For ESL/EFL teachers, one of the more challenging points of lexical feedback concerns the use of conjunctive or connective expressions. The focus here is on misuse (as opposed to over- or under-use). Every writing instructor is familiar with sentences like (1) and (2):

(1) We had a hamburger for lunch. Moreover, we had French fries.

(2) My sisters love the beach. Similarly, I love the beach.

Teachers may respond with, “use also; use so do I”, with no clear means of articulating why the learner’s choice wasn’t quite right. I will argue that much of the problem lies in an overly-simple visualization of connectors as ‘discourse glue’ (cf. Halliday & Hasan 1976 and many composition textbooks) and ignores their cognitive dimension. This synopsis of a presentation at TESOL 2017 is part of a much larger study that includes examination of a two-million-word written corpus of non-fiction prose; I restrict the discussion here to judgment data from university-educated native (and near-native) speakers, a large percentage of them in language-related fields, in which respondents – 50 for each item reviewed – were asked whether a connector ‘fits’ into a slot between two sentences provided to them. I constructed the sentences with an eye to isolating key variables that define the parameters of successful connector use. Many sentence-pairs were very minimally different from each other, and the wealth of respondents (over 200 in all) prevented any individual participant from seeing the variations of an individual item in a single judgment-session, where each session asked for judgments on about ten sentence pairs. Participants were told that the key variables under consideration were not directly related to register (it goes without saying, for example, that expressions like conversely may be more frequent in writing than in spoken discourse). Rather, the variables concern various meaning relations that exist between segments of text and are intended to throw light on why some pairings make sense while others make less sense, or no sense. 

Most treatments of conjunctive expressions assume that a connective marks a relation between the content of two adjacent clauses. Thus, in a sequence like:

(3) This is good coffee. In fact, it’s the best coffee I’ve had all year.

the lexical semantic relation good-best is overtly instantiated across the two clauses and thus justifies the use of in fact, which conventionally signals the existence of an “X…more than X” relation. We need not go much outside the overt content of the sentences to comprehend and approve the connector’s use. 

However, not all connectors work this simply. Some, it appears, require more reader/hearer processing without which the connector’s use may fall flat. In Relevance Theory (Sperber & Wilson 1986, Blakemore 2002), the cognition-oriented pragmatic theory deriving from Grice (1975), connective expressions are not ‘glue’ that holds a text together. Instead, they signal processing instructions to interlocutors; that is, a connector fulfills an utterer’s implicit contractual obligation to an interlocutor by providing a mental shortcut to comprehending the relevance of one sentence to an adjacent one, since the relation is not always self-evident. The juxtaposed sentence-pair,

(4) Joe is on his way over here. My keyboard is malfunctioning.

can be interpreted in multiple ways. (a) Joe’s visit might be seen as a curse that causes the malfunction, or (b) Joe might be a keyboard repairman who will solve my problem; alternatively, (c) both sentences can be taken as parallel evidence that this is not a good day (since I don’t particularly like Joe). A connector helps resolve the ambiguity. If I were to insert moreover (or what’s more), I would be inviting the hearer to derive a conclusion like (c). The accessibility of this conclusion makes moreover appropriate.

To further illustrate the point, consider a few experimental examples with reader ratings taken from Williams (1996). Starting with (5),

(5) Last week, the Oregon Mountaineers Club tackled the ascent of Mt. Hood.
The climbers were not happy with the leaders of the expedition.
___, they were disappointed by the scenery.
So overall, they did not have a pleasantly memorable trip.

Moreover fits? 98%
Moreover fits well? 54%

We get a rather strong fit here. Not so in (6):

(6) Last week, the Oregon Mountaineers Club tackled the ascent of Mt. Hood.
A few people from an Idaho chapter joined the group.
___, a few Californians were there.

Moreover fits? 48%
Moreover fits well? 4%

The relative judgments differ rather strongly. Why? The difference, I argue, is that we were supplied with a third proposition in the first case. Judging from experimental data (and from corpus data which cannot be supplied here), moreover appears to require a propositional structure in which either a conclusion is drawn from the two clauses joined by the connector, or a claim is made and two pieces of evidence connected by moreover are provided. It should then be clear why the ratings in (5) and (6) don’t match. In (5), we see an explicit inference drawn from the second and third sentences. In (6), that inference is missing: the two connected clauses are not processed as support for any kind of claim or inference. True, you could draw the inference Therefore, there were people from at least three states on the hike. But the inference seems trivial, and it’s not likely to be one that yields what Relevance Theory calls contextual effects: it’s not information worth processing. There is therefore no motive to insert a connector.

However, things change if we can make that kind of information worth processing, and that’s all a matter of selecting a context. Consider the next example, (7):

(7) Last week, the Oregon Mountaineers Club tackled the ascent of Mt. Hood.
A few people from an Idaho chapter joined the group.
___, a few Californians were there.
So the Oregon group cannot be accused of excluding outsiders.

Moreover fits? 68%
Moreover fits well? 24%

Here, approval rises – not to the level of (5), but to a higher level than (6). The difference is that we perceive some rationale in juxtaposing the second and third sentences; both provide parallel support for the inference in the final sentence.

Now, briefly involve yourself in the experimental process. How would you rate the use of the connector in (8)? 

(8) Oregon hikers are often accused of insularity.
However, last week the Oregon Mountaineers Club tackled the ascent of Mt. Hood.
A few people from an Idaho chapter joined the group.
___, a few Californians were there.
So the Oregon group cannot be accused of excluding outsiders.

Moreover fits? ____
Moreover fits better than in (7)? ____

If you find that moreover fits better here than it did in (7), you can explain why: the parallelism in the two sentences joined by moreover is seen more easily, because you’ve been cognitively primed to see it by the first sentence in the sequence: a claim is made, a contrast is signaled by however, and you’re cognitively ready to process the next two sentences as clear parallel evidence. In the case of (7), you weren’t primed to expect the final sentence. When presented for judgment to the audience at the TESOL Convention, item (8) approached 100% acceptance when connected with moreover.

This leads to a kind of definition:

The propositional frame for moreover

“X, moreover, Y” is appropriate if sentence X and sentence Y stand in a parallel, non-trivial relation to some proposition Z.

This seems simple enough: we need (at least) three propositions. The kicker is that the third proposition need not be stated for moreover to work for most people. What we do need is the ready accessibility of that third proposition, which may be unstated. In (9), I’ve repeated example (5) above but subtracted the explicit inference about the trip, originally supplied:

(9) Last week, the Oregon Mountaineers Club tackled the ascent of Mt. Hood.
The climbers were not happy with the leaders of the expedition.
___, they were disappointed by the scenery.

Moreover fits? 80% (vs. 98% for (1))
Moreover fits well? 47% (vs. 54% for (1))

We can explain the difference in raters’ votes by the fact that a non-trivial inference can be drawn here – the climbers weren’t happy with their experience – but is not explicitly drawn. Many of the readers of (9) really weren’t sure what the relevance of the juxtaposition of the second and third sentences was, while in (5) it was provided explicitly.

Many additional examples support the given propositional frame, as do written corpus data. If the analysis is correct, it should hopefully be clear why it sounds odd when a student writes a sequence like We had a hamburger for lunch. Moreover, we had French fries. There is no relevance to the juxtaposition apart from the fact that it provides a list of what the writer had for lunch. No relevant third proposition is available. Finally, what holds for moreover seems also to hold for furthermore: judging from written corpus data, the two are more or less interchangeable, with the minor difference that furthermore tends to be more frequent when the list of propositions exceeds three – e.g., when three parallel propositions are set against a fourth.

Though space does not permit elaboration here, the same arguments apply to many other connectors; cf. sequences like (2), repeated below:

(2) My sisters love the beach. Similarly, I love the beach.

Reader approval on similarly is modest for such pairings, but when a third sentence is added as an explicit inferential conclusion, approval rises from 64% to 94%:

(10) My sisters love the beach. Similarly, I love the beach.
So our family would probably not be happy living in Montana.

As was the case for moreover, this third proposition need not be overtly stated as long as a context is easily accessible that prompts the reader to construct the proposition mentally. 

Pedagogical applications are not especially difficult here. Learners need to study examples of connectors in naturalistic prose, of course, but there is often excessive ‘noise’ in such data that needs to be filtered out. The use of simplified constructed frames such as those in the experimental examples provided above can cut to the chase in addressing the basic conditions of felicitous connector use and be applied to examining naturalistic prose later. I have used this procedure for many years in advanced composition courses and can attest to its being well-received by learners.

References

Blakemore, D. (2002). Relevance and Linguistic Meaning. Cambridge, UK: Cambridge University Press.

Grice, H.P. (1975). Logic and conversation. In P.Cole and J.Morgan (eds.) Syntax and Semantics 3: Speech Acts. New York, NY: Academic Press.

Halliday, M.A.K. and Hasan, R. (1976). Cohesion in English. London: Longman.

Sperber, D. and Wilson, D. (1986/1995). Relevance. Cambridge, MA: Harvard University Press.

Williams, H. (1996). An Analysis of English Conjunctive Adverbial Expressions. Unpublished doctoral dissertation, University of California at Los Angeles, Los Angeles, CA.

Howard Williams is Senior Lecturer in the Applied Linguistics & TESOL Program at Teachers College, Columbia University, New York. He taught ESL in community colleges and universities from the mid-1980s onward and currently teaches courses in general and pedagogical linguistics at TC. His research interests lie in grammar, pragmatics, and their pedagogical applications.