Self-reported health is an important outcome in the evaluation of health care but is largely ignored in favor of proxy-based reporting for people with an intellectual disability. This study briefly reviews the role of self-report in health assessment of people with intellectual disability and the challenges and recommendations that have emerged from the considerable body of research on interviewing and self-report. Limitations in current recommendations are addressed from the perspective of the cognition of self-report. The review describes conceptual directions for the reconciliation of the two contradictory themes in the treatment of self-report: the centrality given to personal perceptions and choices and the methodological concerns over the meaningfulness and validity of the self-reporting process.
Self-reported health has emerged as an important outcome in the evaluation of health care and was recently proposed as one of 20 core indicators of national health and the quality of health care in the United States (Institute of Medicine, 2009; McVilly, Burton-Smith, & Davidson, 2000). There are significant merits to the approach. Self-report protocols are flexible. Large numbers of people can be screened efficiently (Patrick & Chiang, 2000). Self-reported health is a good predictor of future morbidity, mortality, and need to access health care (Idler & Benyamini, 1997; McGee, Liao, Cao, & Cooper, 1999). In theory, a subjective report can better reflect an experience of health that integrates physical and psychological status with the individual's social context. The respondent's perspective is crucial if outcomes emphasize personal experiences such as well-being, the reporting of symptoms, or health care preferences (Albrecht, 1996; Patrick et al., 2007).
But what of people with intellectual disability (ID)? The zeitgeist guiding contemporary approaches to ID have placed at their core the principles of empowerment, self-determination, and choice (Parmenter, 2001; Stancliffe, 2001; Wehmeyer & Mithaug, 2006). Although the interpretation and implementation of the concepts are variable and sometimes controversial (F. Brown, Gothelf, Guess, & Lehr, 1998; Stalker & Harris, 1998), there is general consensus on the core objectives: to enable individuals “to act volitionally and to become causal agents in their lives” (Wehmeyer, 2005, p. 120). Although not appropriate for everyone across the spectrum of ID, self-report looms large in the implementation of the key values of choice and self-determination (Stancliffe, 2001; Wehmeyer, 2005).
The manner in which quality of life (QOL) research has addressed self-report is instructive. Although the empirical work largely focused on issues related to the boundaries and measurement of the QOL construct (Schalock, Bonham, & Verdugo, 2008; Wang, Schalock, Verdugo, & Jenaro, 2010), a modest but collateral debate emerged over the role of self-report. At the heart of the debate were the core methodological challenges of interviewing and using verbal reports from individuals with ID. But an epistemological question was also raised: What is the appropriate role of personal experience in the assessment of QOL (Ager & Hatton, 1999; Cummins, 2001, 2002; Hatton, 1998)? At the heart of the critique was the observation that subjective experience is dispositional, often disconnected from external “facts,” for example, the “disability paradox” noted by Albrecht and Devlieger (1999), where QOL is routinely rated higher by people with disabilities even though life conditions would be considered undesirable by most objective criteria and social conventions. Ager and Hatton (1999) and Hatton (1998) observed that service and policy decisions based on dispositional effects can be abused and are more reliably anchored on objective, external referents as guides to “quality.” But the objection is based on the potential misuse of self-report and not on the presumptive value of personal experience in measurement, so long as reliability and validity issues can be addressed.
However, reliability and validity are significant qualifiers. A considerable methodological literature has accumulated over the years on the limitations of self-report as researchers have struggled with the utility of interview data used in research and program evaluation (Finlay & Lyons, 2001; Heal & Sigelman, 1990; Sigelman, Budd, Spanhel, & Schoenrock, 1981; Sigelman, Schoenrock et al., 1981; White-Koning et al., 2005). Similar concerns have been raised with other cognitively impaired groups, for example, individuals with conditions such as dementia (Arlt et al., 2008; Logsdon, Gibbons, McCurry, & Teri, 2002), people in the advanced stages of multiple sclerosis (Marrie, Cutter, Tyry, Vollmer, & Campagnolo, 2006; Mozley et al., 1999; Solari et al., 1998), and survivors of severe strokes (Dorman, Waddell, Slattery, Dennis, & Sandercock, 1997), as well as very young children (Varni, Limbers, & Burwinkle, 2007a, 2007b) . The core concerns are essentially the same across different populations: Can self-report be used effectively? What are the parameters of accurate reporting? What methods enhance accuracy?
These questions are very much related to the global dialogue over right to health and the matter of accountability (Hall, 2008; Hunt, 2007; MacNaughton & Hunt, 2009). The paucity of ID surveillance data generally, and health status information specifically, has led to the proposal of a number of large-scale data collection initiatives in the United States and Europe that propose incorporation of some level of self-report (Fujiura, Rutkowski-Kmitta, & Owen, 2010; Krahn, Fox, Campbell, Ramon, & Jesien, 2010; Pomona, 2008; Walsh, 2008).
In the following review, I attempt to complement, and in some cases reconsider, many of the caveats from the extant literature based on a cognitive perspective of self-report. In an especially wide-ranging review, Finlay and Lyons (2001) argued for greater attention to the reliability and validity of data drawn from interview-based research with people with ID. While acknowledging the value of self-report data, their review underscored the complexity of the interaction between the demands of self-report and the fact of impaired cognitive functioning. Research has largely focused on the identification and control of respondent error. Generally speaking, individuals with mild to moderate levels of cognitive impairment appear capable of responding, but the level of detail and degree of accuracy is influenced by the type of reporting demands, features of the instrumentation, and context of the request (Finlay & Lyons, 2001). Guidelines for structuring questions and response formats and detecting systematic errors such as acquiescence (Heal & Sigelman, 1990) are reasonably well developed if not consistently implemented. As a consequence, self-report as a means of data collection has yielded mixed results. Recommendations include avoiding Likert scales, judgments of magnitude or frequency, personally sensitive content, negative wording, use of modifiers, and the interpretation of abstractions—in short, the content, methodology, and focus of self-reported health assessment.
Such is the conundrum. Honoring personal preferences, self-expression, and choice among people with ID has primacy as both a programmatic guideline and service goal. And yet we lack consensus on how to use self-reported data, the measurement of subjectivity, and its efficacy as a data collection strategy. Perhaps most remarkable is that fact that self-report is routinely used in the face of so many basic conceptual and methodological loose ends. The following review of self-report from a cognitive perspective suggests reasons for both pessimism and hope. Obtaining valid self-reports may be even more challenging than currently characterized, but more inclusive self-report strategies may be possible through a better understanding of the underlying cognitive demands. This is the starting point for our review: wariness of the vagaries of self-report but mindful of possibilities in better understanding it.
Self-Report in ID Health Research
Self-report is rarely used in health-related research on people with ID. Applications include the assessment of health-related quality of life (HRQL; Stade, Stevens, Ungar, Beyene, & Koren, 2006; Zimmermann & Endermann, 2008); measurement of general health status and health behaviors (Boland, Daly, & Staines, 2008; Emerson & Hatton, 2007; Havercamp, Scandlin, & Roth, 2004; Lunsky, Emery, & Benson, 2002; Turk, Geremski, Rosenbaum, & Weber, 1997); and assessment of specific health states, such as pain (Bromley, Emerson, & Caine, 1998; LaChapelle, Hadjistavropoulos, & Craig, 1999), fear (Ramirez & Lukenbill, 2007), and affective status (Benson & Ivins, 1992; Emerson, 2005; Lindsay, Michie, Baty, Smith, & Miller, 1994; Lunsky & Benson, 2001a). Studies using self-report to assess health status are shown in Table 1; characterized in the table is the health indicator and relevant features of the assessment: degree of intellectual impairment among subjects, response format, accommodations for ID, and form of reliability and validity check. Almost all studies excluded subjects with a significant degree of impairment or screened subjects for verbal capacity. In general, self-report scores covaried with proxy or other comparative reports, though there were often differences in the magnitude of the ratings (Bromley, Emerson, & Caine, 1998; Stade et al., 2006; Zimmermann & Endermann, 2008). Responsiveness declines with greater intellectual impairment (Boland et al., 2008). Self-reported data were typically found to be internally consistent (Cui, Stapleton, & Suttle, 2008; Emerson, 2005; Lindsay et al., 1994; Lunsky & Benson, 2001b), though LaChapelle et al. (1999), working with an institutional sample with greater impairment, reported unreliable self-reports of pain.
The vast majority of respondent-based health assessments in the ID literature use proxies in lieu of self-report (Boulton, Haines, Smyth, & Fielder, 2006; Chien, Chou, Ko, & Lee, 2006; Cui et al., 2008; Lau, Chow, & Lo, 2006; Sabaz, Cairns, Lawson, Bleasel, & Bye, 2001). The assumption is that a third-party respondent familiar with the subject can assume their perspective and provide a response on their behalf. Efficacy evidence is mixed. Evaluations of proxy reporting find accuracy varying as a function of the proxy's relationship to the subject (Perkins, 2007; Schwartz & Rabinovitz, 2003) and type of domain being measured (Andresen, Vahle, & Lollar, 2001; Arlt et al., 2008), whereas studies validating specific instruments tend to report good proxy–subject reliabilities (Logsdon et al., 2002; McVilly et al., 2000; Schalock, Bonham, & Marchand, 2000). A convention emerging from the QOL literature suggests “splitting the difference” (McVilly et al., 2000; Nota, Soresi, & Perry, 2006; Perkins, 2007; Perry & Felce, 2002; Stancliffe, 1999), where proxies are limited to objective measurement targets but excluded from reporting the subjective states of others. General validity issues aside, proxy reporting can be considered inappropriate for those tasks where direct representation of subjective states is the point of the analysis (Ramcharan & Grant, 2001; Schalock et al., 2002; Stancliffe, 1999).
As shown in Table 1, self-report is most often accommodated through simplification of the query and response through shortening the process, reducing vocabulary demands, imposing greater response structure, such as dichotomous yes/no answers, or incorporating more visual response formats using pictures. The primary methodological issue related to self-report in these studies is framed in terms of the validity of the response. Thus, most (though not all) investigations include some form of validation procedure: (a) using criterion measures, such as comparison of self-report to a proxy informant (parent, staff person) or a second externally scored standardized measure; (b) establishing the psychometric viability of self-report data via internal consistency or factor structure; (c) using acquiescence items to identify “yea saying”; or (d) screening potential subjects for a minimum level of verbal expression or reading comprehension. In short, the methodological challenge is framed as a simple dichotomy: Is the person with ID capable of using the response format? This is a critical question, of course; however, the matter of how the self-report is generated, that is, how the subjective experience of health is translated into a response, has not yet been directly addressed.
Cognitive Approaches to Self-Report
Although studies have explored how people with ID conceptualize health and its meanings (Fender, Marsden, & Starr, 2007; Fraser & Fraser, 2001; Marks, 1996), little attention has been directed toward understanding and modeling the cognitive processes involved in responding to health-related queries. Invoking cognitive processes will most certainly complicate matters and likely introduce additional reasons for pessimism. But a more nuanced portrait of self-report may also yield new opportunities. Are there avenues for enhancing valid self-report that have yet to be explored? The study of cognitive processes in survey and other forms of self-report provides a useful starting point (Bjorner, Ware Jr, & Kosinski, 2003; Jobe, 2003; Schechter & Herrmann, 1997; Schwarz, 2007; Sudman, Bradburn, & Schwarz, 1996). A number of models have been proposed, all of which attempt to incorporate some variation on understanding the processes associated with interpreting questions, retrieving information, and generating responses. The most widely used framework is Tourangeau's (1984) model for survey responses, which framed the process as four basic cognitive tasks: (a) accurately interpreting the question (comprehension), (b) recalling the requested information (retrieval), (c) resolving ambiguous requests (judgment), and (d) deciding how to respond (response). The limitations commonly observed in self-report applications by people with ID can be described and reevaluated in the context of these cognitive tasks.
The foundational cognitive task in the Tourangeau (1984) formulation is comprehension, which entails understanding the question and correctly interpreting question intent and options for responding. Because impaired cognitive function is a defining feature of ID, it is not surprising that survey and interview modifications for people with ID have largely focused on the need to accommodate comprehension. Common examples include simplified wording and limited response alternatives (Argus, Terry, Bramston, & Dinsdale, 2004; Finlay & Lyons, 2001; Ramirez & Lukenbill, 2007; Sigelman, Schoenrock et al., 1981; Stancliffe, 1999; Stancliffe & Parmenter, 1999). Complexity of vocabulary, sentence structure, and syntax have been found to increase the likelihood of error for the general population; the presence of an ID greatly reduces the margin for error. In retrieval, the model's second stage, health-relevant events and history are recalled. The task typically takes the form of recalling a discrete event, such as an illness or hospitalization, or the generalization of more frequent classes of events, such as degree of activity (Warnecke et al., 1997). The cognition of memory is the key operation in this stage; inaccuracy in self-reported health has long been an issue in assessments requesting detail or events in the distant past (Schechter & Herrmann, 1997). The third stage of responding involves judgment, the use of the information retrieved from memory to formulate a response. A judgment task may be relatively trivial if the requested information is a discrete fact that can be directly retrieved (e.g., how old are you?). However, many health-related questions do not directly map onto readily retrievable information. For example, Blair and Burton (1987) suggested that distinct cognitive processes are involved in responding to questions about low-frequency events (e.g., “how many times were you hospitalized?”) versus high-frequency behaviors (“how often do you watch television?”) because high-frequency events are difficult to enumerate. A response requires an estimate of rates that is a far more sophisticated cognitive task than simple counting. The final cognitive task in Tourangeau's (1984) model is the production of a response. This is an active process in which respondents may choose to omit information or knowingly invent or distort information. In other words, the respondent makes decisions regarding how to answer or what to include in the answer. Decisions reflect an “editing process” that can be influenced by the interview or questionnaire protocol, social desirability, or a conscious effort to maintain consistency with previous responses (Tourangeau & Rasinski, 1988, pp. 300–301).
In their review of interviewing and self-report for individuals with ID, Finlay and Lyons (2001) almost parenthetically noted that “It is possible that people with mental retardation may be too heterogeneous in terms of personal history and linguistic and cognitive abilities for any single questionnaire to be valid for the whole population” (p. 329). The observation is critical for it suggests the futility of asking a global question about the appropriateness of self-report for ID generally. The extraordinary heterogeneity of individuals under the very large ID umbrella will doom any effort to develop a generalized interview protocol; there will always be individuals able to use unmodified protocols, those too cognitively impaired for responses of any kind, and a vast majority who will lie somewhere along the continuum between extremes. Perhaps the research question is not an “either–or” proposition but rather one that can be framed in terms of how best to solicit self-reported information, how much, and from whom? Answering such questions will necessarily require more than identifying those capable of using existing protocols or incorporating methods to detect and control respondent error. The approach reflects both necessity and opportunity: If self-determination lies at the core of the field's operating principles, then we must directly confront the treatment of personal experience and self-report in the face of considerable evidence that its utility is severely limited. A better understanding of the cognitive processes involved in generating responses to questions about health may help us include those currently at the margins of self-expression. We now turn to this theme.
Applications for Self-Report Among People with ID
Simplification—of query and response—is the primary accommodation. But cognitive perspectives on the process suggest that simplification touches only the veneer of responding to survey or interview questions. Information is not passively recalled and given directly to the interviewer. The respondent is an active participant in the process of recognition, recall, interpretation, and formulation of a response. Although cognitive considerations complicate how one approaches self-report, compensatory strategies to maximize comprehension and responses that are congruent with question intent are also suggested. Thus, in addition to the traditional avenues of identifying those “capable” of valid responding and developing specialized “ID” instrumentation, a complementary line for self-report research is to develop an understanding of how questions are perceived, processed, and responded to.
Simplification is necessary but not sufficient. In the Tourangeau (1984) model, comprehension entails not just recognition of the question vocabulary but interpretation of meaning and perception of question intent. Consider, for example, the word “health,” invoked in 7 of the 12 items of the SF-12, one of the most widely used measures of self-reported health (Ware, Kosinski, & Keller, 1996). A cognitive perspective that distinguishes recognition and interpretation must consider variations in personal definitions of the “health” concept. A word may be universally recognized but attached to many different meanings. Some may frame the word “health” narrowly in terms of the absence of illness; others may attach more global meanings, such as well-being or optimal function. In Marks's (1996) exploratory study of the conceptualization of health by adults with ID, subjects across a relatively wide range of cognitive impairment responded to an open-ended format question on health conceptions. Subjects presented multidimensional views of health that encompassed adaptation, ability to fulfill roles, and life satisfaction, as well as more traditional criteria, such as absence of disease or function. If the context for the question is ambiguous, answers may be influenced by a respondent's inference regarding the intent of the question (Barofsky, 2003; Fienberg, Loftus, & Tanur, 1985; Schechter & Herrmann, 1997). Survey and interview design attempts to control intent and enhance question clarity by grouping related questions to establish contextual effects, manipulating word selection and syntax, or using customized rather than standardized questions that take into account experiential differences and backgrounds across subgroups (Fowler, 2009). Consider an item from the Arc's Self-Determination Scale for adolescents (Wehmeyer & Kelchner, 1995): “My friends and I choose the activities we want to do.” How the term friends is used by the respondent could be highly variable depending on the expansiveness of his or her social history. Different versions of the question could be deployed for individuals with vastly different backgrounds, such as a history in the family home versus long-term residential care. A cognitive approach suggests placing as much emphasis on the adequacy of the question asking as is currently devoted to determining the competence and reliability of those providing the answers.
Leveraging Memory Research
Memory is a foundational construct for theories describing the learning and behavioral profiles of people with ID (e.g., Detterman, Gabriel, & Ruthsatz, 2000; Ellis, 1963). As such, there is a long, illustrious, and complex history of research; explanatory models continue to evolve (Edgin, Pennington, & Mervis, 2010). Tourangeau's (1984) caveat regarding the futility of concise syntheses of memory research is embraced here—the implications of this complex body of work for understanding the cognition of survey responding lies well beyond the scope of this article. But engaging models of memory will be necessary in any cognitive portrait of self-report. I touch on only two of many potentially relevant threads of memory research to illustrate the nexus of memory theory and the development of self-report models.
Emerging conceptualizations of memory function involve subsystems interacting with executive function (Henry, Cornoldi, & Mähler, 2010). Both constructs are invoked as the key processes in understanding the role of memory in the cognitive performance of people with ID. Working memory involves the temporary storage and manipulation of information and is postulated to be one of the key systems directing purposeful behavior (Baddeley & Jarrold, 2007; Edgin et al., 2010; Schuchardt, Gebhardt, & Mäehler, 2010; Van der Molen, Van Luit, Jongmans, & Van der Molen, 2009). Research finds great variance in working memory capacity within the ID population (Edgin et al., 2010; Jarrold, Nadel, & Vicari, 2009; Silverman, 2007), and evidence points to its centrality in the performance of daily tasks and, notably, its amenability to intervention. Consistent with concerns over the structure and format of self-report questions, working memory studies have found recall accuracy for people with ID to be far more dependent on question format, novelty of recall task, and complexity of the recalled information (Van der Molen, Van Luit, Van der Molen, & Jongmans, 2010). A second relevant thread of memory-related research is found in the study of individuals with ID as witnesses in legal proceedings. In contrast to laboratory-based memory performance, differences between individuals with and without ID appear less dramatic on more “natural” memory tasks, such as recall of incidental events (Henry & Gudjonsson, 2003). Adequate recall for witnessed events has been reported among people with mild to moderate ID, albeit at a reduced level of detail and accuracy compared with their peers without ID (Agnew & Powell, 2004; Henry & Gudjonsson, 2007; Kebbell & Hatton, 1999; Milne & Bull, 2001). There is, however, far greater variance in recall performance among witnesses with ID (Henry & Gudjonsson, 2003). Of significance for the present discussion are the suggestions of psycho-educational interventions to improve the ability of people with ID to use cognitive strategies, recommendations that have been made based on findings in basic working memory studies as well (e.g., Henry, 2008; Perrig, Hollenstein, & Oelhafen, 2009; Van der Molen et al., 2009; Willner, Bailey, Parry, & Dymond, 2010). The approach would be more consistent with a perspective of memory differences due to deficits in the use of recall strategies than a model of impaired memory “machinery” (Bray, Fletcher, & Turner, 1997, p. 212). In an experimental memory analogue, for example, Bray, Huffman, and Fletcher (1999) noted that children could describe how they recalled items and that the descriptions corresponded well to imputed use of different strategies, suggesting the importance of identifying and modifying inappropriate respondent recall strategies.
Heuristics in Interviewing
However, the utility of strategies extends beyond recall. Cognitive approaches to self-report emphasize the importance of heuristics, or the cognitive “shortcuts” that are used in formulating a response (Collins, 2003). Many of the methodological recommendations that have emerged from the ID-focused self-report and interview literature (Finlay & Lyons, 2001) are consistent with the observations of cognitive researchers suggesting use of heuristics that are faulty or limited in some fashion (Baroff & Olley, 1999; Marquis, Moore, & Huggins, 1990): avoiding question content involving quantitative estimates, time-based judgments, comparisons, generalized evaluations (e.g., “in general” or “usually”), abstractions, or sensitive content. Although these types of queries present significant reliability and validity problems when used with respondents with ID, the challenge is not exclusive to those with cognitive impairments. Consider the cognitive demands imposed on any respondent generating an answer to a question requiring time-based estimates, comparisons, or use of a Likert scale ranking. A prototypical example in self-reported health is the comparison of one's current state of health with a previous point in time (“Are you better, worse, or the same compared with …?”). Generating an answer requires more than accurate recall; strategies must be invoked, such as rescaling the value of different types of health events across time periods or subjectively weighting the importance of a past perception of well-being. Using a Likert response option imposes additional arithmetic operations, which require decisions about how best to represent a subjective estimate on the interviewer's scale (Tourangeau & Rasinski, 1988).
Abstractions compound the challenge. When confronted with requests for information that does not map onto an existing fact, a respondent must engage in multiple levels of processing. Heuristics are used: attaching meaning to nonspecific terminology, the scaling or weighting of values related to the terminology, or the combining of multiple different recall events into a more global response (Lessler & Sirken, 1985; Tourangeau & Rasinski, 1988). More complex strategies may be used in response to queries focused on attitudes and values (“How do you feel about … ?”), causal attributions (“Why did you …?”), and health-related constructs, such as physical or emotional states. Attitudes, values, and our understanding and use of constructs are rarely fixed and often understood only in context. A respondent may use a heuristic that samples from among different hypothetical situations in formulating a response (Tourangeau, Rips, & Rasinski, 2000). Other strategies might involve the averaging of attitudes or the selection of one attitude option over others based on weightings given to a question's wording, available response options, and question context.
Finally, sensitive or emotionally loaded material and intensity of the recalled experience have been recognized in the general population as problematic, influencing both the accuracy and quality of answers (Mitchell et al., 2007; Rasinski, Willis, Baldwin, Wenchi Yeh, & Lee, 1999). A highly emotional event, for example, is often perceived as having occurred more recently than an emotionally neutral event (Bolton, 1993). When reporting sexual activity, there are gender differences in heuristic, with women enumerating partners whereas men tend to estimate (Mitchell et al., 2007). Wyrwich and Tardino (2006) found people with chronic disease conditions more likely to self-report HRQL on the basis of their present experience of health rather than recent past events. The peak-end heuristic (Stone, Schwartz, Broderick, & Shiffman, 2005), in which respondents are biased toward reporting the most intense parts of an experience, has been shown to negatively bias pain reporting. The point is that heuristics are not fixed; their quality and effectiveness are highly variable, regardless of cognitive status. To what extent do poorly developed or erroneous strategies dictate other limitations in self-report among respondents with ID? What forms of heuristics are used? Can problem-solving heuristics be incorporated into the interview process? Again, question simplification is a reasonable start in facilitating self-report, but it is only a first step. Consider the SF-36 question “During the past 4 weeks, to what extent has your physical health or emotional problems interfered with your normal social activities with family, friends, neighbors, or groups?” A better understanding of cognitive strategies involved in judgment might suggest breaking down questions into component demands in order to mimic the structure of a heuristic, each more reliably assessed than the global question left to the vagaries of each person's cognitive strategies: scaling of “extent” to match the five SF-36 options ranging from not at all to extremely, attaching weights to each of the alternative social activities (are interactions with neighbors as important as friends?), and averaging responses to render the judgment. Although this is an oversimplification, the example illustrates how a more systematic application of self-report cognition can be used to supplement the assessment process.
A Harder Look at Acquiescence
The notion of response effects as conceptualized in the Tourangeau model (Kreuter, Presser, & Tourangeau, 2009) has long been a source of concern in the validity of interviewing individuals with ID. Indeed, much of the self-report research conducted on people with ID has focused on procedures to attenuate response bias or minimize distortions associated with acquiescence and social desirability (Finlay & Lyons, 2002; Matikka & Vesala, 1997; Sigelman, Budd et al., 1981). The tendency to provide inconsistent or openly distorted information is well established; the presumption is that these artifacts are a manifestation of social desirability or submissiveness. Agnew and Powell (2004), for example, found discrepancies in the recall of past events across children with and without ID matched on memory and language and suggested the differences were due to the heightened awareness of social clues among the children with ID. Suggestibility has long been the focus of legal investigations and parallels the literature on acquiescence in self-report research (Gudjonsson & Clare, 1995; Henry & Gudjonsson, 2007; Milne, Clare, & Bull, 2002).
Tourangeau's consideration of response processes suggests possible alternative explanations for the inconsistencies most often associated with acquiescence. In the original model, the act of responding was framed in terms of two distinct processes: (a) how respondents select their answers from among different options and (b) how respondents misreport their answers. The ID self-report literature has focused almost exclusively on the latter process. A number of strategies have been developed to identify and control acquiescence through reverse-worded questions, forcing either–or answers or providing “don't know” options. In short, far more attention is directed to the behavior of “yea saying” than to the precedents in the “black box.” Phenomenological studies, although not ostensibly about the cognition of self-report, strongly suggest the active interpretation and reconstruction of questions by people with ID. Using conversational analyses, Antaki and Rapley (1996) and Rapley (2004) challenged the notion of acquiescence, arguing that what often appears to be “yea saying” is often an artifact of context, biased testing, and interview methodologies. They described an interview context in terms of a very dynamic interpersonal exchange, represented in conversation with paraphrasing, feedback, and clarifications by the interviewer and an effort by the interviewee to respond to these shifting demands. This is a highly interactive process that can lead to responses unwittingly shaped by the interviewer. Kravetz, Katz, Alfa-Roller, and Yehoshua (2003), working within a theory of mind paradigm (Leslie, 1987), found that the tendency toward acquiescence in self-report measures of QOL was associated with differences in the use of cognitive strategies. In their analysis, the capacity to reflect on the subjective states and experiences of others was negatively correlated with acquiescence. Kravetz et al. (2003) speculated that the capacity to view the interviewer as a person with factual knowledge and beliefs affected the use of contradictory responses. In a similar vein, Gudjonsson and Clare (1995) found that confabulation, the reporting of false events believed to be true, was largely unrelated to suggestibility or acquiescence, suggesting a more nuanced underlying cognitive process than simple misreporting. The point here is that there is a need to be more expansive in our empirical constructions of the response. Controlling bias is important, of course, but our focus on eliminating misrepresentation may be unnecessarily restricting our perspective on what is possible.
The general thrust of recommendations regarding self-report-based data collection has taken the form of “splitting the difference”—ranging from adopting alternative methods for those who are nonverbal, to using multiple methods (Schalock et al., 2000), mixing self-report with proxies (Nota et al., 2006), using protocols beyond the traditional question-and-answer approach (Mactavish, Lutfiyya, & Mahon, 2000), and imposing greater methodological rigor in how self-reports are obtained (Finlay & Lyons, 2001; Perry & Felce, 2002). The point of the preceding review is to suggest that only the outer shell of self-report has been addressed: reducing response errors via alternative task demands or screening out those people who may be unreliable or unproductive respondents. Cognitively oriented studies suggest that a great deal is yet unknown about the parameters of self-report for individuals with ID. There are direct parallels in the efforts to apply cognitive therapies to people with ID (Sturmey, 2004) and legal witness research (Henry & Gudjonsson, 2007; Milne & Bull, 2001). Realistically speaking, cognitive models will not reveal dramatically new worlds of subjective experience to survey and interview methods. But the approach has the potential of improving the sensitivity, scope, and validity of self-report by identifying strategies for facilitating both response and response accuracy.
To what extent can people with ID accurately describe the cognitive strategies used to generate responses? In addition to having the requisite verbal skills, access will require respondents who can reflect on and express the different components of the response. There is some evidence for the feasibility of conducting meaningful cognitive testing of individuals with ID in the phenomenological studies of meanings, constructions, and conversational analysis (Antaki & Rapley, 1996; Kravetz et al., 2003) and in the studies of people with ID as witnesses in legal proceedings (C. Brown & Edward Geiselman, 1990; Kebbell & Hatton, 1999). The utility of free recall via the cognitive interview or similar approach is a common recommendation emerging from witness research (Colwell, Hiscock, & Memon, 2002; Milne & Bull, 2001; Milne, Clare, & Bull, 1999; Perlman, Ericson, Esses, & Isaacs, 1994). Fisher, Brennan, and McCauley (2002) suggested that the utility of the cognitive interview is due to the core features of the interview process: reviewing the context of the event being queried, adapting to the interviewee's capability in a flexible question-and-answer format, and encouraging the use of modalities for recall, among other reasons. Development and evaluation of cognitive testing protocols represent an important task for future research.
The dialogue over self-report is largely focused on the capacity of the person to fit the measurement task. Perhaps a shift in perspective is necessary. Can the measurement tasks be reevaluated and adapted to the capabilities of the interviewee by better understanding how responses are formulated? Can more flexible designs and interview formats and compensatory cognitive strategies enhance accuracy and expand the pool of potential self-respondents? In other words, is it possible to fit measures to the person? To paraphrase Taylor and Bogdan (1996), the issue is not whether we can reliably measure self-reported health from individuals with ID but what we can learn from their experience.
The contents of this article were developed under a grant from the Department of Education, NIDRR grant numbers H133B040034 and H133B08000. However, these contents do not necessarily represent the policy of the Department of Education, and you should not assume endorsement by the Federal Government. The findings and conclusions in this article are those of the authors and do not necessarily represent the official position of the Centers for Disease Control and Prevention.
The members of the RRTC Expert Panel on Health Measurement are Elena Andresen, Oregon Health and Science University; Bradley J. Cardinal, Oregon State University; Charles E. Drum, University of New Hampshire; Glenn T. Fujiura, University of Illinois at Chicago; Trevor Hall, Oregon Health and Science University; Willi Horner-Johnson, Oregon Health and Science University; Gloria Krahn, Centers for Disease Control and Prevention; Margaret Nosek, Baylor College of Medicine; Jana J. Peterson, University of Missouri-Kansas City; Rie Suzuki, University of Michigan-Flint.
Editor-in-Charge: Susan Parish
Glenn T. Fujiura (e-mail: email@example.com), Department of Disability and Human Development (M/C 626), University of Illinois at Chicago, 1640 W. Roosevelt Road, Chicago, IL 60608, USA; and RRTC Expert Panel on Health Measurement.