, p < .001) self-selected teams were rated significantly more positively than the teams composed by the school administration. In addition, the same was found for the dimension ‘Skills and External Resources’ (t ₍₃₈₎= 4.618 p < .001).

5. Discussion and conclusion

The results confirm the first research hypothesis regarding the sharing of responsibilities. This result indicates that teachers have an implicit knowledge about whom they are compatible with and able to share responsibility in class. As already outlined by Rytivaara and Kershner (2012), this result was in some ways expected because professional teachers know from their own experience what teaching style they are more familiar with. It is obvious that teachers would rather choose teammates with a similar teaching style than those that teach differently. Along these lines, similarities in the conception about teaching also make it much easier to share responsibility when working together in the same classroom.

According to the results of MANCOVA and t-tests, research hypothesis (ii.) – regarding skills and external resources – cannot definitely be answered. Though the means of the ratings of self-selected teams are slightly higher than those of institutionally composed teams, the MANCOVA results failed to reach statistical significance. In contrast, t-tests for dependent samples comparing the ratings of teachers who worked in both kinds of teams revealed significant differences in this dimension as well. From a theoretical basis, the results regarding teachers who work in both kind of teams should be regarded as more valid, because these teachers gained experience in both kind of teams.

MANCOVA results confirm research hypothesis (iii.), regarding expected differences in the dimension ‘Enjoyment with the Co-Teaching Process’. Teachers working in self-selected teams reported considerably more enjoyment than their colleagues in teams composed by the school administration. The contrast between the two groups was highly significant. Moreover, it had also the highest effect size of all tested variables.

In contrast to the existing literature, hypotheses (iv.) and (v.) regarding collective self-efficacy and job satisfaction were not supported by our data. These results were not expected because questioned teachers often mentioned that they felt more effective and more satisfied with their job in self-selected teacher teams. One reason for this result may be found in the circumstance that the existing literature mainly consists of single case studies which rely only on a very limited number of participants. In this regard, it may be possible that only a restricted perspective is provided. Another reason may be the fact that self-selected team composition does not ensure that specific skills necessary for good quality joint teaching are available. In this regard, it is very possible that teaching efficacy is limited. Moreover, the kind of team composition does not have any influence on general job satisfaction. One reason for this result may be found in the circumstance that the questioned teachers only spent a limited time in jointly taught lessons compared to the time they spent with traditional teaching. In this regard, it seems possible that other factors may be more important for the job satisfaction than the kind of team composition.

According to these results and along the lines of the studies by Friend et al. (2010) and Kain (2006), self-selection of team partners is no guarantee for successful and effective work in teacher teams, although the results of the present study suggest a further differentiation of this statement. Given the fact that the results of this study indicate the biggest difference between self-selected and institutionally composed teaching teams in the dimension ‘Enjoyment with the Co-Teaching Process,’ this is somewhat ambivalent. At first glance it seems that Kain’s argument (2006) that self-selected teacher teams run the risk of being more determined by social pleasure and social meeting rather than by purpose is supported because teachers in self-selected teams did not show higher ratings of collective self-efficacy. Nevertheless, as already mentioned, this dimension is ambiguous and enjoyment with the joint teaching process can have positive as well as negative consequences. A closer inspection of the items which represent this dimension reveals that it refers to success in the teacher team, mutual trust and fun with the standards and the joint teaching itself (see above). Therefore, it seems more likely that the dimension ‘Enjoyment with the Co-Teaching Process’ has a positive influence on the quality of the joint teaching and cannot be solely described as enjoyable social interaction. In fact, it seems that the enjoyment dimension examined in this study refers much more to the positive characteristics of enjoyment during the working process and the related positive motivational aspects.

The results of the present study, however, also indicate that teachers know from their own experience and knowledge whom they are more likely compatible with and with whom they will be able to share classroom responsibilities more easily. The sharing of responsibility for a number of actions in classroom – like shared delivery of instruction, shared responsibility for assessment or shared responsibility for the decision how to teach – is considered as essential for high quality joint teaching. Therefore, self-selection of team partners seems to support a higher degree of ‘shared responsibility’ than institutional composition of teams. The professional knowledge and experience of teachers can serve as an important source of information for team composition processes. Administrators should use this experience and knowledge of their teachers to create teacher teams in which the teammates are most likely compatible with each other. Moreover, if this is not possible and teachers are simply put together, these teams should be provided with additional help such as team building measures. According to Biemann and Weckmüller (2012), this would be particularly important in order to clarify roles and help in the negotiation of personal differences. Self-selected teams, on the other hand, should be aware of the risk of being determined rather by fun than by purpose. It seems appropriate that these teams also should seek additional help, particularly regarding the specific joint teaching skills and techniques used in jointly taught classrooms to ensure that high quality teaching is provided and thus the benefits of good quality joint teaching are achieved.

6. Limitations of this study

Because this study is a cross-sectional study, conclusions can only be drawn in terms of correlations. Therefore, the results of this study should be validated in an experimental design to determine causal relations and assess the real effect of different kinds of team compositions on student achievement. Additionally, the national context in which this study was conducted, as always, plays a significant role (Rytivaara and Kershner 2012). Nevertheless, the working reality of teachers in teaching teams in Austria is not that far distinct from teachers in other countries. Therefore, it is very likely, that some general results of this study can be found in other countries as well.

Another important limitation of this study is the fact that the participants were self-selected. This is in fact a weakness of the study because the authors had no control over the sample selection, so the representative nature of the sample is in question. Therefore, it could have happened that mainly teachers participated in the survey who were very satisfied or dissatisfied with joint teaching. However, the range of the answers of the participants revealed that the whole spectrum of the scales was used. Put in another way, teachers who were satisfied as well as those who were dissatisfied with joint teaching participated in the survey. Moreover, most statistical characteristics of the sample regarding socio-demographic variables as well as teaching- and joint teaching experience of the participants are comparable with the characteristics of the relevant teacher population in Austria. Therefore, there seems to be a relatively low risk of a biased sample.

At first glance, the sample size is not very large and group sizes differ considerably. That said, the statistical tests used in this study are usually regarded as appropriate to detect even moderate effects at an α-level of p = .05 (Dattalo 2008). However, it must be kept in mind that sample size estimations for MANCOVAs are more difficult to calculate than for MANOVAs and can rather be seen as approximations (Dattalo 2008).

Finally, it is important to notice that this study is examining perceptions of the questioned teachers and does not observe classroom behaviour of the teachers. Hence, the main dependent variables are based on teacher ratings about their own behaviour and not on observed behaviour of the teachers.

Notes on contributors

Mathias Krammer is a post doc at the University of Graz. His main research interests are in educational assessment, inclusive education and socio-emotional development.

Peter Rossmann is an Associate Professor at the Department of Education, University of Graz, now retired. His main research interests lie in the fields of developmental psychopathology and educational assessment.

Angela Gastager , is a University College Professor for Science of Education and Research on Pedagogical Interaction at the University College of Teacher Education Styria in Austria.

Barbara Gasteiger-Klicpera is a full professor at the University of Graz. Her main research interests are in the fields of inclusive education, reading and teachers’ professional development

Disclosure statement

No potential conflict of interest was reported by the authors.