Strong Untimed, Weak Timed on the LSAT: Is It Knowledge, Decisions or Time?

Run the audit · Classify losses · Worksheets · Worked logs · Choose an action · Try the scenario · Follow-up

By AccessPrep Editorial · Facts checked November 3, 2026

Imani, Callum, their practice logs and the timing scenario are fictional illustrations created by AccessPrep, not real students or results. No LSAT questions are reproduced. The audit, its marks and its thresholds are teaching proposals that have not been tested in a study. AccessPrep and Lovare share a founder, and Lovare offers paid LSAT instruction and admissions consulting.

When your LSAT accuracy drops under a timer, find out where the lost points come from before you drill speed. For two timed sections, write the timer reading and a one-letter mark after every answer. Then, before checking the key, redo the questions you were unsure of, rushed or never reached. Sort each lost point in order: a wrong or unexplained redo goes in the knowledge-check bucket; a defensible redo after a stopped timed attempt goes in time allocation; a defensible redo after a finished timed attempt goes in decision-making. These are working hypotheses, not proven causes. Work on the largest category first, test one change on fresh sections, and compare the targeted count and total losses before deciding whether to keep it.

Who this is for and what success looks like

You do well untimed or in blind review, and noticeably worse on the same kind of questions with the clock running. “Work on timing” is the usual advice, and it is not yet an action. Success after one audit is a sentence with numbers in it, such as “17 lost points across two sections: 11 time allocation, 4 decision, 2 knowledge”, plus one change to test next.

This builds on two AccessPrep guides. The blind-review guide already tells you to flag uncertain and unusually slow decisions and to review guesses, unfinished questions and two-finalist choices first, and its results table sends “wrong timed; correct in review” questions to a comparison of explanations. The plateau guide calls “You can explain your choice before seeing the key, but struggle under time pressure” an execution hypothesis. This audit tests it with two things neither guide logs: each question’s seconds, and fixed marks written as you answer that separate finished answers from ones the clock decided. Together they split the “wrong timed; correct in review” group into decision and time-allocation losses.

Three reasons timed accuracy falls

• Category: Knowledge · What the log shows: Wrong again untimed, or right but you cannot say why the closest rival fails · Hypothesis to investigate: The skill is not reliable yet, with or without the clock · Avoid assuming: That speed alone will repair a reasoning gap

  
  

• Category: Decision-making · What the log shows: Right untimed with a reason, but under the clock you finished thinking and chose wrong · Hypothesis to investigate: You commit too soon, cannot separate two finalists, or lose the thread of the argument · Avoid assuming: That one correct redo proves stable knowledge

  
  

• Category: Time allocation · What the log shows: Right untimed, but under the clock you stopped before finishing: you rushed, moved on or never reached it · Hypothesis to investigate: The minutes went elsewhere, often to a few long questions · Avoid assuming: That every question needs the same pace

  
  

The audit groups observed losses; it cannot isolate their causes. Fatigue, unfamiliar difficulty, a mistaken self-report, or learning during the first attempt can affect the redo. A category tells you what to investigate next, not what is wrong with you. If the log is incomplete, mark the entry unresolved instead of guessing a category.

Know the section you are timing, now and from August 2027

LSAC’s specifications page describes the current four 35-minute sections, and its undated Reading Comprehension page describes the current RC section as four sets of reading questions, each followed by five to eight questions. LSAC announced the August 2027 change on October 26, 2026: unscored questions move inside the three scored sections, question types stay the same, scores still rest on “one section of 27 scored Reading Comprehension questions and two sections of 25 scored Logical Reasoning questions each”, and anyone testing through June 2027 keeps the current format. Its announcement page gives the averages above and says time per question does not change. Four consequences for the audit:

• Do not assume individual pacing carries over. The announced average time per question is similar, but longer sections, different passage mixes and fatigue can change your performance. Keep format-specific baselines.

  
  

• Recompute anything tied to section length. The pace check below uses the section’s own numbers: 2,100 seconds and 25 questions for a current LR section (or the count your practice section shows, if it differs), and 2,700 seconds and 32 questions for a new one.

  
  

• Never skip a question because it might be unscored. LSAC says test takers will not know which questions are unscored and, as with the current format, should treat each question as scored.

  
  

• Recheck on new-format practice. LSAC says new-format practice tests reach LawHub starting in February 2027, and current-format LawHub prep stays available through the June 2027 test. If you will test in August 2027 or later, run at least one audit on new-format practice once it is out.

  
  

The averages are not targets. This audit never asks you to hit a fixed number of seconds. Use the actual duration and question count of your practice section. If you have approved accommodations, use your authorized timing and breaks and compare only like conditions; the standard-time examples are not accommodations guidance.

Run the audit in five steps

1. Pick the material. Use unfamiliar official sections of one type, all LR or all RC. LSAC says four full PrepTests are free through LawHub. Plan on at least two sections.

2. Log while you work. After each answer, write the question number, the timer reading and one mark. The two mark groups carry the audit: did you finish thinking, or did you stop?

• Mark: S · Use it when (decide as you answer): Sure: you finished and would defend the answer · Group: Finished

  
  

• Mark: L · Use it when (decide as you answer): Leaning: you finished with two answers left and picked one · Group: Finished

  
  

• Mark: G · Use it when (decide as you answer): Guess: you finished and still had no basis for any answer · Group: Finished

  
  

• Mark: M · Use it when (decide as you answer): Moved on: you chose to stop early while time remained · Group: Stopped

  
  

• Mark: T · Use it when (decide as you answer): Time forced: the clock made you rush or skip; unreached questions count as T · Group: Stopped

  
  

Logging adds time and may interrupt concentration. Measure that burden in a practice section you will not count; compare logged sections with logged sections under similar conditions, not with unlogged scores. If you return to a question, write a second reading when you leave it again and add the visits. If your practice platform records time per question, copy its numbers instead.

3. Redo before you see the key. Without a timer, redo every question marked L, G, M or T and every question you did not reach, and review each S question too, recording an answer and rival explanation before the key. Otherwise a confidently wrong answer discovered only after scoring has no independent redo record. For each, write your answer and one line on why the closest rival fails, as the blind-review guide’s method also asks. That line keeps a lucky redo from counting as knowledge. After you check the key, code each redo:

• RR: right, and your line explains why the rival fails;

  
  

• RU: right, but you could not write that line;

  
  

• RW: wrong.

  
  

4. Score the section, then classify every lost point with the rule below. A lost point is a question you got wrong or did not answer.

5. Compute three numbers per section, then total the audit.

• Your median seconds on finished questions (marks S, L and G), so rushed answers do not drag it down.

  
  

• Your sinks: questions that took more than twice that median.

  
  

• The pace check: median × the section’s question count, compared with the section’s seconds (2,100 for 35 minutes). Treat it as an optimistic screen, not a completion guarantee: multiplying a median ignores the long tail, passage variation and revisits. Also inspect the actual total time and the slowest questions.

  
  

Add the categories across sections. If the audit has fewer than 12 lost points, log another section first, subject to the four-section stopping rule below. Twelve is an editorial convenience for this worksheet, not a statistically validated sample size. Even a large-looking category can reflect the question mix or your coding choices. Keep the individual entries so another reader can inspect the reasoning.

Classify each lost point

Apply in order and take the first that fits:

1. The redo was RW or RU: Knowledge. This redo did not establish reliable understanding. Timing may still have contributed; begin with the reasoning rather than claiming the clock was irrelevant.

   
   

2. Otherwise, the mark was M or T, or you never reached the question: Time allocation. You produced a defensible redo after stopping under the clock. Investigate allocation, while allowing for second-exposure learning.

   
   

3. Otherwise (redo RR with mark S, L or G): Decision-making, labeled by the mark. S suggests investigating premature commitment or a task misread; L suggests comparing finalists; G suggests checking your understanding of the task and stimulus. Inspect your written explanation before choosing a repair.

   
   

An unreached question you also miss untimed counts as knowledge: this particular untimed attempt did not resolve it; that does not establish what unlimited future practice would do. Correct G or T answers are not lost points. Mark their reasoning as uncertain rather than declaring them pure luck; a rushed answer can still be reasoned.

Copy this timing-and-confidence log

Download the audit worksheets, blank CSVs and complete fictional logs. The ZIP also includes the scenario, action rules and follow-up sheet. No email required.

On small screens, tables scroll sideways.

During the section, keep three columns on your scratch paper:

• Q: 1 · Timer after answer: · Mark (S, L, G, M, T):

  
  

• Q: 2 · Timer after answer: · Mark (S, L, G, M, T):

  
  

• Q: 3 · Timer after answer: · Mark (S, L, G, M, T):

  
  

After the redo and the key, fill five more:

• Q: 1 · Seconds: · Timed (C, W, U): · Redo (RR, RU, RW): · Category (K, D, A):

  
  

• Q: 2 · Seconds: · Timed (C, W, U): · Redo (RR, RU, RW): · Category (K, D, A):

  
  

• Q: 3 · Seconds: · Timed (C, W, U): · Redo (RR, RU, RW): · Category (K, D, A):

  
  

C is correct, W is wrong and U is unanswered. Seconds for the first question are the section’s starting time (35:00 now) minus its reading; for each later question, the previous reading minus this one. The downloadable CSV has 25 starter rows and a rival-line column; add rows to match your section. The companion worksheet includes a section summary. For return visits, use actual chronological exit readings and sum time by question; never subtract readings in question-number order after moving backward.

For Reading Comprehension, add a passage line with the timer reading when you finish each first read. Keep passage lines out of the question median and the sink check. Record revisits to a passage separately and include them in actual total time. The rough pace screen then has two parts: median first read × passages, plus question median × questions, against the section’s seconds.

Fictional RC example. In one current-format RC section (four passage sets, 27 questions, 2,100 seconds), Callum’s first reads took 205, 220, 240 and 215 seconds, and his question median was 52 seconds. Questions alone look comfortable (52 × 27 = 1,404), but with reads the check is 217.5 × 4 + 1,404 = 2,274 seconds, 174 more than the section allows. One section is not an audit, so he logs a second RC section before tallying; if allocation comes out largest, row 5 of the pattern table (pace work, starting with his reads) fits, not a move-on rule.

Worked log 1: Imani loses most points to time allocation

Imani is fictional. Her logs are two current-format LR sections of 25 questions in 35 minutes, matching LSAC’s 25 scored questions per LR section.

• Measure: Timed correct (of 25) · Section 1: 16 · Section 2: 17

  
  

• Measure: Timed correct plus RR redos · Section 1: 24 · Section 2: 24

  
  

• Measure: Median seconds, finished questions · Section 1: 80.5 · Section 2: 80

  
  

• Measure: Sinks (over twice the median) · Section 1: Q5 262 s, Q11 238 s, Q16 231 s · Section 2: Q6 255 s, Q13 247 s, Q17 212 s

  
  

• Measure: T marks plus unreached · Section 1: 7 · Section 2: 6

  
  

• Measure: Knowledge / Decision / Allocation · Section 1: 1 / 2 / 6 · Section 2: 1 / 2 / 5

  
  

The last ten rows of section 1 show the pattern:

Reading the tally. Her audit has 17 lost points: 11 allocation, 4 decision, 2 knowledge. This is enough to choose an allocation experiment under our worksheet rules, not to establish a statistically significant cause.

Reading the seconds. The rough pace screen fits: her combined median of 80 seconds × 25 questions is 2,000 seconds, inside the 2,100 available. The minutes went to sinks. In section 1, three sinks took 731 seconds, about 490 more than three typical questions: roughly six questions’ worth of her time, close to the seven she rushed or never reached. Section 2 repeats it (714 seconds, 474 extra, six rushed or unreached). The sinks themselves cost her 2 points; the rushed and unreached questions at the end cost 11.

Failure and correction. Imani’s first plan was speed drills on every question type. Her log shows why that would misfire: her median finished question fits the rough screen, and pushing them faster risks the sure-but-wrong pattern in Callum’s log below. Her correction is a move-on rule. Twice her median is 2:40; she rounds down to 2:30, an easy reading to catch. When a question reaches 2:30 (her last reading minus 2:30), she selects her best current answer, flags it, marks M and moves on, returning to flagged questions only after question 25.

Worked log 2: Callum loses most points to decisions

Callum is fictional. His logs are two current-format LR sections of 25 questions in 35 minutes.

• Measure: Timed correct (of 25) · Section 1: 17 · Section 2: 18

  
  

• Measure: Timed correct plus RR redos · Section 1: 23 · Section 2: 23

  
  

• Measure: Time left at the end · Section 1: 3:22 · Section 2: 3:09

  
  

• Measure: Median seconds, finished questions · Section 1: 82 · Section 2: 74

  
  

• Measure: T marks plus unreached · Section 1: 0 · Section 2: 0

  
  

• Measure: Knowledge / Decision / Allocation · Section 1: 2 / 6 / 0 · Section 2: 2 / 5 / 0

  
  

Reading the tally. Callum has 15 lost points: 11 decision, 4 knowledge, none from allocation. The contrast suggests a decision check worth testing; it does not establish the cause of his losses. Eight of his 11 decision losses carry an S: sure and wrong. Those eight took a median of 44.5 seconds, against 77.5 for all his finished questions, and they were his eight fastest answers in either section. His fictional redo notes identify plausible contributors: six task misreads, such as a missed EXCEPT or a strengthener chosen for a weaken question, and two choices made before reading all five answers.

Failure and correction. Callum read his gap as a timing problem and planned more timed sections “to get used to the clock”. But the clock never ran him out: he finished both sections with more than three minutes left. His correction is the S check in the table below: before selecting, he names the task in two or three words, reads all five answers, and selects only an answer that passes that task word. It costs a few seconds a question, and he has three minutes to spare. At follow-up he watches one count, his S-marked decision losses: 8 in the baseline, so he is looking for 4 or fewer across two fresh sections. His knowledge losses, two of them sufficient-assumption questions, go on a separate list for untimed review.

Copy this pattern-to-action table

Apply the rows in order and take the first that fits. Test one action at a time.

• Most frequent decision mark (ties: S, then L, then G): S: sure and wrong · Check to add at the point of choice: Name the task in two or three words and read all five answers; select only an answer that passes the task word

  
  

• Most frequent decision mark (ties: S, then L, then G): L: two finalists · Check to add at the point of choice: Write the one difference between the finalists and go back to the argument or passage to settle only that; if it is not settled by about twice your median, choose, flag and mark M

  
  

• Most frequent decision mark (ties: S, then L, then G): G: no basis · Check to add at the point of choice: Name the task and the relevant claims or passage role before rereading the answers. If the stimulus is an argument, state its conclusion and support; not every LR or RC task contains an argument

  
  

Why ties break this way: knowledge first, because faster work on an unreliable skill repeats the error faster; decision before allocation, because a decision check works on the question where the point is lost; S first among marks, because its check costs the fewest seconds.

“An answer for every question” rests on LSAC’s scoring rules: your score is based on the number of questions answered correctly, all questions are weighted the same, and there is no deduction for incorrect answers, so selecting an answer has no wrong-answer penalty; leave enough time to enter one. Row 5 is slower going: speed has to come from fluency with the task types, and no one can promise how quickly. The follow-up compares observed counts; it does not estimate a causal score gain.

Timed-decision scenario: stay, skip or move on?

AccessPrep-created scenario with fictional numbers; it contains no LSAT question content.

Imani is in her fourth logged section, a current-format LR section, on question 10. The timer reads 20:36. She has spent 2:30 on this question, her move-on point. She has eliminated three answers and leans toward one of the two left. Fifteen questions remain; at her median of 80 seconds they need about 20:00. What should she do?

• A. Keep working until one answer is clearly right; in her first section she got a question right after more than four minutes.

  
  

• B. Leave it blank, flag it and move on, so she can return with fresh eyes.

  
  

• C. Select the answer she leans toward, flag it, mark M and move on; return only after question 25.

  
  

• D. Select the answer she leans toward and move on without flagging, since returning wastes time.

  
  

• E. Leave it blank and go back to question 6, where she had two finalists, since that one might be quicker to settle.

  
  

Credited answer: C. Fifteen questions at her typical pace need about 1,200 seconds, and she has 1,236. Moving now leaves 36 seconds of slack; another 90 seconds here would leave her about a minute short, which in her baseline meant rushed answers at the end. LSAC weights every question the same, so this hard one is worth no more than any of the 15 ahead. Selecting her leaning answer keeps whatever her eliminations are worth, and the flag makes the question easy to find later: in the interface used since the August 2026 test, LSAC says flagged questions show on the question bar for navigation.

• A fails because one slow question that happened to be right is not a pattern; in her baseline, sinks cost 2 points directly and the rushed and unreached questions behind them cost 11.

  
  

• B fails because a blank earns nothing if she runs short before returning, and with no deduction for wrong answers, selecting costs nothing.

  
  

• D fails because its reason is wrong: she returns only after question 25, so a return spends leftover time, never time the remaining questions need, and flagging preserves her planned route back without assuming a measured time cost. In her third section she did have time left to revisit the question she had moved on from.

  
  

• E fails because it leaves question 10 blank and spends part of her 36 seconds of slack on question 6 before she has seen the 15 that remain.

  
  

Transfer: the same moment after August 2027. In a new-format LR section (45 minutes, 32 questions, 7 unscored), Imani reaches her move-on point on question 13 with 26:00 left and thinks, “This one is so hard it might be unscored, so I’ll leave it blank.” LSAC says test takers will not know which questions are unscored and should treat each as scored, so “might be unscored” tells her nothing, and a blank still earns nothing. Nineteen questions at 80 seconds need 25:20, leaving 40 seconds of slack. C again.

Follow-up and stopping rules

Use sections of the same type, format and timing conditions. If the action never applied, there is no adherence rate: record “not tested” and gather another comparable section rather than dividing by zero. Unresolved entries or missing baseline values also mean no decision yet.

After you start an action, log the same number of fresh sections as your baseline audit. Check the stopping rules below first; if none fits, apply these in order:

• Order: F1 · If: You used the action on fewer than three of every four questions where it applied · Then: Not a test yet. Simplify it or write it as a cue at the top of your scratch paper, and repeat

  
  

• Order: F2 · If: The count you are checking did not fall to half its baseline or lower · Then: The observation does not meet this worksheet target. Inspect the entries, change one thing if warranted, and repeat

  
  

• Order: F3 · If: That count fell to half or lower, and your total lost points fell · Then: Keep the action. Re-read the pattern table if the new tally has at least 12 lost points; otherwise log more first

  
  

• Order: F4 · If: That count fell to half or lower, but total lost points did not fall · Then: The total did not improve. Inspect the category that grew and possible differences between sections, then adjust; for a move-on rule, try a later point

  
  

With counts this small, even a halving can be chance (Callum’s 8 falling to 4 would not rule it out), so F3 keeps an action without proving it.

Check these stopping rules after every audit, before the pattern table, and take the first that fits:

1. Knowledge is at least as large as each other category in two audits in a row, each with at least 12 lost points: stop timing work and stop logging. Knowledge-check entries are now a joint-largest or largest bucket in this log, which error-log review and a weekly practice plan address better.

   
   

2. Two versions of the same action have failed (F2 twice running): drop that action and re-audit on two new sections. If the same category comes out on top again, have another person read your log.

   
   

3. Four logged sections in a row hold fewer than 12 lost points in total: stop auditing. There are too few timed losses for this method to separate causes; keep blind reviewing.

   
   

If none fits, go on: to the pattern table after a baseline audit, or to the follow-up rules after a follow-up audit.

These rules count sections, not weeks, because no one can say how long a change takes for a particular student.

Imani’s follow-up (fictional). She logged two fresh LR sections using the 2:30 rule.

She used the rule every time it applied, so F1 does not fit; her checked count fell from 13 to 1, below half, and her lost points fell, so F3 applies: she keeps the rule. A before-and-after tally cannot separate the action from practice effects, question mix or other changes. The reduction in T marks plus unreached from 13 to 1 is consistent with her intended mechanism, and she answered every question. These constructed examples demonstrate the rules; they are not evidence of treatment efficacy, and real before-and-after sections need not have equal difficulty. With only 10 lost points in the follow-up tally, she logs another section before switching to decision work, now her largest category.

Troubleshooting

• Problem: I cannot tell L from G · What to try: L means two answers were left and you picked one. G means you could not defend any answer

  
  

• Problem: I have seen some of these questions before · What to try: Familiar questions inflate the redo. Log only unfamiliar sections, and note any question you recognize

  
  

• Problem: My rival lines keep conflicting with the key · What to try: They count as knowledge. The blind-review guide suggests another perspective when explanations repeatedly conflict with the key

  
  

• Problem: Writing readings breaks my concentration · What to try: Practice the habit on one section you will not count, so it takes less attention in the audit

  
  

Turn your timing log into your next study step

Start free: copy the timing-and-confidence log above, use Imani’s ten rows as the model for your own entries, and run it on two unfamiliar sections of one type. Then take the free LSAT Study-Plan Check, which needs no email. If your tally pointed to decisions or time allocation, choose “I can explain my review answers, but lose accuracy under time pressure”. If knowledge came out on top, choose “Even without a timer, I struggle to explain why an answer works”.

If two versions of your action fail and a fresh audit points the same way, an outside reader for your log may help. You can compare current Lovare LSAT programs through Lovare Institut or contact AccessPrep to discuss your plan. Lovare’s Standard, Elite and Founder Private packages currently include 10 tutoring hours (checked November 3, 2026), more than one read of one log, so decide whether you want ongoing tutoring or a single second opinion before you choose.

Continue your preparation

• Blind review a timed section before you check the key

  
  

• Find the bottleneck behind a flat LSAT score

  
  

• Turn your error log into one weekly practice action

  
  

• Decide whether tutoring, a course or self-study fits your problem

  
  

• Build a seven-day plan with the free Study-Plan Check