Likert-type scales are useful when you are measuring latent constructs - that is, characteristics of people such as attitudes, feelings, opinions, etc. Latent constructs are generally thought of as unobservable individual characteristics (meaning that there is no concrete, objective measurement) that are believed to exist and cause variations in behavior (e.g., answer questions on a scale).

Usually, Likert-type scales use statements ("Please rate the extent to which you agree/disagree with the following:") and use 5 or 7-point response scales (most commonly). The response scales use anchors such as 1 = Strongly Disagree, 2 = Disagree, 3 = Neutral, 4 = Agree, 5 = Strongly Agree, etc.

The items should be phrased in a way that only poses one characteristic per item so that it is clear what the person is responding to. For example, "I think politicians are honest and helpful" is not a good item because you are asking about two separate issues (honesty and helpfulness). Also, try to avoid using the word "not" or other negatives directly in these items as it can become confusing about what it means to disagree with a negative.

Likert-type scales can be scored in a variety of ways. Typically, you would score each item so that higher scores always indicate "more" of some characteristic and then take the mean (average) of all of the items. Remember, the numbers will not have any inherent meaning (e.g., if you are measuring attitudes about politicians, scoring a 3.4 doesn't REALLY mean anything except that, on average, that individual was slightly favorable in his/her attitudes and can then be compared to the distribution of the remaining responses).

Finally, you always want to check the reliability of Likert-type scales using Cronbach's Alpha (internal consistency). In general, you want values of .7 and up (the ceiling is 1.0) for good internal consistency. Mathematically, internal consistency is the average of all possible split-half correlations. Conceptually, it measures how welly the items function together (e.g., do people respond consistently with their standing on the construct of interest).

You might want to check out one of the Sage Publications green books on surveys called "Survey Questions" by Converse and Presser. It has a lot of good info on building different types of surveys.

Hope that helps!

Open-Ended Questions

ask participants to come up with their own responses and allow the researcher to document the opinions of the respondent in his or her own words. These types of questions are useful for obtaining in-depth information on facts with which the researcher is not very familiar, opinions, attitudes and suggestions, or sensitive issues. Completely open-ended questions allow the researcher to probe more deeply into issues, thus providing new insights, bringing to light new examples or illustrations, and allowing for different interpretations and a variety of responses. Researchers who utilize open-ended questions must be skilled interviewers since they need to record all information to avoid loss of important information, and the analysis is time-consuming. In addition, open-ended questions can be difficult to analyze statistically because the data is not uniform and must be coded in some manner.

Close-Ended Question

Closed questions have a list of possible answers from which respondents must choose. They can include yes/no questions, true/false questions, multiple choice, or scaled questions. Closed questions can be categorized into 5 different types:(5)

Multiple Choice- this question type is useful when the researcher would like participants to select the most relevant response.

Likert Scale- this question type is appropriate when the researcher would like to identify how respondents feel about a certain issue. The scale typically ranges from extremely not important, not important, neutral, important, to extremely important, or strongly disagree, disagree, neutral, agree to strongly agree.

Numerical- these questions are used when possible responses are numeric in form. For example, these questions are useful for asking someone’s age.

Ordinal- these questions are useful when participants are asked to rank a series of responses.

Categorical- this question type is appropriate when respondents are asked to identify themselves into a specific category. For example, they may be asked if they are male or female.

Closed questions are commonly used for obtaining data on background information such as age, marital status, or education. Closed questions may also be used to assess a respondent’s opinions or attitudes by choosing a rating on a scale. Additionally, closed questions may be used to elicit specific information in an efficient manner.

Likert-type scales are useful when you are measuring latent constructs - that is, characteristics of people such as attitudes, feelings, opinions, etc. Latent constructs are generally thought of as unobservable individual characteristics (meaning that there is no concrete, objective measurement) that are believed to exist and cause variations in behavior (e.g., answer questions on a scale).

Usually, Likert-type scales use statements ("Please rate the extent to which you agree/disagree with the following:") and use 5 or 7-point response scales (most commonly). The response scales use anchors such as 1 = Strongly Disagree, 2 = Disagree, 3 = Neutral, 4 = Agree, 5 = Strongly Agree, etc.

The items should be phrased in a way that only poses one characteristic per item so that it is clear what the person is responding to. For example, "I think politicians are honest and helpful" is not a good item because you are asking about two separate issues (honesty and helpfulness). Also, try to avoid using the word "not" or other negatives directly in these items as it can become confusing about what it means to disagree with a negative.

Likert-type scales can be scored in a variety of ways. Typically, you would score each item so that higher scores always indicate "more" of some characteristic and then take the mean (average) of all of the items. Remember, the numbers will not have any inherent meaning (e.g., if you are measuring attitudes about politicians, scoring a 3.4 doesn't REALLY mean anything except that, on average, that individual was slightly favorable in his/her attitudes and can then be compared to the distribution of the remaining responses).

Finally, you always want to check the reliability of Likert-type scales using Cronbach's Alpha (internal consistency). In general, you want values of .7 and up (the ceiling is 1.0) for good internal consistency. Mathematically, internal consistency is the average of all possible split-half correlations. Conceptually, it measures how welly the items function together (e.g., do people respond consistently with their standing on the construct of interest).

You might want to check out one of the Sage Publications green books on surveys called "Survey Questions" by Converse and Presser. It has a lot of good info on building different types of surveys.

Hope that helps!

One great source for scale development or item construction might be the vast research conducted by Jon Krosnick (you can download most of his papers for free from his website: https://pprg.stanford.edu/krosnick-publications/).

In general, you should also consider what kind of data analysis you want to carry out. Every statistical analysis can only be interpreted in a meaningful way if the level of measurement appropriate (e.g., interval scale for t-Test).

There's lot of discussion about Likert- or Likert-type scales. E.g., whether construct-specific wording should be preferred. Whether they are just measuring on an ordinal level or on a interval level. The only way to be empirically sure about that would be to test the distances of each answer category by applying appropriate item response theory models.

Likert scale could be used in theory of Planed Behavior. Previous answers exemplify very well. Then you can categorize and analyse.

In order to analyse the data, questions like yes or not can be categorize in a numeric scale 1 and 2 and after you can use a quantitative approach. The same with a Likert scale.

Best wishes

Likert type scale are very useful when you would like to measure the "intensity" of an opinion. You can ask to a child if he likes the icecream: the answer can be yes/no. However, you can ask the children to evaluate how much he likes the icecream: you don't like it at all, you don't like it somewhat, neither you dislike nor you like it, you like it somewhat, you like it very much. In the second case you use a Likert type scale. In fact, the use of the 2 way close ended question and a Likert type question is a matter of what you wnat to measure.

Hope this clarifies

Shafig Al-Haddad makes an important point that there is often a confusion between using a Likert-style response format (e.g., from strong agree to strong disagree) for a single item, versus using a number of items scored in this fashion to create a scale.

Technically speaking, a single item with this scoring is ordinal, and thus a poor choice for the kind of statistical analyses that require a numerical or interval level variable (e.g., as the dependent variable in a t-Test).

In contrast, summing together several Likert style variables to create a scale can produce an interval-level variable. (Although, as Brendon Morse points out, you need to check the reliability of such a scale, usually with a Cronbach's alpa.)

Also, going back to your original question about the comparison to yes/no variables, these can also be summed into a scale, but you need more of them (e.g., the total of two such items can only vary between 0-2). One common use for that kind of scale is when you want to count something, and the yes/no responses (e.g., to tell you how how often the respondent experienced various things that all fall into the same domain).

Fun fact: I went to the University of Michigan where Likert taught, and everyone there pronounced his name "Lick-ert" (not Like-ert).

there are different ways to analyze the Likert scale data.

either you can take average of all question and then compare it among different groups by using t test or ANOVA. or you can report individual question one by one by giving the frequency of respondent who gave different responses.

@ Brendan Morse Thanks Sir for the insight. Also for suggesting a book for further reference :)

Thanks @Frank Schneider for the link It can be of great help

I hope this is not too late for you, and it might not align with what you are doing, however, it is a timely reminder that questions are important as well as the measures we use. From my experience, something that is important with questionnaires, is the understand-ability or even acceptability of the question sitting with the scale you wish respondents to use. I observed a number of years ago that some questions are more "understandable" than others.

The Likert Scale like many scales allows you to, with simple coding of the responses, to create histograms of the responses.

While you will often get non normal responses from surveys of even large samples. you get, in my limited experience, extreme scatter of responses to the frustration of a poorly worded question.

This requires a lot more investigation but helps you avoid taking strong views from what might be an artefact of the question. There is limited literature on this issue.

This begs the question (pun not intended :) ) how do we use multiple instruments when measuring a critical component in any research? There are disciplines that do this but there is much research areas that do not..

My apologies for being a little off topic.

The Likert scale is commonly used in public health evaluation. For example, they can be used when evaluating a partnership, conducting a needs assessment about which policies are most pressing in a community, or assessing the public’s knowledge and awareness of a public health campaign. Similarly, the Likert scale is a valuable and important part of survey research, which is commonly used in public health evaluation.

A Likert scale is an ordered scale from which respondents choose one option that best aligns with their view. It is often used to measure respondents' attitudes by asking the extent to which they agree or disagree with a particular question or statement. A typical scale might be “Strongly disagree, Disagree, Neutral, Agree, Strongly agree.” Likert scales may meet your needs when you have attitude, belief, or behavior items. For example, you would not use a Likert scale to assess attributes, such as age, race, and income, but you may use a Likert scale to assess someone’s attitude about a particular topic.

Hi,

I fully agree with my dear colleague Senthil Vel ·

Yours

Maria

Likert scale is commonly used in medical imaging research in order  to assess the  quality of medical images for example mammograms  perceptually. 

What do you then use to measure knowledge questions? For instance if the research wants to know the knowledge of a certain group of people about nutrition? Do you use yes/no/ i don't know? or true or false responses?

I would appreciate some clarification about this.

Thank you.

Shirley, measuring knowledge is pretty much what every teacher has to do with every test, so think in terms of writing test items.  If you have a lot of responses, you won't want to "grade" open responses, so either true/false or multiple choice design tends to work best.  True/false will work if you have knowledge points (statements) that are fairly clear cut, but you get better psychometric properties in your scale if you have multiple choice items (which is why most standardized cognitive tests use them).

Hi Shirley,

My thoughts; a simple test of knowledge asked directly, would be better in many cases.

Self reporting of knowledge is probably not very reliable, except perhaps in technical area where the knowledge boundaries are known by the people you interview e.g. software writers know what they know and don't know because of the demarcation between "trades"/software languages..

So ask a few simple question to test knowledge directly would be ideal in an ideal world.

Self reporting on a complex issue like nutrition seems risky unless you have specific requirements that direct you towards that approach.

I hope that is useful.  Vote me up or down. below i like to know if I make sense :)

best of luck Murray

its really valuable information 

The answer of Brendan is highly interesting and useful. the Likert scale is used when measuring attitudes and opinions. on the other hand, yes/ no scale can be used where specific and accurate answer is measured. for example: did you ever travel to Spain? there is no any other answer than saying yes or no etc. 

Sorry, 

What do you do to calculate knowledge score? For instance if the research wants to know the knowledge of a certain group of people about toxoplasmosis ( the final step to present the data?

I would appreciate some clarification about this.

Thank you.

Many thanks for this information

Thanks for the interesting information.

One useful technique I have used with Likert scale questions is to plot a histogram  of all your responses.

If you see a scatter of responses rather than a bell type curve sitting with or without skew around a point. then you may have a problem, either the issue doesn't matter to them (I.e. level of engagement is low., or the wording is frustrating respondents and they are tending to just pick a point any point on the scale.

This is a very important point and in large surveys makes worthwhile, using pretesting.  Because, questions that require too much cognitive effort/are not resolvable in the respondents mind, give these scatter responses.

Hope that helps you see why Brendan, made the point of taking care with the wording of questions. It gives you a diagnostic to so small degree. 

Murray

Atheer Ibadi, you may want to look into health literacy measurements to evaluate knowledge. Generic (e.g., TOFHLA) and specific measures exist. 

 I  am thinking to suggest the researcher to make this  at  three point scale ( Awareness High = 2, Awareness Low/Poor = 1, Awareness No = 0  then going for calculation , & also try with reliability on this Three point scale Cronbach's Alpha test. Hope this is doable . How do you all suggest ???

Dear Nazneen!!! your opinion is valuable.

What about to make five point scale (None=0. Low=1, Moderate,=2 High=3, Very high=4) for measuring the awareness level of respondents and then try for Reliability on this five point  scale by Cronbach's Alpha test. what do you think???

Thanks all for the clarification on likert scale..

Dear Chopra

The colleagues Brendan J. Morse ·and others before did a good explanation of the subject.

Best Wishes!

Brilliant explanation

its really valuable information

Thanks for all researchers it's really valuable information, each one gave his opinion and experience. At the end, very good group to exchange the information.

Type of test needed to be decided first according to the objectives of the study.

Then, interpretation is according to the output of that test.

Wish to thank each one of you for your worthy and valuable inputs on Likert scale.

Thanks again :)

If you ask a question like to what extent have you used the Internet to promote you group, where 1 is always and 5, never or how likely are you willing to discuss your group on the Internet, where 1 is very likely and 5, not likely. is this a likert-type question? If yes, how do you code it on spss?

Dear Emmanuel,

if I properly understand your question, I would say that you are dealing with a typical Likert type question where you measure the intensity of use of internet according a 5 level scale.

I am not an expert in SPSS (I use Stata) but I guess that you should simply enter the value for each observation (i.e. if the respondent 1 answer 4 you will record 4, if the respondent 2 answered 5 you will enter 5 and so on so forth).

Hope this helps.

Best regards

Andrea

THANK you for your quick replies. On SPSS, your label can be coded with values. for example if respondent 1 ticks 1, it is coded as Always or very likely as the case may be. but if it is 2, 3 or 4, how can you set the values. most videos on likert scale questions tend to only focus on 'strongly disagree, agree. neutral, agree, strongly disagree' examples!

Emmanuel, I think your question type is suitable for semantic differential scales. I'm not an expert yet as I'm still learning too. Try searching for articles on semantic differential scales, you'll surely be satisfied.

Thanks worthy researchers for your contribution and additions to my knowledge.

Really helpful and praiseworthy opinions

Many researchers have abused the use of Likert scale I should say. Many researchers are used or limited to the agreement Likert type of scale and yet there are other types of Likert scale e.g. the likelihood type of scale. Try to read other types then you will understand which one to use and when.