Diffrint
Support
ServicesNot for profit

Classic Inkblots

How it works

The browser uses visible rules to code what you wrote, where you saw it, and which visual clues you confirmed. Unsupported codes stay unknown instead of being guessed.

Try the activity

Answering and clarification

First, you respond to all ten cards without scoring prompts. Then you revisit each answer to mark its location and explain what made the blot look that way.

One answer per card completes the activity. Extra answers are optional. A ten-answer result is labeled a short sample because each response has more influence.

Marking location

You can choose the whole blot, white space, `Not sure`, or tap the parts that contributed. The activity records points on the image. It does not invent formal D or Dd codes from physical size.

Because the cards are symmetrical, a tap can mark the matching area on the other side. You can turn mirroring off, mark several areas, or remove a mark. Rotation is recorded with the original answer.

What the browser codes

Negation and context matter. A knife is not treated as an aggressive act unless the response describes one.

Understanding freeform answers

Exact word banks and language rules handle synonyms, word forms, phrase boundaries, negation, and relationship grammar. These remain the highest-confidence matches.

An optional MiniLM embedding model can recognize similar meanings written in different words. It runs through WebAssembly in the browser, does not generate text, and does not send answers to a model server. Its first download is about 45 MB. If it cannot load, the exact word-bank and grammar layers still produce the report.

Building the interpretation

The interpretation starts with how you handled ambiguity. Mostly using the whole blot supports a broad, organizing reading; mostly using selected areas supports a local, foothold-by-foothold reading. Connected scenes, visual complexity, and movement change what either location pattern means.

Content is interpreted in context, not through a secret-symbol dictionary. Repeated people can make social roles and relationships more central. Animal answers are discounted when they are common responses strongly invited by the card; less-common animals receive more weight when their actions or repeated qualities point toward power, vulnerability, protection, cooperation, or conflict. Bodies, nature, fantasy, art, science, sexuality, food, clothing, and objects are mentioned only when they recur enough to form a theme.

The top paragraphs then combine those content and location patterns with human movement versus color, form-led versus color-led responses, active versus passive movement, relationship themes, and changes across the ten cards. The language stays tentative because these are interpretive hypotheses, not fixed meanings.

Traditional EB and eb components, D and Adjusted D, a Coping Deficit Index estimate, and the Egocentricity Index are shown as experimental estimates. The formulas are established; automatic coding of freeform answers is the uncertain part. Technical calculations and per-response codes stay inside `See the scoring details`.

Vintage inkblot lore

The results include a separate historical section showing what an older Rorschach interpreter might have claimed. It covers genuine ideas from early Rorschach writing, later structural systems, and psychoanalytic content interpretation. It does not feed back into the main interpretation or change any score.

The section can react to animal percentage, whole-blot or selected-area use, movement and color balance, white space, texture, reflections, food, anatomy, blood, damage, masks, sexual content, animal qualities, relationship scenes, and the old reputations assigned to Cards IV, VI, and VII. Every item states which answers or counts triggered it and follows the historical claim with a reality check.

Some historical authors tried to infer diagnoses, intelligence, criminality, gender conformity, or sexual orientation from inkblot content. Those claims were often prejudiced as well as unsupported. They are not personalized here. The point is to show the entertaining old lore without quietly reviving its worst uses.

What is left unknown

Form Quality requires normative tables linking an exact blot area with a specific object. The browser therefore does not calculate FQ, XA%, WDA%, X+%, Xu%, X-%, or conclusions about reality testing.

Formal D and Dd frequency codes, Z scores, GHR/PHR, thought-disorder scores, diagnostic constellations, and risk scores are also left out. Selected areas and themes are reported descriptively instead.

Data and safeguards

The public NaileR dataset contains 600 freeform descriptions. The card and content banks recognize 461, or 77%. Because those responses do not have professional codes, they help with wording rather than clinical labels.

The larger RAP dataset helps check code patterns but does not include the full response and clarification text needed to train a freeform coder. The report also checks response count, content recognition, visual-clue coverage, and location coverage before making claims.

Sources

This is independent educational software and is not affiliated with those authors or any commercial scoring system.

What that means: this is a careful automatic inkblot activity, not a professionally administered Rorschach. It shows its work, leaves gaps visible, and does not turn uncertainty into a diagnosis.