1The question that will be addressed here concerns the link that exists between syntax and semantics. This is not in itself an original topic, but we intend to discuss this phenomenon in relation to the theme of the Conference organised by the LERMA, namely: What relationship exists between form and meaning? And could syntax be deceptive, misleading?
2Syntax and semantics co-participate in the construction of meaning in various ways, but they often oppose each other, because of their own way of functioning: syntax is rather simple, and consists of a small number of rules, compared to the intricacies of the lexicon and its network of semantic features. Syntax is dependent on the human characteristics of sound production, which require that words be pronounced one after the other, hence the linear order of the sentences, regardless of their semantic complexities.
3The question which immediately arises is the following: could syntax be deceptive, misleading? The answer we want to argue here is that syntax is often misleading. Syntax is misleading, because of its own constraints, and in particular owing to the fact, mentioned above, that the words must follow one another, and this constraint forbids the components of the lexicon to relate freely. Our vocal organs function like a violin, and not like a piano, which can play ten notes at a time, just like the lexicon which articulate various shades of meaning and varied combinations of data.
4In our first part, we shall discuss how syntax tackles the problem of the linear order, and why it cannot always mirror semantics. The consequence is that it is misleading in some cases. The examples are taken from English and from French.
5In our second part, we shall see how some semantic data tend to reorganize syntax, making use of patterns not directly ready for them.
6If there seems to exist a kind of war between these two domains, each one prevailing over the other at times, other domains interfere, such as phonetics, pragmatics, the rules of discourse, but we will not have time to take them into consideration here.
7We generally tend to consider that syntax and semantics are working hand in hand, and that the syntactic rules of a language do not contradict themselves in order for the right interpretation to be always construed. But this is only partly true in a simple sentence, and often totally false in a complex one.
8This can be easily exemplified by the interpretation of the negation in the following example:
(1) I didn't say that to hurt your feelings
9In (1) we need to attach the negation to “did”, in accordance with negation rules, but the negation does not actually bear on the structure [I-said that]. It bears on the infinitive. In other words, by uttering: I didn't say that, the speaker is not denying that he said something, but he means, because of the infinitive: I said that, but I didn't mean to hurt your feelings. In syntax, the negation appears in the main clause, but in semantics the interpretation of the negation is on the subordinate clause. This is a very simple example of a deceptive structure, but, as the whole structure is understood globally, and not word after word, nobody even notices the incoherence.
- 1 English and French differ on this point: in English the word order remains the same for the direct (...)
10Yet the linear order is important in a language like English or French where declensions have disappeared, except for personal pronouns1, and where the syntactic function of a Noun Phrase is given by its position before or after the verb. The following sentences are understood differently because of the SVO word order:
(2) John criticized Mary
(3) Mary criticized John
11John is the subject in (2) and the object in (3).
- 2 See Roberto Pellerey for a complete analysis of this evolution.
12The SVO word order gradually became the rule in the Middle English period, and in France the rationalist grammarians of the seventeeth century with Port-Royal insisted on what they considered the natural order of the words: the cause must appear before the effect, hence the subject before the predicate.2 This word-order is the basis of the syntactic rules, and for instance the link between the subject and the verb is marked by the agreement on the verb.
13This brief introduction only recalls well-known facts, but we feel that they are too often disregarded, and this leads to interrogations that might not be entirely relevant.
14Let us compare :
(4) This book is heavy, and
(5) This book is difficult
15In each case we think that a quality of the book is being predicated by the speaker. But are these two qualities understood similarly? They are not. In (5) we are giving our viewpoint on the content of the book, and we mean that it is difficult to read/to understand the book. The quality difficult is predicated of a process: to read this book is difficult, not of the book itself, even if the third person singular agreement suggests that difficult should be semantically linked to the subject.
16Why is it then possible to understand a sentence that is, syntactically speaking, misleading, or at least incomplete?
17It is the semantic features of the adjective difficult and our general cognitive knowledge which tell us that a difficulty can only be the difficulty of a process, because a process requires an Agent acting on an entity, and the difficulty is the difficulty felt by the Agent during the process.
18With modals in their epistemic reading, there is no semantic linking between the subject and the modal, despite the word order. In:
(6) John must have missed the train
- 3 Unlike English, the French “passé composé” is used to refer to past events, with a definite date: (...)
- 4 This epistemic utterance could be replaced by: il doit avoir raté son train, which is similar to (...)
19the speaker assesses the chances that the relation [John – missed the train] might be an accurate description of what happened. The scope of the modal is the predicative relation and not “John”, because only a process can be considered sure, likely, uncertain… The same remark can be made about the French equivalent, which is even more deceptive: Jean a dû rater le train, where the verb devoir is conjugated in the passé composé3, but does not mean that the hypothesis was made in the past. It was made at Speech Time4, but it is the event/eventuality rater le train which belongs to the past, if it did take place.
20The discussion of the modal must suggests that languages need to distinguish between what can be said of individuals/entities, and what can be said of events/processes. Not all nouns can be associated with any verb. It would be odd to say:?My brother took place yesterday, whereas the negociation took place yesterday is perfectly fine. Yet my brother can appear as the Noun Phrase before a verb denoting the existence of an event:
(7) My brother happened to meet Mary on the market yesterday
- 5 John happened makes sense in the following co-text: her life was very sad until John happened, whi (...)
21What “happened” here is the meeting of my brother with Mary: my brother met Mary yesterday. The verb happen encodes the speaker's assessment of this event, which he had not expected. By using happen, he expresses his surprise, but no semantic link exists between “my brother” and “happened”5.
- 6 See Gérard Mélis et Paul Larreya, inter alia, for analyses of this phenomenon.
22A clearer illustration can be given by another well-known “deceptive” structure, the one found with sure6:
(8) John is sure to win his match
(9) The boat is sure to sink
23do not mean, despite the word order and the third-person agreement, that John or the boat are sure of anything. What is sure, according to the speaker is that John will win his match, and that the boat will sink: [John –win his match] is sure/ [the boat-sink]is sure.
24To encode John's certainty, we must use another structure:
(10) John is sure of winning his match
25This means that it is the type of complement structure (To+V / V-ing) that enables us to build the right interpretation.
- 7 We do not have time here to discuss unaccusative verbs, and their lack of subject.
26We have so far discussed the link between the subject and the verb Phrase, and the distinction that must be made between the syntactic subject, which is responsible for the agreement with the verb, and the semantic/notional subject, the entity about which something is said. Nothing very original was presented7 here. But the link between the verb and its complement is worth considering in detail, in particular when this complement receives the Accusative Case.
27As far as the linear order is concerned, a sentence encoding someone's point of view with the verb believe will receive the following pattern, if the infinitive is chosen: [X believes Y to V], and will yield :
(11) John believes Peter to be a doctor
(12) John believes him to be a doctor
28How are we to understand the Accusative Case on him?
29If we compare the sentences with the simple sentences: John believes Peter/John believes him, the interpretation is straightforward and can be paraphrased by: John trusts Peter/him, with the Accusative Case clearly designating Peter as the object of the verb. The interpretation follows from the syntactic rule positing that the subject is before the verb and the object after.
30But can we draw the same conclusion about (12)? In other words, is there a semantic link between the verb and him, namely Peter? Our answer is NO, no matter what the Accusative suggests. The verb believe is a two-place predicate: the subject has a certain belief about the object, but the object of a certain belief can only be a proposition and not an entity. By saying John believes him, the speaker means in fact: John believes that what Peter said is true. Resorting to the simple sentence is a kind of summary, a “shortcut”, whose interpretation can be retrieved from the co-text. What about: John believes him to be a doctor? Is there a notional link between believe and him? If there was one, the structure would be:
31John believes him [him to be a doctor], and the verb believe would be considered as a three-place predicate. which does not really make sense. The structure is then:
John believes [him to be a doctor]
32with him belonging to the subordinate clause, and not to the main clause, contrary to: John asked Mary[for PRO to lock the door], where Mary is at the same time the complement of ask (a three-place predicate) and the subject, via PRO, of lock the door. This structure is derived from the lexicon, from the semantic features of the verb ask.
33This structure raised a problem for the Case theory, as only verbs and prepositions could assign Cases, and ECM (Exceptional Case Marking operation) was then imagined to account for the Accusative Case.
- 8 We shall see that French word order is less fixed.
- 9 They are also dependent for their temporal interpretation. See Girard-Gillet, 2001.
34The hypothesis we propose here is based on the idea that syntax and semantics are not always working hand in hand. Syntax follows its own constraints, and is particularly sensitive to dependency data. In a SVO language, like English8, with a fixed word order, the first Noun Phrase (the first argument) receives the Nominative Case and the second (the second argument) the Accusative Case. The Accusative Case is thus always a dependent Case and it is no wonder that it appears with infinitive clauses, which are dependent clauses9. As the dependency of the clause cannot be ascribed to the clause itself –only Noun Phrases receive Cases- we suggest that it is the first Noun Phrase in the dependent clause that receives a dependent Case, hence the Accusative. But this Case is a syntactic marking, not a semantic one.
35The lack of semantic/notional link between the so-called object of believe and believe can be shown by the following occurrence:
(13) John believes Peter/him to be a liar.
36If there was some kind of link between the subject and the verb, the sentence would imply: John believes him, and we would get a contradictory reading, namely: John believes him -> John believes that what Peter said is true, and John believes that Peter is a liar, hence that what he said is not true…
37The following arguments demonstrate that it is the whole infinitive that is the object of believe:
(14) *John believed Mary himself to be English
38is not felicitous, and must be replaced by (15)
(15) John believed himself Mary to be English
39where Mary is the subject of the infinitive and the first element of the subordinate clause.
40But the (14) pattern is possible with persuade:
(16) John persuaded Mary himself to change jobs
41since Mary is the complement of persuade and thus belongs to the matrix clause. Its Accusative Case is assigned by the verb persuade.
42In Generative Grammar, the verbs persuade and promise were used to show that two kinds of subject interpretation existed in English for infinitive clauses:
(17) John persuaded Mary to come back home early
(18) John promised Mary to come back home early
- 10 For an interesting discussion of the two constructions of help, see Mathilde Pinson's Ph.D Dissert (...)
43Despite the same word order, it is Mary that is understood as the subject of come back in (17) but it is John in (18); hence the idea that there are object-control verbs and subject-control verbs. The problem is that, apart from promise, there does not seem to exist another three-place predicate belonging to the subject-control class of verbs. Does this mean that Semantics (the lexicon) forced its own rules on Syntax, imposing that, for convenience sake, the entity having a notional link with the second verb be closer to this verb than to the matrix verb? This would explain why there are so many three-place predicates functioning as object-control verbs: ask, tell, force, compel, advise, help10, urge, etc. We shall see in our second part how Semantics forces its way into Syntax in resultative constructions.
44It is worth having a look at French data now, and a discuss of the causation pattern with faire will show the same discrepancy between syntax and semantics. Here are two occurrences, one with full Noun Phrases, and one with pronouns:
(19) Jean a fait lire le livre à Marie
Jean has made read the book to Marie (word for word translation)
John made Mary read the book
(20) Jean le lui a fait lire
Jean it her has made read (word for word translation)
John made her read it
45In (19) the sentence is necessarily interpreted as meaning that Marie read the book, despite the fact that Marie is understood syntactically as the indirect object of lire.
46The pattern is in fact the pattern of any sentence with a three-place predicate, such as:
(21) Jean a donné un livre à Marie (John gave a book to Mary)
(22) Jean a reçu une lettre de Marie (John received a letter from Mary
(23) Jean partage ses revenus avec son fils (John shares his income with his son)
47In each case the third argument is introduced by a preposition, and is analysed as an indirect object. In the causative pattern, there are three Noun phrases: Jean, le livre, and Marie. The first receives the Nominative Case –only visible with a pronoun- the second the Accusative Case and the third appears as the complement of the preposition.
- 11 The word order is different for the first and the second person: Jean me le fait lire/Jean te le f (...)
48The use of personal pronouns yields sentence (20), but le is an Accusative and lui a dative11. The same contradiction between the syntactic data and the semantic interpretation exists.
- 12 See Girard-Gillet 2006.
49In English, the SVO order is always maintained: John gave a book to Mary/John gave it to Mary, except for: John gave her the book12.
50In French, the right interpretation is then derived from the lexical features, the theta-roles, and not from the syntactic ones: Marie is a human being, and only Marie can be the semantic/notional subject of read.
51The sentences so far analysed had three Noun Phrases. What happens when only two Noun Phrases are used, as in:
(24) Jean a fait rentrer la poubelle (John had the bin brought back in)
(25) Jean a fait rentrer son fils (John had/made his son come back home)
52We understand from our general knowledge that someone brought the bin in in (24), but that the son came back home by himself in (25). Syntax does not help us here at all; on the contrary, it considers that la poubelle and le fils are direct objects, misleading us completely if syntax was our only clue to the interpretation.
53Here are two other examples:
(26) Jean a fait lire Marie
(27) Jean la fait lire
54They show that when only two Noun Phrases are present, only the Nominative and the Accusative are invited to play a role, which is economical to a certain extent, even if it is rather disturbing for too-logically-minded linguists. And yet, the Agentivity of Marie can be said to be the same in (19) and in (26), with the interpretation that the process took place, with Marie as its Agent. This is another piece of evidence that relying on syntactic Cases can be misleading. Fortunately we know how to resort to syntax when it is necessary, but we also know that we must prefer the semantic data when syntax leads us astray.
55Interestingly, the structure can even be made more complex when the indirect object introduced by à/au is no longer available because of the constraints on the lexical verb, as in:
(28) Jean a lu le conte aux enfants (John read the tale to the children)
56The causative structure needs then to resort to another preposition, namely par:
(29) Jean a fait lire le conte aux enfants par Marie (John had Mary read the tale to the children.
57The “par-complement” can be considered as the equivalent of the Agent in a passive construction, but no passive is present in sentence (29). We could suggest then that it is the semantic use of “par” that syntax adopts to give us a clue about how to construe the whole meaning. And here semantics prevails over syntax, which seems not to have the necessary resources to offer a different type of indirect complement.
58Without “par” the sentence is ambiguous, and can either mean that the children read the tale themselves, or that someone read the tale to them.
59We take these facts to corroborate the view that semantics has to adapt to the possibilities syntax offers, however reduced they may be.
- 13 See Pinker, 1989, and Goldberg, 1995, for discussions on ditransitive syntax.
60We shall not discuss the dative-shift alternation vs the prepositional construction13 when both are possible, and when the choice depends mainly on information packaging data. We shall only analyse the construction that is favoured when give does not involve the transfer of an entity from one individual to another, but when it is close to create, give rise to, bring about.
- 14 See Oehrle, 1976, for the first discussion of this meaning, with the sentence: Nixon gave Mailer a (...)
61In the following example there is no transfer, but something new happens, is created, because of someone else's activity14:
(30) His music gave Mary a headache
62The headache felt by Mary is the consequence of the music, and give only means that the music is responsible for it, in a cause-consequence relation.
63The same can be said about the next examples, taken from Google:
(31) Whispered conversations about money worries, which break off as soon as children come near, may give them the idea that a terrible disaster is about to strike.
(32) The water is so hard that it will give the kettle an internal fur coat in no time.
(33) The fact that Anne now relied on her gave Joan a curious but nonetheless pleasurable sense of power.
(34) a report by the National Council of Public Morals on the Cinema had scrutinized some evidence from children themselves on whether the moving pictures gave them bad dreams.
64Heidi Harley (2003), who worked on the two constructions of give, proposes a locative interpretation for the structure under discussion:
(35) Sandy gave Terry a copy of the new grammar (ex 54):
- 15 For the prepositional construction, Harley opts for:
65 [VP Agent [V' CAUSE [ PP Goal [P' HAVE [DP Theme ]]]]] -> Cause Terry to have15
66We suggest that the dative-shift structure is favoured because there is a closer link between “Terry” and “the copy” in the linear order, and this contributes to our treating together [Terry-the copy], both depending on gave. Semantics seems to be using a device already present in English grammar, namely the dative-shift construction to reach its proper aim: the right interpretation of give, as a causative verb, and it does so by taking advantage of a syntactic choice, and by more or less “twisting” its standard meaning.
- 16 The same impossibility exists with introduce, suggest, explain, which all stress the message.
67Is it syntax that solves the problem of ambiguity? To a certain extent, it is, but if a verb, like the verb give, had not developed a broad range of meaning, syntax would not have any role to play. Likewise, the verbs say and tell exhibit two different structures, and we can wonder whether they are not relying on syntactic possibilities too, in order to differentiate their own meanings. We know that say focuses on the message and tell on the addressee, and interestingly say cannot be constructed as its French equivalent dire with a dative16:
(36) Je lui ai dit qu'il était en retard
(37) *I said him that he was late
68The use of the preposition to,
(38) I said to him that he was late
69could be a sign of a looser link between the verb and the indirect object. In his analysis of say, Goosens (1982:115) found only 5,5% of cases where say was used with to, and the mention of the co-speaker. With say the addressee is superfluous, in a sense, because by reading:
(39) They say they have wiped smallpox off the face of the earth (Google)
70we receive the information mentioned in the subordinate clause, even if we were not the direct addressee of the message.
71When the addressee is present, it is tell that is favoured, and the dative is the case assigned to him/her. With tell the addressee is supposed to know more at the end of the utterance, because the message was meant for him/her:
(40) I was remembering what my father had told me about other conquering heroes who had tried to attack the great bear. (BNC)
(41) She told him he was lucky to be married to a “beautiful, intelligent artist”. (BNC)
72This explains the difference between:
(42) Mary said “hello”,
73where say is used to report someone's words, and:
(43) My neighbour told me his name / he told me what his name was,
74which does not repeat the neighbour's words, but which encodes that the speaker now knows what his name is. In sentence (43) the name itself is not given, and anyone reading the sentence is at a loss to guess it. But as the recipient of the message, the speaker is now aware of it. The transfer of the information took place prior to (43), but in sentence (43) the speaker is only giving the result of the transfer. We suggest then that with tell, there is an interpretation of the words pronounced by the first speaker, and a reformulation of them. That is why we can say: he told me the time, and not: *he said the time to me. Other terms, like “story”, can be also be a complement to tell:
(44) I find it a completely achieved poem; probably Coleridge was lying when he told the story of the person from Porlock, nearly twenty years later. (BNC)
75This does not, of course, disallow an entire message, as in:
(45) Mr Ward told The Independent: ‘Mr Rowland himself telephoned me the next day, said he was appalled at what had happened to my daughter, and gave me his direct line telephone number, saying he would be there every day from 9 to 6. (BNC)
76Note in passing that give requires the Goal, but say does not, as it is retrieved from the co-text.
77Drawing from H.Harley, and her structures for give, we suggest the following structures for say and tell, in at least one of its uses:
78SAY : localisation of the message in the addressee, if any:
79 [VP Agent [V' CAUSE [PP Theme [P' Loc [PP to recipient]]]]]
80TELL : interpretation/reformulation of the message:
81 [VP Agent [V' CAUSE [PP Recipient [P' KNOW [DP Theme ]]]]]
82The different interpretations of the passive: John was said to be rich / John was told to be quieter can be derived from these structures, but a discussion of their characteristics would take us too far from our topic.
83Beth Levin, 1993, 2009, and B. Levin/Malka Rappaport-Hovav, 2001, have been working for many years on the semantic classes of verbs and have shown that the richer a verb is semantically speaking, the more syntactic structures it will develop. They have then proved that sometimes syntax is determined by semantics, which is not what we have seen so far in our first part. But as syntax and semantics must necessarily be related, we can hardly wonder why this should not be so.
84 Manner of motion verbs offer an interesting domain of investigation, and the comparison they made between the verbs run and go prove that run which contains more semantic features than go can appear in many different structures. They compare sentences (46-51) with (52-57) which are deviant, except when go has its basic meaning of a directed movement.
(46) Pat ran (activity)
(47) Pat ran to the beach (directed motion)
(48) Pat ran herself ragged (change of state)
(49) Pat ran her shoes to shreds (change of state)
(50) Pat ran clear of the falling rocks (directed motion)
(51) The coach ran the athletes around the track (causation)
(52) The students went (activity)
(53) The students went to the beach (directed motion)
(54) * the jetsetters went themselves ragged (change of state)
(55) * The runner went his shoes to shreds (change of state)
(56) * The pedestrians went clear of the oncoming car (directed motion)
(57) * The coach went the athletes around the track (causation)
85For Levin and Rappaport-Hovav it seems to be impossible to turn go into a change of state verb or into a causation verb, since its semantic features only enable it to express movement, and to behave as an intransitive verb. To develop the semantic features of run, syntax offers its transitive and resultative patterns, yielding the sentences mentioned above.
86Things are in fact more complex and the semantic features of these two verbs do require more work. An interesting discussion is offered by Kudrnacova (2008), who shows that not all manner of motion verbs behave like run:
(58) *John strutted/staggered/limped himself to the park
(59) * John jogged Michael to the park.
87Her explanation is that these verbs encode movements that are not subject to the executor's control in their entirety (2008:11).
- 17 Kudrnacova disagrees with Huddleston & Pullum, and mentions that for them “assigning a location to (...)
88We do not have time to discuss here the differences between a change of state and a change of location17, but we wish to suggest other semantic characteristics that might account for the syntactic constructions these verbs allow. The verb go seems more abstract and can denote the passage from an absence of movement to the beginning of a movement (inchoation), regardless of the means used, as in:
(60) Ready, Steady, Go!
89No movement in space is implied in:
(61) My alarm clock went off at seven
(62) The bomb went off,
- 18 A comprehensive study of path verbs or manner of motion verbs needs to consider the semantic roles (...)
90but there is nonetheless an abstract passage from a lack of activity to the activity inherent to the object concerned: an alarm clock is supposed to ring at a certain time, and a bomb is meant to explode.18
91The verb go can also refer to a change of state for an object on sale in: going, going, gone, meaning sold in an auction. An anonymous referee mentioned the following sentence: to go red in the face, which is another type of change of state. Google gives occurrences of go ragged, such as:
(63) My breath went ragged in the cold.
92Unlike exemple (54) where the jetsetters would be understood as the agent of the change of state on themselves, the sentence my breath went ragged suggests that the subject is affected by the process, without any agentivity on its part. The ungrammaticality of (54) can then be accounted for by the thematic role of the subject. And interestingly, Kudrnacova distinguishes between verbs encoding the executor's control on the activity and those which do not.
93Unlike go, which has an unaccusative interpretation, run can function as a causative verb with a transitive use:
(64) She ran him a bath (an Agent initiates the movement of the water).
94It can also denote a change of state:
(65) The Worst Case: What If the Water Ran Dry in the Japanese Reactors? (Google)
95These examples show very clearly how semantic features can determine what syntaxcan do. A lot of research is currently going on to define what parameters are mostrelevant. Among them we can note theopposition between internal and external causation.
(66)John talked Mary into resigning
97is a standard construction of a resultative structure, but it is syntactically speaking something of a linguistic monster. Why? Because talk is an intransitive verb, whose complement is introduced by to: John talked to Mary. *John talked Mary is unfelicitous. How can we then account for (66) without endangering the rules of syntax?
98Other verbs appear in this pattern, and among them we find: blackmail, brainwash, threaten, lure, coax, trick, which encode the manner/means resorted to by an individual to get the expected result.
(67) But that's no reason to brainwash them into staying at home. (BNC)
(68) The warmth that had come with his laughter faded, and she felt he had tricked her into saying more than she intended. (BNC)
99But these verbs, unlike talk, are transitive verbs, and they can even be turned into the passive:
(69) He was blackmailed into supporting Castro (Google)
(70) Nine women were tricked into thinking they were reality TV show contestants (Google)
(71) Sirhan now says he was brainwashedinto assassinating Robert Kennedy (Google)
- 19 An anonymous reviewer wondered whether lure, trick and coax necessarily involve something being sa (...)
- 20 Construction Grammars posit that it is the structure itself —a ternary structure here—that is resp (...)
100We can posit that talk, which involves the oral use of language, shares these features with the verbs mentioned above, which all involve that something was said19.The syntactic structure in which these verbs appear is then adopted by the verb talk because it belongs to the same semantic field: it can then introduce its object directly, and, still more important, it expresses, thanks to the preposition into, the process that is brought about by the words pronounced. The whole structure is borrowed, and it is what makes the enrichment of the verb viable.20
101Here we have a case of semantic data imposing their contents and their expressions on syntax.
102To understand the link between form and meaning is what linguistics is about. Syntax cannot alone construe meaning, mainly because of the constraint of the word order, the surface structure, to use a term coined fifty years ago by Chomsky. But traditional grammars had already noticed the absence of a direct link between a syntactic function and a semantic interpretation. It is through the complex interactions between what the words co-occurring in the sentence conjure up and the position they appear in that language enables us to communicate our thoughts. This explains the discussions still taking place between the hypotheses developed by Pustejovsky, Fodor, Lepore, Wechsler among others, about what is syntactically relevant and what is semantically fundamental, as the parts played by the two domains are often difficult to differentiate.