Inductive reasoning — also called abstract reasoning or visual reasoning — is the section most candidates find hardest to prepare for. Unlike numerical or verbal tests, there are no formulas to memorise and no text to read. You're shown a sequence of shapes and must identify the pattern to find what comes next.
SHL, Kenexa, and Cubiks all include inductive reasoning sections. Companies like KPMG, EY, and Accenture use these to assess problem-solving ability that's independent of language or education.
Objects rotate by a consistent angle between each frame. Look for: shapes turning 45°, 90°, or 180° each step. The rotation might be clockwise or anti-clockwise. Sometimes different elements within the same frame rotate in different directions.
Objects move across the frame in a consistent direction. A dot might move one position right each step, or a shape might bounce between corners. Track the position of each element separately.
Elements cycle through states: white → grey → black → white, or filled → empty → striped. Count how many states are in the cycle and predict the next one.
Objects grow or shrink systematically. A circle might get larger each step while a square gets smaller. Or alternating frames show large/small versions.
New elements appear or existing elements disappear following a rule. One new dot is added each step. Or one line is removed each step. Count the elements in each frame and look for arithmetic patterns.
The hardest questions combine multiple rules simultaneously. A shape might rotate 90° AND change colour AND move one position right — all in each step. You need to track all three rules independently.
Step 1: List every attribute. For each element in the sequence, note its: shape, size, position, colour/shading, orientation, and number. Do this for the first 3 frames.
Step 2: Find what changes. Compare frame 1 to frame 2. What's different? Then compare frame 2 to frame 3. Is the change consistent?
Step 3: Apply the rule to predict. Once you've identified the rule, apply it to the last frame to predict the answer. Check your prediction against the options.
Step 4: Eliminate. If you're not 100% sure of the rule, eliminate wrong options. An option that violates ANY rule you've identified is wrong. Even eliminating 2 out of 5 options significantly improves your odds.
Current AI vision models struggle with spatial reasoning. They can recognise objects but have difficulty tracking precise positions, rotations, and counts across a sequence. This is why TestSolve's accuracy on inductive reasoning (72%) is lower than numerical (94%) or verbal (96%). We use object-centric decomposition — breaking each frame into individual objects and tracking their attributes — combined with multi-hypothesis rule testing to maximise accuracy.
Unlike numerical reasoning, you can't grind through formulas. Instead, train your visual pattern recognition:
Timed practice is essential. You get about 60 seconds per question in SHL inductive reasoning tests. Practice identifying patterns quickly — the pattern itself is usually simple, but finding it fast is the skill being tested.
Look at the answers first. In matrix completion questions, looking at the answer options before analysing the pattern can give you clues. If all options have the same shape but different colours, you know the pattern is about colour, not shape.
Don't overthink. The pattern is usually simpler than you expect. If you've been staring at a question for more than 60 seconds, step back and look for the obvious change between frames.
Try TestSolve a free solve — see how our AI handles visual reasoning in real time.
The six pattern families above cover roughly 95% of inductive items. Each works differently, and the diagnostic clue that tells you which type you're looking at is different. These worked examples show the recognition step alongside the solution.
Series: Frame 1 — an arrow pointing right. Frame 2 — the same arrow rotated 45° clockwise. Frame 3 — rotated a further 45° clockwise (now pointing down-right). Frame 4 — rotated again. Find frame 5.
Diagnostic clue: The shape is identical between frames; only orientation changes. That immediately rules out colour, count, position, and replacement patterns.
Approach: Each step rotates 45° clockwise. From frame 4 the arrow is pointing down (270° from the start). Adding another 45° gives down-left (315°). Frame 5: arrow pointing down-left.
Speed tip: For rotation patterns, identify the angle increment (45°, 60°, 90°, 120°) from frames 1→2, then verify by checking 2→3. If both confirm the same increment, project forward.
Series: Frame 1 — one dot. Frame 2 — three dots. Frame 3 — six dots. Frame 4 — ten dots. Find frame 5.
Diagnostic clue: The objects themselves don't change; only how many of them appear. Count, count, count.
Approach: The differences are 1 → 3 → 6 → 10 — these are the triangular numbers (1, 3, 6, 10, 15, ...). Each step adds one more dot than the previous step added: +2, +3, +4, +5. Frame 5: 15 dots.
Speed tip: For count patterns, write the count for each frame above it (1, 3, 6, 10). Then compute the differences. If the differences themselves form a pattern (constant, arithmetic, geometric), you've cracked it.
Series: Frame 1 — three shapes, the leftmost black. Frame 2 — three shapes, the middle one black. Frame 3 — three shapes, the rightmost black. Frame 4 — three shapes, the middle one black again. Find frame 5.
Diagnostic clue: Shapes don't change, count doesn't change — only which one is filled in. The pattern moves left-right-right-middle.
Approach: The black-shape position cycles: 1, 2, 3, 2 — this is a "bouncing" pattern that reverses at the ends. After 2 comes 1 (the bounce back). Frame 5: leftmost shape is black.
Speed tip: When position changes follow no obvious arithmetic, check for bouncing (1,2,3,2,1,2,3...) or cycling (1,2,3,1,2,3...). These are the two most common variants.
Series: Frame 1 — a small square. Frame 2 — a square 1.5× larger. Frame 3 — 2.25× larger than frame 1. Frame 4 — 3.375×. Find frame 5.
Diagnostic clue: Frame-to-frame the shape grows or shrinks by a fixed ratio. The shape, count, colour, and position stay constant.
Approach: 1, 1.5, 2.25, 3.375 — each is 1.5× the previous. Frame 5: square 5.0625× the size of frame 1 (which would visually be ~50% larger than frame 4).
Speed tip: Size patterns usually scale by clean multipliers — 1.5×, 2×, ½×. If the ratio doesn't look clean, you're probably misidentifying which dimension scales (sometimes it's area, not linear size — area doubles when linear size scales by √2 ≈ 1.41).
Series: Frame 1 — triangle, square, circle. Frame 2 — square, circle, triangle. Frame 3 — circle, triangle, square. Find frame 4.
Diagnostic clue: The set of three shapes is consistent across frames, but each frame is a cyclic permutation of the previous.
Approach: Each frame shifts the sequence one position to the left (the leftmost element moves to the rightmost slot). Applied to frame 3 (circle, triangle, square): the circle moves to the end. Frame 4: triangle, square, circle — which is identical to frame 1, confirming a 3-step cycle.
Speed tip: When elements rearrange, check whether the rearrangement is (a) a fixed permutation applied repeatedly, (b) random with one rule (e.g., a specific shape always moves), or (c) external shape replacing internal. The first is most common.
Series: Frame 1 — one black triangle. Frame 2 — two grey squares. Frame 3 — three white circles. Frame 4 — four black hexagons. Find frame 5.
Diagnostic clue: Multiple attributes change between every frame. Don't try to find a single rule — find two or three independent rules and apply each.
Approach: Three attributes vary independently. Count increases by 1 each frame (1, 2, 3, 4 → next is 5). Colour cycles black-grey-white-black-grey (so next is grey). Shape advances in sides (3, 4, 5(circle counts as ∞), 6 → next is 7-sided, a heptagon). Frame 5: five grey heptagons.
Speed tip: Compound patterns are the hardest. Decompose the attributes one at a time — colour, count, shape, size, orientation, position. Solve each independently, then recombine. Trying to spot the "overall pattern" is what causes the 60-second time loss above.
The biggest gain on inductive tests comes from fast pattern classification — knowing which family you're in before you start hunting for the rule. Use this decision tree:
If you can classify the pattern type in 5 seconds, you have 55 seconds to solve. If you spend 30 seconds wondering what kind of pattern you're looking at, you have 30 seconds to solve — and you're already in the time-pressure failure zone. Pre-trained classification is the difference between candidates who finish inductive tests and candidates who don't.