Of further Interestmay bee 3 decades f reseacrh on 4 continentwith stone age people a(Amazon Auca Indians, Brain & Cognition 1989;10:54-750, - fastes ficht-flight reacrion//Aggession & Violn havior, specifically re human facce representation on an early, poss8ibly innaste schema dubbed "Neolirthic Face", see su, in my datatbase on Face represention , a potential innate schema.

Thank you but I'm afraid I cannot understand how this answer is related to my question.

I would advise You to very precisely describe each level in Your case. When You have clear distinctions of what belongs to which level, then You will be able to easily match journal to level.

You are dealing with a classic issue in what is known as qualitative content analysis, which is the difference between manifest and latent content. With manifest content, it would be easy for any two people to agree on whether or not a code was present. With latent content, there is a much high degree of subjective interpretation.

With manifest content, we can expect high inter-rater reliability. In that case, if you had a second person read and rank the journals, you would expect a high level of agreement with your own ranking. But that does not seem to be the case for your data. So, unless you can do what Rūta suggests, and develop very careful rubrics for detecting differences, you may be forced to rely on your own judgement.

One possible option might be to approach this problem "in reverse," so that you would start by sorting the journals from high to low, based on your subjective judgement. And then in the next step, you would do a systematic comparison of the high, medium, and low cases to see if you can find some code-able indicators of the differences.

Thank you very much Rūta Tamošiūnaitė for your constructive advice. I had also thought about this type of operationalization of different levels. Now I'm wondering if some quantification is also needed in my descriptions and data analysis. For example, if counting the number of supporting sentences as indicators of depth of argumentation could be a legitimate idea. 

Thanks a million David L Morgan. You kindly gave me some of key terms that I was not aware of. I will start Googling "qualitative content analysis" and "manifest and latent content" right away, and hopefully, I may be able to find some practical examples in the previous studies. 

The reverse analysis is definitely a good idea, but if I want to stick to my proposal, I think I should go on with deductive analysis of content and classifying journals into the levels that are present in the literature. 

Masoud, you are responding to qualitative data by immediately wanting to codify it, to turn it into quantitative data. While that can be a legitimate activity, the point is that qualitative information involves a search for meaning, to extract a conclusion and the point of your work from data. Numerical or codified data however does need to be qualified into words and meaning, for any numerical data or information to have value. Seen in this way to codify qualitative data is a backward step, one that's automatic for engineers, but in it's philosophy makes no sense. The tools for analysing qualitative data are not the tools of an engineer, they are instead the tools of reason, of inference weak and strong, of evidence for causality, of probabilities, of possibilities like underlying mechanisms.

Analysis and reason are the tools you need to work with words.

You could use the criteria for evaluating qualitative research that is published on the Joanna Briggs Institute website.  If you adopt this approach you could ask a second reviewer to also perform the same activity, then compare and negotiate the ranking.