Online Experiments for Language Scientists

Academic year 2022-2023

For non-Edinburgh people: all the course materials are here and you are welcome to use them with attribution. If you would like us to run something like this for your students or research group, contact Kenny.

This is the webpage for the Honours/MSc course Online Experiments for Language Scientists, running in academic year 2022/2023. I will add links to materials (readings, code) to this page; you will need to use Learn for electronic submission of your assessed work, and to keep an eye on announcements.

Course summary

Many areas in the language sciences rely on collecting data from human participants, from grammaticality judgments to behavioural responses (key presses, mouse clicks, spoken responses). While data collection traditionally takes place face-to-face, recent years have seen an explosion in the use of online data collection: participants take part remotely, providing responses through a survey tool or custom experimental software running in their web browser, with surveys or experiments often being advertised on crowdsourcing websites like Amazon Mechanical Turk (MTurk) or Prolific. Online methods potentially allow rapid and low-effort collection of large samples (and are also useful in situations where face-to-face data collection is not possible, e.g. during a pandemic); however, building and running these experiments poses challenges that differ from lab-based methods.

This course will provide a rapid tour of online experimental methods in the language sciences, covering a range of paradigms, from survey-like responses (e.g. as required for grammaticality judgments) through more standard psycholinguistic methods (button presses, mouse clicks) up to more ambitious and challenging techniques (e.g. voice recording, real-time interaction through text and/or streaming audio, iterated learning). Each week we will read a paper detailing a study using online methods, and look at code (written in javascript using jsPsych) to implement a similar experiment - the examples will skew towards the topics I am interested in (artificial language learning, communication, language evolution), but we’ll cover more standard paradigms too (grammaticality judgments, self-paced reading) and the techniques are fairly general anyway. We’ll also look at the main platforms for reaching paid participants, e.g. MTurk and Prolific, and discuss some of the challenges around data quality and the ethics of running on those platforms.

No prior experience in coding is assumed, but you have to be prepared to dive in and try things out; the assessment will involve elements of both literature review and coding.

The teaching team

The course is co-taught by Kenny Smith and Alisdair Tullo. Kenny (that’s me) is the main lecturer and the course organiser; Alisdair is the PPLS javascript/jsPsych guru and delivers the lab sessions with Kenny. Best way to get in touch with us is in one of the live sessions, see below, or by email to kenny.smith@ed.ac.uk or alisdair.tullo@ed.ac.uk.

We’ll also be supported in lab classes by three excellent tutors: Aislinn Keogh, Vilde Reksnes, and Maisy Hallam.

Class times

The course runs in semester 1. We have lectures 9am-9.50am on Mondays, and lab classes 9am-10.50am on Wednesdays. There are also extra drop-in labs available, see below.

Lectures and labs are both essential to doing well on the course - the assessment involves an understanding both of the literature on online experiments (covered in the readings and lectures) and the practicalities of how to build them (covered in your own work on the practicals, with support available in the labs).

Arrangements for lectures and labs

Lectures take place in person on Monday mornings, 9am-9.50am, in room S1, 7 George Square - as far as I can work out, S indicates Second floor. Labs will take place in on Wednesday mornings, 9am-10.50am, in the MacLaren Stuart room, Old College - this is in the north east corner of the old College quad, go through the door marked “Edinburgh School of Law and the MacLaren Stuart is immediately on your left.

Hopefully you are all familiar with attending in-person classes in covid times, but just as a reminder:

If you are unable to attend in-person lectures or labs due to illness, please see the course page on Learn for instructions on how to participate remotely. Lectures will be streamed live and there is a channel for asking questions in real time. We will use Teams for remote lab attendees, and assign a tutor to monitor remote attenders.

Assessment

There are two pieces of assessment, due on 10th November and 8th December. Assessment 1 is an annotated bibliography reviewing and evaluating 4 articles typically drawn from the course set readings. Assessment 2 is a project where you produce a working experiment implemented in jsPsych and an accompanying report explaining the motivation behind that experiment, justifying important design decisions you took in building the experiment, and appraising the experiment and ways it could be improved/extended. Full details are provided in the assignment brief, and the FAQ (which also features examples of good assignments).

Course Materials

Course content will appear here as we work through the course.

Each week there will be a set reading and a programming assignment. The reading involves a blog post introducing a published paper, you read both the blog and the paper, the lecture then provides an additional brief overview and an opportunity to ask questions/discuss the reading. The programming assignment involves working through a section of the Online Experiments with jsPsych tutorial and/or looking at (and editing) some code which implements a language-related experiment; you can use the lab classes as dedicated time to work on the programming task and get help with programming difficulties or questions you have.

Week 1 (commencing 19th September): Introduction

NB Because of the University closure on 19th September for the Queen’s funeral, there is no Monday lecture; instead we will run a combined lecture/lab in the Wednesday slot, 9am-10.50am, in the MacLaren Stuart room.

Week 2 (26th September): Crowdsourcing experimental data

Week 3 (3rd October): Grammaticality judgements

Week 4 (10th October): Self-paced reading

Week 5 (17th October): Lab catchup week

In week 5 there is no lecture, but we have a lab as usual - use this time to catch up on labs from previous weeks, and/or to catch up on / get ahead with your reading.

Week 6 (24th October): Word learning / frequency learning

Week 7 (31st October): Audio stimuli

Week 8 (7th November): Priming and overspecification

Week 9 (14th November): Iterated Learning

Week 10 (21st November): Participant-to-participant interaction

Week 11 (28th November): Interacting with MTurk

No lecture or lab in week 11, but there are some materials that will be useful for you to read if you are thinking of setting up a real experiment in the wild!

Bonus content

I am sticking some extra documented experiments I have created here, in case they are useful for someone or provide inspiration.

Additional drop-in labs

If you have questions and don’t get a chance to ask them in labs, you can turn up at one of our additional drop-ins, which start in week 5 (week commencing 17th October, until 9th December).

Note that these are not compulsory, and they are drop-ins not extra labs - the idea is that you come along, ask a couple of questions, then go away.

Additional drop-in labs for questions on Assessment 2 code

We will provide some extra drop-in labs to give you an opportunity to get some help with your Assignment 2 code. Obviously we won’t write your code for you, but if you are having trouble interpreting an error message or finding a bug or want some tips on how to achieve a particular effect we can help you figure it out. Note that these are not compulsory, and they are drop-ins not extra labs - the idea is that you come along, ask a couple of questions, then go away. These are available at the following times and locations:

Re-use

All aspects of this work are licensed under a Creative Commons Attribution 4.0 International License.


Course main page

Project maintained by kennysmithed Hosted on GitHub Pages — Theme by mattgraham