The SOLO Taxonomy

 

The Power of the Solo Model to Address Fundamental Measurement Issues

 

John Hattie, University of North Carolina at Greensboro

Nola Purdie, Queensland University of Technology

 

The SOLO model, developed by Biggs and Collis (1982), proposes a structure of learning outcomes, and thus provides a clear basis for a technology of testing within learning and test theory. The chapter by Boulton-Lewis clearly demonstrates the applicability of the SOLO model to teaching and learning in all spheres of higher education. The primary purpose of this chapter is to illustrate the power of the SOLO model with particular reference to measurement issues confronted by students in teacher education programs in universities. We provide practical examples, drawn from teachers' work with primary and secondary school students, of how the model can be used in the classroom to guide lesson plans, model how students learn, model how effectively teachers teach, and construct any form of test item. We believe in the potential to improve the quality of teaching and learning in schools by devising teacher education programs that provide students with a sound understanding of the theoretical basis of the SOLO model, and practical examples of how the model can be implemented in classrooms.

 

The SOLO taxonomy

Biggs and Collis (1982) developed their model from a study of learning outcomes in various school subjects and found that students learn quite diverse material in stages of ascending structural complexity that display a similar sequence across tasks. This led to

the formulation of the SOLO taxonomy (Structure of the Observed Learning Outcome). The taxonomy makes it possible, in the course of learning a subject, to identify in broad terms the stage at which a student is currently operating. In this consistent sequence, or cycle, the following stages occur:

 

Prestructural. There is preliminary preparation, but the task itself is not attacked in an appropriate way.

Unistructural. One aspect of a task is picked up or understood serially, and there is no relationship of facts or ideas.

Multistructural. Two or more aspects of a task are picked up or understood serially, but are not interrelated.

Relational. Several aspects are integrated so that the whole has a coherent structure and meaning.

Extended abstract. That coherent whole is generalised to a higher level of abstraction.

 

Biggs and Collis (1982) based their model on the notion that in any 'learning episode, both qualitative and quantitative learning outcomes are determined by a complex interaction between teaching procedures and student characteristics' (p. 15). They emphasised the roles played by: the prior knowledge the student has of the content relating to the episode; the student's motives and intentions about the learning; and the student's learning strategies. They noted that .‘power.’ factors, such as general ability, operate across the board and thus have little prescriptive value. While quantitative

aspects of evaluating learning are well understood and applied, qualitative aspects have not been researched or applied to nearly the same extent. Such qualitative learning develops in a hierarchy of levels of increasing structural complexity. Biggs and Collis (1982) claimed their model was the 'only instrument available for assessing quality retrospectively in an objective and systemic way that is also easily understandable by both teacher and student.’ (p. xi).

 

The levels are ordered in terms of various characteristics: from the concrete to the abstract; an increasing number of organising dimensions; increasing consistency; and the use of organising or relating principles. It was developed to assess the qualitative

outcomes, of learning range of school and college situations, in most subject areas, hence the title of the taxonomy: Structure of the Observed Learning Outcome.  The model is premised on four factors:

 

Capacity. Each level of the SOLO taxonomy refers to a demand on amount of working memory or attention span. For example, at the unistructural and multistructural levels, a student need only encode the given information and may use a recall strategy to provide

an answer. At the relational or extended abstract level, a student needs to think about more things at once.

 

Relationship. Each level of SOLO refers to a way in which the question and the response interrelate. A unistructural response involves generalising only in terms of one aspect and thus there is little or no relationship involved. The multistructural level involves relationship in terms of a few limited and independent aspects. At the relational level, the student needs to generalise within a given or experienced context, and at the extended abstract level, the student needs to generalise to situations not experienced.

 

Consistency and closure. These refer to two opposing needs felt by the learner. On the one hand, the student wants to come to a conclusion and thus answer the question. But on the other hand, the student wants to be consistent so that there is no contradiction

between the question and answer. Often when there is a greater need for closure, then less information is utilised, whereas a high level of need for consistency is required to utilise more information when conceiving an answer. At the unistructural level, the student often seizes on immediate recall information, but at the extended abstract level, the student leaves room for inconsistency across contexts.

 

Structure. The unistructural response takes one relevant piece of information to link the question to the answer. The multistructural response takes several. The relational response makes more use of an underlying conceptual structure and the extended abstract requires more structure so that the student can demonstrate that he or she can deduce answers beyond the original context.

 

On the basis of the sequencing of student learning according to levels of structural complexity, Collis and Biggs (1982) proposed that there are clear implications for how schools develop programs that enable students to accommodate to the culture of school learning as opposed to that of learning for everyday life. They argue that the 'school program must take into account, among other things, the sequence of four major transitions within the concrete-symbolic mode which define particular educational tasks in content areas determined within the context of the general aims' (p. 193, see also Biggs & Collis, 1989). Students should be assisted to advance in the following way: From pre-structural to unistructural. They argued that the curriculum needed to help students 'join the game' with its new rules and its different way of conceptualising reality. For example, when teaching to read, it is worthwhile to take advantage of the

children.’s iconic mode; that is, to use their interest in listening to stories and extracting meaning from pictures.

 

From unistructural to multistructural. The curriculum needs to concentrate on consolidating and automating the unistructural knowledge and skills, building a store of concrete symbolic knowledge, and encouraging students 'to do more' with their

knowledge base.

 

From multistructural to relational. The task involves more than 'getting to know more about a topic or being adept at following through a sequence of procedures; it includes understanding or integrating what is known into a coherent system wherein the parts are inter-related. This interrelationship comes about as a result of an ability to form an overviewing principle which can be derived from the information given' (p. 196).

 

From relational to extended abstract. This process involves a shift to the formal mode of operating and typically requires dedicated hard work to master abstract concepts and relationships which form the basis of an academic discipline.

 

SOLO and learning processes

An important underlying assumption of the SOLO model is the learning processes used by students when addressing content materials and learning new information. Biggs has published extensively on the development of such learning processes (Biggs, 1985, 1987, 1990). He argued that 'reproducing' or surface approaches depend on an intention that is extrinsic to the real purpose of the task, they usually require investing minimal

time and effort consistent with appearing to meet requirements, invoke relating already understood information, and rarely go beyond the surface of the content. These approaches can increase one.’s knowledge, and involve memorisation and reproducing

as well as the application of facts and procedures in different contexts. The unistructural and multistructural levels are at these surface or reproducing levels. The term 'transforming' or deep approach reflects an intention to gain understanding by relating to

the task in a way that is personally meaningful, or that links up with existing knowledge. The relational and extended abstract levels are at these transforming or deep levels, where the aim is to understand, see something in a different way, and/or change

as a person.

 

This continuum can be traced to Marton and Säljö (1976), who formulated two major levels of learning: surface and deep. A surface approach involves minimum engagement with the task and typically focuses on memorisation or applying procedures that do not

involve reflection, but aim merely to gain a passing grade. The contrasting deep approach involves an intention to understand and impose meaning. The student focuses on relations between various aspects of the content, formulates hypotheses or beliefs about the structure of the problem, and relates more to obtaining an intrinsic interest in learning and understanding. High quality learning outcomes are associated with deep approaches, whereas low quality outcomes are associated with surface ones (see Biggs, 1987; Entwistle, 1988; Harper & Kember, 1989; Marton & Säljö, 1984). There is muchevidence that teachers (both in schools and universities) can also adopt a surface or a deep approach to teaching, and this has consequential effects on what and how students learn (Boulton-Lewis, 1995; Boulton-Lewis, Wilss & Mutch, 1996; Boulton-Lewis, Dart & Brownlee, 1995).

 

Three examples are provided to illustrate more fully the four SOLO levels and the value of the underlying model. The first is found in a meta-analysis of study skills programs by Hattie, Biggs & Purdie (1996). A major issue in that study was the power of the SOLO method to classify interventions. A unistructural study skills intervention was based on one relevant feature or dimension, such as an intervention focused on a single point of change, like coaching on one algorithm, training in underlining, using a mnemonic device, or anxiety reduction. The target parameter may be an individual

characteristic or a skill or technique. The essential feature is that it alone is the focus, independently of the context, or its adaptation to or modification by content. A multistructural intervention involved a range of independent strategies or procedures, but without any integration or orchestration concerning individual differences, or

content or contextual demands. Examples would include typical study skills packages taught directively, without a meta-cognitive or conditional framework. A relational intervention occurred when all the components were integrated to suit the individual's self-assessment, were orchestrated to the demands of the particular task and context, or involved a degree of self-regulation in learning (e.g., meta-cognitive interventions emphasising self-monitoring and self-regulation, and many attribution retraining studies). An extended abstract intervention occurred when the integration achieved in the previous category was generalised to a new domain.

 

Unistructural and multistructural programs were highly effective with virtually all students when studying material requiring only low level cognitive involvement (e.g., memorisation of specific information). Multistructural approaches were most effective with younger rather than older students. Relational programs, integrating the informed

use of strategies to suit the content, and used for near transfer in context, were highly effective in all domains (performance, study skills, and affect) over all ages and ability levels, but were particularly useful with high ability students and older students.

 

A second example of the power of the SOLO taxonomy is its use to classify the behaviour of teachers in classrooms. Hattie, Clinton, Thompson and Schmidt-Davis (1997) observed elementary school teachers who had been certified as 'Highly accomplished' by the National Board for Professional Teaching Standards. This

certification was as a consequence of extended performance evaluations of a large number of teachers and the setting of high and rigorous standards, but did not involve any classroom observations. Hattie et al. visited teachers in their classrooms and, among an array of measures, they were interested in the effects of the teaching on the students. Many researchers have emphasised the importance of 'knowledge' when differentiating experts from novices (Shulman, 1987) and have observed that knowledge that is useful for experts may hold little meaning for novices (deGroot, 1965). Of more importance are the differences in the way knowledge is used in teaching situations. As Chi, Feltovich, & Glaser (1981) noted with physicists, experts were more sensitive to the deep structures of the problems they dealt with, whereas novices were sensitive to surface structures.

 

Given the well known problems of using achievement outcomes as indicators of teaching effectiveness (Haertel, 1986), Hattie et al. used the four levels of the SOLO taxonomy to code students' work as artefacts of the teacher.’s lesson. Such coding related more to the depth of understanding that the teacher could accomplish, rather

than the typical breadth of knowledge assessed by many traditional achievement tests. The following student essay (from a student in one of the classes observed for this study) would be classified as multistructural, as it only contains a series of unrelated ideas:

 

If I could be any tree I would be a Redwood. A Redwood tree can live 4000 years. They are very strong and tall. They even name a forest after me. I think a redwood is a good tree to be. I would be the state tree of my state. I would be known all the way across the state. When somebody cut me down I would fall down real hard on the ground.

 

The following student essay would be classified as at least relational, as the major contributions are the series of related ideas that demonstrate a degree of integration and higher levels of abstraction:

 

I think I would be a willow tree, because I go with the flow like a willow.’s limbs in the wind. I.’m strong to stand up to in hard times and I come out OK. Like a willow in a storm, only the hardest things can get me down. I.’m calm and easy going, part of nature, and cannot be missed.

 

Along with many other attributes, the students of teachers who received certification at the level of 'Highly accomplished' were most likely to exhibit the deep levels of SOLO (47% of the artefacts from the passing teachers versus 27% from the failing teachers were classified as relational or extended abstract). Hattie et al. concluded that expert teachers are more likely to lead students to deep rather than surface learning. These teachers will structure lessons to allow the opportunity for deep processing, set tasks that encourage the development of deep processing, and provide feedback and

challenge for students to attain deep processing (p. 54).

 

A third example is the evaluation of gifted programs by Maguire (1988). He used the SOLO taxonomy as part of an evaluation of programs for bright and gifted students in elementary and junior high school. The students in the program often pursued the

objectives of the program by working independently on projects, working together in small groups, or participating in a mentorship program. This 'diversity in learning activity may lead to uneven levels of knowledge about a particular content domain, inspite of the fact that levels of attainment of higher order objectives such as critical thinking may be uniformly high. In many situations the content provides a vehicle for instruction and may differ across students. It is not easy to find instruments that are relevant to program objectives, flexible enough to capture the creativity and divergence expected in performance from these kinds of students, yet at the same time possesses utility and validity.' (p. 10). Thus, to evaluate the program, Maguire devised two writing and three mathematics tasks, and the answers to these questions were coded into the SOLO levels. It was expected that the students in the gifted program would have a more positively skewed distribution (that there would be more in the higher levels) compared with students of similar ability not in the program.

 

Maguire argued that the SOLO approach seemed to tap a complex of deep understanding, motivation, and intuition as applied to a particular task, thus it was appropriate to assess complex achievements, deep understanding, higher order skills, and strategic flexibility (cf., Snow, 1989). He found that students operating at the higher levels of the SOLO taxonomy (i.e., relational, and extended abstract) tended to have higher scores on deep and achieving styles. Students who gave higher level responses to the SOLO writing tasks were also students who were more deeply engaged in their learning, while students who produced lower level products seemed to have more superficial approaches. When he compared the SOLO profiles from the students in the gifted program with a group of students in the regular classrooms identified as being gifted, and another group identified by the teachers as 'potentially gifted', there were no

discernible differences. As Maguire concluded, the results provide 'a picture of a program that is not yet succeeding' (p. 9). The use of the SOLO levels, however, allowed this researcher to 'put outcomes on a common base while at the same time avoiding the confinement of standardized instruments. ... (SOLO) has been a very useful tool for detecting problem areas' (p. 9).

 

These three examples illustrate the diversity of situations where SOLO can be used. They demonstrate the power of SOLO for assessing interventions, teachers or any observational studies, and for evaluating programs. An under-utilised use of SOLO is in

assessing item construction and test analysis.

 

SOLO and test item construction

A powerful advantage of the SOLO taxonomy is that it can readily be used to devise test items. There are three ways that items can be constructed according to SOLO. Either the questions are worded in an attempt to elicit particular SOLO level-type responses, or answers are scored depending on the evidence at the appropriate SOLO

level, or a combination of these methods. For all three methods, the items can be analysed using traditional psychometrics, or by the classical model approach whereby estimates of reliability, difficulty and discrimination (e.g., point-biserials) are calculated and items are thereby dropped or improved. The alternative item response models can also be used whereby invariant estimates of the difficulty, discrimination and/or guessing can be estimated and items that maximise the desired test information functions are retained.

 

Each level of the taxonomy provides a working principle for an item. (Prestructural is not considered as it involves inappropriate processing that leads to incorrect solutions.) Either a set of up to four items (a testlet), or a series of items that can be coded into one

of the four levels can be written (see Biggs, Holbrook, Ki, Lam, Li, Pong & Stimpson, 1989).

 

Unistructural. Contains one obvious piece of information coming directly from the stem. An answer is based on only one relevant aspect of the presented evidence, so that the conclusion is limited and likely to be dogmatic.

 

Multistructural. Requires using two or more discrete and separate pieces of information contained in the stem.

 

Relational. Uses two or more pieces of information each directly related to an integrated understanding of the information in the stem. Most or all of the evidence is accepted, and attempts are made to reconcile. Conflicting data may be placed into a system that accounts for the given context.

 

Extended abstract. Requires use of an abstract general principle or hypothesis that can be derived from, or suggested by, the information in the stem. There is recognition that the given example or question can lead to a more general case.


Figure 1 illustrates an item constructed according to SOLO (see notes 1).


Figure 1. An item constructed according to SOLO.

`O´ O          O       O      O      O        O         O         O          O

sun Mercury Venus Earth Mars Jupiter Saturn Uranus Neptune Pluto


1. Which is the planet furthermost from the sun? (Unistructural)

2. Which planet is warmer - Venus or Mars? (Multistructural)

3. How does the movement of the Earth relative to the sun define day and night? (Relational)

4. Given the Earth.’s position relative to the sun, in what ways does this affect the Earth.’s climates and seasons? (Extended abstract)

 

This item illustrates the power of the SOLO taxonomy to provide students and teachers with a structure and a process for developing their own questions. In the first question, only a single piece of information is required. It is close to the recall sense of 'knowledge'. In the second question, the student is required to use two separate pieces of information (the position of Venus to the sun, and the position of Mars to the sun) to work out the answer. In the third question, it is necessary that the student sees the connection between the movement of a planet to the sun with the phenomenon of night and day. Finally, in the fourth question, the student has to go beyond the information provided in the item to deduce a more general principle as to the effects of the Earth.’s position to the sun and the effects on the climates and seasons.

 

Although the questions in the example above are open-ended, there is no requirement that a particular form of question is more advantageous at any of the levels of SOLO. Nor is there a requirement that every level of the SOLO taxonomy must be present in every question. Consider the multiple choice item in Figure 2.


Figure 2. A SOLO multiple choice item.


What is the value of D in the following statement? Show all working.

(84 / 42 ) * 7 = (84 * 7) / (D x 7)?

a. 42

b. 14

c. 294

d. 6

The answers, and reasons are:

a. 42 Unistructural There is no 42 on the right side; 2(84/42) =42

b. 14 Unistructural The student has only calculated the left side

where there is all information. In a and b, the

student has only sought one piece of

information.

c. 294 Multistructural A step by step calculation, 2 * 7 = 588 / (D *

7) = 294. The student has used information

[incorrectly] on both sides of the equation.

d. 6 Relational Step by step calculation carried out correctly,

or

Extended abstract Balancing 84 * 7 = 84 * 7

42 D * 7

 

In this item, the highest scored level could be Relational (if 6 is chosen), or the work provided may allow the teacher to score the item at either the Relational or Extended Abstract levels.

 

SOLO items can also be constructed in sets. As in many of the examples (e.g., the relation between the sun and planets), there is one stem, and several items aiming at progressively higher SOLO levels. Thus, there may be context effects as the student is forced to progress through a predetermined path of increasing complexity. The items are grouped into what Wainer and Kiely (1987) called 'testlets'. That is, 'a testlet is a group of items related to a single content area that is developed as a unit and contains a fixed

number of predetermined paths that an examinee may follow' (p. 190). As a consequence, the item at each level is embedded in a testlet, and thus is context bound.

 

There are many advantages when using the testlet approach. First, if the total score of the student is then a score on each testlet (calibrated by either classical or IRT methods), then more stable information is gained (as the testlet score is based on a series of items) and there is information not only relating to the student.’s proficiency on the content matter but also relating to the depth or complexity level of the student's processing proficiency.

 

Second, testlets can improve the computer adaptive testing procedures (CAT). In CAT situations the computer program chooses an item to administer to the student depending on his or her performance on previous items. One approach may be to choose items at the multistructural level first, and then choose items at SOLO levels depending on the student's response to this multistructural item (of course, there can then be multiple

alternative items at each level relating to the same stem). Such a sequential strategy may be most advantageous when the CAT covers materials across diverse subject areas (such as in tests of scholastic aptitude). It may, for example, be advantageous for the test developer to ensure that every student is given items across certain content areas. The differing length of the CAT could then be a function of the number of items within each testlet (i.e., the level of complexity would change, but the coverage of content may be more or less fixed).

 

Third, the items within each testlet can be investigated separately to ensure that they follow the desired patterning, and thus information can be gained about items, the performance of the items within the testlet, the performance of the item within the total test, and the performance of the testlet within the total test (e.g., see Rosenbaum, 1988). This can greatly improve the available diagnostics for developing excellent tests.

 

Several procedures for the analysis of SOLO-based test items have been suggested. Wilson (1989) recommended using the one-parameter partial credit model to analyse SOLO items. Using five in biology and five in chemistry (from Romberg, Collis, Donovan, Buchanan, & Romberg, 1982; Romberg, Jurdack, Collis, & Buchanan, 1982) he reported that this model displayed excellent fit for both persons and items. The levels of the SOLO taxonomy were ordered by item difficulty. Over the total number of items, students with low estimates of ability tended to get the unistructural correct, whereas as ability increased there was a tendency to get higher SOLO level items correct (see Figure 3, adapted from Wilson, 1989). For example, on Item 2 (from the biology test), persons with low ability on the test (-2 to -1 on the total score scale, in logits (see notes2)) tended to get only the Unistructural item correct, whereas those with 'middle' ability (3 logits) tended to also get the Multistructural item correct. As you move up the overall ability level, the probability of passing the higher SOLO levels increases.

 

Figure 3. Ability and success on a SOLO item.

              -2   -1   0   1   2   3   4   5   6   7   8

Item 2         U                    M       R   E


80% mastery ___U__

levels                    _______M____

                                        _______R____

                                                       _____E___

 

A second very useful interpretation procedure was also suggested by Wilson (1989). The 80% mastery levels at the bottom of the above Figure indicate the region where the student is expected to master the questions of a given level in the taxonomy. Thus, if a student.’s ability is within a mastery level (say the score is 5) they have an 80% probability of succeeding at all Unistructural and Multistructural items, and an 80% chance of failing on the Extended Abstract items. This student would be expected to be in the Relational area of learning in biology.


 

Wilson (1989) also demonstrated how fit-statistics could be used to ascertain which items students are not performing to the expected level of their learning capabilities, and demonstrated the paucity of information that is derived when using a Guttman scalogram analysis. In a related paper, Wilson and Masters (1993) illustrated how the partial credit model could be used when there were null categories. For example, a test developer may create a SOLO testlet that does not include an item assessing extended abstract, and thus in the total test there are a series of testlets with most including items at all four levels, and some with a reduced number of levels. Thus, for these latter testlets, there is a 'logically null level'. Wilson and Adams (1993) introduced the ordered partition model, which is akin to the partial credit model, for the analyses of data (like SOLO) in which item responses are categorised and then scored in ordered levels (see also Biggs, 1990; Lam & Foong, 1996; Wilson & Iventosch, 1988).


A final application of the SOLO taxonomy in the realm of testing is its ability to classify items within a published test, regardless of whether it is constructed using the SOLO methods. For example, the questions on an achievement test (like the SAT or a teacher-made achievement test) can be classified into the four SOLO levels and this information used to either re-score the test or to ascertain the depth of learning that the teacher is aiming to assess (or maybe has aimed to teach). Further, we have participated in a school-wide evaluation plan based on such an analyses. The principal wished to

 

know the depth of learning in mathematics in his elementary school (Mort, personal communication). He first analysed the items that were being administered at each grade level to ensure that they covered the four levels of SOLO. The responses to these items

 

were scored according to their SOLO level and a profile graph of each class compiled to show the percentage of students at each level (he also investigated content coverage). The principal and the teachers not only found this most informative, they also then set

 

goals as to percentages desired at each level (and the teachers also profiled each student and set appropriate goals). The effect was a re-focusing of teaching away from exclusive content aims, towards more depth of processing aims, and a consequential leap in the percentage of students at the higher levels of SOLO.


 

SOLO compared with the major alternative: Bloom.’s taxonomy

 

Modern test theory is at a crossroads. The traditional model is rooted in classical test theory and more recently in item response theory. Both theories involve applying mathematical models to 'true scores' or to 'expected responses' from a sample of test items. These hypothetical constructs aim to represent traits or the proficiency of

 

students to answer various items. As Mislevy (1996) has stated, these theories attend to 'the problem stimulus strictly from the assessor.’s point of view, administering the same tasks to all examinees and recording outcomes in terms of behavior categories applied in the same way for all examinees. Behavior constitutes direct evidence about behavioral tendencies' (p. 391). A more defensible model is derived from cognitive psychology which makes claims such as: we interpret experience and solve problems by

 

mapping them to internal models; these internal models must be constructed; and the constructed models result in situated knowledge that is gradually extended and decontexualised to interpret other structurally similar situations (Mislevy, 1996, p. 389).

 

Thus, we need measurement models that acknowledge that different knowledge structures can lead to the same behaviour, where observed behaviour constitutes indirect evidence about cognitive structure, and we need to assess the degree or depth to which the students understand or process. The most dramatic example of a test writing technology based on the earlier 'behavioural tendency' models is Bloom.’s taxonomy, whereas an excellent example of a test writing technology based on the cognitive processing model is Biggs and Collis.’ SOLO taxonomy.


For the past four decades the development of most measures of cognition and achievement have been based on Bloom.’s taxonomy of educational objectives (Bloom, Engelhart, Furst, Hill, & Krathwohl, 1956). This model proposes six levels: knowledge, comprehension, application, analysis, synthesis and evaluation. Knowledge refers to those behaviours and test situations that emphasise th remembering, either by recognition or recall, of ideas, material, or phenomena. Comprehension involves translation, interpretation or extrapolation of knowledge. Application requires the student to know an abstraction well enough that he or she can correctly demonstrate its use when asked to do so. Analysis emphasises the breakdown of material into its constituent parts and detection of the relationships of the parts and of the way they are organised. Synthesis involves the putting together of elements and parts to form a whole. Evaluation is defined as the making of judgements about the value, for some purpose, of ideas, works, solutions, methods, materials, etc. (definitions are taken from Bloom et al., 1956).


The taxonomy was published in 1956, has sold over a million copies, has been translated into several languages, and has been cited thousands of times. The Bloom taxonomy has been extensively used in teacher education to suggest learning and teaching strategies, has formed the basis of many tests developed by teachers (at least while they were in teacher training), and has been used to evaluate many tests. It is thus remarkable that the taxonomy has been subject to so little research or evaluation. Most of the evaluations are philosophical treatises noting, among other criticisms, that there is no evidence for the invariance of these stages, or claiming that the taxonomy is not based on any known theory of learning or teaching (Calder, 1983; Furst, 1981).

 


There are many similarities between the Bloom and SOLO taxonomies. It is necessary when using both taxonomies to know the context of learning, and it is expected that the questions asked follow from some form of instruction or prior exposure to the

 

information required. There is also the premise that the concepts in the instruction are hierarchical