Thomas GIA Tips: Five Tasks, Two Seconds a Question
The Thomas GIA has no fixed number of questions, penalises wrong answers at a different rate on each of its five tasks, and rewards speed on some and accuracy on others. How to pace each one.
Thomas International's General Intelligence Assessment, which Thomas now also calls the Aptitude assessment, is a battery of five very short timed tasks. Almost all the advice you will find about it is Thomas's own line: work quickly and accurately, both matter equally. That is true and it is not much help, because the five tasks do not reward the same balance and two published Thomas sample reports show exactly where each one tips.
The first surprise: there is no number of questions
Most aptitude tests tell you the paper is 20 questions in 20 minutes, and the goal is to finish. The GIA is not that. Thomas publishes only that each of the five tasks lasts between two and five minutes, with the whole series running 30 to 45 minutes including instructions. There is no per-task item count anywhere in Thomas's material, and there is a reason: the GIA descends from a battery built around generating its items by algorithm, so the stream does not run out.
Prep sites confidently quote 130 to 160 questions across the battery. Two Thomas sample reports show candidates completing 178 and 168 items, which tells you that 130 to 160 is a description of an average candidate rather than a ceiling. The practical consequence is simple: stop thinking about finishing. Your score is how many you get right before the clock stops, and nobody reaches the end.
The second surprise: guessing is worth nothing
A real Thomas GIA report prints four columns per task. Done, Right, Wrong, and an Adjusted Score. The adjusted score is your right answers minus your wrong answers divided by one less than the number of answer options on that task, and that formula reproduces the printed figure on every row of both sample reports.
Every penalty is engineered so a blind guess is worth zero on average. That gives you one clear rule, and it is the opposite of the rule on a CCAT or a Wonderlic: guess only once you have ruled an option out. Do not clear the last twenty seconds by clicking, and do not leave an item where you had narrowed five options to two.
The third surprise, and the useful one: the five tasks reward different things
This is where the sample reports earn their keep. Two candidates, two different norm sets, both around the 80th percentile overall, and their subtest rows tell opposite stories depending on which task you look at.
Perceptual Speed is decided by errors, not speed
Candidate A reached 54 items and landed at the 24th percentile. Candidate B reached 56 and landed at the 54th. Two items of extra throughput, thirty percentile points of difference, and the entire difference was errors: twelve wrong against ten. On this task everyone is already working at about two seconds an item, so throughput barely separates anyone and the score falls out of the error rate.
That makes Perceptual Speed the single most counterintuitive task in the battery. It feels like the one to go fast on. Going faster is precisely what will not help, and going slightly slower is the move that raises the score. The rule people most often drop under pressure is the case rule: a capital F and a small f are the same letter and count as a matching pair.
Reasoning is decided by throughput
The reverse. Candidate B reached 38 items at perfect accuracy and scored the 44th percentile. Candidate A reached 53 items with one error and scored the 93rd. Being careful and slow is actively punished here, and one wrong answer out of 53 was nowhere near enough to offset fifteen extra items.
Spatial Visualisation punishes speed bought with errors
And a third pattern again. Candidate B reached 18 items with four wrong and scored the 71st percentile. Candidate A reached 14 items with one wrong and scored the 73rd. Twenty-nine per cent more throughput, and it bought nothing, because four errors at half a mark each wiped out the extra items. If a symbol pair is genuinely ambiguous to you, the honest answer is to take the extra second.
Reasoning hides the statement, which is the whole task
The Reasoning task works in two stages and almost every description of it online misses the second one. A short statement appears comparing two people, for example that one is heavier than the other. You can study it for as long as you need. When you click to continue, the statement disappears, and a question about it appears with two possible answers.
You answer from memory. That hidden step is what makes it a working-memory task rather than a reading task, and it is why Thomas describes the construct as the ability to hold information in short-term memory and solve problems with it. The failure mode is recognisable: long response times and low accuracy usually mean a candidate is trying to reconstruct the relation after the statement has gone.
The fix is to encode the answer before you click through, not the sentence. Decide who is heavier while the statement is still on screen, and then the question is a lookup. Watch for the negative phrasing too, because “Ana is not as heavy as Ben” means Ben is the heavier one, and that inversion is where the harder items live.
What the overall number actually is
A GIA report prints a percentile for each task, an overall percentile, and something labelled GTI. That is the General Trainability Index, and it is a T-score: mean 50, standard deviation 10. A GTI of 60 is roughly the 84th percentile, not “just over half marks”. Candidates who see 60 and assume they did averagely have misread it by about thirty-four percentile points.
Thomas describes the overall as a weighted combination of the five tasks and does not publish the weights, so nobody outside Thomas can tell you how much each task counts. What the report does make clear is what an employer is asked to look at: its interviewer notes are built around probing pronounced highs and lows across the five. An uneven profile is what gets discussed, which is the strongest argument for drilling your weakest task on its own rather than resitting the whole battery.
Practical setup, because on this test it counts
Thomas tells candidates to use a desktop computer or laptop because the assessment will not run on mobile devices such as tablets or smartphones. Its candidate booklet goes further and recommends attaching a real mouse if you normally use one rather than a trackpad, asks left-handed candidates sitting on site to say so in case the mouse buttons need reversing, and notes you may be asked for your screen size so the tasks render at a consistent scale.
None of that is fussiness. On a task scored in units of about two seconds, your pointing device is part of the measurement. Thomas also recommends not sitting it on a work network, and sitting the whole series in one go. Sort all of it out before you start rather than during the first task, because the clock does not wait.
If your invitation says something else entirely
Thomas renamed all four of its instruments and kept the acronyms, so an invitation may say Aptitude assessment, or GIA, or both. It may also be a different Thomas test altogether: the Behaviour assessment, also called the PPA, is the instrument Thomas customers use most and it has no right answers at all. The Thomas practice hub has the full name mapping, the per-task clocks and the drills for each of the five GIA tasks.