Aaron Ecay’s website

Two phases of do auxiliaries in Early Modern English: Evidence from priming data

2018-08-27T00:00:00+01:00

I presented a talk at the 20th International Conference on English Historical Linguistics about the patterns of priming present in Early Modern English auxiliary do usage (with special reference to do in affirmative declaratives).

Link to the slides

Questions for aspiring temporary lecturers

2017-07-26T00:00:00+01:00

Iʼve been thinking lately about temporary teaching jobs. In general, academic departments are built around the assumption that affiliations will be long-lasting and have a personal component. Thus, little attention is paid to the situation of temporary workers vis a vis departmental and university resources: in HR-speak, “onboarding.” Iʼve had mixed experiences with this aspect of jobs, and so I set out to write a list of questions that I think itʼs important for new temporary teachers to have answers to. I take as the paradigmatic case a fixed-term contract to teach one module in linguistics for one term at a UK university, because those are the parameters that most closely match my own experiences and those I know enough about to extrapolate from. Of course, if your own situation is different you might want to change some things, but I feel the principles should hold up fairly broadly.

Before getting on to the list, I should make two things clear. The first is that this post is absolutely not a criticism of my colleagues (or students) who have participated in the experiences underlying this list. I wrote above that the necessity of a list like this comes from an ongoing reorganization of the fundamentals of the academic labor market. This affects both the teachers as well as the departments that use their services.

There are always lessons to be learned in life, but in the context Iʼm writing about my constant experience has been of good and passionate people doing the best that they can in challenging circumstances that have been imposed on them. Some of them have in fact gone far beyond what in turn could be expected of them to help me avoid the sharp edges of many of these questions.

Secondly, what is the purpose of this list? Most obviously, it might help people who are early in their careers and faced with the prospect of one of these jobs to prepare themselves for the demands of the job, and avoid pitfalls obvious in hindsight that can sap time and motivation. But I also hope that it has a broader impact. Permanent teaching staff can make sure that their incoming colleagues have access to this information in an easy-to-reference format. Those with institutional clout can make sure that their institutions address some of the shortcomings that underlie these questions. And those who have never held a similar position can reflect on how what seem like trivial (though doubtless annoying) administrative details from one perspective become more onerous when undertaken without institutional support and on a very limited budget of time, salary, and energy to tackle them.

Without further ado, a list of questions for aspiring temporary lecturers

Are there materials for this module available from past years? Can I be added as an instructor to the VLE(/Blackboard/etc.) for past versions of this module, so that I can access these materials all at once rather than having them trickle in? Is there a policy or guideline about reusing exam questions from year to year (my own questions or those used by other instructors)?
What are studentsʼ expectations about the availability of materials and feedback on the VLE? Will the lectures be recorded? Are students expected to take notes during the lectures? Is there smart board technology that I can use during the lectures?

This might seem like a bizarre and specific question, but one of the biggest difficulties in a recent teaching assignment was retooling my materials and teaching style to accommodate the fact that students expect lecture recordings to be available, and were not able to learn by taking notes during lectures (which is a very complex issue and not one that a temporary lecturer can address). The recording devices in the room where I taught captured only audio and the computer screen – broken smart board and no classroom video, so I had to think hard before each lecture about all the syntax trees I might want to present and make sure to include them on the slides, as the option to draw a tree on the board spontaneously was not available. If I had known this information beforehand, a lot of work revamping lectures – not to mention anxious and critical student feedback – could have been avoided.
What criteria is my teaching being evaluated against? Are there student assessments in the mid-term and/or at the end of the module? Are there other assessment methods (e.g. observed teaching – this might sound intrusive but can actually be an excellent opportunity to get feedback about your teaching from an experienced senior colleague)? Is there a departmental teaching strategy that I should be aware of (especially in light of Tef)?
What are the Universityʼs training requirements for fire safety / information security training / occupational health assessment (the “display screen equipment” assessment; these are three legally required bits of training for new employees)? Have the hours to complete this training been included in my contract?
What is the marking scale for this module? What is the expected distribution of marks? (Marking scales vary drastically in different countries, and less drastically but just as unsystematically across institutions in the UK.)
Are there any deadlines or dates I should be aware of for submission of exams/assessments at the beginning of the module (for the external examiner), the return of marks for assessed work during the term and the final exam, Board of Examiners/Studies meetings, etc.?
What expectations are there of me once term time is over? Attendance at next termʼs BoS meeting(s)? Setting/marking resit exams? Will I be separately paid for these if (as is often the case) these hours are not included in the initial teaching contract?
In light of the rise in mental health morbidity and mortality, what resources does the University provide to staff for supporting students who present to me with signs of such difficulties? How can I access the Universityʼs employee counseling services (for this or any other reason)?
Other than teaching and dealing with students, are there any activities that the department expects me to undertake as part of “good citizenship”? For example, attendance at open days, student presentation days, talks/receptions, etc.
Conversely, what steps will the department take to make me feel included in its intellectual and social life? Have I been added to the email lists or other venues via which announcements are distributed? Will you assign me a “buddy” or mentor on the full-time staff who will make sure to include me in formal and informal department activities (and answer any other procedural/administrative questions I have)?
How do I access the copy machine? Kitchen/break room? Office supplies? Mail facilities? Will I have after-hours access to the building that my office is in?
Will I have the opportunity to access training that the institution provides? For example, teaching techniques, research/grant writing workshops, and other information sessions (recently, legal clinics for EU staff post-Brexit).

And finally, one set of questions that youʼre unlikely to be able to get an official answer to, and even asking will make HR types nervous (at least the clever ones). Nonetheless, itʼs important and worth finding out as soon as you can from unofficial sources:

What do you see the future of this position being? Do you anticipate advertising for another similar temporary position next semester/year? Do you anticipate making any permanent hires in the next year or two?

Disentangling style and priming using Generalized Additive Models

2015-04-14T00:00:00+01:00

This poster, co-written with Christopher Ahern and Meredith Tamminga, was presented at the Workshop on causality in the language sciences in Leipzig in April 2015.

Investigating the history of English do-support using automatically annotated corpora

2015-03-08T00:00:00+00:00

This poster was presented at the British Council’s Researcher Links workshop at the University of Campinas. It touches on work I did with parsed corpora for my dissertation.

Two invited talks on language change at the University of Manchester

2014-10-14T00:00:00+01:00

On October 14 and 15, I gave two invited talks at the University of Manchester. The first was on the topic of my dissertation research into do-support, and the opportunities and challenges that a giga-word corpus brings to historical syntactic investigations. Here’s a link to the slides from that talk.

The second talk was a seminar on the application of the Constant Rate Hypothesis to historical language research. The slides from that talk are also available.

Examining stylistic influences on the evolution of do-support

2014-07-02T00:00:00+01:00

This paper was presented at Diachronic Generative Syntax 16 in Budapest in July 2014. It presents a replication of Warner (2005)’s study of the influence of style, as measured by lexical statistics, on the conditioning of do-support. It also argues that the stylistic makeup of corpora can affect results, and thus should be carefully examined by linguists.

A link to the slides

Shame, 538

2014-06-21T00:00:00+01:00

The point that using spatial data visualizations can easily yield bias is well-rehearsed. (The first time it was made to me was in a middle school algebra textbook.) The basic problem is that, if you are representing the difference between two quantities, only one dimension of your data visualization should covary with the data. So you can have a vertical bar graph, where the width of rectangles is fixed but the height covaries. Here’s an example where our data are the numbers 1 and 2:

Sorry, your browser does not support SVG.

Alternatively, you can have a horizontal bar graph where the width covaries but the height stays the same:

Sorry, your browser does not support SVG.

In both these cases, the area of the “b” bar is twice that of the “a” bar.

On the other hand, if you allow both dimensions of your graph to covary with the data, you are lying:

Sorry, your browser does not support SVG.

Here the underlying data is still 1 and 2, but the “b” bar is four times as big as the “a”. This gives a false impression of the relationship between a and b – that is, it’s a visual lie.

Five Thirty-Eight is a statistics blog, formerly focusing on election prediction but now a sort of general statistics blog. They just published an article by Eno Sarris trying to answer the question of whether Sam Adams is “really” a craft beer company.

The article contains the following image comparing the revenue of various craft beer companies (as of 7:15pm EDT on June 21):

Check out the difference in sizes between the MillerCoors and Sam Adams bottles. Ginormous, right? Now look at the numbers. That visual difference is supposed to represent a 12-fold increase in revenue (9.1B → 9,100M / 759M = 12.0). Does that seem right to you?

It didn’t seem right to me, so I got out my ruler. The shape of the bottles is irregular, but I’ll focus on the region from the bottom up to where the sides begin to taper. This is a rectangle, so it’s easy to measure. Beer bottles are also a standard size (for most American beer, including Sam Adams and Miller), so we can assume that the rest of the bottle is proportional to this region. On my screen the Miller bottle is 5/8 (0.625) inches wide and 5/4 (1.25) inches tall, whereas the Sam Adams bottle is 1/8 (0.125) inches wide and 1/4 (0.25) inches tall. This is a ratio of 5:1 in both directions (Why 5 and not 12, the underlying number? I have no idea.). How does that translate into area? Well, the Miller bottle’s bottom portion has an area of 0.78 in² (= 5/8in × 5/4in), whereas the Sam Adams bottle is 0.031 in² (= 1/8in × 1/4in). That’s a ratio of 25:1. So this visualization overstates the difference in revenue by a factor of roughly 2.

Bonus question: beer bottles are containers, which have volume. What happens if viewers interpret the image in terms of volume? Well, going from one dimension to two squared the ratio, from 5:1 to 25:1. Going to a third dimension causes the base ratio to cube, to 125:1. (A little more uncertainty is also introduced by the irregular shape of the neck of the bottle, and how to take account of the small amount of air at the top.) Nonetheless, there’s the potential for viewers to be misled by a factor of up to 10 (= 125 / 12) by this interpretation.

It’s awful to see one of the most respected mainstream (self-designated) authorities on “data” to be using such crappy visualizations. I said above that this kind of mistake is a lie. In this case, it’s an inept lie since the article’s conclusion (insofar as it has one) is that Sam Adams is not a craft brewery, whereas the effect of this mistake is to overemphasize the difference between the non-craft breweries (Miller and AB InBev) and Sam Adams.

In a way we’re lucky that this isn’t an article about electoral politics, where misleading visualizations are stock in trade for sleazy politicians. Maybe Eno Sarris was out sick on the day that they taught this in middle school. But he – and the editorial staff at 538 – ought to do better.

DOIs come to Github

2014-05-16T00:00:00+01:00

In January, I read an article entitled 3 simple things GitHub can do for science. The title is pretty self-explanatory; it proposes three concrete steps Github can take in order to increase its utility for scientific research. “It’s nice to dream,” I thought. On Wednesday, I read that Github has actually (mostly) implemented one of the three recommendations. On the occasion of this nice surprise, I’d like to reflect on what’s happened, why it’s nice, and what more there remains to do.

Data availability in science

As the technical sophistication of science increases, reproducing results becomes not just a matter of having proper experimental equipment, avoiding mistakes, and technical prowess; but rather a matter of having access to the specific computer files that other researchers used. At least in linguistics, the culture of the field has not entirely kept pace. The upshot is that the half-life of scientific data is 17 years.

Github can help with this problem. It provides a platform on which data can be made available. It has several other desirable features as well. All files on Github have included with them a full version history, which makes it possible to track refinements over time. And Github’s social features allow you to track files as they are edited by multiple people over time.

So it’s good news that it is possible to obtain DOIs for Github repositories, making it easier to cite them in a familiar manner. Indirectly, this will (hopefully) encourage the wider dissemination of data on Github. The system is still not without its kinks: I tried to use it on my one publicly available project (in collaboration with Meredith Tamminga), but failed because the system appears to take account only of new releases, not previously-available ones. I’ve reported this issue, and hopefully it will be resolved in short order.

It’s axiomatic on the internet that technical solutions for social problems almost never work, and ultimately Github’s move is a technical. It’s not going to solve the problem that few people want to be the first penguin off the iceberg in terms of exposing their research to critical examinataion (despite the benefits that open access provides).

And my own experience is that even with the best of intentions it’s hard work to prepare data to be shared. But Github has done an incalculable amount of good to the cause of software freedom, and I’m optimistic to see them engaging with the scientific community. I hope that we continue to see improvements on this front, technical as well as social. You can help by getting DOIs for your scientific code and data that’s already available.

Implications of a three-way analysis of the evolution of English do-support

2014-04-29T00:00:00+01:00

This talk was presented at the CUNY Syntax Supper talk series in April, 2014.

A link to the slides

Algorithmic drawing of isoglosses

2014-04-16T00:00:00+01:00

I’ve been sitting in on Bill Labov’s seminar on dialect geography this semester. One of the things Bill has emphasized to the students in the class is the importance of drawing isoglosses, lines that separate linguistic regions from each other. Traditionally, isoglosses are drawn by hand according to the investigator’s subjective preferences. In the ANAE, Bill and his co-authors introduced a decision procedure for including points in an isogloss region. But it would not be correct to call this an algorithm – important decisions about where to place the drawn lines are left up to the (human) implementor of the procedure.

By way of illustration, here’s an example from the precursor to the ANAE, showing the cot~caught merger (the tendency to pronounce both of those words identically):

As part of my own learning in the class, I decided to take up the question of the algorithmic drawing of isoglosses.

Background

I decided on the Voronoi tessellation as the basis of an isogloss algorithm. This is a fancy name for a very simple concept: each point on the map has a region of influence, defined as the region which is closer to that point than to any other. Below is the Voronoi tessellation of the points in the US sampled in the ANAE

The Delaunay triangulation is the dual (roughly, math-speak for “opposite”) of the Voronoi tessellation. The basic definition is not as simple as its Voronoi counterpart (it involves maximizing angles), but it can be described as “for any two points whose Voronoi regions (= regions of influence) share a border, connect them with a line.” Here’s the Delauany triangulation of the same set of points:

I decided that the regions of the Voronoi tessellation would form the basis of my isoglosses.

Network structure

We can’t just draw lines between every Voronoi cell whose center differs in the value of the variable of interest. We also need to discard noise, for example a divergent point in the middle of a large region (such as the few blue points in MT and CA in the map above). To do this, I took the Delaunay triangulation and treated it as a network. There’s a large body of work on how to carve networks (graphs) up in various ways: separating out unconnected sub-graphs, pruning weak links, etc. I’ve really only scratched the surface by doing clique detection (and small-clique elimination).

Implementation

I implemented this approach in R. I used the deldir package for the tessellation, and the igraph package for graph operations. I decided to package the result as a Shiny webapp. I wanted the tool to be easy to use for my classmates, who are working on papers about various dialect geography phenomena and should probably be worrying about that rather than the right arguments to pass to an R function. And I’ve been looking for a good excuse to learn about Shiny.

The code is on Github, together with instructions to get it running on your machine. If you have geospatial dialect data I encourage you to try the code out and let me know about your experiences.

Implementation tricks

In making implementing this approach, I came up with a few clever hacks I’d like to share.

clipping lines: By default, Voronoi lines at the edges of the map trail off towards infinity, which is not what you want an isogloss to do. I decided to use geographical/political boundaries to truncate the isogloss lines. The functions for this are provided by the rgeos package. I take the political boundary polygons and inflate them, allowing isoglosses to continue into a small buffer zone beyond the political boundaries. Then I trim the lines that extend beyond this buffer region.
input to Shiny: Sadly, Shiny doesn’t seem to offer a good way of providing arguments to apps. I’ve created a Github issue to inquire about the possibility of implementing this. In the meantime, there’s a bit of grim hackery with .GlobalEnv to find the desired dataset.

Breaking: while I was writing this post, Joe Cheng, one of the Shiny developers, got back to me. This reinforces the great impression I’ve gotten from developing with Shiny, and gives some ideas for ways forward.
dynamic behavior in Shiny: I was able to implement a pretty nifty heuristic for automatically finding the latitude and longitude columns in the webapp. For most datasets (hopefully) once you’ve selected the name for the data frame, the latitude and longitude column names should fill in immediately. The trick to doing this is not prominently advertised in the Shiny docs, so if you’re interested check out the call to observe around line 20 of server.R.

Results

What do the results look like? Well, here’s a map of the same isogloss as above:

The match is pretty close, I’d say. We can also apply some smoothing to the lines to reduce the number of zig-zags:

Conclusions and future work

A question that I think has been swirling around this work has been, “can automatically drawn isoglosses replace those drawn by humans?” I’m not sure this is a good question to ask. Isogloss maps are not deductive evidence for a particular conclusion.

Rather, they serve as illustrations of a point, or ways of telling a particular story about a situation. As in history and the social sciences, deductive evidence and storytelling operate together to advance conclusions.

I think the uneasiness about hand-drawn maps comes from a worry that humans will somehow (perhaps without even meaning to) “cook the books” in order to produce evidence that concords with their own hunches rather than the truth. I don’t think automatically drawn isoglosses can fix this problem, just as the use of statistical p-values doesn’t magically convert all the science it touches into objective truth.

What automatically drawn isoglosses can do, however, is to provide another kind of evidence about the spatial distribution of linguistic variation. Investigators can use the output of this algorithm along with traditional (hand-drawn isogloss) and new (spatial autocorrelation) techniques to advance their understanding of phenomena.

Future work

I’d like to end this post with some ideas about future work on improvements to the algorithm.

This paper by Duckham et al. provides a Delauany-based algorithm for drawing shapes which could be a useful refinement
One of the references of that paper (#7, by Alani et al.) suggests adding a ring of points around the area of interest to avoid the problem of Voronoi lines wandering off towards the edge of the map. This would replace the current clipping algorithm. It would also help prevent some odd problems with the current approach, such as the isogloss which wanders over western TX, NM, and AZ in the map above (which is preserved by the small-clique elimination pass because it is continuous, outside the viewable region, with the isogloss separating the western US from the rest of the map).
It would be good to allow isoglosses for multiple features to be drawn on the same map in the interactive interface, and to allow the piecewise deletion of undesired isogloss segments (like the western TX one above)
It would also be nice to package this code as a ggplot primitive, so that an isogloss could be added to a ggplot map simply by appending + geom_isogloss(...) to a plot creation command. There are some not-quite-working experiments with this in the ggplot.R file of the Github repo.