There are many scales for evaluating resposes, one of them is the Likert Scale which consisits of 5 or 4 grades. There are scales, such as semantic differentials that reaches up to 11 grades. To me, I believe Likert scale of 5 grades is good.

@ Mahfuz

Thanks, Can you please pass some information on those similar scales...

Thank you editor for restructuring the question

Likert scale can satisfactorily evaluate responses compared to other similar kinds s/a Semantic differentials, Bogardus scale, Guttman scale etc.

Using a graphical measurement scale may reduce the effects of ELOC 

@ Surbala, Thanks for your inputs on Semantic differentials is Ok between biploar decisions,Bogardus scale is a social distance scale (In / Outsider),

• Guttman scale is a one diemensional, let me look into it....

Please throw some more information on how to use graphical measurements to reduce or eliminate ELOC

@Krishnan, I use a 5 point Likert scale to mean:

1 = never (agree); 2 = rarely; 3 = sometimes; 4 = often and 5 = very often.  This helps me avoid respondents 'sitting on the fence' as you have described, Krishnan.  Thanks.

I think it is difficult to have any universal scale for all situations. Likert has been found to be a better option in a number of situations. However it is necessary to experiment with different scales and options during the pilot stage and come to a satisfactory scale for the specified question.

Please refer to some useful links with likert sclaes.

http://www.ccsr.ac.uk/qmss/seminars/2009-12-03/documents/dusa.pdf

http://business.nmsu.edu/~mhyman/M610_Articles/Albaum_JMRS_1997.pdf?origin=publication_detail

http://ro.uow.edu.au/cgi/viewcontent.cgi?article=1821&context=commpapers

I totally agree with respected Dr. Rao.

I agree with @miranda define lickert scale of 5 is good. But need to see demographic behavior or cultural custom. As people in south asia tends to choose middle number as 3. Too many scale tends to confuse people during questioner especially for those comin from middle to low level of economy.

Inputs of @Nageswara is well taken....@Miranda, point 1 as NEVER is a good customized adoption, there the mid points gets missed...

As @Dony says people normally take a 50:50 position ie., at point 3, this may be

The point 3 on application of "Central Tendencies" have a higher impact on negating the  NEVER, SOMETIMES etc., Therefore it necessitates the requirement of some clarification points which can spread, @Mahfuz suggested semantics scales, which takes more number of columns.... as we discuss we make get more points on wearing a CRITICAL THINKING CAP....

Likert scaling assumes that distances on each item are equal. Five ordered response level seem to be the most common type.

Likert scales may be subject to distortion from several causes. For instance, respondents may avoid using extreme response categories (central tendency bias).

Designing a scale with balanced keying (an equal number of positive and negative statements) can obviate the problem of acquiescence bias. 

I have attached a few relevant ink which discuss the effect of Locus of control: 

http://wiki.answers.com/Q/Likert_scale_questionnaire_for_measuring_internal_and_external_locus_of_control_by_Julian_Rotter

http://eric.ed.gov/?id=ED120214

@ Ismat Beg answered (transferred from Mail box)

21 minutes ago

No. It can be further refined depending on the situation.

I think 7 points are too many for participants, and the two points you are going to add are not equal to the other 5 points. Maybe you can change the scale format in an online form . for example,  you first ask students to choose whether the item is undecided or applicable, then give them 5-point choice if their pre-choice is applicable. online questionnaire can fulfil  the condition control well.

There is a proverb "If you want to make somebody puzzled, give her/him more choices". Therefore, i Think 5 choices are good.

@ Mahfuz

Your point is well taken, but to articulate.....

The Choice of YES or No is a decision.

In the fields like Sociology, Psychology, Human Resources, Marketing, etc., where the study pertains to deal with PEOPLE with the Feelings, Involvement, participation, acceptance etc., ....

However when we collect data the participants CANNOT ALWAYS be in a position to decide. Hence, it is required to give them options, else it would look "autocratic", and may not yield them to participate further....

5 Choices within the limits of YES and NO is OK, but for some people

may also be there options,

Taking marketing,  in Customer Relationship Management (CRM) it is very essential to catch all the fine feelings of customer to make them a POTENTIAL BUYER and latter convert them to be a CONSUMER / End User.

I am very familiar with the issue of odd vs. even (midpoint or no midpoint).  I invite interested parties to read two articles:

Guy, Rebecca F. and Melissa Norvell (1977), "The Neutral Point on a Likert Scale," Journal of Psychology, 95, 199-204.

McCroskey, James C. (1967), "Attitude Intensity and the Neutral Point on Semantic Differential Scales," Psychiatric Quarterly, 642-45.

My personal opinion, supported by these articles and others, is that we use questionnaires to measure perceptions; we analyze that data through variance.  Complete neutrality is highly unlikely.  By eliminating a neutral point, we require respondents to report variance.  Also, I read a very interesting study in which respondents had to report from three reasons (didn't know / no opinion / didn't wish to answer) WHY they chose a neutral point when they did.  Through factor analysis it became clear there were no clear reasons for the choice... it became unexplained variance.  I wish I could find that particular reference, but I can't now.  By the way... I took three full days investigating this during my dissertation proposal stage... by the end of the three full days, the stats specialist on my committee agreed with me that not having a neutral point was preferred.  This same well respected statistician personally did studies where he found that 2- point and 5-point scales provided nearly equal amounts of explained variance in a number of different data collection efforts.

Reply from 

Sev V. Nagalingam

I would rather prefer a qualitative response than selecting a scale between 1 to 5. or a response number with the reasons. since, many tend to select mid scale, if they don't have any opinion or view on that matter.

In my recent study " Organizational Re-building during Recession", a scoring method of 6 point scale was used, participants feedback was positive during filling up of the questionnaire.

Article Organizational Re-building during Recession

Sev, I agree that qualitative responses are valuable... the use of both, in a mixed methods study, achieves triangulation and therefore greater evidence that both uncovers the unexpected (qualitative) and the ability to test validity/reliability/etc. using a quantitative method.

Maritha Purperhart  to you      4 hours ago

I agree with paul just know how to triangulate, when you use mixed methods coding answers (qual) results over yes/no answer can enhance validity.

Development Solutions staff university libraries, college

Sharing link:

https://www.google.es/url?sa=t&source=web&rct=j&url=https://www.acaps.org/sites/acaps/files/resources/files/acaps_technical_brief_questionnaire_design_july_2016_0.pdf&ved=2ahUKEwjr-9OPhMjZAhWC1xQKHYP0DWgQFjANegQIARAB&usg=AOvVaw0ARscnfxXCnSC7WfFaEV6U

Regards!