Loose Associations: Examples, and Why It Is Now Derailment

Loose associations are ideas that slip off the track. Asked a question, the speaker begins to answer and then slides, sentence by sentence, onto something obliquely related, then something unrelated, and ends somewhere the question never pointed. Andreasen’s Scale for the Assessment of Thought, Language, and Communication (TLC) calls this derailment, and defines it as “a pattern of spontaneous speech in which the ideas slip off the track onto another one which is clearly but obliquely related, or onto one which is completely unrelated” (Andreasen, 1986).

The TLC deliberately dropped “loose associations” as a rating term, keeping it only as a parenthetical synonym in the heading of the derailment item, and the reason tells you most of what you need to know about coding derailment, the thought-disorder sign formerly called loose associations.

What are loose associations?

Loose associations, in the TLC’s vocabulary derailment, are a pattern of spontaneous speech in which the ideas slip off the track onto obliquely related or unrelated ones, so that the reply loses its connection to the question (Andreasen, 1986). The TLC’s full description adds the texture a coder needs: “Things may be said in juxtaposition which lack a meaningful relationship, or the patient may shift idiosyncratically from one frame of reference to another. At times there may be a vague connection between the ideas, and at others none will be apparent” (Andreasen, 1986).

The detail that matters most for coding is about pace. “Perhaps the commonest manifestation of this disorder is a slow, steady slippage, with no single derailment being particularly severe, so that the speaker gets farther and farther off the track with each derailment without showing any awareness that his reply no longer has any connection with the question which was asked” (Andreasen, 1986). The sign is often not one dramatic jump but a gradient, and the TLC adds that it “is often characterized by lack of cohesion between clauses and sentences and by unclear pronoun referents.”

Why did Andreasen refuse the familiar term? The TLC explains that “loose associations” was not used “because it is based on an outdated associationist psychology and because it has been used so loosely as to be nearly meaningless,” and that “derailment” was substituted “because it carries a minimum of baggage and because it is graphically descriptive” (Andreasen, 1986). The same passage notes that four other TLC items can also correspond to what people used to call associative loosening: tangentiality, incoherence, illogicality and clanging. The old word, in other words, was covering several distinct signs. The full set is covered in our guide to the Thought, Language and Communication scale.

What do loose associations examples sound like?

Loose associations sound like readable sentences whose connections are loose, so that the reply drifts from the question by steps. The replies below are synthetic, written to show the slow slippage the TLC describes and, in the last, a single hard jump.

Interviewer: How has the week been?

Speaker: Quiet, mostly. My neighbour has a quiet car, electric, you can’t hear them coming. Nobody hears anything any more, that’s the problem with headphones. I was going to buy headphones for my nephew, he’s eleven. Eleven is when it starts.

Interviewer: Have you been eating properly?

Speaker: The shop on the corner shut. They always shut in the end, the corner ones, it’s the rent. My dad paid rent to a man with one arm. You don’t see that now, the arms, they fix everything now.

Interviewer: What did your sister say?

Speaker: She rang on Sunday. The thing about Sunday is the bells, and the bells in that church are electric now, which is cheating. Everything is cheating if you look. The tide is cheating.

In the first two, every link is oblique but visible: quiet to a quiet car, hearing to headphones, headphones to the nephew. That is the slow, steady slippage, and the speaker shows no sign of noticing the question has been left behind. The third has the same pattern and then a hard jump (“The tide is cheating”) with no connection at all. Both forms are derailment.

The TLC’s own published example is worth reading for the same reason. Asked whether they enjoyed something they had done, the speaker drifts, within a few sentences, to putting peroxide on their hair in California, a roommate from Chicago, and tears in the mirror, before asking the interviewer whether they have experienced anything like it (Andreasen, 1986). The sentences are intact; the reply has nothing to do with the question.

What is the difference between loose associations and flight of ideas?

Flight of ideas is loose associations at speed, under pressure of speech, and the TLC does not give it a separate item. In Andreasen’s words, “Flight of ideas is a derailment which occurs rapidly in the context of pressured speech” (Andreasen, 1986). The TLC’s introduction says the term “has been dropped and is now subsumed under the concept of derailment” (Andreasen, 1986); a rater who also sees rapid, voluminous, hard-to-interrupt speech can code pressure of speech separately (our reading, not a TLC instruction), and the combination is what older descriptions called flight of ideas. The pace side of that pair is covered in pressured speech vs flight of ideas.

The TLC also separates derailment from three neighbours by trigger and by level. The table puts the four side by side.

Sign	Trigger	Where speech breaks	Is each sentence readable?
Derailment (loose associations)	The speaker's own spontaneous speech	Between ideas, sentence to sentence	Yes
Tangentiality	A question	Between the question and the reply	Yes
Incoherence (word salad)	None needed	Inside the sentence or clause	No
Flight of ideas	Spontaneous speech, with pressure	Between ideas, rapidly	Yes; coded as derailment plus pressure of speech

The boundary with tangentiality is the question. “Tangentiality has been defined herein as a different phenomenon in that it occurs as the immediate response to a question,” the TLC states, whereas derailment happens inside spontaneous speech (Andreasen, 1986); that pair is covered in tangentiality vs circumstantiality. The boundary with incoherence is the sentence: in derailment each sentence can be paraphrased and the chain cannot, while in word salad the sentence itself has come apart, which is the subject of word salad vs derailment.

Are loose associations specific to schizophrenia?

No. Derailment occurred at similar rates in mania and schizophrenia in Andreasen’s comparison of patients with mania, depression and schizophrenia, alongside tangentiality, incoherence, illogicality and loss of goal; the signs that did separate schizophrenia were poverty of speech and poverty of content (as summarised in secondary sources; Andreasen, 1979b, full text not reviewed). Loosening of associations had long been treated as the characteristic thought disorder of schizophrenia, and the TLC’s stated purpose was to replace that kind of global label with definitions whose diagnostic value could be tested (Andreasen, 1986).

The TLC paper is explicit about the wider point. Language behaviors “such as associative loosening, clanging, blocking, over-concrete or repetitive speech, and poverty of speech also occur in other psychiatric disorders such as mania or depression, and they also occur in the speech of people who do not meet the criteria for any psychiatric diagnosis, particularly when they are fatigued or stressed” (Andreasen, 1986). A tired speaker at the end of a long day drifts. A rater describes the pattern and its frequency; what it means is a clinical judgment made with the rest of the picture. The positive-symptom version of the item, with its 0 to 5 scoring, is in the SAPS positive thought disorder subscale.

How do you code loose associations from a transcript?

Code derailment as spans at the joins: mark the point where one idea slips to the next, then rate severity from the count of slips across the interview. This is instance-level coding, the approach in evidence mode vs instance mode annotation, and it is what lets a second coder agree or disagree with a specific slip rather than an overall impression.

A synthetic exchange with the decisions marked:

Interviewer: What have you been doing in the evenings?

Speaker: Television, mostly. [on track] There’s a programme about canals. [on track: an evening activity] Canals were dug by hand, Irish mostly, my grandmother was Irish. [Derailment, slip 1: canals to the grandmother via “Irish”] She kept a goat. [slip 2: the grandmother to the goat] Goats will eat anything, paper, the lot, which is why you can’t trust what’s written down. [slip 3: the goat to trusting writing] So I don’t read the letters.

Three slips in one reply, each oblique, none dramatic, and by the end the speaker is explaining why they ignore their post. Three rules keep the coding honest. Mark each slip at the join, so the rating points at text. Rate severity from the count across the interview, using the TLC’s frequency bands (once for mild, rising to more than ten, or so often that the interview is incomprehensible, for extreme), rather than from the single most striking jump (Andreasen, 1986). And check the trigger: if the drift began as the immediate reply to a question, code tangentiality for that first move and derailment for what follows.

With this discipline the sign that was once “nearly meaningless” becomes one of the most reliably rated items on the scale, which is exactly what Andreasen set out to achieve by renaming it.

What are the most common mistakes when coding loose associations?

The most common mistakes are rating by the strangest sentence rather than by the slips, coding tangentiality as derailment, and reading schizophrenia into the sign. Each has a fix:

• Rating from one jump. Severity is a count of slips across the interview, and the commonest form is slow slippage with no single severe break (Andreasen, 1986).
• Coding a reply to a question as derailment. If the drift is the immediate response to a question, that first move is tangentiality (Andreasen, 1986).
• Promoting derailment to incoherence. If every sentence can be paraphrased, the sentences are intact; it is derailment.
• Writing “flight of ideas” as a code. The TLC subsumed it under derailment; code derailment, and pressure of speech separately if present (Andreasen, 1986).
• Reading a diagnosis into the sign. Derailment did not separate mania from schizophrenia (Andreasen, 1979b), and mild forms occur in people who are tired or stressed (Andreasen, 1986).
• Using “loose associations” in the methods section. Cite the item you rated: derailment, from the TLC or SAPS, with the version and the anchors.

Pick the instrument for the question, not the other way round. Loose associations is a description that stopped describing; derailment is a definition two raters can apply to the same transcript and reach substantial chance-corrected agreement on (weighted κ = 0.83), and that is the only version of the sign worth putting in a dataset.

References

• Andreasen, N. C. (1986). Scale for the Assessment of Thought, Language, and Communication (TLC). Schizophrenia Bulletin, 12(3), 473–482. doi:10.1093/schbul/12.3.473
• Andreasen, N. C. (1979a). Thought, language, and communication disorders. I. Clinical assessment, definition of terms, and evaluation of their reliability. Archives of General Psychiatry, 36(12), 1315–1321. doi:10.1001/archpsyc.1979.01780120045006
• Andreasen, N. C. (1979b). Thought, language, and communication disorders. II. Diagnostic significance. Archives of General Psychiatry, 36(12), 1325–1330. doi:10.1001/archpsyc.1979.01780120055007

If you code disorganized speech from transcripts, Tagaroo curates the TLC thought, language and communication items and the SAPS positive thought disorder subscale as span-level annotation schemes, with derailment marked at the join between ideas and inter-rater reliability computed as your coders work. The 1979 definitions and reliability study that the TLC grew out of are Andreasen (1979a).