Appendix 7. Papers submitted to journals
Appendix 7.4 - Information seeking and mediated searching. Part 4. Cognitive styles in information seeking analysis
N.J. Ford, T.D. Wilson, A.E. Foster, D. Ellis*,
This is the fourth in a series resulting from a joint, UK/USA research project. The analysis reported here sought to test a number of hypotheses linking global/analytic cognitive styles and aspects of researchers' problem-solving and related information seeking behavior. One hundred and eleven post-doctoral researchers were assessed for Witkin's field-dependence/-independence using Riding's Cognitive Styles Analysis and for Pask's holist/serialist biases using items from Ford's Study Processes Questionnaire. These measures were correlated with the researchers' perceptions of aspects of their problem-solving and information-seeking behavior, and with those of the search intermediary who performed literature searches on their behalf. A number of statistically significant correlations were found. Field-independent researchers were more analytic and active than their field-dependent counterparts. Holists engaged more in exploratory and serendipitous behavior, and were more idiosyncratic in their communication than serialists.
This paper is the fourth in a series reporting the results of a joint research project supported in the USA by the National Science Foundation and in the UK by the British Library. The USA study was undertaken at the University of North Texas by Dr. Amanda Spink (now at The Pennsylvania State University). The UK study, headed by Professor Tom Wilson, was conducted at the University of Sheffield, UK. The theoretical rationale and research design of the project are described in the first paper (Spink, Wilson, Ford, Foster &: Ellis, 2002). The current paper present the results of a sub-project designed to explore the relationship between cognitive styles and problem-solving and its associated information seeking.
Cognitive styles are tendencies displayed by individuals consistently to adopt a particular type of information processing strategy. Many such differences have been identified (Brumby, 1982; Entwistle, 1981; Ford, 1995; Jonassen & Grabowski, 1993; Miller, 1987; Riding & Cheema, 1991; Schmeck, 1988), and a number of them have been studied empirically in terms of their effects on information seeking behavior and performance.
A major focus of research into cognitive styles has been the study of what may be described as a global/analytic dimension of difference; notably work conducted and inspired by Witkin in the USA and Pask in the UK. Witkin investigated global/analytic differences in a very wide range of human activity from basic perception to career choice (Witkin, Moore, Goodenough & Cox, 1977). Pask in the UK studied global/analytic differences in concept acquisition relating to complex academic subject matter (Pask, 1979; 1988). The dimensions of cognitive style identified by Witkin are most generally termed field-dependence and field-independence. Those identified by Pask relate to holist and serialist approaches to information processing.
In a series of experiments (Pask, 1976a,b,c; 1979;1988; Pask & Scott, 1972; 1973), Pask and his colleagues monitored the routes taken by learners through a range of complex academic topics. In these experiments, people used one of two basic approaches. "Holists" tended to adopt a global approach to learning, examining interrelationships between several topics early in the learning process, and concentrating first on building a broad conceptual overview into which detail could subsequently be fitted. "Serialists" tended to use a predominantly local learning approach, examining one thing at a time, and concentrating on separate topics and the logical sequences linking them. The overall picture emerged relatively late in the learning process. When learning material that entailed theoretical and corresponding "real world" examples and applications, the serialist worked through either the theoretical or the real world topics, only bringing them together late in the learning process when absolutely necessary to achieve understanding. The holist, on the other hand, constantly moved between theory and real world right from the start. Holists also tended to look further ahead in the hierarchy of topics making up the subject (Robertson 1977; Entwistle 1981).
The holist is cognitively complex, and likes to have several things "on the go" at the same time. In contrast to the steady "brick by brick" approach of the serialist, the holist adopts what is a comparatively high-risk, exploratory strategy, switching attention across a range of tasks before any one is securely completed and checked as a sure foundation for further progress. The holist progresses in an exploratory fashion compared to the serialist's narrow focus and step-by-step logical progression, making sure to build solid foundations for each next move.
Using the technique of "teachback", Pask and Scott found that extreme holists were distinctive in the personalised, often idiosyncratic way in which they related new information to their existing knowledge, making sense of it in ways often not easily understood by others.
The dimensions of cognitive style identified by Witkin relate to what he termed field-dependence and field-independence. Relatively field-independent individuals are more adept at structuring and analytic activity when compared with relatively field-dependent individuals. Relatively field-dependent individuals thrive more in situations where learning is structured and analysed for them. They tend to prefer a 'spectator' approach to learning rather than the hypothesis-testing approach favoured by more field-independent learners. They operate with a relatively external frame of reference, as opposed to the greater "inner directedness" of the field-independent individual. Field-dependent people tend to be more socially oriented than more field-independent individuals, and this may even be reflected in the type of academic study and employment they choose and in which they excel.
Essentially, field-independent individuals tend to experience the components of a structured field analytically, as discrete from their background, and to impose structure on a relatively unstructured field. By contrast, relatively field-dependent individuals tend to be less good at such structuring and analytic activity, and to perceive a complex stimulus globally as a gestalt. This dimension would seem to extend from perceptual through intellectual and social functioning. Witkin, Moore, Goodenough & Cox (1977) published a detailed review of the educational implications of field-dependence/independence. Riding and Cheema (1991) also include field-dependence/-independence in a comparative review of cognitive styles which also includes Pask's holist/serialist distinction.
Pask developed a series of tests of including the Spy Ring History and the Smugglers tests. These are complex, lengthy to administer, and very demanding on learners. Relatively few studies using these measures (other than those conducted by Pask and his colleagues) have been reported (see for example Coombs, Gibson & Alty, 1982). Entwistle (Entwistle, Hanley & Hounsell, 1979) developed a self-completion inventory, and a shortened version, the Short Inventory of Approaches to Studying (Entwistle, 1981), which was designed to assess, amongst a number of other constructs, Pask's comprehension, operation and versatile learning styles. Although quick and easy to administer, and benefiting from reliability data and norms derived from large-scale studies, this instrument was not designed directly to measure holist and serialist biases (even though arguably they could be inferred from comprehension and operation learning biases). Ford (1985) thus devised a measure specifically designed to assess holist and serialist biases: the Study Processes Questionnaire. Clarke (1993) investigated the reliability of this instrument, and it has been used in a variety of studies (e.g., Ford, 1985; Ford & Chen, 2001). However, its reliability has not been widely studied and no published norms are available.
A number of instruments have been developed to measure field-dependence/-independence, one of the best known of which is Witkin's Group Embedded Figures Test (GEFT). More recently, Riding's (1991) Cognitive Styles Analysis (CSA) measures what the authors refer to as a wholist/analytic dimension, noting that this is equivalent to field-dependence/-independence (Riding & Sadler-Smith, 1992). This instrument offers computerised administration and scoring, and has been designed to overcome a limitation affecting the most widely used measures of field-dependence/-independence. Tests such as Witkin's GEFT derive scores for field-independence by requiring subjects to locate simple shapes embedded in more complex geometrical patterns. However, levels of field-dependence are inferred from poor field-independence performance, that is, from poor performance on this disembedding task.
However, the Cognitive Styles Analysis differs from tests such as the GEFT in that its wholist/analytic test consists of two sub-tests. In the first, subjects are required to judge the similarity of a series of complex geometrical figures, which is a task requiring field-dependent capacity. The second sub-test requires subjects to determine whether a simple shape is contained within a more complex geometrical figure (as in the GEFT), which is a task requiring the disembedding capacity associated with field-independence. In this way, field-dependent competence is positively measured rather than being inferred from poor field-independent capability.
Global/analytic cognitive styles have been linked with various aspects of information seeking behavior.
A study by Ellis, Ford & Wood (1992) investigated hypertext navigation by 40 postgraduate students. Students were tested for Pask's styles using Entwistle's Short Inventory of Approaches to Studying. The Study Preference Questionnaire, a non standardised test devised to assess holist and serialist strategic biases was also used. A hypertext system was used in this experiment, navigation tools being provided in the form of a self-orienting, global concept map, keyword index, menus, and a back-tracking facility. The subject matter of the hypertext was the European Single Market. The students were given the task of using the system to answer a number of questions requiring (a) specific factual recall, and (b) generalisation using information from more than one location in the hypertext. All interactions were automatically logged. Holists made significantly greater use of the global map; serialists of the keyword index. No significant differences were reported for field-dependence.
A later study by Chen & Ford (1998) also investigated hypertext navigation. Twenty postgraduate students were tested using the Cognitive Styles Analysis , then learned from a hypertext system designed to give an introduction to the field of artificial intelligence. Navigation patterns were logged for analysis. It was found that field dependent individuals made significantly greater use of the main menu, their field-independent counterparts making more use of the relatively sequential Previous/Next buttons.
Ford & Ford (1992) conducted an experiment with postgraduate students to discover how they might go about learning from an "ideal" computer-based system. A system was created which preserved the characteristics of a computer-based environment, but freed itself from the constraints of current technology. Although not realising it at the time, 30 postgraduate students were in fact interacting through a computer screen with two human subject experts. The students were asked to interrogate the system, using whatever language and approach they wished, in order to learn about the document indexing system, PRECIS. At the end of the session, they were asked to write what they had learned about the system. The students displayed significant differences in database interrogation strategies, which mapped well on to Pask's holist and serialist distinction.
In a study of online searching behavior (Wood, Ford, Miller, Sobczyk & Duffin, 1996) statistically significant differences were found between global/analytic differences and aspects of information searching including awareness of broadening and narrowing search techniques; levels of satisfaction with search results; number of different terms used in the search formulation; number of new terms introduced during the search; number of relevant references retrieved; and perceived search success. 105 undergraduate students carried out on-line searches of CD-ROM databases for information on topics relating to their coursework. Databases included Inspec, Biological Abstracts, Social Sciences Index, Compendex, ABI-Inform, General Sciences Index and Modern Languages Association. The students completed the Short Inventory of Approaches to Studying and the Cognitive Styles Analysis. Search strategies were logged for analysis.
A previous study (Wood, Ford & Walsh, 1992; Ford, Wood, & Walsh, 1994) had revealed significant links between global/analytic differences and search behavior. Relatively global individuals used significantly broader search strategies than their analytic counterparts. The behavior studied in this experiment related to the searching of a CD-ROM database, containing 105,482 bibliographic records. 67 postgraduate students conducted 275 searches on Silver Platter's CD-ROM-based Library and Information Science Abstracts (LISA) database on subjects related to their coursework. Students were tested using the Short Inventory of Approaches to Studying and the GEFT. Their searching strategies were classified in terms of relative breadth and depth. A high use of the word OR to link keywords represents a relatively broad strategy: a high use of AND a relatively narrow strategy. Other measures of the breadth or narrowness of search included truncation and generic descriptors (which broaden a search), and use of date or language qualifiers (which tend to narrow a search).
The overall research design for the project, including data collection instruments and procedures, is described in detail in the first paper (Spink, Wilson, Ford, Foster & Ellis, 2002). The present paper reports the results of an analysis of a subset of these data. Since, as described below, this part of the study entailed the testing of a number of specific hypotheses, only those variables related to the hypotheses were used. [The research instruments used can be found at the Project Web site.]
Research aim and objectives
The aim of the research reported here was to discover the extent to which cognitive styles may affect an information seeker's perceptions of the nature of his or her research problem and of progress in its solution through information seeking. The objectives were to test the following hypotheses, derived from descriptions of cognitive style constructs in the research literature.
Although the study adopted a hypothesis-testing approach, it was at the same time exploratory in that each dependent concept contained in the hypotheses was not co-extensive with only one variable. For example, "uncertainty" (hypothesis 1d) consisted of a number of measures relating to a range of problem-solving stages (problem recognition, definition, resolution, etc.), taken at different times (before and after the search).
Cognitive style theory is not developed to a sufficiently fine-grained stage to render productive the specification of hypotheses co-extensive with these individual variables. Whilst it may be reasonable, based on cognitive style theory, to hypothesise that styles may affect "levels of uncertainty", it would be less reasonable to generate more specific hypotheses, for example, for each of the proposed stages of problem solving. Therefore, the study sought to discover the nature (as well as the strength) of evidence for each hypothesis, in terms of which (all, some or none) of a range of potentially contributory variables provided support.
Variables may be grouped as follows:
Changes due to interaction in terms of: level of changes in the information seeker's perception of the problem; changes in the question since the outset of the search; changes in personal knowledge of the specific problem at hand, due to the interaction and/or the feedback during the ongoing search; changes in criteria for relevance judgements due to the interaction and /or the feedback during the ongoing search.
Effectiveness of communication, in terms of: the level of effectiveness of the information seeker's explanation to the searcher; the level of understanding by the information seeker of the search procedures being used by the intermediary; understanding by the intermediary of the information seeker's problem (as perceived by the information seeker); the effects (negative or positive) of the intermediary's non-verbal communication as perceived by the information seeker How clear and focused the researcher feels his/her thinking to be, and the extent to which each of Ellis's information seeking activities was engaged in. These are: chaining; browsing; differentiating; maintaining; systematically working through; verifying.
Data were analysed using SPSS for Windows (version 9). Correlations were sought between the independent variables relating to cognitive style, and the dependent variables relating to information seeking and problem solving behavior introduced above. A level of significance of p < 0.05 was adopted for this study.
On hundred and eleven literature search topics were taken through to completion from the subject disciplines shown in Table 1. Age varied widely, from 22 to 76, with a mean value of 39. 42 (37.8%) of the participants were female, 69 (62.2%) male.
TABLE 1: Number of participants by subject discipline
A number of statistically significant correlations were found. These are presented in Tables 2 and 3 below. Since the majority of the variables did not fit the Gaussian distribution, the non-parametric Spearman test was used.
TABLE 2: Correlations with Field-dependence/-independence (Spearman).
TABLE 3: Correlations with Holist/Serialist learning style (Spearman).
The findings relating to the hypotheses concerning field-dependence/-independence are shown in Table 4.
TABLE 4: Evidence supporting the hypotheses concerning field-dependence/-independence.
The finding that field-independent individuals report clearer, more focused thinking is in line with the greater analytic competency associated with them in the research literature. The greater levels of change in perception of the problem and in own personal knowledge reported by the intermediary resulting from the interaction may also be at least circumstantial evidence of more analytic and clearer thinking on the part of the field-independent researcher when describing and explaining the problem to the intermediary.
The tendency of the more analytic field-independent researcher to perceive him/herself to be in an earlier problem-solving phase seems counter-intuitive. However, if indeed the field-independent individual is likely to break a problem down early into its constituent parts, he or she may perceive more clearly the number of sub-tasks requiring attention before the problem can be solved. Thus, it is compatible with theory that field-independent individuals should tend to perceive themselves at an earlier stage of problem solving (being more acutely aware of the complexity) than field-dependent individuals relative to the same time-frame. It may be relevant in this context to note that most work relating field-dependence/-independence to behavior has used objective measures of ability as opposed to measures of perception as used here.
Conversely, the field-dependent individual may have a more fuzzy, less differentiated view of the problem to be solved. At a given point in time, he or she is less likely to be aware of a number of discrete stages of problem solving awaiting solution. Thus, arguably it may be that the field-dependent person perceives himself/herself to be less far away from the goal - possibly therefore as at a less initial stage of problem solving.
There would seem to be some evidence to support the notion that field-independent individuals take a less passive, less reproductive approach to research than their field-dependent counterparts. They report more of Ellis's "engaged differentiating" activity, and the higher reported levels of change in perception of the problem are compatible with the more active transformational engagement with, and questioning of new information characteristic of the relatively field-independent person.
No evidence was found to support the hypotheses relating to uncertainty and serendipity.
Results relating to the holist/serialist dimension of style are shown in Table 5.
TABLE 5: Evidence supporting the hypotheses concerning holist/serialist biases.
The finding that holists reported more activity in relation to Ellis's "exploring" behavior, and greater valuing of serendipity is in line with the notion of the holist's preference for, and greater competence in, engaging in relatively speculative exploration (as opposed to the relatively more secure and sequential behavior that might be associated with Ellis's "chaining" behavior). Their greater valuing of serendipitous information encounters is also in line with theory in that holists are more likely to be open to - indeed seek out - such relatively unplanned encounters, in comparison with their serialist counterparts, who are more likely to prefer a more secure and predictable step-by-step approach. However, the finding that, according to the perception of the intermediary, holists exhibit fewer changes in their questioning does not seem to support the hypothesis, in that changes would seem particularly compatible with the notion of relatively speculative exploration.
The present study complements and extends other work linking cognitive styles with information behavior in that the analysis suggests a tentative mapping of stylistic differences on to a range of factors relating to researchers' problem-solving activities, perceptions and attitudes. A number of significant results have emerged which (a) are generally in accord with the picture emerging from other empirical studies reported in the literature, but (b) extend these studies in that they relate to an area which has not been widely investigated in terms of cognitive styles and other individual differences, namely the effects of literature searching during the academic research process.
It is acknowledged that the study also found a number of non-significant correlations. Although these by no means contradict the significant ones, they must be taken into account in any judgement of the strength of evidence in support (or otherwise) of the hypotheses. Of the two significant findings which were not in accord with the theoretical propositions, one could be explained in terms of a slightly more subtle interpretation of the theory which inspired the original hypothesis.
This was an exploratory study, and as such was designed to elicit what Olaisen (1991) has termed "sensitising" as opposed to "definitive" concepts, which: "... offer a general sense of what is relevant and will allow us to approach flexibility in a shifting, empirical world to 'feel out' and 'pick one's way in an unknown terrain. ... In sum, the on-going refinement, formulation, and communication of sensitizing concepts must inevitably be the building block of our exploratory theory."
The analyses presented here combine with those of a number of other studies to suggest coherent emerging patterns of interactions between cognitive styles and aspects of information behavior. But the evidence is suggestive rather than in any sense conclusive. Ford (2000) has described the nature of evidence to some extent characteristic of so called positivist and illuminative research approaches: "The limitations associated with research may be thought of as a curtain preventing us from viewing the reality beyond, that we seek to understand. Our existing knowledge ranges between two extremes, which to some extent mirror … different research approaches … One may be characterised as scattered pinpricks in the curtain, allowing clear and deep, but narrow and unconnected views through to the reality beyond. The other may be characterised as more extensive areas where the curtain is thinned, allowing complex, inter-connected but hazy shapes to show through, inviting us to trace them onto the curtain, elaborating their detail to represent what we imagine to be their reality." The evidence of the study reported here is very much of the latter type. Nevertheless, such essentially sketched patterns may be useful in the development of models of information behavior that may eventually help us respond more effectively to people's information needs. Such responses may emerge in the form of improved information systems, quality of response by human information intermediaries, and/or helping end-users enhance their own information seeking skills. However, it would seem that the notion of cognitive style is likely to form an important building block in the development of such models.
This project was funded by the British Library Research and Innovation Centre, to whom grateful acknowledgement is made.
This is a draft of a paper published in Journal of the American Society for Information Science and Technology, Volume 53, No. 9, 2002, 728-735
Note: David Ellis is now Professor, University of Wales, Aberystwyth
Uncertainty in information seeking, by Professor Tom Wilson, Dr. David Ellis, Nigel Ford, and Allen Foster
Library and Information Commission Research Report 59
ISBN 1 902394 31 3 ISSN 1466-2949
Grant number LIC/RE/019