I assume the output is from SPSS, which I'm not very familiar with, so the following may not be a correct interpretation of the output.

Both the Chi-square and Likelihood Ratio tests assume that both variables (or both axes on the table of counts) are nominal in nature. You'll note that their results are very similar.

The linear-by-linear test assumes that both variables are ordinal in nature.

You have one ordinal variable and one nominal variable. What to do? In the general case, you could use an extended Cochran-Armitage test.

In your case, because you have a dichotomous variable, the dichotomous variable can be treated as either a nominal variable or an ordinal variable, so you can use the linear-by-linear results.

Tahir, please share the observed counts for the 3x2 table. Thanks.

Dear Bruce Weaver, Here is observed count

Thanks Tahir. I was curious about why the tests that treat the variables as nominal are significant whereas the test of linear-by-linear association is not. I think it is because if you look at the column percentages, you have a U-shaped function in one column and an inverted U in the other. This is clear if you add the /BARCHART option to your CROSSTABS command and look at the graph that is produced (attached). HTH.

Thanks Bruce Weaver,

I excluded the last columns and concluded that creativity only increased till the second story.

Am I doing right? or is there any better way to report it

Tahir, when you say "creativity only increased till the second story", I become suspicious that your Order variable might represent a repeated measures variable. Please tell us more about the design of the study, and clarify whether the 3 levels of Order represent independent groups or repeated test conditions for the same subjects. (I realize that the sample sizes are 100, 101 and 101; but it's possible data were missing for one subject in the First level of Order.) Thanks for clarifying.