Evidence for acquiescence (yea-saying) in interviews with people who have mental retardation is reviewed and the different ways it has been assessed are discussed. We argue that acquiescence is caused by many factors, each of which is detected differentially by these methods. Evidence on the likely causes of acquiescence is reviewed, and we suggest that although researchers often stress a desire to please or increased submissiveness as the most important factor, acquiescence should also be seen as a response to questions that are too complex, either grammatically or in the type of judgments they request. Strategies to reduce acquiescence in interviews are reviewed and measures that can be taken to increase the inclusiveness of interviews and self-report scales in this population suggested.
Editor in charge: Steven J. Taylor
Many investigators have suggested that acquiescence is a potential difficulty when interviewing people with mental retardation. This is often discussed in the context of research showing that people with mental retardation are more suggestible (Clare & Gudjonsson, 1995; Henry & Gudjonsson, 1999) and more responsive to external cues (e.g., Bybee & Zigler, 1992; for reviews see Wong, 1980; Zigler & Hodapp, 1986) than are individuals in the general population. Although there is now much evidence suggesting that acquiescence can be a problem when using self-report scales or other interviews with this population, more research is required on understanding the causes of acquiescence and techniques of minimizing the problem (Matikka & Vesala, 1997; Sigelman, Budd, Spanhel, & Schoenrock, 1981a). In this paper we review evidence for acquiescence in interviews with people who have mental retardation and discuss the different ways in which it has been assessed. We argue that acquiescence is caused by a range of factors, each of which is detected differentially by the methods used. It is important, therefore, when building acquiescence checks into interviews, that through the method used, investigators are able to detect the type of acquiescence that is most likely to arise in the scale or interview schedule in question. We also review strategies used to reduce acquiescence in interviews and suggest measures that can be taken to increase the inclusiveness of interviews and self-report scales in this population.
Acquiescence is defined in the psychometric literature as the tendency to agree with or say yes to statements or questions, regardless of the content of the items (Block, 1965; Couch & Keniston, 1960). Early investigators of acquiescence examined true–false tests in academic contexts and found that students tend to answer “true” when in doubt (Cronbach, 1942, 1950). Scale developers recognized it as an important factor in the development of attitude and personality scales for the general population, and much research has been carried out into ways of detecting and minimizing it (e.g., Block, 1965; Cloud & Vaughn, 1970; Peabody, 1966; Winkler, Kanouse, & Ware, 1982). Acquiescence is one of a number of response sets that becomes more important when items are difficult or ambiguous (Cronbach, 1950; Ray, 1983).
In this paper we address one aspect of acquiescence, the tendency to say yes in answer to questions (“yea-saying”). We note that some writers use the term acquiescence to describe yea-saying only, whereas others use it more broadly to indicate a general submissiveness, including behavioral compliance as well as a wider range of verbal responses (e.g., “true,” “agree”). Because most writers have used the narrower definition, we use the term acquiescence here to refer to yea-saying only, and, therefore, this article is mainly focused on causes of yea-saying that lie in the design of questions and interview schedules and solutions to these aspects. Where writers have used it in a broader sense, this will be indicated. The concept of suggestibility (the tendency to accept messages as true), and the extent to which it overlaps with yea-saying in practice is discussed later. Although many writers have explained yea-saying as due to submissiveness, we suggest here that linguistic and conceptual factors are also important, and it is these that are most amenable to modification on the part of the interviewer. The purpose of this review is, therefore, to provide information on the use of yes/no questions for those who are involved in interviewing people with mental retardation for clinical assessment, service development or planning, or forensic or research purposes. As a result, the recommendations that we offer are focused on ways in which questions and interview schedules should be designed and conducted in order to minimize problems of yea-saying in this population.
In a series of studies using item reversals, questions where the answer should be no, and informant checks, Sigelman and her colleagues found a systematic acquiescence bias in response to yes/no questions (Sigelman & Budd, 1986; Sigelman, Budd, Spanhel, & Schoenrock, 1981a, 1981b; Sigelman, Budd, Winer, Schoenrock, & Martin, 1982; for a review see Heal & Sigelman, 1995). This was found for factual as well as subjective questions. They also found acquiescence to be inversely related to IQ (Sigelman et al., 1981a; see also Burnett, 1989; Gudjonsson, 1990; Shaw & Budd, 1982; for contradictory findings, see Matikka and Vesala, 1997). Response biases were less of a problem for either/or questions, although there was a tendency in these cases for last-choice responding (see also Loper & Reeve, 1983). Yes/no questions were also found to produce the highest levels of responsiveness, followed by either/or questions, with open-ended questions producing the least (Sigelman, Winer, & Schoenrock, 1982). High levels of yea-saying in people with mental retardation have also been found by other authors (Burnett, 1989; Clare & Gudjonsson, 1993; Gerjuoy & Winters, 1966; Heal & Chadsey-Rusch, 1985; Perlman, Ericson, Esses, & Isaacs, 1994).
The suggestion that yea-saying in interviews is a particular problem for people with mental retardation has not been fully supported by other researchers, who find that it is not as common as Sigelman had found (e.g., Booth & Booth, 1994; Conroy & Bradley, 1985; Matikka & Vesala, 1997; Rapley & Antaki, 1996; Wehmeyer, 1994). Differences in the rates of acquiescence found across studies might be due to differences in sample sizes and characteristics and in the techniques used in detection. For example, some of the investigators finding less evidence of acquiescence used qualitative methods with small samples and had no direct measure of acquiescence (e.g., Booth & Booth, 1994a, 1994b; Rapley & Antaki, 1996). Because investigators often find that acquiescence is negatively related to cognitive abilities, it is important to consider the sample selection criteria used in different studies. For example, Matikka and Vesala, who found less evidence of acquiescence in quality of life interviews with over 500 participants, excluded a large number of participants who were not judged to have sufficient communicative or interactive skills. Wehmeyer recruited participants from self-advocacy groups, and they were required to be able to use a pen and paper response format. Booth and Booth pointed out that Sigelman's research was with adults who lived in state institutions and children and that similar results may not be found with adults who live in the community. They suggested that acquiescence may be more adaptive for people living in institutions, and, therefore, it is not having mental retardation in itself that gives rise to this phenomenon, but, rather, the environments in which people live. However, in a review of research with adults living in the community, Heal and Sigelman (1995) also found marked acquiescence, as did Rosen, Floor, and Zistein (1974) and Perlman et al. (1994).
Techniques Used to Detect Acquiescence in Interviews
Although the detection of acquiescence is recognized as a problematic issue in personality and attitude research in general (Block, 1965), it is often treated as a fairly straightforward process in research with people who have mental retardation. Researchers have tended to use four techniques: nonsense questions (where the answer should be no), pairs of questions that are opposite in meaning (e.g., Are you happy/Are you sad?), pairs of questions in which the same question is asked in different formats (e.g., yes/no and either/or), and informant checks. In addition, suggestibility research involves staged events or stories read out loud to participants, followed by a series of leading questions. In this section we suggest that each of these methods involves different problems of interpretation, and the acquiescence that they detect can be explained in different ways. The relevance of any study using such methods for other studies using interviews or self-report scales is, therefore, not a straightforward issue.
Nonsense questions were used by Sigelman et al. (1981a), who asked five questions to which the answer should have been no. These included, “Does it usually snow in the summer here?” and “Do you know how to fly an airplane?” Similarly, in a scale to assess tolerance of rules, Flynn, Reeves, Whelan, and Speake (1985) included questions such as, “Do you think you should always wear sunglasses when walking under a ladder?” Rapley and Antaki (1996) criticized the use of nonsense questions because one cannot assume that answers to such bizarre questions are comparable to those for sensible questions. Certainly, asking absurd questions must create problems regarding participants' interpretations of the research situation, and affirmative answers might indicate the person is sharing a joke, playing along, or is simply perplexed.
Despite these problems, however, Sigelman et al. (1981a) offered some evidence for the validity of this technique because those who scored higher on nonsense question measures of acquiescence also scored higher on reverse wording measures and gave more affirmative answers when asked about their involvement in household chores. This increased involvement was not confirmed by informants, indicating that those participants scoring higher on the acquiescence tests were also more likely to answer factual questions in the affirmative whether or not this was the correct answer (at least according to informants—see later discussion).
A number of investigators have attempted to measure acquiescence through including several pairs of questions in the interview schedule that are logical opposites. Acquiescence is identified if a respondent affirms both versions (e.g., Flynn et al., 1985; Heal & Chadsey-Rusch, 1985; Matikka & Vesala, 1997; Sigelman et al., 1981a, 1981b). The use of reverse wordings, however, is controversial. Block (1965) pointed out that reversing complicated statements is a difficult task and that endorsement of both items often does not represent psychological inconsistency (see Heath, 1986, for examples in attitude research). For example, Matikka and Vesala carried out a study with participants who had mild and moderate mental retardation. Their survey included four pairs of oppositely worded questions (e.g., “Do you get treated fairly/unfairly?” and “Are you stupid/wise?”). They found an average acquiescence rate of 25%, although only 8% of participants answered acquiescently to three or four of the question pairs. The authors pointed out that saying yes to oppositely worded questions might not be illogical because they may elicit different images on which judgments are based. For example, a person may be treated fairly by some people and unfairly by others, or they may see themselves as stupid in some situations and wise in others. Writers in the personality literature have also suggested that acquiescence in the general population can be caused by a confirmation bias, whereby people search their memories for positive examples and ignore disconfirming instances (see Knowles & Condon, 1999).
The same problem can be seen in the Lifestyle Satisfaction Scale. In this instrument, Heal and Chadsey-Rusch (1985) used oppositely worded questions in order to estimate acquiescence. These included, “Would you like to move back to ——?/Do you like living here?” and “Are you happy with what you do in your free time?/Do you wish you could enjoy your free time more?” This technique provided an estimate that the authors used to correct raw scores on their quality of life questionnaire. However, participants could answer yes to both such questions without this indicating a contradiction. People might both enjoy living in their current residence and like the idea of going back to some previous residence to be with old friends. They might enjoy the activities they do as well as wanting to do more. Because saying yes to both questions does not necessarily indicate a contradiction, reverse wordings need to be carefully constructed. The success of reverse wordings not only depends on whether they really are logically inconsistent, but also on the participants being able to attend to the subtleties of phrasing of each form of the question. As a result of these difficulties in interpretation, Matikka and Vesala (1997) suggested that acquiescence should be identified only when people give contradictory answers to most question pairs, rather than just one or two.
The use of reverse wordings is also criticized by Rapley and Antaki (1996), who suggested that inconsistency is being confounded with acquiescence. Inconsistency might be caused by memory failure, lack of attention, or rhetorical reasons. Certainly, the technique assumes that if people think carefully enough, they can have a single, true position on an attitude or personality item, an assumption that often may not be the case (Potter & Wetherell, 1987).
A similar technique for detecting acquiescence is to use different forms of the same question. For example, Conroy and Bradley (1985) carried out a survey of consumer satisfaction using yes/no, either/or, open-ended, and multiple-choice questions. The latter required the person to select from five faces (from a big smile to a big frown), representing a Likert scale. Acquiescence was checked by asking five pairs of similar questions in both yes/no and either multiple choice or either/or formats (e.g., “Do you like living here/Which face is most like how you feel about living here?”) The researchers found acquiescence rates lower than those in the Sigelman studies, and their technique had the advantage of not using nonsense question or reverse wordings, both of which may lead to overestimations of the problems of acquiescence. However, the usefulness of this method depends on the validity of the alternative question formats. Sigelman, Budd et al. (1982) compared responses to yes/no questions to those for open-ended questions when asking about participation in a range of activities. If a person said yes to an activity and then failed to report it when asked an open-ended question, this was taken as an indication of acquiescence. However, such inconsistencies may arise because of problems with the open-ended questions, particularly given the low levels of responsiveness associated with such formats in this population (Sigelman, Budd et al., 1982).
Acquiescence has also been measured by comparing self-reports to those of informants. For example, Sigelman, Budd et al. (1982) asked respondents whether they participated in a range of sports and compared this with informant reports. They found that participants reported significantly more involvement in sports than informants reported. As Sigelman and colleagues pointed out, this technique requires researchers to assume that the informants are correct should there be any discrepancy. Because informants are usually caregivers who interact with the person in limited contexts, it is likely that in some cases it is the informant who is wrong rather than the person with mental retardation. A second problem is that the respondent may interpret terms in a more general or restricted sense compared to the informant or may make a judgment based on different criteria. Investigators have found, for example, that some people with mental retardation include different people in the category “friend” compared to those included by researchers (Barlow & Kirby, 1991) and may use different comparative contexts in making judgments about their level of choice when compared to judgments by support staff (Stancliffe, 1995). When discrepancies are found, then, it may be due to differences in interpretation or in the context in which the judgment is made rather than because of acquiescence. For example, when a child who is institutionalized answers yes to the question “Do you play baseball” (Sigelman, Budd et al., 1982), it may be because they (a) play regularly in organized games, (b) they play once or twice a year in an organized game (e.g., at a party), or (c) they have played informally once or twice in their lives. The informant may use different criteria and only include regular formal participation as sufficient for an affirmative answer.
Causes of Acquiescence in Interviews
Acquiescence is considered to be the result of multiple factors. In this section we show that although the literature on acquiescence by people with mental retardation has tended to concentrate on explanations relating to social desirability or submissiveness, other important determinants of acquiescence, which may have more implications for interviewing practice, have received less attention. These include the methodologies used to detect acquiescence, the type of judgments that are requested, the wording of questions, and the interactional situation to which interviewees respond.
Although it is sometimes assumed that acquiescence is due to impression management concerns, most of the literature on personality and attitude assessment in the general population has focused on cognitive explanations. Acquiescence is more likely to occur (a) when answers are not known (Cronbach, 1942, 1950), (b) when questions are ambiguous (Ray, 1983), and (c) when the person has spent less time or effort in considering a question (Couch & Keniston, 1960; Knowles & Condon, 1999; Knowles & Nathan, 1997). This might be due to a lack of motivation or from limited cognitive abilities (Javeline, 1999). There is also some indication that acquiescence is related to certain cognitive or personality styles, such as greater impulsiveness, extroversion, and stimulus acceptance (Couch & Keniston, 1960) and greater cognitive simplicity, rigidity, and intolerance of alternatives (Knowles & Nathan, 1997). We note, however, that consistent evidence for any of these explanations is lacking (Javeline, 1999; Schuman & Presser, 1981).
The degree of acquiescence found is also dependent on the type of scale under consideration. Ray (1983) found that acquiescence on personality scales is not correlated with that on attitude scales, suggesting that there are different types of acquiescence. Ray proposed that this is because the meaning of items on some scales are clearer and more precise than on other scales, although different groups will find different items ambiguous or difficult to answer.
In the literature on people with mental retardation, researchers often assume that the most important factor in acquiescence is a greater tendency to submissiveness or a greater desire to please. This might be seen as a personality trait or as a learned response that is adaptive in a range of situations in which people with mental retardation commonly find themselves. In an early study Rosen et al. (1974) used the broader definition of acquiescence as “a predisposition to comply or submit to persuasive or coercive attempts by others, even when such attempts are perceived as contrary to the best interests of the individual” (p. 60). This was assessed over a range of tasks, including responses to ambiguous questions (which were repeated using reverse wording); the Asch picture judgment task; a drawing test; and several behavioral compliance tasks, such as being told to take a pill, touch something that would give them an electric shock, sign some documents, and carry out repetitive behaviors. The study involved three groups of participants, institutionalized and community-based people with mental retardation and a mental-age (MA)-matched control group. Effects were found on the yes/no questions, with both groups of people with mental retardation showing more acquiescent responding when measured on both total number of affirmative responses and on contradictions. On the picture-drawing task, both groups showed more stylistic copying of another picture that was present. Measures of behavioral compliance did not show such effects. On two of these measures, the institutionalized participants showed more compliant responding, but those based in the community did not, and on the other measures there were no significant differences. Rosen et al. concluded that people with mental retardation are particularly prone to manipulation by others. This conclusion seems overstated because the effects they found were on measures using yes/no questions and copying on a drawing task, and there was less evidence on behavioral compliance measures. Limited evidence for the latter was found among people who were institutionalized but not among those living in the community.
Many others writers have also suggested that one reason for higher levels of acquiescence in this population is a greater desire to please or conform with expectations (e.g., Perlman et al., 1994; Sigelman et al., 1981a). This tendency is often assumed to result from a greater experience of failure and the greater levels of control exerted by others over the lives of people with mental retardation. For example, Henry and Gudjonsson (1999) suggested that when cognitive factors are not responsible for greater suggestibility, then interrogative pressure creates it because the children in their study had a greater eagerness to please, lower levels of confidence, and a reluctance to disagree with the interviewer. Similarly, explanations for greater outerdirectedness are that the person has a greater history of failure, so has come to distrust their own resources (Zigler & Hodapp, 1986). Perlman et al. (1994) suggested that people with mental retardation tend to agree with authority figures because they have multiple workers in authority over them, and many have participated in training programs to enhance their compliance with the expectations of those in authority. Although these explanations seem plausible, particularly given the unequal distribution of power and resources people with mental retardation often face, we note that they are usually offered as explanations without substantiating evidence being provided.
Although they clearly overlap, the relationship between suggestibility and acquiescence is one that has not been sufficiently elaborated in the literature. Gudjonsson (1990) defined acquiescence as “the tendency of the person to answer questions affirmatively regardless of content,” and suggestibility as “the extent to which people come to accept messages communicated during formal questioning, as a result of which their subsequent behavioral response is affected” (p. 227). Suggestibility, then, refers to the personal acceptance of information as true and is distinguished from compliance, which is when one does not accept the message but nevertheless agrees with the interviewer due to a desire to please (Gudjonsson & Clark, 1986). Despite this conceptual distinction, however, investigators are unable to use measures of suggestibility to identify when a message has been personally, as opposed to publicly, accepted. We argue here that because measures of suggestibility have involved yes/no questions where a yes answer is incorrect, they can often be taken as measures of yea-saying. In such cases, the concept of suggestibility, as an underlying disposition, cannot be used as an explanation for acquiescence in interviews.
Gudjonsson (1983) found that in the general population there is a strong negative correlation between measures of suggestibility and both IQ (see also Gudjonsson, 1990) and memory and a positive correlation with neuroticism and social desirability. In subsequent research, investigators have suggested that the relation between suggestibility and IQ is subject to range effects and is likely to occur for those with below-average IQ rather than those with above-average scores (Gudjonsson, 1988; Henry & Gudjonsson, 1999). Several investigators have found that people with mental retardation are more suggestible than are those in the general population (Clare & Gudjonsson, 1993; Everington & Fulero, 1999; Henry & Gudjonsson, 1999). In these studies, suggestibility is discussed in relation to eye-witness testimony, where an event is presented to a person, either in the form of a staged event or a story read orally to them, and then questions are asked about the details. Suggestibility is identified when misleading questions are used and the person answers in the direction suggested by the question.
Gudjonsson and Gunn (1982) suggested three types of misleading questions that can be used to detect suggestibility: those that require a yes/no answer, either/or questions where both options are wrong, and questions that assume incorrect information (e.g., “What was the name of the man who came into the room?” when no man had come in). It becomes difficult to separate the concepts of acquiescence and suggestibility when the latter is measured using misleading yes/no questions that suggest the answer should be yes (e.g., “The lady wore a white dress, didn't she?” when the lady actually wore a blue dress). In these cases, yea-saying and suggestibility overlap. This is the case in the Gudjonsson Suggestibility Scale-Form 2 (Gudjonsson, 1987). In this scale, suggestibility is measured with 15 misleading questions about a story that has been read to the participant. Ten of these are yes/no questions that suggest the answer is yes, and 5 are either/or questions. Some researchers who find a significant relationship between measures of acquiescence and suggestibility confound their methods in this way (e.g., Clare & Gudjonsson, 1993). This is not a criticism of the studies because the concepts of suggestibility and acquiescence clearly overlap. We simply draw attention to the difficulties in using measures of suggestibility to explain people's tendencies to acquiesce in interviews. When the other types of questions are used, or when yes/no questions that suggest an incorrect negative answer (e.g., “The door wasn't open, was it?” when the door was open) are used, suggestibility can be separated from yea-saying.
The questions used in suggestibility studies also vary in how strongly they suggest an answer (Kebbell & Hatton, 1999). Although some take the form of basic yes/no questions (e.g., those used in the Gudjonsson Suggestibility Scale-Form 2), others take the form of statement questions (e.g., “The stranger knew where to find a key—yes or no?”—Perlman et al., 1994) , and others are significantly more strongly worded than those typically used in interviews or self-report scales (e.g., “The chalk fell in the floor, didn't it?”—Henry & Gudjonsson, 1999). Perlman et al. investigated different types of questions concerning a murder that participants who had mild or borderline mental retardation had seen on a video. The researchers asked a series of short answer questions (e.g., “What was blocking the door?”), yes/no questions, and statement questions. Half of these questions were either misleading or required negative answers. They found that participants were significantly more acquiescent than were control participants on the false statement questions and on misleading yes/no questions, although their performance was better for the latter (84% compared to 52% correct, respectively). The authors pointed out that the statement questions carry a “stronger assertion of veracity” (p. 183).
These studies show that higher rates of acquiescence occur more often for people who score lower on intelligence scales and when questions are more strongly suggestive. Gudjonsson and Clark (1986) proposed two further antecedents to a suggestible response—interpersonal trust (the belief that the interviewer's intentions are genuine) and expectations of success (the belief that the interviewer expects them to know and provide an answer).
Question Content and Linguistic Complexity
The concepts of suggestibility or submissiveness explain acquiescence as being the result of a combination of the dispositions or cognitive abilities of the individual, the interrogative force of questions, and the social relations within the interview. A further explanation, and one that has received little attention to date, lies in the conceptual and grammatical demands of the interview. This distinction is between a person saying yes because they trust the other person to be correct or distrust their own judgments (suggestibility and outerdirectedness), saying yes because they do not want to publicly disagree with the person (submissiveness and compliance), and saying yes because they have not understood the question, have understood the question differently to that intended by the interviewer, or because they do not know the answer. In practice, these aspects are not clearly separable, as was illustrated in the preceding discussion of suggestibility. In particular, yea-saying when one does not know the answer or when one does not understand the question is more likely if a person has an underlying tendency towards compliance or submissiveness or if the situation is one in which power relations are unequal. However, it is useful to discuss these issues separately because a range of actions can be taken to reduce acquiescence once these factors are explicated.
Gudjonsson (1990) stressed the importance of distinguishing between acquiescence as a tendency to answer questions in the affirmative and compliance as the tendency to obey instructions and requests. If yea-saying is found in the absence of behavioral compliance, as it was in Rosen et al.'s (1974) community sample, then it seems likely that acquiescence on yes/no questions might be caused by factors other than, or in addition to, a desire to please or a general submissiveness. It is important to recognize that yea-saying occurs in contexts requiring varying degrees of linguistic skill, comprehension, judgment, and knowledge. In our earlier discussion of detection techniques, we suggested that acquiescence is sometimes a methodological artifact, produced by bizarre questions, reverse wordings that are not mutually exclusive, strongly worded misleading questions, or incompatibility between informant and self-reports. Sigelman et al. (1981a) suggested a number of further reasons for acquiescence, such as it being a strategy when the question is not understood or the answer is not known or being an automatic response. Indeed, they found that acquiescence was lowest on factual questions when answers were most immediate and concrete. The literature on acquiescence in the general population suggests that people use response sets particularly when they do not know the answer to a question (Cloud & Vaughn, 1970; Cronbach, 1942, 1950). Hemdal, Corwin, and Oster (1993) found that this was the case for children and adults with Down syndrome, who were more likely to answer yes when they were uncertain in an olfactory identification task (see also Gerjuoy & Winters, 1961).
The literature on suggestibility can be used to understand yea-saying because suggestibility is measured to a large extent by yes/no questions to which the answer should be no (see earlier discussion). In the following discussion we are, therefore, taking suggestibility as measured in these studies to be an index of yea-saying rather than accepting Gudjonsson's (1990) definition (i.e., personal acceptance of messages communicated during questioning). The evidence here also supports an explanation for acquiescence in terms of the type and form of questions used. In a case study, Gudjonsson and Gunn (1982) found that a witness with mild mental retardation was suggestible in two situations, when questions contained sophisticated or abstract ideas and when she was uncertain about the details. When she was certain about the facts, she was not susceptible to leading questions. Acquiescence, then, may be a response used by some people when they are uncertain of the meaning of the question or when they understand the question but are uncertain of the answer. Investigators have found a relation between initial recall memory and suggestibility in the general population (Gudjonsson, 1983, 1988) and that adults with mental retardation are both more suggestible and have lower recall than do control participants (Clare & Gudjonsson, 1983; Everinton & Fulero, 1999; Perlman et al., 1994). People may be particularly unlikely to admit that they do not know the answer when the factors outlined earlier are also present (i.e., when the question wording is strongly suggestive, when the interviewee trusts the interviewer, and when they believe there is an expectation that they should know the answer).
If acquiescence is more likely when people are uncertain, then rates of acquiescence found in studies of suggestibility are likely to be affected by how much detail has to be remembered, how difficult it has been to encode the details (e.g., eye-witness vs. hearing a story) and how strongly worded the misleading questions are. When details are difficult to remember, and there is strong interrogative pressure, acquiescence might be found to be more of a problem than it is when details are easier for the person to produce or when the wording of the question is less forceful. In a study comparing children ages 11 to 12, with or without mental retardation, and MA-matched controls, Henry and Gudjonsson (1999) found that although suggestibility scores on the Gudjonsson Suggestibility Scale-Form 2 were related to initial memory recall, scores on an eye-witness task were not. This is likely to be due to methodological differences in the two tasks. The Gudjonsson Scale involves listening to and remembering a story, which is likely to require rapid verbal processing and encoding, creating more difficulties than an eye-witness task for those with reduced receptive verbal abilities. In addition, the questions used are only mildly suggestive (e.g., “Was the husband a bank director?”). The eye-witness task, as well as being easier to remember, involved questions that were more strongly worded than those in the Gudjonsson Suggestibility Scale-Form 2 (e.g., “The chalk fell on the floor, didn't it?”). It is likely, then, that interrogative pressure played more of a role in the eye-witness task, and verbal processing/memory problems played more of role in the Gudjonsson Scale task. The measures created different possible causes of acquiescence.
Uncertainty is not just a feature of factual questions, but is likely to be involved when the question requests judgments that are difficult for the individual (e.g., comparisons, socially reflexive questions, generalizations, frequencies), involves abstract concepts (e.g., subjective states), or asks the person to make a decision about an issue that they have never considered in detail before (e.g., some attitude statements). In these cases, the person may not know how to answer and may be unable to come up with an immediate response. These issues would be particularly important for people with mental retardation who may have greater difficulties with a range of question structures and judgments than do members of the general population (Finlay & Lyons, 2001; Matson & Frame, 1986; Prosser & Bromley, 1998) and who may have less well-differentiated articulated attitudes (Peabody, 1966).
Acquiescence may also arise from question structures that are too lengthy or complex for the individual. In the general personality literature, Peabody (1966) argued that response biases are more likely when items are grammatically complex. Investigators have found that some people with mental retardation may be unable to process more than one item or idea from a sentence, with the result that answers may refer to only a limited part of the question (Matson & Frame, 1986; Zetlin, Heriot, & Turner, 1985). This is illustrated by Shaw and Budd (1982), who conducted a study in which people were asked about the permissibility of a range of behaviors in a workshop. Using reverse-wordings of the form “Is it against the rules to——/Are you allowed to ——?” they found that yea-saying was more frequent for socially desirable behaviors, nay-saying was more frequent for undesirable behaviors, and there were significantly more contradictions for people with lower IQs. They concluded that cognitive impairment predisposes people to response biases, and social desirability then determines the direction this will take. People may not be able to fully understand the question structure, although they still respond to the topic: If it was a desirable behavior they endorsed it, if it was undesirable they rejected it, regardless of the form of the question.
Research by Heal and colleagues (reviewed by Heal & Sigelman, 1995) supports the suggestion that people may respond to the topic while not fully understanding the subtleties of question phrasing. They found that the wording of questions was an important factor in producing acquiescence in quality of life interviews. When people were asked “Do you wish you were happier with ——,” there was considerable acquiescence, often followed by the person saying “Yes, I'm happy (with ——)”. In this case, participants responded to the root form of the word rather than its modified meaning (i.e., “happy” rather than “happier”). Heal and Sigelman's study included two pairs of questions about satisfaction with aspects of life (e.g., home, job, friends) designed to detect acquiescence bias (“Are you happy/unhappy about ——?” and “Most of the time do you like/Do you wish you had a different ——?”). These questions were affirmed in both versions by about half of the respondents. Given the findings just described, inconsistency on the former pair of questions might arise if participants responded to the question “Are you unhappy——?” as if it did not contain the modifier un (“Are you [un]happy ——?”) This may be one explanation for the acquiescence found in other studies in which researchers reversed questions by simply adding modifiers. For example, Burnett (1989) reversed the wording of questions by using like/dislike and happy/unhappy. People may say yes to opposite forms of the same question because they are responding to the root form of words or the topic rather than the particular form of the question, and those who have less abilities in receptive language might be particularly vulnerable to this. In addition to modified words creating problems, sentences with modifying clauses can also create problems with understanding. When questions are too long, or the structure too complex, respondents may only focus on some of the words or phrases within the question.
The importance of cognitive skills in acquiescence is also stressed by Gudjonsson (1990), who described acquiescence as involving three stages. The first stage is listening to the question, which involves attention and interest, and the second is understanding it, involving vocabulary, general knowledge, and comprehension. If the person does not understand, then uncertainty is created, and the person may guess, give the most plausible answer, or express their uncertainty. He suggested that acquiescence functions to reduce apparent uncertainty and restore self-esteem. Acquiescence should be seen, then, as a problem of difficult or semantically complicated questions rather than as a problem of yes/no questions per se.
Although this seems a powerful explanation in certain cases of acquiescence, two further developments of the model are needed. Uncertainty can also arise because the person understands the question but does not know the answer, has no firm opinion about it, or is unable to abstract specific instances to make a general judgment (Zetlin et al., 1985). In these cases, answering in the affirmative may be a default. Uncertainty can arise, then, without the difficulties in comprehension Gudjonsson suggested. The second addition to the model is based on the understanding that problems in comprehension are not always recognized by the interviewee. At Stage 2, the person may not experience uncertainty, but may simply respond to the question or subject as they perceive it. If this is different to that intended by the interviewer, then apparent acquiescence might arise in the absence of any uncertainty on the part of the person with mental retardation. Gudjonsson may have, therefore, overemphasized the role of uncertainty and doubt because misunderstandings might not be noticed by the respondent and answers, therefore, were given with some confidence.
In contrast to these explanations, Rapley and Antaki (1996) suggested that the appearance of acquiescence is not necessarily due to the disposition or competence of the interviewee (i.e., an eagerness to please or a lack of understanding). They rejected the idea that the behavior taken to be acquiescent is necessarily due to the submissiveness of people with mental retardation and presented extracts from quality of life interviews that illustrate how apparent inconsistencies and yea-saying can arise as a result of the behavior of the interviewer. For example, this can occur when the interviewer does not accept the initial response because the full question has not yet been read out, because the response does not fit one of the predetermined response categories, or because the interviewer believes he or she knows the right answer and tries to shepherd the respondent towards it. In addition, respondents may use yes to indicate they have understood the question (a back-channel response), and these may be misinterpreted as acquiescent responses. Rapley (1995) suggested that when people change their initial answers in response to multiple choice questions, this can be a conversational strategy adopted in the context of a power asymmetry, particularly when one's initial answer has not been accepted.
One final factor that is important, particularly in relation to acquiescence in false confessions, is that people with mental retardation may not be fully aware of the consequences of acquiescence or confabulation in such situations (Clare & Gudjonsson, 1995; Gudjonsson & MacKeith, 1994).
Apparent acquiescence in interviews can be seen to occur, then, for a number of reasons other than, or in addition to, the participant wanting to agree with the interviewer. These other explanations include that acquiescence is a response strategy when the answer is not known; that it arises when the question structure is too complex for the person, so that the subtleties of phrasing are missed; that it is a methodological artifact of reverse wordings, bizarre questions, or informant checks; and that it arises due to the interactional demands of the interview situation.
Strategies for Reducing Acquiescence in Interviews
There are a range of measures that researchers and professionals can take to combat yea-saying during scale or interview development, during the interview itself, and during analysis. These include using either/or instead of yes/no questions, employing reverse-wordings, developing balanced scales, using content validity procedures during item development, including “don't know” options, simplifying question wording, being aware of which types of judgment may be difficult, asking for examples, and employing statistical corrections. The advantages and disadvantages of these measures are described later. Because the focus of this article is on interview design and procedure, the suggestions do not concern longer term interventions addressing acquiescence as an underlying disposition or a learned behavior nor situations other than those in which verbal questions are used.
Sigelman and her colleagues recommended that yes/no questions should not be used at all (see also Cronbach, 1950) and that either/or questions should be used in preference (Sigelman et al., 1981a, 1981b). This is also recommended in the general social survey research literature (e.g., Cronbach, 1950; Javeline, 1999; Schuman & Presser, 1981). The use of either/or questions creates further problems, however, because such questions are longer and more complex in their phrasing and are, therefore, more difficult to understand. They are unlikely to comply with the recommendation that questions used with this population should be short and simply phrased (e.g., Prosser & Bromley, 1998). For example, as an alternative to yes/no questions, Wehmeyer (1994) used an either/or format in a measure of locus of control. This format had a low test–retest reliability and the author suggested that people found it difficult to understand the questions. Either/or questions are more successful when they involve short or single-word options, although in these cases a last option bias has also been found (Sigelman et al., 1981a, 1981b). If either/or questions are to be used, the options should be short, the items rigorously pilot-tested, and checks included to ensure the interviewee has remembered and understood both options.
If yes/no questions are used, Sigelman et al. (1981a, 1981b) suggested using acquiescence tests, such as reverse wordings, and excluding those who fail. Many researchers now use such screening tests (e.g., Burnett, 1989; Helsel & Matson, 1988; Irene & Kirby, 1990). Because acquiescence is likely to be differentially influential across tests (Ray, 1983), depending on the complexity of both the content and the grammar of items, it is important that any such screening tests are similar in both content and language to the tests that are to be used. The use of techniques such as nonsense questions does not conform to this requirement.
It is, however, usually preferable to exclude as few people as possible, and this can be addressed during the design and administration stages of interviews and self-report scales. Instead of identifying particular people as acquiescent and excluding them from participation, then, we should concentrate on improving question design and interviewing procedures. This can be done by reducing sentence complexity, asking for examples, avoiding the assumption that people can provide immediate answers to all the questions we might ask, and using balanced scales.
Researchers sometimes include reverse wordings of all either/or or yes/no questions at other points in the interview in order to check for consistency (e.g., Shanly & Rose, 1993). Prompts and alternative wordings should be scripted and developed with as much care as the initial items. Antaki (1998) illustrated the difficulties created when interviewers come up with explanations and rewordings spontaneously. Using transcripts of quality of life interviews, he found that, in paraphrasing, interviewers often restricted the meaning of questions to limited activities or events and in doing so “lowered the bar” at which a positive response would be scored. An additional consideration is how interviewees interpret being asked the same question repeatedly in different ways. In order to minimize the risk that they assume they have answered wrongly the first time, the interviewer should explain at the beginning and at points during the interview that this is part of the procedure and does not mean their answers are wrong. If a person contradicts themselves and the interviewer is unable to resolve this, then all answers should be recorded and consultation with other professionals or informants should be carried out later. This is better than making an on-the-spot, arbitrary judgment. The use of tape-recordings and transcriptions are useful in this respect. In many cases, because people's judgments and opinions may not be consistent and unified, it may not be possible to score an answer.
It is also important to include a balanced number of positively and negatively keyed items when constructing a scale (Block, 1965; Cloud & Vaughn, 1970; Peabody, 1966). Negatively keyed items are not simply reverse wordings of positively keyed items, but, rather, items that contraindicate the quality being measured (e.g., “I am happy” in a depression scale). Balanced scales compensate for acquiescence because the tendency should be equal on average over both sets of items, and they allow a scale-specific measure of acquiescence (Ray, 1983). Although this procedure has been recommended for a number of years, many new scales do not conform to this requirement. One reason is that it is often difficult to come up with negatively keyed items that are not complicated in their structure or vocabulary. For example, including a negative (e.g., no, not) in a positive sentence acts as a modifier, and such sentences have been found to be more difficult to respond to by individuals in the general population (Gough, 1965; Slobin, 1966; see review in Barnette, 2000) as well as for people with mental retardation (Wehmeyer, 1994). A good example of a scale that includes both types of items is the self-esteem scale of Szivos-Bach (1993). Negatively keyed items in this scale avoid the use of negative modifiers (e.g., “I forget things/I give up easily/I am slow at work”). The use of balanced scales does not mean, however, that other techniques of minimizing acquiescence can be overlooked. Although such scales compensate, on average, for acquiescence, being useful when studying group differences and correlations, there will be variation for each individual in how much it has affected positively as opposed to negatively keyed items (Javeline, 1999). When the scale does not include a large number of items and one is interested in an individual's score, for diagnostic or other reasons, then it is important to use other techniques, such as asking for examples.
Another strategy used in the item-refinement stage of scale development in the general population is inclusion of a measure of acquiescence response set and removal of those items found to be highly correlated with the acquiescence measure (Winkler et al., 1982). This strategy relies on the existence of a good measure of acquiescence that is relevant to the scale in question. In Winkler's study this required that good item reversals be possible. The difficulties in achieving the latter have already been discussed. Winkler et al. was able to achieve a scale that was less influenced by acquiescence through this method and that was more effective than simply excluding the most acquiescent 5% of the sample. Although this acquiescence scale has been used with people who have mild mental retardation (Clare & Gudjonsson, 1993), it is unlikely to be widely applicable due the complexity of both the sentences and the content (attitudes towards health care delivery). In the absence of good acquiescence measures relevant to the type of scale under development, a balanced set of items allows a measure of acquiescence (Ray, 1983), and those items highly correlated with this measure can be eliminated from the scale. It is likely that problematic items are those that are grammatically complex, ambiguous, or require judgments that people with mental retardation find difficult to make (for a review, see Finlay & Lyons, 2001).
Because acquiescence is likely to occur when the demands of the questions exceed the person's linguistic abilities, it is important to make sure that question wording is simple and clear so that the interviewee is less likely to have focused on only part of the meaning (i.e., avoid modifiers). Following the procedures of content validity (Haynes, Richard, & Cuban, 1995), questions should be developed from pilot work with people who have mental retardation and expert panels so that it is ascertained beforehand that content and grammar are easily understood. In order to further ensure that yes/no questions have been understood as intended, it is advisable to check for the interviewee's understanding of the question by asking for examples (e.g., Smyley & Ellsworth, 1997; Szivos-Bach, 1993; Zetlin et al., 1985; for multiple choice formats see Clare & Gudjonsson, 1995, and Smith & McCarthy, 1996). Zetlin et al., using self-concept scales, found that 70% of responses could be scored into a yes/no format when probing for understanding was carried out. Their sample of 46 people included individuals with mild, moderate, or severe mental retardation. Although this strategy is useful for people with adequate expressive abilities, some people will have receptive abilities sufficient for understanding the question but will not have the expressive abilities to give examples (see Prouty & Strummer, 1994). In this case, further, differently worded closed questions can be asked to check for consistency.
Because acquiescence is more likely to arise when people are uncertain of the answer to a question (Cronbach, 1950; Sigelman et al., 1981a), scales should allow “don't know” options as well as simply yes and no. This option should be included not only for factual questions, but also for those dealing with attitudes, subjective or abstract concepts, and judgments involving comparisons, frequencies, and generalizations. People may be uncertain about the answer to a question if they have never thought about the issue before, if it requires them to make a complex judgment, if it refers to concepts they are unsure about, or if the grammar is too complex. Interviewers should stress at various points throughout the interview the importance of saying “don't know” if this is the case.
A method often used in the general literature is to take measures of acquiescence for individuals and correct their scores using a variety of statistical procedures (e.g., Cronbach, 1950; Hofstee, Ten Berge, & Hendricks, 1998). This technique was used by Heal and Chadsey-Rusch (1985) with people who have mental retardation to correct scores on a preexisting measure of lifestyle satisfaction. The measure of response bias used involved reverse wordings of eight items, with the associated problems of interpretation. A debate on the merits of different statistical procedures is beyond our scope in this article; readers are referred to Hofstee et al. (1998). When this technique is used, the same considerations apply as those used for screening tests (i.e., that the items used for the correction should be similar in both grammatical complexity and content to the test items themselves).
Researchers have found that open-ended, general questions, although producing fewer details, lead to a smaller proportion of errors than do specific or yes/no questions (e.g., Dent, 1986; Perlman et al., 1994). For eyewitness testimony, there is now some evidence that the use of cognitive interviews (where the same incident is approached using a variety of open questions) can increase recall (for a review, see Kebbell & Hatton, 1999). Tully and Cahill (1984), in their study of witness questioning, recommended the use of two interviews, the first for free recall employing carefully worded, open questions. The second interview should include different question wordings to check the information from the previous interview. They also suggested a preinterview during which questions concerning background information the interviewer already knows (from informants) are asked, in order to find out how the person exhibits uncertainty, how easy it is to get them to agree to misleading yes/no questions, and to show that it is all right to disagree or to say they are not sure. The importance of accuracy should be continually stressed.
Finally, there are considerations relating to the social context of the interview. These apply to a wider range of questions than simply those with yes/no formats, such as when criticisms of existing supports or relationships might be implied or when sanctions may be feared by the person with mental retardation. The interviewer should clearly state that a don't know response is acceptable and that he or she does not expect a positive or negative response to every question. Interviewers should be sensitive to the ways in which the word yes can be used in interaction (e.g., to indicate the question has been heard or understood) and the dangers of implicit or explicit shepherding of the interviewee into certain responses because of the interviewer's preexisting knowledge or assumptions about the other person. When the purpose of the interview is to obtain information about satisfaction with existing arrangements or other potentially sensitive issues, efforts should be made to involve interviewers who are not involved in the person's support. This may conflict with the benefit of using interviewers who have an existing rapport with the person and who may be familiar with their communication styles, in which case careful consideration needs to be given to the costs and benefits of different interviewers. In either case, reassurance should be given as to whom the information will and will not be shared with and what the possible consequences will be.
The recommendations given in this paper are based on a clearer conceptual understanding of the causes of acquiescence, and our purpose is to increase both the validity and inclusiveness of self-report scales and interviews with people who have mental retardation. The strategies are aimed at scale developers as well as those who are involved in interviewing people with mental retardation and should be useful in clinical assessment, service evaluation and planning, and forensic contexts. Although there may be individual difference factors that predispose people to yea-saying (e.g., cognitive and linguistic abilities and personality factors), aspects of the interview situation are also important. Yea-saying is more likely to be found in situations where people do not have an opinion, when the question is ambiguous, when they are uncertain, and when the question structure is complex. Much can, therefore, be done to reduce acquiescence in interviews, both at the design, administration, and analysis stages. This includes simplifying sentence structures, recognizing the type of content that makes acquiescence more likely, allowing the expression of uncertainty, and checking answers. We note that many people with mental retardation have no such problems with answering questions; these recommendations are aimed at increasing the inclusiveness of interviews with this population so that more people can have their voices heard.
More research is needed on techniques of improving interviewing with people who have mental retardation. In this review we have focused on yea-saying, but acquiescence in a more general sense might affect other question forms, such as when pictorial cues or scales are offered or when the interviewee assumes the interviewer favors a particular response. Although not the focus of this paper, interactions in interviews can be affected by the relationship between the interviewer and the interviewee, the respective institutional affiliations of the participants, the power dynamics, and the interviewees' understandings of the purpose and confidentiality of the interviews. These issues are particularly important in the context of services in which participants may feel a greater or lesser degree of personal control and, further, fine-grained research, such as that carried out by Antaki and Rapley (Antaki, 1998; Antaki & Rapley, 1996; Rapley & Antaki, 1996), is needed to examine these issues. In addition to this, if we are to understand the importance of underlying personality dispositions or learned behaviors, such as submissiveness, and the extent to which they have effects across a range of contexts, then clearer measures of such concepts are needed. The problems with the concept of suggestibility illustrates that measures need to clearly distinguish between the disposition itself and behaviors that might also be the result of linguistic factors and comprehension.
Scales need to be developed with more consideration of the receptive linguistic abilities of those they are designed for as well as the conceptual problems people may have with certain question types. For the former purpose, it would be useful to have a generally recognized measure of receptive linguistic abilities that could indicate what levels of grammatical complexity people can understand in questions. Measures of verbal IQ or vocabulary are too crude for this. Such a measure would allow scale developers to develop items with particular populations in mind and would specify clear criteria as to the population for whom the test is valid. In addition, more work needs to be carried out in specifying which types of judgments are particularly difficult for people and how to ascertain when this is the case. Asking for explanations or examples is one solution to the latter, but scale developers too often assume that they can ask any type of question without acknowledging that some judgments are much more difficult to make than others.
NOTE: We recognize that the phrase “people with mental retardation” is objectionable to many individuals and use it here reluctantly, only for reasons of clarity for an international audience. In some countries outside the United States, this term is not used. The same population is referred to as “people with learning disabilities” or “people with learning difficulties” in the United Kingdom and “people with intellectual disabilities” in Australia and New Zealand.
Authors: W. M. L. Finlay, PhD, and E. Lyons, PhD, Social Psychology European Research Institute (SPERI), Department of Psychology, University of Surrey, Guildford, Surrey GU2 5XH, England, United Kingdom. firstname.lastname@example.org