gmat test prep - preparing for MBA entry with GMAT books, tutor, course, online review and practice questions

- GMATgym

Sentence Correction Data: A little analysis of Official GMAT questions.

Posted by michael on Fri 26 Sep 08 at 5:00pm

In order to determine which issues are tested most often in Sentence Correction questions, I analyzed every question in the Official Guide, 11th edition, and the Official Verbal Guide. The OG identifies eight types of errors tested in Sentence Correction questions (6th printing of the 11th Official Guide page 631): Agreement, Diction, Grammatical Construction, Idiom
Logical Predication, Parallelism, Rhetorical Construction, Verb Form.

The most common flaw is an idiom flaw, which occurs in 34% of the Sentence Correction questions. Many students might question how valuable it is to memorize idioms for the Sentence Correction section of the GMAT. Since Idiom questions will appear in approximately 1 out of every 3 questions, it seems as though memorizing idioms may be important.

The second most common flaw is Logical Predication, which occurs 28% of the time. In addition, Peter L. Cooper published ‘The Assessment of Writing Ability: A Review of Research” in 1984, which pointed out that questions that test modification (questions referred to in The Official Guide as Logical Predication) are the most difficult and discriminating. This is true with Logical Predication questions in the Official Guide. Although these questions appear 28% of the time in the Sentence Correction section, they appear 37% of the time in the last 1/3 of the questions. Since the questions in the Official Guide are ordered by difficulty, it is clear that on harder question test takers can expect nearly 2 out of 5 questions to contain a flaw in Logical Predication.

The third most common flaw is Parallelism, which appears in 23% of the questions. This is the last remotely significant observation. There is an average of 1.7 flaws per question in the Sentence Correction section of the Official Guide. Thus, the probability of getting any flaw in a question by chance is 21%. Below are the remaining question types with their frequency of appearance.

Verb Form appears 19% of the time.

Grammatical Construction and Agreement both appear 17% of the time.

Diction appears 15% of the time

Rhetorical Construction appears 14% of the time.

Naturally, the usefulness of these observations is limited by the assumptions I have made.

The first assumption is that this is a random sample of GMAT questions. In general GMAC is generous with the amount of information it discloses to test takers. As such, I think it is reasonable to assume that this sample of questions has not suffered from a selection bias.

The second assumption is that GMAT questions have not significantly changed since these questions were printed. This assumption needs a bit more scrutiny.

Many question published in the Verbal Guide to GMAT Review appeared on actual GMAT Exams between 1993 and 1995. This can be verified by cross referencing the questions in the 8th Official Guide with the GMAT Guide to Verbal Review (the purple book). On page 393 of the 8th Official Guide the years these questions appear on GMAT Exams is stated. As a result I have assumed above that GMAT questions have not changed over a period of 15 years. This may be hard to accept as ACT Inc now writes questions for the GMAT Exam instead of ETS.

However, internal memos from ETS have revealed that question writers often relied on different references to write questions. One writer may have used the Merriam-Webster Dictionary and another the American Heritage Dictionary. Given these inconsistencies within ETS, I think it is reasonable to assume that a different company, given the same specifications, could develop questions that are as consistent for the Sentence Correction section as those developed by ETS.

Additionally, GMAC expends a lot of effort ensuring that test results do not change by changing vendors. For example, when GMAC changed from ETS’s Erater to Vantage Learning’s IntelliMectric it ensured the scores would no change (see ‘Use of the Analytical Writing Assessment: Past and Present’ by Karen Owens October 2006.). I think it is fair to assume that before GMAC changed vendors they ensured the questions written by ACT Inc would test the same skills as those tested by ETS.