The Use of the DK Option
Inclusion of a DK option in attitude scales can be problematic when a Likert scale is used because its meaning overlaps with the middle option. As noted above, depending on the label used, respondents interpret the middle option differently, including as a response reflecting an absence of agreement or disagreement with a statement (substantive response) and as a lack of a response (non-substantive response) (Klopfer & Madden, 1980). The former interpretation is consistent with Likert’s intention, a neutral attitude toward the object reflected by the item, and the latter justifies selection of a DK response choice.
Illustrative of this potential confusion, researchers found respondents with clearly defined attitudes who initially provided DK responses to questions related to those attitudes (Gilljam & Granberg, 1993; O’Muircheartaigh, Krosnick, & Helic, 2000). Gilljam and Granberg (1993) found Most (80%) of those subjects indicating DK to the first question took a position on the subsequent two questions, and the positions they took were predictive of subsequent behavior in responding to three questions related to a nuclear referendum. In contrast, Feick (1989) reported DK responses are provided by truly uninformed respondents as well as by respondents providing equivocal responses to attitudinal questions. Schuman and Presser (1996) reported that 30% of respondents would provide a substantive opinion on a law they know nothing about if not given an explicit DK option.
While Alwin and Krosnick (1991) found that allowing DK responses had no effect on the reliability of attitude survey questions, Schuman and Presser (1996) reported mixed findings with respect to the impact of inclusion of the DK option on the correlation between items. They reported that in addition to attitude intensity, education and knowledge levels also affected the use of the DK response, and those with less developed or weaker attitudes used the DK option more often. Francis and Busch (1975) also found increased use of DK with lower educational and SES levels.
These findings raise the question of what DK means. Does a DK response mean a lack of access to information required to formulate and provide a response, uncertainty or a neutral response, or unwillingness to provide a response? Is it interpreted differently by informed and uninformed respondents? Without disentangling the meaning of the DK and the middle-option response, the validity of the research findings is threatened.
In a study with University students (Lam, Allen, & Green, 2011), the authors examined how respondents with (informed) and without (uninformed) the necessary knowledge to form an opinion utilized Neutral and DK response options when presented alone or together and when placement of the option varied—placed as a mid-point or at the end of the scale. Five findings are notable. First, both informed and uninformed respondents used the DK and Neutral anchors interchangeably, implying a blurred semantic distinction between the two terms as suggested by the literature.
Second, when both the DK and Neutral options were provided , uninformed respondents chose the DK more often than the Neutral, and, the proportion of DK response for the uninformed group was much higher than for the informed group. This finding suggests higher validity of responses when both DK and Neutral options are provided than when only one of the two options is provided.
Third, when both the DK and Neutral options were provided , both informed and uninformed respondents selected the DK option more than the Neutral option when either one of these two options was located in the middle of the rating scale. These results are consistent with earlier findings of Krosnick (1999) and Schuman and Presser (1996) who observed greater use of the nonresponse option when located earlier in the response scale; i.e., a primacy effect for options presented in a written format. The meaning of the DK response option is apparent when it is placed physically as the mid-point, and it is chosen as the appropriate response. This finding suggests a placement effect.
Fourth, when offered either a DK or Neutral option but not both, when presented in the middle of the rating scale, both informed and uninformed participants chose that option more often than when presented at the end of the scale. This finding further supports a placement effect.
Fifth, if either a DK or Neutral option was provided, for the uninformed respondents, when the Neutral option was placed at the end of the response scale (with an extra space separating it from the other rating options), they were more likely to interpret it as a non-substantive response option than when the option was placed in the middle of the scale. However, the informed respondents tended to choose either the neutral or DK option more often when placed in the middle rather than at the end of the response scale. This observation implies that informed respondents may rely on placement as a satisficing strategy when they do have the necessary knowledge to respond to an item, but perhaps lack sufficient attitude strength to reply meaningfully (Schuman & Presser, 1996). The uninformed respondents lacked the knowledge to provide substantive responses, meaning they did not need a satisficing strategy. Instead, they simply looked for a way to indicate an absence of attitude or evaluative judgment or opinion by selecting a no response option.