<p>I'm hoping the knowledgeable parents on this board can explain to me how the SAT Subject tests are scored.</p>
<p>As I understand it, the % score is straightforward: It is how you scored relative to other test takers of the same subject test. </p>
<p>But how do they determine the scaled score? Is it what overall SAT test takers would score if they took the test? Or is it also a relative measure of overall subject test takers? </p>
<p>The colleges that ask for 2 or 3 subject test are all very selective, and test takers are all aiming for a 700 or above. So what does the 700 or above mean?</p>
<p>Well, I imagine that tokenadult knows for sure . . .</p>
<p>but as I understand it (roughly), the actual standard deviation of the raw scores on each of the SAT Subject Tests was set to 100, in terms of the scaled score, for the cohort of students who took that subject test in some year or years. Current standard deviations vary somewhat from 100 points, but usually not too much. This probably reflects some shifts in the pool of test takers since the reference group.</p>
<p>So this leaves only the issue of how the mean score is set for each subject test. </p>
<p>I believe that the mean for each subject test is set according to some prescription based on the mean scores on the SAT I of the group taking that SAT subject test. The reference group is probably “historical” by now.</p>
<p>I don’t know whether the mean score on each of the subject tests is based on the overall SAT CR and M scores of the reference group, or on the scores individually. That is, it’s not clear to me whether the mean for Math II C is based only on the SAT I M score of the reference group or on M + CR (and similarly for Literature–whether that mean is based on CR alone, or CR + M). I doubt that W was included in the SAT I when the subject scales were set.</p>
<p>The mean scores do give you some indication of the SAT I level of the students who opt for that subject test–or at least, of the SAT I level of the students in the reference group, who historically opted for that subject test, since there have no doubt been some shifts since then. I believe that this explains why the 50th percentile is pretty high (on the 200 to 800 scale) in physics, for example.</p>
<p>As an addendum: I think the 50th percentile in Chinese is somewhere around 750. I doubt that this means that students who take the SAT II in Chinese have averaged 750 on each section of the SAT I–that seems a bit extreme! I think it might mean that the students currently taking the SAT II in Chinese are shifted up in ability in the language, relative to the reference group on whom the scale is based.</p>
<p>While Quant maybe correct (whose logic makes no sense to me), it’s not worth overthinking the issue. For highly selective colleges (Ivies et al), 700 is the first cut for unhooked candidates, and 750 is a LOT better. The only difference is Math 2 for engineering types, and languages – not sure how much mental discounting occurs for acing a language test in a language spoken at home.</p>
<p>
</p>
<p>It may also indicate that the test is: a) easier relative to the other language tests (and it is reported to be so); 2) the language is spoken at home. Hispanic kids in SoCal can easily score 700 on the Spanish test but typically struggle with SAT I.</p>
It means that you don’t need to retake it. (Unless you are aming at MIT or CalTech, and this is your Math score. In that case you better get 800 or very close to it)</p>
<p>While it’s absolutely true that 700 and above are the scores to aim for if a student is looking at top colleges, I wouldn’t go so far as to claim this:
</p>
<p>I know, personally, numerous students who have gotten into what are pretty generally agreed to be the very top schools with at least one SAT II under 700. They were all extremely competitive candidates with very strong, interesting, varied profiles, but I don’t think any one of them could legitimately be called “hooked.” </p>
<p>An otherwise appropriately competitive application to a top school is not going to be thrown out for one SAT Subject Test score under 700. (Math scores to MIT/Cal Tech or specific engineering programs may be a different story.)</p>
<p>Assuming that the subject test takers tend to be competitive applicants applying to very competitive colleges, then what I am trying to understand is the relative difficulty of getting a 700 or above. Are the scores somehow normed to what these students get on the SAT? Many of them would be in the 700’s/test range for SAT. So does that make it even harder to score in the 700’s or is it normed in such a way that a top test taker who is in the 700’s SAT and knows the material well might also get a 700 on a subject test?</p>
<p>Perhaps I’m over analyzing and don’t really need to know. But if anyone understands how it works, I’d love to know. </p>
<p>Yes, you are definitely over analyzing. For subject tests, you can take CB at its published word. For example, for US hisotry, CB recommends:</p>
<p>
</p>
<p>Thus, for the mean is pegged against what should be learned in an average College Prep course. Of course, many kids take the US History test after completing APUSH. Thus, these kids will likely score much higher than the mean, but the fact is that the AP course is of much higher difficulty than a CP course in the same subject… Ditto for other subject tests which correspond to an AP test. But, those scores have absolutely nothing to do with the SAT I.</p>
<p>Well, don’t overanalyze, but don’t underanalyze, either.</p>
<p>For example, with regard to the U.S. History test, the College Board says that a “one-year, comprehensive course in U.S. History at the college preparatory level is essential.”</p>
<p>But not all courses labeled “U.S. History” are comprehensive. Some schools prefer to choose selected topics in U.S. history and focus on them deeply, perhaps with research papers and other writing assignments, rather than teaching a comprehensive survey course. The SAT Subject Test is very much geared toward a survey course.</p>
<p>It doesn’t hurt to find out what your high school is teaching before making a commitment to take a Subject Test.</p>
<p>Of course, and that is true for EVERY subject at every HS. Not all high schools cover all statistics in PreCalc, for example, but that topic is frequently questioned on the Math 2 subject test. Even an “honors” class might not be comprehensive. </p>
<p>And that is why CB has an average and a curve. But neither have anything to do with the SAT I.</p>
<p>At one point in time (at least) the means on the various GRE subject tests were set with an algorithm that depended on the scores that testers in those subjects had achieved on the general part of the GRE (which, at the time, had verbal and quantitative parts only). This information came from an official information booklet that was issued by ETS and made available at Harvard University (and probably at other universities). My spouse and I both read the booklet rather closely, in 1979. The highest means on the GRE subject tests at that time were in physics and classics.</p>
<p>I would be surprised if ETS is doing something completely different to set the means on the various SAT II subject tests. But perhaps someone can locate ETS information about how the mean scores are set. I glanced quickly at the links posted by tokenadult and didn’t see that information there, but will look again.</p>