The development of an instrument to measure teachers’ inquiry habit of mind

ABSTRACT

In the present study the construct inquiry of mind explored and an instrument to measure it is proposed. Using three different samples, explorative and confirmative factor analyses were performed, resulting in three empirical dimensions that correspond to the three theoretical dimensions: 1) ‘value deep understanding,’ 2) ‘reserve judgment and tolerate ambiguity,’ and 3) ‘taking a range of perspectives and posing increasingly focused questions.’ Our findings suggest the teachers’ inquiry habit of mind scale has good psychometric properties making it useful not only for research that investigates teachers’ research attitude and intention to do research but also as an evaluation tool for the development of an inquiry habit of mind in both student teachers and teacher educators (in teacher education) as well as in experienced teachers (participating in professional development).

KEYWORDS:

• Inquiry habit of mind
• T-IHMS
• teacher professional development
• teacher education

Teacher quality has been identified as the most important factor affecting students’ learning (Cornet et al. 2006; Hattie 2009). Consequently, a lot of effort has been put into the professional development of teachers and teacher educators as they are teachers themselves (Cochran-Smith and Lytle 2009b). These efforts fit within one of the trends over the past 15 years in teacher education and teacher professionalization that is oriented toward research and inquiry (Livingston and Flores 2017; Munthe and Rogne 2015). Conducting research is a form of professionalization because it provides teachers with a deeper insight in the quality of their practices, which may lead to improved practices (Cochran-Smith and Lytle 2009b). An example of such improvement is the use of a particular instructional method as a result of conducting a research to determine more effective instructional methods for increasing students’ motivation. Besides that the school context is important in facilitating research, not all teachers are inclined to conduct research and factors such as a culture of inquiry (Katz, Sutherland, and Earl 2005), collaboration with colleagues (Kroll 2005), support of informed leaders who have sufficient knowledge (Timperley 2010), and – on a personal level – an inquiry habit of mind (Earl and Katz 2002), may underlie their intentions. In the present study, we specifically focused on the latter factor, the inquiry habit of mind (Earl and Katz 2006), which is a set of pre-dispositions, teachers may possess to reflect on certain phenomena and situations inferred from data made available to them. Based on these reflections, teachers can make decisions in order to improve the quality of education whether this is for teacher education or for students’ learning in schools.

Notwithstanding the growing body of literature on inquiry habit of mind – or inquiry as stance (Cochran-Smith and Lytle 2009a) – there still is a lack of a valid instrument that measures it. This inhibits research on how inquiry habit of mind can be positively influenced and how it, in turn, affects, for example, teachers’ research attitude (a favorable versus unfavorable position toward doing research; Ajzan 2005), intention to do research (Fishbein and Ajzen 2010), innovative behavior, and school development. Therefore, the present research is proposing such instrument for measuring teachers’ inquiry habit of mind: the Teachers’ Inquiry Habit of Mind Scale (T-IHMS) that can be used for teacher professional development or added as an evaluation tool in the curriculum in teacher education.

This article starts with a theoretical framework about inquiry habit of mind. We continue by constructing the items for the questionnaire that measures inquiry habit of mind, thereby taking into account the three theoretical dimensions of it as suggested by Earl and Katz (2002). The next section describes how we validated this instrument using explorative and confirmative factor analyses. The article ends with a discussion and conclusion.

Inquiry habit of mind

Although not a new idea, in recent years, an increasing amount of attention has been paid to the importance of teachers developing an inquiry habit of mind in the Netherlands (Imants et al. 2010; Krüger 2010; Leeman and Wardekker 2010; Meijer et al. 2010, 2013), as well as internationally (Cochran-Smith and Lytle 2009a; Earl and Katz 2002; Katz, Sutherland, and Earl 2005; Kroll 2005; So 2013; Zuidema 2012). Presently, there is a general agreement that teachers should not only be knowledge experts in their subject area, but should possess certain behavioural pre-dispositions including an inquiry habit of mind (Hodkinson and Hodkinson 2004; Kroll 2005; Meijer et al. 2013; Osguthorpe 2008; Talbert-Johnson 2006; Thornton 2006). The same arguments are found in the literature on teacher educators (Livingston and Flores 2017; Munthe and Rogne 2015; Tack and Vanderlinde 2014).

The concept of an inquiry habit of mind has first been introduced by Lorna Earl and Steven Katz in the beginning of the 21th century (Earl and Katz 2002, 2006). They refer to it as a way of thinking, represented by a set of pre-dispositions, which are patterns of thinking that determine how a person is disposed to act (Talbert-Johnson 2006; Thornton 2006; Van der Rijst 2009). More specifically, this way of thinking is a data-driven dynamic iterative system to organise ideas, search for information, and move closer to understanding some phenomenon (Earl and Katz 2002, 2006). The concept is closely linked to an inquiry stance, introduced by Cochran-Smith and Lytle (2009a). Moreover, various concepts, very similar to an inquiry habit of mind, have been described such as critical thinking (Abrami et al. 2008), reflective practice (Jay and Johnson 2002; Reid 2004), and a scientific or academic attitude (Gauld 2005; Meijer et al. 2013; Van der Rijst 2009). Although explicitly not the same (see for more information; Cordingley 2003; Reid 2004), inquiry and conducting research are very closely related. The concept of an inquiry habit of mind draws distinct parallels with the definition of scholarship, a term that has mainly been used with regard to university teachers (Andresen 2000; Hutchings 2010; Meijer et al. 2013) We recognize that there are many similarities between these concepts and do not wish to discard any of them. For the remainder of this article, however, Earl and Katz (2002, 2006) terminology and definition of the inquiry habit of mind is used.

An inquiry habit of mind is a set of psychological dispositions that allow teachers to better deal with the complex and ever changing circumstances of school life, and strengthens conditions for improving teaching and continuous professional development (Cochran-Smith and Lytle 2009a; Katz and Earl 2010; Kroll 2005; Van der Donk and van Lanen 2010). An inquiry habit of mind is necessary for teachers to make sense of the vast amount of data, which is increasingly available in schools. Examples are student performances, degree to which ICT is used during lessons, and statistics about lesson cancellations. Data are an important part of a continuing process of analysis, insights, new learning and adjustments in practice. Teachers, themselves, need to be the main consumers of data in the process of decision-making with regard to classroom practices (Earl and Katz 2002, 2006). Besides making sense of the vast amount of data, the necessity for this inquiry habit of mind stems from teachers’ tasks in the 21st century to develop their students’ learning dispositions and capacities to think flexibly and creatively. To be able to teach these 21st century skills teachers and teacher educators themselves should possess and model these capacities (Reid 2004).

For all of these reasons, it is important to investigate the extent in which teachers and teacher educators possess an inquiry habit of mind. Therefore, the main goal of the present study was to develop a valid instrument for measuring the inquiry habit of mind, by means of establishing the psychometric properties in three different samples of teachers and by conducting a broad variety of statistical analyses.

As stated before the term habit of mind is linked to inquiry to underline a way of thinking and that it is a different construct than an academic or research attitude. Whereas inquiry habit of mind is general and closely related to personality traits, research attitude is specific and closely related to intention to perform or not to perform research. Research attitude and intention to perform research has to be understood from the reasoned action approach (RAA) framework of Fishbein and Ajzen (2010). In this framework, attitude toward a specific behavior is seen as one of the factors that influence the intention to perform that specific behavior (Ajzan 2005). Moreover, irrespective of the subject domain, attitude seems to be a strong predictor for intention. For example, Kreijns et al. (2013) proved that teachers’ attitude toward ICT use is an important predictor for the intention to actually use ICT.

Although inquiry habit of mind is closely related to personality traits, it can be influenced (Nelson 2015). By engaging in research, teachers and teacher educators may generally develop a much stronger inquiry habit of mind (e.g. Hall et al. 2006; Meijer et al. 2013; Rust and Meyers 2006; Snoek and Moens 2011; Tack and Vanderlinde 2014; Van der Donk and van Lanen 2010; Zeichner 2003). This reinforcement of inquiry habit of mind, thus, requires that teachers must somehow continuously be immersed in researched activities and be busy with collecting data, interpreting them, and drawing conclusions. To be able to handle the increasing amount of available data in schools, the instrument could be part of the curriculum of teacher education.

Teachers’ inquiry habit of mind scale (T-IHMS)

In line with Earl and Katz (2002, 2006) and Katz, Sutherland, and Earl (2005), we distinguished between three dimensions of inquiry habit of mind, namely 1) ‘value deep understanding,’ 2) ‘reserve judgment and tolerate ambiguity,’ and 3) ‘taking a range of perspectives and posing increasingly focused questions.’ For each dimension corresponding items were constructed based on the specific descriptions of the dimensions (see Table 1). The first dimension ‘value deep understanding’ is reflected by a tendency not to presume a certain outcome, but allow for a range of outcomes and keep searching for increased clarity and understanding (item 1.1 and 1.2) (Earl and Katz 2006, 18). It is important not to interpret new ideas within existing frameworks because it leads to making only superficial changes to practice. For professional learning much deeper changes in practice are required (Muijs et al. 2014), as is searching for other ways to learn from (item 1.3). Besides, asking for feedback from others about what they think of one’s own work and collecting information to evaluate one’s own work are important ways of building deeper understanding (item 1.4 and 1.5). The second dimension ‘reserve judgment and tolerate ambiguity’ refers to keeping an open mind for different explanations, so one should not too quickly support some particular idea or theory. There is a tolerance for uncertainty (item 2.2) in the process of finally finding better answers and deeper understanding (item 2.4) and a willingness to live in the dissonance long enough to investigate ideas until there is some clearness about its possible meaning (item 2.3) (Earl and Katz 2006, 18). A term which has been mentioned in this context is suspension of belief (Gauld 2005), which is required if sufficient evidence to make a decision is not available and so one should not support some particular idea or theory too quickly (item 2.1). This is reflected by a tendency not to presume a certain outcome, but allow for a range of outcomes and keep searching for increased clarity and understanding (item 2.4). The third dimension ‘take a range of perspectives and systematically pose increasingly focused questions’ refers to the systematic consideration of a problem from a range of perspectives through which one gains sufficient evidence to explain, support and challenge points of view and includes the realization that using data almost never provides answers. Instead, it usually leads to a more and more focused investigation and better in depth questions (Earl and Katz 2006, 18). As Kroll (2005) explained, by the cyclical process of raising, refining, and investigating questions and examining data, more thoughtful decisions can be made with regard to changing one’s own practice. By systematically considering a problem from a range of perspectives (item 3.2), instead of approaching problems from one point of view or being prejudiced (item 3.3) one gains sufficient evidence to explain, support and challenge different points of view (item 3.4).

Table 1. The teachers’ inquiry habit of mind scale (T-IHMS).

		TNS/NIPO (n = 1228)		OIG (n = 503)		DUO (n = 1195)
nr item	item	M	SD	M	SD	M	SD
Preamble: When I work at school …
Value deep understanding
1	… I am critical on whether I did the right thing	5.35	1.05	5.55	.91	5.83	1.03
2	… I wonder if I can improve my work	5.32	1.09	5.57	.99	5.92	.89
3	… I watch how colleagues do things in order to learn from them	5.37	1.01	5.32	1.02	5.64	1.04
4	… I ask others what they think of my work	4.65	1.25	4.80	1.20	4.92	1.35
5	… I try to collect information so that I can evaluate my work	4.89	1.13	5.04	1.06	5.13	1.21
	Cronbach α	.74		.74		.76
Reserve judgment and tolerate ambiguity
1	… I refuse to accept unwarranted assertions and explanations irrespective of how plausible they might be	4.52	1.27	4.53	1.37	4.55	1.56
2	… I have a certain tolerance for uncertainties and ambiguities in offered solutions and explanations	4.57	1.12	4.49	1.16	4.61	1.18
3	… I am willing to accept some uncertainty provided that finally there is insight into proven solutions and reasonable explanations	4.68	1.18	4.74	1.14	4.82	1.29
4	… I can deal with situations wherein solutions and explanations are not yet available	4.95	1.10	5.17	.98	5.04	1.20
	Cronbach α	.74		.74		.75
Take a range of perspectives and systematically pose increasingly focused questions
1	… I try, when it comes to sort out things, to pose increasingly better and more targeted questions	4.91	1.07	5.22	1.01	5.48	.96
2	… I try to view things from other perspectives	5.13	1.04	5.44	.91	5.64	.97
3	… I try to avoid prejudices with regard to solutions and explanations	5.00	1.06	5.28	.95	5.48	1.03
4	… I try, by means of a systematic approach to investigations, to find evidence for solutions and explanations	4.64	1.19	4.96	1.16	5.24	1.13
	Cronbach α	.73		.79		.75
	Cronbach α (total scale)	.86		.83		.81

Notes:

• Judgements were made on 7-point Likert scales (1 = not applicable at all on me, …, 7 = totally applicable on me).

• Grey text represents items that were ultimately removed and, therefore, were not included in the final T-IHMS.

• Cronbach α was only calculated for the items that were included in the final T-IHMS.

The resulting T-IHMS is seen in Table 1. In this table, some findings of our statistical analyses (see the next section) have already been incorporated to reduce the number of tables. In particular, the grey items represent items that were ultimately removed thereby constructing the final scale. Responsibility for the three samples were respectively the research agencies TNS/NIPO (http://www.tnsglobal.com/), OIG (http://oig.nl/), and DUO (http://duo.nl). For each sample the means and standard deviations are reported as well as the internal consistency of the items (Cronbach α) of each dimension of the final scale.

Method

Participants and procedure

Three samples of teachers were recruited as participants for the study (respectively, n = 1228; n = 503, and n = 1195) over a period of three years by three different research agencies (respectively, TNS/NIPO, OIG, and DUO). The recruited teachers belonged to one of the agency’s panels. These panels consisted of teachers who on a regular basis voluntarily answer questionnaires on various subjects for a small fee. All demographic data are shown in Table 2. Teachers received an email with an invitation to participate in the study with a link to the online questionnaire. Finishing the complete questionnaire took approximately 20 minutes, the subsection of the questionnaire representing T-IHMS scale in the present study took approximately three minutes. The items comprising the three dimensions of T-IHMS were intermixed before administering them to participants. By intermixing items, it was avoided that otherwise measurement bias would occur, such as higher reliability scores (Goodhue and Loiacono 2002). Indeed, research of Davis and Venkatesh (1996), and the replica of their study by Goodhue and Loiacono (2002), found fairly strong evidence that intermixing items created a slightly higher actual reliability but with Cronbach’s alpha’s showing lower values. Intermixing items, however, might have confused respondents (Davis and Venkatesh 1996).

Table 2. Demographics of the participants of the three samples.

				gender		sector
nr sample	research agency	n	age M(SD)	male	female	primary education	secondary education	vocational education	other*
1	TNS/NIPO	1228	44.45(12.49)	457	771	533	492	121	218
2	OIG	503	48.43(9.90)	256	247	70	365	54	17
3	DUO	1195	46.16(10.87)	573	756	533	502	321	-

*‘other’ included higher education, university, special education, or not known.

Analysis

Mplus version 7.31 (Muthén & Muthén, 1998–2012) was used to perform explorative and confirmative factor analyses on the items of the raw and final T-IHMS respectively. The objective of these analyses was to confirm that the items loaded on factors whose interpretations corresponded with the three theoretical derived dimensions. To that end, we first used two of the three samples for carrying out exploratory factor analyses (EFA’s) in order to create a set of ‘approved’ items and to determine the number of empirical dimensions; it was expected that there would be three dimensions. The last sample was then used for carrying out a confirmatory factor analysis (CFA) on the set of approved items to confirm that the theoretically derived three-factor structure is empirically supported. Finally, the reliability of the T-IHMS was established (see Table 1).

Results

Explorative factor analyses (sample 1)

EFA’s with Oblimin rotation were performed on the first sample (n = 1228). Oblimin rotation was chosen because of the expectation that factors would correlate (Tabachnick and Fidell 2007). To validate this assumption, the correlations among the factors are given within Table 3. Four different solutions were compared: the 1-factor solution, the 2-factor solution, the 3-factor solution and the 4-factor solution. The EFA’s used all items¹ of the raw Teachers’ Inquiry Habit of Mind Scale. The first theoretical dimension was represented by the items 1.1–1.5, the second one by the items 2.2–2.4, and the last one by the items 3.1–3.4 (see also Table 1). Table 3 depicts the results for each EFA. Note that we used a cut-off value of .40 (Comrey and Lee 1992; Stevens 1992) for the factor loadings; that is, only items that load higher than .40 were the items that retained in the pool of items that represent a factor.

Table 3. Factor loadings for respectively the 1-, the 2-, the 3-, and the 4-factor solution (n = 1228).

		1-factor solution	2-factor solution		3-factor solution			4-factor solution
no. item	item				1	2	3	1	2	3	4
1.1	… I am critical on whether I did the right thing	.687*	.796*	−.099*	.684*	−.035	.126*	.640*	− .023	.151*	.036
1.2	… I wonder if I can improve my work	.601*	.665*	−.054	.698*	.007	−.033	.693	.013	−.009	− .012
1.3	… I watch how colleagues do things in order to learn from them	.608*	.600*	.026	.602*	.099*	−.024	.559*	.108*	− .032	.109
1.4	… I ask others what they think of my work	.587*	.559*	.048	.359*	.102*	.200*	.279*	.112*	.101	.339*
1.5	… I try to collect information so that I can evaluate my work	.710*	.722*	.008	.304*	.047	.464*	.177*	.038	.445*	.299*
2.2	… I have a certain tolerance for uncertainties and ambiguities in offered solutions and explanations	.590*	−.043*	.768*	−.068*	.775*	.008	.074*	.775*	− .017	−.010
2.3	… I am willing to accept some uncertainty provided that finally there is insight into proven solutions and reasonable explanations	.566*	.037	.640*	.012	.645*	.018	.026	632*	.027	−.094*
2.4	… I can deal with situations wherein solutions and explanations are not yet available	.597*	.079	.625*	.038	.639*	.030	.057	.627*	.079	.087
3.1	… I try, when it comes to sort out things, to pose increasingly better and more targeted questions	.714*	.699*	.037	.255*	.055	.515*	.187	.030	.654*	−.087
3.2	… I try to view things from other perspectives	.752*	.514*	.291*	.391*	.354*	.105*	.361*	.350*	.140*	.016
3.3	… I try to avoid prejudices with regard to solutions and explanations	.677*	.376*	.363*	.334*	.420*	.016	.329*	.419*	.032	.002
3.4	… I try, by means of a systematic approach to investigations, to find evidence for solutions and explanations	.650*	.598*	.078	−.065*	.023	.837*	− .093*	.042	.788*	.044
	factor
	1		-	.71*	-	.63*	.67*	-	.67*	.60*	.34
	2				-	-	.63*	-	-	.62*	.32
	3							-	-	-	.40*

* significant at 5% level.

Notes:

• Judgements were made on 7-point Likert scales (1 = not applicable at all on me, …, 7 = totally applicable on me).

• Grey numbers represent items with factor loadings < .40.

• Bold numbers represent items with factor loadings > .40 on their target factor.

• Bold italic numbers represent items with factor loadings > .40 but not on their target factor.

The significant correlations between the factor solutions (see Table 3) justified the use of the Oblimin rotation (Tabachnick and Fidell 2007). Comparing the four different factor solutions, the factor loadings on the 1- and 2-factor solution were relatively high. Though it was hypothesized that the 3-factor solution would best fit the theory, this solution was statistically not optimal because of items that loaded on other factors than the target factor (i.e. with loadings above .40) (item 1.5 and 3.3) or had low loadings on the target factor (i.e. below .40) (item 1.4). When inspecting the factor loadings of the 4-factor solution, it can be seen that all item loadings on the fourth factor did not exceeded .40 indicating poor loadings and, in addition, these loadings were mostly not significant at the 5% level. Therefore, this fourth factor of the 4-factor solution had no value for interpretation. Succinctly, these EFA’s favoured a 2-factor solution.

Also, the eigenvalues for the sample correlation matrix for the EFA’s were calculated (see Table 4). Only two factors showed eigenvalues greater than 1 suggesting that the T-IHMS has two instead of three dimensions. Thus, here too, a 2-factor solution was favoured.

Table 4. Eigenvalues for sample correlation matrix (n = 1228).

factor	eigenvalues	factor	eigenvalues
1	5.600	7	.501
2	1.071	8	.480
3	.812	9	.451
4	.672	10	.422
5	.620	11	.412
6	.578	12	.382

To find more support for the 2-factor solution, the fit indices for the different factor solutions were investigated. First for the model fit the χ² statistic was used. However, when model complexity increases and with increasing sample size, the χ² statistic increases (Hu and Bentler 1999). For better comparison the comparative fit index (CFI; Hoyle 1995; Marsh, Balla, and Hau 1996), the Tucker-Lewis index (TLI; Tucker and Lewis 1973), the root mean square error of approximation (RMSEA; Browne and Cudeck 1989), and the standardized root mean square residual (SRMR; Bentler 1995) were used. Criteria for the fit indices were based on Browne and Cudeck (1989). They stated that RMSEA should have values between .06 and .08 for a good fit and values of .05 and less for a very close fit. CFI and the TLI should have a value of .95 or above to indicate a good fit. Finally, a value less than .08 for SRMR is generally considered a good fit (Hu and Bentler 1999).

When fit indices for the different factor solutions were investigated, we looked at changes in fit indices for each of the four EFA’s (see Table 5). These changes (i.e. big changes in RMSEA, CFI and TLI) suggested a much better fit for the 3-factor solution rather than the 2-factor solution also reflected in the ${\chi }^2$-statistics: ${\chi }_{change}^2=250.267-92.285=157.982$; $d{f}_{change}=43-33=10$; both critical values (cv) were respectively $c{v}_{p=.05}=18.31$ and $c{v}_{p=.01}=23.21$ so both these cv’s were $\ll 157.982$ indicating a big improvement. The 4-factor solution seemed to give even a better fit than the 3-factor solution, but this improvement was not dramatic (i.e. small changes of RSMEA, CFI and TLI) also reflected in the ${\chi }^2$-statistics: ${\chi }_{change}^2=92.285-58.839=33.446$; $d{f}_{change}=33-24=9$; both cv’s were respectively $c{v}_{p=.05}=16.92$ and $c{v}_{p=.01}=21.67$ so both these cv’s were $<33.839$ indicating an improvement but not as big as the transition from the 2-factor solution to the 3-factor solution. Therefore, this analysis on changes in fit indices favoured the 3-factor solution. In conclusion, the first sample was inconclusive whether the 2-factor or the 3-factor solution is the favoured solution based on statistical grounds.

Table 5. Goodness of fit EFA’s for the four factor solutions (n = 1228).

	factor solutions
test	1 factor solution	2 factor solution	3 factor solution	4 factor solution	criteria
Chi-Square Test of Model Fit
value	555.235	250.267	92.285	58.839
egrees of freedom	54	43	33	24
p-value	.000	.000	.000	.000
RMSEA
estimate	.087	.063	.038	.034	should be <.06 for good/close fit
90% C.I.	.080, .094	.055, .070	.029, .048	.023, .046	C.I should encompass values <.06
probability RMSEA ≤ .05	.000	.003	.982	.990
CFI/TLI
CFI	.914	.964	.990	.994	should be > .95 for good fit
TLI	.894	.945	.980	.983	should be > .95 for good fit
Standardized Root Mean Square Residual
value	.046	.027	.015	.012	should be <.08 for good fit

Note: Grey italic numbers represent values that do not meet the criteria for good/close fit

Explorative factor analyses (sample 2)

The second sample was used to further explore to determine whether the 2-factor or the 3-factor solution is favored. EFA’s with Oblimin rotation were performed on the second sample (n = 503). From the analyses on the first sample, the items 1.4, and 3.2 were not included in these EFA’s, because of too low loadings (< .40) on all factors in the 3-factor solution of the previous analysis on the first sample. Two different solutions were compared: the 2-factor and the 3-factor solution, see Table 6. The eigenvalues for the sample correlation matrix for the EFA’s are shown in Table 7. In this sample, three factors have eigenvalues above 1 suggesting that T-IHMS comprised three dimensions. Nevertheless, the question remained whether the items within each factor correspond with the three theoretical derived dimensions proposed by Earl and Katz (2002, 2006).

Table 6. Factor loadings for respectively the 2- and the 3-factor solution (n = 503).

		2-factor solution		3-factor solution
no. item	item	1	2	1	2	3
1.1	… I am critical on whether I did the right thing	.642*	.134*	.648*	−.017	.158*
1.2	… I wonder if I can improve my work	.798*	−.038	.780*	.022	.027
1.3	… I watch how colleagues do things in order to learn from them	.629*	.008	.606*	.080	−.014
1.5	… I try to collect information so that I can evaluate my work	.274	.469*	.270*	.117*	.422*
2.2	… I have a certain tolerance for uncertainties and ambiguities in offered solutions and explanations	.051	.484*	−.012	.644*	.122
2.3	… I am willing to accept some uncertainty provided that finally there is insight into proven solutions and reasonable explanations	.139*	.430*	.055*	.886*	−.056*
2.4	… I can deal with situations wherein solutions and explanations are not yet available	−.150*	.732*	−.149*	.372*	.491*
3.1	… I try, when it comes to sort out things, to pose increasingly better and more targeted questions	.149*	.707*	.171*	−.006	.743*
3.3	… I try to avoid prejudices with regard to solutions and explanations	.028	.674*	.052	.176	.546*
3.4	… I try, by means of a systematic approach to investigations, to find evidence for solutions and explanations	−.048	.748*	−.040	.069*	.844*
	factor	Oblimin factor correlations
	1	-	.54*	-	.37*	.47*
	2			-	-	.43*

* significant at 5% level.

Notes:

• Judgements were made on 7-point Likert scales (1 = not applicable at all on me, …, 7 = totally applicable on me).

• Grey numbers represent items with factor loadings < .40.

• Bold numbers represent items with factor loadings > .40 on their target factor.

• Bold italic numbers represent items with factor loadings > .40 but not on their target factor.

Table 7. Eigenvalues for sample correlation matrix (n = 503).

factor	eigenvalues	factor	eigenvalues
1	4.300	7	.440
2	1.331	8	.396
3	1.107	9	.354
4	.693	10	.324
5	.568
6	.488

The factor loadings shown in Table 6 show that all the items of the three theoretical derived dimensions loaded on their target factors, except for item 1.5 and item 2.4 in case of the 3-factor solution. But as item 1.5 also did not load in the first sample on its target factor, whereas item 2.4 did, we only removed item 1.5 for the next analyses, and preserved item 2.4. To conclude our analyses with this second sample, the fit indices for the two different factor solutions were investigated, see Table 8. The fit indices suggested that the 3-factor solution was a clear better solution (model) than the 2-factor solution. Therefore, it was not needed to also investigate the changes in fit indices as was done in the first sample.

Confirmatory factor analyses (sample 3)

CFA was used on the third sample (n = 1195). A 1st- and 2nd-order model of the T-IHMS was tested (see Table 9 for the factor loadings and Table 10 for the fit indices). Items not included were the items, 1.4, 1.5, 2.1 and 3.2 because of the previous analyses on the first and second sample. With respect to the 2nd-order model, Mplus suggested a better fit if item 2.2 would covariate with item 2.3 and 2.4; the results for this modification are also shown in Tables 9 and 10.

Table 8. Goodness of fit EFA’s for the two factor solutions (n = 503).

	factor solutions
test	2 factor solution	3 factor solution	criteria
Chi-Square Test of Model Fit
value	264.897	45.149
degrees of freedom	26	18
p-value	.000	.000
RMSEA
estimate	.135	.055	should be <.06 for good/close fit
90% C.I.	.121, .150	.035, .075	C.I should encompass values <.06
probability RMSEA ≤ .05	.000	.320
CFI/TLI
CFI	.869	.985	should be > .95 for good fit
TLI	.773	.963	should be > .95 for good fit
Standardized Root Mean Square Residual
value	.059	.020	should be <.08 for good fit

Note: Grey italic numbers represent values that do not meet the criteria for good/close fit.

Table 9. Factor loadings for 3-factor solution (using STDYX standardization (n = 1195).

		3-factor Solution (1st-order without modification)			3-factor Solution (2nd-order without modification			3-factor Solution (2nd-order with modification)
no. item	item	1	2	3	1	2	3	1	2	3
1.1	… I am critical on whether I did the right thing	.771*	-	-	.771*	-	-	.771*	-	-
1.2	… I wonder if I can improve my work	.892*	-	-	.892*	-	-	.893*	-	-
1.3	… I watch how colleagues do things in order to learn from them	.544*	-	-	.544*	-	-	.544*	-	-
2.2	… I have a certain tolerance for uncertainties and ambiguities in offered solutions and explanations	-	.594*	-	-	.594*	-	-	.597*	-
2.3	… I am willing to accept some uncertainty provided that finally there is insight into proven solutions and reasonable explanations	-	.843*	-	-	.843*	-	-	.708*	-
2.4	… I can deal with situations wherein solutions and explanations are not yet available	-	.704*	-	-	.704*	-	-	.842*	-
3.1	… I try, when it comes to sort out things, to pose increasingly better and more targeted questions	-	-	.785*	-	-	.785*	-	-	.783*
3.3	… I try to avoid prejudices with regard to solutions and explanations	-	-	.667*	-	-	.667*	-	-	.671*
3.4	… I try, by means of a systematic approach to investigations, to find evidence for solutions and explanations	-	-	.674*	-	-	.674*	-	-	.672*
	Percentage of explained variance R^2 **	-	-	-	.347*	.337*	.946*	.337*	.366*	.973*

* significant at 1% level.

** A 2nd-order CFA allowed to calculation of the percentage of explained variance R² for each latent variable (i.e. factor).

Note:

• Judgements were made on 7-point Likert scales (1 = not applicable at all on me, …, 7 = totally applicable on me).

Table 10. Goodness of fit EFA’s for the three factor solutions (maximum likehood) (n = 1195).

	factor solutions
test	3-factor Solution (1st-order without Modification)	3-factor Solution (2nd-order without modification	3-factor Solution (2nd-order with modification)	criteria
Chi-Square Test of Model Fit
value	150.035	150.035	76.679
degrees of freedom	24	24	22
p-value	.000	.000	.000
RMSEA
estimate	.066	.066	.046	should be <.06 for good/close fit
90% C.I.	.056, .077	.056, .077	.035, .057	C.I should encompass values <.06
probability RMSEA ≤ .05	.004	.004	.725
CFI/TLI
CFI	.964	.964	.984	should be > .95 for good fit
TLI	.946	.946	.975	should be > .95 for good fit
Standardized Root Mean Square Residual
value	037	.037	.028	should be <.08 for good fit

Note: Grey italic numbers represent values that do not meet the criteria for good/close fit.

The factor loadings in Table 9 and the fit indices in Table 10 show that there is no difference between the 1st-order and 2nd-order 3-factor solution when no modification is applied. Fit indices were better when the modifications suggested by Mplus were applied, leading to our final model.

Conclusions and discussion

Teachers inquiring their own practice is important for effective teaching (Muijs et al. 2014). More and more data become available in schools requiring teachers to act accordingly. Even though, the school context is an important factor in whether teachers actually do research making use of the data, teachers themselves need to have an inquiry habit of mind (Earl and Katz 2006). Indeed, classroom practitioners who inquire into their own practice need to employ an inquiry habit of mind as ‘an on-going process of using evidence to make decisions’ (Earl and Timperley 2008, 4). Experienced teachers and teacher educators can develop this inquiry habit of mind by means of professional development programs, however student teachers can develop it during their teacher education.

Following Earl and Katz (2002, 2006), we defined an inquiry habit of mind as a way of thinking that is a data-driven dynamic iterative system to organise ideas, search for information, and move closer to understanding some phenomenon. Earl and Katz (2002, 2006) first described three theoretical dimensions that constitute an inquiry habit of mind: 1) ‘value deep understanding,’ 2) ‘reserve judgment and tolerate ambiguity,’ and 3) ‘taking a range of perspectives and posing increasingly focused questions.’

However, up until now, no instrument was available to determine whether teachers have such an inquiry habit of mind. Therefore, in the present study, we developed an instrument to measure teachers’ inquiry habit of mind: the Teachers’ Inquiry Habit of Mind Scale (T-IHMS). This instrument can be used during professional development and can be part of the curriculum during teacher education. The raw T-IHMS consisted of thirteen different items, (respectively 5, 4, and 4 items per theoretical dimension). Whether these items indeed represented the three theoretical dimensions as proposed was examined by means of exploratory factor analyses (EFAs) and confirmative factor analyses (CFAs) in three samples of teachers.

When performing EFAs on the first sample, item 2.1 was removed a-priori because all preliminary analyses pointed this item to be problematic. The EFAs showed that a 2- as well as a 3-factor solution would be possible. Both solutions had advantages and disadvantages. The 2-factor model did not fit the theoretical dimensions of the inquiry habit of mind and the 3-factor solution was problematic because of the low loadings of items on the three target factors. On the other hand, the 3-factor solution had a better a fit on the data as expressed by the fit indices. Therefore, it was decided to perform a second series of EFAs on the second sample but where the items 1.4 and 3.2 were removed because of their low loadings (< .04). In the second sample, the T-IHMS consisted of 10 items (13 minus the problematic item 2.1 and minus the two items, item 1.4 and 3.2). The T-IHMS was analysed once more on a 2- or 3-factor solution with EFA. This time, the EFAs showed that the 3-factor solution had the best fit, although the loadings of some of the items 1.5 and 2.4 was problematic. It was decided to remove item 1.5 but to remain 2.4 as this item was not problematic when performing EFAs on the first sample, whereas item 1.5 was. The remaining nine items of the T-IHMS were further tested by performing a series of CFAs. This final analysis showed good fit indices and more than sufficient factor loadings of the items on their target factors. The three factors form the empirical dimensions of the T-IHMS which corresponded with the theoretical derived dimensions of the inquiry habit of mind.

However, the removal of four items (item 1.4, 1.5, 2.1, and 3.2) from the raw T-IHMS means that there is a potential danger of underrepresenting the respective dimensions. For example, in the first dimension (‘value deep understanding’), item 1.4 ‘… I ask others what they think of my work’ and item 1.5 ‘… I try to collect information so that I can evaluate my work’ represent important areas of this dimension but were removed because of statistical reasons. The two items refer to seeking feedback from the environment (opinion of others and data to make more objective judgments), a feature that is necessary to interpret one’s own behavior in relation to others in order to learn and to make informative decisions. Feedback seeking behavior has important consequences for the adaptation, learning and performance of individuals (Crommelinck and Anseel 2013) and therefore, our future research will focus on creating alternative items that carry the same semantic meaning but will be part of the first dimension when performing statistical analyses. In the same manner we seek for alternative items for item 2.1 ‘… I refuse to accept unwarranted assertions and explanations irrespective of how plausible they might be’ and item 3.2 ‘… I try to view things from other perspectives’ because both items represent of what an inquiry habit of mind in essence is; item 2.1 is typical for the part of ‘reserve judgement’ and item 3.2 is typical for ‘taking a range of perspectives’. Thus, our future research should focus on how to compensate for the removed items by proposing alternative items or by proposing additional dimensions if necessary.

Nevertheless, we may conclude that the present final T-IHMS has good psychometric properties making it useful for research that investigates, amongst others, teachers’ and teacher educators’ research attitude, intention to do research, innovative behavior, and school development. In addition, T-IHMS is useful as an evaluation tool for the development of an inquiry habit of mind in both student teachers and teacher educators (in teacher education) as well as in experienced teachers (participating in professional development).

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Notes on contributors

Karel Kreijns

Karel Kreijns is professor of educational sciences at the Welten Institute of the Open University of the Netherlands. His primary research interests are computer-supported collaborative learning (CSCL), motivation, open educational resources (OER), open education (e.g., MOOCs) and teachers’ adoption of ICT. Furthermore, he is interested in school leadership and inquiry habit of mind.

Marjan Vermeulen

Marjan Vermeulen is professor of educational sciences with a focus on teacher professionalisation and director of the master Educational science at the Open University in the Netherlands. She is interested in individual and collective learning within the workplace, innovative behaviour of teachers, teams and schools, HRM/HRD, school leadership and organisations.

Arnoud Evers

Arnoud Evers is assistant professor at the Welten Institute, Research Centre for Learning, Teaching and Technology at the Open University of the Netherlands. His research areas are: Learning at work; innovative behavior; and (distributed) leadership. He published, among others, in Review of Educational Research and Studies in Continuing Education.

Celeste Meijs

Celeste Meijs is assistant professor educational sciences at the Welten Institute of the Open University in the Netherlands. Her focus was on professional development of teachers by means of networked learning and inquiry habit of mind and she developed instruments for teachers that can be used during their professional development.

Notes

1. One item, item 21 ‘I refuse to accept unwarranted assertions and explanations irrespective of how plausible they might be’, appeared in all our pre-analysis to be troublesome. For that reason, we excluded this item in all our analyses reported here. As a result, the raw T-IHMS did not include this troublesome item.