Visualizing vowel spaces in R: from points to contour maps

Typically when linguists wish to examine the vowels of a language, they plot the vowels in an F1xF2 space, which approximates a relative articulatory position of the vowels. Now, there are certainly problems with this approach (lack of F3, possible nasal formants, dynamic movement). Yet, despite these drawbacks, visualizing vowels this way is relatively standard and has the advantage of being understood by a wide audience. In R, there are several methods one might use to plot vowels in a space like this. I will discuss three here, two of which are clearly less than ideal and another which I am in the process of learning still. I will be relying on a data set of Arapaho vowels from elicitation sessions from three speakers. Given the nature of the data, I had to analyze a different number of vowels per speaker, so that one speaker is over-represented (1290 vowels) and two others are underrepresented (600-700 vowels each).

I am interested in visualizing the quality differences between long and short vowels in the language. Arapaho has short, long, and extra long vowels, though I only really have enough data to analyze long and short vowels, so I am sticking to that. I am looking at the monophthongs /i, ɛ, ɔ, u/, which are realized as more centralized variants when they are short. Here is a sample of my data:

Vquality Label Length Vaccent seg_Start  seg_End Duration Time       F1       F2       F3
8         o    v2  short       H  19985.83 20088.10  102.277    2 673.0124 1116.783 2887.543
9         o    v2  short       H  18887.63 18990.38  102.753    2 682.5495 1204.757 2636.614
10        o    v2  short       H  10048.30 10152.09  103.789    2 679.4601 1077.850 2910.295

I am leaving out several columns here (including speaker, word, etc.), but all data is coded for vowel, written i, e, o, u, and for length (short, long).

1. The first way to visualize a vowel space given data like this is to use R's plot function. The default here is to built up a plot by adding individual elements. With the data formatted the way it is, it would be necessary to create several subsets and plot each separately. For instance, if we restrict ourselves to just the long vowels, one could do the following:

> long <- subset(formant_data, Length=="long")
> u.long <- subset(long, Label=="u")
> i.long <- subset(long, Label=="i")
...

The result of this will be 4 different data frames, each of which could be plotted separately as points, as follows:

> par(mar=c(5, 4, 2, 2))

> plot(F1~F2, data=i.long, ylim=c(1000, 200), xlim=c(3000, 600), pch=1, col="red")

> points(F1~F2, data=e.long, pch=2, col="blue", add=T)

> points(F1~F2, data=o.long, pch=3, col="green", add=T)

> points(F1~F2, data=u.long, pch=4, col="black", add=T)

> legend("top", horiz=TRUE, c("i", "\u025B", "\u0254", "u"), pch=c(1, 2, 3, 4), col=c("red", "blue", "green", "black"), x.intersp=0.8)

This produces the following:

Fig 1

This looks good enough to plot observations, but what if one wants to get an idea about where the averages lie? It might be easy to imagine an average "i" here, but the other vowels seem somewhat dispersed throughout the vowel space (the short vowels are even more so). So, this is harder. 

2. One solution is to plot the vowel data using the vowelplot() function in the vowels package. This package can both draw circles around the vowel area and compute mean values for each of the vowels. However, it requires the user to reformat his/her data to fit a template used by the package. Depending on one's data organization, this can be cumbersome. The format of their template is a data frame of 9 columns, which includes: speaker_id, vowel_id, context, F1, F2, F3, F1_glide, F2_glide, and F3_glide. Plotting requires fewer commands, but the options within each command are limited. If we try the following:

> vowelplot(long, color="vowels", ylim=c(1000, 200), xlim=c(2600, 600)); it produces:

Fig 2

This figure is interesting insofar as it color codes the vowels and divides up the space by speaker. It does all this with a single command too. However, there is no way within the package to avoid dividing the data into speakers (as you might notice, I did not specify this in the plot command) when showing individual data points. 

Another advantage of this package is the ability to display vowel ellipses around the plotted data. We can do this with the following command added on:

> vowelplot(long, color="vowels", ylim=c(1000, 200), xlim=c(3000, 500))

> add.spread.vowelplot(long, ellipsis=TRUE, labels="vowels")

Fig 3

Ack! Yes, this is indeed very ugly. Unfortunately, the vowelplot() package always assumes that you want individual speakers. One could simply eliminate speaker differences in the data frame and replot the data with single ellipses. Alternately, one could plot the ellipses without the observations. I won't go into how to do this here. Instead, I will show one of the interesting advantages of using the vowelplot package: the ability to extract and plot mean formant values. So far, I have not plotted the short vowels because the degree of overlap would have been particularly large. I will do so here though:

These commands calculate the mean values for the first two formants among the short and long vowels.

> vlong <- compute.means(mono.long)

> vshort <- compute.means(mono.short)

These commands plot the data:

> vowelplot(vlong, label="vowels", ylim=c(800, 300), xlim=c(2400, 800), title="Vowel space in Arapaho")

> add.spread.vowelplot(vshort, labels="vowels")

This produces the following:

Fig 4

This looks much cleaner, but averages always look cleaner. The vowels with the dots represent the long vowels, while the vowels without the dots represent the short vowels. You'll notice that the short vowels are more centralized than the long vowels, though the back low vowel doesn't really change in quality. So far, so good.

Yet, as phoneticians (or as linguists/speech scientists), we are often more interested in the distribution of the data than the average values. Yet, if we plot ellipses here, it looks just as chaotic as Figure 3. This is because an ellipse contains two mutually perpendicular axes about which the ellipse is symmetric. These axes are the two dimensions (F1 and F2) which position the vowel in the vowel space. However, actual observations are not elliptically symmetrical around the center. Thus, ellipses might tend to overestimate the actual degree of overlap in a vowel space.

3. One solution to using the vowelplot package is to use ggplot2(). This very modern plotting software allows us a larger set of data visualization techniques. One way that I might think about plotting the distribution of my vowel data is with a two-dimensional contour map. Contour maps include three dimensions, with density as a "higher" point. They rely on kernel density estimation (KDE), which is a non-parametric way to estimate the probability density function of a random variable. Given that a prior distribution is not assumed, they also have the advantage of non-symmetry. As far as I know, I have not seen these applied to vowel spaces before. We can plot our data as follows:

> f.plot <- ggplot(formant_data, aes(x = F2, y = F1, color=factor(Vquality))) + geom_density2d(aes(label= factor(Vquality))) + scale_y_reverse() +  scale_x_reverse() + ylim(900, 200) + xlim(2800, 500)+ theme_bw() + scale_color_hue(name="Vowel quality", breaks=c("i", "e", "o", "u"), labels=c("i", "\u025B", "\u0254", "u"))

> f.plot 

This produces the following figure:

What this figure reveals that is so often left out of vowel plots is a clearer sense of the concentration of observations. One can observe a somewhat bimodal distribution for /u/, one concentrated with an F2 around 1000 Hz and another with an F2 around 1500 Hz. These probably reflect differences among speakers, but they may also reflect a difference of context (there is substantial alveolar fronting). If we wish to plot both short and long vowels, we can do so by using a facet_wrap() function. 

> f.plot <- ggplot(form.mono2, aes(x = F2, y = F1, color=factor(Vquality))) + geom_density2d(aes(label= factor(Vquality))) + scale_y_reverse() +  scale_x_reverse() + ylim(900, 200) + xlim(2800, 500)+ theme_bw() + scale_color_hue(name="Vowel quality", breaks=c("i", "e", "o", "u"), labels=c("i", "\u025B", "\u0254", "u")) + facet_wrap(~Length)

> f.plot 

This produces the following:

Now, we observe not only the tightness of observations around the median for the long vowels, but the asymmetrical ways in which the vowel space changes as a function of length. There are clear realizations of short /i/ which match those of long /i/ in quality. However, there are also a larger number which encroach into the center of the vowel space (though significantly more along the F1 dimension).

The advantage of using ggplot2() to show this data is that one can represent most of the observations and simultaneously observe non-linearities in the shape of the distribution. Outliers are more naturally excluded since they do not contribute to the estimated density function. By contrast, ellipses simply expand to symmetrically cover the entire space of the observations (or at least a space determined by symmetries inherent to normal distributions).

I think I am a fan of this method for vowel visualization, but I am unsure if this is the right way to go about things. Thus, any commentary is welcome.

It was never just about marriage.

Within the debate on same-sex marriage in the United States we have countless times heard the refrain that traditionalists are not homophobic, but rather wish the preserve the sanctity of the religious institution of marriage. This argument, in fact, was the main crux of the defense of Proposition 8 when it was debated in California in 2010. Now, as state after state considers its laws in light of the recent rulings that have severely weakened DOMA, one anticipates this argument to continue to come up on the state level. This argument is legally important, as it allows one to separate the preservation of the sanctity of marriage from the animus associated with homophobia. As the Proposition 8 and Supreme Court trials demonstrated, animus can not be used as a justification for the support of a law.

There is a dirty conceit that conservatives would do well to admit though: such a separation is simply an argument of convenience. To show this, let's begin by assuming that one can separate homophobia from a religious justification for traditional marriage. If this were true, one might expect that in states where traditional marriage was outlawed, other laws providing rights to gay and lesbian citizens would find more favor among local governments and voters. Yet, what we observe is the opposite: in states where gay marriage is forbidden, there has been substantial opposition to many other rights as well.

Simplifying a bit, the types of rights that gay and lesbian people have sought usually keep to the following trajectory. First, the right to consensual relations was sought. Second, non-discrimination in education, housing, or the workplace was sought. Third, some type of legal recognition (domestic partnership or civil union) for one's relationship was sought. Finally, the right to marry was sought. In those same states where same-sex marriage is outlawed and where its opponents argue that their position isn't driven by animus, one finds resistance to non-discrimination ordinances or even to consensual relations among same-sex partners (as in much of the southern US).

If it were just about the institution of marriage, and nothing else, then what on earth is the motivation for resisting the passage of all the other possible rights for gay and lesbian people? I don't know the conservative response to this question, but I would like to hear some rational argument if it exists. The notion that one can separate upholding the religious institution of marriage from a motivation from animus may sound good in theory, but actual practice shows us otherwise. And as a true empiricist (and positivist), I trust my observations.

If I were to call a spade a spade, I would say the separation argument is simply an attempt to save face in social circles in light of broader social acceptance of gay people and their relationships. After all, one can "hate the wedding but not the wedd-er" and still be in the in-crowd. Taken as such though, this sounds more like a social coping mechanism for someone uncomfortable with gay people than a legal argument used to oppose same-sex marriage.

R scripting problem

Maybe it's just something silly I can't figure out, but I've been banging my head at my computer for the past couple hours. So, I thought I would put this up on the web to elicit help. And yes, I have looked at stackoverflow and other sites for answers, but I've come up short so far.

Here's the issue:

Assume you have a data frame where "Time" values range from 1:10 and you have 3 measures (F1, F2, F3) for each 10 time point, e.g. F1 at time 1, F1 at time 2, etc. The goal is to take this data frame and simply print the mean value for each of the measures at each time point. If so, it should be possible to simply create subsets at each time point and then extract mean values for each measure. So, I wrote a script that does just this. It doesn't work though:

ts.obj <- ts(array(data=NA, dim=c(10, 3)))     #Create a time series for the output.
{for (i in 1:10)                            #Run a loop through each of the time points.
obj.i <- subset(object, Time==i)
mnF1 <- mean(obj.i$F1)           #Get mean values for each measure at time point "i."
mnF2 <- mean(obj.i$F2)
mnF3 <- mean(obj.i$F3)
ts.obj[i,1] <- mnF1                    #Place these mean values into the time series.
ts.obj[i,2] <- mnF2
ts.obj[i,3] <- mnF3 }

Any suggestions?

Interacting with a big academic publisher

Publishing in academia is an odd activity. Unless you are writing a book, you receive no royalties. The only "royalty" is the joy you experience having produced (hopefully) original research that will interest those who work in your academic discipline. The fact that many academic journal publishers persist in asking large fees for work that is essentially done for free is an ongoing point of contention in many circles. Nevertheless, many of us have accepted this problem in exchange for getting one's work published in a prestigious journal. And face it, most of the most prestigious peer-reviewed journals are owned by big academic publishers.

Since so much effort and personal branding is placed in one's work in academia, it stands to reason that you might be interested in how popular your work is. So, several weeks ago, I decided to find out  how much some of my recent work had been downloaded. I figured that Elsevier, the publishing company who owns Journal of Phonetics, might have an answer for me. I contacted them through the contact information on the Elsevier website. Here is my first email (on 8/16/12):

To whom it may concern,  

I have a couple publications now with Elsevier. On my personal website, it is possible for me to tally the number of people that download particular articles of mine (as well as the location of their ISP). I imagine that Elsevier keeps a similar background service running on their journal websites to determine popularity of different articles. Would it be possible for Elsevier to share any of this information with the authors? I understand if the ISP location is considered private. I am specifically interested in the number of downloads.  

Cordially, Christian DiCanio

I thought this question simple enough. A day later, I received a reply.

Dear Dr DiCanio, 

Thank you for your e-mail. We wish to advise that the statistics on ScienceDirect are recorded on the account level for the institutions to see their usage of the material on our platform, but, unfortunately, they are not recorded for specific articles, and it is not possible for us to provide authors with the download statistics of their articles. I hope this is of assistance to you. 

Meanwhile, the following Elsevier Customer Support solution may be of interest: http://support.elsevier.com/app/answers/detail/a_id/263. Please ensure that the reference number remains in the subject line when responding to this email. 

Why not also visit our self-help site at http://support.elsevier.com? Here you can find FAQs, online tutorials and instructions relating to manuscript submissions and articles in production. You will also find 24/7 support contact details, including live chat, should you require further assistance. 

Kind regards, Miss Jane Doe, EP Customer Support 

This had me a bit confused. On most journal websites, there is a box called "Most Downloaded Articles". Unless this information here is a lie, the journal must keep a record of how many people download certain articles. In fact, any good business would want to know which of their products are selling well. As for the "customer support solution", it simply brought me to a general help page. So, I replied to the customer support specialist.

Dear Miss Jane Doe, 

I apologize, but I do not believe you that Elsevier does not record the usage of specific articles. For most journals, there is a section called "Most downloaded articles." If you do not believe me, click here: http://www.journals.elsevier.com/journal-of-phonetics/. Unless each journal is specifically inventing these values, the data must be recorded. It would be bad business of you not to record which of your products gets the most hits or sales. I can not imagine that it is asking too much of you to offer one of your authors access to a service that you already possess. 

Cordially, Christian DiCanio

They didn't shut me up with their first reply, so I got another response a day later:

Dear Dr DiCanio, 

Thank you for your e-mail. I have sent this to the ScienceDirect team for further advise. I will contact you as soon as I have received a response from them. Your patience in this matter is highly appreciated. (etc.) 

Kind regards, Miss Jane Doe

A few days later, I received another email telling me "With regard to your below query, I am still waiting for a response. I have now made a follow up. Your patience in this matter is highly appreciated." I felt  a bit elated. Could it be that I was causing a stir at Elsevier and that they might change their policy to actually provide a service to the authors? I looked every day at the Elsevier website, curious as to whether "New Products" might include this type of service. After several days, I received the following email:

Dear Dr DiCanio, 

Further to my below e-mail, I have received a response from ScienceDirect team and they ask you to contact them via the "Contact Us button" found via the following link: http://www.sciencedirect.com/science/contactus. I hope this helps.

Kind regards, Miss Jane Doe

Ack! They essentially just want me to resend my original message with hopes that I might get tired of this issue and stop sending them emails asking them to do something about it. Well, I did resend my message, but stated more succinctly. I received the following reply a few days later:

Dear Christian, 

Thank you for contacting Elsevier's e-Helpdesk. I apologize for the inconvenience this may cause you, but it is not ScienceDirect policy to provide that data. 

Please visit the following site to read more: http://www.info.sciverse.com/sciencedirect/using/Make-ScienceDirect-yours/top25. Please feel free to contact us with any further issues. 

Sincerely, Joe Schmo

So, apparently there is some policy about not providing this information that is used as a justification (er, excuse) to avoid doing anything about it. I was rather more brusque in my reply to "Joe Schmo".

Dear Joe Schmo, 

I understand that it is not ScienceDirect's policy to provide this data. I am actually asking why this is the policy and why Elsevier could not just decide to offer such a service to authors. It is rather absurd to say that this simply is not part of your policies. If I were a novelist and I wrote a prolific novel which sold millions of copies, I could certainly find out how many copies were sold. Indeed, I would be able to find out this information while also earning a royalty for my work. When I publish with Elsevier, I earn no such royalties. Yet, apparently it is problematic to ask for information that you already collect. 

There is a deeper, more important goal here. Sometimes academics want to get a sense of whether their work is particularly popular or relevant within a particular journal. If it is not, they may choose to change where they submit their work. Such information is useful. 

Cordially, Christian DiCanio 

Well, perhaps my tone was less than cordial. I had received rather lame responses though. I think I had all but given up on Elsevier to actually address my question. Yet, out of the blue tonight, I received the following short email:

Hi Mr. DiCanio, 

Please find the attached full text download of your articles published in Journal of Phonetics. And if you have any more concerns, please let us know, thanks! 

Best regards, Jim Schmo

An attached spreadsheet contained precisely the information I had been requesting, with the number of downloads listed for every month the articles had been in press. Of course Elsevier had this information. It just took some persistence to actually get it.

While I was happy to find this out, it actually appears to be just a quick band-aid. I don't imagine that Elsevier will change their policy and offer such a service to their authors. Though, perhaps they will think about doing this if enough people request it. I would be persistent if I were you.

The value of being lazy about your education

About a week ago, the NY Times published an opinionator piece with David Brooks and Gail Collins on the value of higher education. It took me a week to see it, but I finally got around to reading the debate and plenty of the commentary. It's essentially the same old argument, rehashed. It goes like this: "All these online courses will cause a major paradigm shift in higher education. As more students just take online courses, the need for so many major, expensive universities will decline. College will become more affordable. After all, college is expensive because of all the money those greedy professors make."

Now, I'm a postdoc, not a professor. Though, I hope to have an academic position as a professor some day in the near future. However, these debates consistently strike me as extremely short-sighted. The source for this myopia is a very odd view of what role a professor serves at a university. The debaters seem to assume that the only role of the professor is to stand in front of a classroom and lecture with no interruptions. Furthermore, there is a very odd public idea about how information transfer, i.e. "learning" takes place.

To counter the first assumption, consider for a minute what is involved in teaching. Certainly syllabi are created and lesson plans are devised. That is a given. Yet, in most courses, a substantial number of assignments and tests are also given. Unless the professor's course is sufficiently large so as to require graders, the professor will do all the grading. In the entire debate in the NY Times, no one ever asked how grading gets done in online courses. I suppose that if every course were a literature lecture on how wonderful Jane Eyre is, then perhaps very little homework could be required. Yet, in every linguistics course, in every math course, in every biology, chemistry, and physics course, there are homework assignments. In a world where 10,000 people (instead of 100) sit in on a physics class via the web, who grades the homework assignments? It seems as if Brooks and Collins assume that one can get a college degree in a technical field just by listening and not through any sort of practice.

Moreover, learning a technical discipline frequently involves using symbols and formulas that are not easily typable. This makes doing any sort of automated grading (or even online homework submission) near impossible, lest you think of some way to automate it. In fact, such automatic grading methods are quite problematic in mathematics, where wrong answers due to small arithmetic errors are evaluated identically as wrong answers due to not understanding a theorem.

I'd also point out that universities benefit quite handsomely from grants which accomplished faculty bring in. These grants often pay professors salaries, in part or in total. In a university system where there are simply fewer professors, the university makes less money. This is a major source of income for major research universities.

The second issue which I'd like to comment on is the idea about how information transfer takes place. In typical courses, professors take time to answer student questions and to evaluate counterexamples to certain claims. There is no perfect course which is entirely clear to everyone. But, how do questions get answered in an online course? If they are not answered in real-time, understanding can be dramatically stalled. There is a false assumption that many people make about learning. It goes something like this "If you see someone do X, you learn to do X." For certain types of simple tasks which involve simple repetition, this might be true. Yet, for methods applied to novel problems, you often only learn if you practice the skill and are evaluated for it (either by yourself or others). Calculus is an ideal example of this. You only really understand integration after having done lots of it.

I will end with an anecdote. Many non-linguists assume that children are aided learning language by watching television. Yet, 40 years of research in language acquisition (and dynamics) has shown that humans require actual interaction (with others or with the world) to learn linguistic skills. Why do those who wish to revolutionize higher education seem so ignorant of the fact that learning requires doing? Listening about how something is done via your computer screen is often insufficient. The idea that one could simply learn a college course well with cheap online learning is certainly attractive. Yet, there is a pretense that the authors would do well to admit. It's just attractive because you get to be passive.

What constitutes sufficient knowledge?

I was inspired by a comment that my colleague Bonnie left on her facebook today. It read (if I may, Bonnie, even though you are not here) "Research = good. Researching the heck out of a topic you feel less competent in however leads to a rather staggering amount of info/papers to winnow down into 2 50 min lesson plans. Sigh...."

There is a general sentiment that she captured here that I've often encountered in doing a google search for research related to my own work. If I find a couple of recent articles that are strongly related to my topic, I can usually navigate to other research by searching through their references. However, when exactly does one know when to stop? If there is an active theoretical debate related to the topic, then you can easily find articles in each camp and talk about the relative merits of each in the introduction to an article that you're writing or for a class you're teaching. If there is no big debate, or if your work does not directly test someone's hypothesis, then it isn't so clear how much background information you need to research.

This is one of the reasons why google searches for articles related to my research always cause me to feel depressed. What constitutes "knowing something" enough to discuss it? For the more esoteric topics that I work on (Oto-Manguean languages, phonation type), the literature is relatively sparse and controlled. I feel like I have a command of it because I've read the major findings. But for a topic that is new, it is difficult to get a sense that you have the right perspective. Having read the 5 out of 7 articles for one topic makes you an informed academic. Having read only 5 out of 100 articles on a different topic makes you seem like an amateur. It's simply more likely that you've missed something important. How do you know when you have enough mojo to actually talk about something authoritatively?

This ends up being only part of the problem that I have when I write about a new topic. The other, perhaps more major, issue is language. If you are deeply embedded in a certain realm of research, it becomes easier to use its associated technical jargon comfortably. When you are just trying to publish in a new area though, the language barrier seems insurmountable. It's not that I don't understand the language that is being used, it's just that my memory doesn't think about things in the argot that is used to describe them. When it comes to a new area of research, I often feel like I am describing it with a 10 word lexicon. "Ba ba ba, um, perceptual integration, der, hum, hum, blech."

About a year ago, I just started, out of the blue, to write up all my notes on Trique morphophonology. The pages just kept flowing and flowing. When I finished, I had a 40 page manuscript which really summed up lots of what I knew about the language. I was proud of myself. It felt easy to write. I probably felt this way because I was academically-raised speaking "linguistics" but my work over the past couple years requires me to speak "psychologist." This still feels like an L2 (second language), like I am an imposter striving to put simple phrases together. This is why writing becomes so difficult for me. I can write a nice intro, but then when I have to describe the literature on a topic, I feel dumb.

If there were a book "How to talk like a cognitive psychologist", I would certainly buy it. Yet, there is just a screen with a half-written article and groups of random pdfs in front of me, laying like a few pieces to a 500 piece jigsaw puzzle whose main image I may never glimpse at. I know that all academic work requires that you have some background. I just wish that there was a better way to translate comprehension into intelligent prose. And with that, a way to know that you know enough to translate in the first place.

Clitics, suffixes, and incorporateability

In much of the literature on Oto-Manguean languages, linguists discuss what are known, across different families, as the "personal enclitics." Yet, the evidence in favor of considering person marking to be clitic-like rather than affixal is always a bit spotty. I've been debating the status of person marking in Itunyoso Trique ever since I started studying it almost 8 years ago. This is the first time I'm making this debate manifest though, at least in writing. (Caveat: Please don't read this like you are peer-reviewing it. I am just putting some ideas down in my blog.)

A typical verb structure found in Oto-Manguean languages is shown in (1) and structure typical of nouns is shown in (2).

(1) CAUS-ASP1-ASP2-VERB-PERSON
(2) NOUN-PERSON

In Trique, the structure is a little different. It is shown in (3) and (4)

(3) CAUS/DETRANS-ASP-VERB-PERSON.SUBJ-PERSON.OBJ
(4) GEN-NOUN-PERSON

Itunyoso Trique Subject Enclitics

To give you a bit of a background, Trique has 9 tonal contrasts. There are four level tones (/1/ is low): /1/, /2/, /3/, /4/; two rising tones: /45/, /13/; and three falling tones: /43/, /32/, /31/. Syllable structure is simple, with no codas but two laryngeal codas, /h/ and /ʔ/. The two rising tones only appear in syllables with a coda /h/. No contour tones occur before a coda /ʔ/.

Regular alienably-possessed nouns are marked with either a genitive prefix /si³-/. Inalienably-possessed nouns never take a genitive prefix. Consider the paradigm shown below in (5).

(5) ka³siʔ³ 'honey' | si³-ka³sih⁵ 'my honey' | si³-ka³siʔ³-reʔ¹ 'your honey' | si³-ka³siʔ³-sih³ 'his honey'

Note that the /h/ of the 1.SG replaces the glottal stop on the stem and causes tone raising. The morph of the first person singular is not /h/ though. Rather, if a stem does not contain a coda /h/, one is added. If a stem does  contain a coda /h/, it is deleted. Consider the paradigm in (6). The 2.SG and 3.SG person markers are less remarkable for our purposes here.

(6) tʃa³kih³ 'ear' | tʃa³ki⁴³ 'my ear' | tʃa³kih⁴-reʔ¹ 'your ear' | tʃa³kih³-sih³ 'his ear'

This toggling process is completely regular on all Itunyoso Trique words. I am not putting more examples here, but you can read my morphophonology squib for plenty of cases. There are a number of tonal alternations involved with the 1.SG toggle, far far too many to discuss here, but again, you are welcome to read about them in my morphophonology squib. I am just providing this information as background.

Generally speaking, linguists have said that person markers like these are enclitics. The argument is, as per Zwicky & Pullum (1983), that these enclitics are not picky as to the part of speech of their host. Thus, observe the following words with personal enclitics in Trique.

(7) ni³ʔi³-sih³ know-3sg, 'he knows'
(8) si³-ka²-sih³ GEN-corn.tassle-3sg, 'his corn tassle'
(9) ru³ku³-sih³ behind-3sg, 'behind him' / 'his back'
(10) ŋga¹-sih³ and/with-3sg, 'with him / and him'

In (7), we observe the clitic on a verb. In (8) and (9), it attaches to a noun and a relational noun, respectively. In (10), the clitic attaches onto a conjunction, or what we might call a simple comitative particle. The stem in (10) is not a relational noun and has no other interpretation. Considering the different parts of speech in Trique, person markers are not very selective as to their host. Generally speaking, this is the main criterion for "clitic-hood" that people consider.

Hollenbach (1984) in her dissertation on Copala Trique, also notes that clitics apply freely to adverbs immediately following verbs. So, a structure like VERB-ADVERB-PERSON is possible in Trique. One might not expect such a structure to occur if person marking were suffixal. Generally speaking, additional stems do not usually intervene between a stem and an affix. Observe this pattern in (11), below (from Itunyoso Trique):

(11)
(a.) ka³-tʃi⁴nih⁴ PERF-get.drunk.1sg, 'I got drunk.'
(b.) ka³-tʃi⁴ni⁴³-yũh³⁵ PERF-get.drunk-again.1sg, 'I got drunk again.'

In (11a), the first person is marked with a coda /h/ on the verb root. Yet, once the adverb /yũ⁴/ 'again' is included, person marking applies to this word. Examples like these convince Hollenbach that person marking is a clitic in Trique.

This evidence has never struck me as very strong. First of all, several clitics condition substantial tonal allomorphy on stems. Second, one of the exponents of the 1.SG clitic is the deletion of stem material (half of the "toggle"). These strong phonological patterns are more typical of affixes than clitics. Though, as far as I know, phonological evidence doesn't play a role in decisions of "clitic-hood."

Turning to syntactic criteria, there is other evidence that person marking may be suffixal. While Hollenbach makes a blanket statement that person marking may apply to adverbs, she doesn't specify which ones. As it turns out, they are rather restricted. Observe the data in (12)-(13).

(12)
(a.) ko³ʔo³²-sih³ drink-3sg, 'He drinks'

(b.) ko³ʔo³² ni²ʔrua³²-sih³ drink much-3sg, 'He drinks a lot'


(13)
(a.) a³kĩh³⁵-sih³-yũh² call-3sg-1sg, 'He calls me'
(b.) a³kĩh³⁵ ni²na²-sih³-yũh² call always-3sg-1sg, 'He always calls me'

The words 'much' and 'always' are the only other two adverbs that may precede person marking. The word 'slow' is ungrammatical in a similar construction, shown in (14).

(14)


(a.) ka³-tʃi⁴nih⁴ na²nah² PERF-get.drunk.1sg slowly, 'I got drunk slowly.'
*(b.) ka³-tʃi⁴ni⁴³ na²na² PERF-get.drunk slowly.1sg, 'I got drunk slowly.'

It strikes me as odd that the only adverbs that are permitted to precede person marking are words for 'again', 'always', and 'a lot.' These each seem eerily similar to verbal aspect markers (frequentative, intensificational, etc.). So, perhaps the adverbial restriction that Hollenbach discussed is not a very convincing argument. Yet, if person marking is actually suffixal, then these words would be examples of incorporated adverbials. I don't know if this analysis is equally wrought with problems.

There are some other data that are relevant here too, involving verb concatenation, but that will wait until tomorrow. Any thoughts so far?