Terms of Use

CROCUFID food and object pictures

Standardized images were taken from the CROCUFID database available from the OSF repository at https://osf.io/5jtqx.

Used in the Food-Picture Dot-Probe, the Food Cue Reactivity Task, the Evaluative Conditioning Task, the Food Choice Task and the Two-Tone Pictures Task. ConductCognition resized the pictures to 600 × 400 pixels (JPEG) and made no other changes to them. The Two-Tone Pictures Task shows changed versions of 13 of them, made by code: each food blurred and turned into black and white patches in a frame of 480 × 370 pixels, some also warped so that the patches no longer form a food.

Citation: Toet A, Kaneko D, de Kruijf I, Ushiama S, van Schaik MG, Brouwer AM, Kallen V, van Erp JBF. CROCUFID: A Cross-Cultural Food Image Database for Research on Food Elicited Affective Responses. Front Psychol. 2019;10:58. doi:10.3389/fpsyg.2019.00058

License: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/)

Face Research Lab London Set face photographs

Face photographs: Face Research Lab London Set by Lisa DeBruine and Benedict Jones (figshare, https://doi.org/10.6084/m9.figshare.5047666), CC BY 4.0.

Used in the Face Memory Test, the Face-name Memory Task, the Composite Face Task, the Choice Blindness Task, the Face Matching Task, the Reverse Correlation Task, the Chimeric Faces Task, the Part-Whole Face Task, the Thatcher Illusion Task, the Two-Tone Pictures Task, the Verbal Overshadowing Task, the Face Impressions Task, the Real or AI Face Task, the Uncanny Valley Task, the Face Aftereffect Task and the Group Faces Task. ConductCognition aligned each photo on the eyes, cropped it to an oval on black (280 × 320 pixels, JPEG), and made some grayscale with matched brightness and contrast. The Reverse Correlation Task's base face is an average of these photos, the Chimeric Faces Task's pictures join halves of two photographs, the Part-Whole Face Task's composites join one person's face outline with the eyes, nose and mouth of two others, the Thatcher Illusion Task's changed pictures turn the eyes and mouth upside down within the face, the Two-Tone Pictures Task's pictures turn blurred three-quarter views into black and white patches, some also warped so that the patches no longer form a face, the Verbal Overshadowing Task adds grayscale copies of 13 color front photographs, and the Face Aftereffect Task expands or contracts the central features of six grayscale front photographs by fixed amounts, saved as PNG files. The Group Faces Task shows a grayscale front photograph alone or beside two others, unchanged. While a test runs, it may add visual noise to a picture or join the top half of one face to the bottom half of another.

Citation: DeBruine L, Jones B. Face Research Lab London Set. figshare. Dataset. 2017. doi:10.6084/m9.figshare.5047666.v5

License: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/)

Generated face pictures

Fictional adult portraits generated by ConductScience with Stable Diffusion 1.5, revision 451f4fe16113bff5a5d2269ed5ad43b0592e9a14 (https://huggingface.co/stable-diffusion-v1-5/stable-diffusion-v1-5/tree/451f4fe16113bff5a5d2269ed5ad43b0592e9a14), under CreativeML OpenRAIL-M.

Used in the Real or AI Face Task and the Uncanny Valley Task. ConductScience generated the portraits locally on a Windows CPU with the safety checker enabled, without real-person names or reference photographs. Each portrait was aligned on the eyes, converted to grayscale, matched for brightness and contrast, and cropped to an oval on black (280 × 320 pixels, JPEG). The Uncanny Valley Task also uses flat stylizations and blends made locally from those portraits. London Set photographs appear separately and enter no generated portrait or blend. The task end screens identify generated pictures. The model terms, generation details and changes are retained at https://conductcognition.com/stimuli/generated-faces/NOTICE.txt and https://conductcognition.com/stimuli/generated-faces/MODEL-LICENSE.txt.

Citation: ConductScience. ConductCognition generated faces 1.0.0. Stable Diffusion 1.5, revision 451f4fe16113bff5a5d2269ed5ad43b0592e9a14.

License: CreativeML OpenRAIL-M (https://conductcognition.com/stimuli/generated-faces/MODEL-LICENSE.txt)

Noto Emoji pictures

Emoji pictures: Noto Emoji by Google, from the GitHub repository at https://github.com/googlefonts/noto-emoji.

Used in Emoji Emotion Recognition, the Picture Naming Test, the Director Task, the Self-Ordered Pointing Task, the Pavlovian-Instrumental Transfer Task, the Directed Forgetting Task, the Go/No-Go Association Task, the Implicit Relational Assessment Task, the Size Order Memory Test, the Picture Order Memory Test, the Mnemonic Similarity Task, Evaluative Priming, the Willingness to Pay Task, the Think and No-Think Task, the Associative Recognition Test, the Brief Implicit Association Test, the Two-Step Task, the Object Location Memory Task, the Remember and Know Test, the Perspective Pointing Test, the Affect Misattribution Task, the Misinformation Effect Task, the Repetition Priming Task, the Triplet Odd-One-Out Task, the Keep-Track Task, the Picture Superiority Task, the Change Blindness Task, the Source Memory Task, the Semantic Triads Task, the Concealed Information Test, the Picture Matching Task, the Emotional Face Matching Task, the Slips-of-Action Task, the Repeat Detection Task, the Helicopter Task, the Acquired Equivalence Task, the Gaze and Expression Theory of Mind Task, the Real or Made-Up Animal Task, the Overlapping Outlines Task, the Hindsight Bias Task, the Structural Priming Task, the Decoy Choice Task, the Endowment Effect Task, the Cue Blocking Task, the Mouse-Tracking Choice Task, the Associative Inference Task, the Sensory Preconditioning Task, the Memory Conformity Task, the Preference Change Task, the Planning Fallacy Task, the Zeigarnik Effect Task, the Serial Naming Task, the Mental Scanning Task, the Relational and Item Encoding Task, the Attentional Boost Task, the Grocery Trip Game, the Picture Partners Game and the Week Later Game. ConductCognition made no changes to the pictures (512 × 512 pixels, PNG). The Real or Made-Up Animal Task shows silhouettes, and the Overlapping Outlines Task outlines, that ConductCognition traced from some of the pictures by code: changed versions of those pictures. The Planning Fallacy and Zeigarnik Effect tasks show each picture in a frame cut into 3 by 3 tiles for a picture puzzle.

Citation: Google. Noto Emoji. GitHub repository, commit 41e31b110b4eb929dffb410264694a06205b7ad7, 15 February 2024.

License: Apache License 2.0 (https://www.apache.org/licenses/LICENSE-2.0)

OASIS affective pictures

Pictures: Open Affective Standardized Image Set (OASIS) by Benedek Kurdi, Shayn Lozano and Mahzarin R. Banaji (https://benedekkurdi.com/oasis.php), photographs from Pixabay.

Used in the Emotion Regulation Task, which shows 33 of the set's photographs: scenes such as rubbish, insects, floods, fire, storms and prison cells, which are at most moderately unpleasant, and everyday objects. The set's authors offer its pictures free of charge for research, and OASIS does not own them: each keeps the terms of its own source. ConductCognition uses only pictures whose source is Pixabay, each published there before 9 January 2019, which Pixabay's terms call CC0 Content, for all personal and commercial purposes with no credit required. Each picture's Pixabay page, published date and uploader are listed in the pictures' notice at https://app.conductcognition.com/stimuli/oasis/NOTICE.txt. ConductCognition re-saved each 500 × 400 pixel OASIS picture as a JPEG and made no other changes.

Citation: Kurdi B, Lozano S, Banaji MR. Introducing the Open Affective Standardized Image Set (OASIS). Behav Res Methods. 2017;49(2):457-470. doi:10.3758/s13428-016-0715-3

License: CC0 Content under Pixabay's terms (https://pixabay.com/service/terms/)

Unsplash scene photographs

Scene photographs: Unsplash photographs under CC0, from Wikimedia Commons (https://commons.wikimedia.org).

Used in the Boundary Extension Task, which shows 21 photographs of everyday scenes with no person in them, such as a bench before a hedge, two chairs on a lawn, a teapot on a balcony and a tire swing over tall grass: 20 in the test and 1 as its example. Each was published on Unsplash before 5 June 2017, while Unsplash still released its photographs under the Creative Commons CC0 1.0 Universal Public Domain Dedication, which lets anyone copy, change and share them, even for commercial purposes, without asking permission, and was copied to Wikimedia Commons, whose file page for it carries that dedication. Each photograph's Commons and Unsplash pages, photographer and earliest archived Unsplash page are listed in the pictures' notice at https://app.conductcognition.com/stimuli/unsplash-scenes/NOTICE.txt. ConductCognition cut each photograph to its widest 4:3 frame, resized it to 1024 × 768 pixels and saved it as a JPEG with no metadata. The test shows each file whole as its wide view and its middle enlarged 1.2 times as its close view: changed versions of the photographs.

Citation: Wikimedia Commons. Unsplash photographs taken from 2012 to 2017, by the photographers named in the pictures' notice.

License: CC0 1.0 Universal Public Domain Dedication (https://creativecommons.org/publicdomain/zero/1.0/)

Pixabay films

Films: Pixabay CC0 Content by XiaosKitchen, padrinan and AlekseiyPovar, from https://pixabay.com.

Used in the Event Segmentation Task, which plays four films of hands at everyday tasks, with no face shown: making a spring roll, folding a paper boat and finishing a stew in the test, and folding a paper plane for practice. Each was published on Pixabay before 9 January 2019, which Pixabay's terms call CC0 Content, for all personal and commercial purposes with no credit required. Each film's Pixabay page, published date and uploader are listed in the films' notice at https://app.conductcognition.com/stimuli/pixabay-films/NOTICE.txt. ConductCognition made no changes to the films: each is Pixabay's smallest version of the film (640 × 360 pixels, H.264 video in MP4, with no sound track), played muted.

Citation: Pixabay videos 11641 (XiaosKitchen, 2017), 9082 and 9081 (padrinan, 2017) and 15736 (AlekseiyPovar, 2018).

License: CC0 Content under Pixabay's terms (https://pixabay.com/service/terms/)

CREMA-D emotional speech recordings

Contains information from CREMA-D, the Crowd-sourced Emotional Multimodal Actors Dataset (https://github.com/CheyneyComputerScience/CREMA-D), which is made available under the Open Database License (ODbL).

Used in the Vocal Emotion Task, which plays 54 of its clips: recordings of actors speaking English sentences with anger, disgust, fear, happiness, sadness or no emotion, 9 clips for each. ConductCognition chose the clips by a fixed rule, filtered out sound below 60 Hz, trimmed each clip to its speech, set all of them to one speech level and saved them as FLAC. This 54-clip subset, with its rating fields, stays under the Open Database License 1.0, and its contents under the Database Contents License 1.0 (https://opendatacommons.org/licenses/dbcl/1-0/). The Vocal Emotion Task is a Produced Work of it.

Citation: Cao H, Cooper DG, Keutmann MK, Gur RC, Nenkova A, Verma R. CREMA-D: Crowd-sourced Emotional Multimodal Actors Dataset. IEEE Trans Affect Comput. 2014;5(4):377-390. doi:10.1109/TAFFC.2014.2336244

License: Open Database License 1.0 (https://opendatacommons.org/licenses/odbl/1-0/)

Synthetic speech

Speech: ConductCognition's own texts, spoken by open text-to-speech voices, each listed with its terms and training data in the clips' notice at https://app.conductcognition.com/stimuli/speech/NOTICE.txt.

Used in the Dichotic Listening Task, the Digits-in-Noise Task, the Listening Span Task, the Dual N-Back Task, the Cross-Modal Priming Task, the Categorical Perception Task, the Speech-to-Song Task, the Phoneme Monitoring Task, the Phonemic Restoration Task, the Lexical Tone Task, the Spelling to Dictation Task and the Voice Identity Task. These voices are synthetic: ConductCognition wrote the digits, sentences, words and phrases, and open text-to-speech models spoke them, one voice for each language and two for Spanish. Each voice was trained on recordings in the public domain or under an open license: English, the LJ Speech dataset (public domain); Spanish, carlfm01's speech dataset (public domain) for the sentences and the CSS10 LibriVox readings by Tux (public domain) for the digits; French, the CSS10 LibriVox readings by Gilles G. Le Blanc (public domain); German, the Thorsten-Voice dataset (CC0); Italian, Riccardo Fasol's LibriVox readings in the M-AILABS Speech Dataset (public domain, with the dataset's own license); Portuguese, one speaker of Common Voice (CC0); Chinese, the AISHELL-3 corpus (Apache 2.0); Japanese, ekzemplaro's LibriVox readings of Aozora Bunko texts in the Kokoro Speech Dataset (public domain). The voice models are under the MIT, Apache 2.0 and BSD-3-Clause licenses. The Voice Identity Task adds six voices in each language, four for the test and two for practice: rule-based formant voices of eSpeak NG 1.52.0, with no recording, voice model or imitated person. eSpeak NG is under GPL-3.0-or-later, which does not cover the speech it outputs. The Dichotic Listening Task's six syllables and the Categorical Perception Task's nine steps from ba to da were made by ConductCognition's own formant synthesis after Klatt (1980), with no voice model or recording, and the Digits-in-Noise Task's noise from the long-term spectrum of each language's digits or sentences. The Phonemic Restoration Task lays white noise made by ConductCognition's code over one sound of each spoken word, or puts the noise in that sound's place: changed versions of those clips.

Citation: Klatt DH. Software for a cascade/parallel formant synthesizer. J Acoust Soc Am. 1980;67(3):971-995.