ImageGuesser
M03. Physics

Checking light, shadows and reflections

A strange shadow can be a useful clue, but first you need a plausible account of the scene. Several lamps, a sloping surface or a reflection can explain a detail that initially looks impossible. This lesson starts with an actual example of that problem.

Worked example · Published puzzle, 2026-09-28

A real terrace with several sources of light

The September 28 puzzle includes a night photograph and a generated daylight terrace. The photograph is a useful counterexample to the rule that every shadow must point in one direction: this scene contains several lamps.

These are two candidates from the same archive puzzle. Their labels are shown because this is a lesson. Numbered boxes identify the details discussed below.

Night photograph of a rooftop terrace; boxes mark hanging lamps and light reflected on decking.
Photograph used as the real answerArchive image 2

Multiple hanging lamps, wall lights and small table lights illuminate the real terrace.

Photo by Brands&People on Unsplash.

AI-generated daylight terrace with a wooden table; boxes mark lanterns and the deck around the table.
AI-generatedThis image was generated by artificial intelligence.Archive image 1

A generated alternative uses daylight, recognizable lanterns and a shallow-focus appearance.

  1. 1Count sources before tracing shadows

    Two bright hanging lamps are visible near the left post, with more lamps farther along the terrace. Each source can illuminate an object from a different direction.

    Limit: Different shadow directions can be consistent with a real scene containing several lights.

  2. 2Bright patches need context

    The decking and tables catch pools of warm light. Their brightness varies with location, surface angle and the surrounding lamps.

    Limit: A bright patch is not automatically a lighting error. Try to identify what could illuminate it.

  3. 3Believable objects do not prove a photograph

    The generated terrace includes lanterns with recognizable paper-like ribs. At this scale, those familiar details can make the scene feel photographic.

    Limit: Familiar materials and plausible lighting are within the capabilities of image generators.

  4. 4Blur is an observation, not a verdict

    The synthetic scene has a sharper central table and softer surroundings. That arrangement resembles photographic shallow focus.

    Limit: Real lenses, editing and generated images can all produce selective blur. It cannot settle this comparison.

What establishes the answer here?

Image 2 is the source photograph credited to Brands&People; image 1 is a generated candidate. The day/night difference is not the answer key: these are different scenes. The source record establishes the puzzle label, while the comparison teaches why the number of lights matters before judging shadows.

The labels come from this published puzzle's answer key and linked photo source. They do not come from an automated detector or a claim that one visual detail proves authenticity.

Explore all six images from this puzzle →

Map the light sources before judging shadows

Start with what the image shows: windows, lamps, signs, skylight, reflected light, and any flash or fill light that could sit outside the frame. Note whether the scene is indoors, outdoors, in daylight or at night.

Under one distant light, shadows on a flat surface can be compared for consistency. Their appearance in a photograph also depends on perspective, the surface, and the objects' positions and heights. Equal-looking objects do not automatically need equal-length shadows in the image.

For a street at night or a room with several lamps, the one-light assumption is a poor starting point. The terrace example above shows why. Explain the visible pools of light before deciding that a dark patch points the wrong way.

Compare reflections with their geometry

A useful comparison

Find an object that should be identifiable in a mirror or other reflective surface. Compare distinctive details such as clothing, a sign, or the arrangement of nearby objects. An unexplained extra object or changed detail is worth investigating.

An important limit

The reflected viewpoint differs from the camera's. Curved glass, rippling water, tinted windows and overlapping reflections can distort or hide details. A reflection does not have to repeat the camera's direct view.

Follow structure across the scene

Trace a railing, table edge or row of tiles through the frame. Look at the junctions where objects overlap. If a line reappears in an unexpected position, check whether a change of angle, a different object, lens distortion or an obstruction explains it.

For materials, ask how the surface could be lit or treated. Cloth can be glossy, wood can be varnished, and metal can be matte. A surprising appearance is not enough to establish that an object was generated.

Check blur with the same care. Different depths, movement and editing can affect sharpness. Generated pictures can imitate these effects. Record a specific unexplained mismatch; avoid turning “looks too sharp” or “looks too blurry” into a rule.

A lighting and structure check

Before you guess, run through these checks:

  • 1
    List plausible light sources, including ones outside the frame and reflected light.
  • 2
    Decide whether a one-light, flat-surface comparison actually fits the scene.
  • 3
    Compare recognizable reflected details while accounting for the reflected viewpoint.
  • 4
    Follow edges and junctions, allowing for perspective, obstruction and lens distortion.
  • 5
    Use anomalies to guide a source search; consistent-looking lighting does not prove a photograph.

Sources and further checks

Our worked examples explain observations in published ImageGuesser puzzles. For provenance tools and their limits, consult the organizations that maintain them:

Examples and guidance updated . A visual clue is a reason to investigate; it is not a universal authenticity test.