Download as pdf or txt
Download as pdf or txt
You are on page 1of 15

Copyright © The British Psychological Society

Reproduction in any form (including the internet) is prohibited without prior permission from the Society

405

The
British
Psychological
British Journal of Educational Psychology (2006), 76, 405–419
q 2006 The British Psychological Society
Society

www.bpsjournals.co.uk

Confirmatory factor analysis of the Approaches


to Teaching Inventory

Michael Prosser1 and Keith Trigwell2*


1
University of Sydney, Australia
2
University of Oxford, UK

Background. The Approaches to Teaching Inventory (ATI) was developed to


explore the relationship between students’ approaches to learning, and teachers’
approaches to teaching, in the physical sciences in higher education. It is increasingly
being used in other contexts.
Aim. To analyse the factor structure of the ATI.
Method. Confirmatory factor analysis.
Sample. Inventory responses from over 1,000 university teachers, from 4 different
countries and a wide range of disciplines, experience, and teaching level (year of
students’ study) were used as the sample.
Results. Cronbach’s alpha values of the 2-scale inventory were acceptable for scale
reliability, but not for the 4-scale structure. A good fit to the intended 2-factor
structure was obtained by confirmatory factor analysis. Both Information
transmission/teacher-focused and conceptual change/student-focused scales had well
identified intention and strategy subscales.
Conclusion. The ATI has been developed as a relational indicator of 2 dimensions
of approaches to teaching. It is not intended for use in gathering a full, rich self-report of
teaching, or in non-relational contexts. However, it has now been used in a number
of different contexts, including different subject areas and cultures. There is evidence in
this paper that in those contexts its factor structure is, in the main, consistent with the
original design. This analysis also suggests ways that the ATI could be improved.

The Approaches to Teaching Inventory (ATI) was originally designed as part of a study
examining the relationship between students’ approaches to learning, and teachers’
approaches to teaching, in the physical sciences in higher education (Trigwell, Prosser, &
Waterhouse, 1999). Since then, it has been used with some success in a number of studies
in disciplines other than the physical sciences (Gibbs & Coffey, 2004; Trigwell, 2002).

* Correspondence should be addressed to Dr Keith Trigwell, Oxford Learning Institute, University of Oxford, Littlegate House,
St Ebbe’s Street, Oxford OX1 1PT, UK (e-mail: keith.trigwell@learning.ox.ac.uk).
Michael Prosser is now at the Higher Education Academy, UK.

DOI:10.1348/000709905X43571
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society

406 Michael Prosser and Keith Trigwell

There has been, however, no published analysis of the factor structure of the instrument
outside its original administration. In this paper we describe the background to the
development of the ATI, what it was designed to measure, and an analysis of the factor
structure based upon data provided to us by others who have used the ATI. We also
indicate where it should not be used, how it is being improved, and address some recent
criticisms.

Background to the ATI


A very substantial amount of empirical research and theory development in the field of
student learning in higher education, from a student learning perspective, has been
conducted in the last two decades (Marton & Booth, 1997; Marton, Hounsell, & Entwistle,
1997; Prosser & Trigwell, 1999). That research has identified a fundamental qualitative
distinction between a surface and a deep approach to study. Students’ approaches to
study are defined as being context dependent and related to the ways students’ perceive
the teaching and learning context within which they are engaged, rather than
representing stable orientations within cognitive structure (Ramsden, 2003).
The contextual dependency of these approaches has been confirmed empirically
(Eley, 1993; Laurillard, 1997) but it has been understood in two different ways: one from a
systems, or constructivist, perspective (Biggs, 1993; Biggs, Kember, & Leung, 2001), and
the other from a relational perspective (Marton & Booth, 1997). In both cases the
approach to learning is conceived of in terms of the relationship between the student and
the teaching and learning context.
The outcomes of much of the research into approaches to study can by modelled in
terms of the top half of the diagram shown in Fig. 1. The model, a form of the 3P
(presage, process, product) model developed originally by Dunkin and Biddle (1974),
and more recently by Biggs (1993), shows the relationship between students’ prior
characteristics, the designed teaching and learning context, students’ perceptions of
that context, their approaches to study in that context, and their outcomes of learning in
that context (Prosser & Trigwell, 1999).

Figure 1. Model of teaching and student learning.


Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society

Approaches to Teaching Inventory 407

Towards the end of the 1980s, researchers in teaching and learning in higher
education began to investigate university teachers’ experiences of teaching using the
methods developed in the student learning research (see Kember, 1997; Samuelowicz &
Bain, 2001 for recent reviews). One such set of studies was conducted by the authors
with the purpose of addressing the question of how variation in university teachers’
approach to teaching relates to variation in their students’ approach to learning.
That research was conducted in several stages, and included:

. A phenomenographic study of university teachers’ approaches to teaching in the


physical sciences, aimed at identifying the qualitatively different key aspects of the
variation in approaches (not rich descriptions of approaches, but decontextualized
key aspects of the variation; Trigwell, Prosser, & Taylor, 1994).
. The development of an ATI, based upon the key aspects of the variation, and to be
used in conjunction with Biggs’ Study Process Questionnaire (SPQ; Biggs, 1987) to
study the relationship between approaches to teaching and approaches to learning
(Prosser & Trigwell, 1992).
. A large scale empirical study of teachers’ approaches to teaching and students
approaches to learning in large first-year physical science classes (Trigwell et al., 1999).

Based upon the research into student learning, and using the 3P model which
summarized that research, we hypothesized that a similar model for teaching could be
developed and used to help explore the relationship between approaches to teaching
and approaches to learning (as shown in the bottom half of Fig. 1). We hypothesized that
approaches to teaching would be context dependent, and that an instrument could be
developed with a structure similar to the SPQ to provide indicators of key aspects of
variation in approaches to teaching.

What does the ATI measure?


The phenomenographic study of university physical science teachers’ approaches to
teaching resulted in the constitution of a phenomenographic outcome space designed
to describe the key aspects of the variation (Trigwell et al., 1994). Table 1 reproduces
that outcome space.

Table 1. Key aspects of the variation in University physical science teachers’ approaches to teaching

Strategy (Act)

Intention (Indirect object) Teacher-focused Student/teacher interaction Student-focused

Information transmission Approach A


Concept acquisition Approach B Approach C
Conceptual development Approach D
Conceptual change Approach E

The outcome space was conceived of in hierarchically inclusive terms, with some
teachers describing Approach A only, with no reference to other approaches, while other
teachers describing approaches higher up the hierarchy mentioned aspects lower down
the hierarchy. From this analysis, we judged that a key qualitative variation in approach
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society

408 Michael Prosser and Keith Trigwell

to teaching was between an information transmission/teacher-focused approach to


teaching and a conceptual change/student-focused approach.
This study has since been replicated by the authors and our colleagues, with some
amendments to the outcome space, but maintaining the key variation between
a conceptual development and change/student-focused perspective and an information
transmission/teacher-focused perspective (Martin et al., 2000). As well, this fundamental
distinction has been confirmed in two substantial studies by Samuelowicz and Bain (1992,
2001). They identify a teacher-centred and a student-centred orientation, aimed at
transmitting and imparting information on the one hand and changing students’
conceptions on the other. They base their distinction within an Ausubelian framework,
arguing that the teacher-centred orientation focused on students’ assimilation of
knowledge, and the student-centred orientation on student accommodation of
knowledge. Thus from both a phenomenographic perspective and a cognitivist
perspective, this fundamental distinction has been identified. Kember’s (1997) review of
the literature on what he called academics’ conceptions of teaching came to a similar
conclusion about the fundamental distinction between teacher-centred/content-
oriented conceptions, and student-centred/learning-oriented conceptions.
The ATI was designed to provide indicators of this fundamental qualitative variation,
with the structure of the resulting inventory being similar to the SPQ developed by Biggs
(1993) to look at approaches to student learning. This design approach is in contrast to
the development of an inventory aimed at capturing a fuller empirical variation in ways
of teaching. A design of that type would have parallels with Entwistle’s ASI or ASSIST
inventories of student learning (Entwistle, Tait, & McCune, 2000; Tait, Entwistle, &
McCune, 1998), rather than Biggs’ SPQ.

Development of the ATI


In developing the ATI we had several key issues in mind. Firstly, it had to measure the
key variation between an information transmission/teacher-focused view of teaching
and a conceptual change/student-focused view of teaching. Secondly, the items had to
be phrased such that teachers could see in them their own experience of teaching.
Thirdly, it had to have sufficiently few items to encourage busy university teachers to
complete it, yet enough items to capture the key variation. A full description of this
development and the resulting ATI have been published elsewhere (Trigwell & Prosser,
2004). A summary follows.
The categories of description representing a key aspect of the qualitative variation in
the approaches to teaching formed the basis for the development of the inventory.
For the intention and strategy components of each of these categories, we sought a
range of items based upon the component descriptions and on the transcripts. The items
themselves were written so as to be consistent with our understanding of the meaning
of the components, recognizable by teachers as things other teachers might have said,
and phrased to evoke a range of responses to create variance in the responses. The items
were not meant to represent the empirical variation in the transcripts, rather to focus
more on the development of scales to represent the key aspects of the qualitative
variation described in the categories of description. This resulted in an initial pool of
104 statements. From this pool, 74 items were selected as representing the four
intentions and three strategies in the approaches to teaching outcome space, now
constituted as seven subscales. This pool of 74 items was further reduced to a pool of 49
items in five subscales by deleting the concept acquisition and conceptual development
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society

Approaches to Teaching Inventory 409

subscales. They were deleted after analysis revealed that the items in the conceptual
development subscale could not be differentiated from those in the conceptual change
subscale, and the items in the concept acquisition subscale could not be differentiated
from those in the information transmission subscale. This 49-item version was then
taken back to 11 of the staff, representing all five approaches to teaching, who were
originally interviewed. After comparing their responses to the items with our analyses of
their transcript, a further 10 items were eliminated, resulting in a 39-item version.
This version was tested using a sample of 58 university physics and chemistry
teachers, including 11 of the original interviewees (Trigwell & Prosser, 1996b).
Following a principal components analysis with varimax rotation, and a test of alpha
reliability, it was reduced to a 22-item version, in five subscales (while it might be argued
that a maximum likelihood analysis with oblique rotation is preferable, Preacher and
MaCallum (2003) note that methods similar to ours were used and published at the
time). The principal components analysis of the five subscales produced results
consistent with the theoretical model underlying the development of the inventory, and
with the congruence of the relationship between intention and strategy found in the
research from which it derived. That analysis also showed that the student/teacher
interaction subscale loaded heavily on the Student-focused subscale, and so we
eliminated the former subscale.
This 19-item version was then administered to a new sample of 46 university physics
and chemistry teachers (Prosser & Trigwell, 1997). Further principal components
analyses and test of alpha reliability finally resulted in a 16-item version with four items
per subscale and eight items per scale (Prosser & Trigwell, 1999; Trigwell et al., 1999).
At that stage, the inventory was conceived of as a two scale questionnaire; an
information transmission/teacher-focused scale and a conceptual change/student-
focused scale. This study also included a crucial test of the validity of the questionnaire.
The theory underlying its development suggested that an information transmission/
teacher-focused approach to teaching should be related to a surface approach to
learning; while a conceptual change/student-focused approach should be related to a
deep approach to learning. That study, using both principal components analysis and
cluster analysis, clearly showed that approaches to teaching as measured by the ATI from
self-reports of staff were associated with approaches to learning as measured by the SPQ
through the self-reports of students (Trigwell et al., 1999).
In late 1999, the wording of some of these inventory items was modified to
accommodate more flexible learning contexts than those found in the (science)
environments from which the items evolved. This more generalized version of the
inventory is included as an Appendix. Items 5, 8, 15, and 16 come from the intention
subscale, and with items 3, 6, 9, 14 from the strategy subscale, they form the conceptual
change/student-focused (CCSF) approach scale. Items 2, 4, 11, 13, and 1, 7, 10, and 12
are derived from the intention and strategy subscales, respectively, and form the
Information transmission/teacher-focused (ITTF) approach scale. Like the SPQ
(Biggs et al., 2001) responses to all items are on a 5-point scale from only rarely true
to almost always true, and all items are scored positively.

Testing the ATI


The final version of the questionnaire was tested on a new sample of 1,023 university
teachers from a number of universities in the UK, the USA, Scandinavia, and Hong Kong,
and from a range of academic disciplines. Data have been provided to the authors of this
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society

410 Michael Prosser and Keith Trigwell

paper by those administering the questionnaire. Nearly half of this sample was
engineering academics, mainly from Sweden, who were teaching across the range of
years. The rest are teachers from all key disciplinary areas and study years. The present
analysis does not include any of the data used to develop the original version, nor does it
include data collected by either author in subsequent administrations.
The structure of the analysis reported here is based upon the recent analysis of the
revised SPQ reported by Biggs et al. (2001). Confirmatory factor analysis using AMOS 5
was used. In the first part, the unidimensionality and reliability of the inventory’s scales
and subscales were examined. In the second part, the latent factor structure of the
scales based upon individual item and subscale analyses was examined.
The goodness of fit of the confirmatory factor structure was assessed by the
following fit indices: Comparative Fit Index (CFI), Tucker-Lewis Index (TLI), the RMSEA
and the SRMR. CFI and TLI values greater than .90 and RMSEA and SRMR values less than
.08 are used as indicators of a relatively good fit of the data to the hypothesized models.

Unidimensionality and reliability


Tables 2 and 3 show the results of separately testing each of the subscales and scales.
The results show relatively good fits of the separate factor structures to the observed data
for information transmission, teacher-focused, and conceptual change subscales. The
student-focused subscale showed a relatively good fit with one constraint imposed upon
the model: an error covariance between items 3 and 6 (Table 2). The Information
Transmission/teacher-focused and conceptual change/student-focused scales also showed
relatively good fit of the data to the model, both of which were improved by imposing one
constraint (an error covariance) in the former, and two constraints (2 error covariances) in
the latter. Nevertheless, the fits were relatively good without the constraints ( Table 3).
Having shown that the unidimensionality was supported by the CFA, the alpha reliabilities
were then examined. They showed that the reliabilities for the 4-item subscales ranged from
.61 to .40, while for the 8-item scales the reliabilities were .74–.66. We are aware of the
debate in the literature about the level of alpha values that are to be considered acceptable
(see Kember, Biggs, & Leung, 2004), including the effects of multidimensionality, but we
consider these scale reliabilities to be on the low side of acceptable, and any subsequent
revisions of the inventory would need to address this issue.
The multidimensionality, or hypothesized subscale structure of each of the two
scales was also tested. The results are also shown in Table 3. They show relatively

Table 2. Unidimensionality and reliability of ATI subscales

Scale/subscale Model CFI TLI RMSEA SRMR a

Information transmission One factor no error 1.00 1.00 .000 .005 .54
(4 items) covariance
Teacher-focused One factor no error .961 .884 .048 .022 .40
(4 items) covariance
Conceptual change One factor no error .989 .968 .044 .018 .59
(4 items) covariance
Student-focused One factor no error .899 .696 .154 .61
(4 items) covariance
One factor 1 error 1.00 .998 .013 .008
covariance
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society

Approaches to Teaching Inventory 411

Table 3. Dimensionality and reliability of ATI scales

Scale/subscale Model CFI TLI RMSEA SRMR a

Conceptual change/ One factor model .944 .922 .060 .037 .74
student-focused (8 items) no error covariance
One factor 1 error .969 .955 .046 .029
covariance
Two factor no error .946 .921 .060 .037
covariance
Two factor 1 error .970 .953 .047 .029
covariance
Information transmission/ One factor model .939 .914 .049 .034 .66
teacher-focused (8 items) no error covariance
One factor 1 error .970 .956 .035 .026
covariance
One factor 2 error .991 .987 .019 .021
covariances
Two factor no error Not
covariances admissible
Two factor 2 error .991 .984 .021 .021
covariances

good fits of the subscale structure of the conceptual change/student-focused scale to


the model. This was improved by adding one constraint: an error covariance. The
subscale structure of the information transmission/teacher-focused scale was initially
inadmissible, but with the addition of two error covariances the data showed a
relatively good fit to the model. In each of the two scales, there is a very high
correlation between the intention subscale and the strategy subscale. These
correlations (.96 and .97) suggest the ATI may be better conceptualized as a two-
factor rather than a four-factor instrument. This issue will be pursued in the analysis
in the next section.

Hypothesized factor models


The hypothesized factor structure of the ATI has been examined at the individual item
level and at the subscale level. The results of these analyses are shown in Table 4.
The first thing to note is that at the individual item level, a single factor structure does
not fit the data (CFI ¼ .654, TLI ¼ .601, RMSEA ¼ .088, SRMR ¼ .085).
Subsequently, two models were tested at the individual item level. The first tested
the four subscale structure of the inventory, and the second the two scale structure.
The unconstrained four-factor model was not admissible. With the addition of four
constraints (4 error covariances), the model showed reasonable fit to the data
(CFI ¼ .934, TLI ¼ .915, RMSEA ¼ .041, SRMR ¼ .043). All path coefficients were
statistically significant at the .001 level. Again, very high correlations between the
intention and strategy subscales in the two scales (.97 and .92) were found, suggesting
that the two scale version may be preferable.
The unconstrained two-factor model showed a relatively poor fit to the data
(CFI ¼ .865, TLI ¼ .843, RMSEA ¼ .055, SRMR ¼ .056). With the addition of four
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society

412 Michael Prosser and Keith Trigwell

Table 4. Latent factor structure of ATI

Scale/subscale Model CFI TLI RMSEA SRMR

Full ATI at item One factor no error .654 .601 .088 .085
level (16 items) covariances
Four factor no correlations Not
admissible
Four factor model 4 error .934 .915 .041 .043
covariances
Two factor model no error .865 .843 .055 .056
covariances
Two factor model 3 error .931 .916 .040 .043
covariances and 1 cross link
Full ATI at subscale One factor model no error .646 2.061 .367 .138
level (4 subscales) covariances
Two factor model no error .998 .989 .037 .008
covariances

constraints (three error covariances and one cross link), the model showed relatively
good fit to the data (CFI ¼ .931, TLI ¼ .916, RMSEA ¼ .040, SRMR ¼ .043). This model is
shown in Fig. 2. Again, each of the standardized path coefficients is statistically significant
at the .001 level.
As a way of testing that the constraints in the two-factor model shown in Fig. 2 are
not sample specific, the sample was randomly divided into two subsamples and a
multiple group analysis was conducted using the two subsamples (Byrne, 2001).
The model was first tested on each subsample to ensure that it was properly identified
with no minimization problems. The model was tested for the equivalence of covariates,
regression weights, and variances. The model showed good fit for the equality of these
parameters across subsamples (CFI ¼ .922, TLI ¼ .905, RMSEA ¼ .027, SRMR ¼ .053).
This provides some support that the constraints are not sample specific.
Finally, at the subscale level, the one-factor four subscale model showed relatively
poor fit to the data (CFI ¼ .646, TLI ¼ 2 .061, RMSEA ¼ .367, SRMR ¼ .138). The two-
factor, four subscale model showed relatively good fit to the data (CFI ¼ .998,
TLI ¼ .989, RMSEA ¼ .037, SRMR ¼ .008). This model is shown in Fig. 3, with each of
the standardized path coefficient statistically significant at the .001 level.

Discussion and conclusions


The discussion is structured around the four areas of significance that arise from this
analysis. They are: a summary of the analysis, a discussion of the constraints, further
development of the inventory, and uses and misuses of the ATI.

Summary of the analysis


The analysis has shown that the four subscales of the inventory show good
unidimensionality, with low to moderate alpha reliabilities. The two scales show
good unidimensionality and acceptable alpha reliabilities. The correlations between the
intention and strategy subscales within each scale are very high and suggest that the two
scale solution is more justifiable.
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society

Approaches to Teaching Inventory 413

The analysis at the individual item level also showed that the four-factor subscale and
the two-factor scale models showed reasonable fit to the data. Again, in the four-factor
subscale model, correlations between the intention and strategy subscales suggest that
the two-factor version is more justifiable.
Finally, the analysis at the subscale level showed that the four-factor subscale model
showed good fit to the data. This multidimensionality in the instrument is similar to that
described by Biggs et al. (2001) and Kember et al. (2004) in the SPQ and LPQ,
respectively. The goal, motive, or intention teachers have for teaching is a crucial
component of their approach, just as students’ intrinsic interest, commitment to work,
fear of failure and a concentration on qualification are crucial components of their
approaches to learning.

Figure 2. Latent structure of hypothesized two-scale model of ATI.


Note. Variables in circles are latent constructs and variables in squares are observed variables.
ccsf ¼ conceptual change/student-focused, ittf ¼ information transmission/teacher-focused.
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society

414 Michael Prosser and Keith Trigwell

Overall, the analysis confirmed the hypothesized structure of the inventory, but with
low to moderate reliabilities at the subscale level and acceptable reliabilities at the scale
level. The analyses suggest that the two-scale version, rather than the four subscale
version of the questionnaire is to be preferred, and that approaches can be described in
terms of two main factors which have a hierarchical substructure.
Correlations between the two scales are low and negative, as is generally found
between deep and surface approaches to learning. While the intention component of
the original phenomenographic categories, from which these two scales derive, form an
inclusive hierarchy (Trigwell & Prosser, 1996a), the items taken from those categories to
develop the scales are based on the focus of that category. For example, presenting (as in
presenting a lot of facts) may have been mentioned in both phenomenographic
categories, but a focus on transmission is part of Approach A only (Table 1) and has been
included only in the ITTF scale.

Model constraints
In order to achieve reasonable fits of the data to the models, some constraints need to be
imposed. Here we focus on the analysis of the two-scale version of the questionnaire.
The constraints included three error covariances and one cross link between scales.
The error covariances were between items 3 and 6, 2 and 10, and 9 and 12
(see Appendix for item wording). The cross link was between the conceptual
change/student-focused variable, and Item 7 from the Information transmission/teacher-
focused scale. A multi-group analysis of the data divided randomly in half provided
empirical support to these constraints not being sample specific.
With respect to the cross link, Item 7 refers to ‘information available from a good
textbook’. This item reflects the physical science origins of the questionnaire. Academics
from the humanities and social sciences may be equally likely to disagree with this item

Figure 3. Latent structure of hypothesized two-factor four-subscale model.


Note. cc ¼ conceptual change, sf ¼ student-focused, it ¼ information transmission, tf ¼ teacher-
focused, ccsf ¼ conceptual change/student-focused, ittf ¼ information transmission/teacher-focused.
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society

Approaches to Teaching Inventory 415

whether they are adopting more conceptual change/student-focused or information


transmission/teacher-focused approaches. That is, the idea that ‘information’ may be
available from one textbook is not the way humanities and social science academics
conceive of their reading lists.
Turning to the three error covariances, Byrne indicates that error covariances may
reflect ‘perceived redundancy in item content’ (Byrne, 2001, p. 134). Thus, one way of
substantively explaining error covariances in models is if it can be argued that there may
be some perceived redundancy in the items. Items 2 and 10 both refer to describing and
structuring, not teaching, the subject around ‘formal assessment items’. It may be that
a response to Item 2 would condition a response to Item 10. A positive response to the
first may lead to a more positive response to the second due to a perceived overlap of
the items. Items 9 and 12 take a directly opposite perspective on teaching. Item 9 ‘uses
difficult or undefined examples to encourage debate’ and item 12 ‘only provides the
student with information they will need to pass the exam’. It is hard to imagine that the
response to Item 9 does not condition in some way a response to Item 12. If your
intention is only to provide information, then you may well be more likely to disagree
with an item using difficult and undefined examples, than if the first item was not there.
Again there is an overlap in the meanings of the items. Finally, Item 3 refers to the
teacher developing ‘a conversation with students’, and Item 6 refers to ‘students can
discuss among themselves’. These two items are in the same subscale, and it may be that
the item on developing a conversation with students triggers a similar response to
the item on encouraging discussions between students.
Given that there was a good fit between the re-specified models and the data for the
two random subsamples, that the cross link is explicable in terms of disciplinary
differences, and that there may be some redundancy in the items with substantial error
covariances, we believe that this constitutes a reasonable argument for the constraints
not being sample specific.

Further development of the ATI


The ATI was made publicly available for use in 1999 (Prosser & Trigwell, 1999) as a two-
scale inventory containing 16 items (eight items in each of a conceptual change/student-
focused and information transmission/teacher-focused approach to teaching scale).
Since this publication, 5 of the 16 items have been revised slightly to accommodate
more flexible learning environments, such as on-line learning. For example, Item 6 was
changed from ‘we take time out in classes so that the students can discuss, among
themselves, the difficulties that they encounter studying this subject’ to ‘I set aside some
teaching time so that the students can discuss, among themselves, the difficulties that
they encounter studying this subject.’
While the confirmatory factor analysis supports this two-scale (two-factor) version of
the ATI, this analysis has also highlighted areas that may benefit from further
development. The two-factor model shown in Fig. 2 yielded a relatively poor fit to the data
before the addition of the four constraints already mentioned. The cross linked item ‘in
this subject I concentrate on covering the information that might be available from a good
textbook’ has a positive regression weighting (0.48) on the CCSF factor and a negative
regression weighting (2 0.28) on the ITTF factor. The reference in this item to
information being available from a good book has been reported by administrators
to cause difficulty in some areas where no good books are available, and in further
development work, consideration might be given to the use of ‘key texts’ or similar
wording.
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society

416 Michael Prosser and Keith Trigwell

As noted above, the six items involved in the three error covariances constraints show
evidence that a response to one of a pair might lead to an effect on the response to the
other. We propose in our further development of the inventory, to trial a different order of
these items (as suggested by Rubio & Gillespie, 1995) and the addition of nine new items,
which will also be used as a source of replacement items to enhance scale reliabilities.
Finally, Item 1 ‘I design my teaching in this subject with the assumption that most of
the students have very little useful knowledge of the topics to be covered’ has the lowest
scale regression weighting in any ATI scale and is reported by administrators to be
a difficult starting item. This suggests that in our further development of the Inventory,
Item 1 might be removed or moved away from the start of the inventory.

Uses and misuses of the ATI


There are three aspects of the ATI that may lead to misuse. First, the ATI has been
developed from research using a relational perspective. From this perspective,
approaches to teaching (or learning) are seen as being contextual or relational, and the
approach adopted by a teacher in one context may not be the same as the approach the
same teacher would adopt in a different context. For this reason, the ATI cannot be used
to classify teachers as being teacher-focused or student-focused. Second, while it can be
used to order or rank teachers in similar environments on these two key, and relatively
independent, dimensions of teaching, scale scores for any one teacher, are not
comparable. Third, the ATI has been developed as an indicator of variation in two
dimensions of approaches to teaching. It is not intended for use in gathering a full, rich
self-report of all aspects of teaching.
In using the ATI, we specifically ask respondents to describe their teaching context
(subject, year, type of course), and we note that the responses they give to that context
may not be the same as the responses they might give to other contexts. We also suggest
that administrators modify the items to reflect that context. An example of item
modification is in reference to the level of the context. For example, in the case of a
modular course, the item ‘I structure this subject to help students to pass the formal
assessment items’ would be changed to ‘I structure this module to help students to pass
the formal assessment items’. ATI studies that gather respondents’ experience of a range
of contexts, or do not have the same (or similar) context in component aspects of the
study, may not observe relations between the target variables of the magnitudes, or even
in the same directions, as those described in studies reported by the authors of this paper.
The studies using the ATI reported above have all used this instrument in one of two
ways. One use is as a way to monitor changes in a teacher’s approach to teaching in a
subject area over time. Gibbs and Coffey (2004) have used it in this way to look at the
effects, on teaching approach, of teaching development programmes. The second way,
and what we see as being the main use of the ATI, has been as a way of collecting data for
the analysis of relationships between approaches to teaching and other elements of the
same teaching-learning environment. For example, the study for which the inventory was
initially designed looked at the relationships between a teacher’s approach to teaching
and the approaches to learning of the students in the class of that teacher (Trigwell et al.,
1999). The same teachers, in another subject, may adopt a different approach to teaching,
and their students may adopt a different approach to learning.
While the ATI was designed for a specific study (investigating the relationship
between university teachers’ approach to teaching and their students’ approach to
learning), it has now been used in a number of different contexts. There is evidence in
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society

Approaches to Teaching Inventory 417

this paper that in those studies its factor structure is, in the main, consistent with our
original design.

Acknowledgements
Data used in the analyses described here were supplied by Greg Light, Graham Gibbs, Martin
Coffey, Li-Fang Zhang, and Peter Arvidson. Paul Ginns assisted in the interpretation of the
CFA results.

References
Biggs, J. B. (1987). Student approaches to learning and studying. Hawthorne, Victoria: Australian
Council for Educational Research.
Biggs, J. B. (1993). From theory to practice: A cognitive systems approach. Higher Education
Research and Development, 12, 73–85.
Biggs, J., Kember, D., & Leung, D. Y. P. (2001). The revised two-factor Study Process Questionnaire:
R-SPQ-2F. British Journal of Educational Psychology, 71, 133–149.
Byrne, B. M. (2001). Structural equation modelling with AMOS. New Jersey: Lawrence Erlbaum
Associates.
Dunkin, M. J., & Biddle, B. J. (1974). The study of teaching. New York: Holt, Rinehart and Winston.
Eley, M. G. (1993). Differential adoption of study approaches within individual students. Higher
Education, 23, 231–254.
Entwistle, N. J., Tait, H., & McCune, V. (2000). Patterns of response to an approaches to studying
inventory across contrasting groups and contexts. European Journal of the Psychology of
Education, 15, 33–48.
Gibbs, G., & Coffey, M. (2004). The impact of training of university teachers on their skills, their
approaches to teaching and the approach to learning of their students. Active Learning, 5,
87–100.
Kember, D. (1997). A reconceptualisation of the research into university academics’ conceptions
of teaching. Learning and Instruction, 7, 255–275.
Kember, D., Biggs, J., & Leung, D. Y. P. (2004). Examining the multidimensionality of approaches to
learning through the development of a revised version of the Learning Process Questionnaire.
British Journal of Educational Psychology, 74, 261–280.
Laurillard, D. (1997). Styles and approaches in problem-solving. In F. Marton, D. Hounsell &
N. J. Entwistle (Eds.), The experience of learning: Implications for teaching and studying in
higher education (2nd ed., pp. 126–144). Edinburgh: Scottish Academic Press.
Martin, E., Prosser, M., Trigwell, K., Ramsden, P., & Benjamin, J. (2000). What university teachers
teach and how they teach it. Instructional Science, 28, 387–412.
Marton, F., & Booth, S. (1997). Learning and awareness. Mahwah, NJ: Lawrence Erlbaum
Associates.
Marton, F., Hounsell, D., & Entwistle, N. (1997). The experience of learning (2nd ed.). Edinburgh:
Scottish Academic Press.
Preacher, K. J., & MacCallum, R. C. (2003). Repairing Tom Swift’s electric factor analysis machine.
Understanding Statistics, 2, 13–43.
Prosser, M., & Trigwell, K. (1992). Development of an approaches to teaching questionnaire.
Research and Development in Higher Education, 15, 468–473.
Prosser, M., & Trigwell, K. (1997). Relations between perceptions of the teaching environment
and approaches to teaching. British Journal of Educational Psychology, 67, 25–35.
Prosser, M., & Trigwell, K. (1999). Understanding learning and teaching: The experience in
higher education. Buckingham: SRHE and Open University Press.
Ramsden, P. (2003). Learning to teach in higher education. London: Routledge Falmer.
Rubio, D. M., & Gillespie, D. F. (1995). Problems with error in structural equation models.
Structural Equation Modeling, 2(4), 367–378.
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society

418 Michael Prosser and Keith Trigwell

Samuelowicz, K., & Bain, J. D. (1992). Conceptions of teaching held by academic teachers. Higher
Education, 24, 93–111.
Samuelowicz, K., & Bain, J. D. (2001). Revisiting academics’ beliefs about teaching and learning.
Higher Education, 41, 299–325.
Tait, H., Entwistle, N. J., & McCune, V. (1998). ASSIST: A reconceptualisation of the Approaches to
Studying Inventory. In C. Rust (Ed.), Improving students as learners. Oxford: Oxford Brookes
University, The Oxford Centre for Staff and Learning Development.
Trigwell, K. (2002). Approaches to teaching design subjects: A quantitative analysis. Art, Design
and Communication in Higher Education, 1, 69–80.
Trigwell, K., & Prosser, M. (1996a). Changing approaches to teaching: A relational perspective.
Studies in Higher Education, 21, 275–284.
Trigwell, K., & Prosser, M. (1996b). Congruence between intention and strategy in science
teachers’ approach to teaching. Higher Education, 32, 77–87.
Trigwell, K., & Prosser, M. (2004). Development and use of the Approaches to Teaching Inventory.
Educational Psychology Review, 16(4), 409–424.
Trigwell, K., Prosser, M., & Taylor, P. (1994). Qualitative differences in approaches to teaching first
year university science. Higher Education, 27, 75–84.
Trigwell, K., Prosser, M., & Waterhouse, F. (1999). Relations between teachers’ approaches to
teaching and students’ approaches to learning. Higher Education, 37, 57–70.

Received 18 December 2003; revised version received 5 November 2004

Appendix
Approach to Teaching Inventory scales

Scale: Conceptual change/student-focused (CCSF) approach Item no.

Subscale: Conceptual change intention items


I feel that the assessment in this subject should be an opportunity for ATI05
students to reveal their changed conceptual understanding of the subject
I encourage students to restructure their existing knowledge in terms of the new ATI08
way of thinking about the subject that they will develop
I feel that it is better for students in this subject to generate their own notes rather ATI15
than always copy mine
I feel a lot of teaching time in this subject should be used to question students’ ideas ATI16
Subscale: Student-focused strategy items
In my interactions with students in this subject I try to develop a conversation with ATI03
them about the topics we are studying
I set aside some teaching time so that the students can discuss, among themselves, ATI06
the difficulties that they encounter studying this subject
In teaching sessions for this subject, I use difficult or undefined examples ATI09
to provoke debate
I make available opportunities for students in this subject to discuss their changing ATI14
understanding of the subject
Copyright © The British Psychological Society
Reproduction in any form (including the internet) is prohibited without prior permission from the Society

Approaches to Teaching Inventory 419

Appendix (Continued)

Scale: Information transmission/teacher-focused (ITTF) Item no.

Subscale: Information transmission intention items


I feel it is important that this subject should be completely described in terms ATI02
of specific objectives relating to what students have to know for formal
assessment items
I feel it is important to present a lot of facts to students so that they know ATI04
what they have to learn for this subject
I think an important reason for running teaching sessions in this subject is to give ATI11
students a good set of notes
I feel that I should know the answers to any questions that students may put to ATI13
me during this subject
Subscale: Teacher-focused strategy items
I design my teaching in this subject with the assumption that most of the ATI01
students have very little useful knowledge of the topics to be covered
In this subject I concentrate in covering the information that might be available ATI07
from a good textbook
I structure this subject to help students to pass the formal assessment items ATI10
When I give this subject, I only provide the students with the information they ATI12
will need to pass the formal assessments

You might also like