AI RABBIT HOLES | FRIDAY, OCTOBER 30, 2026 | EXPEDITION #155

Featuring Professor Parallax, White Rabbit, and Hatta

What if a recording contains a voice that nobody remembers hearing? What if an artificial intelligence confidently identifies words buried beneath static? And how can we determine whether those words were actually spoken or whether our brains and machines have discovered patterns that were never meaningful speech?

Imagine entering an empty room.

The windows are closed.

The furniture is motionless.

There is no one else present.

You place a digital recorder upon a table.

You press RECORD.

For thirty seconds, nothing remarkable happens.

Then you return to your laboratory and play the recording.

At first, you hear only faint background noise.

A hum.

A soft hiss.

Perhaps the distant movement of air through ventilation equipment.

But then you notice something.

A whisper.

Or what sounds remarkably like one.

You listen again.

This time, the sound seems almost recognizable.

You think you hear a word.

You play the recording for a friend.

Your friend hears something different.

Then you ask an artificial intelligence system to analyze the recording.

And it produces a sentence.

What happens next?

Have you discovered a mysterious speaker?

A recording artifact?

An auditory illusion?

Or a computer-generated interpretation that is more convincing than the evidence supports?

Welcome to Expedition #155.

On the night before Halloween, we're investigating one of the strangest intersections of human perception and artificial intelligence.

The mystery of hearing words where words may not exist.

And Professor Parallax has a laboratory full of suspicious recordings.

Naturally, Hatta has brought a ghost.

Or so he claims.

Professor Parallax stood beside an old-fashioned reel-to-reel tape recorder.

The laboratory lights were low.

A large computer monitor displayed an audio waveform.

White Rabbit entered carrying a flashlight and three pocket watches.

"Professor, why are the lights dimmed?"

"To establish the atmosphere."

"Is that scientifically necessary?"

"No."

"Then why have you done it?"

"Because tomorrow is Halloween."

The Rabbit looked uneasy.

"Very well. But I should prefer proper illumination."

At that moment, the laboratory door opened.

Hatta entered carrying a white sheet draped over an object.

"Professor," he announced, "I have brought our special guest."

Parallax examined the sheet.

"What special guest?"

"A ghost."

White Rabbit's ears stood straight up.

"A ghost?"

"Indeed."

"Where did you find it?"

"In the attic."

The Professor folded his arms.

"Hatta, what is underneath that sheet?"

"A deeply mysterious presence."

"Remove the sheet."

Hatta hesitated.

"Are you certain?"

"Entirely."

With enormous ceremony, Hatta lifted the fabric.

Beneath it stood a coat rack.

Several hats hung from its branches.

White Rabbit stared.

"That's a coat rack!"

"Not just any coat rack," replied Hatta. "A haunted coat rack."

"What makes it haunted?"

"It has been making strange noises."

Parallax approached.

He gently moved one of the hooks.

Squeeeeeak.

Hatta nodded solemnly.

"There. The voice of the unknown."

The Professor inspected the rack.

"The joints require oiling."

"An embarrassingly materialistic explanation."

"Nevertheless, a testable one."

Parallax returned to his recording equipment.

"Today, gentlemen, we shall examine a genuine scientific question."

White Rabbit opened his notebook.

"What question?"

"Why do people sometimes hear speech in ambiguous sounds?"

Hatta leaned forward.

"And can artificial intelligence identify the speaker?"

"Not necessarily."

"Excellent. I shall reserve judgment until the ghost has retained legal counsel."

The Professor approached the blackboard.

And wrote:

A PROFOUND PARADOX, MY PEEPS! WHAT IF THE VOICE WE HEAR IS SOMETIMES CREATED NOT BY AN UNKNOWN SPEAKER, BUT BY THE WAY OUR MINDS INTERPRET UNCERTAIN SOUNDS?

And down the Rabbit Hole we went.

Consider what happens when someone speaks to you.

Sound waves enter your ears.

The auditory system transforms those vibrations into neural signals.

Your brain processes the information.

You recognize words.

You identify the speaker.

And you interpret the message.

The process feels automatic.

Yet it is extraordinarily complicated.

The brain must distinguish speech from other sounds.

It must recognize phonetic patterns.

It must account for accents, speaking speed, and background noise.

And it must interpret words within their surrounding context.

Suppose someone says:

"Please put the book on the..."

Even before the final word arrives, you may anticipate a likely ending.

Table.

Shelf.

Desk.

Your expectations help you understand speech efficiently.

When the recording is clear, the incoming sound ordinarily provides strong evidence of what was spoken.

But what happens when the sound is incomplete?

Or distorted?

Or almost unintelligible?

Your brain may rely more heavily upon what it expects to hear.

This can sometimes be useful.

It helps us recognize familiar words in imperfect conditions.

But it also creates a possibility.

An expectation can influence what we believe we heard.

And in some circumstances, that influence can produce a mistaken perception.

You may have heard of people recognizing faces in clouds.

Or seeing familiar figures in rock formations.

Or noticing animals within the shapes of tree branches.

These experiences are commonly discussed under the term pareidolia.

It involves perceiving meaningful patterns within ambiguous sensory information.

Something similar can happen with sound.

A person may hear what seems to be a word within a burst of static.

Or detect apparent speech in flowing water.

A repetitive mechanical noise may begin to resemble a familiar phrase.

This is called auditory pareidolia.

An ambiguous recording may contain complicated acoustic patterns.

Those patterns can resemble some of the frequency and timing characteristics of speech.

The listener's brain attempts to interpret them.

Sometimes a word seems to emerge.

But when another person listens, the same sound may suggest something entirely different.

That disagreement is important.

If a recording contains a clearly spoken sentence, independent listeners will often agree about the words, provided the audio is intelligible and the language is familiar.

If the recording contains highly ambiguous noise, listeners may produce conflicting interpretations.

The difference helps researchers investigate whether a proposed message is actually supported by the acoustic evidence.

White Rabbit studied the Professor's waveform.

"So the sound is real?"

"The recording contains real acoustic information."

"But the words may not be?"

"Correct."

Hatta nodded.

"An unfortunate situation. I have experienced it during several speeches."

Parallax raised an eyebrow.

"How so?"

"People heard words. They simply refused to recognize their wisdom."

In 2014, researchers James Nees and Charlotte Phillips published a study examining auditory pareidolia and supposedly paranormal recordings.

The research investigated what people perceived when listening to several types of audio.

These included purported electronic voice phenomena, sometimes abbreviated EVP.

EVP refers to recordings that some people interpret as containing voices of spirits or other paranormal entities.

The researchers also examined actual speech, degraded speech, and acoustic noise.

Different participants received different descriptions of the experiment.

Some were told they were taking part in research involving paranormal voice recordings.

Others received a more ordinary explanation involving speech intelligibility.

The recordings themselves did not acquire new voices.

But the framing changed what listeners expected.

The researchers found that paranormal framing increased the proportion of trials in which participants reported hearing voices in certain ambiguous materials.

And when listeners identified apparent messages in EVP recordings, agreement about the content was low.

The results suggested that expectations can influence how people interpret unclear acoustic information.

They did not demonstrate that every unexplained recording has the same cause.

Nor did they establish a supernatural explanation.

Instead, the experiment showed why investigators must be careful about suggesting what listeners ought to hear.

Once someone tells you that a whisper says a particular sentence, you may have difficulty hearing the sound without that suggestion influencing you.

That is one reason independent, blinded listening tests are valuable.

Science advances through replication and critical examination.

A later study published in PLOS ONE in 2022 investigated related questions about paranormal framing, sensory sensitivity, and auditory pareidolia.

Researchers again compared responses to ambiguous recordings.

The results were more nuanced than a simple claim that mentioning ghosts always makes people hear more voices.

The investigators did not reproduce every effect reported in the earlier study.

Some comparisons showed differences within conditions, but the overall between-group effects did not consistently match the earlier findings.

This matters.

It reminds us that psychological effects may depend upon experimental design, participants, stimuli, and analysis methods.

A single study can produce an interesting finding.

But researchers should investigate whether the effect appears reliably across different circumstances.

It is tempting to tell a dramatic story.

Someone hears a mysterious whisper.

A scientist explains it.

The mystery is solved forever.

But real research rarely works so neatly.

Investigators need to distinguish several questions.

Does the recording contain speech?

Can independent listeners identify the same words?

Does presenting a suggested interpretation change what they hear?

And can the findings be repeated under controlled conditions?

The scientific value lies not in forcing an exciting answer, but in discovering which explanations survive careful testing.

A particularly relevant study appeared in Consciousness and Cognition in August 2024.

Its title was:

"Effects of Expectations and Sensory Unreliability on Voice Detection: A Preregistered Study."

The researchers investigated how expectations influence people's judgments about whether voices are present.

They also examined what happens when the sensory information becomes unreliable.

The study found that expectations influenced voice-detection decisions.

The effects were especially pronounced when the acoustic evidence was less reliable.

This supports the broader idea that perception depends upon an interaction between incoming sensory information and the brain's expectations.

Think about what that means.

In a clear recording, a recognizable word may strongly constrain interpretation.

In a noisy recording, the listener may have less reliable evidence.

An expectation can then exert a stronger influence.

That can help explain why ambiguous sounds sometimes seem to contain meaningful speech.

It does not establish that every unusual voice-hearing experience results from the same process.

It does not imply that someone reporting a strange sound is being dishonest.

And it does not prove or disprove every supernatural claim.

It investigates a particular aspect of human auditory perception.

The discovery is useful because it helps explain why two people can sincerely disagree about an ambiguous recording.

A person can genuinely believe they heard a word even when the available evidence does not establish that the word was spoken.

Professor Parallax prepared three recordings.

"First," he said, "we shall hear a clear example of speech."

A voice emerged from the speakers.

"Good evening."

White Rabbit smiled.

"Easy."

Hatta bowed.

"An excellent greeting."

The Professor activated the second recording.

A faint, heavily distorted voice sounded beneath crackling noise.

The Rabbit leaned forward.

"I think I heard something."

"What?"

"Perhaps 'good morning.'"

Hatta shook his head.

"Clearly 'more pudding.'"

Parallax activated the third recording.

It contained only an ambiguous mixture of sound.

Hatta listened carefully.

"There! I heard it!"

"What?" asked the Rabbit.

"'Hatta is magnificent.'"

The Professor turned.

"Did you hear that before you knew what you wanted the recording to say?"

Hatta considered.

"Professor, you are asking exceedingly personal questions about my expectations."

The experiment illustrated three possibilities.

A recording may contain intelligible speech.

It may contain real speech that is difficult to recognize.

Or it may contain non-speech sound that listeners interpret as words.

Those are different situations.

A scientific investigation must determine which one the evidence supports.

And now artificial intelligence introduces another complication.

Modern artificial intelligence systems can perform sophisticated audio analysis.

Depending upon their design, they may detect speech.

Recognize words.

Separate voices from background noise.

Identify speakers.

Or generate transcripts.

These capabilities can be extremely useful.

But a system's output is not automatically an accurate record of what a person said.

Speech-recognition systems are trained using examples of audio and corresponding language.

They learn relationships between acoustic patterns and words.

When presented with new recordings, they estimate likely text.

But what happens if the input contains highly degraded speech?

Or unintelligible sounds?

Or no speech at all?

The system may fail to produce a useful transcript.

Some models may also produce mistaken words that seem plausible in context.

In certain circumstances, speech-recognition systems can generate apparent text that is unsupported by the actual audio.

That phenomenon is especially concerning when the recording is ambiguous.

A transcript may appear authoritative.

Yet it may reflect the model's learned expectations more than any genuine spoken message.

A machine-generated transcript must not be treated as proof that a voice was present.

The underlying recording remains the evidence.

The transcript is an interpretation.

And its reliability must be assessed.

Hatta had been examining Professor Parallax's speech-recognition software.

"Professor, I have made an extraordinary discovery."

"What discovery?"

"This machine can transcribe voices."

"Yes."

"And you have a library of suspicious recordings."

"Yes."

"And therefore..."

Hatta paused dramatically.

"I have established the world's first automated ghost-detection bureau."

White Rabbit looked alarmed.

"What does it do?"

"It analyzes recordings and prints whatever the ghosts are saying."

Parallax approached the computer.

"Hatta, have you tested it against recordings known to contain no speech?"

"Certainly."

"And what happened?"

"It produced seven messages."

"From pure noise?"

"Apparently very talkative noise."

"What messages?"

Hatta unfolded a sheet.

"One requested tea."

"Of course."

"Another asked for a larger hat."

"Naturally."

"And the third announced that my musical talent was tragically underappreciated."

The Rabbit folded his arms.

"Hatta, you entered suggestions into the system, didn't you?"

Hatta looked offended.

"Only to provide context."

Parallax examined the settings.

He found a prompt instructing the software to interpret every recording as a message from a mysterious speaker.

"Hatta, you've biased the analysis."

"Improved the hospitality."

"No. You've encouraged it to invent messages."

Hatta considered.

"Then perhaps we should rename the bureau."

He crossed out the title.

HATTA'S AUTOMATED GHOST DETECTION BUREAU

And wrote:

HATTA'S AUTOMATED GHOST INTERPRETATION AND OCCASIONAL INVENTION BUREAU.

White Rabbit sighed.

"That's at least more accurate."

Friday, October 30, has a particularly appropriate place in broadcasting history.

On this date in 1938, Orson Welles and the Mercury Theatre on the Air presented a radio adaptation of H. G. Wells's The War of the Worlds.

The production used fictional news-style bulletins and dramatic interruptions to depict a Martian invasion.

The format created a striking sense of immediacy.

Some listeners who encountered the broadcast without its opening context misunderstood portions of the program.

But the familiar story that the broadcast caused widespread nationwide panic has been significantly exaggerated.

Historical investigations have questioned the scale of the supposed mass hysteria.

The program demonstrated how presentation conventions shape the way audiences interpret information.

A fictional announcement can seem more convincing when delivered in a format associated with genuine news.

A dramatic recording can create a powerful impression.

And listeners may bring their own assumptions into the experience.

The lesson applies equally to modern synthetic media.

A believable voice.

An urgent announcement.

A realistic soundtrack.

Or a fabricated eyewitness account.

These elements may cause people to accept a story before verifying its source.

Sound can create an impression of authenticity. Authenticity still requires evidence.

On the eve of Halloween, it is worth remembering that some of the most compelling audio illusions do not require anything supernatural.

They require talented performers, persuasive techniques, and a receptive audience.

Why do expectations matter so much?

Because perception is not simply a mechanical recording of the external world.

The brain continuously interprets incomplete and changing information.

It uses prior experience.

Context.

Recognition.

And learned patterns.

Suppose you hear a faint sound late at night.

If you believe a friend is arriving, the sound may resemble footsteps.

If you believe the building is haunted, you might interpret the same noise very differently.

The underlying acoustic event may remain unchanged.

But the meaning you assign to it can shift.

One influential framework in cognitive science describes perception as involving ongoing predictions about incoming sensory information.

The brain does not merely receive signals.

It also uses expectations to help interpret them.

When the signal is strong and clear, incorrect expectations may be corrected.

When the signal is ambiguous, interpretation can become more uncertain.

This does not mean people simply choose what they hear.

Much of perception operates automatically.

Nor does it mean that every unusual auditory experience has the same explanation.

It means that expectations are among the factors researchers must consider when investigating what people perceive.

And the same general concern arises when AI models analyze ambiguous data.

Yes.

But the methods matter.

Imagine receiving a recording that supposedly contains a mysterious voice.

A careful investigation might proceed through several stages.

Keep an untouched copy of the recording.

Do not replace it with a heavily processed version.

Where was the microphone placed?

What equipment was used?

Was there a nearby radio, television, telephone, ventilation system, or other source of sound?

Could electrical interference have affected the recording?

Use appropriate tools to inspect the waveform and frequency content.

Look for patterns consistent with speech, mechanical noise, or recording artifacts.

Allow listeners to hear the recording without telling them what words they are expected to identify.

If different people hear different phrases, that disagreement is relevant evidence about ambiguity.

Speech-detection or transcription models may contribute useful information.

But their results should be treated as hypotheses requiring verification.

An unexplained sound is not the same as a demonstrated supernatural message.

Failure to identify a source does not establish that no ordinary source exists.

And a confident transcript does not establish that the audio contained the words.

A responsible investigator should be willing to conclude that the recording remains uncertain.

Remember our earlier expedition into historical audio restoration?

We learned that excessive processing can change a recording.

The same principle applies here.

Suppose a faint sound occurs beneath heavy static.

An AI-enhancement system attempts to make it clearer.

The processed version may become easier to listen to.

But it may also contain artifacts.

Some original frequencies may be suppressed.

Other patterns may be exaggerated.

And certain generative tools may introduce details that were not reliably present in the source.

This creates a serious interpretive danger.

A heavily enhanced version may appear to contain recognizable speech.

Yet the apparent words might partly reflect the processing method rather than the original recording.

Was the sound actually preserved in the original?

Or did the model produce a plausible approximation?

For creative projects, generating an atmospheric ghost voice can be perfectly legitimate.

For historical, journalistic, or scientific claims, the standard must be different.

A dramatized reconstruction should be labeled.

An enhanced recording should be distinguished from its original.

And a synthetic voice should not be presented as authentic evidence of a real speaker.

Improving the experience of listening is not necessarily the same as improving the reliability of the evidence.

For fiction and entertainment, modern generative audio tools offer remarkable possibilities.

A creator can develop eerie whispers.

Echoing voices.

Unusual vocal textures.

Distant footsteps.

Creaking doors.

Or an entire supernatural soundscape.

The results can be extraordinarily convincing.

Yet their realism does not make the fictional events real.

This distinction matters even more as generative audio becomes increasingly sophisticated.

A voice can sound familiar without belonging to the apparent speaker.

A recording can seem old without having been made decades ago.

And a fictional conversation can be presented with all the acoustic qualities of documentary sound.

Imagine producing an original Halloween audio drama.

A mysterious mansion.

A forgotten radio broadcast.

A recording found in an abandoned archive.

A character who hears a whisper no one else can identify.

AI tools can assist with sound effects, voice performances, and musical atmosphere.

But the strongest creative work still requires thoughtful direction.

What should the audience believe at a particular moment?

Which details should remain mysterious?

When should information be revealed?

And how can suspense develop without confusing fictional storytelling with genuine evidence?

The purpose of fictional horror is to create an experience. The purpose of investigation is to determine what actually happened.

Those aims can coexist.

They simply should not be confused.

Professor Parallax returned to his blackboard.

He wrote:

DEMONSTRATED. DEVELOPING. NOT ESTABLISHED.

Under DEMONSTRATED, he listed:

Under DEVELOPING:

Under NOT ESTABLISHED:

White Rabbit examined the list.

"Professor, so it's possible to hear a voice that wasn't actually spoken?"

"Yes, under some conditions."

"And a computer can make a similar kind of mistake?"

"It can produce unsupported interpretations through different computational mechanisms."

"And that doesn't mean either the person or machine is deliberately lying?"

"Correct."

Hatta raised his hand.

"Professor?"

"Yes?"

"Could a ghost nevertheless be extremely good at hiding from scientific instruments?"

"That is not a hypothesis we can test merely by inventing explanations for every failure to find evidence."

Hatta sighed.

"An unusually demanding profession, this science."

The laboratory had grown quiet.

Outside, the evening wind moved through the trees.

Professor Parallax switched off the main lights.

White Rabbit carefully labeled the recordings.

Hatta approached carrying a small speaker.

"Professor, I have prepared one final experiment."

Parallax looked suspicious.

"What kind of experiment?"

"A scientific demonstration of auditory expectation."

The Rabbit raised his eyebrows.

"Have you followed the proper procedures?"

"Almost certainly."

Hatta placed the speaker upon a table.

"First, gentlemen, I want you to listen to this recording without knowing what it contains."

He pressed PLAY.

A faint hiss emerged.

Then a series of clicks and mechanical vibrations.

Professor Parallax listened.

White Rabbit tilted his head.

The recording ended.

"What did you hear?" asked Hatta.

"Background noise," said the Professor.

"Some irregular mechanical sounds," added the Rabbit.

Hatta smiled.

"Excellent. Now I shall provide a suggestion."

He held up a card.

It read:

THE GHOST WHISPERS: HATTA DESERVES MORE PUDDING.

The Rabbit groaned.

"Hatta!"

"Please. This is an experiment."

Hatta played the recording again.

The same hiss.

The same clicks.

The same mechanical vibrations.

White Rabbit looked startled.

"Professor..."

"Yes?"

"I think I can almost hear the words now."

Hatta beamed.

"SUCCESS!"

"Not so fast," said Parallax.

He examined the speaker.

Then traced its cable to a small recorder.

The Professor pressed a button.

The original recording stopped.

But another sound continued.

A faint, synthesized whisper.

More pudding...

Parallax turned toward Hatta.

"You added a second audio track."

Hatta looked wounded.

"Only a little."

"Then you've changed the stimulus!"

"I've enhanced the ghost."

"You've invented the ghost."

White Rabbit began laughing.

Hatta sighed.

"Science refuses to appreciate theatrical improvement."

The Professor removed the synthetic track.

He replayed the genuine original.

This time, the laboratory fell silent.

Then the Rabbit spoke.

"Professor, I have a question."

"Yes?"

"If someone genuinely believes they heard something mysterious, what should we tell them?"

Parallax considered.

"We should listen carefully to what they experienced."

"And then?"

"Investigate the evidence respectfully."

"And if we cannot determine what caused the sound?"

"Then we say we do not know."

Hatta looked toward the darkened window.

"An honest answer."

"Often the best one."

The wind moved through the trees again.

The old coat rack creaked.

Nobody spoke.

Then Hatta whispered:

"More pudding."

And all three burst into laughter.

For centuries, human beings have heard mysterious sounds.

Footsteps in empty corridors.

Whispers in the wind.

Voices in running water.

And apparently meaningful words hidden within static.

Some of these experiences have ordinary explanations.

Others remain uncertain because the available evidence is incomplete.

Science can investigate how people perceive ambiguous sounds.

It can test whether expectations influence what listeners report.

It can examine the acoustic structure of recordings.

And artificial intelligence can assist with pattern recognition and signal analysis.

But powerful technology also introduces new risks.

A model may generate a confident interpretation of uncertain input.

An enhancement system may create artifacts.

And a synthetic voice may convincingly imitate a person who never spoke the words.

The challenge is therefore not simply to develop machines capable of hearing more.

It is to develop methods for determining what the evidence actually supports.

We must distinguish sounds from interpretations.

Interpretations from verified speech.

And creative invention from authentic documentation.

Yet none of this diminishes the wonder of the human auditory imagination.

The brain's remarkable ability to recognize patterns helps us understand language, music, and our surroundings.

Sometimes that same ability leads us astray.

And perhaps that is one reason we remain so fascinated by mysterious voices.

We are creatures who search for meaning.

Even in the darkness.

Even in uncertainty.

Even in the faintest whisper.

So the question at the bottom of today's Rabbit Hole is not merely:

It is:

Somewhere in a quiet room, an audio recorder captures the faint hum of an electrical appliance.

A computer examines the signal.

A listener hears a possible whisper.

Another listener hears nothing resembling speech.

The sound itself has not changed.

The interpretations have.

And perhaps the most important discovery is not a voice hidden within the recording.

It is the extraordinary power of the mind to find meaning, even when meaning remains uncertain.

🐰🕳️♾️

Have you ever heard a familiar word, voice, or melody hidden in an unrelated sound?

Perhaps in running water?

A mechanical hum?

An old recording?

Or a noisy radio signal?

And here's our Halloween question:

If three people listen to the same mysterious recording and hear three different sentences, what would convince you that one interpretation is correct?

Would you trust the majority?

An AI transcript?

A scientific acoustic analysis?

Or would you want stronger evidence before deciding?

Tell us what you think.

Hatta has offered to analyze your recordings.

Unfortunately, his machine believes every unexplained sound contains a request for pudding.

1. Szymanek, P., Homan, M., van Elk, M., and Hohol, M. (2024). "Effects of Expectations and Sensory Unreliability on Voice Detection: A Preregistered Study." Consciousness and Cognition, 123, 103718.

Peer-reviewed experimental research demonstrating the influence of expectations on voice-detection judgments, particularly when acoustic evidence is unreliable.

https://doi.org/10.1016/j.concog.2024.103718

2. Nees, M. A., and Phillips, C. (2015). "Auditory Pareidolia: Effects of Contextual Priming on Perceptions of Purportedly Paranormal and Ambiguous Auditory Stimuli." Applied Cognitive Psychology.

Research investigating the influence of paranormal framing on reports of voices in ambiguous recordings.

https://doi.org/10.1002/acp.3068

3. Williams, J. M., and Blagrove, M. (2022). "Paranormal Experiences, Sensory-Processing Sensitivity, and the Priming of Pareidolia." PLOS ONE.

A follow-up investigation of auditory pareidolia and paranormal framing, reporting nuanced findings and limitations on replication.

https://doi.org/10.1371/journal.pone.0274595

4. Laloyaux, J., and colleagues (2022). "Eliciting False Auditory Perceptions Using Speech Frequencies and Semantic Priming: A Signal Detection Approach." Cognitive Neuropsychiatry.

Research on how speech-related noise and semantic expectations interact in auditory perception.

https://doi.org/10.1080/13546805.2022.2031945

5. Alderson-Day, B., and colleagues (2022). "Susceptibility to Auditory Hallucinations Is Associated with Spontaneous but Not Directed Modulation of Top-Down Expectations for Speech." Neuroscience of Consciousness.

Research examining the relationship between prior knowledge, ambiguous acoustic information, and speech perception.

https://doi.org/10.1093/nc/niac002

6. Smithsonian Magazine. Historical coverage of Orson Welles and the 1938 War of the Worlds broadcast.

Context for the use of radio-news conventions in dramatic fiction and the distinction between documented listener confusion and exaggerated accounts of nationwide panic.

https://www.smithsonianmag.com/

7. Library of Congress, National Recording Preservation Program.

Resources about preserving authentic recordings, investigating audio history, and maintaining access to original recorded materials.

https://www.loc.gov/programs/national-recording-preservation-board/

Research note: Studies of auditory pareidolia investigate perception and expectations under experimental conditions. They do not establish the existence of supernatural speakers. AI-generated transcripts and enhanced recordings must be evaluated against original audio evidence. The laboratory scenes and conversations involving Professor Parallax, White Rabbit, and Hatta are fictional educational dramatizations.

AI RABBIT HOLES | EXPEDITION #155

Where curiosity listens in the darkness, evidence separates whispers from imagination, and Hatta's ghost-detection bureau remains suspiciously well supplied with pudding.

#AIRabbitHoles #ArtificialIntelligence #AuditoryPareidolia #Halloween2026 #GhostVoices #AudioScience #CognitiveScience #Psychology #AITranscription #SyntheticAudio #MediaLiteracy #WarOfTheWorlds #OrsonWelles #ProfessorParallax #WhiteRabbit #Hatta #Wondermage

📜White Rabbit 🐰 • Hatta 🎩 • Professor Parallax 🔭 • Q~ ♾️👈

Sometimes the strangest language is another way of seeing the world. 🌎👀

💡❔ Where curiosity goes slightly sideways, then comes back carrying a lantern. 🏮

🐰🕳️ Follow the White 🐰 Rabbit: AIRabbitHoles.com

🟨 Walk the Road: YellowBrickRoadtoAI.com

© 2026 AIRabbitHoles.com 🐰🕳️🕳️🕳️🕳️🕳️